A cost effectiveness_model_of_screening_strategies.16

ORIGINAL ARTICLE

A Cost-Effectiveness Model of ScreeningStrategies for Amblyopia and Risk Factors and

Its Application in a German Setting

AFSCHIN GANDJOUR, MD, PhD, STEFANIE SCHLICHTHERLE, MD,ANTJE NEUGEBAUER, MD, WALTER RÜSSMANN, MD, and

KARL WILHELM LAUTERBACH, MD, DSs

Institute of Health Economics and Clinical Epidemiology, University of Cologne, Cologne, Germany (AG, SS, KWL), Department ofOphthalmology, University of Cologne, Cologne, Germany (AN, WR)

ABSTRACT: Purpose. To develop a general setting–independent decision-analytical model that determines the costs,effectiveness, and cost-effectiveness of four screening strategies to detect amblyopia or amblyogenic factors inpre-school children and to apply the model in a German setting. Methods. The general setting–independent decision–analytical model was developed from the perspective of society and the statutory health insurance was developed.Outcomes were the total number of newly detected true positive cases of amblyopia and the costs per newly detectedtrue positive case of amblyopia. Strategies were screening of high-risk children up to the age of 1 year (ophthalmol-ogists), screening of all children up to the age of 1 year (ophthalmologists), screening of all children aged 3 to 4 years(pediatricians or general practitioners), and screening of children aged 3 to 4 years visiting kindergarten (orthoptists).For the application example in a German setting, data from the published medical literature were used. Results. In thebase-case analysis of the application example, screening high-risk children by opthalmologists had the lowest averagecost per case detected but became dominated (less effective and more costly than an alternative) if a low (5.3%)probability of familial clustering of strabismus was assumed. Considering the various assumptions tested in thesensitivity analysis, screening of all children up to the age of 1 year by opthalmologists was the only strategy notdominated by others. Detection rates, including cases detected before screening, were between 72% and 78% for thestrategies that screen for all children. Conclusions. The model suggests that in Germany, both from a cost-effectivenessand a pure effectiveness point of view, screening all children up to the age of 1 year by opthalmologists is the preferredstrategy to detect amblyopia or amblyogenic factors. All strategies left a significant portion of children undetected.(Optom Vis Sci 2003;80:259–269)

Key Words: amblyopia, children, cost-effectiveness analysis, Germany, screening

Amblyopia is defined as a unilateral or bilateral decrease invisual acuity for which no cause can be found through thephysical examination of the eye.1 Risk factors associated

with amblyopia are anisometropia, hyperopia, and strabismus. Theprevalence rate of amblyopia among children aged 3 to 4 years isbetween 1% and 5%, depending on the threshold value of visualacuity.2 The prevalence rates of anisometropia, hyperopia, andstrabismus are 3.1%,3 5.5%,4 and 5.3%,5 respectively.

Pre-school screening for amblyopia is recommended by a largenumber of professional societies and experts in the United States,Canada, and Germany.2, 6–14 Recent evidence supports the expertrecommendations. A controlled trial comparing amblyopia screen-

ing and treatment before the age of 37 months with screening andtreatment at 37 months demonstrated the superiority of an earlyapproach.15 Similarly, a prospective cohort study on occlusiontherapy for amblyopia showed greater improvement in visual acu-ity among children 5 years of age or younger.16

Still, the question remains as to whether screening for amblyopiais economically attractive. The high prevalence rate of amblyopiathat leads to high costs of detection and treatment makes thisquestion economically relevant. In this regard, it is also importantto know whether alternative screening options that are currentlynot recommended would be more economically attractive.

Screening by orthoptists has been shown to be effective in the

1040-5488/03/8003-0259/0 VOL. 80, NO. 3, PP. 259–269OPTOMETRY AND VISION SCIENCECopyright © 2003 American Academy of Optometry

Optometry and Vision Science, Vol. 80, No. 3, March 2003

U.K.15, 17–21 Introducing screening for amblyogenic factors up tothe age of 1 year could be an alternative to screening for manifestamblyopia between the ages of 3 and 4 years because the treatmentof amblyogenic factors is assumed to be shorter in duration.22

However, ophthalmologists* would need to perform such a screen-ing if pediatricians and orthoptists lacked the skill to perform theappropriate tests up to the age of 2 (retinoscopy plus examinationof ocular media and fundus). Among strategies for screening foramblyogenic factors, the screening of high-risk children, such aspremature infants and children with a familial predisposition foramblyopia, could be particularly economically attractive due to thelow number of children that need to be screened to detect oneamblyopia case.

To the best of our knowledge Barry et al.,23 and König et al.,24

and König and Barry25 have performed the only economic evalu-ations of amblyopia screening so far. These studies, however, useddifferent assumptions compared with our model and did not alsoevaluate screening strategies for amblyogenic factors.

The goal of this article was to present a general setting–indepen-dent decision-analytical model that permits determination of thecosts, effectiveness (total number of newly detected true-positiveamblyopias), and cost-effectiveness (costs per newly detected true-positive amblyopia) of four screening strategies to detect amblyo-pia or amblyogenic factors in pre-school children. In this article, westress factors and assumptions that are important to construct themodel. We then use a German setting as an application example.

In general, a decision-analytical model is able to overcome thecomplexity of decision making by structuring the problem clearly.Furthermore, it permits identification of parameters that have astrong influence on results and thus may be candidates for futureresearch.

METHODSModel Overview

Our model on the cost-effectiveness of screening for amblyopiaor amblyogenic factors compares the following strategies:1. Screening of high-risk children up to the age of 1 year

(ophthalmologist).2. Screening of all children up to the age of 1 year

(ophthalmologist).3. Screening of all children aged 3 to 4 years (pediatrician or

general practitioner).4. Screening of children aged 3 to 4 years visiting kindergarten

(orthoptist).“No screening program” should be thought of as a further alter-

native; it may be difficult, however, to assess screening patterns ina natural setting and associated costs. In addition, screening byoffice staff may be considered as an alternative to screening bypediatricians or general practitioners if adequate training can beassumed. The same consideration holds for screening by nonoph-thalmologists as an alternative to screening by ophthalmologists.

Screening up to the age of 1 year is directed to amblyogenicfactors including anisometropia (�1 D), strabismus, and high de-

grees of hyperopia (�3.5 D) because a meaningful vision test atthis age is afflicted by high uncertainty and because amblyopiapossibly has not yet developed. Screening between the ages of 3 and4 years has the aim of detecting amblyopia, defined in our model asa visual acuity of �0.63 decimal units (�20/32 Snellen units).

In the model, pediatricians, general practitioners, and orthop-tists examine children using a visual acuity test, a stereo acuity test,a cover test, and a Hirschberg test. Ophthalmologists, however,perform the following diagnostic procedures at initial visits and forreferral patients: retinoscopy (after the application of cycloplegicdrops), cover test, Hirschberg test, and an examination of theocular media and fundus.

All children with abnormal results from screening performed bya pediatrician, general practitioner, or orthoptist are referred to anophthalmologist. If these children are not confirmed to have am-blyopia, they are not invited for a second visit to an ophthalmolo-gist. Children who are mistakenly diagnosed by ophthalmologistsof having amblyopia (false-positive cases) are identified within 1year after the initial diagnosis and after two additional visits to anophthalmologist. At their first visit to an ophthalmologist after theinitial diagnosis, false-positive cases receive a retinoscopy, a covertest, a Hirschberg test, and an examination of the ocular media andfundus. At their second visit, they have the same examinationsexcept the retinoscopy. Furthermore, they receive a pair of glassesand daily patches (occlusion therapy) for 120 days.

Determining Cost-Effectiveness

Economic evaluations of the four approaches to screening areconducted from a societal viewpoint because this perspective is themost comprehensive, considering all direct and indirect costs, andis also recommended by the U.S. Panel on Cost-Effectiveness inHealth and Medicine.26 Simultaneously, the economic evaluationsalso take the perspective of a third-party payer because the third-party payer covers the costs of the screening programs.

All economic evaluations are performed as cost-effectivenessanalyses. The time horizon is 1 year. The outcome measures arenewly detected true-positive cases of amblyopia requiring treat-ment. Thus, for screening strategies up to the age of 1 year, onlycases that develop an amblyopia at a later point of time are consid-ered. Measures such as life-years saved, quality of life, or quality-adjusted life years were not used because data on the life expectancyand the quality of life of treated and untreated patients are cur-rently unavailable.

The four screening programs are compared using the incremen-tal cost-effectiveness ratio. This ratio is defined as the additionalcost of a specific screening strategy divided by its additional clinicalbenefit compared with the next less-expensive strategy. Screeningstrategies that are less effective and more costly than an alternative(dominated) and strategies with a higher cost-effectiveness ratiothan a more-effective alternative strategy (extended dominance)are ruled out.26

Clinical and Epidemiological Data

Tables 1 to 4 list clinical and epidemiological parameters re-quired for the model (all values provided in tables refer to theapplication example described below). Adjustment of data on the

* In the United States and other countries outside Germany optometrists may beconsidered as an alternative to opthalmologists. In order for the model to accommodatefor optometrists, however, it might need some adjustments.

260 Cost-Effectiveness of Screening for Amblyopia—Gandjour et al.


sensitivity and specificity (accuracy) of the screening proceduresneeds to be considered if there is a lack of data on the accuracy (1)of a combination of the above tests and (2) of different examiners.

In addition, the model requires estimates or data on the propor-tion of patients that have already been identified at the time ofscreening as having amblyopia because of obvious squints and/or apositive family history.

Important clinical, epidemiological, and cost data that are un-available may be determined by a two-round modified Delphipanel.38 The Delphi technique uses an iterative questionnaire tomeasure consensus among individual responses. There is no inter-action between responder and interviewer.

Cost Data

Cost parameters needed to construct the model are provided byTable 5. All evaluations include direct medical and nonmedical(transportation) costs. The societal viewpoint considers, in addi-tion, the productivity losses and transportation costs of caregivers.These incur when caregivers accompany children to and from aphysician’s office. The reason for including these costs is that asocietal viewpoint considers all costs nontrivial in magnitude.

In the model, productivity losses are assessed by the human

capital approach using the average labor cost of an employee as anapproximate measure. For determining the percentage of workingcaregivers missing work, it is assumed that all caregivers are moth-ers. The time cost of nonworking caregivers for accompanyingchildren to and from a physician’s office is estimated by the re-placement cost approach.48

The follow-up costs for false-positive diagnoses are added to thecost per true-positive case detected. Follow-up costs are not dis-counted because they incur during the first year after the initialdiagnosis.

Sensitivity Analysis

A sensitivity analysis examines how results change when input dataare varied. Whereas a univariate sensitivity analysis alters the value ofone parameter at a time, a multivariate sensitivity analysis alters thevalues of two or more parameters simultaneously. Our model uses aMonte Carlo simulation, which is a type of multivariate sensitivityanalysis. This technique runs a large number of simulations by repeat-edly drawing samples from probability distributions of input param-eters. Thus, it derives a probability distribution for the results andforecasts the probability that certain strategies are dominated. Further-more, it helps to identify parameters that have a strong association

TABLE 1.Sensitivity and specificity stratified by age and type of examiner.a

Age(yr)

Population Examiner Sensitivity (Range)Specificity

(Range)

�1 High risk Ophthalmologist 0.95 (0.8–1.0) 0.95 (0.7–1.0)�1 Non-high risk Ophthalmologist 0.95 (0.8–1.0) 0.95 (0.8–1.0)3–4 Total Pediatrician/GPb 0.75 (0.5–0.9) 0.8 (0.7–0.9)3–4 Total Orthoptist 0.9 (0.65–1.0) 0.9 (0.8–1.0)3–4 Referrals from pediatricians/GP’s/orthoptists Ophthalmologist 0.95 (0.85–1.0) 0.95 (0.8–1.0)

a All data were estimated.b GP, general practitioner.

TABLE 2.Rates for participation, referral, and compliance.

ActivityBase-Case Rate

(Range)Reference Source and Comments

High-risk children aged �1 yrParticipation in the recommended pediatric exam U6a 0.91 27 (1996 data)Referral to an ophthalmologist 0.95 (0.9–1.0) EstimateCompliance with referral 0.90 (0.8–1.0) Estimate

All children aged �1 yrParticipation in the recommended ophthalmologic exam 0.65 (0.5–0.85) Estimate combining the rates for participation

in the exam U6, ophthalmologic referral,and compliance with referral

All children aged 3–4 yrParticipation in the screening program U8 0.80 27 (1996 data)Participation of a child in the kindergarten screening

program0.91 (0.85–0.97) 28, 29 (weighted average)

Referral to an ophthalmologist after positive eye exam 0.90 (0.8–1.0) EstimateCompliance with referral 0.90 (0.8–1.0) Estimatea U6, Untersuchung 6, the sixth preventative health examination in a series of 10 preventive examinations carried out by

pediatricians and general practitioners for children from birth to 13 years of age.

Cost-Effectiveness of Screening for Amblyopia—Gandjour et al. 261


with the average cost per case detected. Such parameters can be furtherevaluated in a univariate sensitivity analysis as well as a worst-case andbest-case scenario. In the latter, all parameter values are set for worstcase and best case, respectively.

RESULTS

In this article, we provide an application example of themodel in a German setting. We focus on deviations from theabove model description and the specific requirements of theGerman setting.

In this setting search for visual defects including strabismus andrefractive amblyopia is part of a series of 10 preventive examina-tions carried out by pediatricians and general practitioners for chil-dren from birth to 13 years of age. The program is funded by thestatutory health insurance.

Model Overview

The possibility of having no screening program was not consid-ered as an alternative because of the difficulty of assessing screeningpatterns in a natural setting and their associated costs. We did not

consider screening by office staff as an alternative to screening bypediatricians or general practitioners because the office staff inGermany is not qualified to perform the examinations. For thesame reason, we did not consider the possibility of optometrists asan alternative to ophthalmologists.

In the model, all screenings took place at outpatient offices(separate from hospitals in Germany) except for the screening con-ducted in kindergartens by orthoptists.

Data Sources

We identified sources of data by searching MEDLINE for arti-cles in English and German that were published up to February2000. We handsearched review articles and book chapters for ad-ditional sources. A panel of three ophthalmologists and one phy-sician specializing in health economics estimated any data thatwere unavailable in the literature.

The percentage of time during the U8 examination devoted toamblyopia screening was determined by a two-round modifiedDelphi panel38 because this figure had been assumed to be a majordriver of the cost of pediatric examinations. Twelve pediatricians

TABLE 4.Miscellaneous data.

Parameter Base-Case Value (Range) Reference Source and Comments

Number of children aged 3–4 attending kindergarten 374,000 33 (1998 data)Probability of patients with strabismus developing amblyopia 0.53 (0.42–0.64) 37

TABLE 3.Data for calculating the number of children at high risk.

ParameterBase-Case Value

(Range)Reference Source and Comments

Hereditary risk factors for amblyopiaPrevalence rate of strabismus among children 0.052 (0.047–0.057) Weighted average from Graham,5 Haase et

al,30 and Kendall et al.31

Risk of strabismus if one parent or one siblingis affected

0.176 (0.053–0.3) 32

Number of households with two childrenbelow age 18

3,536,000 33 (1998 data)

Number of households with three childrenbelow age 18

863,000 33 (1998 data)

Number of households with four childrenbelow age 18

249,000 33 (1998 data)

Pre-term children (birthweight �1500 g)Prevalence rate at birth 0.01 34 (1997 data)Probability of anisometropia (�1 D) and/or

strabismus and/or hyperopia (�3.5 D)0.233 (0.183–0.283) Estimated from the prevalence rate of

strabismus among pre-term children35 andof anisometropia and hyperopia among allchildren3, 4

First year mortality rate 0.176 (0.168–0.185) 34, 36 (1997 data)Pre-term children (birthweight 1500–2500 g)

Prevalence rate at birth 0.052 34 (1997 data)Probability of anisometropia and/or strabismus 0.15 (0.11–0.19) EstimateFirst year mortality rate 0.014 (0.013–0.015) 34, 36 (1997 data)



performing a visual acuity test and a stereo acuity test participatedin the first round; eight thereof participated in the second roundwith the opportunity to change their score in view of the group’sresponse.

Data Analysis

We performed the analysis using Microsoft Excel 97 and CrystalBall 4.0 (Decisioneering, Denver, CO), an Excel add-in programthat performs Monte Carlo simulations.

Clinical and Epidemiological Data

Tables 1 to 4 list clinical and epidemiological data used in themodel. Ranges were defined as four standard errors of the mean or, ifunavailable, as reasonable estimates covering the complete distribution.Our estimates on the sensitivity and specificity (accuracy) of the screeningprocedures were based on published data. These data, however, had to beadjusted for our study due to a lack of data on the accuracy (1) of acombination of the above tests and (2) of different examiners.

The size of the child population in different age groups was

TABLE 5.Data for calculating costs.

ItemBase-Case Value

(Range)Reference Source and

Comments

Direct costsTotal point score of one exam at an ophthalmologist (items 1,

1200, 1202, 1216, 1242)456 39

Total point score of a complete U8 exam (pediatrician or GP)a 650 40Point value €0.0422b 41Percentage of time of the U8 exam devoted to amblyopia

screening0.25 (0.18–0.29) Modified Delphi panel

Cost of one glasses lens (to the statutory health insurance) €12.02 42Cost of one glasses lens (to society) €12.02 (10.00–12.02) The statutory health insurance

was assumed to reimburse thewholesale price.

Cost of a glasses frame €25.00 (12.50–37.50) Estimate of the wholesale priceDaily cost of an eye patch €0.50 (0.40–0.60) Estimate of the wholesale priceAverage annual part-time salary of an orthoptist including

social security payment of the employer (according to classVa of the German Federal Employee Tariffs)

€15 727c 43

Number of kindergartens visited by an orthoptist each day 2 EstimateTime required to organize an orthoptist’s visit to a kindergarten 30 min (20–40) Estimate; organization was

assumed to take place on aseparate day

Time required to prepare the examination in a kindergarten 20 min (10–30) EstimateDuration of an orthoptist’s exam in a kindergarten 8 min (5–12) 24; Range estimateDaily driving time of orthoptist 45 min (30–60) EstimateDaily driving distance of orthoptist 45 km (30–60) EstimateCost per kilometer of orthoptist €0.25 44

Indirect costsDriving distance of caregiver to the physician office 10 km (5–15) EstimateCost per kilometer of caregiver €0.25 44Average annual full-time labor cost of an employee €42 236d 45Average annual labor cost of the civil service €8743 46Percentage of working mothers 0.47 47Percentage of working mothers who work part-time �20h/wk 0.41 47Percentage of working mothers who work 21–35h/wk 0.16 47Percentage of working mothers who work �35h/wk 0.42 47Percentage of mothers working �20h/wk to miss work 0.10 (0.0–0.2) EstimatePercentage of mothers working 21–35h/wk to miss work 0.5 (0.3–0.7) EstimatePercentage of mothers working �35h/wk to miss work 0.8 (0.6–0.9) EstimateProportion of a full-time working day spent for a child’s health

care visit0.25 (0.1–0.3) Estimate

a U8, Untersuchung 8; GP, general practitioner.b Average point value for preventive services of the statutory health insurance in Northrhine Westfalia.c Salaries in the newly-formed German states and the old West German states are weighted according to the number of children at

the age of 3–4 years.d The 1996 figure is multiplied by the average increase in the gross income of employees in the industrial and service sector in 1997

and 1998 (3.6%).



taken from the Federal Office of Statistics in Germany (1997 da-ta).33 The prevalence rate of amblyogenic risk factors was esti-mated to be 10% (range, 7 to 13) considering the prevalence rate ofeach risk factor3–5 as well as the joint occurrence. The prevalencerate of amblyopia was calculated as the weighted average of threestudies defining amblyopia as a level of visual acuity �0.63 (3.1%;95% confidence interval, 0.028 to 0.034).49–51 This is a ratherconservative estimate because the prevalence study by Attebo etal.,51 which was performed in adults, did not include cases that hadbeen treated successfully. Weights were obtained by dividing thesample size of one study by the total sample size of the three studies.

We assumed that at the time of screening a proportion of pa-tients had already been identified as having amblyopia because ofobvious squints and/or a positive family history. The percentagesof amblyopia cases identified before screening at age under 1 yearfor high-risk children, age less than one year 1 for non-high-riskpatients, and age 3 to 4 for all children were estimated to be 0.5(range, 0.3 to 0.6), 0.25 (range, 0.1 to 0.35), and 0.6 (range, 0.4 to0.7), respectively. The estimates present rather high values; thus,estimates on the cost-effectiveness of screening strategies are rela-tively conservative.

Cost Data

Cost data were calculated based on the data in Table 5. All costswere reported in 1999 Euro (1 Euro � 0.91 US$). To calculatedirect costs from the societal viewpoint, reimbursement fees wereused assuming that they are equal to opportunity costs. To calcu-late the cost of an ophthalmologist’s examination, we used theprivate health insurance fee-for-service items list (Gebührenord-nung für Ärzte), which contains a more detailed listing of therelevant items than the statutory health insurance’s price scale(Einheitlicher Bewertungsma�stab), and, thus, enables a more ac-curate portrayal of the costs of ophthalmologic examinations. Weincluded items 1, 1200, 1202, 1216, and 1242 from the privatehealth insurance fee-for-service list, which add up to 456 points.39

In the sensitivity analysis, we applied a total point score of 890resulting from items 1, 1216, 1219, and 1242 of the statutoryhealth insurance’s price scale.40 In both cases, we used the averagepoint value for preventive services from all statutory health insurersin the state of Northrhine Westfalia (€0.0422).41

We used the labor cost of civil service as the market value of thetime cost of nonworking caregivers for accompanying children toand from a physician’s office. In calculating the cost of amblyopia

screening as part of the preventative health examination U8 doneby a general practitioner or pediatrician, we did not consider pro-ductivity losses and time costs because they were not influenced bythe absence or presence of amblyopia screening.

Sensitivity Analysis

A Monte Carlo simulation with 10,000 iterations was per-formed. For this purpose, the parameters listed in Tables 1 to 5were simultaneously varied. Ranges were entered as four standarderrors of the mean or, if unavailable, as reasonable estimates cov-ering the complete distribution. Distributions were assumed to beeither normal or lognormal. A multiple regression analysis acrosssimulation iteration results determined the association betweenmodel parameters and the average cost per case detected. The fourparameters with the strongest association plus the higher cost esti-mate for ophthalmologists’ examinations were further evaluated ina univariate sensitivity analysis as well as a worst-case and best-casescenario.

Results of the Application Example

The two-round modified Delphi panel suggested that the me-dian percentage of the U8 examination devoted to amblyopiascreening was 25 (95% percentile range, 18 to 29). The medianpercentage of the four pediatricians who participated only in thefirst round was 15 (95% confidence interval, 5 to 39).

Table 6 shows the cost per screening examination depending onperspective (societal or statutory health insurance) and the type ofexaminer. The costs of the first follow-up visit to an ophthalmol-ogist for false-positive diagnoses were identical to those shown inthe table. The direct medical costs of the second follow-up exam-ination were €12.82.

Table 7 shows the clinical and economic outcomes of screeningstrategies from the societal perspective. The screening of high-riskchildren up to the age of 1 year had the lowest average cost per casedetected. The total number of cases detected through orthopticscreening was much lower than the number for general practitio-ners or pediatricians mainly because less than half of all children atthe age of 3 to 4 visit kindergarten in Germany.

The probability that screening by orthoptists was less econom-ically attractive than or dominated by screening of high-risk chil-dren was 62%. The probability that screening by general practitio-ners or pediatricians was less economically attractive than or

TABLE 6.Cost items of one screening exam.a

ExaminerDirect

MedicalCost

Travel Costof Examiner

Total DirectCostb

Travel Cost ofCaregiver

Loss ofCaregiver’sProductivity

Time Cost ofCaregiver

Total Direct andIndirect Costc

Ophthalmologist 19.23 — 19.23 2.45 10.35 7.79 39.83Pediatrician/GPd 6.85 — 6.85 — — — 6.85Orthoptist 5.55 0.34 5.90 — — — 5.90

a All estimates are in 1999 EUR.b Equivalent to costs from the viewpoint of the statutory health insurance.c Equivalent to costs from the viewpoint of society.d GP, general practitioner.



dominated by screening of all children up to the age of 1 year was12%. The total detection rates (including prescreening cases iden-tified) for screening at age less than 1 year for high-risk children,age less than 1 year for non-high-risk patients, age 3 to 4 years forpediatrician or general practitioner examination, and age 3 to 4years for orthoptist examination were 24.4%, 75.7%, 78.2%, and71.6%, respectively.

Table 8 shows the clinical and economic outcomes of screeningstrategies from the perspective of the statutory health insurance.Again, the screening of high-risk children up to the age of 1 yearhad the lowest average cost per case detected. The screening ofchildren aged 3 to 4 years by orthoptists (general practitioners andpediatricians) was dominated by screening by ophthalmologists ofhigh-risk children (all children) up to the age of 1 year.

Average costs from the societal perspective were higher thanthose from the perspective of the statutory health insurance (€643per case detected if the mean cost difference of all four strategies

was calculated). Table 9 shows the annual loss of working days tocaregivers employed outside the home and of time to caregivers notemployed outside the home.

The four parameters with the strongest association with theaverage cost per case detected were prevalence rate of amblyopia,prescreening detection rate, specificity, and risk of strabismus ifone parent or one sibling is affected. These parameters explainedbetween 48% and 94% of the variation in the average cost per casedetected.

Tables 10 to 12 show the results of the univariate and the mul-tivariate sensitivity analysis. The number in the first row and col-umn of Table 10 says that it costs €1203 per newly detected true-positive case of amblyopia (compared with no screening) if wescreen high-risk children up to the age of 1 year while assuming ahigh value for the prevalence rate of amblyopia (3.4%). In Table12, the number in the first row and column is read as follows: itcosts €475 per newly detected true-positive case of amblyopia

TABLE 7.Mean � SD of clinical and economic outcomes of screening strategies (societal perspective).a

Intervention Total Cost, €Total

No. of CasesAverage Cost

(€)/CaseIncrease in

Cost, €

Increase inNo. of Cases

Increase in Cost(€)/Increase inNo. of Cases

Screening of high-riskchildren aged �1 yr

4,556,426 � 827,933 3,682 � 1,161 1238 � 797 — — —

Screening of all childrenaged �1 yr

25,391,226 � 3,362,855 10,694 � 994 2374 � 371 16,164,039 6287 2571

Screening of all childrenaged 3–4 yr(GP/pediatrician)b

9,227,188 � 1,214,275 4,406 � 777 2094 � 370 4,670,762 725 6445

Screening of all childrenaged 3–4 yr(orthoptist)

3,910,210 � 688,914 2,809 � 420 1392 � 278 — — Dominated

a Standard deviations were obtained through a Monte Carlo simulation. They are not shown for incremental values because thesimulation did not account for the different causes of positive or negative incremental values which would have distorted standarddeviations.52 Calculations were made under consideration of decimal places (not shown).

b GP, general practitioner.

TABLE 8.Mean � SD of clinical and economic outcomes of screening strategies (perspective of the statutory health insurance).a

Intervention Total Cost, €Total

No. of CasesAverage Cost

(€)/CaseIncrease in

Cost, €

Increase inNo. of Cases

Increase in Cost(€)/Increase inNo. of Cases

Screening of high-riskchildren aged �1 yr

2,053,781 � 463,448 3258 � 1027 630 � 496 — — —

Screening of allchildren aged �1 yr

11,457,113 � 1,749,289 9464 � 879 1211 � 220 9,403,333 6206 1515

Screening of allchildren aged 3–4 yr(GP/pediatrician)b

5,996,990 � 734,290 3900 � 688 1538 � 279 — — Dominated

Screening of allchildren aged 3–4 yr(orthoptist)

2,852,873 � 502,807 2486 � 372 1147 � 234 — — Dominated

a Standard deviations were obtained through a Monte Carlo simulation. They are not shown for incremental values because thesimulation did not account for the different causes of positive or negative incremental values which would have distorted standarddeviations.52 Calculations were made under consideration of decimal places (not shown).

b GP, general practitioner.



(compared with no screening) if we screen high-risk children up tothe age of 1 year while assuming the most favorable estimates forprevalence rate of amblyopia, prescreening detection rate, specific-ity, and risk of strabismus if one parent or one sibling is affected.

In the sensitivity analysis, screening of all children up to the ageof 1 year was the only strategy not dominated by others. Screeninghigh-risk children was dominated in the worst-case scenario and ifthe lower bound of the 95% confidence interval of the risk ofstrabismus if one parent or one sibling is affected was used. Screen-ing by general practitioners or pediatricians and orthoptists wasdominated using most of the assumptions.

DISCUSSION

This article presents a decision-analytical model to evaluate thecost-effectiveness of screening strategies for amblyopia and its risk

factors in different settings or countries. One of the major advantagesof modeling is the identification of parameters with a strong influenceon results that may be candidates for future research. When applyingthe model to Germany, the four parameters with the strongest associ-ation with the average cost per case detected were prevalence rate ofamblyopia, prescreening detection rate, specificity, and risk of strabis-mus if one parent or one sibling is affected.

The model suggests that screening all children in Germany up tothe age of 1 year by ophthalmologists is the most economicallyattractive strategy to detect amblyopia or amblyogenic factors. Thisconclusion considers the various assumptions tested in the sensi-tivity analysis. Screening high-risk children up to the age of 1 yearshowed lower average costs per case detected than screening allchildren up to the age of 1 year in the base-case analysis but becamedominated if a low (5.3%) probability of familial clustering ofstrabismus was assumed.

TABLE 10.Results of the univariate sensitivity analysis (societal perspective).a

Strategy

PrevalenceRate of

Amblyopia/AmblyogenicRisk Factors

SpecificityPrescreening

Detection Rates

Probabilityof Familial

Clustering ofStrabismus

Cost of anOphthalmologist’s

Exam

High valuesScreening of high-risk

children aged �1 yr1,203b 1312c 1,523b 769b 2242c


2,320c 2318c 2,794c 4620c 3913c

Screening of all childrenaged 3–4 yr(GP/pediatrician)d

4,500c 3579c 13,413c Dominated 6689c

Screening of all childrenaged 3–4 yr (orthoptist)

Dominated 936b Dominated Dominated 1599b

Low valuesScreening of high-risk

children aged �1 yr1,274b 2301b 911b Dominated NA


2,882c 3918c 2,372c 2394c NA

Screening of all childrenaged 3–4 yr(GP/pediatrician)

11,376c 9946c 3,107c 3329c NA

Screening of all childrenaged 3–4 yr (orthoptist)

Dominated Dominated Dominated 1392b NA

a Values indicate the cost (€) per newly detected true-positive case of amblyopia.b Average cost-effectiveness ratio (compared to no screening).c Incremental cost-effectiveness ratio (compared with the next least-expensive strategy).d GP, general practitioner; NA, not applicable.

TABLE 9.Loss of working days and time according to the type of strategy.

Strategy Annual Loss of Working DaysAnnual Loss of Time ofNonworking Caregivers

(in Working Days)

Screening of high-risk children aged �1 yr 5,572 20,258Screening of all children aged �1 yr 30,984 112,641Screening of all children aged 3–4 yr (GP/pediatrician) 39,301 142,878Screening of all children aged 3–4 yr (orthoptist) 19,355 70,364



In contrast, screening all children when they are between 3 and4 years of age for amblyopia is less economically attractive bothfrom the perspective of the statutory health insurance and from theperspective of society. The main reason for inefficiency is the highpercentage of amblyopia cases already identified before screeningleading to a high number of children that need to be screened todetect one amblyopia case. However, using cost-effectiveness as thekey criterion, the ultimate choice of a strategy depends on thewillingness of decision makers or insured persons to pay for anadditional case detected.

Apart from cost-effectiveness, effectiveness is an important cri-terion for evaluating screening strategies. Effectiveness was mea-sured by the percentage of all true-positive cases detected includingcases identified before screening. In the model, the most effectivestrategies were screening of all children up to the age of 1 year andat the age of 3 to 4 years with detection rates between 72% and

78%. Thus, all strategies left a significant portion of childrenundetected.

In our model only 31% of true-positive cases with amblyogenicfactors were true-positive cases of amblyopia as well. Thus, 69% oftrue-positive cases with amblyogenic factors were unnecessarilytreated for preventing amblyopia because they would not havemanifested amblyopia. On the other hand, we believe that mostchildren with amblyogenic factors require glasses for correctinganisometropia and/or hyperopia. Moreover, those 31% whowould have manifested an amblyopia may benefit from a reducedduration of treatment.22 The net effect (lower quality of life andadditional costs for children receiving unnecessary treatment vs.the increased quality of life of true positive amblyopia cases whobenefit from a reduced duration of treatment) cannot be deter-mined due to a lack of data.

The results of this study need to be interpreted with caution.

TABLE 11.Results of the univariate sensitivity analysis (perspective of the statutory health insurance).a

Strategy

PrevalenceRate of

Amblyopia/AmblyogenicRisk Factors

SpecificityPrescreening

Detection Rates

Probabilityof Familial

Clustering ofStrabismus

Cost of anOphthalmologist’s

Exam

High valuesScreening of high-risk children aged �1 yr 613b 495b 776b 391b 1144b

Screening of all children aged �1 yr 1327c 1182c 1736c 2358c 2324c

Screening of all children aged 3–4 yr(GP/pediatrician)d

Dominated Dominated Dominated Dominated 6392c

Screening of all children aged 3–4 yr(orthoptist)

Dominated Dominated Dominated Dominated Dominated

Low valuesScreening of high-risk children aged �1 yr 649b 1307b 463b Dominated NAScreening of all children aged �1 yr 1766c 2515c 1276c 1165c NAScreening of all children aged 3–4 yr

(GP/pediatrician)Dominated Dominated Dominated Dominated NA

Screening of all children aged 3–4 yrs(orthoptist)

Dominated Dominated Dominated 1147b NA

a Values indicate the cost (€) per newly detected true-positive case of amblyopia.b Average cost-effectiveness ratio (compared to no screening).c Incremental cost-effectiveness ratio (compared with the next least-expensive strategy).d GP, general practitioner; NA, not applicable.

TABLE 12.Results of the multivariate sensitivity analysis.a

InterventionBest Case

(Societal Perspective)

Best Case(Statutory Health

Insurance)

Worst Case(Societal Perspective)

Worst Case(Statutory Health

Insurance)

Screening of high-risk children aged �1 yr 475b 229b Dominated DominatedScreening of all children aged �1 yr 2636c 1273c 5010c 2630c

Screening of all children aged 3–4 yr(GP/pediatrician)d

Dominated Dominated Dominated Dominated

Screening of all children aged 3–4 yr (orthoptist) Dominated Dominated 3468b 2512b

a Values indicate the cost (€) per newly detected true-positive case of amblyopia.b Average cost-effectiveness ratio (compared to no screening).c Incremental cost-effectiveness ratio (compared with the next least-expensive strategy).d GP, general practitioner.



First, some parameters used in the model had to be estimated.Sensitivity analysis was used to modify the assumptions. In thesensitivity analysis the relationships among strategies in terms ofdominance were fairly robust, although some uncertainty remains,particularly with regard to the size of the cost-effectiveness ratios.

Second, the cost-effectiveness analysis did not take into accountthe number of amblyopia cases detected before screening and thecosts of identifying them. The difficulty here was to assess screen-ing patterns in a natural setting, for example the likelihood ofrepeat or inadequate examinations (with inadequate examinationsresulting in a high number of false-positive cases). If the costs perdetected amblyopia case were higher for unsystematic screeningthan for systematic screening, screening at 3 to 4 years of age wouldbecome even less economically attractive.

Third, the estimate on the number of cases detected throughscreening high-risk children up to the age of 1 year is very conser-vative because non-high-risk children identified through unsys-tematic screening beyond the age of 1 year were not taken intoaccount.

Fourth, we assumed ideal conditions in the base case for theestimates of the examiners’ test statistics. We did this because webelieved that it was ethically justified to demand an optimal level ofmotivation from the examiners. On the other hand, reducing thespecificity of diagnostic procedures to the low end of the range didnot change our main conclusion.

Fifth, like many other cost-effectiveness analyses on screeningand diagnostic tests, this study used an intermediate outcome, thenumber of cases detected. Including information on the costs andbenefits of treatment would provide a more complete picture of thecost-effectiveness of screening for amblyopia and its risk factors.However, data on important treatment outcomes such as the qual-ity of life are lacking.

Given these limitations of the study, we regard the conclusionsfrom our model as preliminary. Combining the results of the cost-effectiveness and the pure effectiveness analysis, screening all chil-dren up to the age of 1 year by opthalmologists is the preferredstrategy. However, the denominator chosen for this analysis doesnot permit the deduction of whether this strategy is economicallyattractive compared with other interventions in the health caresystem. Furthermore, the transferability of the conclusions of thisstudy to situations in other countries is limited, for example, bydifferences in costs, clinical management, epidemiology, or de-mography as well as in the diagnostic accuracy and organization ofhealth care providers. In the U.S., for example, optometrists arewell prepared to perform screening for amblyogenic factors, fol-low-up investigations, and treatment. In Germany, optometristslack the necessary skills and, thus, were not considered in theapplication example.

To be able to make final recommendations regarding a favorablescreening model, additional research is needed in several areas. Inparticular, we recommend (1) studies on the sensitivity and spec-ificity of combined screening tests stratified for the age of the childand the type of examiner, (2) prevalence studies on amblyogenicfactors and amblyopia, and (3) further investigation on the effec-tiveness of treatment for amblyopia including the impact of ther-apy on the quality of life.

ACKNOWLEDGMENTS

Parts of this article were presented at the joint meeting of the BielschowskySociety for Strabismus Research and the Association of German Orthoptistsheld November 19–21, 1999, in Cologne, Germany. The authors would liketo thank Noelle Aplevich, M.A., for her very valuable comments on a prelim-inary draft of this article.

Received June 29, 2002; revision received November 12, 2002.

REFERENCES

1. Burian HM, von Noorden GK. Burian-von Noorden’s BinocularVision and Ocular Motility: Theory and Management of Strabismus,2nd ed. St. Louis: Mosby, 1980.

2. Ciuffreda KJ, Levi DM, Selenow A. Amblyopia: Basic and ClinicalAspects. Boston: Butterworth-Heinemann, 1991.

3. Fern KD, Manny RE, Garza R. Screening for anisometropia in pre-school children. Optom Vis Sci 1998;75:407–23.

4. Atkinson J, Braddick O, Robier B, Anker S, Ehrlich D, King J,Watson P, Moore A. Two infant vision screening programmes: pre-diction and prevention of strabismus and amblyopia from photo- andvideorefractive screening. Eye 1996;10:189–98.

5. Graham PA. Epidemiology of strabismus. Br J Ophthalmol 1974;58:224–31.

6. U.S. Preventive Services Task Force. Guide to Clinical PreventiveServices: an Assessment of the Effectiveness of 169 Interventions.Baltimore: Williams & Wilkins, 1989.

7. American Academy of Ophthalmology. Preferred Practice Pattern:Comprehensive Pediatric Eye Evaluation. San Francisco: AmericanAcademy of Ophthalmology, 1992.

8. American Academy of Family Physicians. Age Charts for PeriodicHealth Examination. Reprint no. 510. Kansas City: American Acad-emy of Family Physicians,1994.

9. American Optometric Association. Recommendations for RegularOptometric Care. St. Louis: American Optometric Association,1994.

10. Green M. Bright Futures: Guidelines for Health Supervision of In-fants, Children and Adolescents. Arlington, VA: National Center forEducation in Maternal and Child Health, 1994.

11. Canadian Task Force on the Periodic Health Examination. CanadianGuide to Clinical Preventive Care. Ottawa: Canada CommunicationGroup, 1994.

12. Haase W. Amblyopie. In: Kaufmann H, de Decker W, eds. Strabis-mus, 2nd ed. Stuttgart: Enke-Verlag, 1995.

13. American Academy of Pediatrics Committee on Practice and Ambu-latory Medicine, Section on Ophthalmology. Eye examination andvision screening in infants, children, and young adults. Pediatrics1996;98:153–7.

14. Hartmann EE, Dobson V, Hainline L, Marsh-Tootle W, Quinn GE,Ruttum MS, Schmidt PP, Simons K. Preschool vision screening:summary of a task force report on behalf of the Maternal and ChildHealth Bureau and the National Eye Institute Task Force on VisionScreening in the Preschool Child. Pediatrics 2000;106:1105–16.

15. Williams C, Northstone K, Harrad RA, Sparrow JM, Harvey I. Am-blyopia treatment outcomes after screening before or at age 3 years:follow up from randomised trial. BMJ 2002;324:1549.

16. Stewart CE, Fielder AR, Moseley MJ, Stephens DA, MOTAS Coop-erative. Dose-response relationship of occlusion therapy foramblyopia: initial results from the MOTAS study (Abstract 2939).Association for Research in Vision and Ophthalmology, 2002.

17. Bolger PG, Stewart-Brown SL, Newcombe E, Starbuck A. Visionscreening in preschool children: comparison of orthoptists and clini-cal medical officers as primary screeners. BMJ 1991;303:1291–4.



18. Jarvis SN, Tamhne RC, Thompson L, Francis PM, Anderson J,Colver AF. Preschool vision screening. Arch Dis Child 1991;66:288–94.

19. Milne C. An evaluation of cases referred to hospital by the Newcastlepreschool orthoptic service. Br Orthop J 1994;51:1–5.

20. Williamson TH, Andrews R, Dutton GN, Murray G, Graham N.Assessment of an inner city visual screening programme for preschoolchildren. Br J Ophthalmol 1995;79:1068–73.

21. Newman DK, Hitchcock A, McCarthy H, Keast-Butler J, MooreAT. Preschool vision screening: outcome of children referred to thehospital eye service. Br J Ophthalmol 1996;80:1077–82.

22. Hubel DH. Auge und Gehirn: Neurobiologie des Sehens.Heidelberg: Spektrum der Wissenschaft-Verlagsgesellschaft, 1989.

23. Barry JC, Hartmann A, Pongs UM, Jockel M. Model for cost-benefitrelations of amblyopia screening. Ophthalmologe 1998;95:19–27.

24. König HH, Barry JC, Leidl R, Zrenner E. Cost effectiveness of massorthoptic screening in kindergarten for early detection of develop-mental vision disorders. Gesundheitswesen 2000;62:196–206.

25. König HH, Barry JC. Economic evaluation of different methods ofscreening for amblyopia in kindergarten. Pediatrics 2002;109:E59.

26. Gold MR. Cost-effectiveness in Health and Medicine. New York:Oxford University Press, 1996.

27. AOK Bundesverband. Statistische Informationen. Bonn: 1998.28. Russmann W, König U, Schlimbach K, Pawlowska-Seyda D, Wir-

batz B. Refractive errors, strabismus and amblyopia in pre-schoolscreening: experiences using a vision test in kindergarten. OffentlGesundheitswes 1990;52:77–84.

29. Kasmann-Kellner B, Heine M, Pfau B, Singer A, Ruprecht KW.Screening for amblyopia, strabismus and refractive abnormalities in1,030 kindergarten children. Klin Monatsbl Augenheilkd 1998;213:166–73.

30. Haase W, Muhlig HP. The incidence of squinting in school begin-ners in Hamburg (author’s translation). Klin Monatsbl Augenheilkd1979;174:232–5.

31. Kendall JA, Stayte MA. Ocular defects in children from birth to 6years of age. Br Orthop J 1989;46:3.

32. Aurell E, Norrsell K. A longitudinal study of children with a familyhistory of strabismus: factors determining the incidence of strabis-mus. Br J Ophthalmol 1990;74:589–94.

33. Statistisches Bundesamt. Statistisches Jahrbuch 1999. Stuttgart: Met-zler-Poeschel, 1999.

34. Statistisches Bundesamt. Fachserie 1, R1. Stuttgart: Metzler-Poeschel, 1997.

35. McGinnity FG, Bryars JH. Controlled study of ocular morbidity inschool children born preterm. Br J Ophthalmol 1992;76:520–4.

36. Statistisches Bundesamt. Sondertabelle. Stuttgart: Metzler-Poeschel,1997.

37. Abrahamsson M, Fabian G, Sjostrand J. Refraction changes in chil-

dren developing convergent or divergent strabismus. Br J Ophthal-mol 1992;76:723–7.

38. Evans C, Crawford B. Expert judgement in pharmacoeconomicstudies: guidance and future use. Pharmacoeconomics 2000;17:545–53.

39. Hess R. Gebührenordnung für Ärzte (GOÄ). Köln: Deutscher ÄrzteVerlag, 1996.

40. Kassenärztliche Bundesvereinigung. Einheitlicher Bewertungsma�stab(EBM). Köln: Deutscher Ärzte Verlag, 1999.

41. Kassenärztliche Vereinigung Nordrhein (1999) HonorarabrechnungKV Nordrhein 1/2000. Available at: http://www.kvno.de/. Accessedon September 18, 2000.

42. Bundesministerium der Justiz. Bekanntmachung über die Festset-zung von Festbeträgen für Sehhilfen gemä� § 36 Abs. 2; fünftes BuchSozialgesetzbuch (SGB V) für Nordrhein-Westfalen. Bundesanzeiger1997;135:9170–1.

43. Deutsche Forschungsgemeinschaft (1999). Merkblatt Gerhard Hess-Programm. Available at: http://www.dfg.de/. Accessed on September18, 2000.

44. Landesregierung Nordrhein-Westfalen. Sammlung des bereinigtenGesetz- und Verordnungsblattes für das Land Nordrhein-Westfalen(Stand 1.1999). Düsseldorf: 1999.

45. Statistisches Bundesamt. Wirtschaft und Statistik. Stuttgart: Metzler-Poeschel, 1999;2:121.

46. Bundesamt für den Zivildienst. Leitfaden für die Durchführung desZivildienstes. Köln: Bundesamt für den Zivildienst; 1999.

47. Statistisches Bundesamt. Fachserie1, R3. Stuttgart: Metzler-Poeschel; 1998.

48. Drummond MF, O’Brien BJ, Stoddart GL, Torrance GW. Methodsfor the Economic Evaluation of Health Care Programmes, 2nd ed.Oxford: Oxford University Press, 1997.

49. McNeill NL. Patterns of visual defects in children. Br J Ophthalmol1955;39:688–701.

50. Flom MC, Neumaier RW. Prevalence of amblyopia. Am J OptomArch Am Acad Optom 1966;43:732–51.

51. Attebo K, Mitchell P, Cumming R, Smith W, Jolly N, Sparkes R.Prevalence and causes of amblyopia in an adult population. Ophthal-mology 1998;105:154–9.

52. Briggs A. Handling uncertainty in economic evaluation and present-ing the results. In: Drummond MF, McGuire A, eds. EconomicEvaluation in Health Care: Merging Theory with Practice. Oxford:Oxford University Press, 2001:172–214.

Dr. Afschin GandjourInstitut für Gesundheitsökonomie und Klinische Epidemiologie

Universität zu KölnGleueler Stra�e 176-178

50935 Köln, Germanye-mail: [email protected]



A cost effectiveness_model_of_screening_strategies.16

Health & Medicine

Transcript of A cost effectiveness_model_of_screening_strategies.16