Valuation of Morbidity Losses: Meta-Analysis of ...(RIAs), CFSAN must have at its di sposal reliable...
Transcript of Valuation of Morbidity Losses: Meta-Analysis of ...(RIAs), CFSAN must have at its di sposal reliable...
March 2003
Valuation of Morbidity Losses: Meta-Analysis of Willingness-to-Pay
and Health Status Measures
Final Report Task Order 2
Prepared for
Amber Jessup Food and Drug Administration
Center for Food Safety and Applied Nutrition HFS-726
5100 Paint Branch Parkway College Park, MD 20740
Prepared by
George Van Houtven Matthew Rousu Jui-Chen Yang
Charles Pringle Wanda Wagstaff
Jason DePlatchett RTI
Health, Social, and Economics Research Research Triangle Park, NC 27709
Contract Number 223-01-2466
RTI Project Number 08184.002
Contract Number 223-01-2466 RTI Project Number 08184.002
Valuation of Morbidity Losses: Meta-Analysis of Willingness-to-Pay
and Health Status Measures
Final Report Task Order 2
March 2003
Prepared for
Amber Jessup
Food and Drug Administration Center for Food Safety and Applied Nutrition
HFS-726 5100 Paint Branch Parkway
College Park, MD 20740
Prepared by
George L. Van Houtven Matthew Rousu Jui-Chen Yang
Charles Pringle Wanda Wagstaff
Jason DePlatchett RTI
Health, Social, and Economics Research Research Triangle Park, NC 27709
iii
Contents
1. Introduction 1-1
1.1 Background....................................................................... 1-2
1.2 Summary of Results ........................................................... 1-3
2. Conceptual Framework for Morbidity Valuation 2-1
2.1 Basic Framework for Health Valuation............................... 2-1
2.2 Extensions of the Basic Framework .................................... 2-3
2.2.1 Health Production ................................................. 2-4
2.2.2 Uncertainty ........................................................... 2-6
2.2.3 Lifetime Utility and QALYs .................................... 2-8
2.3 Determinants of Health Values ........................................ 2-13
2.3.1 Measures of Health Changes ................................ 2-13
2.3.2 Study Population Characteristics .......................... 2-19
2.3.3 Price Effects ......................................................... 2-20
2.3.4 Valuation Method................................................ 2-20
2.4 Summary ........................................................................ 2-22
3. Analytical Approach—Meta-Analysis 3-1
3.1 Meta-Analysis in Nonmarket Valuation.............................. 3-1
3.2 Procedures for Conducting Meta-Analyses ......................... 3-3
4. Data Collection and Evaluation 4-1
4.1 Review and Selection of Studies on WTP for Improved Health............................................................................... 4-2
4.1.1 Literature Search and Screening ............................. 4-2
4.1.2 Annotated Bibliography of Selected Studies............ 4-3
iv
4.2 WTP for Health Database .................................................. 4-4
4.2.1 Database Design.................................................... 4-5
4.2.2 Data Summary and Evaluation ............................. 4-20
4.3 Database of Health Status Measures................................. 4-23
4.3.1 Overview of HSMs .............................................. 4-23
4.3.2 MAUS Database Description................................ 4-31
5. Meta-Analysis Results 5-1
5.1 Meta-Analysis of Value Estimates for Acute Effects ............. 5-1
5.1.1 Data Selection and Description.............................. 5-2
5.1.2 Meta-Regression Models and Results...................... 5-6
5.1.3 Implications of Results for Benefit Transfer ........... 5-16
5.2 Meta-Analysis of Value Estimates for Chronic Effects ........ 5-22
5.2.1 Data Selection and Description............................ 5-22
5.2.2 Meta-Regression Models and Results.................... 5-27
5.3 Summary and Conclusions .............................................. 5-30
6. Summary and Discussion of Results 6-1
6.1 Illustrative Applications of the Estimated Benefit Transfer Function for Acute Effects..................................... 6-3
6.2 Conclusions ...................................................................... 6-6
References R-1
Appendixes
A Bibliography and Summary of Morbidity Valuation Studies ..............................................................................A-1
B Annotated Bibliography of Publications Included in the Morbidity Value Meta-Analyses ......................................... B-1
C Summary Statistics for the Morbidity Value Database .........C-1
v
Figures
Figure 4-1 Three-Level Database Design ..................................................... 4-5
vi
Tables
Table 2-1 Comparison of Health Status Measures..................................... 2-14
Table 4-1 Number of Publications per Study .............................................. 4-6
Table 4-2 Number of Value Estimates per Publication ................................ 4-6
Table 4-3 Study-Level Data Fields (Spreadsheet 1)..................................... 4-6
Table 4-4 Publication-Level Data Fields (Spreadsheet 2)............................. 4-7
Table 4-5 Value-Level Data Fields (Spreadsheet 3.1) .................................. 4-9
Table 4-6 Value-Level Data Fields (Spreadsheet 3.2) ................................ 4-11
Table 4-7 Value-Level Data Fields (Spreadsheet 3.3) ................................ 4-13
Table 4-8 Value-Level Data Fields (Spreadsheet 3.4) ................................ 4-16
Table 4-9 Value-Level Data Fields—Stated Preference Methods (Spreadsheet 3.5) ..................................................................... 4-18
Table 4-10 Value-Level Data Fields—Hedonic Method (Spreadsheet 3.6) ..................................................................... 4-19
Table 4-11 Value-Level Data Fields—Averting Behavior Method (Spreadsheet 3.7) ..................................................................... 4-20
Table 4-12 Number of Publications by Year............................................... 4-21
Table 4-13 Number of Publications by Type of Publication ........................ 4-21
Table 4-14 Valuation Methods Used (Number of Value Estimates per Method.................................................................................... 4-21
Table 4-15 Number of Value Estimates by Country .................................... 4-22
Table 4-16 Number of Value Estimates by Type of Health Condition Valued..................................................................................... 4-23
Table 4-17 Symptom and Problem Complexes (CPX) for the Quality of Well-Being Scale ..................................................................... 4-25
Table 4-18 Dimensions, Function Levels, and Weights of the Quality of Well-Being Scale ..................................................................... 4-26
Table 4-19 Multiattribute Health Status Classification System: Health Utilities Index Mark 3 (HUI-3) .................................................. 4-28
vii
Table 4-20 The EuroQol Descriptive System .............................................. 4-31
Table 4-21 Summary of MAUS Studies and Scores for Selected Health Conditions ............................................................................... 4-33
Table 5-1 Descriptions of Variables Used in the Meta-Analysis................... 5-4
Table 5-2 Summary Statistics for Variables Used in the Meta-Analysis ........ 5-5
Table 5-3 Meta-Regression Results—WTP for Avoided Acute Effects Using the Total QWB Score ....................................................... 5-9
Table 5-4 Meta-Regression Results—WTP for Avoided Acute Effects Using the Total QWB Score ..................................................... 5-10
Table 5-5 Meta-Regression Results—WTP for Avoided Acute Effects Using the Four-Dimensional QWB Scores ................................ 5-14
Table 5-6 Meta-Regression Results—WTP for Avoided Acute Effects and Four-Dimensional QWB Scores ......................................... 5-15
Table 5-7 Benefit Transfer Function Estimates........................................... 5-18
Table 5-8 Out-of-Sample WTP Predictions with BT Function 1................. 5-19
Table 5-9 Out-of-Sample WTP Predictions with BT Function 2................. 5-20
Table 5-10 Chronic Health Effect Descriptions and Scores.......................... 5-24
Table 5-11 Descriptions of Variables Used in the Meta-Analysis................. 5-26
Table 5-12 Summary Statistics for Variables Used in the Meta-Analysis ...... 5-27
Table 5-13 Meta-Regression Results—WTP for Avoided Acute Effects and Total QWB Score .............................................................. 5-28
Table 6-1 Three Illustrative Applications of the Meta-Analytic Benefit Transfer Function for Acute Effects.............................................. 6-5
1-1
1 Introduction
A primary objective of the Food and Drug Administration’s (FDA’s) Center for Food Safety and Applied Nutrition (CFSAN) is to protect and improve public health through a variety of food safety regulations. It is well recognized, however, that regulatory actions will typically impose both costs and benefits on society. Therefore, CFSAN has the responsibility to develop methods for accurately assessing these costs and benefits.
The purpose of this project is to assist CFSAN in strengthening its capabilities for assessing the health benefits, in monetary terms, of its regulatory alternatives. To conduct regulatory impact analyses (RIAs), CFSAN must have at its disposal reliable and cost-effective benefits assessment methods. Unfortunately, CFSAN does not have the resources to conduct original research on the value of all of the health outcomes affected by its actions. Consequently, it must make best use of the existing research on health valuation to inform its decisions. In other words, it must rely to a large extent on “benefit transfer” approaches.
This report summarizes RTI’s efforts and results in developing a systematic benefit transfer method for valuing changes in morbidity. This method is based on an integrated statistical analysis (“meta-analysis”) of results from the existing health valuation literature. By combining the findings from multiple studies, we are able to specify and demonstrate a benefit transfer function for acute health effects. We are also able to compare value estimates based on this function with estimates based on separate benefit transfer approaches that CFSAN has used for recent regulatory analysis. In addition, the
The purpose of this project is to assist CFSAN in strengthening its capabilities for assessing the health benefits, in monetary terms, of its regulatory alternatives.
Valuation of Morbidity Losses: Meta-Analysis of Willingness-to-Pay and Health Status Measures
1-2
meta-analysis for acute effects lays an important foundation for assessing values for chronic health effects as well.
1.1 BACKGROUND Regulatory action by CFSAN can protect against a wide variety of adverse health outcomes. Morbidity outcomes can range from short-term acute conditions, such as incidents of food poisoning or allergic reactions, to long-term chronic conditions, such as reactive arthritis from food contamination or diabetes associated with poor nutrition. These outcomes can also vary considerably in terms of their severity, and in more extreme cases, they can progress beyond illness (morbidity) and cause death (mortality).
Assessing the health benefits of its regulations is a challenge for CFSAN in part because the monetary values for avoiding many of these health effects have not been well quantified. This lack of value information is particularly the case for morbidity outcomes. Although valuation of mortality effects continues to be somewhat controversial and contains important areas of uncertainty, compared to morbidity valuation it is relatively well researched and summarized (Viscusi, 1993; Mrozek and Taylor, 2002).
A number of general methods for valuing morbidity effects exists; however, each has important disadvantages. As we discuss in more detail in Section 2 of this report, it is generally accepted that a complete accounting of losses due to ill health must capture direct costs (e.g., medical expenditures), indirect costs (e.g., lost income or productivity), and nonpecuniary losses such as those from pain and suffering. Cost of illness (COI) methods are often used to monetize losses from illness, but these methods do not capture the potentially important category of nonpecuniary losses. As we also discuss in Section 2, methods that focus on measuring individual’s willingness to pay (WTP) for avoiding illness are considered to be more conceptually correct and comprehensive. However, one of the main drawbacks of these methods is that they are relatively expensive to implement.
To address resource constraints and make best use of existing research, CFSAN and other regulatory agencies must often use benefit transfer approaches. These approaches involve identifying, selecting, and adapting value estimates from studies conducted in
To address resource constraints and make best use of existing research, CFSAN and other regulatory agencies must often use benefit transfer approaches.
Section 1 — Introduction
1-3
one context and applying them to estimate the benefits of effects (usually policy related) in a separate context. Transferring values from one context to another inevitably introduces additional uncertainties into the benefit estimation process; therefore, it is particularly important to evaluate these transfer methods carefully.
In recent years, CFSAN has predominantly used one benefit transfer approach for valuing changes in morbidity. This approach combines nonmonetary measures of the severity and duration of illness—quality-adjusted life years (QALYs) lost—with monetary measures of the value of avoided mortality—value of statistical life (VSL) estimates. As we discussed in a previous report (RTI, 2002), this approach has a number of appealing qualities, but it also imposes a number of relatively stringent assumptions regarding individuals’ preferences for health. For this reason and because of FDA’s interest in strengthening its regulatory decisions, CFSAN has asked RTI to reexamine this approach and to assist them in continuing to develop sound benefit transfer approaches for morbidity valuation.
1.2 SUMMARY OF RESULTS To address CFSAN’s needs, RTI conducted the following activities:
Z developed a conceptual framework that describes the microeconomic foundations for health valuation and identifies the key expected determinants of health values (Section 2);
Z reviewed the empirical literature on health valuation (including over 600 publications) and compiled a detailed bibliography of the most relevant 136 publications (Section 4);
Z selected WTP estimates from these studies and constructed a database of health values, which currently contains 389 WTP estimates (and corresponding data) from 44 publications (Section 4);
Z used meta-analysis methods to analyze subsets of the database—236 WTP estimates for avoiding acute illness and 38 WTP estimates for avoiding chronic illness (Section 5); and
Z applied the meta-analysis results to specify benefit transfer functions for acute illnesses (Section 5).
Valuation of Morbidity Losses: Meta-Analysis of Willingness-to-Pay and Health Status Measures
1-4
Through this process, we also tested hypotheses regarding the determinants of WTP estimates. The results of the meta-analysis indicate that WTP estimates for avoided acute effects vary in systematic and expected ways with respect to key explanatory variables. We found a strong statistical relationship between the value estimates and corresponding measures of the severity and duration of the health effects. In addition, we find generally positive and significant income effects and age effects.
The meta-analysis results also provide a simple but informative test of the assumptions underlying the QALY valuation approach for assessing morbidity values. The results strongly reject the assumption of a constant value per QALY and the assumption that the duration and the severity of illness have equivalent and proportional effects on WTP.
In Section 6, we illustrate how the results of the analysis can be applied to estimate the benefits of avoiding specific acute conditions often associated with foodborne illness. We also discuss other implications and limitations of the analysis.
The data and analyses assembled for this project provide a foundation for health benefits analysis that should extend beyond this report. The bibliography and databases described in Section 4 should serve as general resources for identifying, summarizing, or transferring estimates from the health valuation literature. The databases also provide a structure for organizing data that can easily be used to include information from more studies. As such, they should support additional analyses (including meta-analyses) of the health valuation literature and continued development of benefit transfer tools.
The results of the meta-analysis indicate that WTP estimates for avoided acute effects vary in systematic and expected ways with respect to key explanatory variables.
2-1
Conceptual Framework for 2 Morbidity Valuation
To provide CFSAN with a morbidity valuation approach that is not only practical to use but is also theoretically sound, it is important to establish an appropriate conceptual foundation for the approach. In this section, we
Z present a relatively simple theoretical model describing how measures of morbidity can be related to individuals’ preferences, and
Z use this framework to identify and describe the main factors expected to explain and influence estimates of morbidity values.
In other words, we describe the expected relationships and linkages between key variables, and we identify key hypotheses to be tested in the statistical analysis (described in Section 5).
2.1 BASIC FRAMEWORK FOR HEALTH VALUATION To formalize the way in which individuals derive value from changes in health, we begin with a simple conceptual framework that links an individual’s private utility (U) with his/her health status (H). This simple single period framework, which assumes that H is exogenously determined and that there is no uncertainty regarding H or other determinants of utility, is summarized in the following utility function:
U = U(H, X) (2.1)
In this function, X represents a vector of other goods that contribute to utility. Individuals are assumed to maximize this utility function
Valuation of Morbidity Losses: Meta-Analysis of Willingness-to-Pay and Health Status Measures
2-2
subject to the budget constraint Y = PX, where Y is exogenously determined income and P is a vector of prices corresponding to the X goods.
The indirect utility function associated with this maximization process can be written as follows:
V = V(H, Y, P) (2.2)
The monetary value, or WTP, associated with an improvement in health from initial health, H0, to new health, H1, can be expressed by the compensating variation term (CV) in the following equation:
V(H0, Y, P) = V(H + ∆H, Y – CV, P) = V(H1, Y – CV, P) (2.3)
CV represents the reduction in income that would exactly offset the increase in utility resulting from health improvement, such that there would be no net gain in utility. Similarly, the WTP associated with avoiding a decline in health from H0 to H2 can be expressed by the equivalent variation term (EV) in the following equation:
V(H0 – ∆H, Y, P) = V(H2, Y, P) = V(H0, Y – EV, P) (2.4)
EV represents the reduction in income that would reduce utility by exactly the same amount as the health decline.
Both CV and EV are considered to be conceptually correct welfare measures for changes in health; however, even for equivalent gains or losses in health (i.e., holding ∆H constant), they will not necessarily have the same value.1 In particular, if the marginal
utility of health decreases with H ( 0H
V2
2
<∂∂ ) or if the marginal
utility of income increases with H ( 0HY
V2
>∂•∂
∂ ), EV is generally
expected to be greater than CV. Nonetheless, for small changes in health, the two measures should be roughly equivalent. Large differences between CV and EV are possible, but they may also
1CV can also be expressed as the increase in income that would exactly offset the
utility loss associated with a decline in health (i.e., an individual’s minimum willingness to accept [WTA] compensation for a decline in health). Although much attention has been devoted in the literature to discrepancies between WTP and WTA measures (see for example, Hanemann [1991] and Johansson [1995], in practice relatively few studies have estimated WTA values for health changes.
Section 2 — Conceptual Framework for Morbidity Valuation
2-3
indicate deviations from the standard utility model, such as the loss aversion and reference dependence models proposed by Tversky and Kahneman (1991). According to this alternative framework, individuals value changes with respect to a reference point (such as their status quo condition) and place a much higher (negative) weight on losses than they do on equivalent gains.2
Both Eqs. (2.3) and (2.4) can be rearranged to derive the following corresponding value functions:
CV = CV(H0, H1, P, Y) (2.3’)
EV = EV(H0, H2, P, Y) (2.4’)
These equations are essentially “variation” or WTP functions with respect to a change in health status. Consequently, they provide a conceptual basis for constructing and statistically estimating the meta-analytic functions of health values, which are described in Section 5. The expected properties of these functions and their implications for meta-analysis are described in more detail below.
2.2 EXTENSIONS OF THE BASIC FRAMEWORK At least three extensions of this basic framework are useful for establishing the conceptual basis for health values. The first extension involves using a “health production” function (HPF) to account for how individuals’ actions influence their health outcomes. Including a health production function allows for a more explicit link between individuals’ WTP to avoid a particular health outcome and the losses—both pecuniary and nonpecuniary—associated with that outcome. The second extension incorporates uncertainty into the model, through the use of an expected utility (EU) framework. This extension is particularly helpful for conceptualizing values that are based on changes in the probability (i.e., risk) rather than the certainty of a particular health outcome. The third extension introduces a temporal dimension to health-related utility. A specific and frequently used framework for
2Note that with a standard utility model, an individual’s WTP to avoid a loss from
H0 to H2 (EV in Eq. [2.4]) should be identical to his/her WTP for an increase in health (CV) going from H2 to H0. However, with reference dependence—shifting the status quo from H0 to H2—the gain may be treated differently from the avoided loss.
The basic framework can be usefully extended by including Z a health production
framework,
Z uncertainty regarding health outcomes, and
Z a temporal dimension to health outcomes.
Valuation of Morbidity Losses: Meta-Analysis of Willingness-to-Pay and Health Status Measures
2-4
characterizing health-related utility over time is the quality-adjusted life year (QALY) method. Although QALYs have important practical appeal, their relationship to WTP measures of health is a complex one. Each of these three extensions is discussed in more detail below.
2.2.1 Health Production
First developed by Grossman (1972), the HPF is now commonly used as the conceptual basis for explaining health behaviors and values. In contrast to the basic framework described above, where health outcomes are assumed to be exogenously determined, the HPF clearly distinguishes between exogenous and endogenous determinants of health. As a simple example, we define the following HPF:
S = S(H, M) (2.5)
As in Eq. (2.1), H is an exogenous variable, which can be thought of as a measure of a person’s health status (or accumulated health capital) at a point in time. For example, H could be a measure of the presence and/or severity of a chronic illness such as asthma. M is a choice variable representing purchasable goods, such as medication, which can be used to alleviate the symptoms of adverse health conditions associated with H. These are sometimes referred to as “mitigating” goods/activities. However neither H nor M directly affects utility. Rather, they jointly determine the health outcome (S) that does matter to an individual.
In other words, S is the health measure that directly affects a person’s level of utility such that Eq. (2.1) can be reformulated as follows:
U = U(S(H, M), X) (2.6)
For this example, we assume that S is a measure of the number of sick days experienced. In this case, the budget constraint can then be defined as follows:
Y = I + w(T – L – S) = X + PmM (2.7)
where I is exogenous income, w is the wage rate, T is total time, L is leisure time, S is sick time (such that T – L – S is labor time), and Pm is the unit price of M. The price of X is normalized at 1.
Section 2 — Conceptual Framework for Morbidity Valuation
2-5
Given this formulation, it can be shown (see, for example, Freeman [1993]) that the individual’s WTP for an improvement in health status (the reduction in I that would exactly offset the utility gain from a change in H) can be expressed by the following equation:
WTP = dIdH = Pm
∂M∂H + w
dSdH –
∂U/∂Sλ
dSdH (2.8)
The first term on the right-hand side of this equation represents the change (savings) in expenditures on M associated with an improvement in H. These savings represent avoided direct costs. The second term represents the avoided loss of wage income (opportunity cost) due to the reduction in sick days. It captures avoided indirect costs. The last term represents the nonpecuniary utility gain (e.g., avoided pain and suffering) due to improved health, which is converted to monetary terms through the marginal utility of income term ( ).
This decomposition of WTP highlights at least two important points. First, it emphasizes the distinction between the pecuniary effects (direct and indirect costs), which are typically captured by COI measures, and nonpecuniary effects (disutility effects), which are not. WTP estimates are interpreted to be more comprehensive than COI estimates because they include both effects.
Second, this decomposition implies that WTP is positively related to both the monetary costs of illness avoided and the amount of pain and suffering avoided. It is important to note, however, that an individual’s private WTP is only related to the portion of the costs that he or she bears. Through public and private health insurance and sick leave policies, individuals may only bear a small portion of the direct and indirect costs associated with an additional day of illness. Under these conditions, costs are shifted to others in society (e.g., employers or taxpayers), and the effective marginal costs faced by individuals for mitigating goods and sick days are less than their full prices (Pm and w, respectively). As can be seen in Eq. (2.8), lowering the prices of mitigating goods and sick days decreases WTP. When marginal costs of illness are subsidized in this way, private WTP to avoid an illness is less than the full societal WTP.
If individuals’ marginal costs of illness are subsidized—e.g., through sick leave or health insurance policies—then private WTP to avoid illness will be less than the full societal WTP.
Valuation of Morbidity Losses: Meta-Analysis of Willingness-to-Pay and Health Status Measures
2-6
2.2.2 Uncertainty
The previously described models treat changes in health as if they were known with certainty; however, it is often the case that individuals make choices and reveal or express values for changes that are not known with certainty. The traditional approach for including uncertainty in a conceptual model of health is to treat health outcomes (and thus utility) in a probabilistic manner. Using an EU framework, we assume that there are multiple (N) possible health states, that each has a defined probability (πi) of occurring, and that all of these probabilities sum to 1. As in the following expression, EU can be expressed as the probability weighted average utility associated with the various possible health states:
EU = ∑i=1
N πiV(Hi, Y, P) (2.9)
where πi is the probability associated with health state i.3
The EU framework has been widely questioned and tested, and several empirical violations of EU have been noted in the academic literature (Weber and Camerer, 1987). Despite its shortcomings, the simplicity and generalizability of EU continue to make it most useful (and most widely used) as a basic conceptual structure. Most violations of EU have been attributed to either systematic biases in individual risk judgments or to behavior that contradicts the assumption that utility is linearly related to the probabilities of outcomes, as expressed in Eq. (2.9). Alternatives to the EU framework have been proposed, such as theories involving decision weights, reference dependent preferences, and nontransitive and nonmonotonic preferences; however, none of these alternatives has emerged as a clearly superior framework for explaining individual decision making under uncertainty. For more on the arguments against the EU framework and some non-EU models, we refer the interested reader to Starmer (2000).
Eq. (2.9) can be used to develop an adapted measure of compensating (or equivalent) surplus. In this instance we are interested in the value of a reduction in health risk (i.e., a reduction in the probability of an adverse health outcome) rather than a
3For simplicity, we revert to the previous form of the model (without the HPF) and
assume here that health and income are exogenous.
Despite its shortcomings, the simplicity and generalizability of EU continue to make it most useful (and most widely used) as a basic conceptual structure.
Section 2 — Conceptual Framework for Morbidity Valuation
2-7
certain improvement in health status. For simplicity, we assume that there are two possible and mutually exclusive health states, good health, Hg, and bad health, Hb. We further assume that the initial risk associated with Hb is πb
0. WTP for a reduction in risk to πb
1 (< πb0) can be expressed by the CV measure in the following
equation:
πb0V(Hb, Y, P) + (1 – πb
0)V(Hg, Y, P) =
πb1V(Hb, Y – CVπ, P) + (1 – πb
1)V(Hg, Y – CVπ, P) (2.10)
Similarly, WTP to avoid an increase in risk to πb2(> πb
0) can be expressed by the EV measure in the following equation:
πb0V(Hb, Y – EVπ, P) + (1 – πb
0)V(Hg, Y – EVπ, P) =
πb2V(Hb, Y, P) + (1 – πb
2)V(Hg, Y, P) (2.11)
Both of these equations can also be rearranged to derive the following corresponding WTP functions with respect to change in health risks, health outcomes, income, and prices:
CVπ = CVπ (πb0, πb
1, Hg, Hb, P, Y) (2.10′)
EVπ = EVπ (πb0, πb
2, Hg, Hb, P, Y) (2.11′)
In this context, CVπ and EVπ are referred to as ex ante (or “option price”) health values because they are defined from a perspective where the relevant health outcomes have not yet been resolved. In contrast, the CV and EV measures in Eqs. (2.3) and (2.4) are ex post health values because they are defined for health changes that are known with certainty.
Because WTP for a specific health change can be expressed and measured either as an ex post or ex ante value, it is important to consider how the two measures are related. For example, one approach to estimating the value of avoiding a chronic illness is to estimate how much an individual with a chronic illness would be willing to pay for a cure. This approach provides an ex post CV measure of going from Hb to Hg (with πb
0 = 1 and πb1 = 0). An
alternative approach is to estimate how much an individual who is at risk of contracting the chronic illness would be willing to pay for a risk reduction. This approach provides an ex ante CVπ measure, such as in Eq. (2.10’), where πb
0 < 1 and πb1 � 0. Dividing CVπ by
Valuation of Morbidity Losses: Meta-Analysis of Willingness-to-Pay and Health Status Measures
2-8
the risk reduction (CVπ /(πb0 – πb
1)) in effect rescales the ex ante measure to provide an estimate of WTP for avoiding a “statistical” case of the illness.
Both approaches provide conceptually valid estimates of private WTP to avoid a case of the illness, but they will not necessarily generate equivalent values. Johannesson (1996) shows that when the ex ante measure is based on reducing the risk of illness to zero (πb
1 = 0), it should be less than or equal to the ex post measure, as long as the individual is risk averse with respect to income (i.e., the marginal utility of income declines as income increases).4 However, ex ante measures may exceed ex post measures in cases where risks are not reduced to zero and where the marginal utility of income is higher in the better health state. Therefore, in many cases it is difficult to establish strong priors regarding the relative magnitudes of the two measures.
2.2.3 Lifetime Utility and QALYs
The QALY framework has been developed primarily to address the need for a simplified and summary measure of individuals’ health-related quality of life (HRQL) over time. Measures of this type are often needed to compare and evaluate health outcomes of alternative treatments or public health programs.
Measuring health status, particularly over the long term, is complicated by the fact that individuals generally experience a variety of “health states” over the course of their life span. The time path of these health states can be captured in a “health profile,” which describes the sequence of health states across time periods. According to the QALY framework, the HRQL corresponding to any possible health state (i = 1 to N) can be represented by a single numerical index value (qi), typically ranging between 0 (death) to 1 (prefect health). If Ti presents the number (or fraction) of life years spent in each health state, i, then the number of QALYs corresponding to a lifetime health profile can be expressed as
QALY = ∑i=1
N qi * Ti (2.12)
4This conclusion is based on a standard expected utility framework. It does not
allow, for example, for the presence of a “certainty premium,” whereby individuals are willing to pay a premium to reduce risks all the way to zero (Viscusi, 1989).
The QALY framework has been developed primarily to address the need for a simplified and summary measure of individuals’ health-related quality of life (HRQL) over time.
Section 2 — Conceptual Framework for Morbidity Valuation
2-9
Because of their relative simplicity as a measure of individuals’ health over an extended period, QALYs have been widely used to compare and evaluate health interventions. Estimates of the number of QALYs gained due to an intervention provide a convenient summary measure of effectiveness; consequently, QALY-based measures have been particularly applied in cost-effectiveness analyses (CEA).
Despite the popularity and relative simplicity of QALYs as a measure of effectiveness, a number of questions have arisen regarding their normative implications. In particular, under what conditions is maximizing QALY gains equivalent to maximizing human welfare? That is, to what extent can the QALY equation described above represent an actual utility function. This question is fundamentally the same as examining the equivalence between CEA and cost-benefit analysis (CBA).
The general conclusion from the literature exploring these issues is that QALYs only represent a valid utility function under very restrictive conditions. Many of the original studies on this topic began with the assumption that preferences over health and longevity could be defined independently of income and other personal characteristics. Even under these conditions, the validity of QALYs as a utility function requires strong assumptions.
Pliskin, Shepard, and Weinstein (1980) define a lifetime utility function, U(q,L), where q represents a scalar index of long-term (e.g., average lifetime) health status and L represents longevity (e.g., number of years in one’s life span). They conclude that the QALY model is consistent with this utility formulation if three main assumptions hold:
Z “Risk neutrality” over life expectancy, which implies, for example, that one is indifferent between (1) living 25 more years with certainty and (2) a gamble offering 50 percent chance of living 50 more years and a 50 percent chance of dying immediately.
Z “Constant proportional trade-off” of longevity for health, which implies, for example, that if one is willing to give up 10 out of 50 years remaining in life for a specific improvement in health, then one should be willing to give up 1 year for the same health improvement if one’s remaining life is 5 years.
Z “Mutual utility independence” between life years and health status, which implies that (1) preferences between lotteries
The general conclusion from the literature exploring these issues is that QALYs only represent a valid utility function under very restrictive conditions.
Valuation of Morbidity Losses: Meta-Analysis of Willingness-to-Pay and Health Status Measures
2-10
involving different health statuses for the same life span do not depend on the length of the life span and (2) preferences between lotteries on life span in a constant health status do not depend on the level of the health status.
A number of studies have demonstrated empirical violations of these assumptions; however, these findings are most likely not sufficient by themselves to invalidate the use of QALYs, at least as an approximation of utility.
Establishing consistency between QALYs and individual preferences is even more difficult if one allows for discounting of future utilities. Johannesson, Pliskin, and Weinstein (1994), for example, use a multiperiod discounted utility model of the form
U = ∑t=0
T qt / (1–d)t (2.13)
where d is a constant discount rate, qt is the HRQL index for period t, and T is a fixed remaining life span. They demonstrate that the QALY approach is not consistent with this commonly assumed lifetime utility structure (with discounted and additive utility). Consistency will only be maintained if one adapts the QALY model to include discounted life years in each health state.
Bleichrodt and Quiggin (1999) extend these analyses even further by evaluating QALYs under conditions where lifetime utility depends not only on health status, but also on consumption (and thus income or wealth) as well. Using a life-cycle model, they conclude that CEA and CBA are mutually consistent (QALY maximization is consistent with utility maximization) if lifetime utility is additive over time and is multiplicative in the utility of consumption and the utility of health status, and if the utility of consumption is constant over time:
U = ∑t=0
T 1 / (1 – d)t * u(c) * qt. (2.14)
If individuals are allowed to optimize lifetime utility by selecting consumption levels in each period, then the main condition under which consumption levels will be constant over time is when the rate of individual time preference is equal to the discount rate. By including consumption in the lifetime utility function, Bleichrodt
Section 2 — Conceptual Framework for Morbidity Valuation
2-11
and Quiggin are also able to explore the relationship between WTP for health improvements and QALY gains. One of the main implications of their analysis is that only under conditions of constant consumption over time is it possible to define a constant WTP per QALY gained, which is necessary for the equivalence between QALY maximization and utility maximization for an individual.5
Given the rather strong assumptions that are required to establish equivalence between QALY maximization, utility maximization, and WTP, a slightly different question is whether QALYs can be included at all within a utility theoretic framework that includes both health status and consumption. In other words, rather than defining conditions under which the QALY function is a utility function or conditions under which QALY maximization is equivalent to utility maximization, is it possible to define a utility-theoretic preference structure that includes both consumption and QALYs as an argument? If so, then how is WTP related to QALY gains?
Hammitt (2002) uses a simplified lifetime utility structure (similar to Pliskin, Shepard, and Weinstein [1980])) to define utility functions for health, longevity, and wealth that are “admissible” if one assumes that preferences for health and longevity can be represented by QALYs. To define admissible functions, he introduces the concept of “HRQL invariance,” which essentially means that it is possible to define a utility index for health that is independent of wealth. For example, he defines the following lifetime utility function with respect to health (H), longevity (L), and wealth (w):
U(H,L,w) = [q(H)L]r a(w) + b(w) (r > 0). (2.15)
Note that, within preference specification, the marginal utility of wealth is not independent of health, nor is the marginal utility of health independent of wealth. Even so, if q(death) = 0 and q(full health) = 1, then q(H*)r will be the utility index (i.e., QALY weight)
5More recently, Dolan and Edlin (2002) have extended this model to include other
nonhealth and nonwealth factors in the utility function. They also take a societal perspective rather than an individual perspective in comparing QALY maximization and welfare maximization. They conclude that in a broad welfare economic framework it is essentially impossible to link CBA and CEA.
Valuation of Morbidity Losses: Meta-Analysis of Willingness-to-Pay and Health Status Measures
2-12
for health state H* estimated by a standard gamble (see Section 2.3.1.1), regardless of wealth level.
Defining QALYs as Q = q(H)*L, Hammitt then explores how the marginal WTP per QALY gain is affected by total QALYs, wealth, and the parameter r, which is like a risk aversion coefficient. The main conclusion is that WTP per QALY is not constant. Under most conditions, WTP per QALY is diminishing with respect to total QALYs and increasing with respect to wealth.
Klose (2002) investigates similar issues as Hammitt (2002) using a multiperiod model of utility. This model defines lifetime utility as the discounted sum utilities across time periods:
U = ∑t=0
T 1 / (1 – d)t *u(Ht, wt) (2.16)
Klose does not specify a functional form for within-period utility, u(w,h); however, he does specify conditions under which it is possible to derive utility indexes for health states (QALY weights) that are consistent with utility theory and independent of wealth level. As in Hammitt’s analysis, defining “wealth-standardized” QALY weights that are independent of wealth does not require the marginal utility of wealth to be independent of health status.
Klose also examines the relationship between WTP and QALY gains and comes to the same fundamental conclusion: WTP per QALY gain is not constant. In particular, as long as health has a positive effect on the marginal utility of wealth, the WTP per QALY gain decreases with health status and with the size of the gain.
In summary, the QALY framework continues to be widely used because of its simplicity and intuitive appeal, but its use as a tool for welfare analysis has raised a number of important concerns. Most analyses indicate that QALYs are not equivalent to the preferred welfare measure, WTP. For this analysis, a more important issue than the equivalence of WTP and QALYs, is whether it is possible to define a consistent relationship between WTP and QALYs. The recent work by Hammitt and Klose, which we have described above, come closest to addressing this issue. The potentially testable hypotheses raised by their analyses are discussed in more detail below.
For this analysis, a more important issue than the equivalence of WTP and QALYs, is whether it is possible to define a consistent relationship between WTP and QALYs.
Section 2 — Conceptual Framework for Morbidity Valuation
2-13
2.3 DETERMINANTS OF HEALTH VALUES Starting with the framework described above, we expect health value estimates to be primarily influenced by
Z the magnitude of changes in health outcomes or health risks;
Z the characteristics of the study population, including
X income effects,
X health status, and
X other socio-demographic characteristics;
Z price effects; and
Z the valuation method used.
Below we describe how these determinants of health values can be measured and how they are expected to affect the magnitude of value estimates.
2.3.1 Measures of Health Changes
As shown by Eqs. (2.3′), (2.4′), (2.10′), and (2.11′), the magnitude of changes in health outcomes (∆H) and/or changes in health risks (∆π) is a key determinant of health values. Therefore, even though health is a complex and multidimensional concept, for the purposes of economic evaluation, it is necessary to develop somewhat simplified characterizations and measures of these factors.
Severity Measures
To characterize and assess health status (i.e., HRQL) in a systematic and standardized way, health economists, psychometricians, and other health experts have developed a wide array of health status measures (HSMs). HSMs generally use standardized questionnaires to assess various aspects of illness or disability. Although some of these measures have been developed for specific diseases, many are designed for more generic use.
Table 2-1 describes how some of the more widely used and broadly applicable (i.e., generic) HSMs are used to classify health status. In each case, they characterize health status (or severity of illness) according to multiple health dimensions, including physical, mental, and emotional elements. In many cases, they also divide each of these dimensions into discrete levels. For example, one component of the Quality of Well-Being (QWB) asks respondents to indicate whether (two levels) they have certain symptoms/problems,
Valuation of Morbidity Losses: Meta-Analysis of Willingness-to-Pay and Health Status Measures
2-14
Table 2-1. Comparison of Health Status Measures
HSM Health Dimension Levels Health States Preference-Based Index
SF-36 Physical functioning, role limitations due to physical health, role limitations due to emotional health, social functioning, bodily pain, mental health, vitality, general health
NA NA No
QWB Mobility, physical activity, social functioning 3 1,170 Yes
27 symptoms/problems 2
HUI-III Vision, hearing, speech, ambulation, dexterity, emotion, cognition, pain/discomfort
5 to 6 972,000 Yes
EuroQol Mobility, self-care, usual activity, pain/discomfort, anxiety/depression
3 243 Yes
Source: Adapted from Brazier, J., M. Deverill, C. Green, R. Harper, and A. Booth. 1999. “A Review of the Use of Health Status Measures in Economic Evaluation.” Health Technology Assessment 3(9).
such as a cough or the need for eyeglasses or contact lenses. A component of the HUI-III asks respondents to rate their speech according to three levels—ability to be (1) completely understood by strangers and friends, (2) partially understood by strangers and completely understood by friends, or (3) partially understood by all. These various dimensions and levels can combine to define a multitude of health states. For example, with five to six possible levels for eight health dimensions, the HUI-III defines as many as 972,000 unique health states.
In addition to these methods for categorizing health states, a number of survey-based scoring techniques have also been developed for measuring and comparing the severity of health states. These techniques use preference elicitation methods to produce “utility weights”—typically varying between 0 (immediate death) and 1 (perfect health)—for specifically defined health states. Three of the most commonly used scoring techniques can be briefly described as follows:
Z Time trade-off (TTO)—asks respondents how many years of perfect health would be equivalent to a given number of years of compromised health.
Z Standard gamble (SG)—asks respondents the risk of death that would make them indifferent between a lottery between death and perfect health relative to the certainty of a given level of compromised health.
Section 2 — Conceptual Framework for Morbidity Valuation
2-15
Z Visual analog scale (VAS)—asks respondents to rate their health on a visual representation of the zero to one scale, commonly in a thermometer-type format.
The last column in Table 2-1 highlights a key difference between the SF-36 and the other HSMs. SF-36 does not provide a “preference-based” measure of health. It does not make use of any of the preference-based scoring techniques described above, and, as a result, it does not provide a systematic way to aggregate the various health dimensions into a single utility index.6 In contrast, the other HSMs in Table 2-1 can all be characterized as preference-based measures (or multiattribute utility scales [MAUSs]) because they have used one or more of these techniques to develop utility scores for specifically defined health states. For instance, the EuroQol team used a combination of VAS and TTO techniques to elicit preferences from a sample population for approximately 45 of the health states. The sample weights were then used to extrapolate and define utility scores for each of the 243 possible health states described by the EuroQol (Gudex et al., 1997).
The advantages and limitations of the various scoring techniques and MAUSs have been widely analyzed in the health economics literature (see Brazier et al. [1999] for a good summary). None of these methods has emerged as a clearly superior alternative for measuring health, but as a group they provide the most promising set of tools for quantifying changes in the severity of health outcomes.
Duration Measures
In addition to severity, the duration of illness can also be an important factor affecting utility. Fortunately, duration is much more straightforward to measure than severity, but selection of the appropriate time scale is still an issue. For example, it may be critical to distinguish between acute and chronic morbidity. Whereas chronic morbidity refers to long-term or recurring conditions that can extend over several years or even a person’s lifetime, acute morbidity refers to discrete and more short-term health events, sometimes lasting 1 day or less. Furthermore, in some cases, acute conditions, such as asthma attacks, may be
6Scoring systems have been developed for the SF-36, for each of its eight health
dimensions and for composite physical and mental health dimensions; however, the resulting scores cannot be interpreted as a utility index.
Valuation of Morbidity Losses: Meta-Analysis of Willingness-to-Pay and Health Status Measures
2-16
directly related to underlying chronic conditions. In other cases, such as foodborne illness, there may be no relation to a long-term condition. Therefore, measures of duration changes may need to account for how both long-term and short-term conditions are affected.
Combined Measures of Severity and Duration
Severity and duration measures, such as the ones described above, can be combined in a multitude of ways to describe specific health outcomes, but the most common method is the QALY approach. As described above in Eq. (2.12), this approach assumes a simple multiplicative relationship between duration and severity.
A QALY measure for a specified health state is constructed by multiplying the time spent (in years) in the health state by its corresponding utility weight. This weight can be estimated directly for a specific health state (or health profile) using one of the preference scoring techniques described above. Alternatively, a weight can be derived from one of the MAUSs, by mapping the health state to the MAUS classification system and calculating or selecting the corresponding weight.
As discussed in detail above, the relative simplicity of the QALY approach has advantages in terms of understandability and ease of use, but it also imposes strict and perhaps unrealistic assumptions about individual preferences. Somewhat less restrictive, but slightly more complex, versions of QALYs have been proposed. For example, Pliskin, Shepard, and Weinstein (1980) define two forms of “risk-adjusted” QALYs, which include a risk-aversion parameter, r:
RA-QALY1 = (q(H)*L)r (2.17a)
RA-QALY2 = q(H)*(L)r (2.17b)
Unless r = 1, both of these forms relax the assumption of linearity with respect to the duration of the health state. Another adaptation of the QALY approach is to include a discounting factor for future time periods. “Discounted QALYs” are calculated by replacing the duration of the health state—number of life-years—with discounted life years (Johannesson, Pliskin, and Weinstein, 1994). These adapted versions are less commonly used because they require
Section 2 — Conceptual Framework for Morbidity Valuation
2-17
additional assumptions about the risk aversion or discount factors; however, they do provide potentially useful alternatives to the simple QALY approach.
Risk Measures
Under conditions of uncertainty, expected utility can be affected, not only by changes in health outcomes, but also by changes in the risk of experiencing poor health states (see Eqs. [2.10′] and [2.11′]). Defining an appropriate metric for health risks is, in general, more straightforward than for health outcomes. In the EU framework, risks are uniquely characterized and measured by the mathematical probabilities (π) associated with each health state. Whether one relies on subjective estimates of these probabilities or more scientifically based (“objective”) estimates is often an issue in measuring risk preferences, but the risk metric is the same in both cases.
Hypotheses for Health Change Measures
To the extent that health changes can be characterized in the three dimensions described above—severity, duration, and risk—economic theory and the conceptual model described in Sections 1 and 2 help to define hypotheses regarding how these dimensions should affect health values. More formally, they inform our expectations about the first and second derivatives of H and π with respect to WTP in Eqs. (2.3′), (2.4′), (2.10′), and (2.11′).
In particular, we expect larger reductions in the severity, duration, and/or risk of illness to generate larger values (positive first derivatives). The test of this relationship is commonly referred to as the “scope test.”
The assumptions underlying the QALY model also define hypotheses regarding how changes in severity and duration should affect values. As discussed in Section 2.2.3, under the most restrictive set of assumptions where QALYs are interpreted as a utility function, WTP should increase in direct proportion to the gain in QALYs. This implies the following expression:
WTP = α * (∆QALY)β, where β = 1 (2.18)
In this expression, α can be interpreted as a constant value per QALY gain. In practice, a constant value per QALY or “QALY
In particular, we expect larger reductions in the severity, duration, and/or risk of illness to generate larger values (positive first derivatives). The test of this relationship is commonly referred to as the “scope test.”
Valuation of Morbidity Losses: Meta-Analysis of Willingness-to-Pay and Health Status Measures
2-18
valuation” approach has been frequently used (see, for example, Zarkin et al. [1993] and Cutler and Richardson [1997]) for valuing national-level health changes. In most of these applications, the value per QALY is derived by annualizing empirical estimates of the value per statistical life (VSL) (Viscusi, 1993), resulting in values of roughly $100,000 per QALY.
Maintaining the constant value per QALY assumption, if the QALY change is composed of a specific change in HRQL (∆q) over a specific time period (∆t), then (Eq. 2.18) can be further decomposed into the following expression:
WTP = α * (∆q *∆t) β, where β =1 (2.19)
This formulation implies that the elasticities of WTP with respect to ∆q and ∆t are both equal to 1.
As shown by Hammitt (2002) and Klose (2002), it is possible to include QALYs (or QALY weights) in a utility theoretic framework without imposing assumptions that imply a constant WTP per QALY. In other words, QALY weights that are invariant to wealth and longevity can be derived from plausible preference structures that include both health and consumption. Under these conditions WTP is systematically but not linearly related to QALY gains. In particular, under most plausible conditions, WTP per QALY gain is declining with respect to health status and the size of the QALY gain.
For ex ante values, the relationships between WTP and risk reductions can be derived from the EU framework. This framework assumes a linear relationship between risks (for specific health states) and EU. Therefore, for relatively small changes in risk (and/or less severe health outcomes), we expect a roughly proportional relationship between WTP and risk changes.7
7With EU, nonlinearity of WTP with respect to risk changes is primarily attributable
to differences in the marginal utility of income across health states. Consequently, one would only expect to observe significant nonlinearities if one is dealing with relatively large changes in health status and relatively large changes in the risk of severe health outcomes.
Section 2 — Conceptual Framework for Morbidity Valuation
2-19
2.3.2 Study Population Characteristics
Values for health changes are also expected to depend importantly on the characteristics of the individual in question, including income, age, and other characteristics.
Income/wealth effects. Income and wealth are primarily expected to have a positive effect on health values. First, individuals’ WTP for health changes is constrained by their available budget; therefore, other things equal and as long as health is a normal good, it is expected that individuals with higher incomes will also have higher values for health changes. Second, as shown in Eq. (2.8), a potentially critical component of WTP is the avoided opportunity cost (lost wages) associated with illness. To the extent that higher incomes imply higher opportunity costs, there is further reason to expect a positive relationship between income and WTP for health.
Age. An individual’s age can potentially affect his/her WTP for health improvements in a number of ways, both positive and negative. First, in general terms, age is negatively related to health status. As discussed above, WTP for a specific health improvement is generally expected to lower when starting from a higher level of health status; therefore, to the extent that age serves as an indicator (inverse) of health status, a positive relationship between WTP and age is expected. Second, age can also affect the nature of the health improvement in a way that decreases WTP. For example, because of a shorter remaining life expectancy, an older individual with moderate arthritis may have a lower total WTP for curing his/her illness than a younger individual with the same condition. In this case, age may serve as a proxy for the duration of the health improvement, in which case it would have a negative effect on WTP. Third, age may capture aspects of socio-economic status that are not captured by annual income measures. For example, elderly individuals may have higher accumulated wealth relative to younger individuals with comparable income. If wealth measures are not adequately controlled for, then age may increase WTP for health improvements through a wealth effect.
Other characteristics. Several other factors can influence people’s perceptions and/or preferences for health and, in turn, their WTP. It is not feasible to catalog all of these possible effects, but a few are worth considering for meta-analysis. First, experience and
Valuation of Morbidity Losses: Meta-Analysis of Willingness-to-Pay and Health Status Measures
2-20
familiarity with a health condition can certainly influence a person’s WTP to avoid the condition, but the effect of these factors on WTP are not always predictable. For example, with experience may come adaptation to the condition and thus a lower WTP. On the other hand, experience may create a greater appreciation for the potential pain and discomfort associated with the condition and thus a higher WTP. Second, education and cognitive abilities can also influence one’s ability to comprehend the risks and outcomes associated with illness. The direction of these effects, however, will depend on the context.
2.3.3 Price Effects
As shown in Eq. (2.8), at least two price effects are potential factors influencing WTP: wages and prices of mitigating goods. Above, in the section on income effects, we discussed how wages can represent opportunity costs of illness. Prices of mitigating goods, such as medication and treatment, can affect the direct cost of illness. Consequently, as shown in Eq. (2.8), higher prices for these goods are expected to increase the WTP to avoid or reduce morbidity. It is important, however, to emphasize that these prices may be partially or totally covered by employers or health insurers, in which case the marginal effect of these prices on private WTP should be smaller.
2.3.4 Valuation Method
The magnitude of value estimates for changes in health is also expected to depend on the way in which they are estimated. A number of nonmarket valuation approaches are potentially applicable for assessing WTP for changes in morbidity outcomes or risks, but most fall under the categories of either stated preference (SP) and revealed preference (RP) approaches.
SP methods such as the contingent valuation method (CVM) and, to a lesser extent, conjoint analysis (CA) are most commonly used for morbidity valuation. Using surveys of individuals, these methods present respondents with hypothetical scenarios involving tradeoffs between monetary gains/losses and health gains/losses. Respondents are then asked to rate, rank, or select their preferred options, or they are asked to directly state their maximum WTP (or minimum WTA) relating to the proposed scenarios. A variety of statistical and econometric techniques can then be used to analyze
Section 2 — Conceptual Framework for Morbidity Valuation
2-21
responses and estimate the size, distribution, and key determinants of values for specified health changes.
In contrast, RP methods rely on information from actual human behavior to estimate individuals’ tradeoffs between monetary gains/losses and health gains/losses. For example, the hedonic approach examines the relationship between market wages and job risks (for injury and fatality) across occupational categories to infer how much income individuals are on average willing to accept/forgo to accept higher/lower risks. The household production approach typically examines individuals’ purchase behavior with respect to goods that allow them to avoid or mitigate adverse health effects (e.g., water filters to avoid drinking water risks or medication to treat symptoms). These behaviors are also used to infer WTP or WTA for health changes. Although value estimates based on RP methods are sometimes given more credence because they are not based on hypothetical settings, in practice they are less widely used. Compared to SP methods, it is generally more difficult with RP methods to acquire the necessary data and to control conditions necessary to isolate the monetary-health tradeoffs of interest.
A number of features and issues associated with SP methods have the potential to influence the values that are estimated with these approaches. The implications for designing and interpreting SP studies have been widely debated and studied in the literature. A complete listing and evaluation of these issues is beyond the scope of this report; however a few of them are worth highlighting:
Z Elicitation format—a key distinction is between open-ended (OE) and dichotomous choice (DC) formats, both of which can induce respondent bias.
Z Protest responses—stated values of zero may reflect scenario rejection rather than true WTP, in which case WTP is downward biased.
Z Response rates—low response rates, particularly for the WTP or choice questions, can induce bias in WTP estimation if nonrespondents have unobserved differences from those who do respond.
Z Payment method—stated values may be influenced by the way in which payment is made (e.g., out of pocket payment versus insurance premium).
Valuation of Morbidity Losses: Meta-Analysis of Willingness-to-Pay and Health Status Measures
2-22
Z Scenario description—the type and level of detail regarding the health effect of interest and the hypothetical choice scenario may influence responses.
Z Scenario comprehension—internal consistency checks can be used to identify and address “invalid” responses that reflect a lack of understanding of the scenario.
Through the application of meta-analysis, it is possible to test for systematic biases associated with some of these points. In particular, the effects of elicitation format can be examined. The relative advantages of OE and DC question formats for CVM studies have been widely discussed and analyzed. Carson (2000), for example, argues in favor of the DC format. He suggests that individuals will tend to understate their WTP with the OE format because respondents do not have the incentive and are not accustomed to conditions where they have to “find” their maximum point. However, if respondents are answering questions strategically, OE responses may overstate or understate WTP (Smith, 2000). For example, if respondents feel that their response will influence the availability of a public or private good without actually affecting the price they have to pay, the OE format offers more of an opportunity to overstate WTP. Meta-analysis offers a framework for testing whether systematic differences do exist between DC- and OE-based estimates of health values.
2.4 SUMMARY This section provides a simple theoretical framework describing how measures of morbidity can be related to individuals’ preference. We have used this framework to identify and describe the factors that are primarily expected to explain and influence estimates of morbidity values. As a result, the framework provides a basis for our empirical model, which is described in Section 5, and for identifying potential hypotheses to be tested in the analysis.
3-1
Analytical Approach— 3 Meta-Analysis
Over the last two decades, a large and diverse body of empirical research in the area of health valuation has emerged in the economics and health literature. In particular, an increasing number of studies have focused on measuring individuals’ willingness to pay for specifically defined health improvements. All of these studies provide potentially useful information for evaluating the benefits of policies designed to improve public health; however, finding ways to integrate all of this information in a systematic way presents a key challenge to policymakers. Meta-analysis provides an approach for addressing this challenge.
Below, we define meta-analysis and describe how it has typically been used to integrate research findings in the area of nonmarket (including health) valuation. Importantly, meta-analysis can be used to define a benefit transfer function. This function can then be applied to estimate values associated with a wide variety of health improvements, including those associated with food safety policies and programs.
3.1 META-ANALYSIS IN NONMARKET VALUATION Meta-analysis refers to the practice of using a collection of formal and informal statistical methods to synthesize the results found in a well-defined class of empirical studies. Glass (1976), in an early review of the method, described it as “the statistical analysis of a large collection of results from individual studies for the purposes of integrating the findings. It connotes a rigorous alternative to the
Valuation of Morbidity Losses: Meta-Analysis of Willingness-to-Pay and Health Status Measures
3-2
casual, narrative discussion of research studies, which typify our attempts to make sense of the rapidly expanding research literature.” It is an analytical approach that has primarily evolved and been most commonly applied in the area of health sciences; however, it is increasingly being used in social sciences, including in the field of economics (Stanley, 2001).
Smith and Pattanayak (2002) provide a recent summary and evaluation of how this set of techniques has been used in the area of nonmarket valuation, including health valuation. They argue that meta-analyses in this field have generally served three main purposes:
Z research synthesis,
Z hypothesis testing, and
Z prediction (benefit transfer).
Research synthesis is the most common objective of these analyses. In contrast to most literature reviews, however, the approach is to define quantitative measures that can be defined consistently across studies and to then provide statistical summaries of these measures. In these cases, the primary measure of interest is the monetary value of a particular welfare effect, such as the average WTP to reduce one’s mortality risk. For example, Viscusi (1993) provides an often cited summary of the VSL literature, which defined the most relevant range of VSL estimates for policy evaluation purposes.
Hypothesis testing takes the statistical analysis of the meta-data one step further. Depending on the context, a number of statistical methods can be used to test a variety of hypotheses. Regression analysis (“meta-regression”) is a tool that is particularly used in economic applications of meta-analysis. In contrast to clinical trial analyses, for example, where conditions are carefully controlled and duplicated across multiple studies, economic studies typically vary in several respects, each of which may exert a significant influence on the result of interest. Regression analysis offers a way to simultaneously control for a variety of factors and to test for how these factors influence the results. In Section 2 of this report, we identified a number of hypotheses regarding factors that we expect to influence health values. In Section 5, we report on the results of meta-regressions which we have used to test some of these hypotheses.
Section 3 — Analytical Approach—Meta-Analysis
3-3
Through the process of meta-regression and hypothesis testing, it is possible to determine whether, across studies, there exists a systematic relationship between key explanatory variables and the primary measure of interest. In the area of nonmarket (e.g., health) valuation, meta-analyses have typically focused on whether WTP estimates vary in systematic and expected ways with respect to the “commodity” of interest and the characteristics of the study population and study methods. To the extent that one is able to uncover such a systematic relationship, the results provide a basis for specifying values for benefit transfer.
One benefit transfer approach is to use the results to define an unconditional average unit-value (with confidence interval), such as the mean value of statistical life (Mrozek and Taylor, 2002). This unit value benefit transfer approach is possible when the commodity of interest is well-defined and relatively homogeneous, and the statistical analysis supports the validity of the underlying value estimates (i.e., to demonstrate that they do not simply represent random “noise”).
Alternatively, it may be possible to use the regression results to define a mean value that is conditional on the type of change and the context of interest. This benefit function transfer approach may be more appropriate when there is substantial heterogeneity across values and the meta-regression analysis is able to account for this variation in a statistically significant manner. For example, Johnson, Fries, and Banzhaf (1997) have conducted a meta-analysis of morbidity valuation studies, and used the results to define a benefit transfer function, which uses measures of duration and severity of illness as explanatory variables. The analysis described in Section 5 of this report uses a similar approach.
3.2 PROCEDURES FOR CONDUCTING META-ANALYSES The steps most commonly used in conducting a meta-analysis are similar to those used in most forms of empirical research with primary data. Researchers have proposed various clusters of essentially the same sets of activity that constitute a meta-analysis (Cooper, 1988; Rosenthal, 1991). Cooper’s approach, perhaps the most used categorization, includes the following primary steps:
Valuation of Morbidity Losses: Meta-Analysis of Willingness-to-Pay and Health Status Measures
3-4
Z problem formulation
Z data collection
Z data evaluation
Z analysis and interpretation
Z public presentation
These are precisely the steps which guide the analysis described in this report. In Sections 1 and 2, we have formulated the primary issues to be addressed and hypotheses to be examined. The next section describes the processes we have used to identify, collect, and organize the valuation data. Selecting and preparing the data for meta-analysis has also required several steps of screening and study evaluation. The process of results of this data evaluation process is also described in Section 4. In Section 5, we discuss and interpret the results of two meta-analyses of value estimates, one for acute health effects and another for chronic effects. This report addresses the final step of the process by communicating and presenting the results of our analysis.
4-1
Data Collection and 4 Evaluation
This section describes the main data collection and evaluation activities that have been conducted to support the meta-analysis and to provide CFSAN with tools that can be used in assessing the health benefits of its regulations. The first major group of activities has been to develop a comprehensive bibliography of health valuation studies from the existing literature and to screen and evaluate these studies with respect to their usefulness for this project. These activities are described in Section 4.1. Next, selecting from the set of studies in our bibliography, we have created a database of health-based WTP estimates. Section 4.2 describes the design and contents of this database.
On a separate but related track, we have also collected information on nonmonetary measures of health. In particular, we have compiled a bibliography of studies that have used established MAUS techniques, such as QWB and EuroQol, to estimate health indexes for selected health effects and populations. We have also created a database that summarizes the results of these studies. The objective of these data collection activities, which are described in Section 4.3, is to provide a consistent set of HSMs that can be used to predict WTP for avoided selected health effects.
Although the bibliographies and databases described in this section have primarily been assembled to support the analysis described in Section 5, their applicability should extend beyond this report. The information currently contained in these files can serve as a more general resource for identifying, summarizing or transferring estimates from the health valuation literature. The databases also
Although the bibliographies and databases described in this section have primarily been assembled to support the analysis described in Section 5, their applicability should extend beyond this report.
Valuation of Morbidity Losses: Meta-Analysis of Willingness-to-Pay and Health Status Measures
4-2
provide organizing structures that can easily be used to include information from more WTP and HSM studies. As such, they can support additional and expanded analyses of the health valuation literature.
4.1 REVIEW AND SELECTION OF STUDIES ON WTP FOR IMPROVED HEALTH To begin identifying and collecting estimates on the WTP for health improvements, we conducted a through literature search and developed an annotated bibliography of WTP for health improvements. This bibliography provides general descriptions of the methods used and issues addressed in the studies. It is primarily designed to screen and evaluate studies for potential inclusion in a meta-analysis.
This section describes the literature search and screening methods we used to identify and select studies. It also provides an overview of the 136 publications that are included in the bibliography.
4.1.1 Literature Search and Screening
Beginning with RTI’s current bibliography of nonmarket and health valuation studies, we systematically identified as many empirical studies as possible that have been applied to or relate in some way to WTP for health improvements. We initially started by looking at previous literature reviews (e.g., Johnson, Fries, and Banzhaf, 1997; Diener, O’Brien, and Gafni, 1998; Olsen and Smith, 2001). We then expanded the search using several search engines, including
Z PubMed—a service of the National Library of Medicine that provides access to MEDLINE citations and life science journals
Z Ingenta—a website specifically designed for searching and delivering research articles (http://www.ingenta.com)
Z Econlit—a detailed indexed bibliography with selected abstracts in economics
These searches were conducted using several key words such as WTP, WTA, health value, health valuation, contingent valuation, and conjoint analysis. We then scanned the reference lists of these articles and used our personal contacts to identify additional candidate studies.
Section 4 — Data Collection and Evaluation
4-3
In screening and selecting studies for inclusion in our bibliography, we focused on empirical studies that examined an individual’s WTP to improve his/her own health. Our focus was on the WTP to reduce morbidity. Thus, many potential candidate studies were excluded, including studies that focused only on the WTP to avoid death (mortality), studies that were purely theoretical, review/summary articles, studies that did not use WTP to value health improvements, studies that primarily valued health-related information, and studies that focused strictly on the COI.
4.1.2 Annotated Bibliography of Selected Studies
We examined over 600 studies to compile this bibliography. Of the publications screened, we currently have 136 in our bibliography.
Appendix A contains an annotated bibliography of the 136 publications; each publication is separately identified and described according to the following fields:
Z Priority Code: This field includes a numerical indicator of the priority that is currently being given to the publication for more thorough review and inclusion in the WTP for health database (see Section 4.2 for more details). The code values have the following meaning:
1. Value estimates from these publications have been included in the WTP for improved health database. (1a indicates publications that were also used in the Johnson, Fries, and Banzhaf [1997] study). They were identified as the most likely candidates for inclusion in a meta-analysis.
2. Value estimates from these publications are the next “in line” to be included in the WTP for health database. For various reasons (e.g., lack of specificity regarding the health change evaluated), these studies are less likely candidates for meta-analysis.
3. These are documents with supporting information to supplement data from other publications (with priority code equal to 1 or 2) in the same study, but it does not provide additional independent value data (i.e., additional records/observations) to the WTP for health database.
Z Study ID Number and Publication ID number: Each study has a separate ID number, and, to the extent that individual studies have resulted in multiple publications, each publication is also numbered. The combination of these two numbers provides a unique identifier for each publication. For example, Mark Dickie and his coauthors conducted a
Valuation of Morbidity Losses: Meta-Analysis of Willingness-to-Pay and Health Status Measures
4-4
single study (study ID = 7) that yielded four publications (pub ID = 1–4), all of which are in our bibliography.
Z Publication Name: This field includes standard bibliographic reference information, including authors names, date, title, etc.
Z Valuation Method: This field describes the empirical valuation method applied in the publication. The studies either use the CVM, conjoint analysis, the hedonic method, or the averting behavior method.
Z Health Effect/Change: This field describes the type of health change that each study was evaluating.
Z Risk Based: This field indicates whether the WTP was for a reduction in the risk of a health outcome (ex ante analysis), or a change in the health outcome itself (ex post analysis).
Z Country: This field indicates the country where the study was conducted.
Z HSM: This field indicates which, if any, nonmonetary HSMs were used in the study to evaluate the same health effect.
The assembled bibliography, summarized in Appendix A, reveals a large and diverse body of empirical literature on this topic. The large number of studies/publications is encouraging, especially when one considers the multiple value estimates that many of the studies contain. This large number of studies makes meta-analysis possible.
4.2 WTP FOR HEALTH DATABASE After collecting and screening the health valuation studies, the next step in preparing for meta-analysis is to design a database that efficiently stores and codes detailed information from each publication. The key field in this database contains the WTP estimate(s) for health improvements from each publication. Several other fields are included to further characterize the valuation methods, sample population, type of illness being considered, and whether the WTP was for a reduction in the probability of an illness (ex ante) or the elimination of an illness once obtained (ex post). This section describes the design of the database and the data entry and verification process. It also summarizes the contents of the database, which includes 389 values from 44 separate publications.
Section 4 — Data Collection and Evaluation
4-5
4.2.1 Database Design
This database is based on a three-level nested structure, as shown in Figure 4-1. We first distinguish between studies and publications, because one study can lead to many publications. We then distinguish between publications and values, because one publication can have many values.
Figure 4-1. Three-Level Database Design
Study i
Publication i1
Publication ij
Publication iM
Value i11
Value i10
Value ij1
Value ijP
Value iM1
Value iMQ
•••
•••
•••
•••
•••
This database currently contains value information from 35 studies and 44 publications. The frequency distribution of publications per study is shown in Table 4-1. Eight of these studies (23 percent) have more than one publication included in the database. The frequency distribution of value estimates per publication is shown in Table 4-2. Most publications (59 percent) have from 1 to 5 value estimates.
The first spreadsheet contains study-level information for each of the publications included in the spreadsheet file. As shown in Table 4-3, this information includes a study identification number, names of the key authors, and names of the primary sponsors/funders of the study (if available and relevant).
Valuation of Morbidity Losses: Meta-Analysis of Willingness-to-Pay and Health Status Measures
4-6
Number of Publications per Study
Number of Studies in Category Percent
1 27 77.1
2 7 20.0
3 1 2.9
Total 35 100.0
Number of Value Estimates per Publication
Number of Publications in Category Percent
1 to 5 26 59.1
6 to 10 6 13.6
11 to 15 2 4.5
16 to 20 6 13.6
21 to 25 1 2.3
26+ 3 6.8
Total 44 100.0
Description Field Type Field Name
Study ID Number integer studyid
Key Author 1 (last name) character kauthor1
Key Author 2 (last name) character kauthor2
Key Author 3 (last name) character kauthor3
Study Sponsor/Funder 1 character sponsor1
Study Sponsor/Funder 2 character sponsor2
Study Sponsor/Funder 3 character sponsor3
The second spreadsheet contains publication-specific information for each of the publications included in the spreadsheet file. This information is listed in Table 4-4. In addition to study and publication identification numbers, it includes author identifiers and information about the data and type of publication.
Table 4-1. Number of Publications per Study
Table 4-2. Number of Value Estimates per Publication
Table 4-3. Study-Level Data Fields (Spreadsheet 1)
Section 4 — Data Collection and Evaluation
4-7
Description Field Type Field Name
Study ID Number integer studyid
Publication ID Number integer pubid
First Author Last Name character pubauthor1
First Author First Name Initial character pubauthorfn1
Second Author Last Name character pubauthor2
Second Author First Name Initial character pubauthorfn2
Third Author Last Name character pubauthor3
Third Author First Name Initial character pubauthorfn3
Fourth Author Last Name character pubauthor4
Fourth Author First Name Initial character pubauthorfn4
Total Number of Authors integer numauthor
Publication Year integer pubyr
Publication Type pubtype
Journal article (peer reviewed) dummy pubjrl
Book dummy pubbk
Book chapter dummy pubbkchap
Technical report dummy pubtech
Working paper dummy pubwp
Ph.D. dissertation dummy pubphdd
Master’s thesis dummy pubmt
Conference presentation dummy pubconf
Other dummy pubother
Other description character pubdes
The remaining seven spreadsheets (3.1 through 3.7) contain information that is specific to the value estimates. This information is split into multiple sheets, primarily to facilitate the entry and viewing of data. The first five data fields in each of these studies are identical—study ID number, publication ID number, value ID number, lead author last name, and publication year. Together, these fields create a unique identifier for each value estimate. In each of the seven sheets, value-specific information is entered, by column, beneath this identifier. Each of these seven sheets has the same value identifiers in each column (e.g., column F contains the same value identifier and includes information about the same value estimate in each sheet).
Table 4-4. Publication-Level Data Fields (Spreadsheet 2)
Valuation of Morbidity Losses: Meta-Analysis of Willingness-to-Pay and Health Status Measures
4-8
The first value-level sheet (spreadsheet 3.1) contains information about the value estimate of WTP for health improvements. The specific data fields are listed in Table 4-5. The WTP for improved health can either be entered as a mean or a median (or both) in dollar terms, and the range and error estimates can also be included. The currency and currency year are also specified here. Farther down, fields are included to distinguish between the value concept. That is, was this the WTP to avoid an illness or the willingness of a person to accept an illness for a certain dollar amount (equivalent or compensating variation measures).
The second value-level sheet (spreadsheet 3.2) contains information about the changes in the health outcome. There are several measures for the change in the intensity/severity of the illness. There are also descriptors to measure the frequency and the duration of the illness. For studies that measured the WTP to reduce the risk of illness (ex ante WTP), there are several descriptors to measure the change in the probability of getting ill. These fields are listed in Table 4-6.
The third value-level sheet (spreadsheet 3.3) contains information on the specific characteristics of the illness. In particular, there is the type of illness, the specific symptoms, and the cause of the illness. These fields are listed in Table 4-7.
The fourth value-level sheet (spreadsheet 3.4) contains information on how the WTP data were collected. It specifies the analysis period and the WTP methods applied. Many types of valuation methods could be used, among them the CVM and the hedonic pricing method. This spreadsheet also specifies the size, socio-demographic, and other characteristics of the study sample/population, along with the methods used to recruit and gather information from this population. These fields are listed in Table 4-8.
Section 4 — Data Collection and Evaluation
4-9
Table 4-5. Value-Level Data Fields (Spreadsheet 3.1)
Description Field Type Field Name
Study ID Number integer studyid
Publication ID Number integer pubid
Value ID Number integer valueid
Lead Author Last Name character pubauthor1
Publication Year integer pubyr
Value
Mean real mean
Outliers trimmed dummy trimmean
Nonresponse corrected dummy corrected
Protest response corrected dummy protcorrected
Inconsistency corrected dummy inconcorrected
Turnbull lower-bound estimate dummy turnbull
Median real median
Low (95%CI) real lowerci
High (95%CI) real upperci
Standard error real stderr
Currency character currency
Currency year integer currencyyr
Payment time frame
Frequency
Every X days integer tfday
One time dummy tfmonth
Duration
X days integer tfyear
Duration of illness dummy tfduration
Permanent dummy tfpermanent
Present value dummy tfpv
Discount rate (%) integer tfpvdiscount
Number of years integer tfpvyrs
(continued)
Valuation of Morbidity Losses: Meta-Analysis of Willingness-to-Pay and Health Status Measures
4-10
Table 4-5. Value-Level Data Fields (Spreadsheet 3.1) (continued)
Description Field Type Field Name
Other dummy tfother
Other description character tfotherdes
Value concept
WTP (compensating variation) dummy wtp
WTP to avoid (equivalent variation) dummy wtpavoid
WTA (equivalent variation) dummy wta
WTA to forgo (compensating variation) dummy wtaforgo
Marginal rate of substitution dummy mrs
Other dummy other
Other description character valueotherdes
Health Change Valued
Description character hcvdes
General health effect
Mortality dummy hcvmortality
Acute morbidity/disability dummy hcvacutemorb
From chronic condition dummy hcvchrocon
From treatment dummy hcvtrtmt
Chronic morbidity/disability dummy hcvchronic
Section 4 — Data Collection and Evaluation
4-11
Table 4-6. Value-Level Data Fields (Spreadsheet 3.2)
Description Field Type Field Name
Study ID Number integer studyid
Publication ID Number integer pubid
Value ID Number integer valueid
Lead Author Last Name character pubauthor1
Publication Year integer pubyr
Ex Post Health Change
Health outcome character exposthlthoc
Number of health outcome measures changed integer hlthocnumchg
Change in intensity/severity dummy sevchgint
Severity measure character sevmsr
Before change level real sevbefore
Before change description character sevbeforedes
After change level real sevafter
After change description character sevafterdes
Change in severity measure character sevchgmsr
Change real sevchg
Severity change description character sevchgdes
Change in duration dummy durationchg
Duration measure character durmsr
Before change level real durbefore
Before change description character durbeforedes
After change level real durafter
After change description character durafterdes
Change in duration measure character durchgmsr
Change real durchg
Duration change description character durchgdes
Change in frequency dummy frequencychg
Frequency measure character freqmsr
Before change level real freqbefore
Before change description character freqbeforedes
After change level real freqafter
After change description character freqafterdes (continued)
Valuation of Morbidity Losses: Meta-Analysis of Willingness-to-Pay and Health Status Measures
4-12
Table 4-6. Value-Level Data Fields (Spreadsheet 3.2) (continued)
Description Field Type Field Name
Change in frequency measure character freqchgmsr
Change real freqchg
Frequency change description character freqchgdes
Ex Ante Health Change dummy exante
Risk-related health outcome character riskhlthoc
Risk measure character riskmsr
Before change risk real riskbefore
After change level real riskafter
Change in risk measure character riskcghmsr
Change real riskchg
Health Status Measure
Visual Analog Scale (VAS) dummy vas
Average baseline real vasavgbasln
Average with change real vasavgwchng
Standard Gamble (SG) dummy sg
Time frame used character sgtimeframe
Average baseline real sgbasln
Average with change real sgavgwchg
Time Trade Off (SG) dummy tto
Time frame used character ttotimeframe
Average baseline real ttobasln
Average with change real ttoavgwchg
Other dummy hsmother
Other description character hsmotherdes
Average baseline real hsmotherbasln
Average with change real hsmotheravgwchg
Section 4 — Data Collection and Evaluation
4-13
Table 4-7. Value-Level Data Fields (Spreadsheet 3.3)
Description Field Type Field Name
Study ID Number Integer studyid
Publication ID Number integer pubid
Value ID Number integer valueid
Lead Author Last Name character pubauthor1
Publication Year integer pubyr
Illness Category
AIDS/HIV dummy aids
Alcoholism and related diseases dummy alcohol
Allergies dummy allergy
Back and neck injuries dummy backneck
Birth defects dummy birfthdef
Blindness dummy blind
Blood disorders (anemia) dummy blood
Bone diseases dummy bone
Bowel syndromes dummy bowel
Cancer dummy cancer
Circulatory/metabolic disorder dummy circulat
Dental conditions dummy dental
Depressive disorders dummy depress
Diabetes diabetes
Digestion and nutrition disorders dummy digest
Ear and eye disorders dummy eareye
Eating disorders dummy eating
Epilepsy dummy epilepsy
Gastrointestinal diseases dummy gastroint
Genetic disorders dummy genetic
Glaucoma dummy glaucoma
Heart (cardiovascular) diseases, stroke, aneurysm dummy heart
Hepatitis dummy hepatitis
High cholesterol dummy cholesterol
Hypertension, anemia, etc. dummy hypertense
Infertility dummy infertility (continued)
Valuation of Morbidity Losses: Meta-Analysis of Willingness-to-Pay and Health Status Measures
4-14
Table 4-7. Value-Level Data Fields (Spreadsheet 3.3) (continued)
Description Field Type Field Name
Leprosy dummy leprosy
Liver and kidney diseases dummy liverkidney
Lupus dummy lupus
Malaria, polio, smallpox, etc. dummy malaria
Meningitis dummy meningitis
Mental disorders dummy mental
Mood disorders dummy mood
Multiple sclerosis (MS) dummy ms
Nervous system disorder dummy nervoussys
Neurological diseases dummy neurological
Obesity dummy obesity
Osteoporosis dummy osteoporosis
Palsy (cerebral and Bell), facial paralysis dummy palsy
Parkinson’s diseases dummy parkinsons
Pregnancy-associated conditions dummy pregnancy
Psoriasis dummy psoriasis
Respiratory diseases dummy respiratory
Schizophrenia dummy schizo
Sexually transmitted diseases (STDs) dummy sexual
Skin diseases dummy skindis
Sleep disorders dummy sleepdis
Speech disorders dummy speechdis
Sports injuries dummy sportsinjuries
Substance abuse and addiction dummy substanceabuse
Thyroid disease dummy thyroid
Not specified dummy illnessnotspec
Other dummy illnessother
Other specify character illnessotherdes
Specific Symptoms
Cough dummy cough
Pain dummy pain
Headache dummy headache
Nausea/stomach upset dummy nausea
Vomiting dummy vomit (continued)
Section 4 — Data Collection and Evaluation
4-15
Table 4-7. Value-Level Data Fields (Spreadsheet 3.3) (continued)
Description Field Type Field Name
Fever/aching dummy fever
Disorientation/light-headedness dummy disorient
Chest pain dummy chestpain
Shortness of breath dummy shortbreath
Throat irritation dummy throat
Eye irritation dummy eyeirritation
Itching dummy itching
Not specified dummy sympnotspec
Other dummy sympother
Other specify character symotherdes
Cause of Health Effect
Environment/air dummy environair
Environment/water dummy environwater
Food/nutrition dummy food
Occupational dummy occupation
Product safety dummy productsafety
Substance abuse dummy causesubstance
Transportation dummy transportation
Natural disaster dummy naturaldisaster
Genetic/hereditary dummy causegenetic
Infectious disease dummy infectious
Not specified dummy causenotspec
Other dummy causeother
Other specify character causeotherdes
Valuation of Morbidity Losses: Meta-Analysis of Willingness-to-Pay and Health Status Measures
4-16
Table 4-8. Value-Level Data Fields (Spreadsheet 3.4)
Description Field Type Field Name
Study ID Number integer studyid
Publication ID Number integer pubid
Value ID Number integer valueid
Lead Author Last Name character pubauthor1
Publication Year integer pubyr
Analysis Period
Year begin integer apyrbegin
Year end integer apyrend
Valuation Method
Contingent valuation dummy cv
Hedonic dummy hedonic
Conjoint/paired comparison dummy conjoint
Averting behavior dummy mktv
Other dummy othervm
Other specify character othervmspec
Study Sample/Population
Unit/number of observations
Individuals/households dummy person
Number integer numperson
Choice occasions dummy choice
Number integer numchoice
Other dummy othersample
Other description character othersampledes
Number integer numothersample
Sample size integer samplesize
Response rate (%) integer responserate
Total number of observations integer numobs
Sampling approach
Country
U.S. dummy sampleus
Canada dummy samplecan
Other dummy sampcntryother
Other specify character sampcntryotherspec
City/state character samplecitystate
Recruitment
Random digit dial dummy randial
Random mailing dummy ranmail (continued)
Section 4 — Data Collection and Evaluation
4-17
Table 4-8. Value-Level Data Fields (Spreadsheet 3.4) (continued)
Description Field Type Field Name
Mall intercept dummy mall
Patient list dummy patientrecruit
Other dummy otherrecruit
Other description character otherrecruitdes
Inclusion criteria
Age dummy inclcritage
Gender dummy inclcritgender
Parent dummy inclcritparent
Race dummy inclcritrace
Health condition dummy inclcrithlthcond
Health condition specify character inclcrithlthspec
Other dummy inclcirtother
Other specify character inclcritotherspec
Sample characteristics
Income
Mean integer incomemean
Median integer incomemedian
Currency year integer incomeyr
Gender (% male) integer gendermale
Ethnicity (% white) integer racewhite
Age
Mean integer agemean
Median integer agemedian
Minimum integer agemin
Maximum integer agemax
Education (average number of years) integer avgedu
Survey method
Mail dummy mail
In-person interview dummy inperson
Telephone dummy phone
Computer interview dummy computer
Internet dummy internet
Other dummy othersurvey
Other description character othersurveydes
Valuation of Morbidity Losses: Meta-Analysis of Willingness-to-Pay and Health Status Measures
4-18
The final three sheets are used to summarize information that is specific to one of the three valuation method categories—stated preference, hedonic, or averting behavior methods. Data are only entered in these fields if the valuation method was used to derive the value estimate. As shown in Table 4-9, the fifth value level sheet (spreadsheet 3.5) contains information for stated preference method applications. In this sheet, the value estimates are categorized according to the manner in which they were elicited from survey respondents and the method proposed in the survey to pay for the improvement in health.
Table 4-9. Value-Level Data Fields—Stated Preference Methods (Spreadsheet 3.5)
Description Field Type Field Name
Study ID Number integer studyid
Publication ID Number integer pubid
Value ID Number integer valueid
Lead Author Last Name character pubauthor1
Publication Year integer pubyr
Elicitation Format
Open ended dummy openend
Closed ended dummy closedend
Dichotomous choice dummy dichochoice
Double bounded dummy doublebond
Iterative bidding dummy bidding
Open-ended follow up dummy followup
Payment card dummy card
Ranking dummy ranking
Rating dummy rating
Other dummy otherformat
Other description character otherformatdes
Payment Vehicle
Tax dummy payctax
Out-of-pocket payment dummy payoutofpock
Insurance payment dummy payinsurance
Cost of living dummy paycsurcharge
User fee dummy paycfee
Voluntary contribution dummy payv
Not specified dummy paynotv
Other dummy paycother
Other specify character paycotherpaydes
Section 4 — Data Collection and Evaluation
4-19
The sixth value level sheet (spreadsheet 3.6) contains information for the hedonic pricing method. This includes information on job characteristics and wage of participants in the sample population. It also includes some methodological variables, including dummy variables to indicate what type of econometric techniques were used. These fields are listed in Table 4-10.
Table 4-10. Value-Level Data Fields—Hedonic Method (Spreadsheet 3.6)
Description Field Type Field Name
Study ID Number integer studyid
Publication ID Number integer pubid
Value ID Number integer valueid
Lead Author Last Name character pubauthor1
Publication Year integer pubyr
Job characteristics
% union integer jobpercunion
Average wage real jobavgwage
Average wage income (per yr) real jobavgwageinc
Average nonwage income (per yr) real jobavginc
Average hrs worked (per yr) real jobavghrs
Other character jobother
Specification
Linear dummy linear
Log-linear (Cobb-Douglas) dummy loglinear
Semi-log dummy semilog
rhs dummy rhs
lhs dummy lhs
Number of risk characteristics integer numriskchar
Number of other job characteristics integer numjobchar
Estimation Method
First Stage dummy stage1
Ordinary least squares (OLS) dummy stage1ols
Other dummy stage1other
Other specify character stage1otherdes
Second Stage dummy stage2
OLS dummy stage2ols
Other dummy stage2other
Other specify character stage2otherdes
Labor income/supply data source character laborincomedata
Job risk data source character jobriskdata
Valuation of Morbidity Losses: Meta-Analysis of Willingness-to-Pay and Health Status Measures
4-20
The seventh value level sheet (spreadsheet 3.7) contains information for the averting behavior method. This includes information on the types of goods purchased, prices, and quantity purchased. It also includes information on total expenditures on goods. These fields are listed in Table 4-11.
Table 4-11. Value-Level Data Fields—Averting Behavior Method (Spreadsheet 3.7)
Description Field Type Field Name
Study ID Number integer studyid
Publication ID Number integer pubid
Value ID Number integer valueid
Lead Author Last Name character pubauthor1
Publication Year integer pubyr
Averting good characteristics
Type of good character typeofgood
Average price per unit real avgunitprice
Average units purchased (per yr) real avgunitpurch
Average expenditure (per yr) real avgexpperyr
Other character otheravert
4.2.2 Data Summary and Evaluation
As previously indicated, the database currently contains value information from 44 publications (and 35 studies), and a brief description of these publications can be found in Appendix A. In this section, we summarize and evaluate the compiled data from these studies. Additional descriptive statistics for the variables included in the database are included in Appendix C.
The distribution of these publications by year and by type of publication is shown in Tables 4-12 and 4-13. The earliest publication date is 1979, and 73 percent of the publications have been published in the last 8 years. Thirty-six of the articles are from peer-reviewed journal articles (81.8 percent).
The database contains a total of 389 WTP estimates from these publications. As shown in Table 4-14, these estimates have been generated using different valuation techniques; however, a large majority have used CVM—77 percent. Due to the focus of our
Section 4 — Data Collection and Evaluation
4-21
Publication Year Number of Publications Percent
1975–1979 1 2.3
1980–1984 0 0.0
1985–1989 8 18.2
1990–1994 3 6.8
1995–1999 17 38.6
2000–2002 15 34.1
Total 44 100.0
Publication Type Number of Publications Percent
Journal article (peer reviewed) 36 81.8
Technical report 5 11.4
Working paper 3 6.8
Total 44 100.0
Valuation Method Number of Value Estimates Percent
Contingent valuation 299 76.9
Conjoint 83 21.3
Averting behavior 7 1.8
Total 389 100.0
analysis, relatively little data from RP studies have thus far been collected and added to the database. Spreadsheets 3.6 and 3.7 include very few data entries as a result. Nonetheless, the database is designed to provide an organizing structure that can be used to include RP data and to support broader analyses of the valuation literature.
Table 4-12. Number of Publications by Year
Table 4-13. Number of Publications by Type of Publication
Table 4-14. Valuation Methods Used (Number of Value Estimates per Method
Valuation of Morbidity Losses: Meta-Analysis of Willingness-to-Pay and Health Status Measures
4-22
This database contains studies from 11 countries, and the distribution of value estimates to each country is shown in Table 4-15. Almost 45 percent of the value estimates are from the United States. Canada has the second highest percentage with 20 percent. Seven European countries—Denmark, Great Britain, Netherlands, Norway, Portugal, Spain, and Sweden—have 29 percent of the value estimates. Australia and Taiwan have the remaining 6 percent of the estimates.
Country Number of Value Estimates Percent
U.S. 177 44.9
Canada 79 20.0
Norway 37 9.4
Sweden 28 7.1
Great Britain 18 4.6
Netherlands 18 4.6
Australia 16 4.1
Taiwan 7 1.8
Portugal 6 1.5
Spain 5 1.3
Denmark 3 0.8
Totala 394 100.0
aFive value estimates were based on samples from both the U.S. and Canada.
The type of health condition that is valued is also considered. The majority of value estimates are considering acute morbidity (79 percent). The distribution of value estimates is shown in Table 4-16. Two publications also included value estimates for avoided mortality. For completeness, these 3 estimates were also included in the database.
Table 4-15. Number of Value Estimates by Country
Section 4 — Data Collection and Evaluation
4-23
Table 4-16. Number of Value Estimates by Type of Health Condition Valued
Health Condition Valued Number of Value Estimates Percent
Acute morbidity 308 79.2
Acute morbidity from chronic condition 31
Acute morbidity from treatment 1
Chronic morbidity 78 20.1
Mortality (only) 3 0.8
Total 389 100.0
4.3 DATABASE OF HEALTH STATUS MEASURES As described above, the database of WTP estimates includes values covering a wide range of acute and chronic health conditions. To integrate and compare these values in a systematic way it is important to identify a common metric for characterizing the various health states. In Section 2, we introduced three possible candidates for such a metric—the QWB Index, the HUI, and the EuroQol. Below we describe each of the MAUSs in more detail.
The QWB Index is of particular interest because this is the metric we have used as a severity index in our meta-analysis of WTP estimates for acute health effects (see Section 5). Nevertheless, we have examined each of these three measures as a way of characterizing chronic health effects. To support our evaluation of these measures, we compiled a bibliography of studies that have applied these measures for selected health effects, and we developed a database that summarizes the results of these studies. Section 4.3.2 provides a description of this database.
4.3.1 Overview of HSMs
QWB
Developed at the University of California San Diego, the QWB grew out of the initial endeavor of creating the General Health Policy Model. The goal was to find a way to assess quality of life using a general instrument, because most of the available instruments at the time were condition specific. Researchers, notably Bush, Kaplan,
Valuation of Morbidity Losses: Meta-Analysis of Willingness-to-Pay and Health Status Measures
4-24
Anderson, and colleagues, began development of the QWB, formerly the Index of Well-Being, in the early 1970s.
Scoring Approach. Health is given a value on a 0.0 (death) to 1.0 (perfect health) point scale, with negative values representing health states worse than death. Individual health is rated on three dimensions: Mobility, Physical Activity, and Social Functioning. Each of these dimensions has three levels on which health can be assessed, varying from no limitations to extreme debilitation. Each level for each dimension corresponds to a weight, which is subtracted from 1.0 (perfect health). Weights increase as disability increases. Another element to the QWB is the list of Symptoms/Problems used to further describe the health state. The worst symptom present out of the 27 possible symptoms determines which weight is used. Included in the list are death (highest weight) and no symptoms (lowest weight). The numerous combinations of dimensions, levels, and symptoms result in a total possible 1,170 health states.
Tables 4-17 and 4-18 help describe the dimensions of the QWB Index. Table 4-17 describes the symptom and problem complexes, which range from excessive worry or anxiety to death. Table 4-18 describes the dimensions for the mobility scale, the physical activity scale, and the social activity scale.
Construction of the Weights. The weights used for the current version of the QWB were developed using an ethnically representative sample of the general community in the San Diego, CA region. Trained scorers interviewed 866 people from 1974 to 1975. Surveys used the VAS, such as a rating scale in the form of a thermometer, to rate respondent valuations of 343 case descriptions. Approximately 100 respondents valued each health state. These values were used to determine the values for other health states using a method of modeling (Spilker, 1996).
Current Use. The current questionnaire is given by interview and measures individual health experienced in the past 6 days. A large sample group allows researchers to find an accurate value for average utility for that health state. The QWB is constantly being tested for validity and reliability (Anderson et al., 1989). One issue with the QWB is that the interviewer-administered version, which takes 15 to 35 minutes to complete, is too cumbersome to use in a
Section 4 — Data Collection and Evaluation
4-25
Table 4-17. Symptom and Problem Complexes (CPX) for the Quality of Well-Being Scale
CPX No. CPX Description Weight
1 Death (not on respondent’s card) –0.727
2 Loss of consciousness such as seizure (fits), fainting, or coma (out cold or knocked out)
–0.407
3 Burn over large areas of face, body, arms, or legs –0.387
4 Pain, bleeding, itching, or discharge (drainage) from sexual organs—does not include normal menstrual bleeding
–0.349
5 Trouble learning, remembering, or thinking clearly –0.340
6 Any combination of one or more hands, feet, arms, or legs either missing, deformed (crooked), paralyzed (unable to move), or broken—includes wearing artificial limbs or braces
–0.333
7 Pain, stiffness, weakness, numbness, or other discomfort in chest, stomach (including hernia or rupture), side, neck, back, hips, or any joints or hands, feet, arms or legs
–0.299
8 Pain, burning, bleeding, itching, or other difficulty with rectum, bowel movements, or urination (passing water)
–0.292
9 Sick or upset stomach, vomiting or loose bowel movement, with or without chills, or aching all over
–0.290
10 General tiredness, weakness, or weight loss –0.259
11 Cough, wheezing or shortness of breath, with or without fever, chills, or aching all over
–0.257
12 Spells of feeling upset, being depressed, or of crying –0.257
13 Headache, or dizziness, or ringing in ears, or spells of feeling hot, nervous or shaky
–0.244
14 Burning or itching rash on large areas of face, body, arms, or legs –0.240
15 Trouble talking, such as lisp, stuttering, hoarseness, or being unable to speak –0.237
16 Pain or discomfort in one or both eyes (such as burning or itching) or any trouble seeing after correction
–0.230
17 Overweight for age and height or skin defect of face, body, arms, or legs, such as scars, pimples, warts, bruises or changes in color
–0.188
18 Pain in ear, tooth, jaw, throat, lips, tongue; several missing or crooked permanent teeth—includes wearing bridges or false teeth
–0.170
19 Took medication or stayed on a prescribed diet for health reasons –0.144
20 Wore eyeglasses or contact lenses –0.101
21 Breathing smog or unpleasant air –0.101
22 No symptoms or problems (not on respondent’s card) –0.000
23 Standard symptom/problem –0.257
24 Trouble sleeping –0.257 (continued)
Valuation of Morbidity Losses: Meta-Analysis of Willingness-to-Pay and Health Status Measures
4-26
Table 4-17. Symptom and Problem Complexes (CPX) for the Quality of Well-Being Scale (continued)
CPX No. CPX Description Weight
25 Intoxication –0.257
26 Problems with sexual interest or performance –0.257
27 Excessive worry or anxiety –0.257
Note: Reproduced with permission from an original supplied by Dr. Kaplan.
Table 4-18. Dimensions, Function Levels, and Weights of the Quality of Well-Being Scale
Step Step Definition Weight
Mobility Scale (MOB)
5 No limitations for health reasons –0.000
4 Did not drive a car, health related; did not ride in a car as usual for age (younger than 15 years), health related and/or did not use public transportation, health related; or had or would have used more help than usual for age to use public transportation, health related
–0.062
2 In hospital, health related –0.090
Physical Activity Scale (PAC)
4 No limitations for health reasons –0.000
3 In wheelchair, moved or controlled movement of wheelchair without help from someone else; or had trouble or did not try to lift, stoop, bend over, or use stairs or includes, health related; and/or had any other physical limitation in walking, or did not try to walk as far as or as fast as others the same age are able, health related
–0.060
1 In wheelchair, did not move or control the movement of wheelchair without help from someone else, or in bed, chair, or couch for most or all of the day, health related
–0.077
Social Activity Scale (SAC)
5 No limitations for health reasons –0.000
4 Limited in other (e.g., recreational) role activity, health related –0.061
3 Limited in major (primary) role activity, health related –0.061
2 Performed no major role activity, health related, but did perform self-care activities
–0.061
1 Performed no major role activity, health related, and did not perform or had more help than usual in performance of one or more self-care activities, health related
–0.106
Note: Reproduced with permission from an original supplied by Dr. Kaplan.
Section 4 — Data Collection and Evaluation
4-27
survey with a large sample size. A self-administered version of the QWB (QWB-SA) that takes only 10 minutes to complete has been developed to solve this problem (Kaplan et al., 1996).
Health Utility Index (HUI)
Designed by Torrance et al. in Canada in 1982, the HUI now encompasses versions 1, 2, and 3. The preference-based scoring system is the component of the HUI that generates the health utility score. The HUI Mark 3 (or HUI-3) is the most often used today.
Scoring Approach. The HUI-3 uses eight dimensions: Vision, Hearing, Speech, Ambulation, Dexterity, Emotion, Cognition, and Pain/Discomfort. The five to six levels of each dimension range from full ability to complete lack of ability. The description of individual health depends on the utility values described by the level in each dimension. All together there are 972,000 health states with utility values ranging from 0.0 to 1.0. Table 4-19 describes the states available in each of the eight dimensions.
Construction of the Weights. The surveys used to determine the weights for the HUI-3 were administered in Hamilton, Ontario, Canada. The VAS and SG were used, through a two-sided feeling thermometer and flip chance board, respectively. Two surveys, the HUI-3 Modelling Survey given to 256 respondents and the HUI-3 Direct Measurement Survey given to 248 respondents, were presented to different groups of people through interviewers (Furlong et al., 1998). Respondents were instructed to assume that the health conditions were chronic and would persist throughout the rest of their lives. Remaining life expectancy was determined and recorded by respondents. A total of 74 health states values were gathered. The MAUF, used to generate utility scores for the remaining health states, was derived from these data using an algebraic method (Furlong et al., 1998).
Current Use. The HUI-3 is the most current version. The versions differ greatly from one another, such as in the number of dimensions used to describe the health states. Many national health surveys have included some form of the HUI, which makes it a valuable tool in gathering large-scale health utilities for a population.
Valuation of Morbidity Losses: Meta-Analysis of Willingness-to-Pay and Health Status Measures
4-28
Table 4-19. Multiattribute Health Status Classification System: Health Utilities Index Mark 3 (HUI-3)
Attribute Description
Vision 1. Able to see well enough to read ordinary newsprint and recognize a friend on the other side of the street, without glasses or contact lenses.
2. Able to see well enough to read ordinary newsprint and recognize a friend on the other side of the street, but with glasses.
3. Able to read ordinary newsprint with or without glasses but unable to recognize a friend on the other side of the street, even with glasses.
4. Able to recognize a friend on the other side of the street with or without glasses but unable to read ordinary newsprint, even with glasses.
5. Unable to read ordinary newsprint and unable to recognize a friend on the other side of the street, even with glasses.
6. Unable to see at all.
Hearing 1. Able to hear what is said in a group conversation with at least three other people, without a hearing aid.
2. Able to hear what is said in a conversation with one other person in a quiet room without a hearing aid, but requires a hearing aid to hear what is said in a group conversation with at least three other people.
3. Able to hear what is said in a conversation with one other person in a quiet room with a hearing aid, and able to hear what is said in a group conversation with at least three other people, with a hearing aid.
4. Able to hear what is said in a conversation with one other person in a quiet room, without a hearing aid, but unable to hear what is said in a group conversation with at least three other people even with a hearing aid.
5. Able to hear what is said in a conversation with one other person in a quiet room with a hearing aid, but unable to hear what is said in a group conversation with at least three other people even with a hearing aid.
6. Unable to hear at all.
Speech 1. Able to be understood completely when speaking with strangers or friends.
2. Able to be understood partially when speaking with strangers but able to be understood completely when speaking with people who know me well.
3. Able to be understood partially when speaking with strangers or people who know me well.
4. Unable to be understood when speaking with strangers but able to be understood partially by people who know me well.
5. Unable to be understood when speaking to other people (or unable to speak at all).
Ambulation 1. Able to walk around the neighborhood without difficulty, and without walking equipment.
2. Able to walk around the neighborhood with difficulty; but does not require walking equipment or the help of another person.
3. Able to walk around the neighborhood with walking equipment, but without the help of another person.
(continued)
Section 4 — Data Collection and Evaluation
4-29
Table 4-19. Multiattribute Health Status Classification System: Health Utilities Index Mark 3 (HUI-3) (continued)
Attribute Description
Ambulation (continued)
4. Able to walk only short distances with walking equipment, and requires a wheelchair to get around the neighborhood.
5. Unable to walk alone, even with walking equipment. Able to walk short distances with the help of another person, and requires a wheelchair to get around the neighborhood.
6. Cannot walk at all.
Dexterity 1. Full use of two hands and ten fingers.
2. Limitations in the use of hands or fingers, but does not require special tools or help of another person.
3. Limitations in the use of hands or finders, is independent with use of special tools (does not require the help of another person).
4. Limitations in the use of hands or fingers, requires the help of another person for some tasks (not independent even with use of special tools).
5. Limitations in use of hands or fingers, requires the help of another person for most tasks (not independent even with use of special tools).
6. Limitations in use of hands or fingers, requires the help of another person for all tasks (not independent even with use of special tools).
Emotion 1. Happy and interested in life.
2. Somewhat happy.
3. Somewhat unhappy.
4. Very unhappy.
5. So unhappy that life is not worthwhile.
Cognition 1. Able to remember most things, think clearly, and solve day-to-day problems.
2. Able to remember most things, but have a little difficulty when trying to think and solve day-to-day problems.
3. Somewhat forgetful, but able to think clearly and solve day-to-day problems.
4. Somewhat forgetful, and have a little difficulty when trying to think or solve day-to- day problems.
5. Very forgetful, and have great difficulty when trying to think or solve day-to-day problems.
6. Unable to remember anything at all, and unable to think or solve day-to-day problems.
Pain/ Discomfort
1. Free of pain and discomfort.
2. Mild to moderate pain that prevents no activities.
3. Moderate pain that prevents a few activities.
4. Moderate to severe pain that prevents some activities.
5. Severe pain that prevents most activities.
Source: Feeny, D., W. Furlong, M. Boyle, and G.W. Torrance. 1995. “Multi-attribute Health Status Classification Systems.” Pharmaco Economics 7(6):490-502.
Valuation of Morbidity Losses: Meta-Analysis of Willingness-to-Pay and Health Status Measures
4-30
EuroQol
The creation of the EuroQol occurred because of a joint effort among many European researchers to develop a general index to measure quality of life. The EuroQol Group began developing the index in 1990. Generally speaking, EuroQol refers to the combination of a five-item questionnaire (the EQ-5D) and a visual analogue rating scale (EQ-VAS).
Scoring Approach. The EQ-5D, the questionnaire portion of the EuroQol, is used to generate a health utility score between 0.0 and 1.0. The five dimensions tested by the questionnaire are meant to cover all aspects of health: Mobility, Self-Care, Usual Activity, Pain/Discomfort, and Anxiety/Depression. The dimensions are subdivided into three levels each, from no problems (level 1) to extreme problems (level 3). A score of 11111 describes perfect health, and a score of 33333 describes the worst health state possible. There are 243 health states defined by the EuroQol. Table 4-20 describes the EuroQol dimensions.
Construction of the Weights. The weights used for the EuroQol (or EQ-5D) are based on the results of a representative survey consisting of 3,395 people in the UK conducted from August to December of 1993. Face-to-face interviews by trained interviewers in respondents’ homes lasting approximately 1 hour were used to present the survey. Respondents were asked to imagine themselves in a given health state for a time period of 10 years. They then rated that health condition using the VAS and TTO methods. Values were elicited for 45 health states (15 states per respondent). Utility weights for the remaining health states were identified using modeling and generalized least squares regression (Dolan, 1997).
Current Use. Both the EQ-5D and EQ-VAS are available for use, although only the EQ-5D produces utility scores. The EQ-5D is self-completed by the respondent and can be completed in a few minutes. The EuroQol is routinely updated and validated by members of the EuroQol group, which has grown to include people in other European countries and around the world.
Section 4 — Data Collection and Evaluation
4-31
Table 4-20. The EuroQol Descriptive System
Attribute Description
Mobility 1. No problems walking about
2. Some problems walking about
3. Confined to bed
Self-Care 1. No problems with self-care
2. Some problems washing or dressing self
3. Unable to wash or dress self
Usual Activities 1. No problems with performing usual activities (e.g., work, study, housework, family, or leisure activities)
2. Some problems with performing usual activities
3. Unable to perform usual activities
Pain/Discomfort 1. No pain or discomfort
2. Moderate pain or discomfort
3. Extreme pain or discomfort
Anxiety/Depression 1. Not anxious or depressed
2. Moderately anxious or depressed
3. Extremely anxious or depressed
Note: For convenience each composite health state has a five-digit code number relating to the relevant level of each dimension, with the dimensions always listed in the order given above. Thus 11223 means: 1 No problems walking about 1 No problems with self-care 2 Some problems with performing usual activities 2 Moderate pain or discomfort 3 Extremely anxious or depressed
Source: Dolan, Paul. 1997. “Modeling Valuations for EuroQol Health States.” Medical Care 35(11):1095-1108.
4.3.2 MAUS Database Description
Each of the previously described measures has been used to quantify, on a zero to one scale, health status associated with a wide variety of health conditions. Therefore, each one provides a potential metric for severity of illness that can used in a meta-analysis of WTP estimates. For example, Section 5 describes how we specifically use the QWB Index in a meta-analysis of WTP estimates for avoiding acute effects.
Because of data limitations and the heterogeneity of individuals’ experiences with long-term illness, identifying an appropriate index for chronic effects is particularly challenging. To address this issue, we searched the relevant literature and compiled a database of estimated MAUS scores for selected illnesses.
Valuation of Morbidity Losses: Meta-Analysis of Willingness-to-Pay and Health Status Measures
4-32
To identify potentially applicable empirical studies we relied on several sources, including existing reviews of the literature (Brazier et al., 1999), on-line databases (primarily PubMed), and reference lists from acquired articles. We focused our data collection efforts on studies that have applied at least one of these measures to estimate an average health index for individuals with selected health conditions. The health conditions that are of particular interest to us are (1) those for which we have corresponding WTP estimates and (2) those that CFSAN has identified as priorities, including diabetes, reactive arthritis, and peanut allergies.
The database contains over 700 scores collected from over 60 studies. For each record, the database includes fields to describe the corresponding
Z health condition (including an ICD-9 code);
Z study reference;
Z MAUS method;
Z mean score, as well as the corresponding ranges (i.e., low and high score, and confidence interval);
Z sample size;
Z average age (and range);
Z gender distribution;
Z study country; and
Z study year.
Table 4-21 provides summary statistics for a selected subset of the database, focusing on 35 health conditions.1 Although the QWB Index, followed by EQ-5D, has been applied in the largest number of studies, the largest number of scores have been estimated using HUI-3, followed by EQ-5D. Partly for this reason, HUI-3 is a primary candidate to be used as a metric in the meta-analysis of WTP estimates for avoiding chronic conditions (see Section 5 for details).
1The database includes a number of estimated scores using EQ-VAS, HUI-1, and
HUI-2, and it includes scores for a wide range of other health conditions that were assigned lower priority. These observations are not included in the statistics for Table 4-21.
Section 4 — Data Collection and Evaluation
4-33
Table 4-21. Summary of MAUS Studies and Scores for Selected Health Conditions
Number of Studies Number of Scores ICD 9 Illness
Category Health Condition QWB HUI-3 EQ-5D QWB HUI-3 EQ-5D 153 Colon cancer 1 0 25 0 201 Hodgkin’s Disease 1 0 0 5 239 AIDS/cancer 2 1 2 2 11 14 246 Thyroid disorder 1 1 0 0 250 Diabetes 3 1 7 15 11 57 277 Cystic fibrosis (CF) 1 1 0 0 340 Multiple sclerosis 1 1 1 0 6 345 Epilepsy 2 2 0 12 3 401 Hypertension 2 1 1 8 11 3 410 Myocardial infarction 1 1 0 0 413 Angina pectoris 1 1 3 0 3 428 Congestive heart failure 2 2 0 0 429 Heart disease 1 0 11 0 436 Stroke 1 0 1 0 490 Bronchitis/emphysema 1 0 11 0 491 Chronic bronchitis 1 1 0 0 492 Emphysema 1 1 0 0 493 Asthma 1 1 2 1 11 9 496 Chronic obstructive pulmonary
disease 2 6 0 0
518 Interstitial lung disease 2 4 0 0 531 Ulcer 1 1 1 11 0 533 Drug rash or diarrhea 1 1 0 0 555 Colitis 1 1 0 0 627 Hormone replacement therapy 1 2 0 0 706 Acne 1 1 0 5 4 714 Rheumatoid arthritis 1 3 1 0 11 715 Osteoarthritis 2 3 5 0 4 716 Arthritis/rheumatism 2 2 1 10 12 1 719 Rheumatic disorder 2 0 0 6 729 Fibromyalgia 1 1 1 0 1 797 Old age disability 1 0 18 0 995 Miscellaneous allergies 1 1 0 0 V15 Food allergy 1 0 11 0 V15 Other allergy (not food) 1 0 11 0 V15 Peanut allergy 0 0 0
Total: 32 17 28 70 172 127
5-1
Meta-Analysis 5 Results
Using information from the collection of valuation studies described in the previous section, RTI conducted two meta-analyses of WTP values for health improvements. The first meta-analysis builds on work by Johnson, Fries, and Banzhaf (1997). It focuses on values for acute conditions and uses regression analysis to explain variation in the value estimates. The second meta-analysis uses a similar approach to analyze value estimates for chronic conditions.
This section presents the results of the two meta-analyses. In both cases, we begin by discussing how the data were selected and analyzed. We then describe in detail the results and implications of the analysis. The findings of the meta-analyses indicate that, for the most part, WTP estimates for avoided adverse health effects vary in systematic and expected ways with respect to key explanatory variables. However, these results are stronger and more conclusive for acute effects than for chronic effects. We describe how the results can be applied for benefit transfer and discuss what they imply about the use of a QALY valuation approach.
5.1 META-ANALYSIS OF VALUE ESTIMATES FOR ACUTE EFFECTS Building on work by Johnson, Fries, and Banzhaf (1997), the analysis described in this section focuses on values for acute conditions. Most importantly, the analysis finds a strong statistical relationship between WTP and measures of severity and duration of acute illness. As a result, the estimated meta-regression functions
Valuation of Morbidity Losses: Meta-Analysis of Willingness-to-Pay and Health Status Measures
5-2
also provide a basis for specifying benefit transfer functions for estimating WTP to avoid acute effects.
5.1.1 Data Selection and Description
As Smith and Pattanayak (2002) note in their review of meta-analyses in nonmarket valuation, “Synthesis requires the ability to define a common concept to be measured” (p. 274). Unfortunately, because of the heterogeneity in valuation approaches and health effects across value estimates (see Section 4), defining such a common concept for all 389 value estimates currently included in the value database is difficult. Instead, using an approach similar to Johnson, Fries, and Banzhaf (1997), we selected a subset of more closely related values and used these in a meta-analysis.
In particular, we selected WTP values if
Z they were estimated for well-defined acute health effects,
Z they were estimated using stated preference methods,
Z the severity of the acute health effect could be expressed as and converted to a QWB score, and
Z the change in duration or frequency of the acute effect could be quantified in terms of discrete days and/or episodes.
We began by selecting the 53 observations included in the Johnson, Fries, and Banzhaf study. These values (see Table A-1 in Appendix A) were taken from five CV studies conducted in the United States in the late 1970s and 1980s. The studies were predominantly conducted for cardio-respiratory health effects associated with air pollution.
We then supplemented these values with 183 additional values taken from 12 other studies. These additional studies were, for the most part, conducted after 1990, and they include research conducted both in the United States and in other countries. In Appendix B, Table B-1 provides more detailed descriptions of some of the key characteristics of these studies.
As is commonly done in meta-regressions analyses, we did not define strict quality criteria for including or excluding studies from the analysis. Rather, we followed the recommendation of Stanley (2001) who writes “when in doubt, it is best to err on the side of inclusion…. Differences in quality, data, or methods do not provide a valid justification for omitting studies [from meta-analysis]” (p.
Section 5 — Meta-Analysis Results
5-3
135). Excluding studies based on quality would require making judgments that might introduce more bias into the results than it would avoid. Moreover, to the extent possible, we have included explanatory variables in the analysis to control for methodological differences.
For the 236 selected values, Tables 5-1 and 5-2 describe and summarize the main variables that we used in the meta-analysis. WTPACUTE was the key variable of interest. It represents individuals’ WTP to avoid or to reduce the duration or frequency of a specific acute condition over the course of a year. All WTP estimates were converted to 2000 dollars using the consumer price index (CPI) and, if they were originally measured in a foreign currency, we first converted the estimates to dollars using the purchasing power parity (PPP) index. Most of the selected studies estimate and report average WTP values. If only median WTP values were reported, we included these estimates in WTPACUTE.
Following the approach used by Johnson, Fries, and Banzhaf, we characterized the change in acute health outcomes associated with each WTP value in two main dimensions. First, we created the variable ∆DAYS to capture changes in the duration or frequency of the health effect. In most cases, this variable represents the reduction in the number of days, over the course of a year, that one experiences a given condition, such as shortness of breath, nausea, or headache. In a relatively small number of cases (N = 17), this variable represents a reduction in the number of acute events, such as asthma, angina, or allergy attacks.1
Second, we used the QWB index to characterize the severity of the acute health effect. As described in Section 4, the QWB index characterizes health outcomes in four dimensions—symptoms, mobility, social activity, and physical activity—each of which can be scored separately. Using the health effect descriptions from the valuation studies and following the approach by Johnson, Fries, and Banzhaf, we assigned each health effect to a defined level for all four dimensions. We then used the premeasured QWB weights (see Tables 4-17 and 4-18) to assign a numerical score to each level.
1In statistical tests of the regression results, the effect of ∆DAYS on WTP was not
found to be significantly different for these observations.
Valuation of Morbidity Losses: Meta-Analysis of Willingness-to-Pay and Health Status Measures
5-4
Table 5-1. Descriptions of Variables Used in the Meta-Analysis
Variables Description
WTPACUTE Mean WTP for health change (in 2000 dollars)a
∆DAYS Reduction in duration (in days) or number of episodes of acute effect
∆QWB Improvement in health-related quality of life on affected days (= 1 – total QWB index)
QWBSYMSCORE QWB Symptom Score (27 symptoms)
QWBMOBSCORE QWB Mobility Score (3 levels)
QWBSACSCORE QWB Social Activity Score (3 levels)
QWBPACSCORE QWB Physical Activity Score (3 levels)
INCOME Mean household income (in 2000 dollars)a
AGE Mean age
% MALE Percent male
US = 1 if study was conducted in the U.S.
WTPAVOID = 1 if value was stated for avoiding a decrease in health
OPEN ENDED = 1 if an open-ended value elicitation method was used
PAYMENT CARD = 1 if a payment card value elicitation method was used
IN PERSON = 1 if survey was conducted as an in-person interview
JOURNAL = 1 if publication was published in a peer-reviewed journal
SAMPLE SIZE Number of respondents used to estimate the WTP value
aConverted to dollars using purchasing power parity (PPP) if in foreign currency; median WTP used if mean value not reported.
The resulting four scores were captured in the variables QWBSYMSCORE, QWBMOBSCORE, QWBSACSCORE, and QWBPACSCORE. A higher score in each of these cases indicates a more severe condition.2 As shown in Table 5-2, the symptom scores are on average higher than the other three, but the standard deviation of this score is slightly less.3
The variable ∆QWB provided a summary measure of the reduction in severity of the health effect. It was calculated as the sum of the four scores. This additive assumption is somewhat arbitrary, but it is
2The absolute values of the weights are used in the analysis. They represent the
amount that is deducted from 1.0 (perfect health) to arrive at the specified level of health.
3For one study (Liu et al., 2000), the health effect of interest could not be mapped into the four separate scores, but an overall QWB score was estimated in the study. As a result, one observation is missing for each of the four scores.
Section 5 — Meta-Analysis Results
5-5
Table 5-2. Summary Statistics for Variables Used in the Meta-Analysis
Variables N Mean SD Min Median Max
WTPACUTE 236 270.22 322.06 2.70 145.62 2927.69
∆DAYS 236 11.9 20.5 1 5 90
∆QWB 236 0.37 0.11 0.17 0.36 0.57
QWBSYMSCORE 235 0.26 0.03 0.17 0.257 0.30
QWBMOBSCORE 235 0.03 0.03 0 0 0.09
QWBSACSCORE 235 0.05 0.04 0 0.061 0.11
QWBPACSCORE 235 0.03 0.03 0 0 0.08
INCOME 236 46,348 13,618 21,891 47,067 88,020
AGE 236 45.2 6.8 35.4 44.5 68
% MALE 236 49.25 14.74 0 47.7 100
US 236 0.39 0.49 0 0 1
WTPAVOID 236 0.88 0.33 0 1 1
OPEN ENDED 236 0.14 0.35 0 0 1
PAYMENT CARD 236 0.29 0.46 0 0 1
IN PERSON 236 0.42 0.49 0 0 1
JOURNAL 236 0.67 0.47 0 1 1
SAMPLE SIZE 236 316.04 151.91 20 399 832
consistent with the way a total QWB index is typically estimated for a specified health condition.4
The variables INCOME, AGE, %MALE, and US were included to account for potentially influential characteristics of the study population. Most studies report summary statistics for these characteristics, but this information is not always reported for the specific subsample that is used to calculate the WTP value. When the subsample information was not available, we used summary statistics for the full study sample.
4The total QWB score is typically calculated as 1 – QWBSYMSCORE –
QWBMOBSCORE – QWBSACSCORE – QWBPACSCORE, such that a higher index represents better health. Therefore, ∆QWB is equal to 1 – the total QWB score.
Valuation of Morbidity Losses: Meta-Analysis of Willingness-to-Pay and Health Status Measures
5-6
The dummy variable WTPAVOID was specified to distinguish between WTP values that were estimated for avoiding a decline in health, as opposed to those for improving health from current conditions. A positive effect for this variable would be consistent with declining marginal utility of health. However, as discussed in Section 2, a large positive effect for this variable may also indicate deviations from the standard utility model, such as the loss aversion and reference dependence models proposed by Tversky and Kahneman (1991).
The remaining dummy variables were included to control for study design effects and for potential publication bias. Finally, we included SAMPLE SIZE, not as an explanatory variable, but as a weighting variable, so that WTP estimates based on larger samples could be given more weight in the regression analysis. The results of including these variables in meta-regressions are described below.
As discussed in Section 2, other factors not included in our list of variables may influence WTP. For example, prices for medical care, wages (opportunity cost of sick time), and average education are all potentially influential factors. Unfortunately, the amount of information reported in the original studies is not adequate to include these factors in the analysis.
5.1.2 Meta-Regression Models and Results
Tables 5-3 through 5-6 describe the regression results for several model specifications, all of which share the same basic structure. We included a measure of WTP in all models as the dependent variable. Measures of the change in duration and severity of the corresponding health effect were included as explanatory variables. Additional explanatory variables include characteristics of the study population, valuation method, study design, and publication outlet.
To explore the robustness of the results across model specifications, 16 sets of results are reported in these tables.5 Four model estimation/specification issues in particular are addressed in the
5The robustness of the results described in these tables is confirmed by the fact that
the size and statistical significance of the key variables change very little when the two highest WTP estimates (over $1,000 each) are excluded from the regression analysis.
Four model estimation/ specification issues in particular were addressed in conducting the meta-analysis: Z functional form
specification
Z health index specification
Z regression weights
Z panel data clustering effects
Section 5 — Meta-Analysis Results
5-7
regression results tables. Each of these issues is described separately below.
Functional form specification. To evaluate the robustness of model results, we applied linear, semi-log, and log-linear specifications to analyze the data. Although all of these approaches are reasonable for approximating the relationship between WTP estimates and the other variables described in Table 5-1, the log-linear approach has a few conceptual advantages. First, it implies that, as changes in severity and duration of illness (and income) approach zero, WTP also approaches zero. Second, it implies that the marginal effect of income on WTP and the marginal effect of the severity/duration change on WTP are not mutually independent. So, for example, the additional WTP associated with a larger health improvement is not assumed to be independent of an individual’s budget. Third, as discussed in more detail below, the log-linear form allows for a more explicit statistical test of the QALY-based valuation approach.
Health index specification. All of the results reported in these tables rely on the QWB index to describe the severity of acute illness. However, using the composite QWB index by itself implies that the four subcomponent scores—the symptoms, mobility, social activity, and physical activity scores—each have the same marginal effect on WTP. A less restrictive specification allows the four scores to enter the functional relationship separately and independently. In this way it is possible to directly test the simple additive assumption underlying the composite QWB score.
Regression weights. For each functional form and health index specification, we used two different weighted regression approaches. In the first approach (No Weight), each value estimate was weighted equally in the regression. This approach is perhaps too restrictive because it does not account for the fact that some estimates are based on larger sample sizes. Larger samples are likely to contribute more “information” to the analysis. Therefore, in the second approach (Sample Size), we weighted each observation in direct proportion to its sample size. Although there is no explicit test of these alternative approaches, we believe that the assumptions underlying the second weighting approach are inherently more defensible.
Valuation of Morbidity Losses: Meta-Analysis of Willingness-to-Pay and Health Status Measures
5-8
Panel data clustering effects. To account for the panel nature of the data, we estimated all of the models using clustered robust regression. Because the data used in the analysis are characterized by multiple observations from individual studies, they are likely to violate the OLS assumptions of independent and identically distributed errors. Therefore, we used clustered regressions, by study ID, to correct the standard error estimates in a way that accounts for error correlation within study clusters and unequal variance of errors across clusters.
Tables 5-3 and 5-4 report results for linear, log-linear, and semi-log specifications of WTPACUTE with respect to ∆DAYS and the single composite QWB score (∆QWB). All of the models show a reasonably good fit, with R-squared statistics between 38 and 65 percent. More importantly, several of the coefficients have the expected sign and are statistically significant.
The results indicate that the WTP estimates “pass” the scope test. The coefficients for ∆DAYS and ∆QWB, in both linear and log forms, are consistently positive and predominantly significant across specifications at a 0.05 level. In other words, average WTP to avoid acute effects increases with the number of days/episodes avoided and with the severity of the conditions avoided. The coefficients for the linear specification with sample size regression weighting (second specification in Table 5-3) indicate that WTP increases by an average of $4.80 for each additional day/episode avoided and increases by $16 for each 0.01 change (between 0 and 1) in the total QWB index. Using a linear specification implies that these marginal values for ∆DAYS and ∆QWB are constant and independent of one another; therefore, these estimates must be interpreted as average effects across the range of ∆DAYS and ∆QWB. The log-linear specification discussed below relaxes these restrictions.
As expected, income also has a consistently positive and, for the most part, significant effect on WTP. The income coefficient in the log-linear specification (last specification in Table 5-4) can be interpreted as an elasticity of WTP with respect to income. The elasticity estimate is 0.7, which implies that income has a positive but relatively inelastic effect on WTP. Notably, controlling for this
Section 5 — Meta-Analysis Results
5-9
Table 5-3. Meta-Regression Results—WTP for Avoided Acute Effects Using the Total QWB Score
Dependent Variable: WTPACUTE (N = 236)
Regression Weight: No Weight Sample Size No Weight Sample Size
Explanatory Variable Coef. t-stata Coef. t-stata Coef. t-stata Coef. t-stata
∆DAYS 5.88 2.78 4.78 3.59
LN(∆DAYS) 106.69 5.51 97.40 4.48
∆QWB 1,577.65 2.97 1,613.27 2.82
LN(∆QWB) 428.20 2.36 454.20 2.23
INCOME 0.00 0.43 0.00 0.57
LN(INCOME) 249.99 2.02 197.20 1.61
AGE 5.19 0.32 –0.96 –0.05
LN(AGE) 432.17 0.71 –58.22 –0.08
%MALE –3.63 –0.93 –2.90 –0.70 –2.64 –0.84 –1.61 –0.58
US 79.35 0.53 18.87 0.14 22.60 0.17 –41.88 –0.38
WTPAVOID –67.07 –0.20 –122.39 –0.34 17.06 0.06 –43.27 –0.15
OPEN ENDED 28.03 0.17 51.77 0.28 57.91 0.43 93.63 0.84
PAYMENT CARD –92.48 –1.20 –15.06 –0.16 –28.85 –0.53 31.94 0.51
IN PERSON –21.50 –0.17 –42.98 –0.33 –48.24 –0.44 –124.55 –1.08
JOURNAL –36.08 –0.24 43.13 0.26 –22.74 –0.15 7.91 0.05
CONSTANT –445.16 –0.46 –256.01 –0.22 –3,623.47 –1.10 –1,168.54 –0.31
R2 38.41% 44.47% 42.88% 48.00%
aBased on robust standard error estimates, corrected for clustering by study ID.
income effect, the dummy variable for studies done in the United States does not have a significant effect (at a 0.05 level) on WTP in any of these specifications.
The average age of the sample also tends to have a positive effect, although the significance of this variable varies across specifications. To the extent that the age is an inverse proxy for health status, this result is consistent with declining marginal utility of health. In other words, this result suggests that older and less healthy individuals are willing to pay more for an increment in health. The estimated size of this age effect is surprisingly large,
Valuation of Morbidity Losses: Meta-Analysis of Willingness-to-Pay and Health Status Measures
5-10
Table 5-4. Meta-Regression Results—WTP for Avoided Acute Effects Using the Total QWB Score
Dependent Variable: LN(WTPACUTE) (N = 236)
Regression Weight: No Weight Sample Size No Weight Sample Size
Explanatory Variable Coef. t-stata Coef. t-stata Coef. t-stata Coef. t-stata
∆DAYS 0.03 3.05 0.02 4.36
LN(∆DAYS) 0.54 10.33 0.50 12.59
∆QWB 6.59 5.69 7.23 4.89
LN(∆QWB) 1.70 3.59 1.97 3.26
INCOME 0.00 1.34 0.00 1.82
LN(INCOME) 0.69 2.17 0.70 2.13
AGE 0.09 2.56 0.09 2.20
LN(AGE) 3.55 3.30 2.56 1.78
%MALE –0.03 –2.12 –0.03 –2.05 –0.01 –1.85 –0.01 –1.36
US –0.11 –0.37 –0.19 –0.64 –0.33 –1.41 –0.41 –1.48
WTPAVOID 1.02 1.63 0.75 0.99 1.12 2.48 0.78 1.33
OPEN ENDED 0.25 0.55 0.33 0.85 0.17 0.60 0.20 0.76
PAYMENT CARD –0.33 –1.30 –0.06 –0.22 –0.21 –1.36 –0.02 –0.09
IN PERSON –0.17 –0.45 –0.11 –0.32 –0.46 –1.29 –0.47 –1.28
JOURNAL –0.84 –1.78 –0.54 –1.37 –0.94 –2.27 –0.71 –1.88
CONSTANT –1.69 –0.73 –1.75 –0.64 –14.26 –2.21 –10.34 –1.29
R2 55.92% 57.19% 65.20% 64.48%
aBased on robust standard error estimates, corrected for clustering by study ID.
however. In the log-linear specification, the elasticity of WTP with respect to age is 2.56.
The coefficient on WTPAVOID is generally not statistically significant. Its sign varies across specifications, although it is more often positive. Therefore, these results do not indicate the presence of loss aversion or strong reference effects.
None of the other variables characterizing the study approach show significant effects on WTP. However, in a few specifications, the JOURNAL coefficient is negative and significant at a 0.10 level,
The assumptions implicit in the QALY valuation approach are not supported by the meta-analysis results.
Section 5 — Meta-Analysis Results
5-11
which suggests that values published in peer-reviewed journals are generally lower.
The results of the log-linear specification (last specification in Table 5-4) can also be used to test the assumptions of the QALY valuation approach for estimating morbidity values. The results do not support these assumptions.
According to the QALY valuation approach, WTP is assumed to increase in direct proportion to the gain in QALYs. This relationship can be expressed as
WTP = α * (∆QALY). (5.1)
In this expression, α can be interpreted as the unit value per QALY (which is often assumed to be approximately $100,000).
Under less restrictive assumptions, the relationship between WTP and QALYs can be expressed as
WTP = α * (∆QALY)β. (5.2)
In other words, the QALY valuation approach restricts β to be equal to one. If β is greater than or equal to one, then WTP will increase more or less than proportionately with respect to QALY gains.
If we assume that the QWB index is an appropriate health utility index, and we assume that the QALY gain of interest is brought about by avoiding a health effect of specific duration and severity, then Eq. (5.2) can be expanded as follows:
WTP = α * (∆DAYS/365 * ∆QWB)β. (5.3)
In this expression, the QALY gain is expressed as the constant per-period utility gain (from less than 1 to 1) times the duration of the utility gain (converted to years). In log-linear form, this equation becomes
ln(WTP) = ln(α) + β*ln(1/365) + β*ln(∆DAYS) +
β*ln(∆QWB). (5.4)
This expression implies that, with a log-linear model, the restrictions imposed by the QALY valuation approach can be directly tested. The null hypothesis is that coefficients on ln(∆DAYS) and ln(∆QWB) are both equal to one. An F-test of this restriction for the log-linear
Valuation of Morbidity Losses: Meta-Analysis of Willingness-to-Pay and Health Status Measures
5-12
specification in Table 5-4 can be strongly rejected (p – value < 0.001). In the last specification, the coefficient estimate for ln(∆DAYS) is considerably less than one (0.5), and the coefficient estimate for ln(∆QWB) is considerably greater than one (1.97). Therefore, even the less restrictive assumptions implied by Eq. (5.2)—that coefficients on ln(∆DAYS) and ln(∆QWB) are equal to one another (but not necessarily equal to one)—can be strongly rejected (p – value = 0.026).
In contrast to the assumptions of Eqs. (5.1) through (5.4), the results of the log-linear specification in Table 5-4 indicate that WTP increases less than proportionately with respect to changes in duration, as measured by ∆DAYS. This result is consistent with a declining marginal disutility with respect to duration of illness. Furthermore, the coefficient for ln(∆QWB) implies that WTP increases more than proportionately with the severity of illness. This result is consistent with a decreasing marginal utility with respect to health status.
Tables 5-5 and 5-6 report results for very similar specifications; however, the main difference is that the QWB score is decomposed into its four components—scores for symptoms, mobility, social activity, and physical activity.
For the most part, the results using the four QWB scores continue to show significant scope effects. The estimated coefficients for the scores are mostly positive and, particularly for the mobility score, are often statistically significant. For example, the results of the second specification in Table 5-5 indicate that average annual WTP increases by $24 and $29 for each 0.01 increment in the physical activity and mobility scores respectively.
The decomposition of ∆QWB in these specifications also allows us to test one of the underlying assumptions of the QWB index (i.e., that each of the four separate dimensions of the index contribute equally to utility [and thus WTP]). In the linear specifications (first two specifications in Table 5-5) and the semilog specification using ln(WTP) as the dependent variable (first two specifications in Table 5-6), ∆QWB is linearly decomposed. Separate coefficients are estimated for each of the four scores.
The results of the decomposed model indicate that the mobility score and the physical activity score have larger and consistently
The meta-analysis results indicate that the QWB mobility and physical activity scores have larger and more statistically significant effects on WTP than the social activity and symptom scores.
Section 5 — Meta-Analysis Results
5-13
significant and positive effects on WTP. The symptom score and, to a lesser extent, the social activity score have lower and generally insignificant effects on WTP. It is worth reemphasizing that the regressions estimate the marginal effects of these scores on private WTP. If individuals are directly compensated for lost work time through sick leave policies, then the additional WTP of employers or society to avoid these conditions is not captured by these measures. If, in contrast, the costs of lost work time were to be fully internalized by the individuals experiencing the health effect, then it is likely that the mobility, physical activity, and social activity scores would have a larger effect on private WTP. It is also possible that the relative effects of these scores would be different (e.g., social activity restrictions might have a larger effect on WTP relative to physical activity restrictions).
This finding—that the different scores have different effects on individuals’ utility and WTP—contradicts the implicit assumptions of the composite ∆QWB score. F-tests of the restriction that all four of the scores have the same marginal effect on WTP can be rejected at a 0.05 level for each of these specifications.
A comparable analysis of the other specifications is somewhat more complicated. In these cases, a linear decomposition of ∆QWB results in nonlinear models. For example, the simple log-linear model can be written as
ln(WTP) = *ln(∆QWB) + α*X, (5.5)
where X represents the vector of all the other explanatory variables. When ∆QWB is linearly decomposed, it results in the following nonlinear form:
ln(WTP) = β*ln( 1*QWBSYMSCORE + 2*QWBMOBSCORE +
3*QWBSACSCORE + 4*QWBPACSCORE ) + α*X, (5.6)
We estimated the nonlinear models using the QWB subcomponent scores using maximum likelihood estimation and report the results in the last four columns of Table 5-5 and Table 5-6. In each of these specifications, one of the coefficients (in this case 4) is restricted to be equal to one. This restriction implies that, if the marginal effects of the four scores on WTP are the same, then the other three coefficients will also equal one. Furthermore, if these
Valuation of Morbidity Losses: Meta-Analysis of Willingness-to-Pay and Health Status Measures
5-14
Table 5-5. Meta-Regression Results—WTP for Avoided Acute Effects Using the Four-Dimensional QWB Scores
Dependent Variable: WTPACUTE (N=235)
Regression Weight: No Weight Sample Size No Weight Sample Size
Explanatory Variable Coef. t-stata Coef. t-stata Coef. t-stata Coef. t-stata
∆DAYS 5.69 2.83 4.66 3.70
LN(∆DAYS) 104.88 7.66 95.09 4.22
βb 221.95 1.40 289.81 1.65
QWBSYSCORE –336.28 –0.51 –88.88 –0.11 0.51c 0.58 0.50c 0.86
QWBMOBSCORE 3,176.32 4.45 2,896.72 5.86 5.10c 0.68 2.78c 1.34
QWBSACSCORE 1,646.13 2.61 880.54 1.15 0.91c 0.59 0.31c 0.93
QWBPACSCORE 1,698.67 1.30 2,380.64 1.86
INCOME 0.00 0.51 0.00 0.62
LN(INCOME) 302.17 2.60 240.60 1.86
AGE 9.74 0.60 1.93 0.10
LN(AGE) 514.78 1.66 52.72 0.07
%MALE –4.59 –1.21 –5.47 –1.39 –3.86 –1.90 –3.87 –1.33
US 126.32 0.84 40.36 0.30 50.83 1.01 –24.52 –0.21
WTPAVOID –110.36 –0.32 –205.62 –0.55 5.15 0.06 –65.84 –0.22
OPEN ENDED 18.21 0.11 40.00 0.22 75.12 0.99 121.37 1.13
PAYMENT CARD –64.28 –0.85 10.67 0.12 –3.30 –0.07 63.47 0.99
IN PERSON 37.25 0.30 –14.55 –0.11 –29.90 –0.43 –119.09 –0.91
JOURNAL –20.31 –0.14 44.73 0.29 0.81 0.01 19.41 0.12
CONSTANT –187.30 –0.18 209.85 0.17 –4,623.95 –2.06 –1,978.06 –0.52
R2 41.19% 47.07% NA NA
aBased on robust standard error estimates, corrected for clustering by study ID. bSee Eq. (5.5) for interpretation of this coefficient. cRefers to η coefficient in Eq. (5.5).
coefficients are all equal to one, then the specification in Eq. (5.6) simplifies to the one Eq. (5.5).
The results of these nonlinear specifications again cast doubt on the assumption underlying the composite QWB score. The coefficients on the mobility and physical activity scores are distinctly larger than for the other two QWB scores. To test the restrictions implied by
Section 5 — Meta-Analysis Results
5-15
Table 5-6. Meta-Regression Results—WTP for Avoided Acute Effects and Four-Dimensional QWB Scores
Dependent Variable: LN(WTPACUTE) (N=235)
Regression Weight: No Weight Sample Size No Weight Sample Size
Explanatory Variable Coef. t-stata Coef. t-stata Coef. t-stata Coef. t-stata
∆DAYS 0.03 3.13 0.02 4.67
LN(∆DAYS) 0.53 10.87 0.49 13.15
βb 0.80 1.88 1.18 1.61
QWBSYSCORE –0.38 –0.17 1.09 0.39 0.26c 0.78 0.47c 0.95
QWBMOBSCORE 11.03 8.58 10.94 9.93 3.00c 1.96 2.73c 3.92
QWBSACSCORE 5.05 1.53 4.88 1.24 0.38c 1.55 0.45c 1.24
QWBPACSCORE 10.16 4.34 10.49 3.65
INCOME 0.00 1.40 0.00 2.08
LN(INCOME) 0.83 2.25 0.88 2.46
AGE 0.11 3.10 0.10 2.70
LN(AGE) 3.90 3.42 3.04 2.16
%MALE –0.03 –2.52 –0.04 –3.84 –0.02 –2.63 –0.03 –3.51
US 0.05 0.17 –0.12 –0.40 –0.22 –0.90 –0.35 –1.27
WTPAVOID 0.86 1.35 0.30 0.43 1.06 2.47 0.57 1.03
OPEN ENDED 0.21 0.53 0.33 0.98 0.24 0.97 0.33 1.48
PAYMENT CARD –0.25 –1.11 0.08 0.29 –0.11 –0.75 0.14 0.68
IN PERSON 0.01 0.03 0.02 0.04 –0.40 –1.06 –0.43 –1.07
JOURNAL –0.82 –1.71 –0.50 –1.31 –0.87 –2.20 –0.65 –1.79
CONSTANT –0.60 –0.26 0.22 0.09 –17.18 –2.24 –13.58 –1.59
R2 58.48% 59.46% NA NA
aBased on robust standard error estimates, corrected for clustering by study ID. bSee Eq. (5.5) for interpretation of this coefficient. cRefers to η coefficient in Eq. (5.5).
using the sum of the four scores (∆QWB) as a composite index, in these cases we can compare the log-likelihood values for the restricted models (in Tables 5-3 and 5-4) with those for the unrestricted models (in Tables 5-5 and 5-6). Based on likelihood
Valuation of Morbidity Losses: Meta-Analysis of Willingness-to-Pay and Health Status Measures
5-16
ratio tests, the restricted models can all be rejected at a 0.05 level of significance.
In the models reported in Tables 5-5 and 5-6, the effects of duration income, age, and other variables are, for the most part, consistent with the previous results. The coefficients on ∆DAYS (linear and log form) are virtually unchanged and consistently positive and significant. The effect of LN(INCOME) is also consistently positive and significant, although the coefficient is somewhat larger. The final specification in Table 5-6 implies an income elasticity of almost 0.9. The estimated elasticity of WTP with respect to age continues to be positive and significant, and once again it is surprisingly large (i.e., greater than 3 in Table 5-6). The effect of gender composition (%MALE) continues to be negative but is somewhat larger (in absolute value) and more significant.
The final (log-linear) specification in Table 5-6 can also be used to test the restrictions implied by the QALY valuation approach. By replacing ∆QWB in Eq. (5.4) with a reweighted composite score based on the results in Table 5-6 ( 1*QWBSYMSCORE +
2*QWBMOBSCORE + 3*QWBSACSCORE + 1*QWBPACSCORE), it is again possible to test the restriction that the coefficient for LN(DAYS) is equal to the parameter. Using a likelihood ratio test, this restriction can again be rejected at a 0.05 level of significance.
5.1.3 Implications of Results for Benefit Transfer
The overall goodness of fit and statistical significance of the models reported in Tables 5-3 through 5-6 suggest that they provide a reasonable foundation for constructing a predictive WTP function. In other words, the models can be used to develop and test alternative benefit transfer functions.
The advantage of being able to specify a WTP function is that it can be used to extrapolate beyond the existing set of WTP estimates in the literature. That is, it can be used to estimate values for any number of avoided acute illnesses, as long as these illnesses can be described according to their severity (in this case, using the QWB classification system) and their duration. In addition, depending on the model specification and econometric results, it can be used to tailor WTP estimates according to the socio-demographic
The overall goodness of fit and statistical significance of the models suggest that they provide a reasonable foundation for constructing a predictive WTP function.
Section 5 — Meta-Analysis Results
5-17
characteristics of the affected population (e.g., average age, gender composition, and/or average income).6
To demonstrate and evaluate how the results can be used to predict WTP for avoiding acute effects under specific conditions, we selected two specifications—the log-linear specification using the composite QWB score (last specification with sample size weighting in Table 5-4) and the log-linear form using the four separate scores (last specification in Table 5-6). We selected these specifications because, as argued previously, they are most defensible from an a priori standpoint and because they provide strong empirical results in terms of goodness-of-fit and statistical significance.
In both cases, we conducted specification tests and dropped the least significant variables. The revised specifications are reported in Table 5-7 as BT Function 1 and BT Function 2, respectively. For BT Function 1, the three variables related to survey method (OPEN ENDED, PAYMENT CARD, and IN PERSON) and the gender proportion variable (%MALE) were dropped from the original specification. Based on a F-test, the coefficients for these four variables were jointly not significantly different from zero. For BT Function 2, only the three survey method variables were dropped from the original specification. Using a likelihood ratio test, a restricted model that also held the %MALE coefficient at zero could not be rejected at a 0.05 level of significance.
To explore the implications of the BT functions under selected out-of-sample conditions, we applied each functions to estimate average individual WTP under eight scenarios. The scenarios and results are summarized in Tables 5-8 and 5-9. The scenarios are defined by specifying values for the relevant explanatory variables.
Specifically, eight combinations of ∆DAYS (1 or 10), ∆QWB (0.1 or 0.4), and age (40 or 60) were used. In all cases, income is set to $45,000 per year, and for BT Function 2 the %MALE is 50 percent. The values are all estimated for the U.S. population (US = 1) and for avoiding health declines (WTPAVOID = 1). Furthermore, assuming
6Developing value estimates that vary according to income levels can raise difficult
ethical issues. Nonetheless, it may be useful, for example, to estimate WTP for a person with average income or to estimate how average WTP will be affected by growth in per-capita income levels.
Valuation of Morbidity Losses: Meta-Analysis of Willingness-to-Pay and Health Status Measures
5-18
Table 5-7. Benefit Transfer Function Estimates
Dependent Variable: LN(WTPACUTE)
BT Function: (Regression Weight):
BT Function 1a (Sample Size)
BT Function 2b (Sample Size)
Explanatory Variable Coefficient t-statc Explanatory Variable Coefficient t-statc
LN(∆DAYS) 0.501 13.07 LN(∆DAYS) 0.477 12.93
LN(∆QWB) 2.339 6.23 βd 1.239 2.22
QWBSYSCORE ( 1)d 0.357 1.31
QWBMOBSCORE ( 2)d 2.120 4.45
QWBSACSCORE ( 3)d 0.476 2.22
QWBPACSCORE ( 4)d 1.000
LN(INCOME) 0.777 3.01 LN(INCOME) 0.833 3.96
LN(AGE) 2.591 2.06 LN(AGE) 2.987 2.49
%MALE –0.017 –2.61
US –0.181 –1.52 US –0.101 –0.73
WTPAVOID 0.799 1.72 WTPAVOID 0.516 1.25
JOURNAL –0.357 –2.05 JOURNAL –0.382 –1.96
CONSTANT –12.031 –1.66 CONSTANT –13.293 –2.02
R2 63.50% NA
N 236 235
aEstimated with Weighted Least Squares Regression (SAMPLE SIZE weight). bEstimated with Weighted Maximum Likelihood Regression (SAMPLE SIZE weight). cBased on robust standard error estimates, corrected for clustering by study ID. dSee Eq. (5.5) for interpretation of coefficient.
that values published in peer-reviewed articles are most defensible, we set JOURNAL equal to one as well.
For BT Function 1, the predicted mean WTP values range from $2.88 (95% C.I.: $1.05 – $4.1) to $669.25 (95% C.I.: $268.61 – $863.03).7 The 10-fold increase in ∆DAYS has a positive but relatively small effect on predicted WTP. In contrast, increasing ∆QWB by a factor of four increases predicted WTP substantially, by
7To appropriately transform the model prediction of mean WTP in logarithmic form
to mean WTP in nonlogarithmic form, we applied a smearing factor to the model predicted values (Duan, 1983). The smearing factor is equal to the average of the exponentiated residuals for the 235-6 observations in the estimation sample.
Generally speaking, the WTP estimates based on the QALY valuation approach are larger than WTP estimates based on the meta-analysis benefit transfer functions.
Section 5 — Meta-Analysis Results
5-19
Ta
ble
5-8
. O
ut-
of-
Sa
mp
le W
TP
Pre
dic
tio
ns
wit
h B
T F
un
cti
on
1
Sc
enar
ios
1
2 3
4 5
6 7
8
∆DA
YS
1 10
1
10
1 10
1
10
∆QW
B
0.1
0.1
0.4
0.4
0.1
0.1
0.4
0.4
INC
OM
E $4
5,00
0 $4
5,00
0 $4
5,00
0 $4
5,00
0 $4
5,00
0 $4
5,00
0 $4
5,00
0 $4
5,00
0
AG
E 40
40
40
40
60
60
60
60
US
1 1
1 1
1 1
1 1
WTP
AV
OID
1
1 1
1 1
1 1
1
JOU
RN
AL
1 1
1 1
1 1
1 1
95%
CI L
ower
$1
.05
$3.5
4 $3
3.48
$1
01.8
8 $1
.45
$4.8
6 $8
2.33
$2
68.6
1
90%
CI L
ower
$1
.17
$3.9
0 $3
6.01
$1
10.3
1 $1
.81
$6.0
2 $9
0.69
$2
94.5
9
Pred
icte
d W
TP
$2.8
8 $9
.15
$73.
81
$234
.03
$8.2
5 $2
6.16
$2
11.0
6 $6
69.2
5
90%
CI U
pper
$3
.68
$11.
09
$78.
29
$256
.99
$19.
42
$58.
77
$254
.21
$786
.94
95%
CI U
pper
$4
.10
$12.
23
$84.
21
$278
.24
$24.
26
$72.
79
$280
.06
$863
.03
QA
LY v
alua
tion
estim
ates
a $2
7.40
$2
73.9
7 $1
09.5
9 $1
,095
.89
$27.
40
$273
.97
$109
.59
$1,0
95.8
9
a (∆D
AY
S/36
5) •
(∆Q
WB
) • $
100,
000.
Valuation of Morbidity Losses: Meta-Analysis of Willingness-to-Pay and Health Status Measures
5-20
Ta
ble
5-9
. O
ut-
of-
Sa
mp
le W
TP
Pre
dic
tio
ns
wit
h B
T F
un
cti
on
2
Sc
enar
ios
1
2 3
4 5
6 7
8
∆DA
YS
1 10
1
10
1 10
1
10
QW
BSY
SCO
RE
0.02
5 0.
025
0.1
0.1
0.02
5 0.
025
0.1
0.1
QW
BM
OB
SCO
RE
0.02
5 0.
025
0.1
0.1
0.02
5 0.
025
0.1
0.1
QW
BSA
CSC
OR
E 0.
025
0.02
5 0.
1 0.
1 0.
025
0.02
5 0.
1 0.
1
QW
BPA
CSC
OR
E 0.
025
0.02
5 0.
1 0.
1 0.
025
0.02
5 0.
1 0.
1
INC
OM
E $4
5,00
0 $4
5,00
0 $4
5,00
0 $4
5,00
0 $4
5,00
0 $4
5,00
0 $4
5,00
0 $4
5,00
0
AG
E 40
40
40
40
60
60
60
60
%M
ALE
50
50
50
50
50
50
50
50
US
1 1
1 1
1 1
1 1
WTP
AV
OID
1
1 1
1 1
1 1
1
JOU
RN
AL
1 1
1 1
1 1
1 1
95%
CI L
ower
b $4
.88
$16.
64
$59.
09
$166
.78
$11.
51
$39.
30
$220
.41
$672
.81
90%
CI L
ower
b $6
.09
$20.
35
$65.
23
$185
.86
$15.
19
$50.
81
$239
.30
$728
.32
Pred
icte
d W
TP
$26.
77
$80.
21
$149
.17
$446
.97
$89.
86
$269
.26
$500
.79
$1,5
00.5
1
90%
CI U
pper
b $6
4.43
$1
73.2
1 $1
86.9
1 $5
88.9
1 $2
91.2
9 $7
81.8
2 $5
74.1
7 $1
,693
.68
95%
CI U
pper
b $8
0.41
$2
11.8
0 $2
06.3
3 $6
56.3
0 $3
84.4
5 $1
,010
.71
$623
.37
$1,8
33.4
2
QA
LY v
alua
tion
estim
ates
a $2
7.40
$2
73.9
7 $1
09.5
9 $1
,095
.89
$27.
40
$273
.97
$109
.59
$1,0
95.8
9
a (∆D
AY
S/36
5) •
(∆Q
WB
) • $
100,
000.
b C
onfid
ence
inte
rval
s w
ere
estim
ated
usi
ng th
e B
ernd
t, H
all,
Hal
l, an
d H
ausm
an (1
974)
asy
mpt
otic
est
imat
or o
f the
cov
aria
nce
mat
rix.
Section 5 — Meta-Analysis Results
5-21
a factor of more than 25. A change in ∆QWB from 0.1 to 0.4 is comparable to the difference between avoiding breathing unpleasant air and avoiding a severe angina attack. Increasing age from 40 to 60 also has a large impact on WTP, increasing it by a factor of roughly 3.
For comparative purposes, value estimates based on a simple QALY valuation approach (using QWB as the health utility index) are also included in Table 5-8 for each scenario. In seven of the eight scenarios, these QALY-based estimates are higher. They are particularly large compared to the BT function estimates when the duration change is large (e.g., 10 days). This difference occurs because the QALY-based approach assumes that values increase in direct proportion to duration, whereas the BT function has a lower elasticity with respect to duration. Also, in contrast to the BT function estimates, the QALY-based values for acute effects do not increase with respect to age. As a result, the QALY-based estimates are less likely to exceed the BT function estimates in the higher age scenarios.
As reported in Table 5-9, similar scenarios are run for BT Function 2, but in this case values for each of the four QWB component scores are specified. Each score is assumed to vary by the same amount, such that the total ∆QWB (sum of the four scores) again varies between 0.1 and 0.4 across scenarios. For these scenarios, the BT Function 2 predictions are consistently higher than the BT Function 1 estimates.8 The predicted WTP values range from $26.77 (95% C.I.: $4.88 – $80.41) to $1500.51 (95% C.I.: $672.81 – $1833.42). The 10-fold increase in ∆DAYS has a similar positive effect with this function. Increasing overall ∆QWB by a factor of four again increases predicted WTP substantially, by a factor of more than 6. Increasing age from 40 to 60 increases WTP by a factor of roughly 3.4.
Table 5-9 also includes comparisons with QALY-based estimates. These estimates are the same as in Table 5-8 because the overall ∆QWB scores are the same for each scenario. In this table, the QALY-based values exceed the BT function values in only four of
8Compared to BT Function 1, BT Function 2 tends to generate higher estimates in
the upper range of WTP and lower estimates in the lower range. The eight scenarios described here tend toward the upper range of WTP.
Valuation of Morbidity Losses: Meta-Analysis of Willingness-to-Pay and Health Status Measures
5-22
the eight scenarios—those with larger duration changes and lower age.
5.2 META-ANALYSIS OF VALUE ESTIMATES FOR CHRONIC EFFECTS Conducting a comparable analysis of WTP estimates for avoiding chronic health effects is hampered by two main factors. First, relatively few WTP estimates are available in the literature. Of the 389 estimates identified and included in our health value database, less than 75 were for avoiding or reducing the risk of chronic health conditions. Second, it is difficult to define a single index (such as the QWB for acute effects) that can be used to characterize all of the chronic effects addressed by these few estimates.
In spite of these limitations, we are able to conduct a small-scale meta-analysis. Through this analysis, we find that WTP estimates for chronic effects are related in expected and statistically significant ways with respect to a few explanatory factors. Although these results are somewhat limited for the purposes of benefit transfer, they do provide a foundation for developing functions to predict WTP for avoiding chronic effects.
5.2.1 Data Selection and Description
To conduct a meta-analysis of values for chronic effects, we initially identified 74 WTP estimates from 15 studies that were candidates for inclusion in the analysis. That is, they estimated individuals’ WTP to either (1) avoid a specific chronic condition with certainty (e.g., a cure for asthma) or (2) reduce the probability of acquiring or experiencing a chronic condition in the future. In other words, we included both ex ante and ex post estimates.
For a variety of reasons, several of the initially identified values needed to be excluded from further consideration. For example, O’Brien and Viramontes (1994) estimate WTP for a treatment of chronic obstructive pulmonary disease (COPD); however, the treatment also would entail a small increase in mortality risk. Because it is not possible to separate the mortality and morbidity values from the resulting WTP estimates, they could not be included. In another case, Krabbe, Essink-Bot, and Bonsel (1997) estimate values for 13 broadly defined conditions; however, their
A meta-analysis of WTP estimates for chronic effects is limited by Z a lack of WTP data
and
Z difficulties in defining a single common health index.
Section 5 — Meta-Analysis Results
5-23
analysis was based on a sample of university students. We judged this sample to be too specialized for inclusion in the meta-analysis.
After this second round of screening, we ended up with 38 value estimates from 10 studies (12 publications). A description of key characteristics of these studies is included in Appendix B, Table B-2. Table 5-10 summarizes the health changes associated with these 38 WTP estimates.
Table 5-10 also includes health index estimates for the chronic health effects addressed in the 38 WTP estimates. Whereas it was relatively straightforward to use a MAUS like the QWB to generate severity scores for selected acute effects, it is generally more difficult to assign severity scores for chronic conditions. One reason is that for many chronic illnesses, such as asthma, the severity (i.e., disutility) of illness can vary substantially across time and across individuals. Fortunately, as described in Section 3, a number of empirical studies have administered surveys using MAUS methods and developed average health scores for a wide variety of health conditions. Three of the most commonly used MAUS methods are the EQ-5D, the HUI-3, and the QWB. Unfortunately, none of the three MAUS methods has been applied to more than 5 of the roughly 10 health conditions addressed by the 38 WTP estimates. As a result, it is not possible to assign illness-specific severity scores to each of the conditions using a common, standardized method.
As an alternative, it is possible to group illnesses more broadly according to three categories of illness—mild, moderate, and severe—using an approach based on Kopec et al. (2000). It is also possible, using the results of that study, to assign average severity scores for these three broad categories. Kopec et al. used a sample of 11,372 Canadians drawn from the 1994/5 National Population Health Survey. This survey included health status questions that corresponded to the HUI classification system. It also inquired about the presence of 20 chronic health conditions. Kopec et al. used the responses to first group individuals into the three general health categories (plus a “none” category). They then used the HUI responses to generate average HUI-3 scores for each category (0.93 for mild, 0.89 for moderate, and 0.806 for severe). As shown in Table 5-10, we used this same basic approach to categorize (and score) the illnesses corresponding to the 38 WTP estimates.
Valuation of Morbidity Losses: Meta-Analysis of Willingness-to-Pay and Health Status Measures
5-24
Table 5-10. Chronic Health Effect Descriptions and Scores
Average MAUS Scores
Publication Country Health Change Description Number of Estimates Severity EQ5D HUI QWB
Blumenschein and Johannesson (1998)
U.S. Cure for asthma 2 Moderate 0.79a 0.84b 0.68c
Kartman et al. (1996) Sweden Reduced risk of experiencing symptoms of reflux oesophagitis
4 Mild
Krupnick and Cropper (1992)
U.S. Reduced risk of chronic bronchitis
2 Severe 0.79d 0.67 c
Reduced risk of chronic respiratory disease as experienced by relative
1 Severe
Lundberg et al. (1999) Sweden Cure for psoriasis 2 Mild 0.92d
Cure for atopic eczema 2 Mild 0.92d
Sloan et al. (1998) U.S. Reduced risk of acquiring multiple sclerosis
8 Severe 0.185e 0.6f
Stavem (1999) Norway Cure for epilepsy 1 Moderate 0.78d
Stavem (2002) Norway Cure for COPD 1 Severe 0.61g
Thompson (1986) U.S. Cure for rheumatoid arthritis (for patients able to climb steps without any difficulty)
1 Moderate 0.73h 0.78d 0.60i
Cure for rheumatoid arthritis (for patients able to climb steps with some difficulty)
1 Moderate 0.47h 0.78d 0.60i
Cure for rheumatoid arthritis (for patients able to climb steps with much difficulty)
1 Severe 0.24h 0.78d 0.60i
Cure for rheumatoid arthritis (for patients unable to climb steps)
1 Severe 0.02h 0.78d 0.60i
Viscusi et al. (1991) U.S. Reduced chronic bronchitis risk
1 Severe 0.79 d 0.67 c
Zethraeus (1998) Sweden Hormone replacement therapy
2 Mild
(continued)
Section 5 — Meta-Analysis Results
5-25
Table 5-10. Chronic Health Effect Descriptions and Scores (continued)
Average MAUS Scores
Publication Country Health Change Description Number of Estimates Severity EQ5D HUI QWB
Zethraeus et al. (1997) Sweden Hormone replacement therapy (for mild menopause)
1 Mild
Hormone replacement therapy (for severe menopause)
1 Moderate
Zillich et al. (2002) U.S. Cure for mild asthma 2 Mild 0.79a 0.84b 0.68c
Cure for moderate asthma 2 Moderate 0.79a 0.84b 0.68c
Cure for severe asthma 2 Severe 0.79a 0.84b 0.68c
aGarratt, A.M., A. Hutchinson, and I. Russell. 2000. “Patient-Assessed Measures of Health Outcome in Asthma: A Comparison of Four Approaches.” Respiratory Medicine 94(6):597-606.
bLeidy, N.K., and C. Coughlin. 1998. “Psychometric Performance of the Asthma Quality of Life Questionnaire in a U.S. Sample.” Quality of Life Research 7(2):127-134.
cFryback, D.G., W.F. Lawrence, P.A. Martin, R. Klein, and B.E. Klein. 1993. “The Beaver Dam Health Outcomes Study: Initial Catalog of Health-State Quality Factors.” Medical Decision Making 13:89-102.
dMittmann, N., K. Trakas, N. Risebrough, and B.A. Liu. 1999. “Utility Scores for Chronic Conditions in a Community-Dwelling Population.” Pharmacoeconomics 15(4):369-376.
eForbes, R.B., A. Lees, N. Waugh, and R. J. Swingler. 1999. “Population Based Cost Utility Study of Interferon Beta-1b in Secondary Progressive Multiple Sclerosis.” British Medical Journal 319(7224):1529-1533.
fSchwartz, C.E., R.M. Kaplan, J.P. Anderson, T. Holbrook, and M.W. Genderson. 1999. “Covariation of Physical and Mental Symptoms Across Illnesses: Results of a Factor Analytic Study.” Annals of Behavioral Medicine 21(2):122-127.
gKaplan, R.M., C.J. Atkins, and R. Timms. 1984. “Validity of a Quality of Well-Being Scale as an Outcome Measure in Chronic Obstructive Pulmonary Disease.” J Chronic Dis 37(2):85-95.
hHurst, N.P., P. Kind, D. Ruta, M. Hunter, and A. Stubbings. 1997. “Measuring Health-Related Quality of Life in Rheumatoid Arthritis: Validity, Responsiveness and Reliability of EuroQol (EQ-5D).” British Journal of Rheumatology 36(5):551-559.
iBombardier, C., and J. Raboud, The Auranofin Cooperating Group. 1991. “A Comparison of Health-Related Quality-of-Life Measures for Rheumatoid Arthritis Research.” Control Clin Trials 12(4 Suppl):243S-256S.
Table 5-11 describes the variables that were created to describe the 38 WTP estimates and were used in a small-scale meta-analysis. In this case, WTPCHRONIC is the key variable of interest. It represents individuals’ WTP to avoid or to reduce the risk of chronic illness. As with WTPACUTE, all WTP estimates (and mean INCOME values) were converted to 2000 dollars using the CPI and, if they were originally measured in a foreign currency, we first converted the estimates to dollars using the PPP index. Special attention was also given to the time and risk dimensions of the
Valuation of Morbidity Losses: Meta-Analysis of Willingness-to-Pay and Health Status Measures
5-26
Table 5-11. Descriptions of Variables Used in the Meta-Analysis
Variables Description
WTPCHRONIC Mean WTP for avoiding chronic health condition (in 2000 dollars)a
MODERATE = 1 if classified as a severe condition (see Table 5-10)
SEVERE = 1 if classified as a moderate condition (see Table 5-10)
HUI-3 Mean HUI (Mark III) score for mild, moderate, and severe conditions (based on Kopec et al., 2000)
INCOME Mean household income (in 2000 dollars)a
AGE Mean age
EXANTE = 1 if value was ex ante
SAMPLESIZE Number of survey respondents used in calculating WTP
aConverted to dollars using PPP if in foreign currency; median WTP used if mean value not reported.
health change. All values were converted to annual WTP to avoid a lifetime case (statistical or actual) of illness. One-time payments were annualized assuming a 20-year payment period and a 5 percent discount rate.9 To approximate WTP for avoiding a “statistical” case of illness, ex ante values for risk reductions were divided by the corresponding reduction in probability of illness. Table 5-12 provides summary statistics for the variables described in Table 5-11. WTPCHRONIC varies considerably, from as little as $595 per year for mild asthma (Zillich et al., 2002) to almost$200,000 per year to avoid a statistical case of chronic bronchitis (Krupnick and Cropper, 1992).
Almost 45 percent of the WTP estimates are for health effects classified as severe, and 21 percent are for moderate effects. Using the results from Kopec et al., this translates to an overall average HUI-3 score of 0.87. About 42 percent of the estimates are ex ante values. The SAMPLESIZE variable, which is used as one of the regression weights, varies between 26 and 400 observations.
9Annual payments for contemporaneous annual risk reductions were assumed to
be equivalent to a one-time payment for a lifetime risk reduction, and they were annualized in the same manner.
Section 5 — Meta-Analysis Results
5-27
Table 5-12. Summary Statistics for Variables Used in the Meta-Analysis
Variables N Mean SD Min Median Max
WTPCHRONIC 38 18,206.53 44,655.65 595.36 3,328.78 194,153.70
MODERATE 38 0.21 0.41 0 0 1
SEVERE 38 0.45 0.50 0 0 1
HUI-3 38 0.87 0.06 0.81 0.89 0.93
INCOME 38 39,185.25 17,437.45 16,308.18 37,765.73 78,545.42
AGE 38 46.21 9.68 24.36 48.9 60
EXANTE 38 0.42 0.50 0 0 1
SAMPLESIZE 38 153.52 125.82 26 87 400
5.2.2 Meta-Regression Models and Results
Table 5-13 describes regression results for a number of model specifications. All of the models were estimated using weighted least squares, and the standard error estimates were corrected to account for clustering by study ID.
Two types of regression weights were included in this analysis. First, as in the analysis of acute effects, estimates based on larger samples were assumed to contain relatively more information. Thus, each estimate was assigned a weight equal to the size of the sample used to estimate the value (weight1 = SAMPLESIZE). Second, several of the WTP studies for chronic effects included multiple estimates for the same health change and the same sample of respondents (i.e., same “group”). For example, in some cases, different values were calculated for the same sample of individuals using different model specifications or assumptions. To avoid assigning too much weight to each of these estimates, they were down-weighted in proportion to the number of estimates coming from the same group (i.e., same sample and health change) (weight2 = 1/GROUPSIZE). The results reported in Table 5-13 include and combine both weights (as the product, weight1*weight2).
As might be expected because of the relatively small number of observations, the regression results are quite sensitive to model specification. Nevertheless, the results do provide some evidence of a systematic and theoretically consistent relationship between
Valuation of Morbidity Losses: Meta-Analysis of Willingness-to-Pay and Health Status Measures
5-28
Table 5-13. Meta-Regression Resultsa—WTP for Avoided Acute Effects and Total QWB Score
Dependent Variable (N=38):
WTPCHRONIC LN(WTPCHRONIC)
(1) (2) (3) (4) (5) Explanatory Variable Coef. t-stata Coef. t-stata Coef. t-stata Coef. t-stata Coef. t-stata
Moderate –16447 –1.19 0.11 0.3
Severe –5182 –0.35 0.65 1.23
HUI-3 –13849 –0.13 –5.60 –1.47
Ln(HUI3) –4.15 –1.3
Income 1.376 3.07 1.054 3.03 0.0001 6.92 0.0001 7.95
Ln(Income) 1.65 6.33
Age –1275 –2.98 –1419 –3.46 0.0027 0.13 0.0018 0.09
Ln(Age) –0.72 –1.01
Ex Ante 12741 1.11 19803 2.2 0.34 0.75 0.39 1.03 0.45 1.24
Constant 32810 1.71 55122 0.5 6.30 6.03 11.56 3.02 –6.38 –1.76
R2 72.41% 70.72% 65.80% 65.74% 64.31%
aWeighted least squares with robust standard error estimates, corrected for clustering by study ID.
WTP and a certain explanatory factors. These relationships are examined in specifications (1) and (2) using WTPCHRONIC as the dependent variables and in specifications (3), (4), and (5) using WTPCHRONIC in logarithmic form.
To test for scope effects with respect to the severity of the chronic condition, we use two approaches. Specifications (1) and (3) control for differences in the severity of illness using two dummy variables—MODERATE and SEVERE. None of the estimated coefficients for these variables are statistically significant at a level less than 0.25, and they have unexpected signs in the first specification. In contrast, specifications (2), (4), and (5) use HUI-3 as an index of severity. In accordance with expectations, the sign of this coefficient is always negative—lower scores imply more severe conditions and higher WTP—and the level of significance varies between 0.18 and 0.22 when WTPCHRONIC is expressed in logarithmic form. Thus, there is some, albeit weak, evidence that the variation in WTP across studies is related to the severity of illness.
Section 5 — Meta-Analysis Results
5-29
The results show consistently positive and statistically significant income effects. Specification (5) estimates an income elasticity of 1.65, which suggests a high sensitivity of WTP to income. Although the level of significance is somewhat lower, the EXANTE coefficient is also consistently positive across specifications. The sign of this effect may be interpreted in different ways. On the one hand, it may relate to cognitive differences in how people perceive of and respond to small changes in probabilities. For example, respondents may have upwardly biased perceptions of small risks and thus overstate their WTP. It may also relate to individuals’ experiences with disease. Ex post valuations are typically asked of individuals with the illness, whereas the opposite is true of ex ante valuations. Therefore, if individuals adapt well to having an illness, this may have a dampening effect on ex post WTP. Furthermore, as discussed for example by Johannesson (1996), the sign of the difference between ex post and ex ante WTP to avoid illness depends importantly on whether the individual is risk averse with respect to income. A positive sign for EXANTE suggests risk aversion.
The sign and size of the age effect vary considerably across specifications. In the first two regression equations, age has a negative and statistically significant effect on WTP. This is the opposite effect from what was observed for acute effects. One reason may be that “background” quality of life is lower for older individuals, such that avoiding a chronic illness would have a lower net positive effect on their utility compared to younger people. Another reason may be that older individuals have lower life expectancy and thus fewer years to enjoy the improved health. When the logarithmic form of WTPCHRONIC is used as a dependent variable in specifications (3) – (5), the effect of age is indeterminate; therefore, it is difficult to draw strong conclusions about how age affects WTP to avoid chronic illness.
The results show consistently positive and statistically significant income effects.
Valuation of Morbidity Losses: Meta-Analysis of Willingness-to-Pay and Health Status Measures
5-30
5.3 SUMMARY AND CONCLUSIONS Generally speaking, the results of our meta-analyses indicate that WTP estimates for avoided adverse health effects vary in systematic and expected ways with respect to key explanatory variables. These results are stronger and more conclusive for acute effects than for chronic effects; however, this difference may be largely due to the larger sample size and the ability to more easily characterize the severity and duration of illness for acute effects.
The analysis of acute effects was based on over 230 WTP estimates from 17 separate stated preference studies, most of which were conducted in the United States, Canada, and Northern Europe. We found a strong statistical relationship between the value estimates and corresponding measures of the severity and duration of the health effects. These results provide evidence to support the theoretical validity of WTP estimates in the literature. Furthermore, we found generally positive and significant income effects and age effects. Positive income effects support the hypothesis that health is a normal good, and positive age effects are consistent with declining marginal utility with respect to health status. However, in most cases, the estimated magnitude of the age effect on WTP is surprisingly high, with an elasticity of around 3.
The regression results for acute effects also provide a simple test of the assumptions underlying the QALY valuation approach for assessing morbidity values. If the composite QWB score is accepted as an appropriate health utility index for calculating QALYs, then the elasticities of WTP with respect to the change in QWB and the change in duration should both be equal to one. This parameter restriction was tested and could be strongly rejected. However, at the same time, the regression results cast doubt on the QWB as an appropriate health utility index. The four component scores were found to have statistically different effects on WTP, which contradicts the assumptions underlying the QWB index.
The meta-analysis for chronic effects is inherently more limited due primarily to data limitations. Nevertheless, based on 38 estimates from 10 studies in the United States, Sweden, and Norway, we did find preliminary indications of systematic and theoretically consistent relationships between WTP and other factors. In particular, we found a strong positive and statistically significant
Generally speaking, the results of our meta-analyses indicate that WTP estimates for avoided adverse health effects vary in systematic and expected ways with respect to key explanatory variables.
Section 5 — Meta-Analysis Results
5-31
effect of income. Using a crude measure of differences in HUI-3 scores between health conditions, we found that WTP estimates were positively related to severity of illness but with low levels of statistical significance.
Although both of these meta-analyses warrant further investigation, the results for acute effects provide a foundation for developing benefit transfer functions to predict WTP. In particular, we have proposed two benefit transfer functions for acute effects and examined how they perform under alternative scenarios. Developing comparable functions for chronic effects will require additional analysis and, most likely, additional data.
6-1
Summary and Discussion of 6 Results
The main purpose of this analysis has been to assist CFSAN in strengthening its capabilities for valuing the health benefits of its regulatory alternatives. Because of constraints on the availability of resources for regulatory analysis, the most practical approach for systematically evaluating a wide range of morbidity impacts is to develop of a flexible and broadly applicable benefit transfer method.
This report discusses the steps that RTI has taken to develop such a benefit transfer tool. These steps include
Z developing a conceptual framework that describes the microeconomic foundations for health valuation, identifies the key expected determinants of health values, and explores the conceptual links between HSMs (and QALYs) and WTP measures;
Z reviewing the empirical literature on health valuation and compiling a detailed bibliography of the most relevant 136 publications;
Z selecting WTP estimates from these studies and constructing a database of health values, which currently contains 389 WTP estimates (and corresponding data) from 44 publications;
Z using meta-regressions to analyze subsets of the database—236 WTP estimates for avoiding acute illness and 38 WTP estimates for avoiding chronic illness; and
Z applying the meta-regression results to specify benefit transfer functions. These functions can be used to predict WTP for avoiding acute illness, based on characteristics of the illness (through the QWB scoring system) and
Valuation of Morbidity Losses: Meta-Analysis of Willingness-to-Pay and Health Status Measures
6-2
characteristics of the affected population (age, gender, income).
Through this process, we have also been able to test hypotheses regarding the determinants of WTP estimates. The results of the meta-analysis indicate that WTP estimates for avoided acute effects vary in systematic and expected ways with respect to key explanatory variables. We find a strong statistical relationship between the value estimates and corresponding measures of the severity (using the QWB scale) and duration of the health effects. In addition, we find generally positive and significant income effects and age effects. Positive income effects support the hypothesis that health is a normal good, and positive age effects are consistent with declining marginal utility with respect to health status.
The meta-analysis results also provide a simple but informative test of the assumptions underlying the QALY valuation approach for assessing morbidity values. The results strongly reject the assumption of a constant value per QALY and the assumption that the duration and the severity of illness have equivalent and proportional effects on WTP.
The results also indicate that the four health utility scores underlying the QWB index have statistically different effects on WTP. The mobility and physical activity dimensions were found to have stronger effects on WTP than the symptoms or social activity dimensions. This finding contradicts the equal weighting assumption, which is typically used in constructing the composite QWB health status index.
It should be emphasized that the strength of these results depends in large part on the strength of the underlying WTP estimates and QWB scores. It is encouraging that the convergent validity of the two measures is supported by the finding that WTP is positively and significantly related to QWB scores. Nevertheless, the results would be weakened if either of these two preference measures includes systematic biases. Although the empirical accuracy and reliability of both methods have been questioned (see Hausman, [1993] for a critique of CVM methods and Brazier [1999] for a review of QWB evaluations), there is no conclusive evidence that such systematic biases exist.
The databases developed for this report can be expanded to include more WTP and HSM studies. Consequently, they should support the continued development of benefit transfer tools.
Section 6 — Summary and Discussion of Results
6-3
Below, we illustrate how the results of the analysis can be applied to estimate the benefits of avoiding specific acute conditions often associated with foodborne illness. However, the data and analyses assembled for this project should extend beyond this report in other ways. The information currently contained in the bibliography and databases described in Section 4 can serve as more general resources for identifying, summarizing, or transferring estimates from the health valuation literature. The databases also provide organizing structures that can easily be used to include information from more WTP and HSM studies. As such, they should support additional analyses (including meta-analyses) of the health valuation literature and continued development of benefit transfer tools.
6.1 ILLUSTRATIVE APPLICATIONS OF THE ESTIMATED BENEFIT TRANSFER FUNCTION FOR ACUTE EFFECTS The final objective of this analysis is to demonstrate how our findings can be used to estimate values for avoiding specific illnesses of interest. At the outset of this project, CFSAN identified four health conditions of particular concern:
Z acute symptoms of foodborne illness,
Z reactive arthritis,
Z diet-related diabetes, and
Z peanut allergy.
Based on our assessment of the data, we have concluded that it is not feasible at this stage to specify a reliable benefit transfer function for avoiding chronic conditions. As discussed in Section 5, the availability of WTP estimates for avoiding chronic illness is currently too limited for this purpose. Consequently, the current analysis does not support estimation of values for avoiding long-term conditions, such as arthritis, diabetes, or allergies.
However, all of the four conditions listed above involve acute outcomes. To the extent that CFSAN’s activities are helpful in limiting the incidence of these acute effects (as opposed to necessarily preventing the associated chronic illness), the benefit transfer functions described in Section 5 can be used to evaluate their benefits.
Valuation of Morbidity Losses: Meta-Analysis of Willingness-to-Pay and Health Status Measures
6-4
BT Function 2 is particularly suited for estimating the benefits of avoiding these types of acute effects (as shown more generally in Table 5-9). According to this function, WTP can be defined as a function of the four QWB scores, duration of illness, income, age, plus a number of other factors. Through our review of the HSM literature, we were not able to identify any studies that have specifically assigned QWB scores for these four categories of foodborne illness. Nevertheless, the descriptions of the QWB health dimensions and levels (see Tables 4-17 and 4-18) are general enough to be used in a few illustrative case examples.
Table 6-1 summarizes the results of applying BT Function 2 to estimate benefits for three illustrative cases:
Z avoiding 10,000 cases of acute gastrointestinal illness (GI), lasting on average 5 days;
Z avoiding 10,000 severe allergic reactions/attacks requiring 5 days of hospitalization; and
Z reducing by 10 the number of days with moderate arthritis symptoms for 5,000 older individuals (averaging 60 years old).
The four QWB scores for each case were assigned using the descriptions in Table 4-17. For acute GI illness we assign QWB symptom (“CPX No.”) 9, corresponding to general stomach ailments, and assume moderate mobility, physical activity, and social activity restrictions. For allergy attack, we assign QWB symptom 2, which includes loss of consciousness, and assume hospitalization with severe mobility, physical activity, and social activity restrictions. For moderate arthritis symptoms, we assign QWB symptom 7, including joint pain, and assume no mobility restrictions (i.e., able to drive as usual) and moderate physical and social activity restrictions.
In all three cases, we assume that the average annual household income of the affected population is roughly equal to the national average in 2000.1 We also assume an equal distribution between males and females affected. For GI illness and allergy attack we assume that the average age of the affected population is close to
1Even if the average income of the affected population is lower or higher than the
U.S. average, it is arguably most appropriate to apply average U.S. income in the transfer function for estimating national benefits.
Section 6 — Summary and Discussion of Results
6-5
Table 6-1. Three Illustrative Applications of the Meta-Analytic Benefit Transfer Function for Acute Effects
Acute Gastro-
intestinal Illness Severe Allergy Attack Moderate Arthritis
Symptom Days
DAYS 5 5 10
QWBSYSCOREa 0.29 (9) 0.407 (2) 0.299 (7)
QWBMOBSCOREa 0.062 (4) 0.09 (2) 0 (5)
QWBSACSCOREa 0.06 (3) 0.077 (1) 0.06 (3)
QWBPACSCOREa 0.061 (3) 0.106 (1) 0.061 (3)
INCOME $57,000 $57,000 $57,000
AGE 40 40 60
%MALE 50 50 50
US 1 1 1
WTPAVOID 1 1 1
JOURNAL 1 1 1
Number of affected individuals 10,000 10,000 5,000
Mean Annual WTP/Personb $306.37 $496.00 $767.71
95% CI Lower $1,185,827 $1,366,987 $3,163,076
90% CI Lower $1,313,872 $1,598,176 $3,470,165
Total Annual Benefitsb $3,063,700 $4,960,000 $3,838,600
90% CI Upper $3,913,835 $8,433,439 $9,305,009
95% CI Upper $4,336,451 $9,859,722 $10,208,391
aNumber in parentheses corresponds to the level of the QWB health dimension (see Table 4-17). bIn 2000 dollars.
the U.S. adult average (40 years old); however, for arthritis, we assume an older population.
By transforming (out of logarithmic form) the results reported in Table 5-7, median WTP per person can be calculated as follows:
Median WTPACUTE = EXP {–13.29 + 0.477*LN(DAYS) +
1.239*LN[QWBPACSCORE + 0.357*QWBSYSSCORE
+ 2.12*QWMOBSCORE + 0.476*QWBSACSORE] +
0.833*LN(INCOME) + 2.987*LN(AGE) – 0.017*%MALE -
0.101*US + 0.516* WTPAVOID – 0.382*JOURNAL} (6.1)
Valuation of Morbidity Losses: Meta-Analysis of Willingness-to-Pay and Health Status Measures
6-6
To properly estimate the mean (expected) WTP reported in Table 6-1, we multiplied the median WTP estimate by the smearing factor of 1.351 (see footnote 5 in Section 5).
Total benefits were then calculated by simply multiplying the estimated mean WTP by the size of the affected population.
6.2 CONCLUSIONS The three case examples summarized in Table 6-1 demonstrate how the estimated meta-regression function can provide CFSAN with a flexible benefit transfer tool for assessing benefits of avoided acute morbidity. This function can be used to assess benefits for any number of avoided health impacts, as long as one is able to specify values for the explanatory variables—in particular the QWB health dimensions and duration of illness—and the size of the affected population.
A number of limitations and uncertainties must also be recognized in applying this function. First, the function provides point estimates of median and/or mean individual WTP. Confidence intervals for the model predictions can also be estimated by taking into account the variance of the estimated parameters. However, other types and sources of uncertainty, such as model specification uncertainty (e.g., functional form) or sample selection effects, will inevitably be present, and they are much more difficult to quantify. As the number of published WTP studies increases, this function can (and should) be re-estimated and refined. This process should improve the precision and accuracy of the WTP predictions, but it will not eliminate the various sources of uncertainty.
Second, the function does not directly address WTP to avoid chronic illness. In our judgment, the number of available WTP estimates is currently too limited to develop a comparable meta-analytic function for chronic illness. Two of the illustrative examples described above—allergy attacks and arthritis symptoms—demonstrate how the transfer function can be used to assess acute effects associated with chronic illness. However, extending this approach to assess long-term changes in acute effects (i.e., over 1 year in duration) would most likely entail an unreasonable extrapolation beyond the range of the data. Furthermore, important differences in values may exist between
Section 6 — Summary and Discussion of Results
6-7
(1) preventing or curing a chronic illness and (2) limiting the number and severity of the acute conditions associated with the chronic illness. As more WTP estimates for avoiding acute and chronic conditions become available, it may eventually be possible to pool these data and formally test this hypothesis.
Third, it is important to reemphasize that the meta-analysis function only includes estimates of adults’ WTP to avoid morbidity outcomes. As a result, it may not include all of the value components that are of potential interest to policy makers. To the extent that costs of illness are externalized to other members of society, for example, through social insurance, this analysis will not capture the benefits of avoiding these external costs. Moreover, the benefit transfer function cannot be applied to specifically estimate values for protecting children’s health. In other words, it would certainly not be appropriate to specify values of less than about 20 years old for the AGE variable in Eq. (6.1).2 Doing so would again entail an unreasonable extrapolation beyond the range of the data.3 Values for children’s health are best measured by assessing parents’ WTP to protect their children. Although the number of studies using this approach is still limited, it would eventually be of interest to include these types of estimates in a meta-analysis.
Finally, the data and analyses assembled for this report have intentionally focused on morbidity-related values, rather than values for reducing mortality risks.4 However, morbidity- and mortality-related values are not always entirely separable. This is especially the case for more serious acute and chronic conditions. Although it is reasonable to assume that the values included in our meta-analysis for acute effects and the WTP predictions based on these functions do not include significant mortality-related values, a more cautious approach will be required for chronic effects. As new WTP data become available for chronic illnesses, it should eventually be possible to develop reliable benefit transfer functions for this category of morbidity. However, to apply these functions
2In the absence of more precise estimates for children, however, it may be useful to
estimate values for young or average age adults and apply them, with all the necessary caveats, to children.
3Note from Table 5-2 that the minimum average age from the estimation sample is 35 years old.
4For good summaries of the mortality valuation literature see, for example, Viscusi (1993) and Mrozek and Tayler (2002).
Valuation of Morbidity Losses: Meta-Analysis of Willingness-to-Pay and Health Status Measures
6-8
for benefits analysis, it will most likely be necessary to separate the mortality-related values from these WTP estimates. Through the continued use of meta-analysis, it should be possible to control for mortality effects.
R-1
References
Anderson, J.P., R.M. Kaplan, C.C Berry, J.M. Bush, and R.G. Rumbaut. 1989. “Interday Reliability of Function Assessment for a Health Status Measure: The Quality of Well-Being Scale.” Medical Care 27(11):1076-1083.
Berndt, E., B. Hall, R. Hall, and J. Hausman. 1974. “Estimation and Inference in Nonlinear Structural Models,” Annals of Economic and Social Measurement, 3/4, 653-665.
Bleichrodt, H., and J. Quiggin. 1999. “When is Cost-Effectiveness Analysis Equivalent to Cost-Benefit Analysis?” Journal of Health Economics 18:681-708.
Blumenschein, K., and M. Johannesson. 1998. “Relationship Between Quality of Life Instruments, Health State Utilities, and Willingness to Pay in Patients with Asthma.” Annals of Allergy, Asthma, and Immunology 80:189-194.
Bombardier, C., and J. Raboud, The Auranofin Cooperating Group. 1991. “A Comparison of Health-Related Quality-of-Life Measures for Rheumatoid Arthritis Research.” Control Clin Trials 12(4 Suppl):243S-256S.
Brazier, J., M. Deverill, C. Green, R. Harper, and A. Booth. 1999. “A Review of the Use of Health Status Measures in Economic Evaluation.” Health Technology Assessment 3(9).
Carson, R.T. 2000. “Contingent Valuation: A User’s Guide.” environ. Sci. Technol. 34:1413-1418.
Cooper, Harris M. 1988. “Organizing Knowledge Synthesis: A Taxonomy of Literature Reviews.” Knowledge in Society 1:104-126.
Cutler, D.M., and E. Richardson. 1997. “Measuring the Health of the U.S. Population.” Brookings Papers on Economic Activity: Microeconomics.
Valuation of Morbidity Losses: Meta-Analysis of Willingness-to-Pay and Health Status Measures
R-2
Diener, A., B. O’Brien, and A. Gafni. 1998. “Health Care Contingent Valuation Studies: A Review and Classification of the Literature.” Health Economics 7:313-326.
Dolan, P. 1997. “Modeling Valuations for EuroQol Health States.” Medical Care 35(11):1095-1108.
Dolan, P., and R. Edlin. 2002. “Is It Really Possible to Build a Bridge Between Cost-Benefit Analysis and Cost-Effectiveness Analysis?” Journal of Health Economics 21(5):827-843.
Duan, N. 1983. “Smearing Estimate: A Nonparametric Retransformation Method.” Journal of the American Statistical Association 78(383):605-610.
Feeny, D., W. Furlong, M. Boyle, and G.W. Torrance. 1995. “Multi-attribute Health Status Classification Systems.” Pharmaco Economics 7(6):490-502.
Forbes, R.B., A. Lees, N. Waugh, and R. J. Swingler. 1999. “Population Based Cost Utility Study of Interferon Beta-1b in Secondary Progressive Multiple Sclerosis.” British Medical Journal 319(7224):1529-1533.
Freeman, A. Myrick III. 1993. The Measurement of Environmental and Resource Values: Theory and Methods. Resources for the Future.
Fryback, D.G., W.F. Lawrence, P.A. Martin, R. Klein, and B.E. Klein. 1993. “The Beaver Dam Health Outcomes Study: Initial Catalog of Health-State Quality Factors.” Medical Decision Making 13:89-102.
Furlong, W., D. Feeny, G.W. Torrance, C. Goldsmith, S. DePauw, Z. Zhu, M. Denton, and M. Boyle. 1998. “Multiplicative Multi-Attribute Utility Function for HUI3: A Technical Report.” McMaster University: CHEPA Working Paper Series.
Garratt, A.M., A. Hutchinson, and I. Russell. 2000. “Patient-Assessed Measures of Health Outcome in Asthma: A Comparison of Four Approaches.” Respiratory Medicine 94(6):597-606.
Glass, G.V. 1976. “Primary, Secondary, and Meta-Analysis.” Educational Researcher 5:3-8.
Grossman, Michael. 1972. “On the Concept of Health Capital and the Demand for Health.” Journal of Political Economy 80:223-255.
Gudex, C., P. Dolan, P. Kind, and A. Williams. 1997. “Valuing Health States: Interviews with the General Public.” European Journal of Public Health 7:441-448.
References
R-3
Hammitt, J.K. 2002. “How Much is a QALY Worth? Utility Functions for Health and Wealth.” Unpublished manuscript.
Hanemann, W. Michael. 1991. “Willingness to Pay and Willingness to Accept: How Much Can They Differ?” American Economic Review 81(3):635-647.
Hausman, J.A. 1993. Contingent Valuation: A Critical Assessment. Amsterdam: Elsevier Science B.V.
Hurst, N.P., P. Kind, D. Ruta, M. Hunter, and A. Stubbings. 1997. “Measuring Health-Related Quality of Life in Rheumatoid Arthritis: Validity, Responsiveness and Reliability of EuroQol (EQ-5D).” British Journal of Rheumatology 36(5):551-559.
Johannesson, Magnus. 1996. “A Note on the Relationship Between Ex Ante and Expected Willingness to Pay for Health Care.” Social Science and Medicine 42(3):305-311.
Johannesson, M., J.S. Pliskin, and M.C. Weinstein. 1994. “A Note on QALYs, Time Tradeoff, and Discounting.” Medical Decision Making 14:188-193.
Johansson, Per-Olov. 1995. Evaluating Health Risks: An Economic Approach. Cambridge, MA: Cambridge University Press.
Johnson, F.R., E.E. Fries, and H.S. Banzhaf. 1997. “Valuing Morbidity: An Integration of the Willingness-to-Pay and Health-Status Index Literatures.” Journal of Health Economics 16:641-665.
Kaplan, R.M., C.J. Atkins, and R. Timms. 1984. “Validity of a Quality of Well-Being Scale as an Outcome Measure in Chronic Obstructive Pulmonary Disease.” J Chronic Dis 37(2):85-95.
Kaplan, R.M., J.E. Alcaraz, J.P. Anderson, and M. Weisman. 1996. “Quality-Adjusted Life Years Lost to Arthritis: Effects of Gender, Race,and Social Class.” Arthritis Care Res 9(6):473-482.
Kartman B., et al. 1996. ”Valuation of Health Changes with the CV Method.” Health Economics 5:531-41.
Klose, Thomas. 2002. “A Utility-Theoretic Model for QALYs and Willingness to Pay.” Health Economics.
Kopec J.A., J.I. Williams, T. To, and P.C. Austin. 2000. “Measuring Population Health: Correlates of the Health Utilities Index among English and French Canadians.” Canadian Journal of Public Health 91(6):465-70.
Valuation of Morbidity Losses: Meta-Analysis of Willingness-to-Pay and Health Status Measures
R-4
Krabbe, Paul F., Marie-Louise Essink-Bot, and Gouke J. Bonsel. 1997. “The Comparability and Reliability of Five Health-State Valuation Methods.” Social Science and Medecine 45(11):1641-1652.
Krupnick, Alan J., and Maureen Cropper. 1992. “The Effect of Information on Health Risk Valuations.” Journal of Risk and Uncertainty 5(1):29-48.
Leidy, N.K., and C. Coughlin. 1998. “Psychometric Performance of the Asthma Quality of Life Questionnaire in a U.S. Sample.” Quality of Life Research 7(2):127-134.
Liu, J.-T., J.K. Hammitt, J.-D. Wang, and J.-L. Liu. 2000. “Mother’s Willingness to Pay for Her Own and Her Child’s Health: A Contingent Valuation Study in Taiwan.” Health Economics 9:319-326.
Lundberg L., et al. 1999. “Quality of Life, Health-State Utilities and WTP in Patients with Psoriasis and Atopic Eczema.” British Journal of Dermatology 141(6):1067-75.
Mittmann, N., K. Trakas, N. Risebrough, and B.A. Liu. 1999. “Utility Scores for Chronic Conditions in a Community-Dwelling Population.” Pharmacoeconomics 15(4):369-376.
Mrozek, J., and L. Taylor. 2002. “What Determines the Value of Life? A Meta Analysis.” Journal of Policy Analysis and Management 21(2):253-270.
O’Brien, Bernie J., and Jose Luis Viramontes. 1994. “Willingness to Pay: A Valid and Reliable Measure of Health State Preferences?” Medical Decision Making 14(3):289-297.
Olsen, J.A., and R.D. Smith. 2001. “Theory Versus Practice: A Review of ‘Willingness-to-Pay’ in Health and Health Care.” Health Economics 10:39-52.
Pliskin, J.S., D.S. Shepard, and M.C. Weinstein. 1980. “Utility Functions for Life Years and Health Status.” Operations Research 28:206-224.
Rosenthal, R. 1991. Meta-analytic Procedures for Social Research. Newbury Park, CA: Sage Publications.
RTI. March 2002. Valuation of Morbid Loss. Interim report prepared for the Food and Drug Administration. Research Triangle Park, NC: RTI.
Schwartz, C.E., R.M. Kaplan, J.P. Anderson, T. Holbrook, and M.W. Genderson. 1999. “Covariation of Physical and Mental Symptoms Across Illnesses: Results of a Factor Analytic Study.” Annals of Behavioral Medicine 21(2):122-127.
References
R-5
Sloan, Frank A., W. Kip Viscusi, Harrell W. Chesson, Christopher J. Conover, and Kathryn Whetten-Goldstein. 1998. “Alternative Approaches to Valuing Intangible Health Losses: The Evidence for Multiple Sclerosis.” Journal of Health Economics 7:475-497.
Smith, R.D. 2000. “The Discrete-Choice Willingness-to-Pay Question Format in Health Economics: Should We Adopt Environmental Guidelines?” Medical Decision Making 20:194-206.
Smith, V.K., and S.K. Pattanayak. 2002. “Is Meta-Analysis the Noah’s Ark for Non Market Valuation?” Environmental and Resource Economics 22(1-2):271-296.
Spilker, B. 1996. “The General Health Policy Model: An Integrated Approach.” In Quality of Life and Phamacoeconomics in Clinical Trials, 2nd ed., R.M. Kaplan and J.P. Anderson, (eds.), pp. 309-322.
Stanley, T.D. 2001. “Wheat from Chaff: Meta-Analysis as Quantitative Literature Review.” The Journal of Economic Perspectives 15(Summer):131-150.
Starmer, C. 2000. “Developments in Non-Expected Utility Theory: The Hunt for a Descriptive Theory of Choice under Risk.” Journal of Economic Literature 38:332-382.
Stavem K. 1999. ”WTP: A Feasible Method for Assessing Treatment Benefits in Epilepsy?” Seizure 8:14-19.
Stavem, K. 2002. “Association of Willingness to Pay with Severity of Chronic Obstructive Pulmonary Disease, Health Status and Other Preference Measures.” International Journal of Tuberculosis and Lung Disease 6(6):542-549.
Thompson, Mark S. 1986. “Willingness to Pay and Accept Risks to Cure Chronic Disease.” American Journal of Public Health 76(4):392-396.
Tversky, Amos, and Daniel Kahneman. 1991. “Loss Aversion in Riskless Choice: A Reference-Dependent Model.” Quarterly Journal of Economics 106:1039-1061.
Viscusi, W. Kip. 1989. “Prospective Reference Theory: Toward an Explanation of the Paradoxes.” Journal of Risk and Uncertainty 2:297-323.
Viscusi, W. Kip, Wesley A. Magat, and Joel Huber. 1991. “Pricing Health Risks: Survey Assessments of Risk-Risk and Risk-Dollar Tradeoffs.” Journal of Environmental Economics and Management 21(1):32-51.
Valuation of Morbidity Losses: Meta-Analysis of Willingness-to-Pay and Health Status Measures
R-6
Viscusi, W. Kip. 1993. “The Value of Risks to Life and Health.” Journal of Economic Literature 31:1912-1946.
Weber, Martin, and Colin Camerer. 1987. “Recent Developments in Modelling Preferences Under Risk.” OR Spektrum 9:129-151.
Zarkin, Gary A., Nancy Dean, Josephine A. Mauskopf, and Richard Williams, Jr. 1993. “Potential Health Benefits of Nutrition Label Changes.” American Journal of Public Health 83(5):717-724.
Zethraeus N., et al. 1997. “The Impact of Hormone Replacement Therapy on Quality of Life and WTP.” British Journal of Obstetric Gynaecology 104(10):1191-5.
Zethraeus, Niklas. 1998. “Willingness to Pay for Hormone Replacement Therapy.” Health Economics 7:31-38.
Zillich, A.J., K. Blumenschein, M. Johannesson, and Patricica Freeman. 2002. “Assessment of the Relationship Between Measures of Disease Severity, Quality of Life, and Willingness to Pay in Asthma.” Pharmacoeconomics 20(4):257-265.
Appendix A: Bibliography and Summary of Morbidity Valuation Studies
A-1
Ta
ble
A-1
. B
iblio
gra
ph
y a
nd
Su
mm
ary
of
Mo
rbid
ity
Va
lua
tio
n S
tud
ies
Stud
y ID
Pu
b ID
Pr
iori
ty
Stud
y N
ame
Val
uati
on
Met
hod
Hea
lth
Effe
ct/C
hang
e R
isk-
base
d C
ount
ry
Hea
lth
Stat
us
Mea
sure
6 2
1a
Che
stnu
t, La
urai
ne G
., St
even
D. C
olom
e, L
. Rob
in
Kel
ler,
Will
iam
E. L
ambe
rt, B
art O
stro
, Rob
ert D
. Row
e,
and
Sand
ra L
. Woj
ciec
how
ski.
198
8. H
eart
Dis
ease
Pa
tient
s’ A
vert
ing
Beh
avio
r, C
osts
of I
llnes
s, a
nd
Will
ingn
ess
to P
ay to
Avo
id A
ngin
a Ep
isod
es.
Fina
l re
port
pre
pare
d fo
r U
.S. E
nvir
onm
enta
l Pro
tect
ion
Age
ncy.
Doc
umen
t No.
EPA
-230
-10-
88-0
42.
CV
M
Ang
ina
atta
cks
No
US
7 4
1a
Dic
kie,
M.,
Ger
king
, S.,
McC
lella
nd, G
., Sc
hulz
e, W
. 19
88.
”Con
tinge
nt V
alua
tion:
The
Val
ue o
f For
mat
ion
Proc
ess.
” U
npub
lishe
d m
anus
crip
t. C
VM
A
cute
sym
ptom
s N
o U
S
29
1 1a
Loeh
man
, E.T
., S.
V. B
erg,
A.A
. Arr
oyo,
R.A
. Hed
inge
r,
J.M. S
chw
artz
, M.E
. Sha
w, R
.W. F
ahie
n, V
.H. D
e, R
.P.
Fish
e, D
.E. R
io, W
.F. R
ossl
ey, a
nd A
.E.S
. Gre
en.
1979
. “D
istr
ibut
iona
l Ana
lysi
s of
Reg
iona
l Ben
efits
and
Cos
t of
Air
Qua
lity
Con
trol
.” J
ourn
al o
f Env
iron
men
tal
Econ
omic
s an
d M
anag
emen
t 6:2
22-2
43.
CV
M
Acu
te s
ympt
oms
No
US
30
1 1a
Row
e, R
ober
t D.,
and
Laur
aine
G. C
hest
nut.
198
5.
Oxi
dant
s an
d A
sthm
atic
s in
Los
Ang
eles
: A
Ben
efits
A
naly
sis.
Fin
al r
epor
t pre
pare
d fo
r U
.S. E
nvir
onm
enta
l Pr
otec
tion
Age
ncy.
Doc
umen
t No.
EPA
/230
/7-8
5/01
0.
CV
M
Ast
hma
atta
cks
No
US
32
1 1a
Tolle
y, G
eorg
e, L
yndo
n B
abco
ck, e
t al.
198
6.
“Val
uatio
n of
Red
uctio
ns in
Hum
an H
ealth
Sym
ptom
s an
d R
isk.
” In
Vol
. 3:
Con
tinge
nt V
alua
tion
Stud
y of
Li
ght S
ympt
oms
and
Ang
ina.
Fin
al r
epor
t pre
pare
d fo
r U
.S. E
nvir
onm
enta
l Pro
tect
ion
Age
ncy.
CV
M
Acu
te s
ympt
oms
No
US
1 1
1
Alb
erin
i, A
., an
d A
. Kru
pnic
k. 1
998.
“A
ir Q
ualit
y an
d Ep
isod
es o
f Acu
te R
espi
rato
ry Il
lnes
s in
Tai
wan
Citi
es:
Evid
ence
from
Sur
vey
Dat
a.”
Jour
nal o
f Urb
an
Econ
omic
s 4
4(1)
:68-
92.
CV
M
Acu
te r
espi
rato
ry
sym
ptom
s N
o Ta
iwan
(con
tinue
d)
Valuation of Morbidity Losses: Meta-Analysis of Willingness-to-Pay and Health Status Measures
A-2
Ta
ble
A-1
. B
iblio
gra
ph
y a
nd
Su
mm
ary
of
Mo
rbid
ity
Va
lua
tio
n S
tud
ies
(co
nti
nu
ed
)
Stud
y ID
Pu
b ID
Pr
iori
ty
Stud
y N
ame
Val
uati
on
Met
hod
Hea
lth
Effe
ct/C
hang
e R
isk-
base
d C
ount
ry
Hea
lth
Stat
us
Mea
sure
1 2
1
Alb
erin
i, A
nna,
Mau
reen
Cro
pper
, Tsu
-Tan
Fu,
Ala
n K
rupn
ick,
Jin-
Tan
Liu,
Dai
gee
Shaw
, and
Win
ston
H
arri
ngto
n. 1
997.
“V
alui
ng H
ealth
Effe
cts
of A
ir
Pollu
tion
in D
evel
opin
g C
ount
ries
: Th
e C
ase
of
Taiw
an.”
Jou
rnal
of E
nvir
onm
enta
l Eco
nom
ics
and
Man
agem
ent
34:1
07-2
6.
CV
M
Acu
te r
espi
rato
ry
sym
ptom
s N
o Ta
iwan
2 1
1
Bal
a, M
ohan
V.,
Lisa
L. W
ood,
Gar
y A
. Zar
kin,
Edw
ard
C. N
orto
n, A
mir
am G
afni
, and
Ber
nie
O’B
rien
. 19
98.
“Val
uing
Out
com
es in
Hea
lth C
are:
A C
ompa
riso
n of
W
illin
gnes
s to
Pay
and
Qua
lity-
Adj
uste
d Li
fe-Y
ears
.”
Jour
nal o
f Clin
ical
Epi
dem
iolo
gy 5
1(8)
:667
-676
.
CV
M
Pain
from
shi
ngle
s N
o U
S SG
, Q
ALY
32
2 1
Ber
ger,
Mar
k C
., G
lenn
C. B
lom
quis
t, D
on K
enke
l, an
d G
eorg
e S.
Tol
ley.
198
7. “
Val
uing
Cha
nges
in H
ealth
R
isks
: A
Com
pari
son
of A
ltern
ativ
e M
easu
res.
”
Sout
hern
Eco
nom
ic Jo
urna
l 53
(4):9
67-9
84.
CV
M
Acu
te s
ympt
oms
No
US
4 1
1
Blu
men
sche
in, K
., Jo
hann
esso
n, M
. 19
98.
“Rel
atio
nshi
p B
etw
een
Qua
lity
of L
ife In
stru
men
ts,
Hea
lth S
tate
Util
ities
, and
Will
ingn
ess
to P
ay in
Pat
ient
s w
ith A
sthm
a.”
Ann
als
of A
llerg
y, A
sthm
a, a
nd
Imm
unol
ogy
80:
189-
194.
CV
M
Ast
hma
cure
N
o U
S V
AS,
SG
, TT
O
5 1
1
Car
thy,
Tre
vor,
Sus
an C
hilto
n, Ju
dith
Cov
ey, L
orra
ine
Hop
kins
, Mic
hael
Lee
-Jon
es, G
raha
m L
oom
es, N
ick
Pidg
eon,
and
Ann
e Sp
ence
r. 1
999.
“O
n th
e C
ontin
gent
V
alua
tion
of S
afet
y an
d th
e Sa
fety
of C
ontin
gent
V
alua
tion:
Par
t 2—
The
CV
/SG
‘Cha
ined
’ App
roac
h.”
Jo
urna
l of R
isk
and
Unc
erta
inty
17(
3):1
87-2
13.
CV
M
Hos
pita
lizat
ion
for
inju
ry
No
UK
SG
6 1
1
Che
stnu
t, La
urai
ne G
., L.
Rob
in K
elle
r, W
illia
m E
. La
mbe
rt, a
nd R
ober
t D. R
owe.
199
6. “
Mea
suri
ng H
eart
Pa
tient
s’ W
illin
gnes
s to
Pay
for
Cha
nges
in A
ngin
a Sy
mpt
oms.
” M
edic
al D
ecis
ion
Mak
ing
16:
65-7
7.
CV
M
Ang
ina
atta
cks
No
US
(con
tinue
d)
Appendix A — Bibliography and Summary of Morbidity Valuation Studies
A-3
Ta
ble
A-1
. B
iblio
gra
ph
y a
nd
Su
mm
ary
of
Mo
rbid
ity
Va
lua
tio
n S
tud
ies
(co
nti
nu
ed
)
Stud
y ID
Pu
b ID
Pr
iori
ty
Stud
y N
ame
Val
uati
on
Met
hod
Hea
lth
Effe
ct/C
hang
e R
isk-
base
d C
ount
ry
Hea
lth
Stat
us
Mea
sure
8 1
1
Dic
kie,
M.,
and
V. U
lery
. 20
02.
“Par
enta
l Altr
uism
and
th
e V
alue
of C
hild
Hea
lth:
Are
Kid
s W
orth
Mor
e Th
an
Pare
nts?
” R
epor
t pre
pare
d fo
r U
.S. E
nvir
onm
enta
l Pr
otec
tion
Age
ncy.
CV
M
Acu
te s
ympt
oms
No
US
7 2
1
Dic
kie,
Mar
k, S
helb
y G
erki
ng, D
avid
Bro
oksh
ire,
Don
C
ours
ey, W
illia
m S
chul
ze, A
nne
Cou
lson
, and
Don
ald
Tash
kin.
198
7. “
Rec
onci
ling
Ave
rtin
g B
ehav
ior
and
Con
tinge
nt V
alua
tion
Ben
efit
Estim
ates
of R
educ
ing
Sym
ptom
s of
Ozo
ne E
xpos
ure.
” In
Impr
ovin
g A
ccur
acy
and
Red
ucin
g C
osts
of E
nvir
onm
enta
l Ben
efit
Ass
essm
ent.
Was
hing
ton,
DC
: U
.S. E
nvir
onm
enta
l Pr
otec
tion
Age
ncy.
CV
M/A
VB
A
cute
sym
ptom
s N
o U
S
7 3
1
Dic
kie,
Mar
k, S
helb
y G
erki
ng, W
illia
m S
chul
ze, A
nne
Cou
lson
, and
Don
ald
Tash
kin.
198
6. “
Val
ue o
f Sy
mpt
oms
of O
zone
Exp
osur
e: A
n A
pplic
atio
n of
the
Ave
rtin
g B
ehav
ior
Met
hod.
” In
Impr
ovin
g A
ccur
acy
and
Red
ucin
g C
osts
of E
nvir
onm
enta
l Ben
efit
Ass
essm
ents
, U
.S. E
nvir
onm
enta
l Pro
tect
ion
Age
ncy.
AV
B
Acu
te s
ympt
oms
No
US
9 1
1 G
an T
., F.
Slo
an, e
t al.
2001
. “H
ow M
uch
Are
Pat
ient
s W
TP to
Avo
id P
osto
pera
tive
Nau
sea
and
Vom
iting
?”
Ane
sthe
sia
& A
nalg
esia
92(
2):3
93-4
00.
CV
M
Post
-ope
rativ
e na
usea
N
o U
S
10
1 1
Hen
son,
Spe
ncer
. 19
96.
“Con
sum
er W
illin
gnes
s to
Pay
fo
r R
educ
tions
in th
e R
isk
of F
ood
Pois
onin
g in
the
UK
.”
Jour
nal o
f Agr
icul
tura
l Eco
nom
ics
47(
3):4
03-4
20.
CV
M
Food
poi
soni
ng
Yes
U
K
39
1 1
Jaco
bs, R
. Jak
e, R
onal
d J.
Mol
eski
and
Alle
n S.
M
eyer
hoff.
200
2. “
Val
uatio
n of
Sym
ptom
atic
Hep
atiti
s A
in A
dults
.” P
harm
acoe
cono
mic
s 2
0(11
):739
-47.
C
VM
H
epat
itis
A
Yes
U
S
(con
tinue
d)
Valuation of Morbidity Losses: Meta-Analysis of Willingness-to-Pay and Health Status Measures
A-4
Ta
ble
A-1
. B
iblio
gra
ph
y a
nd
Su
mm
ary
of
Mo
rbid
ity
Va
lua
tio
n S
tud
ies
(co
nti
nu
ed
)
Stud
y ID
Pu
b ID
Pr
iori
ty
Stud
y N
ame
Val
uati
on
Met
hod
Hea
lth
Effe
ct/C
hang
e R
isk-
base
d C
ount
ry
Hea
lth
Stat
us
Mea
sure
13
1 1
Kar
tman
B.,
et a
l. 1
996.
”V
alua
tion
of H
ealth
Cha
nges
w
ith th
e C
V M
etho
d.”
Hea
lth E
cono
mic
s 5
:531
-41.
C
VM
R
eflu
x oe
soph
agiti
s Y
es
Swed
en
12
1 1
Kar
tman
, Ber
nt, F
redr
ik A
nder
sson
, and
Mag
nus
Joha
nnes
son.
199
6. “
Will
ingn
ess
to P
ay fo
r R
educ
tions
in
Ang
ina
Pect
oris
Atta
cks.
” M
edic
al D
ecis
ion
Mak
ing
16
(3):2
48-2
53.
CV
M
Ang
ina
atta
cks
No
Swed
en
14
1 1
Kei
th P
.L.,
et a
l. 2
000.
”A
Cos
t-B
enef
it A
naly
sis
Usi
ng
a W
TP Q
uest
ionn
aire
of I
ntra
nasa
l Bud
eson
ide
for
Seas
onal
Alle
rgic
Rhi
nitis
.” A
nn A
llerg
y A
sthm
a Im
mun
ol 8
4:55
-62.
CV
M
Alle
rgic
rhi
nitis
N
o U
S
22
1 1
Kra
bbe,
Pau
l F.,
Mar
ie-L
ouis
e Es
sink
-Bot
, and
Gou
ke J.
B
onse
l. 1
997.
“Th
e C
ompa
rabi
lity
and
Rel
iabi
lity
of
Five
Hea
lth-S
tate
Val
uatio
n M
etho
ds.”
Soc
ial S
cien
ce
and
Med
icin
e 4
5(11
):164
1-52
.
CV
M
Bro
adly
def
ined
he
alth
sta
tes
No
Net
herl
and
s SG
, TTO
23
1 1
Kru
pnic
k, A
lan
J., a
nd M
aure
en C
ropp
er.
1992
. “T
he
Effe
ct o
f Inf
orm
atio
n on
Hea
lth R
isk
Val
uatio
ns.”
Jo
urna
l of R
isk
and
Unc
erta
inty
5(1
):29-
48.
Con
join
t B
ronc
hitis
Y
es
US
37
1 1
Lee,
Pat
rick
Y.,
Dav
id M
atch
ar, D
enni
s C
lem
ents
, Joe
l H
uber
, Joh
n H
amilt
on, a
nd E
ric
Pete
rson
. 20
02.
“Eco
nom
ic A
naly
sis
of In
fluen
za V
acci
natio
n an
d A
ntiv
iral
Tre
atm
ent f
or H
ealth
y W
orki
ng A
dults
.”
Ann
als
of In
tern
al M
edic
ine
137
(4):2
25-3
1.
Con
join
t
One
day
of
influ
enza
sy
mpt
oms
no
US
15
1 1
Liu
Jin-T
an, e
t al.
200
0. ”
Mot
her’
s W
TP fo
r H
er O
wn
and
Her
Chi
ld’s
Hea
lth:
A C
ontin
gent
Val
uatio
n St
udy
in T
aiw
an.”
Hea
lth E
cono
mic
s 9
:319
-26.
C
VM
C
old
No
Taiw
an
QW
B
35
1 1
Lund
berg
L.,
et a
l. 19
99.
“Qua
lity
of L
ife, H
ealth
-Sta
te
Util
ities
and
WTP
in P
atie
nts
with
Pso
rias
is a
nd A
topi
c Ec
zem
a.”
Br
J Der
mat
ol 1
41(6
):106
7-75
. C
VM
Ps
oria
sis
and
atop
ic e
czem
a N
o Sw
eden
SG, T
TO,
VA
S,
Euro
Qol
, 15
D,S
F-36
(c
ontin
ued)
Appendix A — Bibliography and Summary of Morbidity Valuation Studies
A-5
Ta
ble
A-1
. B
iblio
gra
ph
y a
nd
Su
mm
ary
of
Mo
rbid
ity
Va
lua
tio
n S
tud
ies
(co
nti
nu
ed
)
Stud
y ID
Pu
b ID
Pr
iori
ty
Stud
y N
ame
Val
uati
on
Met
hod
Hea
lth
Effe
ct/C
hang
e R
isk-
base
d C
ount
ry
Hea
lth
Stat
us
Mea
sure
16
1 1
O’B
rien
, Ber
nie
J., a
nd Jo
se L
uis
Vir
amon
tes.
199
4.
“Will
ingn
ess
to P
ay:
A V
alid
and
Rel
iabl
e M
easu
re o
f H
ealth
Sta
te P
refe
renc
es?”
Med
ical
Dec
isio
n M
akin
g 14
(3):2
89-2
97.
CV
M
Hea
lthy
lung
fu
nctio
n Y
es
Can
ada
SG
26
1 1
Perr
eira
, Kri
sta
M.,
and
Fran
k Sl
oan.
200
2. “
Livi
ng
Hea
lthy
and
Livi
ng L
ong:
Val
uing
the
Non
pecu
niar
y Lo
ss fr
om D
isab
ility
and
Dea
th.”
The
Jour
nal o
f Ris
k an
d U
ncer
tain
ty 2
4(1)
:5-2
9.
US
Dis
abili
ty
Yes
U
S
17
1 1
Rea
dy, R
icha
rd C
., St
ale
Nav
rud,
and
W. R
icha
rd
Dub
ourg
. 20
01.
“How
do
Res
pond
ents
with
Unc
erta
in
Will
ingn
ess
to P
ay A
nsw
er C
ontin
gent
Val
uatio
n Q
uest
ions
?” L
and
Econ
omic
s 7
7(3)
:315
-26.
CV
M
Acu
te s
ympt
oms
No
Nor
way
17
2 1
Rea
dy, R
icha
rd, S
tale
Nav
rud,
Bre
tt D
ay, R
icha
rd
Dub
ourg
, Fer
nand
o M
acha
do, S
usan
a M
oura
to, F
rank
Sp
anni
nks,
and
Mar
ia X
ose
Vaz
quez
Rod
rigu
ez.
1999
. “B
enef
it Tr
ansf
er in
Eur
ope:
Are
Val
ues
Con
sist
ent
Acr
oss
Cou
ntri
es?”
Wor
king
Pap
er.
CV
M
Acu
te s
ympt
oms
No
Nor
way
, N
ethe
rlan
ds,
Port
ugal
, Sp
ain,
En
glan
d
27
1 1
Sloa
n, F
rank
A.,
W. K
ip V
iscu
si, H
arre
ll W
. Che
sson
, C
hris
toph
er J.
Con
over
, and
Kat
hryn
Whe
tten-
Gol
dste
in.
1998
. “A
ltern
ativ
e A
ppro
ache
s to
Val
uing
Inta
ngib
le
Hea
lth L
osse
s: T
he E
vide
nce
for
Mul
tiple
Scl
eros
is.”
Jo
urna
l of H
ealth
Eco
nom
ics
7:4
75-4
97.
Con
join
t M
ultip
le S
cler
osis
Y
es
US
19
1 1
Slot
huus
U.,
et a
l. 2
000.
“W
TP fo
r A
rthr
itis
Sym
ptom
A
llevi
atio
n.”
Inte
rnat
iona
l Jou
rnal
of T
echn
olog
y A
sses
smen
t in
Hea
lth C
are
16(
1):6
0-72
. C
VM
A
rthr
itis
sym
ptom
s N
o D
enm
ark
19
2 1
Slot
huus
U.,
et a
l. 2
000.
WTP
in A
rthr
itis:
A D
anis
h C
ontr
ibut
ion.
” R
heum
atol
ogy
(Oxf
ord)
39(
7):7
91-9
. C
VM
A
rthr
itis
No
Den
mar
k
(con
tinue
d)
Valuation of Morbidity Losses: Meta-Analysis of Willingness-to-Pay and Health Status Measures
A-6
Ta
ble
A-1
. B
iblio
gra
ph
y a
nd
Su
mm
ary
of
Mo
rbid
ity
Va
lua
tio
n S
tud
ies
(co
nti
nu
ed
)
Stud
y ID
Pu
b ID
Pr
iori
ty
Stud
y N
ame
Val
uati
on
Met
hod
Hea
lth
Effe
ct/C
hang
e R
isk-
base
d C
ount
ry
Hea
lth
Stat
us
Mea
sure
28
1 1
Stav
em K
. 19
99. ”
WTP
: A
Fea
sibl
e M
etho
d fo
r A
sses
sing
Tre
atm
ent B
enef
its in
Epi
leps
y?”
Sei
zure
8:
14-1
9.
CV
M
Epile
psy
cure
N
o N
orw
ay
SG,
TTO
, V
AS,
Eu
roQ
ol,
15D
34
1 1
Stav
em, K
. 20
02.
“Ass
ocia
tion
of W
illin
gnes
s to
Pay
w
ith S
ever
ity o
f Chr
onic
Obs
truc
tive
Pulm
onar
y D
isea
se, H
ealth
Sta
tus
and
Oth
er P
refe
renc
e M
easu
res.
”
Inte
rnat
iona
l Jou
rnal
of T
uber
culo
sis
and
Lung
Dis
ease
6(
6):5
42-5
49.
CV
M
CO
PD c
ure
No
Nor
way
SG,
TTO
, V
AS,
SF-
36
33
1 1
Thom
pson
, Mar
k S.
198
6. “
Will
ingn
ess
to P
ay a
nd
Acc
ept R
isks
to C
ure
Chr
onic
Dis
ease
.” A
mer
ican
Jo
urna
l of P
ublic
Hea
lth 7
6(4)
:392
-396
.
CV
M
Art
hriti
s cu
re
No
US
SG
38
1 1
Torr
ance
, Geo
rge,
Val
lery
Wal
ker,
Ron
ald
Gro
ssm
an,
Jaya
nti M
ukhe
rjee
, Dav
id V
augh
an, J
aque
s La
For
ge,
and
Noe
l Lam
pron
. 19
99.
“Eco
nom
ic E
valu
atio
n of
C
ipro
floxa
cin
Com
pare
d w
ith th
e U
sual
Ant
ibac
teri
al
Car
e fo
r th
e Tr
eatm
ent o
f Acu
te E
xace
rbat
ions
of
Chr
onic
Bro
nchi
tis in
Pat
ient
s Fo
llow
ed fo
r 1
Yea
r.”
Ph
arm
acoe
cono
mic
s 1
6(5
Pt. 1
):499
-520
.
CV
M
Acu
te
exac
erba
tion
of
chro
nic
bron
chiti
s no
C
anad
a
21
1 1
Vis
cusi
, W. K
ip, W
esle
y A
. Mag
at, a
nd Jo
el H
uber
. 19
91.
“Pri
cing
Hea
lth R
isks
: Su
rvey
Ass
essm
ents
of
Ris
k-R
isk
and
Ris
k-D
olla
r Tr
adeo
ffs.”
Jou
rnal
of
Envi
ronm
enta
l Eco
nom
ics
and
Man
agem
ent
21(1
):32-
51.
Con
join
t C
hron
ic b
ronc
hitis
Y
es
US
31
2 1
Zet
hrae
us N
., et
al.
199
7. “
The
Impa
ct o
f Hor
mon
e R
epla
cem
ent T
hera
py o
n Q
ualit
y of
Life
and
WTP
.” B
r J
Obs
tet G
ynae
col
104(
10):1
191-
5.
CV
M
Hor
mon
e re
plac
emen
t th
erap
y N
o Sw
eden
V
AS,
TT
O
31
1 1
Zet
hrae
us, N
ikla
s. 1
998.
“W
illin
gnes
s to
Pay
for
Hor
mon
e R
epla
cem
ent T
hera
py.”
Hea
lth E
cono
mic
s 7:
31-3
8.
CV
M
Hor
mon
e re
plac
emen
t th
erap
y N
o Sw
eden
V
AS,
TT
O
(con
tinue
d)
Appendix A — Bibliography and Summary of Morbidity Valuation Studies
A-7
Ta
ble
A-1
. B
iblio
gra
ph
y a
nd
Su
mm
ary
of
Mo
rbid
ity
Va
lua
tio
n S
tud
ies
(co
nti
nu
ed
)
Stud
y ID
Pu
b ID
Pr
iori
ty
Stud
y N
ame
Val
uati
on
Met
hod
Hea
lth
Effe
ct/C
hang
e R
isk-
base
d C
ount
ry
Hea
lth
Stat
us
Mea
sure
2
Age
e, M
.D.,
and
T.D
. Cro
cker
. 19
96.
“Par
enta
l A
ltrui
sm a
nd C
hild
Lea
d Ex
posu
re:
Infe
renc
es fr
om th
e D
eman
d fo
r C
hela
tion
Ther
apy.
” Jo
urna
l of H
uman
R
esou
rces
31(
3):6
77-6
91.
AV
B
Che
latio
n th
erap
y Y
es
US
2
App
el, L
.J., S
tein
berg
E.P
., Po
we
N.R
., et
al.
199
0.
“Ris
k R
educ
tion
From
Low
Osm
ality
Con
tras
t Med
ia:
Wha
t Do
Patie
nts
Thin
k It
Is W
orth
?” M
edic
al C
are
28:3
24.
CV
M
Side
effe
cts
from
ra
diol
ogy
Yes
U
S
2
Ara
na, J
orge
E.,
and
Car
mel
o J.
Leon
. 20
02.
“Will
ingn
ess
to P
ay fo
r H
ealth
ris
k R
educ
tion
in th
e C
onte
xt o
f Altr
uism
.” H
ealth
Eco
nom
ics
in p
ress
. C
VM
R
educ
e pr
obab
ility
of
get
ting
flu
Yes
Sp
ain
2
Ari
stid
es, M
ike,
Jack
Che
n, M
ak S
chul
z, E
ve
Will
iam
son,
and
Ste
phen
Cla
rke.
200
2. “
Con
join
t A
naly
sis
of a
New
Che
mot
hera
py.”
Ph
arm
acoe
cono
mic
s 20
(110
:775
-84.
C
onjo
int
2 Ty
pes
of m
outh
ul
cera
tions
from
ch
emot
hera
py
Yes
A
ustr
alia
2 B
iddl
e, J.
E.,
and
G. Z
arki
n. 1
988.
“W
orke
r Pr
efer
ence
s an
d M
arke
t Com
pens
atio
n fo
r Jo
b R
isk.
” R
evie
w o
f Ec
onom
ics
and
Stat
istic
s 7
0(4)
:660
-667
. H
edon
ic
Inju
ry
Yes
U
S
2
Blu
men
sche
in K
., et
al.
200
1. ”
Hyp
othe
tical
ver
sus
Rea
l Will
ingn
ess
to P
ay in
the
Hea
lth C
are
Sect
or:
Res
ults
from
a F
ield
Exp
erim
ent.”
J H
ealth
Eco
nom
ic
20:4
41-4
57.
CV
M
Ast
hma
man
agem
ent
prog
ram
N
o U
S
2 C
ross
M.J.
, et a
l. 2
000.
”D
eter
min
ants
of W
TP fo
r H
ip
and
Kne
e R
epla
cem
ent S
urge
ry fo
r O
steo
arth
ritis
.”
Rhe
umat
olog
y (O
xfor
d) 3
9(11
):124
2-8.
C
VM
H
ip a
nd k
nee
repl
acem
ent
surg
ery
No
Aus
tral
ia
2 C
unni
ngha
m S
.J., a
nd N
.P. H
unt.
200
0. ”
Rel
atio
nshi
p B
etw
een
Util
ity V
alue
s an
d W
TP in
Pat
ient
s U
nder
goin
g Tr
eatm
ent.”
Com
mun
ity D
enta
l Hea
lth 1
7(2)
:92-
6.
CV
M
Ort
hagn
atic
tr
eatm
ent
U
K
SG
(con
tinue
d)
Valuation of Morbidity Losses: Meta-Analysis of Willingness-to-Pay and Health Status Measures
A-8
Ta
ble
A-1
. B
iblio
gra
ph
y a
nd
Su
mm
ary
of
Mo
rbid
ity
Va
lua
tio
n S
tud
ies
(co
nti
nu
ed
)
Stud
y ID
Pu
b ID
Pr
iori
ty
Stud
y N
ame
Val
uati
on
Met
hod
Hea
lth
Effe
ct/C
hang
e R
isk-
base
d C
ount
ry
Hea
lth
Stat
us
Mea
sure
2
Dic
kie,
Mar
k, a
nd S
helb
y G
erki
ng.
1996
. “D
efen
sive
A
ctio
n an
d W
illin
gnes
s to
Pay
for
Red
uced
Hea
lth R
isk:
In
fere
nces
from
Act
ual a
nd C
ontin
gent
Beh
avio
r.”
Pa
per
pres
ente
d at
the
1996
AER
E W
orks
hop,
Tah
oe
City
, CA
.
AV
B
Skin
can
cer
Yes
U
S
2
Dic
kie,
Mar
k, a
nd S
helb
y G
erki
ng.
1996
. “F
orm
atio
n of
Ris
k B
elie
fs, J
oint
Pro
duct
ion
and
Will
ingn
ess
to P
ay
to A
void
Ski
n C
ance
r.”
Rev
iew
of E
cono
mic
s an
d St
atis
tics
451
-463
.
AV
B
Skin
can
cer
Yes
U
S
2 D
iez
L. 1
998.
”A
sses
sing
the
Will
ingn
ess
of P
aren
ts to
Pa
y fo
r R
educ
ing
Post
oper
ativ
e Em
esis
in C
hild
ren.
”
Phar
mac
o-Ec
onom
ics
13(
5 pa
rt 2
):589
-596
. C
VM
Po
st o
pera
tive
emes
is in
chi
ldre
n Y
es
UK
2
Don
alds
on, C
., T.
Map
p, M
. Rya
n, a
nd K
. Cur
tin.
1996
. “E
stim
atin
g th
e Ec
onom
ic B
enef
its o
f Avo
idin
g Fo
od
Bor
ne R
isk:
Is
Will
ingn
ess
to P
ay F
easi
ble?
” E
pide
mio
l. In
fect
. 11
6:28
5-94
.
CV
M
Elim
inat
e ri
sk o
f po
ultr
y bo
rne
illne
ss
Yes
U
K
2
Don
alds
on, C
am, a
nd P
hil S
hack
ley.
199
7. “
Doe
s ‘P
roce
ss U
tility
’ Exi
st?
A C
ase
Stud
y of
Will
ingn
ess
to
Pay
for
Lapa
rosc
opic
Cho
lecy
stec
tom
y.”
Soc
ial S
cien
ce
and
Med
icin
e 4
4(5)
:699
-707
.
CV
M
Gal
l bla
dder
tr
eatm
ent
No
UK
2
Dra
nits
aris
G.,
et a
l. 2
000.
”Th
e Ec
onom
ic V
alue
of a
N
ew In
sulin
Pre
para
tion,
Hum
alog
Mix
25.
Mea
sure
d by
a W
TP A
ppro
ach.
” P
harm
aco-
Econ
omic
s
18(3
):275
-87.
CV
M,
Con
join
t N
ew in
sulin
fo
rmul
atio
n Y
es
Can
ada
2
Dra
nits
aris
G.
1997
. ”A
Pilo
t Stu
dy to
Eva
luat
e th
e Fe
asib
ility
of U
sing
WTP
as
a M
easu
re o
f Val
ue in
C
ance
r Su
ppor
tive
Car
e: A
n A
sses
smen
t of A
mifo
stin
e C
ytop
rote
ctio
n.”
Sup
port
Car
e C
ance
r 5
(6):4
89-9
9.
CV
M
Che
mot
hera
py
toxi
city
Y
es
Can
ada
(con
tinue
d)
Appendix A — Bibliography and Summary of Morbidity Valuation Studies
A-9
Ta
ble
A-1
. B
iblio
gra
ph
y a
nd
Su
mm
ary
of
Mo
rbid
ity
Va
lua
tio
n S
tud
ies
(co
nti
nu
ed
)
Stud
y ID
Pu
b ID
Pr
iori
ty
Stud
y N
ame
Val
uati
on
Met
hod
Hea
lth
Effe
ct/C
hang
e R
isk-
base
d C
ount
ry
Hea
lth
Stat
us
Mea
sure
2 Ea
stau
gh S
.R.
1991
. “V
alua
tions
of t
he B
enef
its o
f Ris
k-Fr
ee B
lood
.” I
nter
natio
nal J
ourn
al o
f Tec
hnol
ogy
Ass
essm
ent
7:51
. C
VM
R
isk
free
blo
od
Yes
C
anad
a
2 Ea
stau
gh, S
.R.
2000
. ”W
TP in
Tre
atm
ent o
f Ble
edin
g D
isor
ders
.” I
ntl J
Tec
hnol
ogy
Ass
essm
ent i
n H
ealth
C
are
16(
2):7
06-1
0.
CV
M
Ble
edin
g di
sord
er
trea
tmen
t Y
es
US
2
Eber
hart
, L.H
.J., M
. Mau
ch, A
. M. M
orin
, H. W
ulf,
and
G. G
eldn
er.
2002
. “I
mpa
ct o
f Mul
timod
al A
nti-
Emet
ic
Prop
hyla
xis
on P
atie
nt S
atis
fact
ion
in H
igh-
Ris
k Pa
tient
s fo
r Po
stop
erat
ive
Nau
sea
and
Vom
iting
.” A
naes
thes
ia
57:1
022-
27.
CV
M
Red
uce
risk
of
post
-ope
rativ
e na
usea
Y
es
UK
2
Evan
s, W
illia
m N
., an
d W
. Kip
Vis
cusi
. 19
91.
“Est
imat
ion
of S
tate
-Dep
ende
nt U
tility
Fun
ctio
ns U
sing
Su
rvey
Dat
a.”
Rev
iew
s of
Eco
nom
ics
and
Stat
istic
s 73
:94-
104.
CV
M
Occ
upat
iona
l in
jury
ris
k Y
es
US
2
Gay
er, T
ed, J
ames
T. H
amilt
on, a
nd W
. Kip
Vis
cusi
. 20
00.
“Pri
vate
Val
ues
of R
isk
Trad
eoffs
at S
uper
fund
Si
tes:
Hou
sing
Mar
ket E
vide
nce
on L
earn
ing
abou
t R
isk.
” T
he R
evie
w o
f Eco
nom
ics
and
Stat
istic
s 82
(3):4
39-5
1.
Hed
onic
(P
rope
rty
Val
ue)
Can
cer
Yes
U
S
2
Gay
er, T
ed, J
ames
T. H
amilt
on, a
nd W
. Kip
Vis
cusi
. 20
02.
“The
Mar
ket V
alue
of R
educ
ing
Can
cer
Ris
k:
Hed
onic
Hou
sing
Pri
ces
with
Cha
ngin
g In
form
atio
n.”
So
uthe
rn E
cono
mic
Jour
nal
69(2
):266
-289
.
Hed
onic
(P
rope
rty
Val
ue)
Can
cer
Yes
U
S
2
Gra
nber
g, M
aria
, Mat
ts W
ikla
nd, L
ars
Nils
son,
and
Lar
s H
ambe
rger
. 19
95.
“Cou
ple’
s W
illin
gnes
s to
Pay
for
IVF/
ET.”
Act
a O
bste
tric
ia e
t Gyn
ecol
ogic
a Sc
andi
navi
ca
74:1
99-2
02.
2
Gyl
dmar
k, M
arle
ne, a
nd G
wen
doly
n C
. Mor
riso
n.
2001
. “D
eman
d fo
r H
ealth
Car
e in
Den
mar
k: R
esul
ts o
f a
Nat
iona
l Sam
ple
Surv
ey U
sing
Con
tinge
nt V
alua
tion.
”
Soci
al S
cien
ce a
nd M
edic
ine
53:
1023
-36.
CV
M
Insu
ranc
e to
cov
er
hosp
ital c
osts
for
four
chr
onic
co
nditi
ons
Yes
D
enm
ark
Valuation of Morbidity Losses: Meta-Analysis of Willingness-to-Pay and Health Status Measures
A-10
(con
tinue
d)
Ta
ble
A-1
. B
iblio
gra
ph
y a
nd
Su
mm
ary
of
Mo
rbid
ity
Va
lua
tio
n S
tud
ies
(co
nti
nu
ed
)
Stud
y ID
Pu
b ID
Pr
iori
ty
Stud
y N
ame
Val
uati
on
Met
hod
Hea
lth
Effe
ct/C
hang
e R
isk-
base
d C
ount
ry
Hea
lth
Stat
us
Mea
sure
2
Joha
nnes
son,
M. a
nd B
. Fag
erbe
rg.
1992
. “A
Hea
lth
Econ
omic
Com
pari
son
of D
iet a
nd D
rug
Trea
tmen
t in
Obe
se M
en w
ith M
ild H
yper
tens
ion.
” Jo
urna
l of
Hyp
erte
nsio
n 1
0:10
63-7
0.
CV
M
Die
t and
/or
drug
tr
eatm
ent t
o re
duce
ris
k of
ca
rdio
vasc
ular
di
seas
e fr
om
obes
ity
Yes
Sw
eden
V
AS
2
Joha
nnes
son,
Mag
nus,
Ben
gt Jö
nsso
n, a
nd L
ars
Bor
gqui
st.
1991
. “W
illin
gnes
s to
Pay
for
Ant
ihyp
erte
nsiv
e Th
erap
y: R
esul
ts o
f a S
wed
ish
Pilo
t St
udy.
” Jo
urna
l of H
ealth
Eco
nom
ics
10:
461-
473.
CV
M
Ant
ihyp
erse
nsiti
ve
ther
apy
No
Swed
en
2
Joha
nnes
son,
Mag
nus,
H. A
berg
, L. A
greu
s, L
. Bor
quis
t, an
d B
. Jon
sson
. 19
91.
“Cos
t-B
enef
it A
naly
sis
of N
on-
phar
moc
olog
ical
Tre
atm
ent o
f Hyp
erte
nsio
n.”
Jour
nal
of In
tern
al M
edic
ine
230
:307
-312
.
CV
M
Ant
ihyp
erse
nsiti
ve
ther
apy
No
Swed
en
VA
S
2
Joha
nnes
son,
Mag
nus,
Per
-Olo
v Jo
hans
son,
Ben
gt
Kri
strö
m, a
nd U
lf-G
. Ger
dtha
m.
1993
. “W
illin
gnes
s to
Pa
y fo
r A
ntih
yper
tens
ive
Ther
apy—
Furt
her
Res
ults
.”
Jour
nal o
f Hea
lth E
cono
mic
s 1
2:95
-108
.
CV
M
Ant
ihyp
erse
nsiti
ve
ther
apy
No
Swed
en
VA
S
2
Joha
nnes
son,
Mag
nus,
R.M
. O’C
onor
, G. K
obel
t-N
guye
n, a
nd A
. Mat
tiass
on.
1997
. “W
illin
gnes
s to
Pay
fo
r R
educ
ed In
cont
inen
ce S
ympt
oms.
” B
ritis
h Jo
urna
l of
Uro
logy
80:
557-
562.
CV
M
Red
uctio
n in
Le
akag
es
No
Swed
en
SF-3
6
2 Jo
hann
esso
n, M
agnu
s. 1
992.
“Ec
onom
ic E
valu
atio
n of
H
yper
tens
ion
Trea
tmen
t.” I
nter
natio
nal J
ourn
al o
f Te
chno
logy
Ass
essm
ent i
n H
ealth
Car
e 8(
3):5
06-5
23.
C
VM
A
ntih
yper
sens
itive
th
erap
y N
o Sw
eden
2 Jo
hann
esso
n, M
agnu
s. 1
992.
“Ec
onom
ic E
valu
atio
n of
Li
pid
Low
erin
g—A
Fea
sibi
lity
Test
of t
he C
ontin
gent
V
alua
tion
App
roac
h.”
Hea
lth P
olic
y 2
0:30
9-32
0.
CV
M
Trea
tmen
t to
redu
ce c
hole
ster
ol
leve
ls to
nor
mal
fo
r lif
e
No
Swed
en
(con
tinue
d)
Appendix A — Bibliography and Summary of Morbidity Valuation Studies
A-11
Ta
ble
A-1
. B
iblio
gra
ph
y a
nd
Su
mm
ary
of
Mo
rbid
ity
Va
lua
tio
n S
tud
ies
(co
nti
nu
ed
)
Stud
y ID
Pu
b ID
Pr
iori
ty
Stud
y N
ame
Val
uati
on
Met
hod
Hea
lth
Effe
ct/C
hang
e R
isk-
base
d C
ount
ry
Hea
lth
Stat
us
Mea
sure
2
John
son,
F. R
eed,
and
Kri
sty
E. M
atth
ews.
200
1.
“Sou
rces
and
Effe
cts
of U
tility
-The
oret
ic In
cons
iste
ncy
in
Stat
ed-P
refe
renc
e Su
rvey
s.”
Am
eric
an Jo
urna
l of
Agr
icul
tura
l Eco
nom
ics
83(
5):1
328-
33.
Con
join
t G
luco
se c
ontr
ol
Yes
U
S
2
Kle
inm
an, L
eah,
Em
ma
McI
ntos
h, M
andy
Rya
n, Jo
rdan
Sc
hmie
r, Jo
seph
Cra
wle
y, G
. Ric
hard
Loc
ke, a
nd
Gre
gory
de
Liss
ovoy
. 20
02.
“Will
ingn
ess
to P
ay fo
r co
mpl
ete
Sym
ptom
Rel
ief o
f Gas
troe
soph
agea
l Ref
lux
Dis
ease
.” A
rch
Inte
rn M
ed 1
62:1
361-
66.
Con
join
t
Rel
ief o
f ga
stro
esop
hage
al
reflu
x di
seas
e sy
mpt
oms
No
US
2
Kob
elt,
Gis
ela.
199
7. “
Econ
omic
Con
side
ratio
ns a
nd
outc
ome
Mea
sure
men
t in
Urg
e In
cont
inen
ce.”
Uro
logy
50
(6A
):100
-107
.
CV
M
25 a
nd 5
0%
impr
ovem
ent i
n in
cont
inen
ce
sym
ptom
s N
o Sw
eden
Eu
roQ
OL
2
Kup
perm
ann
M.,
et a
l. 2
000.
”Pa
rent
s’ P
refe
renc
es fo
r O
utco
mes
Ass
ocia
ted
with
Chi
ldho
od V
acci
natio
ns.”
Pe
diat
r In
fect
Dis
19:
129-
33.
2 Le
e S.
J., e
t al.
199
8. ”
Perc
eptio
ns a
nd P
refe
renc
es o
f A
utol
ogou
s B
lood
Don
ors.
” T
rans
fusi
on 3
8(8)
:757
-63.
2
Lee,
S.J.
, B. L
iljas
, W.H
. Chu
rchi
ll, M
.A. P
opov
sky,
C.P
. St
owel
l, M
.E. C
anno
n, a
nd M
. Joh
anne
sson
. 19
98.
“Per
cept
ions
and
Pre
fere
nces
of A
utol
ogou
s B
lood
D
onor
s.”
Tra
nsfu
sion
38:
757-
763.
CV
M
Fee
for
patie
nt
stor
age
of th
eir
own
bloo
d
Red
uce
risk
of
infe
cted
blo
od
No
US
2
Lee,
Ste
phan
ie J.
Jun
e 19
98.
“Pat
ient
s’ W
illin
gnes
s to
Pa
y fo
r A
utol
ogou
s B
lood
Don
atio
n.”
Ris
k in
Pe
rspe
ctiv
e 6(
6):.
(c
ontin
ued)
Valuation of Morbidity Losses: Meta-Analysis of Willingness-to-Pay and Health Status Measures
A-12
Ta
ble
A-1
. B
iblio
gra
ph
y a
nd
Su
mm
ary
of
Mo
rbid
ity
Va
lua
tio
n S
tud
ies
(co
nti
nu
ed
)
Stud
y ID
Pu
b ID
Pr
iori
ty
Stud
y N
ame
Val
uati
on
Met
hod
Hea
lth
Effe
ct/C
hang
e R
isk-
base
d C
ount
ry
Hea
lth
Stat
us
Mea
sure
2
Lee,
Ste
phan
ie, J
., Pe
ter
J. N
eum
ann,
W. H
allo
wel
l C
hurc
hill,
Mar
ie E
. Can
non,
Milt
on C
. Wei
nste
in, a
nd
Mag
nus
Joha
nnes
son.
199
7. “
Patie
nts’
Will
ingn
ess
to
Pay
for
Aut
olog
ous
Blo
od D
onat
ion.
” H
ealth
Pol
icy
40:1
-12.
CV
M
Red
uce
risk
of
infe
cted
blo
od
No
US
2 Li
eu T
A, e
t al.
200
0. ”
The
Hid
den
Cos
ts o
f Inf
ant
Vac
cina
tion”
Vac
cine
19(
1):3
3-41
. C
VM
Num
ber
of
vacc
ine
inje
ctio
ns
and
adve
rse
sym
ptom
s
No
US
2
Lipt
on, R
icha
rd B
., Sa
ndra
W. H
amel
sky,
and
Jeffr
ey M
. D
ayno
. 20
02.
“Wha
t do
Patie
nts
with
Mig
rain
e W
ant
from
Acu
te M
igra
ine
Trea
tmen
t?”
Hea
dach
e 4
2(S1
). C
VM
M
igra
ine
no
US
2 Lo
ngo
C.J.
199
9. ”
Cho
ices
of M
etho
dolo
gy in
Ph
arm
acoe
cono
mic
s St
udie
s.”
(bas
ed o
n ab
stra
ct) M
ed
Car
e 3
7(4
Supp
l Lill
y):A
S32-
5.
CV
M
Neu
trop
enia
, ne
urot
oxic
ity,
neph
roto
xici
ty
Yes
C
anad
a
2
Mag
at, W
esle
y A
., W
. Kip
Vis
cusi
, and
Joel
Hub
er.
1988
. “P
aire
d C
ompa
riso
n an
d C
ontin
gent
Val
uatio
n A
ppro
ache
s to
Mor
bidi
ty R
isk
Val
uatio
n.”
Jour
nal o
f En
viro
nmen
tal E
cono
mic
s an
d M
anag
emen
t 15
(4):3
95-
411.
CV
M
Con
sum
er p
rodu
ct
safe
ty (a
ccid
ents
/ po
ison
ing)
Y
es
US
2
Mag
at, W
esle
y, a
nd W
. Kip
Vis
cusi
. 19
92.
“Inf
orm
atio
nal A
ppro
ache
s to
Reg
ulat
ion.
” R
egul
atio
n of
Eco
nom
ic A
ctiv
ity S
erie
s, V
ol 1
9. C
ambr
idge
and
Lo
ndon
: M
IT P
ress
.
CV
M
Inju
ry
Yes
U
S
2
Mat
thew
s, D
ebor
a, A
ngel
a R
occh
i, an
d A
mir
am G
afni
. 20
02.
“Put
ting
You
r M
oney
Whe
re Y
our
Mou
th Is
: W
illin
gnes
s to
Pay
for
Den
tal G
el.”
Ph
arm
acoe
cono
mic
s 2
0(4)
:245
-55.
CV
M
Non
-inj
ecte
d (n
o pa
in) d
enta
l an
aest
hetic
Y
es
Can
ada
(con
tinue
d)
Appendix A — Bibliography and Summary of Morbidity Valuation Studies
A-13
Ta
ble
A-1
. B
iblio
gra
ph
y a
nd
Su
mm
ary
of
Mo
rbid
ity
Va
lua
tio
n S
tud
ies
(co
nti
nu
ed
)
Stud
y ID
Pu
b ID
Pr
iori
ty
Stud
y N
ame
Val
uati
on
Met
hod
Hea
lth
Effe
ct/C
hang
e R
isk-
base
d C
ount
ry
Hea
lth
Stat
us
Mea
sure
2
Mitc
hell,
Rob
ert C
., an
d R
icha
rd T
. Car
son.
198
6.
“Val
uing
Dri
nkin
g W
ater
Ris
k R
educ
tions
Usi
ng th
e C
ontin
gent
Val
uatio
n M
etho
d: A
Met
hodo
logi
cal S
tudy
of
Ris
ks fr
om T
HM
and
Gia
rdia
.” P
repa
red
for
the
U.S
. En
viro
nmen
tal P
rote
ctio
n A
genc
y un
der
Coo
pera
tive
Agr
eem
ent C
R81
0466
-01-
6.
CV
M
Can
cer
Yes
U
S
2 M
ulle
r A
., an
d T.
J Reu
tzel
. 19
84.
”WTP
for
Red
uctio
n in
Fat
ality
Ris
k: A
n Ex
plor
ator
y Su
rvey
.” A
m J
Publ
ic
Hea
lth 7
4(8)
:808
-12.
C
VM
In
jury
and
fata
lity
Yes
U
S
2 N
arbr
o K
., an
d Sj
ostr
om L
. 200
0. ”
WTP
for
Obe
sity
Tr
eatm
ent.”
Int
l J T
echn
olog
y A
sses
smen
t in
Hea
lth
Car
e 1
6(1)
:50-
59.
CV
M
Rel
ieve
obe
sity
-re
late
d pr
oble
ms
No
Swed
en
2
Neu
man
n, P
eter
J.,
and
Joha
nnes
son,
Mag
nus.
199
4.
“The
Will
ingn
ess
to P
ay fo
r In
Vitr
o Fe
rtili
zatio
n: A
Pilo
t St
udy
Usi
ng C
ontin
gent
Val
uatio
n.”
Med
ical
Car
e 32
(7):6
86-6
99.
2 N
orin
der,
A.,
et a
l. 2
001.
”Sc
ope
and
Scal
e In
sens
itivi
ties
in a
CV
Stu
dy o
f Ris
k R
educ
tions
.” H
ealth
Po
licy
57(
2):1
41-5
3.
CV
M
Fata
l and
non
-fat
al
Yes
Sw
eden
2
O’B
rien
, Ber
nie
J., R
on G
oere
e, A
mir
am G
afni
, Geo
rge
W. T
orra
nce,
Mar
k V
. Pau
ly, H
aim
Erd
er, J
im
Rus
thov
en, J
ane
Wee
ks, M
iliss
a C
ahill
, and
Bru
ce
LaM
ont.
199
8. “
Ass
essi
ng th
e V
alue
of a
New
Ph
arm
aceu
tical
: A
Fea
sibi
lity
Stud
y of
Con
tinge
nt
Val
uatio
n in
Man
aged
Car
e.”
Med
ical
Car
e.
CV
M
Neu
trop
enia
from
ch
emot
hera
py
Yes
U
S
2
O’B
rien
, Ber
nie
J., S
neza
na N
ovos
el, G
eorg
e To
rran
ce,
and
Dav
id S
trei
ner.
199
5. “
Ass
essi
ng th
e Ec
onom
ic
Val
ue o
f a N
ew A
ntid
epre
ssan
t: A
Will
ingn
ess-
to-P
ay
App
roac
h.”
Pha
rmac
oEco
nom
ics
8(1
):34-
45.
CV
M
Adv
erse
effe
cts
of
antid
epre
ssan
t Y
es
Can
ada
(con
tinue
d)
Valuation of Morbidity Losses: Meta-Analysis of Willingness-to-Pay and Health Status Measures
A-14
Ta
ble
A-1
. B
iblio
gra
ph
y a
nd
Su
mm
ary
of
Mo
rbid
ity
Va
lua
tio
n S
tud
ies
(co
nti
nu
ed
)
Stud
y ID
Pu
b ID
Pr
iori
ty
Stud
y N
ame
Val
uati
on
Met
hod
Hea
lth
Effe
ct/C
hang
e R
isk-
base
d C
ount
ry
Hea
lth
Stat
us
Mea
sure
2
O’B
ryne
, Pau
l, La
uren
Cud
dy, D
. Way
ne T
aylo
r,
Step
hen
Bir
ch, J
oann
e M
orri
s, a
nd Je
rry
Syro
tuik
. 19
96.
“Effi
cacy
and
Cos
t Ben
efit
of In
hale
d C
ortic
oste
roid
s in
Pa
tient
s C
onsi
dere
d to
Hav
e M
ild A
sthm
a in
Pri
mar
y C
are
Prac
tice.
” C
anad
ian
Res
pira
tory
Jour
nal
3(3)
:169
-17
5.
CV
M
Ast
hma
Yes
C
anad
a
2
O’C
onor
, Ric
hard
M.,
and
Gle
nn C
. Blo
mqu
ist.
199
7.
“Mea
sure
men
t of C
onsu
mer
-Pat
ient
Pre
fere
nces
Usi
ng a
H
ybri
d C
ontin
gent
Val
uatio
n M
etho
d.”
Jour
nal o
f H
ealth
Eco
nom
ics
16:
667-
683.
CV
M
Ast
hma
drug
ef
ficac
y, d
eath
Y
es
US
2 O
’Con
or, R
.M.,
et a
l. 1
998.
”U
rge
Inco
ntin
ence
: Q
ualit
y of
Life
and
Pat
ient
s’ V
alua
tion
of S
ympt
om
Red
uctio
n.”
Pha
rmac
o-Ec
onom
ics
14(
5):5
31-5
39.
CV
M
Mic
turi
tions
and
le
akag
es
Yes
U
S SF
-36
2 O
lson
, C.A
. 19
81.
“An
Ana
lysi
s of
Wag
e D
iffer
entia
ls
Rec
eive
d by
Wor
kers
on
Dan
gero
us Jo
bs.”
Jou
rnal
of
Hum
an R
esou
rces
16:
167-
185.
he
doni
c In
jury
Y
es
US
2
Ort
ega,
Ana
, Geo
rge
Dra
nits
aris
, and
Ani
tash
a L.
V.
Puod
siun
as.
1998
. “W
hat A
re C
ance
r Pa
tient
s W
illin
g to
Pay
for
Prop
hyla
ctic
Epo
etin
Alfa
? A
Cos
t-B
enef
it A
naly
sis.
” C
ance
r 8
3(12
):258
8-25
96.
CV
M
Blo
od tr
ansf
usio
n fo
r ca
ncer
ane
mia
Y
es
Can
ada
2
Penn
ie, .
R.A
, et a
l. 1
991.
”Fa
ctor
s In
fluen
cing
the
Acc
epta
nce
of H
epat
itis
B V
acci
ne b
y St
uden
ts in
H
ealth
Dis
cipl
ines
in O
ttaw
a.”
Can
J Pu
blic
Hea
lth
82(1
):12-
5.
CV
M
Hep
atiti
s B
Y
es
Can
ada
2
Poin
er, T
.F.,
et a
l. 2
000.
”Pa
tient
Atti
tude
s to
Top
ical
A
ntip
sori
atic
Tre
atm
ent w
ith C
alci
potr
iol a
nd
Dith
rano
l.” J
of t
he E
urop
ean
Aca
dem
y of
Der
mat
olog
y &
Vin
ereo
logy
14(
3):1
53-8
.
CV
M
antip
sori
atic
tr
eatm
ent
Yes
U
K
(con
tinue
d)
Appendix A — Bibliography and Summary of Morbidity Valuation Studies
A-15
Ta
ble
A-1
. B
iblio
gra
ph
y a
nd
Su
mm
ary
of
Mo
rbid
ity
Va
lua
tio
n S
tud
ies
(co
nti
nu
ed
)
Stud
y ID
Pu
b ID
Pr
iori
ty
Stud
y N
ame
Val
uati
on
Met
hod
Hea
lth
Effe
ct/C
hang
e R
isk-
base
d C
ount
ry
Hea
lth
Stat
us
Mea
sure
2 R
amse
y, S
.D.,
et a
l. 1
997.
”W
TP fo
r A
ntih
yper
tens
ive
Car
e: E
vide
nce
from
a S
taff-
Mod
el H
MO
.” (
base
d on
ab
stra
ct)
Soc
Sci M
ed 4
4(12
):191
1-7.
C
VM
A
ntih
yper
sens
itive
th
erap
y Y
es
US
2
Rea
d, D
anie
l, an
d N
icol
eta
Lilia
na R
ead.
200
1. “
An
Age
-Em
bedd
ing
Effe
ct: T
ime
Sens
itivi
ty a
nd T
ime
Inse
nsiti
vity
Whe
n Pr
icin
g H
ealth
Ben
efits
.” A
cta
Psyc
holo
gica
108
:177
-136
.
CV
M
Cur
e fo
r co
nditi
on
that
wou
ld li
mit
driv
ing
abili
ty
No
UK
2
Rya
n, M
. 19
98.
“Val
uing
Psy
chol
ogic
al F
acto
rs in
the
Prov
isio
n of
Ass
iste
d R
epro
duct
ive
Tech
niqu
es U
sing
th
e Ec
onom
ic In
stru
men
t of W
illin
gnes
s to
Pay
.”
Jour
nal o
f Eco
nom
ic P
sych
olog
y 1
9:17
9-20
4.
2
Rya
n, M
andy
, and
Fer
nand
o Sa
n M
igue
l. 2
000.
“T
estin
g fo
r C
onsi
sten
cy in
Will
ingn
ess
to P
ay
Expe
rim
ents
.” J
ourn
al o
f Eco
nom
ic P
sych
olog
y 2
1:30
5-17
.
CV
M
Men
orha
gia
trea
tmen
t Y
es
UK
2
Rya
n, M
andy
, and
Julie
Rat
cliff
e. 2
000.
“So
me
Issu
es
in th
e A
pplic
atio
n of
Clo
sed-
Ende
d W
illin
gnes
s to
Pay
St
udie
s to
Val
uing
Hea
lth G
oods
: A
n A
pplic
atio
n of
A
nten
atal
Car
e in
Sco
tland
.” A
pplie
d Ec
onom
ics
32
:643
-51.
CV
M
Ant
enat
al c
are
No
UK
2 R
yan,
Man
dy.
1996
. “U
sing
Will
ingn
ess
to P
ay to
Pay
to
Ass
ess
the
Ben
efits
of A
ssis
ted
Rep
rodu
ctiv
e Te
chni
ques
.” E
cono
mic
Eva
luat
ion
5:5
43-5
58.
2
Sans
om, S
teph
anie
L.,
L. B
arke
r, P
.S. C
orso
, C. B
row
n,
and
R. D
euso
n. “
Rot
avir
us V
acci
ne a
nd
Intu
ssus
cept
ion:
How
Muc
h R
isk
Will
Par
ents
in th
e U
nite
d St
ates
Acc
ept t
o O
btai
n V
acci
ne B
enef
its.”
A
mer
ican
Jour
nal o
f Epi
dem
iolo
gy 1
54(1
1):1
077-
1085
.
CV
M
Side
effe
ct (b
owel
bl
ock)
from
chi
ld
vacc
ine
Yes
U
S
(con
tinue
d)
Valuation of Morbidity Losses: Meta-Analysis of Willingness-to-Pay and Health Status Measures
A-16
Ta
ble
A-1
. B
iblio
gra
ph
y a
nd
Su
mm
ary
of
Mo
rbid
ity
Va
lua
tio
n S
tud
ies
(co
nti
nu
ed
)
Stud
y ID
Pu
b ID
Pr
iori
ty
Stud
y N
ame
Val
uati
on
Met
hod
Hea
lth
Effe
ct/C
hang
e R
isk-
base
d C
ount
ry
Hea
lth
Stat
us
Mea
sure
2
Scha
fer,
T.,
A. R
iehl
e, H
.-E.
Wic
hman
n, a
nd J.
Rin
g.
2002
. “A
ltern
ativ
e M
edic
ine
in A
llerg
ies
- Pr
eval
ence
, Pa
ttern
s of
Use
, and
Cos
ts.”
Alle
rgy
57:
694-
700.
C
VM
C
ompl
ete
heal
ing
of a
llerg
y sy
mpt
oms
No
Ger
man
y
2
Schw
ab-C
hris
te, N
atha
lie G
., an
d N
ils C
. Sog
uel.
199
6.
“The
Pai
n of
Roa
d-A
ccid
ent V
ictim
s an
d th
e B
erea
vem
ent o
f the
ir R
elat
ives
: A
Con
tinge
nt-V
alua
tion
Expe
rim
ent.”
Jou
rnal
of R
isk
and
Unc
erta
inty
13:
277-
291.
CV
M
Red
uced
ris
k of
ro
ad a
ccid
ent f
or
self
and/
or
rela
tive
Yes
Sw
itzer
land
2
Sevy
, Ser
ge, K
ay N
atha
nson
, Cly
de S
chec
hter
, and
G
eorg
e Fu
lop.
200
1. “
Con
tinge
ncy
Val
uatio
n an
d Pr
efer
ence
s of
Hea
lth S
tate
s A
ssoc
iate
d w
ith S
ide
Effe
cts
of A
ntip
sych
otic
Med
icat
ions
in S
chiz
ophr
enia
.”
Schi
zoph
reni
a B
ulle
tin 2
7(4)
:643
-52.
CV
M
All
side
effe
cts
from
an
tipsy
chot
ic
med
icat
ion
no
US
2
Slot
huus
, Ulla
, Met
te L
. Lar
sen,
and
Pet
er Ju
nker
. 20
02.
“The
Con
tinge
nt R
anki
ng M
etho
d- A
Fea
sibl
e an
d V
alid
M
etho
d W
hen
Elic
iting
Pre
fere
nces
for
Hea
lth C
are?
”
Soci
al S
cien
ce M
edic
ine
54:
1601
-160
9.
Con
tinge
nt
rank
ing
33%
, 66%
, 100
%
impr
ovem
ent o
f rh
eum
atoi
d ar
thri
tis s
ympt
oms
No
Den
mar
k
2
Soru
m P
. 199
9. ”
Mea
suri
ng P
atie
nt P
refe
renc
es b
y W
illin
gnes
s to
Avo
id:
The
Cas
e of
Acu
te O
titis
Med
ia.”
M
edic
al D
ecis
ion
Mak
ing
19(
01):2
7-37
.
2
Stie
b, D
., P.
Civ
ita, F
. Joh
nson
, M. M
anar
y, A
. Ani
s.
“Val
uing
Acu
te C
ardi
ores
pira
tory
Mor
bidi
ty A
ssoc
iate
d w
ith A
ir P
ollu
tion
Usi
ng C
ost o
f Illn
ess
and
Indi
vidu
al
Will
ingn
ess
to P
ay.”
Con
join
t C
ardi
ores
pira
tory
m
orbi
dity
N
o C
anad
a
2 Ta
mbo
ur M
., an
d N
. Zet
hreu
s. 1
998.
”N
onpa
ram
etri
c W
TP M
easu
res
and
Con
fiden
ce S
tate
men
ts.”
Med
ical
D
ecis
ion
Mak
ing
18(
3):3
30-3
36.
CV
M
Hor
mon
e re
plac
emen
t th
erap
y Y
es
Swed
en
(con
tinue
d)
Appendix A — Bibliography and Summary of Morbidity Valuation Studies
A-17
Ta
ble
A-1
. B
iblio
gra
ph
y a
nd
Su
mm
ary
of
Mo
rbid
ity
Va
lua
tio
n S
tud
ies
(co
nti
nu
ed
)
Stud
y ID
Pu
b ID
Pr
iori
ty
Stud
y N
ame
Val
uati
on
Met
hod
Hea
lth
Effe
ct/C
hang
e R
isk-
base
d C
ount
ry
Hea
lth
Stat
us
Mea
sure
2
Tang
, J.,
B. W
ang,
P.F
. Whi
te, M
.F. W
atch
a, J.
Qi,
and
R.H
. Wen
der.
199
8. “
The
Effe
ct o
f Tim
ing
of
Oda
nset
ron
Adm
inis
trat
ion
on it
s Ef
ficac
y, C
ost-
Effe
ctiv
enes
s, a
nd C
ost-
Ben
efit
as a
Pro
phyl
actic
A
ntie
met
ic in
the
Am
bula
tory
Set
ting.
” A
nest
hesi
a A
nalg
esia
86:
274-
82.
CV
M
post
ope
rativ
e em
esis
(nau
sea,
vo
miti
ng)
No
US
2 Te
lser
, Har
ry, a
nd P
eter
Zw
eife
l. 2
002
“M
easu
ring
W
illin
gnes
s-to
-Pay
for
Ris
k R
educ
tion:
An
App
licat
ion
of C
onjo
int A
naly
sis.
” H
ealth
Eco
nom
ics
11:
129-
139.
C
onjo
int
Elde
rly
fem
ur
frac
ture
Y
es
Switz
erla
nd
34
1 2
Thom
pson
, M.S
., J.L
. Rea
d, a
nd M
. Lia
ng.
1984
. “F
easi
bilit
y of
Will
ingn
ess
to P
ay M
easu
rem
ent i
n C
hron
ic A
rthr
itis.
” M
edic
al D
ecis
ion
Mak
ing
4:1
95-
215.
CV
M
Art
hriti
s N
o U
S
2
Vis
cusi
, W. K
ip, a
nd W
illia
m N
. Eva
ns.
1990
. “U
tility
Fu
nctio
ns th
at D
epen
d on
Hea
lth S
tatu
s: E
stim
ates
and
Ec
onom
ic Im
plic
atio
ns.”
Am
eric
an E
cono
mic
Rev
iew
80
(3):3
53-3
74.
Hed
onic
In
jury
Y
es
US
2 V
iscu
si, W
. Kip
. 19
78.
“Lab
or M
arke
t Val
uatio
ns o
f Li
fe a
nd L
imb:
Em
piri
cal E
stim
ates
and
Pol
icy
Impl
icat
ions
.” P
ublic
Pol
icy
26(
3):3
59-3
86.
Hed
onic
In
jury
Y
es
US
2
Vis
cusi
, W. K
ip.,
and
C. O
’Con
nor.
198
4. “
Ada
ptiv
e R
espo
nses
to C
hem
ical
Lab
elin
g: A
re W
orke
rs B
ayes
ian
Dec
isio
n M
aker
s?”
Am
eric
an E
cono
mic
Rev
iew
74
(5):9
42-9
56.
Hed
onic
In
jury
Y
es
US
2
Vis
cusi
, W.K
., W
. Mag
at, a
nd A
. For
rest
. 19
88.
“Altr
uist
ic a
nd P
riva
te V
alua
tions
of R
isk
Red
uctio
n.”
Jo
urna
l of P
olic
y A
naly
sis
and
Man
agem
ent
7(2)
:227
-24
5.
CV
M
Pair
ed e
ffect
: In
hala
tion,
ski
n-po
ison
ing
Yes
U
S
(con
tinue
d)
Valuation of Morbidity Losses: Meta-Analysis of Willingness-to-Pay and Health Status Measures
A-18
Ta
ble
A-1
. B
iblio
gra
ph
y a
nd
Su
mm
ary
of
Mo
rbid
ity
Va
lua
tio
n S
tud
ies
(co
nti
nu
ed
)
Stud
y ID
Pu
b ID
Pr
iori
ty
Stud
y N
ame
Val
uati
on
Met
hod
Hea
lth
Effe
ct/C
hang
e R
isk-
base
d C
ount
ry
Hea
lth
Stat
us
Mea
sure
2
Wer
ner,
P.,
and
I. V
ered
. 20
02.
“Wom
en’s
WTP
Out
-of
-Poc
ket f
or D
rug
Trea
tmen
t for
Ost
eopo
rosi
s B
efor
e an
d A
fter
the
Enac
tmen
t of R
egul
atio
ns P
rovi
ding
Pub
lic
Fund
ing:
Evi
denc
e fr
om a
Nat
ural
Exp
erim
ent i
n Is
rael
.”
Osp
eopo
rosi
s In
t 13
:228
-34.
CV
M
Ost
eopo
rosi
s: h
ip
frac
ture
Y
es
Isra
el
1 3
3
Alb
erin
i, A
., an
d A
. Kru
pnic
k. 2
000.
“C
ost-
of-I
llnes
s an
d W
illin
gnes
s-to
-Pay
Est
imat
es o
f the
Ben
efits
of
Impr
oved
Air
Qua
lity:
Evi
denc
e fr
om T
aiw
an.”
Lan
d Ec
onom
ics
76(
1):3
7-53
.
CV
M
Res
pira
tory
illn
ess
No
Taiw
an
2 2
3
Bal
a, M
ohan
V.,
Lisa
L. W
ood,
and
Gar
y A
. Zar
kin.
19
97.
Val
uing
Out
com
es in
Hea
lth C
are:
A
Com
pari
son
of W
illin
gnes
s to
Pay
and
Qua
lity-
Adj
uste
d-Li
fe Y
ears
.” W
orki
ng P
aper
.
CV
M
Pain
from
shi
ngle
s N
o U
S SG
, Q
ALY
29
2 3
Gre
en, A
.E.S
., et
al.
197
8. “
An
Inte
rdis
cipl
inar
y St
udy
of th
e H
ealth
, Soc
ial a
nd E
nvir
onm
enta
l Eco
nom
ics
of
Sulfu
r O
xide
Pol
lutio
n in
Flo
rida
.” I
nter
disc
iplin
ary
Cen
ter
for
Aer
onom
y an
d (o
ther
) Atm
osph
eric
Sci
ence
s.
Prep
ared
for
Flor
ida
Sulfu
r O
xide
s St
udy,
Inc.
CV
M
Acu
te s
ympt
oms
No
US
3 M
orri
s J.,
Per
ez D
. 20
00.
”WTP
for
New
C
hem
othe
rapy
for
Adv
ance
d O
vari
an C
ance
r.”
New
Z
eala
nd M
edic
al Jo
urna
l 11
3(11
08):1
43-6
. C
VM
C
hem
othe
rapy
N
o N
ew
Zea
land
30
3 3
Row
e, R
ober
t D. a
nd L
aura
ine
G. C
hest
nut.
198
4.
“Val
uing
Cha
nges
in M
orbi
dity
: WTP
ver
sus
CO
I M
easu
re.”
Pap
er p
repa
red
for
the
Am
eric
an E
cono
mic
A
ssoc
iatio
n/A
ssoc
iatio
n of
Env
iron
men
tal a
nd R
esou
rce
Econ
omic
s Jo
int M
eetin
gs, D
alla
s, T
exas
12/
27-1
2/30
.
CV
M
Ast
hma
No
US
30
2 3
Row
e, R
ober
t D.,
and
Laur
aine
G. C
hest
nut.
198
6.
Add
endu
m to
: Oxi
dant
s an
d A
sthm
atic
s in
Los
Ang
eles
: A
Ben
efits
Ana
lysi
s. R
epor
t pre
pare
d fo
r U
.S. E
PA.
Doc
umen
t No.
EPA
-230
-09-
86-0
17.
CV
M
Ast
hma
No
US
a Inc
lude
d in
the
John
son,
Fri
es, a
nd B
anzh
af (1
997)
met
a-an
alys
is.
Appendix B: Annotated Bibliography of Publications Included in the Morbidity Value Meta-Analyses
B-1
Ta
ble
B-1
. A
nn
ota
ted
Bib
lio
gra
ph
y o
f S
tud
ies
Inc
lud
ed
in
th
e A
cu
te E
ffe
cts
Me
ta-A
na
lysi
s
Stud
y ID
Pu
b
ID
Aut
hors
A
rtic
le T
itle
Y
ear
Sour
ce
Val
ue
Elic
itat
ion
Met
hod
Mod
e of
A
dmin
istr
atio
n an
d Y
ear
Com
mod
ity
Def
init
ion
Targ
et
Popu
lati
on
Bas
elin
e H
ealt
h of
St
udy
Popu
lati
on
Que
stio
n Fo
rmat
W
TP E
stim
ate
# of
Val
ues
in
Acu
te M
eta-
Ana
lysi
s
1 1
Alb
erin
i, A
an
d K
rupn
ick,
A “A
ir Q
ualit
y Ep
isod
es o
f Acu
te
Res
pira
tory
Illn
ess
in T
aiw
an C
ities
: Ev
iden
ce fr
om
Surv
ey D
ata.
”
1998
Jou
rnal
of
Urb
an
Econ
omic
s 44
: 68-
92 C
V
Tele
phon
e an
d In
-pe
rson
inte
rvie
w;
Nov
embe
r 19
91 -
Ja
nuar
y 19
92;
Sept
embe
r 19
92.
Mos
t rec
ent
expe
rien
ce o
f ac
ute
resp
irat
ory
illne
ss a
s de
fined
by
the
resp
onde
nt
Gen
eral
po
pula
tion
of
Taip
ei,
Kao
hsiu
ng,
and
Hua
lien,
Ta
iwan
52.3
3% o
f re
spon
dent
s re
port
ed
sym
ptom
s du
ring
th
e st
udy
peri
od
with
a m
ean
of
0.96
epi
sode
s pe
r re
spon
dent
and
a
mea
n du
ratio
n of
3.
97 d
ays
Dic
hoto
-m
ous
choi
ce
The
mea
n W
TP
ran g
ed fr
om $
26
to $
54 (1
992
US$
) to
avoi
d fr
om 1
to 1
0 sy
mpt
om d
ays
3
1 2
Alb
erin
i, A
et
al.
“Val
uing
Hea
lth
Effe
cts
of A
ir
Pollu
tion
in
Dev
elop
ing
Cou
ntri
es: T
he
Cas
e of
Tai
wan
”
1997
Jou
rnal
of
Envi
ron-
men
tal
Econ
omic
s an
d M
anag
eme
nt 3
4: 1
07-
126
CV
In
per
son
inte
rvie
w;
Sept
embe
r 19
92.
Mos
t rec
ent
expe
rien
ce o
f ac
ute
resp
irat
ory
illne
ss a
s de
fined
by
the
resp
onde
nt
Gen
eral
po
pula
tion
of
Taip
ei,
Kao
hsiu
ng,
and
Hua
lien,
Ta
iwan
Med
ian
dura
tion
of a
n ep
isod
e w
as
4 da
ys a
nd m
ean
was
6.8
day
s.
Dur
ing
an e
piso
de
the
med
ian
num
ber
of
sym
ptom
s ex
peri
ence
d w
as
1 an
d th
e m
ean
was
2.2
.
Dic
hoto
mo
us c
hoic
e M
edia
n W
TP to
av
oid
a re
curr
ence
of t
he
aver
age
epis
ode
was
$39
(199
2 U
S$),
Mea
n W
TP fo
r th
e av
era g
e of
1 a
nd
5 da
y ep
isod
es
whe
re th
e ep
isod
e is
or
is
not a
col
d w
as
$42
(199
2 U
S$)
2
2 1
Bal
a M
V, e
t al
. “V
alui
ng
Out
com
es in
H
ealth
Car
e: a
C
ompa
riso
n of
W
TP a
nd Q
ALY
s”
1998
Jou
rnal
of
Clin
ical
Ep
idem
i-ol
ogy
51(8
): 66
7-67
6
CV
C
ompu
ter
inte
ract
ive
inte
rvie
w, y
ear
not
avai
labl
e
WTP
for
3 di
ffere
nt
trea
tmen
ts: T
1:
no tr
eatm
ent =
m
ild p
ain
for
2 w
eeks
; tr
eatm
ent =
no
pain
. T2
: no
trea
tmen
t =
seve
re p
ain
for
2 w
eeks
fo
llow
ed b
y m
ild p
ain
for
1 w
eek;
trea
tmen
t =
mild
pai
n fo
r 65
-70
year
ol
ds in
Sa
raso
ta a
nd
Ft. M
yers
, FL
12%
had
pri
or
shin
gles
ex
peri
ence
and
ov
er 8
0% h
ad a
n ac
quai
ntan
ce w
ho
had
expe
rien
ced
shin
gles
Dou
ble-
boun
ded
dich
oto-
mou
s ch
oice
Med
ian
WTP
ra
nged
from
$3
79 (T
reat
men
t 1)
to $
1,19
8 (T
reat
men
t 3)
(199
6 U
S$)
1
(con
tinue
d)
Valuation of Morbidity Losses: Meta-Analysis of Willingness-to-Pay and Health Status Measures
B-2
Ta
ble
B-1
. A
nn
ota
ted
Bib
lio
gra
ph
y o
f S
tud
ies
Inc
luded in t
he A
cute
Eff
ects
Me
ta-A
na
lysi
s (c
on
tin
ue
d)
Stud
y ID
Pu
b
ID
Aut
hors
A
rtic
le T
itle
Y
ear
Sour
ce
Val
ue
Elic
itat
ion
Met
hod
Mod
e of
A
dmin
istr
atio
n an
d Y
ear
Com
mod
ity
Def
init
ion
Targ
et
Popu
lati
on
Bas
elin
e H
ealt
h of
St
udy
Popu
lati
on
Que
stio
n Fo
rmat
W
TP E
stim
ate
# of
Val
ues
in
Acu
te M
eta-
Ana
lysi
s
Bal
a M
V, e
t al
. (co
nt’d
)
2 w
eeks
. T3
: no
trea
tmen
t =
seve
re p
ain
for
2 m
onth
s fo
llow
ed b
y m
ild p
ain
for
1 m
onth
; tr
eatm
ent =
se
vere
pai
n fo
r 2
wee
ks
follo
wed
by
mild
pai
n fo
r 1
wee
k
6 1
Che
stnu
t LG
, et a
l. “M
easu
ring
Hea
rt
Patie
nts’
WTP
for
Cha
nges
in A
ngin
a Sy
mpt
oms”
1996
M
edic
al
Dec
isio
n M
akin
g 16
: 65-
77 C
V
Tele
phon
e in
terv
iew
. 198
6 A
void
ance
of
eith
er 4
or
8 ad
ditio
nal
angi
na e
piso
des M
ale
angi
na
patie
nts
who
ha
d be
en
trea
ted
at a
m
edic
al c
ente
r or
a V
A
hosp
ital i
n th
e Lo
s A
ngel
es
area
.
43 s
ubje
cts
wer
e cu
rren
tly
expe
rien
cing
an
gina
and
all
subj
ects
had
ex
peri
ence
d an
gina
with
in th
e pr
evio
us tw
o ye
ars.
Bid
ding
ga
me
follo
wed
by
open
-end
ed M
ean
WTP
to
avoi
d 4
or 8
ad
ditio
nal
angi
na
epis
odes
ra
nged
from
$2
03 to
$21
8 w
ith a
med
ian
of $
100
(US$
)
2
6 2
Che
stnu
t LG
, et
al.
“Hea
rt D
isea
se
Pati
ents
’ Ave
rtin
g B
ehav
ior,
Cos
t of
Il
lnes
s, a
nd
Will
ingn
ess
to P
ay
to A
void
Ang
ina
Epis
odes
”
1988
U
.S. E
PA
Rep
ort
CV
Te
leph
one
inte
rvie
w; 1
986
Avo
idan
ce o
f an
gina
sy
mpt
oms
for
eith
er o
ne o
r tw
o ep
isod
es
Mal
e an
gina
pa
tien
ts w
ho
had
been
tr
eate
d at
a
med
ical
cen
ter
or a
VA
ho
spit
al in
the
Lo
s A
ngel
es
area
.
43 s
ubje
cts
wer
e cu
rren
tly
expe
rien
cing
an
gina
and
all
subj
ects
had
ex
peri
ence
d an
gina
wit
hin
the
prev
ious
tw
o ye
ars.
Dic
hoto
-m
ous
choi
ce
wit
h op
en-
ende
d fo
llow
up
Mea
n W
TP t
o av
oid
one
epis
ode
was
$1
00 a
nd $
165
to a
void
tw
o ep
isod
es (
US$
)
2
7 4
Dic
kie
M,
Ger
king
S,
McC
lella
nd
G, S
chul
ze
W
“Con
ting
ent
Val
uati
on: T
he
Val
ue o
f Fo
rmat
ion
Proc
ess”
1988
U
.S. E
PA
Rep
ort
CV
Te
leph
one
inte
rvie
w; 1
986
Avo
id
sym
ptom
s of
O
3 ex
posu
re
and
othe
r sy
mpt
oms
for
a du
rati
on o
f one
da
y
Part
icip
ants
of
the
CO
RD
st
udy
from
the
Lo
s A
ngel
es,
CA
are
a co
mm
unit
ies
of G
lend
ora
and
Bur
bank
30%
of
resp
onde
nts
had
been
dia
gnos
ed
wit
h ch
roni
c re
spir
ator
y ai
lmen
ts.
Ope
n-en
ded
Mea
n W
TP
rang
ed f
rom
$2
to
$27
per
sym
ptom
day
(1
986
US$
)
12
(con
tinue
d)
Appendix B — Annotated Bibliography of Publications Included in the Morbidity Value Meta-Analysis
B-3
Ta
ble
B-1
. A
nn
ota
ted
Bib
lio
gra
ph
y o
f S
tud
ies
Inc
luded in t
he A
cute
Eff
ects
Me
ta-A
na
lysi
s (c
on
tin
ue
d)
Stud
y ID
Pu
b
ID
Aut
hors
A
rtic
le T
itle
Y
ear
Sour
ce
Val
ue
Elic
itat
ion
Met
hod
Mod
e of
A
dmin
istr
atio
n an
d Y
ear
Com
mod
ity
Def
init
ion
Targ
et
Popu
lati
on
Bas
elin
e H
ealt
h of
St
udy
Popu
lati
on
Que
stio
n Fo
rmat
W
TP E
stim
ate
# of
Val
ues
in
Acu
te M
eta-
Ana
lysi
s
8 1
Dic
kie,
M
and
V
Ule
ry.
“Par
enta
l Altr
uism
an
d th
e V
alue
of
Chi
ld H
ealth
: Are
K
ids
Wor
th M
ore
Than
Par
ents
?”
2002
U
.S. E
PA
Rep
ort
CV
In
-Per
son
Inte
rvie
ws.
June
-Ju
ly, 2
000.
Avo
id
com
bina
tions
of
sym
ptom
s (c
ough
, sh
ortn
ess
of
brea
th, c
hest
pa
in, f
ever
); du
ratio
n (o
ne
day,
one
wee
k,
and
one
mon
th);
pers
on
expe
rien
cing
it
(chi
ld o
r pa
rent
).
Gen
eral
po
pula
tion
of
pare
nts
in
Hat
tiesb
urg,
M
issi
ssip
pi
Phys
icia
n-di
agno
sed
asth
ma
10%
, cou
gh w
ith
phle
gm in
pas
t ye
ar 4
6%,
shor
tnes
s of
br
eath
with
w
heez
ing
in p
ast
year
13%
, che
st
pain
whe
n co
ugh
or b
reat
he d
eep
in
past
yea
r 25
%,
feve
r in
pas
t yea
r 46
%, m
ean
sym
ptom
day
s 11
.16,
mea
n co
ugh
days
4.4
8,
mea
n sh
ortn
ess
of
brea
th w
ith
whe
ezin
g da
ys
2.74
, mea
n ch
est
pain
day
s 2.
72,
mea
n fe
ver
days
2.
73
Dou
ble-
boun
ded
dich
oto-
mou
s ch
oice
w
ith o
pen-
ende
d fo
llow
up.
Mea
n W
TP to
av
oid
diffe
rent
sy
mpt
om/d
urat
ion
co
mbi
natio
ns
rang
ed fr
om
$53
to $
218
(200
0 U
S$)
15
11
1 Jo
hnso
n FR
, et
al.
“WTP
for
Impr
oved
R
espi
rato
ry a
nd
Car
diov
ascu
lar
Hea
lth:
A
Mul
tiple
-For
mat
, St
ated
-Pre
fere
nce
App
roac
h”
2000
H
ealth
Ec
onom
ics
9: 2
95-
317.
CA
C
ompu
ter
inte
rvie
w; M
arch
–Ju
ly, 1
997
Avo
id
com
bina
tions
of
sym
ptom
s (7
), D
urat
ion
(3)
and
activ
ity
leve
ls (4
).
Gen
eral
po
pula
tion
of
the
Toro
nto
area
Not
spe
cifie
d G
rade
d-pa
ir
and
disc
rete
-ch
oice
Mea
n W
TP to
av
oid
the
diffe
rent
co
mbi
natio
ns o
f sy
mpt
oms,
du
ratio
n, a
nd
activ
ity le
vels
ra
nged
from
$3
to $
1002
(199
7 C
an$)
67
(con
tinue
d)
Valuation of Morbidity Losses: Meta-Analysis of Willingness-to-Pay and Health Status Measures
B-4
Ta
ble
B-1
. A
nn
ota
ted
Bib
lio
gra
ph
y o
f S
tud
ies
Inc
luded in t
he A
cute
Eff
ects
Me
ta-A
na
lysi
s (c
on
tin
ue
d)
Stud
y ID
Pu
b
ID
Aut
hors
A
rtic
le T
itle
Y
ear
Sour
ce
Val
ue
Elic
itat
ion
Met
hod
Mod
e of
A
dmin
istr
atio
n an
d Y
ear
Com
mod
ity
Def
init
ion
Targ
et
Popu
lati
on
Bas
elin
e H
ealt
h of
St
udy
Popu
lati
on
Que
stio
n Fo
rmat
W
TP E
stim
ate
# of
Val
ues
in
Acu
te M
eta-
Ana
lysi
s
12
1 K
artm
an B
, et
al.
“WTP
for
Red
uctio
ns in
A
ngin
a Pe
ctor
is
Atta
cks”
1996
M
edic
al
Dec
isio
n M
akin
g 16
(3):
248-
53
CV
Te
leph
one
Inte
rvie
ws,
199
3-19
94
Red
uctio
ns
(50%
, 25%
, 75
%) i
n an
gina
pe
ctor
is w
eekl
y at
tack
rat
e ov
er
a pe
riod
of 3
m
onth
s
Swed
ish
angi
na p
ecto
ris
patie
nts
rece
ivin
g dr
ug
trea
tmen
t in
tend
ed fo
r an
gina
pec
tori
s91.2
% o
f pat
ient
s ha
d st
able
ang
ina
and
the
rem
aini
n g
8.8%
had
un
stab
le a
ngin
a.
The
mea
n w
eekl
y at
tack
rat
e w
as
4.8
atta
cks
per
wee
k.
Dis
cret
e ch
oice
with
a
bidd
ing
gam
e fo
llow
up
Mea
n W
TP
rang
ed fr
om
1145
to 3
350
(199
4 SE
K) f
or
vary
ing
perc
ent
redu
ctio
ns
12
14
1 K
eith
PL,
et
al.
“A C
BA
Usi
ng a
W
TP
Que
stio
nnai
re o
f In
tran
asal
B
udes
onid
e fo
r Se
ason
al A
llerg
ic
Rhi
nitis
”
2000
A
nn
Alle
rgy
Ast
hma
Imm
un
84: 5
5-62
CV
Se
lf-ad
min
iste
red
ques
tionn
aire
. 199
3 In
tran
asal
be
deso
nide
for
trea
tmen
t of
seas
onal
al
lerg
ic r
hini
tis
Patie
nts
olde
r th
an 1
8 w
ith
seas
onal
al
lerg
y sy
mpt
oms
(pos
itive
ski
n pr
ick
for
ragw
eed)
Mod
erat
e se
ason
al a
llerg
y sy
mpt
oms
Ope
n-en
ded
Prio
r to
tr
eatm
ent m
ean
WTP
$1
5.89
/wee
k,
follo
win
g tr
eatm
ent
$12.
95/w
eek
(199
3 C
an$)
.
1
15
1 Li
u Jin
-Tan
, et
al.
“Mot
her’
s W
TP fo
r H
er O
wn
and
Her
C
hild
’s H
ealth
: a
Con
tinge
nt
Val
uatio
n St
udy
in
Taiw
an”
2000
H
ealth
Ec
onom
ics
9: 3
19-2
6
CV
In
-per
son
surv
ey,
1995
Pr
even
tion
of
mot
hers
and
th
eir
child
ren
from
sym
ptom
s eq
uiva
lent
to
the
mos
t rec
ent
cold
ex
peri
ence
d by
th
e re
spon
dent
.
Mot
hers
of
prim
ary
scho
ol
stud
ents
in
Taiw
an
Perc
ent o
f re
spon
dent
s w
ith
sym
ptom
s du
ring
la
st c
old:
57.
5%
with
hea
dach
e,
61.5
% w
ith
coug
h, 1
2% w
ith
feve
r, 7
2.6%
with
a
doct
or v
isit,
12
% lo
st a
day
of
wor
k. A
vera
ge
QW
B w
as 0
.66.
Th
e av
erag
e du
ratio
n of
the
last
col
d w
as 6
.48
days
.
Dic
hoto
-m
ous
with
2
follo
w-u
p qu
estio
ns
yiel
ding
a
trip
le-
boun
ded
bina
ry
choi
ce
form
at
Med
ian
WTP
to
avoi
d th
e sy
mpt
oms
and
dura
tion
of th
e la
st c
old
was
$3
7.3
(199
5 U
S$).
Med
ian
WTP
to in
crea
se
QW
B in
dex
from
0.6
6 to
1.
00 w
as $
37.5
(1
995
US$
).
1
(con
tinue
d)
Appendix B — Annotated Bibliography of Publications Included in the Morbidity Value Meta-Analysis
B-5
Ta
ble
B-1
. A
nn
ota
ted
Bib
lio
gra
ph
y o
f S
tud
ies
Inc
luded in t
he A
cute
Eff
ects
Me
ta-A
na
lysi
s (c
on
tin
ue
d)
Stud
y ID
Pu
b
ID
Aut
hors
A
rtic
le T
itle
Y
ear
Sour
ce
Val
ue
Elic
itat
ion
Met
hod
Mod
e of
A
dmin
istr
atio
n an
d Y
ear
Com
mod
ity
Def
init
ion
Targ
et
Popu
lati
on
Bas
elin
e H
ealt
h of
St
udy
Popu
lati
on
Que
stio
n Fo
rmat
W
TP E
stim
ate
# of
Val
ues
in
Acu
te M
eta-
Ana
lysi
s
17
1 R
eady
RC
, N
avru
d S,
an
d W
R
Dub
org.
“How
do
Res
pond
ents
with
U
ncer
tain
W
illin
gnes
s to
Pa
y A
nsw
er
Con
tinge
nt
Val
uatio
n Q
uest
ions
.”
2001
La
nd
Econ
omic
s 77
(3):
315-
26.
CV
In
-Per
son
Inte
rvie
ws,
199
8
Avo
id: a
) one
da
y of
cou
gh b
) th
ree
days
of f
lu
and
c)
Res
pira
tory
sy
mpt
oms
requ
irin
g ho
spita
lizat
ion
for
3 da
ys
follo
wed
by
5 da
ys a
t hom
e in
be
d
Gen
eral
po
pula
tion
of
Osl
o, N
orw
ay N
ot s
peci
fied
Dic
hoto
-m
ous
choi
ce a
nd
Paym
ent
Car
d fo
rmat
s w
ith
cert
aint
y-fo
llow
up
Mea
n W
TP to
av
oid
a) o
ne d
ay
of c
ough
ran
ged
from
93
NO
K to
14
3 N
OK
; b)
thre
e da
ys o
f flu
ra
n ged
from
380
N
OK
to 6
29
NO
K; a
nd c
) re
spir
ator
y sy
mpt
oms
rang
ed fr
om
1016
NO
K to
10
86 N
OK
(1
998
valu
es)
10
17
2 R
eady
RC
, N
avru
d S,
D
ay B
, D
ubor
g W
R,
Mac
hado
F,
et a
l.
Ben
efit
Tran
sfer
in
Euro
pe: A
re
Val
ues
Con
sist
ent
Acr
oss
Cou
ntri
es?
1999
W
orki
ng
Pape
r C
V
In-p
erso
n in
terv
iew
s, 1
998.
A
dditi
onal
day
s of
sev
en li
ght
heal
th
sym
ptom
s (c
ough
ing,
sin
us
cong
estio
n,
thro
at
cong
estio
n, e
ye
irri
tatio
n,
head
ache
, sh
ortn
ess
of
brea
th a
nd
acut
e br
onch
itis)
and
as
thm
a.
Gen
eral
po
pula
tion
of
Am
ster
dam
, N
ethe
rlan
ds;
Osl
o, N
orw
ay;
Vig
o, S
pain
; Li
sbon
, Po
rtug
al; a
nd
Engl
and
Not
spe
cifie
d Pa
ymen
t ca
rd,
itera
tive
bidd
ing
WTP
to a
void
va
riou
s co
mbi
natio
ns o
f sy
mpt
oms
and
dura
tions
ran
ged
from
13.
56 to
42
5.7
(199
8 B
ritis
h po
unds
)
27
25
1 N
avru
d S.
“V
alui
ng H
ealth
Im
pact
s fr
om A
ir
Pollu
tion
in
Euro
pe.”
2001
En
viro
n-m
enta
l an
d R
esou
rce
Econ
omic
s 20
:305
-29.
CV
In
-per
son
inte
rvie
ws.
199
6.
Avo
id
addi
tiona
l day
s (1
or
14 d
ays)
of
seve
n lig
ht
heal
th
sym
ptom
s (c
ough
ing,
sin
us
cong
estio
n,
thro
at
Rep
rese
ntat
ive
sam
ple
of
Nor
weg
ians
ol
der
than
15
year
s of
age
.
Not
spe
cifie
d O
pen-
ende
d M
ean
WTP
to
avoi
d ad
ditio
nal
days
of o
ne o
f se
ven
sym
ptom
s or
ast
hma
rang
ed fr
om 9
9 N
OK
to 1
772
NO
K (1
996
NO
K)
18
(con
tinue
d)
Valuation of Morbidity Losses: Meta-Analysis of Willingness-to-Pay and Health Status Measures
B-6
Ta
ble
B-1
. A
nn
ota
ted
Bib
lio
gra
ph
y o
f S
tud
ies
Inc
luded in t
he A
cute
Eff
ects
Me
ta-A
na
lysi
s (c
on
tin
ue
d)
Stud
y ID
Pu
b
ID
Aut
hors
A
rtic
le T
itle
Y
ear
Sour
ce
Val
ue
Elic
itat
ion
Met
hod
Mod
e of
A
dmin
istr
atio
n an
d Y
ear
Com
mod
ity
Def
init
ion
Targ
et
Popu
lati
on
Bas
elin
e H
ealt
h of
St
udy
Popu
lati
on
Que
stio
n Fo
rmat
W
TP E
stim
ate
# of
Val
ues
in
Acu
te M
eta-
Ana
lysi
s
Nav
rud
S.
(con
t’d)
co
nges
tion,
eye
ir
rita
tion,
he
adac
he,
shor
tnes
s of
br
eath
and
ac
ute
bron
chiti
s) a
nd
asth
ma.
29
1 Lo
ehm
an
ET, e
t al
. “D
istr
ibut
iona
l A
naly
sis
of
Reg
iona
l Ben
efit
s an
d C
ost
of A
ir
Qua
lity
Con
trol
”
1979
Jo
urna
l of
Envi
ron-
men
tal
Econ
omic
s an
d M
anag
e-m
ent
6:
222-
243.
CV
M
ail q
uest
ionn
aire
, 19
77
Avo
id
sym
ptom
s of
ex
posu
re t
o ai
r po
lluti
on
vary
ing
in
seve
rity
and
du
rati
on
Gen
eral
po
pula
tion
of
Tam
pa B
ay
area
Not
spe
cifie
d C
lose
d-en
ded;
pa
ymen
t ca
rd
Med
ian
WTP
to
avoi
d sy
mpt
oms
of v
aryi
ng
seve
rity
and
du
rati
on r
ange
d fr
om $
2.31
to
$493
.16
(US$
).
36
30
1 R
owe
RD
an
d C
hest
nut
L
“Oxi
dant
s an
d A
sthm
atic
s in
Los
A
ngel
es: A
Ben
efit
A
naly
sis”
1985
U
.S. E
PA
Rep
ort
CV
In
per
son
inte
rvie
w; 1
983
Fift
y pe
rcen
t re
duct
ion
in
bad
asth
ma
days
Ast
hma
pati
ents
who
ha
d pa
rtic
ipat
ed in
th
e C
OR
D
stud
y an
d liv
e in
Gle
ndor
a,
CA
All
pati
ents
wit
h as
thm
a C
lose
d-en
ded;
pa
ymen
t ca
rd
Mea
n W
TP w
as
$401
(19
83
US$
)
1
32
1 To
lley
G
and
Bab
cock
L
“Val
uati
on o
f R
educ
tion
s in
H
uman
Hea
lth
Sym
ptom
s an
d R
isks
”
1986
U
.S. E
PA
Rep
ort
CV
In
per
son
inte
rvie
w; 1
985
Ligh
t sy
mpt
om
redu
ctio
ns a
nd
angi
na r
elie
f for
va
ryin
g du
rati
ons
Ran
dom
se
lect
ion
of
resi
dent
s of
C
hica
go I
L an
d D
enve
r C
O
Not
spe
cifie
d C
lose
d en
ded
iter
ativ
e bi
ddin
g w
ith
open
-en
ded
follo
wup
Mea
n W
TP t
o av
oid
rang
ed
from
$25
.2 t
o $8
68.8
9 fo
r va
riou
s sy
mpt
oms
and
dura
tion
s
20
37
1 Le
e PY
, M
atch
ar D
, C
lem
ents
D
, Hub
er J,
et
al.
Econ
omic
Ana
lysi
s of
Influ
enza
V
acci
natio
n an
d A
ntiv
iral
Tr
eatm
ent f
or
Hea
lthy
Wor
king
A
dults
2002
A
nnal
s of
In
tern
al
Med
icin
e 13
7(4)
: 22
5-31
.
Con
join
t an
alys
is
In p
erso
n in
terv
iew
, ye
ar n
ot a
vaila
ble
One
day
of
relie
f fro
m
influ
enza
, na
usea
, and
di
zzin
ess
Adu
lts a
ged
18-5
0 in
Nor
th
Car
olin
a w
ith
flu e
xper
ienc
e an
d no
ser
ious
ch
roni
c co
nditi
on
No
sign
ifica
nt
com
orbi
d co
nditi
ons
Perc
enta
ge
chan
ce th
at
resp
onde
nt
wou
ld
choo
se a
pa
rtic
ular
op
tion
Mea
n W
TP to
av
oid
influ
enza
w
as $
15.4
9,
naus
ea $
61.7
9,
and
dizz
ines
s $5
6.39
(200
1 U
S$)
3
(con
tinue
d)
Appendix B — Annotated Bibliography of Publications Included in the Morbidity Value Meta-Analysis
B-7
Ta
ble
B-1
. A
nn
ota
ted
Bib
lio
gra
ph
y o
f S
tud
ies
Inc
luded in t
he A
cute
Eff
ects
Me
ta-A
na
lysi
s (c
on
tin
ue
d)
Stud
y ID
Pu
b
ID
Aut
hors
A
rtic
le T
itle
Y
ear
Sour
ce
Val
ue
Elic
itat
ion
Met
hod
Mod
e of
A
dmin
istr
atio
n an
d Y
ear
Com
mod
ity
Def
init
ion
Targ
et
Popu
lati
on
Bas
elin
e H
ealt
h of
St
udy
Popu
lati
on
Que
stio
n Fo
rmat
W
TP E
stim
ate
# of
Val
ues
in
Acu
te M
eta-
Ana
lysi
s
38
1 To
rran
ce G
, W
alke
r V
, G
ross
man
R
, et a
l.
Econ
omic
Ev
alua
tion
of
Cip
roflo
xaci
n C
ompa
red
with
th
e U
sual
A
ntib
acte
rial
Car
e fo
r th
e Tr
eatm
ent
of A
cute
Ex
acer
batio
ns o
f C
hron
ic
Bro
nchi
tis in
Pa
tient
s Fo
llow
ed
for
1 Y
ear
1999
Ph
arm
aco
econ
omic
s 16
(5 P
t. 1)
: 499
-52
0.
CV
Q
uest
ionn
aire
at
phys
icia
n vi
sit,
Nov
embe
r, 1
993–
June
, 199
4.
Sym
ptom
day
s of
acu
te
exac
erba
tion
of
chro
nic
bron
chiti
s
Out
patie
nt
adul
t men
and
w
omen
age
d 18
or
olde
r w
ith c
hron
ic
bron
chiti
s pa
rtic
ipat
ing
in
a cl
inic
al tr
ial
for
cipr
oflo
xaci
n
18%
mild
chr
onic
br
onch
itis,
82%
m
oder
ate,
15
%
seve
re
Ope
n-en
ded
with
ce
rtai
nty
follo
wup
.
Mea
n W
TP
amon
g th
ose
in
the
trea
ted
with
ci
prof
loxa
cin
grou
p w
as
$1,2
35 (1
994
Can
$) a
nd $
868
(199
4 C
an$)
am
ong
thos
e in
th
e us
ual c
are
grou
p
2
39
1 Ja
cobs
R,
Mol
eski
RJ,
and
AS
Mey
erho
ff
Val
uatio
n of
Sy
mpt
omat
ic
Hep
atiti
s A
in
Adu
lts
2002
Ph
arm
aco
econ
omic
s 20
(11)
: 73
9-47
.
CV
M
ail s
urve
y, 2
001.
R
isk
free
pr
even
tion
of
all d
isea
se
sym
ptom
s fr
om
hepa
titis
A
Gen
eral
po
pula
tion
of
Uni
ted
Stat
es
31%
had
a
pers
onal
or
fam
ily
hist
ory
of
hepa
titis
. Th
e m
ean
SF-3
6 ph
ysic
al h
ealth
sc
ore
was
48
and
the
mea
n SF
-36
men
tal h
ealth
sc
ore
was
50.
Paym
ent
card
M
ean
WTP
is
$3,0
11 (2
001
US$
)
1
Not
e: S
tudi
es in
clud
ed in
the
John
son
(199
7) M
eta-
Ana
lysi
s ar
e in
bol
d.
Valuation of Morbidity Losses: Meta-Analysis of Willingness-to-Pay and Health Status Measures
B-8
Ta
ble
B-2
. A
nn
ota
ted
Bib
lio
gra
ph
y o
f S
tud
ies
Inc
lud
ed
in
th
e C
hro
nic
Eff
ec
ts M
eta
-An
aly
sis
Stud
y ID
Pu
b ID
A
utho
rs
Art
icle
Tit
le
Yea
r So
urce
V
alue
El
icit
atio
n M
etho
d
Mod
e of
A
dmin
istr
atio
n an
d Y
ear
Com
mod
ity
Def
init
ion
Targ
et
Popu
lati
on
Bas
elin
e H
ealt
h of
Stu
dy
Popu
lati
on
Que
stio
n Fo
rmat
W
TP E
stim
ate
# of
Val
ues
in
Chr
onic
Met
a-A
naly
sis
4 1
Blu
men
sche
in,
K a
nd
Joha
nnes
son,
M “R
elat
ions
hip
betw
een
qual
ity o
f life
in
stru
men
ts,
heal
th s
tate
ut
ilitie
s an
d w
illin
gnes
s to
pa
y in
pa
tient
s w
ith
asth
ma.
”
1998
A
nn A
llerg
y A
sthm
a Im
mun
80:
18
9-94
.
CV
In
-Per
son
Inte
rvie
ws,
ye
ar n
ot
avai
labl
e
Trea
tmen
t tha
t w
ill c
ure
resp
onde
nts
from
ast
hma
for
one
mon
th
Ast
hma
patie
nts
in
Cen
tral
and
Ea
ster
n K
entu
cky
age
18 a
nd o
lder
All
patie
nts
have
ast
hma
Dic
hoto
-m
ous
choi
ce
alon
e an
d di
chot
omou
s ch
oice
w
ith b
iddi
n g
gam
e fo
llow
-up.
Mea
n W
TP
rang
es fr
om
$189
to $
343
(US$
)
2
4 2
Zill
ich
AJ,
Blu
men
sche
in
K, J
ohan
ness
on
M, a
nd
Free
man
P
“Ass
essm
ent
of th
e R
elat
ions
hip
Bet
wee
n M
easu
res
of
Dis
ease
Se
veri
ty,
Qua
lity
of
Life
, and
W
illin
gnes
s to
Pay
in
Ast
hma”
2002
Ph
arm
aco-
econ
omic
s 20
(4):
257-
265.
CV
In
-Per
son
Inte
rvie
ws,
ye
ar n
ot
avai
labl
e
Cur
e fo
r as
thm
a cl
assi
fied
as
mild
, m
oder
ate,
or
seve
re.
Ast
hma
patie
nts
18
year
s or
old
er
in C
entr
al a
nd
East
ern
Ken
tuck
y.
All
patie
nts
with
a
diag
nosi
s of
as
thm
a ob
ject
ivel
y cl
assi
fied
as
mild
, mod
erat
e or
sev
ere,
re
ceiv
ing
an
inha
led
B2
agon
ist
med
icat
ion
with
or
with
out
inha
led
cort
icos
tero
id
med
icat
ion.
Dic
hoto
-m
ous
choi
ce
Mea
n W
TP
ran g
es fr
om $
48
to $
331
(US$
) fo
r a
cure
for
mild
, mod
erat
e,
or s
ever
e as
thm
a
6
13
1 K
artm
an B
, et
al.
“Val
uatio
n of
H
ealth
C
han g
es w
ith
the
CV
M
etho
d”
1996
H
ealth
Ec
onom
ics
5: 5
31-4
1
CV
Te
leph
one
inte
rvie
w.
Oct
ober
, 199
4 –S
epte
mbe
r,
1995
(1) A
sho
rt-
term
trea
tmen
t th
at in
crea
ses
the
prob
abili
ty
of b
eing
free
fr
om
sym
ptom
s af
ter
4 w
eeks
; (2)
a
long
-ter
m
trea
tmen
t tha
t re
duce
s th
e ri
sk o
f hav
ing
a re
laps
e on
ce
reco
vere
d; (3
) a
med
icat
ion
Patie
nts
with
re
flux
oeso
phag
itis
in
Swed
en
All
patie
nts
diag
nose
d w
ith
and
rece
ivin
g m
edic
atio
n fo
r re
flux
oeso
phag
itis
Dic
hoto
-m
ous
choi
ce w
ith
open
-end
ed
follo
w-u
p
Mea
n W
TP
rang
ed fr
om
431
to 1
023
(SEK
199
5) fo
r re
duct
ion
in r
isk
of s
ympt
oms
and
from
261
to
912
(SEK
199
5)
for
redu
ctio
n in
ri
sk o
f rel
apse
4
(con
tinue
d)
Appendix B — Annotated Bibliography of Publications Included in the Morbidity Value Meta-Analysis
B-9
Ta
ble
B-2
. A
nn
ota
ted
Bib
lio
gra
ph
y o
f S
tud
ies
Inc
luded in t
he C
hro
nic
Eff
ects
Me
ta-A
na
lysi
s (c
on
tin
ue
d)
Stud
y ID
Pu
b ID
A
utho
rs
Art
icle
Tit
le
Yea
r So
urce
V
alue
El
icit
atio
n M
etho
d
Mod
e of
A
dmin
istr
atio
n an
d Y
ear
Com
mod
ity
Def
init
ion
Targ
et
Popu
lati
on
Bas
elin
e H
ealt
h of
Stu
dy
Popu
lati
on
Que
stio
n Fo
rmat
W
TP E
stim
ate
# of
Val
ues
in
Chr
onic
Met
a-A
naly
sis
Kar
tman
B, e
t al
. (co
nt’d
)
that
can
be
take
n w
ith
mea
ls, a
s co
mpa
red
with
a
med
icat
ion
that
mus
t be
take
n at
leas
t 1
hour
bef
ore
a m
eal.
21
1 V
iscu
si W
K,
Ma g
at W
A, a
nd
J Hub
er.
“Pri
cing
En
viro
nmen
t-al
Hea
lth
Ris
ks: S
urve
y A
sses
smen
ts
of R
isk-
Ris
k an
d R
isk-
Dol
lar
Trad
e-O
ffs fo
r C
hron
ic
Bro
nchi
tis.”
1991
Jo
urna
l of
Envi
ron-
men
tal
Econ
omic
s an
d M
anag
e-m
ent 2
1:
32-5
1.
CJ/P
aire
d-C
ompa
riso
n In
tera
ctiv
e co
mpu
ter
ques
tionn
aire
, ye
ar n
ot
avai
labl
e.
Ris
k of
co
ntra
ctin
g ch
roni
c br
onch
itis
and
risk
of f
atal
au
to a
ccid
ent.
Gre
ensb
oro
mal
l sho
pper
s N
ot s
peci
fied
Itera
tive
bidd
ing
R
espo
nden
ts
wer
e w
illin
g to
su
bstit
ute
a m
ean
cost
of
livin
g in
crea
se
of $
8.83
(US$
) pe
r ye
ar fo
r a
redu
ctio
n in
the
risk
of g
ettin
g ch
roni
c br
onch
itis
and
of $
81.8
4 (U
S$)
per
year
to
redu
ce th
e ri
sk
of g
ettin
g in
a
fata
l aut
o ac
cide
nt
1
23
1 K
rupn
ick
AJ
and
M C
ropp
er “T
he E
ffect
of
Info
rmat
ion
on H
ealth
R
isk
Val
uatio
ns.”
1992
Jo
urna
l of
Ris
k an
d U
ncer
tain
ty
5: 2
9-48
.
CJ/P
aire
d-C
ompa
riso
n C
ompu
ter
inte
rvie
w, y
ear
not a
vaila
ble
Ris
k of
co
ntra
ctin
g ch
roni
c br
onch
itis
Was
hing
ton,
D
C a
rea
resi
dent
s ov
er
age
18 th
at
had
a re
lativ
e ov
er 2
1 ye
ars
old
who
had
a
chro
nic
resp
irat
ory
cond
ition
.
Non
e of
the
resp
onde
nts
had
a hi
stor
y of
ch
roni
c re
spir
ator
y co
nditi
on
Itera
tive
bidd
ing
M
ean
cost
of
livin
g in
crea
se
rang
ing
from
$1
1 to
$21
(U
S$)
3
(con
tinue
d)
Valuation of Morbidity Losses: Meta-Analysis of Willingness-to-Pay and Health Status Measures
B-10
Ta
ble
B-2
. A
nn
ota
ted
Bib
lio
gra
ph
y o
f S
tud
ies
Inc
luded in t
he C
hro
nic
Eff
ects
Me
ta-A
na
lysi
s (c
on
tin
ue
d)
Stud
y ID
Pu
b ID
A
utho
rs
Art
icle
Tit
le
Yea
r So
urce
V
alue
El
icit
atio
n M
etho
d
Mod
e of
A
dmin
istr
atio
n an
d Y
ear
Com
mod
ity
Def
init
ion
Targ
et
Popu
lati
on
Bas
elin
e H
ealt
h of
Stu
dy
Popu
lati
on
Que
stio
n Fo
rmat
W
TP E
stim
ate
# of
Val
ues
in
Chr
onic
Met
a-A
naly
sis
27
1 Sl
oan
FA, e
t al.
“Alte
rnat
ive
App
roac
hes
to v
alui
n g th
e In
tang
ible
H
ealth
Lo
sses
: The
Ev
iden
ce o
f M
ultip
le
Scle
rosi
s.”
1998
Jo
urna
l of
Hea
lth
Econ
omic
s 17
: 475
-97.
CJ/P
aire
d-C
ompa
riso
n In
-per
son
and
com
pute
r in
terv
iew
s,
1995
Ris
k-ris
k an
d ri
sk d
olla
r tr
adeo
ffs fo
r de
ath
and
mul
tiple
sc
lero
sis
Mal
l int
erce
pt
in G
reen
sbor
o,
NC
re
pres
enta
tive
of th
e ge
nera
l po
pula
tion;
M
embe
rs o
f th
e Ea
ster
n N
orth
Car
olin
a M
ultip
le
Scle
rosi
s So
ciet
y in
O
rang
e an
d D
urha
m
Cou
ntie
s, N
C.
13%
hav
e m
ultip
le
scle
rosi
s. T
he
mea
n ra
nk o
f cu
rren
t hea
lth
(on
a sc
ale
of 0
w
orst
to 1
00
best
) was
73.
79. Ite
rativ
e bi
ddin
g
Med
ian
WTP
pe
r ye
ar r
ange
d fr
om $
419,
000
to $
510,
000
for
the
low
-pr
obab
ility
sc
enar
io in
the
gene
ral s
ampl
e an
d fr
om
$346
,000
to
$420
,000
for
the
high
-pr
obab
ility
sc
enar
io in
the
gene
ral s
ampl
e.
It ra
nged
from
$5
83,0
00 to
$8
81,0
00 fo
r th
e lo
w-
prob
abili
ty
scen
ario
in th
e M
S sa
mpl
e an
d fr
om $
375,
000
to $
566,
000
for
the
high
-pr
obab
ility
sc
enar
io in
the
MS
sam
ple
(199
6 U
S$)
8
28
1 St
avem
K
“Will
ingn
ess
to P
ay: A
Fe
asib
le
Met
hod
for
Ass
essi
ng
Trea
tmen
t B
enef
its in
Ep
ileps
y?”
1999
Se
izur
e 8:
14
-19.
C
V
In-p
erso
n in
terv
iew
s,
year
not
av
aila
ble
Perm
anen
t cu
re fo
r ep
ileps
y
Patie
nts
aged
18
-67
who
ha
d be
en
adm
itted
to th
e ou
tpat
ient
cl
inic
for
epile
psy
at th
e C
entr
al
Hos
pita
l of
Aer
shus
in
Nor
way
be
twee
n 19
87
and
1994
24.5
% h
ad
seiz
ures
last
ye
ar,7
5.4%
us
ing
med
icat
ion
Ope
n-en
ded
Med
ian
WTP
is
$20,
000
(US$
) to
cur
e ep
ileps
y.
1
(con
tinue
d)
Appendix B — Annotated Bibliography of Publications Included in the Morbidity Value Meta-Analysis
B-11
Ta
ble
B-2
. A
nn
ota
ted
Bib
lio
gra
ph
y o
f S
tud
ies
Inc
luded in t
he C
hro
nic
Eff
ects
Me
ta-A
na
lysi
s (c
on
tin
ue
d)
Stud
y ID
Pu
b ID
A
utho
rs
Art
icle
Tit
le
Yea
r So
urce
V
alue
El
icit
atio
n M
etho
d
Mod
e of
A
dmin
istr
atio
n an
d Y
ear
Com
mod
ity
Def
init
ion
Targ
et
Popu
lati
on
Bas
elin
e H
ealt
h of
Stu
dy
Popu
lati
on
Que
stio
n Fo
rmat
W
TP E
stim
ate
# of
Val
ues
in
Chr
onic
Met
a-A
naly
sis
31
1 Z
ethr
aeus
N
“Will
ingn
ess
to P
ay fo
r H
orm
one
Rep
lace
men
t Th
erap
y.”
1998
H
ealth
Ec
onom
ics
7: 3
1-38
.
CV
In
-per
son
inte
rvie
ws.
19
95-1
996.
Hor
mon
e re
plac
emen
t th
erap
y
Wom
en
betw
een
age
45-6
5 th
at h
ad
been
trea
ted
with
HR
T fo
r at
leas
t 1
mon
th.
Rec
ruite
d fr
om
the
depa
rtm
ent
of G
ynec
olog
y at
the
Sode
rtal
e H
ospi
tal i
n Sw
eden
Not
spe
cifie
d C
lose
d-en
ded
with
ce
rtai
nty
follo
w u
p
Mea
n W
TP fo
r H
RT
rang
ed
from
365
1 SE
K
to 3
772
SEK
(1
996
valu
es)
2
31
2 Z
ethr
aeus
N,
Joha
nnes
son
M,
Hen
riks
son
P,
and
RT
Stra
nd
“The
Impa
ct
of H
orm
one
Rep
lace
men
t Th
erap
y on
Q
ualit
y of
life
an
d W
illin
gnes
s to
Pay
.”
1997
H
ealth
Ec
onom
ics
7: 3
1-38
.
CV
In
-per
son
inte
rvie
ws.
19
95-1
996.
Hor
mon
e re
plac
emen
t th
erap
y fo
r a
redu
ctio
n in
m
enop
ause
sy
mpt
oms
Wom
en
betw
een
age
45-6
5 th
at h
ad
been
trea
ted
with
HR
T fo
r at
leas
t 1
mon
th.
Rec
ruite
d fr
om
the
depa
rtm
ent
of G
ynec
olog
y at
the
Sode
rtal
e H
ospi
tal i
n Sw
eden
Self-
rate
d (b
ased
on
inte
rvie
wer
de
scri
ptio
n)
mild
or
seve
re
men
opau
sal
sym
ptom
s, 5
6 w
ith m
ild a
nd
48 w
ith s
ever
e sy
mpt
oms
Clo
sed-
ende
d w
ith
cert
aint
y fo
llow
up
Mea
n W
TP fo
r H
RT
and
a re
duct
ion
in
mild
sym
ptom
s w
as 2
346
SEK
an
d 48
38 S
EK
for
seve
re
sym
ptom
s (1
996
valu
es)
2
33
1 Th
omps
on M
S,
et a
l. “
WTP
and
A
ccep
t Ris
ks
to C
ure
Chr
onic
D
isea
ses”
1986
A
mer
ican
Jo
urna
l of
Publ
ic
Hea
lth
76(4
):392
-39
6
CV
In
-per
son
Inte
rvie
w, y
ear
not a
vaila
ble
Cur
e fo
r rh
eum
atoi
d ar
thri
tis w
ith a
ri
sk o
f dea
th
Patie
nts
with
ad
ult-
onse
t rh
eum
atoi
d ar
thri
tis,
enro
lled
in a
ra
ndom
ized
co
ntro
lled
drug
tria
l, w
ho
had
mai
ntai
ned
for
at le
ast t
hree
m
onth
s on
ba
sic
All
patie
nts
with
ad
ult-
onse
t, un
rem
ittin
g rh
eum
atoi
d ar
thri
tis
Ope
n-en
ded
Mea
n W
TP
rang
ed fr
om
$4,4
75 to
$7
,752
(US$
)
4
(con
tinue
d)
Valuation of Morbidity Losses: Meta-Analysis of Willingness-to-Pay and Health Status Measures
B-12
Ta
ble
B-2
. A
nn
ota
ted
Bib
lio
gra
ph
y o
f S
tud
ies
Inc
luded in t
he C
hro
nic
Eff
ects
Me
ta-A
na
lysi
s (c
on
tin
ue
d)
Stud
y ID
Pu
b ID
A
utho
rs
Art
icle
Tit
le
Yea
r So
urce
V
alue
El
icit
atio
n M
etho
d
Mod
e of
A
dmin
istr
atio
n an
d Y
ear
Com
mod
ity
Def
init
ion
Targ
et
Popu
lati
on
Bas
elin
e H
ealt
h of
Stu
dy
Popu
lati
on
Que
stio
n Fo
rmat
W
TP E
stim
ate
# of
Val
ues
in
Chr
onic
Met
a-A
naly
sis
Thom
pson
MS,
et
al.
(con
t’d)
cons
erva
tive
prog
ram
s in
clud
ing
rest
an
d ph
ysic
al
ther
apie
s,
salic
ylat
es,
nons
tero
idal
an
ti-in
flam
mat
ory
drug
s
34
1 St
avem
K
“Ass
ocia
tion
of w
illin
gnes
s to
pay
with
se
veri
ty o
f ch
roni
c ob
stru
ctiv
e pu
lmon
ary
dise
ase,
he
alth
sta
tus,
an
d ot
her
pref
eren
ce
mea
sure
s”
2002
In
tern
atio
nal
Jour
nal o
f Tu
berc
ulos
is
and
Lung
D
isea
se
6(6)
: 542
-54
9.
CV
In
-Per
son
Inte
rvie
ws,
19
94-1
995
Cur
e fo
r C
OPD
with
out
side
-effe
cts
Patie
nts
aged
18
-67
with
C
OPD
see
n at
th
e C
entr
al
Hos
pita
l of
Ake
rshu
s,
Nor
way
be
twee
n 19
94
and
1995
.
Patie
nts
with
C
OPD
, for
ced
expi
rato
ry
volu
me
in o
ne
seco
nd <
70
per
cent
of
pred
icte
d,
impr
ovem
ent
afte
r in
hala
tion
of b
eta-
2 a g
onis
t < 1
5 pe
r ce
nt in
FEV
1 or
pr
evio
usly
un
know
n
Paym
ent
card
M
edia
n W
TP o
f 20
0,00
0 N
OK
(1
994
valu
e)
1
35
1 Lu
ndbe
rg L
, et
al.
“Qua
lity
of
life,
hea
lth-
stat
e ut
ilitie
s an
d w
illin
gnes
s to
pa
y in
pa
tient
s w
ith
psor
iasi
s an
d at
opic
ec
zem
a”
1999
B
ritis
h Jo
urna
l of
Der
mat
olog
y 14
1:
1067
-107
5. C
V
In-P
erso
n In
terv
iew
s.
Nov
embe
r 19
96 -
D
ecem
ber
1997
.
Cur
e fo
r ps
oria
sis
and
atop
ic e
czem
a w
ithou
t sid
e-ef
fect
s
Patie
nts
aged
17
-73
with
ps
oria
sis
or
atop
ic e
czem
a w
ho h
ad
atte
nded
the
derm
atol
ogy
outp
atie
nt
clin
ic a
t the
U
nive
rsity
ho
spita
l in
Upp
sala
, Sw
eden
from
N
ovem
ber
1996
to
Dec
embe
r 19
97.
Mea
n he
alth
st
ate
utili
ty w
as
0.69
(rat
ing
scal
e), 0
.88
(tim
e tr
ade-
off),
an
d 0.
97
(sta
ndar
d ga
mbl
e) fo
r pa
tient
s w
ith
psor
iasi
s an
d 0.
73 (r
atin
g sc
ale)
, 0.9
3 (ti
me
trad
e-of
f),
and
0.98
(s
tand
ard
gam
ble)
for
patie
nts
with
at
opic
ecz
ema.
Dic
hoto
mou
s ch
oice
and
bi
ddin
g ga
me.
Mea
n W
TP
rang
ed fr
om
960.
2 SE
K to
19
55.9
SEK
4
Appendix C: Summary Statistics for the Morbidity Value Database
C-1
Table C-1. Summary Statistics for the Morbidity Value Database
Variable Name Number of
Observations Mean Standard Deviation Minimum Maximum
numauthor 388 5.332474 4.495795 1 18
pubyr 389 1,994.47 7.424692 1,979 2,002
pubjrl 268 1 0 1 1
pubbk 0
pubbkchap 0
pubtech 59 1 0 1 1
pubwp 62 1 0 1 1
pubphdd 0
pubmt 0
pubconf 0
pubother 0
valueid 389 13.01542 14.26208 1 67
mean 336 866.6709 2,540.2 0.3 3,3746
trimmean 54 1 0 1 1
corrected 7 1 0 1 1
protcorrected 68 1 0 1 1
inconcorrected 40 1 0 1 1
turnbull 10 1 0 1 1
median 129 2,384.436 17,736.46 0 200,000
lowerci 7 15,001.86 37,490.76 24 100,000
upperci 7 43,754.29 11,2999.7 55 300,000
stderr 121 384.7199 1,061.301 0.26 10,310
currencyyr 304 1,994.497 4.62023 1,983 2,001
tfday 88 155.6705 156.5961 7 365
tfonetime 301 1 0 1 1
tfyear 2 42 0 42 42
tfduration 11 1 0 1 1
tfpermanent 1 1 . 1 1
tfpv 0
tfpvdiscount 0
tfpvyrs 0
tfother 2 1 0 1 1
wtp 109 1 0 1 1
wtpavoid 273 1 0 1 1
wta 2 1 0 1 1
(continued)
Valuation of Morbidity Losses: Meta-Analysis of Willingness-to-Pay and Health Status Measures
C-2
Table C-1. Summary Statistics for the Morbidity Value Database (continued)
Variable Name Number of
Observations Mean Standard Deviation Minimum Maximum
wtaforgo 0
mrs 5 1 0 1 1
other 0
hcvmortality 6 1 0 1 1
hcvacutemorb 308 1 0 1 1
hcvchrocon 31 1 0 1 1
hcvtrtmt 1 1 . 1 1
hcvchronic 78 1 0 1 1
hlthocnumchg 195 2.066667 1.377532 1 15
sevchgint 181 1 0 1 1
sevbefore 16 1.791 0.744197 0.656 3
sevafter 16 2.5625 1.931105 0 4
sevchg 0
durationchg 300 1 0 1 1
durbefore 213 10.60909 21.40374 1 90
durafter 227 0.729339 3.497302 0 33.15
durchg 83 –55.1139 440.17 –4016 –1
frequencychg 28 1 0 1 1
freqbefore 18 7.477778 16.71732 0 74
freqafter 7 11.25714 7.694772 0 16
freqchg 15 35.86667 36.13836 –50 75
exante 46 1 0 1 1
riskbefore 28 0.232381 0.349496 1.67E–07 0.8
riskafter 17 0.188378 0.211651 0 0.5
riskchg 42 –0.06944 0.146584 –0.5 0.01
vas 23 1 0 1 1
vasavgbasln 23 35.25522 29.623 0.32 76
vasavgwchng 7 43.33714 53.00325 0.82 100
sg 17 1 0 1 1
sgbasln 14 25.63857 40.91571 0.13 95
sgavgwchg 3 100 0 100 100
tto 11 1 0 1 1
ttobasln 28 3.993929 17.05241 0.18 91
ttoavgwchg 4 0.9225 0.020616 0.9 0.95
hsmother 6 1 0 1 1
hsmotherbasln 5 0.748 0.087293 0.69 0.9
(continued)
Appendix C — Summary Statistics for the Morbidity Value Database
C-3
Table C-1. Summary Statistics for the Morbidity Value Database (continued)
Variable Name Number of
Observations Mean Standard Deviation Minimum Maximum
hsmotheravgwchg 0
aids 0
alcohol 0
allergy 7 1 0 1 1
backneck 0
birfthdef 0
blind 0
blood 0
bone 0
bowel 0
cancer 0
circulat 0
dental 0
depress 0
diabetes 0
digest 0
eareye 2 1 0 1 1
eating 0
epilepsy 1 1 . 1 1
gastroint 8 1 0 1 1
genetic 0
glaucoma 0
heart 89 1 0 1 1
hepatitis 1 1 . 1 1
cholesterol 0
hypertense 0
infertility 0
leprosy 0
liverkidney 0
lupus 0
malaria 0
meningitis 0
mental 5 1 0 1 1
mood 0
ms 8 1 0 1 1
(continued)
Valuation of Morbidity Losses: Meta-Analysis of Willingness-to-Pay and Health Status Measures
C-4
Table C-1. Summary Statistics for the Morbidity Value Database (continued)
Variable Name Number of
Observations Mean Standard Deviation Minimum Maximum
nervoussys 0
neurological 0
obesity 0
osteoporosis 0
palsy 0
parkinsons 0
pregnancy 0
psoriasis 0
respiratory 165 1 0 1 1
schizo 0
sexual 0
skindis 0
sleepdis 2 1 0 1 1
speechdis 1 1 . 1 1
sportsinjuries 0
substanceabuse 0
thyroid 0
illnessnotspec 125 1 0 1 1
illnessother 81 1 0 1 1
cough 90 1 0 1 1
pain 25 1 0 1 1
headache 39 1 0 1 1
nausea 16 1 0 1 1
vomit 4 1 0 1 1
fever 51 1 0 1 1
disorient 12 1 0 1 1
chestpain 21 1 0 1 1
shortbreath 71 1 0 1 1
throat 17 1 0 1 1
eyeirritation 21 1 0 1 1
itching 1 1 . 1 1
sympnotspec 71 1 0 1 1
sympother 0
environair 132 1 0 1 1
environwater 0
(continued)
Appendix C — Summary Statistics for the Morbidity Value Database
C-5
Table C-1. Summary Statistics for the Morbidity Value Database (continued)
Variable Name Number of
Observations Mean Standard Deviation Minimum Maximum
food 8 1 0 1 1
occupation 0
productsafety 0
causesubstance 0
transportation 5 1 0 1 1
naturaldisaster 0
causegenetic 0
infectious 7 1 0 1 1
causenotspec 220 1 0 1 1
causeother 1 1 . 1 1
apyrbegin 360 1,991.681 6.933481 1,977 2,001
apyrend 360 1,991.842 6.973482 1,977 2,001
cv 299 1 0 1 1
hedonic 0
conjoint 83 1 0 1 1
mktv 7 1 0 1 1
othervm 19 1 0 1 1
person 373 1 0 1 1
numperson 389 268.3599 217.1858 5 1,250
choice 30 1 0 1 1
numchoice 30 5.966667 2.470283 1 8
othersample 0
numothersample 0
samplesize 236 566.178 585.609 50 1,800
responserate 117 71.95218 31.73554 6.1 99
numobs 385 1,192.2 2,018.049 3 5,504
sampleus 177 1 0 1 1
samplecan 79 1 0 1 1
sampcntryother 141 1 0 1 1
randial 27 1 0 1 1
ranmail 45 1 0 1 1
mall 17 1 0 1 1
patientrecruit 66 1 0 1 1
otherrecruit 136 1 0 1 1
inclcritage 91 1 0 1 1
(continued)
Valuation of Morbidity Losses: Meta-Analysis of Willingness-to-Pay and Health Status Measures
C-6
Table C-1. Summary Statistics for the Morbidity Value Database (continued)
Variable Name Number of
Observations Mean Standard Deviation Minimum Maximum
inclcritgender 10 1 0 1 1
inclcritparent 17 1 0 1 1
inclcritrace 1 1 . 1 1
inclcrithlthcond 0
inclcirtother 186 1 0 1 1
incomemean 311 84,235.82 90,009.96 13,577 334,080
incomemedian 22 47,002.42 1,611.329 42,553.19 52,500
incomeyr 288 1,991.698 7.583385 1977 2,001
gendermale 310 49.43839 19.48298 0 100
racewhite 50 77.66 7.215658 57 90
agemean 307 45.96344 7.386145 24.36 68
agemedian 4 53.5 5 46 56
agemin 11 43 1.949359 41 45
agemax 11 68.72727 7.072353 65 83
avgedu 185 13.67982 1.462819 9.098 16.1
mail 45 1 0 1 1
inperson 205 1 0 1 1
phone 74 1 0 1 1
computer 87 1 0 1 1
internet 0
othersurvey 13 1 0 1 1
openend 91 1 0 1 1
closedend 209 1 0 1 1
dichochoice 75 1 0 1 1
doublebond 21 1 0 1 1
bidding 73 1 0 1 1
followup 63 1 0 1 1
card 79 1 0 1 1
ranking 1 1 . 1 1
rating 67 1 0 1 1
otherformat 106 1 0 1 1
payctax 1 1 . 1 1
payoutofpock 316 1 0 1 1
payinsurance 0
paycsurcharge 16 1 0 1 1
(continued)
Appendix C — Summary Statistics for the Morbidity Value Database
C-7
Table C-1. Summary Statistics for the Morbidity Value Database (continued)
Variable Name Number of
Observations Mean Standard Deviation Minimum Maximum
paycfee 0
payv 0
paynotv 49 1 0 1 1
paycother 5 1 0 1 1
jobpercunion 0
jobavgwage 0
jobavgwageinc 0
jobavginc 0
jobavghrs 0
linear 0
loglinear 0
semilog 0
rhs 0
lhs 0
numriskchar 0
numjobchar 0
stage1 0
stage1ols 0
stage1other 0
stage2 0
stage2ols 0
stage2other 0
avgunitprice 0
avgunitpurch 7 7.515714 2.2867 4.16 11.35
avgexpperyr 7 27.83857 32.10604 11 100