Adoption of Pharmaceutical Innovation and the Growth of Drug Expenditure

download Adoption of Pharmaceutical Innovation and the Growth of Drug Expenditure

of 11

Transcript of Adoption of Pharmaceutical Innovation and the Growth of Drug Expenditure

  • 8/3/2019 Adoption of Pharmaceutical Innovation and the Growth of Drug Expenditure

    1/11

    Adoption of Pharmaceutical Innovation and the Growth ofDrug Expenditure in Taiwan: Is It Cost Effective?

    Chee-Ruey Hsieh, PhD,1 Frank A. Sloan, PhD2

    1The Institute of Economics,Academia Sinica,Taipei,Taiwan; 2Duke University, Durham, NC, USA

    A B S T R A C T

    Objectives: To investigate the impact of adopting pharma-

    ceutical innovations on the growth of pharmaceutical expen-

    ditures, focusing specifically on Taiwans experience.

    Methods: We first provide a descriptive analysis of cost

    impacts of introducing new drugs into Taiwans national

    formulary using data from Taiwan. We then use a statistical

    method to decompose the growth of pharmaceutical expen-

    ditures during 19972001 into three components: 1) treat-

    ment expansion; 2) treatment substitution; and 3) priceeffect. By incorporating the estimated benefit from prior

    studies, we calculate the incremental cost-effectiveness ratio

    for new drugs as a whole.

    Results: We find that from 1997 to 2001 public expenditures

    on pharmaceuticals grew 57%. The primary drivers of this

    expenditure growth were treatment expansion and treatment

    substitution. Prices declined by 18%. Cost per life-year

    gained resulting from introduction of new drugs was

    US$1053 (in 2003 dollars) from the perspective of the public

    payer and US$1824 from the perspective of society as a

    whole.

    Conclusions: Overall, our analysis provides evidence with

    previous studies that the drug reimbursement price is not the

    primary driver of increased spending. Rather the introduc-

    tion of new drugs into the formulary leading to expansion of

    treatment, expansion and substitution of the new drugs forexisting drugs may increase spending. Although the adoption

    of pharmaceutical innovation is costly, the estimated benefit

    of adopting pharmaceutical innovation generally far exceeds

    the cost, indicating that the adoption of pharmaceutical inno-

    vation is on the whole worthwhile.

    Keywords: drug expenditure, pharmaceutical innovation,

    treatment expansion, treatment substitution.

    Introduction

    Technological change in medicine has been a majorcause of rising health-care expenditures in manycountries [1]. A substantial amount of technologicalprogress in medicine has taken the form of pharma-ceutical innovation. Consequently, both spending onprescription drugs and the share of drug expenditure intotal health-care expenditures have increased rapidlyin recent years [24]. Around the world, governmentshave used a number of regulatory mechanisms tocontrol spending on pharmaceuticals [4]. Nevertheless,there is a growing body of empirical evidence thattechnological advances in the form of new prescrip-tion drugs have made a substantial contribution toincreased longevity and improved quality of life [57].Thus, there is an inherent conflict between the regula-tory goal of controlling health-care budgets andimproving population health.

    The purpose of this study is to use the experienceof Taiwan as an example to investigate the impact of

    adopting pharmaceutical innovations on the growth of

    pharmaceutical expenditures and whether pharmaceu-tical innovation is worth the increased cost. There areseveral advantages of using Taiwanese data to quantifythe impact of pharmaceutical innovation on health-care costs. First, Taiwan has a social insurance systemproviding universal insurance coverage. The NationalHealth Insurance (NHI) plan offers comprehensivebenefits, including physician services, hospital care,and prescription drugs. To control the cost of publicinsurance, the government in Taiwan regulates theprice paid by the single health insurance plan for indi-vidual drugs. The single-payer system allows policy-makers as well as researchers to trace the impact of

    introducing new drugs on national health costs.Second, under a system of NHI, Taiwan has estab-

    lished a national formulary (positive list), whichincludes all pharmaceuticals subject to reimbursementby NHI. Because the government imposes direct pricecontrols on pharmaceuticals by fixing the pricesproduct by product, the numbers of drugs included inthe formularies are extraordinary large. There aremore than 21,000 drugs included in the NHI formu-laries in Taiwan. The detailed list of drug formularyallows the researcher to decompose the source ofexpenditure growth on prescription drugs.

    Address correspondence to: Chee-Ruey Hsieh, The Institute ofEconomics, Academia Sinica, 128, Section 2, Academia Road,Taipei 115, Taiwan. E-mail: [email protected]

    10.1111/j.1524-4733.2007.00235.x

    Volume 11 Number 2 2008

    V A L U E I N H E A L T H

    334 2008, International Society for Pharmacoeconomics and Outcomes Research (ISPOR) 1098-3015/08/334 334344

    mailto:[email protected]:[email protected]
  • 8/3/2019 Adoption of Pharmaceutical Innovation and the Growth of Drug Expenditure

    2/11

    Third, Taiwan spends about one quarter of health-care expenditures on pharmaceuticals. As comparedto other developed countries, the higher spendingmakes prescription drugs be more likely to becomethe target of cost containment. As a result, the expe-rience of Taiwan provides a valuable insight on

    understanding how policymakers struggle with theconflict between short-run cost impact and long-runhealth benefit.

    The focal points of this article are threefold. We firstbriefly discuss institutional details of Taiwans phar-maceutical policies and describe the cost implicationsof introducing new prescription drugs. We thendecompose the growth of pharmaceutical expendituresduring 19972001 into three components. The first isthe growth of expenditures on prescription drugsattributable to an increase in quantity of prescribeddrugsthe treatment expansion effect. Second, tech-nological change may lead to shifts in drug use within

    specific therapeutic category, especially to more expen-sive productsthe treatment substitution effect. Thethird is the effect of price changes on pharmaceuticalexpenditurepure price effect.

    By incorporating the estimated benefit from theexisting study into our cost measurement, we alsocalculate the incremental cost-effectiveness ratio fornew drugs as a whole. Specifically, we compare spend-ing on new drugs per capita with average annualincrease in life expectancy attributable to new druglaunches.

    Institutional Background

    Characteristics of Pharmaceutical Market

    There are two distinguishing characteristics of health-care system that shape the structure of pharmaceuticalmarket in Taiwan. First, physicians dispense the drugsthat they prescribe. Thus, they are in a position toprofit directly from the sales of prescription drugs. Thegovernment regulates the reimbursement price, i.e.,the price paid on behalf of patients, but do not regulatethe acquisition prices that hospitals or physicians prac-ticing in clinics purchase drugs from the pharmaceuti-cal manufacturers. Thus, hospitals or clinic physicians

    receive the profit margin between the reimbursementand the acquisition prices.

    Second, the medical staff is employed by the hospi-tal. Also, patients are free to choose their own provid-ers; there is no gatekeeper mechanism. Under thissystem, except for emergencies, the main source ofinpatients is the outpatient department at the samehospital. Thus, hospitals in Taiwan have a strongincentive to operate large outpatient departments toincrease their inpatient flows. Larger hospitals oftenpossess substantial bargaining power with pharmaceu-tical manufacturers and can obtain drugs at a lower

    acquisition price. In general, the profit margin betweenthe reimbursement price and the acquisition price rep-resents a major source of hospital revenue.

    These two characteristics have had several impor-tant consequences in the pharmaceutical markets inTaiwan. First, the existence of profit margins for

    pharmaceuticals has led to distortions in relativeprices of prescription drugs and other health-care ser-vices. Compared to prescription drugs, other health-care services, such as production of diagnosticinformation and surgical treatment, are very laborintensive. In contrast to physical products, there is nowholesale market for services, and hence providersdo not have an opportunity to earn a profit marginbetween the reimbursement and the acquisitionprices. With a positive margin on prescription drugs,providers have a financial incentive to substitute pre-scription drugs for other inputs, such as time spent indiagnosis or in surgical treatment. Spending on phar-

    maceuticals consequently accounts for one quarter ofhealth-care expenditures in Taiwan. In 2004, theshare of health-care expenditures spent on drugsranged from 9% to 30% among OECD countries [8].Although many other factors also account for theinternational variation in the share of pharmaceuticalspending, the existence of profit incentives in pre-scribing drugs provides a plausible explanation forthe relatively higher share of drug expendituresobserved in Taiwan.

    Second, although reimbursement prices are fixed bygovernment, there is price competition in the whole-sale market. Given the profit that providers can earn

    from the sale of drugs, the profit margin between thereimbursement and the acquisition price becomes theimportant factor in the drug prescribing decisions.Since there are many drugs within a therapeutic group,the pharmaceutical manufacturers compete by cuttingacquisition prices. Of course, the lower limit on suchprices is marginal cost of manufacturing and distrib-uting the drugs. Providers in turn only select the drugswith higher profit margins into their prescriptionformularies.

    Price competition in the wholesale market hasdriven many products from the market (Table 1).Although there are 21,931 drugs listed in the Taiwans

    national formulary, only about 75% of products areactually sold in a given year. In addition, the distribu-tion of market size (in terms of annual expenditures)by product is highly skewed. About 3700 drugs, or17% of products listed in the national formulary,account for about 97% of expenditures on pharma-ceuticals. The prescribers profit incentive shifts pricecompetition from the retail to the wholesale marketwith the result that only about 3700 drugs remainactively sold in the market. In monetary terms, theannual market size is greater than one million Taiwandollars (or about US$30,000).

    Pharmaceutical Innovation and Drug Expenditure 335

  • 8/3/2019 Adoption of Pharmaceutical Innovation and the Growth of Drug Expenditure

    3/11

    Pricing and Cost Impact of New DrugsNew pharmaceutical products are introduced annuallyas a consequence of technological advances. In addi-tion to obtaining authorization to market a new drug,Taiwan, like other countries with direct price controls

    on pharmaceutical products, requires that the manu-facturer of a new drug obtain approval for coverageand a price for reimbursement by the Bureau ofNational Health Insurance (BNHI) [4,9,10]. Taiwanhas not formally employed economic evaluation toestablish prices for new drugs to be included on itsformulary. Rather, BNHI adopts a mix of strategies todetermine the reimbursement price for new drugs,including reference to existing products and to inter-national comparisons, and the drugs therapeuticvalue.

    BNHI uses the median price of international com-parisons to set the cap on regulated prices for branded

    drugs still on patent. For off-patent branded drugs, theprice cap is 85% of international median prices. Forgeneric drugs, the price cap is the mean price of thebranded price in the same therapeutic group if the drughas passed a bio-equivalent test and 85% of the meanprice of the branded drugs in the same therapeuticgroup if a bio-equivalent test is not conducted for thedrug. The BNHI selects 10 countries as the referencegroup for international comparisons, including

    Australia, Belgium, Canada, France, Germany, Japan,Sweden, Switzerland, United Kingdom, and the UnitedStates [11].

    Many forms of new prescription drugs are addedto the national formulary annually, including new

    molecular entities, formulations, combinations, andindications. As a single payer, BNHI can trace allexpenditures on drugs as well as other health-careservices through its electronic administrative system.Table 2 summarizes the expenditure and number ofnew drugs included in the national formulary between1996 and 2003. A new drug is defined here as one thatwas included in the national formulary after 1996.

    During 19962003, the annual number of newdrugs included in the national formulary has rangedfrom 13 to 75. On average, each drug cost about NT$2159 million annually (1 US$ equaled about 27.534.6 NT$ during this period). Judging from the

    average market size per drug in Taiwan, the annualmean cost per new drug is very significant. Forexample, as shown in Table 1, there were only 717products in Taiwans market (3.2% of total number ofproducts listed in the formulary) for which annualexpenditure on individual drug exceeded NT$20million. Thus, the cost impact of introducing newdrugs was substantial relative to the mean nationallevel. Furthermore, the average annual growth rate of

    Table 1 The distribution of pharmaceutical spending by market size

    Market sizein 2001(NT dollar)

    Numberof drugs

    Percentage oftotal numberof drug (%)

    Pharmaceuticalspending (in

    million NT dollar)

    Percentage of totalpharmaceutical

    spending (%)

    Greater than 100 million 129 0.59 25,666 37.9050100 million 178 0.81 12,230 18.062050 million 410 1.87 12,780 18.75

    1020 million 474 2.16 6,617 10.38110 million 2,536 11.56 9,005 13.49Less than 1 million 12,684 57.84 1,382 2.070 (no sale) 5,520 25.17 0 0Total 21,931 100.00 66,730 100.00

    Source: Bureau of National Health Insurance,Taipei,Taiwan.Note:The exchange rate of US$/NT$ was 34.42 in 2003.

    Table 2 Expenditures and number of new drugs included in the National Health Insurance formulary*

    YearNumber ofnew drugs

    Mean annual expenditureper drug between

    introduction year and2003 (in million NT dollar)

    Mean annual growthrate of expenditure

    between introductionyear to 2003 (%)

    Total expenditureon new drugs

    (in millionNT dollar)

    Spending on newdrugs as percent

    of total spending inpharmaceuticals

    1996 13 59 40 1997 54 21 24 473 0.991998 75 47 49 1,933 3.491999 42 30 64 4,063 6.422000 36 54 190 6,121 9.232001 73 40 229 8,431 12.452002 56 32 216 12,863 16.672003 50 16,694 20.92

    Source: Bureau of National Health Insurance,Taipei,Taiwan.Note:1.The definition of new drugs includes 1) new molecular entity;2) new formulation;3) new combination; and 4) new indication.2.The exchange rate of US$/NT$ was 34.42in 2003.

    336 Hsieh and Sloan

  • 8/3/2019 Adoption of Pharmaceutical Innovation and the Growth of Drug Expenditure

    4/11

    drug expenditure for new products between the yearsof introduction to 2003 ranged form 24% to 229%(Table 2). A shorter observational period was associ-ated with a higher growth rate. This suggests that theintroduction of new drug is not only costly in the shortrun, but also in the longer run. For example, in 1998,

    75 new drugs were included into the formulary and themean annual growth rate of expenditure was 49%from 1998 to 2003, substantially more than the meanannual growth rate of overall expenditures on phar-maceuticals (6.4%). By 2003, total spending on 399new drugs introduced into the national formulary after1996 was about NT$17 billion, accounting for 21% oftotal drug expenditures in that year. This indicates thaton average introducing a new drug into the formularycosts about NT$42 million (17,000/399) per year.

    As suggested by Cutler and McClellan [12], newtechnology in medicine affects the cost of health carethrough the two channels: a treatment substitution

    effect and a treatment expansion effect. In the contextof pharmaceuticals, a treatment substitution effectdescribes substitution of new drugs for older ones fortreating established patients. Because of the high costsof pharmaceutical innovation which is partly reflectedin high prices because of the patent system, the price ofnew drug is higher than for older drugs, at least someof which are no longer patented. Thus, the treatmentsubstitution effect always leads to increased pharma-ceutical expenditures [13,14]. The treatment expan-sion effect indicates that the introduction of new drugsinto the formulary leads more people to be treated fordisease. Hence, the treatment expansion effect also

    leads to increased pharmaceutical expenditure. In thenext section, we will use a statistical method to decom-pose the growth of pharmaceutical expenditure intotreatment substitution effect and expansion effect.

    Decomposition of the Source ofExpenditure Growth

    Previous Studies

    The rapid increase in pharmaceutical expenditure hasled to a search for the primary determinants of thisincrease. By definition, pharmaceutical expendituresare the product of prices and demand (quantity of

    prescription pharmaceuticals consumed). The demandfor drug in turn depends on several factors, such asdrug insurance benefit coverage, income, the incidenceand duration of chronic diseases, and technologicaladvances in medicine.

    In practice, the relative importance of the above-mentioned factors varies with country-specific aspectsof the health-care system and pharmaceutical policy ineach country. For example, in a country in which gov-ernment does not regulate pharmaceutical pricesdirectly, such as the United States, price increases mayhave a relatively important role in accounting for

    expenditure growth. There is evidence that some phar-maceutical manufacturers adopt a penetration strategyto launch their new drugs that leads to an upwardsloping price curve over time [15].

    By contrast, in some countries in which govern-ments impose direct controls on product prices, such

    as Japan and Sweden, price-cap regulation rules outthe use of penetration strategies [16,17]. Real (relativeto a consumer price index for all goods and services)prices of drugs often fall over time as a result of gov-ernment regulation and price-cutting. In these coun-tries, price increases have not been an important causeof increased expenditures. Similarly, the introductionof insurance coverage for prescription drugs is animportant factor in accounting for expenditure growthin countries in which private and/or public drug insur-ance benefit coverage has gradually been added. Forexample, in 1965, only 3.5% of the US prescriptiondrug expenditure was paid by private insurance. By

    1998, the private plus Medicaid insurance shareincreased to 69.8% [3]. The empirical evidence showsthat the growth in the US pharmaceutical expenditureclosely paralleled this expanded drug insurance cover-age [18]. But health insurance is not the primary driverof increased spending in countries that insurance cov-erage for prescription drugs had remained constantover time, such as Japan and Taiwan.

    Given that the major cause or causes of increasedspending varies across countries and over time, severalstudies have sought to identify the underlying driversof spending trends by decomposing the sources ofexpenditure growth [2,3,19,20]. Dubois et al. [2] used

    large claims databases from managed care and theemployer-sponsored health benefit plans in the UnitedStates to disaggregate the growth of drug spendingbetween 1994 and 1998 into several price and volumefactors. They decomposed the change in price intothree components: 1) pure inflation (measuring by anindex of change in actual transaction prices); 2) pricechange because of the change in dosage strength andtherapeutic mix; and 3) change in mean price per dayon account of the introduction of new drugs. Forvolume, they calculated these three measures: 1)changes in the number of prescriptions per person; 2)changes in the number of days supplied per prescrip-

    tion; and 3) changes in the number of users and poten-tial users of prescription drugs per thousand planmembers, which they used as a measure of prevalenceof the disease. In their study, users were defined aspatients treated with a particular drug and potentialusers were defined as patients with a diagnosis corre-sponding to an approved indication for that drug butno drug use. They found that change in volume wasthe primary driver of increased spending for the sevendiseases they studied. In particular, increase in diseaseprevalence and in the number of prescriptions perpatient for new drugs were the two most important

    Pharmaceutical Innovation and Drug Expenditure 337

  • 8/3/2019 Adoption of Pharmaceutical Innovation and the Growth of Drug Expenditure

    5/11

    factors accounting for the growth in pharmaceuticalexpenditures.

    Berndt [3] used aggregate sales data to decomposepharmaceutical expenditure growth in the USA intothree components: 1) price growth of incumbent prod-ucts; 2) quantity growth of incumbent products; and

    3) expenditures on new products. Between 1987 and1994, price growth of incumbent products accountedfor about half of the total expenditure growth. Never-theless, from 1994 through 2000, price growthaccounted for only 20% of expenditure growth. Theremaining 80% was attributable to quantity growth ofincumbent drugs and expenditures on new drugs. Hisfinding suggests that price increases have become arelatively less important factor in explaining thegrowth of the US pharmaceutical expenditures sincethe mid-1990s. Rather, quantity growth, either fromincreased demand for incumbent or new drugs, hasbecome the primary determinant of expenditure

    growth, which is consistent with the studies by Duboiset al. [2]. Berndts analysis further indicated that thisquantity growth was driven by increased drug insur-ance benefit coverage and enhanced marketing efforts,especially direct-to-consumer advertising.

    Morgan [19] used Canadian data to decompose thechange in per capita prescription drug expendituresfrom 1985 to 1999 into four components: 1) thepattern of exposure to drugs across therapeutic catego-ries; 2) the mix of drugs used within therapeutic cat-egories; 3) the price of unchanged products; and 4) therate of generic drug product selection. He foundthat the above-mentioned first three factors worked

    together led to a per capita expenditure growth from$49 to $150 per quarter over the period of study.Nevertheless, generic substitution resulted in a savingin per capita spending of $14 per quarter during thestudys observational period.

    Thus, net increased spending was $87 per quarter.For the expenditure growth of $101 which was drivenby increases in both price and quantity, the increase inprice accounted for 22% of increased spending, indi-cating that price increases have also had a relativelyminor role in explaining growth of pharmaceuticalexpenditure in Canada. Rather the primary cause ofincreased spending was quantity growth, a result con-

    sistent with other studies [2,3]. Morgan emphasizedthat the major economic forces behind the quantitygrowth were an increase in the rate of exposure to agiven therapeutic category of drugs and changes in themix of drugs used within therapeutic categories withthese factors being of almost have equal importanceand together accounting for 78% of the expendituregrowth in Canada.

    Addis and Magrini [20] used data from Italy todecompose the growth of pharmaceutical expendituresinto three components: 1) quantity (expressed in termof defined daily doses, DDDs); 2) price; and 3) a

    change in product mix, a change in expendituresbecause of the shift within the same therapeutic groupof drugs toward more or less expensive products. InItaly, drug expenditure increased 13.5% from 2000 to2001. Their decomposition of this growth rate showedthat the increase in quantity of prescription drugs con-

    sumed (measured by DDDs) led to a 9.5% increase inexpenditure and mix effects led to the expendituregrowth by 4.8%. Price changes led to a decrease indrug expenditure of 1%.

    In spite of the wide variation in statistical methodsand data sources, existing studies provide fairly con-sistent evidence that price is not the primary determi-nant of increased spending in prescription drugs, butrather the quantity growth is. Existing studies alsohave identified several common factors underlyingquantity growth, including an increase in diseaseprevalence, an increase in drug utilization, and achange in the drug mix within therapeutic categories,

    both working to expand treatment and to substitutedrug therapy for other forms of therapy.

    Method

    For this study, we employed the concepts of treatmentexpansion and substitution effects and the statisticalmethod developed by Addis and Magrini [20] todecompose the growth of pharmaceutical expenditurein Taiwan. The growth of pharmaceutical expendituresfrom the base period (0) to the current period (1) canbe expressed as:

    P QP Q

    QQ

    P QP Q

    P Q

    QP Q

    Q

    i i

    i i

    i

    i

    i i

    i i

    i i

    i

    i i

    1 1

    0 0

    1

    0

    1 0

    0 0

    1 1

    1

    1 0

    =

    ii0

    (1)

    In equation 1, Qi1 and Qi

    0 represent quantities of theith prescription drug dispensed in periods 1 and 0respectively, measured in terms of DDD. As men-tioned, the government in Taiwan regulates reimburse-ment prices product by product. Thus, our analysisclassified drugs by brand name. Drugs with the sameingredient but produced by different manufacturers aretreated as different products. Pi

    1 and Pi0 represent the

    prices of the ith drug per DDD in periods 1 and 0,

    respectively. Summing the product of prices by quan-tities for all drugs yields the total expenditure on pre-scription drugs. The left hand side of equation 1therefore represents the growth in pharmaceuticalexpenditures from the base period (0) to the currentperiod (1).

    Expenditure growth can be decomposed into threecomponents on the right side of equation 1. The firstratio on the right side is an index of quantity growthbetween two periods. We used the DDD as the stan-dard unit of prescription drug and hence the quantitiesof prescription drugs could be summed and compared

    338 Hsieh and Sloan

  • 8/3/2019 Adoption of Pharmaceutical Innovation and the Growth of Drug Expenditure

    6/11

    across different drugs and time periods. The growth ofthe pure quantity of drug consumed represents thetreatment expansion effect that arises from morepeople being treated or the increase in treatment inten-sity for the same group of treated patients, such asoccurs when prescriptions per patient and/or days per

    prescription increased.The second ratio represents the Laspeyres priceindex, using quantities of drugs consumed in the baseyear as weights. This index measures the pure pricechange between current period (1) and base period (0).

    The third ratio represents an index for the mixeffect, which equals the ratio of two weighted averageprices per DDD in the current period. The numeratorof this ratio uses the quantity of drug consumed atcurrent period (1) as the weight to calculate the meanprice per DDD. By contrast, the denominator of thisratio uses the quantity of drug consumed in baseperiod (0) as the weight to calculate the mean price per

    DDD. The third ratio is not equal to 1 as long as thequantity mixes of all drugs used in current period ( Qi1)

    are different to those of the base period (Qi0). Since the

    drug prices within the same therapeutic group is notuniform in Taiwan, the drug substitution that physi-cians prescribe different brands of drugs for treatingthe same disease would lead to a change in price perDDD. Thus, this ratio represents the change in themean price per DDD from shifts within the same thera-peutic groups of drugs toward more or less expensiveproducts. If the ratio is greater than 1, it indicates thatthere is an increase in mean price per DDD resultingfrom the change in the mix of quantities between the

    current period and the base period. Thus, this indexrepresents expenditure growth because of treatmentsubstitution.

    Data Sources

    Data for this study came from two sources. First, dataon the utilization of prescription drugs came frominsurance claim files obtained from Taiwans BNHI(hereafter referred to as NHI claims data). After NHIwas implemented in March 1995, and ever since, theBNHI has maintained a national, population-basedclaims database. The NHI claims data contain detailed

    records on utilization of personal health-care services,including outpatient visits, hospital admissions, andprescription drugs. The data on prescription drugsprovide information identifying the drug (drug ID), theanatomical therapeutic chemical (ATC) classificationsystem, the reimbursement price, quantity of utiliza-tion, total spending, and such drug characteristics asingredient name, brand name, dosage, and manufac-turers for all items of drugs listed in NHI formulary.We used NHI claims data for prescription drugs during19972001 to investigate the sources of expendituregrowth between these two periods.

    Second, data on DDD came from Kao et al. [21]. Inthat study, DDD for prescription drugs used in Taiwanwas classified according to the guidelines developed bythe World Health Organization (WHO) CollaboratingCenter for Drug Statistics Methodology [22]. TheDDD is defined as the assumed mean maintenance

    dose per day for a drug used for its main indication inadults [22]. This provided a standard unit of measure-ment invariant to price and formulation. Nevertheless,DDDs are not established for all drugs. There are noDDDs for certain drugs, such as topical products, sera,vaccines, antineoplastic agents, allergen extracts,general and local anesthetics and contrast media [22].Thus, we only included drugs with a DDD assignmentinto our analysis. Among 21,931 items of drugs (inbrand names) listed in NHI formulary, Kao et al. [21]assigned DDDs for 13,295 drugs.

    After merging NHI claims data with DDD data byindividual drug ID, we first transformed the quantity

    of drug consumption in terms of DDD according to thefollowing equation:

    Q q DDDi i i= (2)

    where qi indicates the quantity of drug utilization forthe ith drug (by brand name) in terms of its packageunit obtained from NHI claims data, and Qi indicatesthe quantity of drug utilization for the ith drugexpressed in terms of DDD; using the concept of DDD,we could aggregate across drugs. We then calculatedprice per DDD using

    P E Qi i i= (3)

    where Ei indicates the expenditure for the ith drugobtained from NHI claims data, and Pi represents theprice per DDD for the ith drug.

    After computing quantity and price expressed interms of DDD, we applied the method depicted inequation 1 to decompose the sources of expendi-ture growth into three components: 1) price; 2) quan-tity; and 3) mix effects. In our analysis, the baseperiod refers to 1997 and the current period refers to2001.

    Results

    Table 3 shows our results. For simplicity, we set theindex of base year (1997) equal 1. For all drugs, theindex of drug expenditure grew to 1.56 in 2001, indi-cating that nominal pharmaceutical expendituresincreased 56% during the 5-year period or, equiva-lently, an average annual rate of growth of 11.86%.This growth rate far exceeded the annual growth rateof overall health-care expenditures and gross domesticproduct in Taiwan during this period.

    The decomposition reveals that the price indexdecreased from 1 in 1997 to 0.82 in 2001 or an 18%decrease in the price per DDD. As mentioned above,

    Pharmaceutical Innovation and Drug Expenditure 339

  • 8/3/2019 Adoption of Pharmaceutical Innovation and the Growth of Drug Expenditure

    7/11

    Taiwan adopted a system of direct controls on reim-bursement prices for pharmaceuticals. During ourstudy period, the BNHI cut the reimbursement pricetwice (in 1999 and 2001) for nearly 10,000 drug prod-ucts [11]. BNHI had two justifications for its policy: 1)to offset increasing expenditures from adding newdrugs to its formulary; and 2) to reduce profit marginsearned by providers. Therefore, it is not surprising thatprices decreased.

    In contrast to the decreases in prices, the quantity

    index increased from 1 in 1997 to 1.2 in 2001, imply-ing that the quantity of prescription drugs consumed(measured in terms of DDD) increased 20% duringour study period. This quantity growth comes fromtwo sources: 1) incumbent products; and 2) new prod-ucts. Nevertheless, the approach we used could notseparate these two sources.

    Further, as seen in Table 3, the mix effect was 1.59in 2001, suggesting that the weighted average price perDDD increased 59% during 19972001 as a result ofdrug substitution within therapeutic groups of drugs.Thus, the treatment substitution effect is the dominantdriver of increased spending in Taiwan.

    There are two major factors that led to treatmentsubstitution. First, during the study period, the BNHIcut the reimbursement price twice for selected drugs.The reduction in reimbursement prices for selectedproducts in turn led to a change in profit marginearned by the provider and hence created an incentivefor providers to change the prescription formulary orto renegotiate the acquisition price with the manufac-turers. To increase profit margins, providers have anincentive to substitute the incumbent drugs with higherreimbursement prices for those with lower reimburse-ment prices, other things being equal. This substitution

    may occur within the same ingredient but differentbrands. In addition, it may occur among products withdifferent ingredients but within the same therapeuticcategory. Our approach could not separate these twotypes of substitution effect. Second, the provider alsosubstituted new drugs for the incumbent drugs afterthe new drugs were added to the formulary.

    As Addis and Magrini [20] noted, the approach weused does not take into account the impact of intro-ducing new drugs. Thus, we cannot know how much

    that the treatment expansion (quantity growth) andtreatment substitution (mix effect) came from thissource. For this reason, we further decomposed thesource of expenditure growth by drug vintage andATC classification. First, we categorize drugs into twogroups: old and new with new drugs being thoseincluded on the formulary after 1996. As shown inTable 3, among 13,295 drugs in our analysis, 13,068drugs are old and 227 drugs are new. The index ofoverall expenditure grew from 1 in 1997 to 25.22 in2001 for new drugs. By contrast, during the sameperiod, the overall expenditure only increased 38% forold drugs that were included in the national formulary

    before 1996.The decomposition reveals that the quantity index

    for 227 new drugs increased from 1 in 1997 to 315 in2001, representing a 314 fold increase in quantity ofconsumption. Although one would expect that theincrease for new drugs would be high, it is extremelyhigh. Since our base period is 1997, some of the 227new drugs reported in Table 3 may not be introducedon the formulary at that time. In this case, the quan-tities of consumption (in terms of DDD) for thosedrugs in 1997 were zero. Thus, the very large numberof quantity index for new drugs reported in Table 3

    Table 3 Index of NHI pharmaceutical expenditures in 2001 (base year 1997 = 1)

    Classification of drugsNumber of

    i tems Total spe nding

    Decomposition of increased spending

    Price Quantity Mix

    All drugs 13,295 1.56 0.82 1.20 1.59By anatomical therapeutic chemical

    Alimentary tract and metabolism 2,108 1.39 0.81 1.33 1.29Blood and blood forming organs 637 1.71 0.82 1.13 1.84Cardiovascular system 1,454 1.88 0.80 1.17 2.00Dermatologicals 48 2.09 0.76 1.85 1.49Genito urinary system and sex hormones 618 1.66 0.90 1.56 1.18Systemic hormonal preparations, excl. sex hormones and insulins 653 1.83 0.92 1.27 1.57Antiinfectives for systemic use 2,540 1.25 0.81 0.79 1.96Antineoplastic and immunomodulating agents 120 2.39 0.88 1.39 1.96Musculo-skeletal system 1,424 1.39 0.77 1.22 1.47Nervous system 1,795 1.76 0.85 1.34 1.55Antiparasitic products, insecticides and repellents 198 1.42 0.48 1.19 2.50Respiratory system 1,568 1.34 0.86 1.03 1.51Sensory organs 98 1.61 0.99 0.98 1.66Various 34 3.10 0.56 4.83 1.15

    By drug vintageOld 13,068 1.38 0.82 1.17 1.44New* 227 25.22 0.91 315.45 0.09

    *New drugs indicate those included in the NHI formulary after 1996.

    NHI, National Health Insurance.

    340 Hsieh and Sloan

  • 8/3/2019 Adoption of Pharmaceutical Innovation and the Growth of Drug Expenditure

    8/11

    not only reflects the growth of quantity for the existingnew drugs in 1997 but also includes the effect ofadding more new drugs on the formulary between1998 and 2001. It is apparent that the introduction ofnew drugs led to a very rapid increase in pharmaceu-tical expenditures and expenditure growth of new

    drugs mainly came from quantity growth.Second, we categorize drugs into 14 groups by ATCclassification. The results for subgroup analysis consis-tently indicate that treatment expansion and substitu-tion effects were the major causes of increasedpharmaceutical expenditures. In particular, the growthrate of expenditure for antineoplastic and immuno-modualating agents was the highest among 14 sub-groups, except for the group of various. Other groupswith relatively high rates of growth in expenditureswere drugs for cardiovascular system and nervoussystem. These three groups of drugs share twocommon characteristics: 1) they are for chronic dis-

    eases; and 2) there was relatively more pharmaceuticalinnovation in these subgroups. These two factors areimportant for the treatment expansion and substitu-tion effects to occur. As population ages, prevalencerate of several chronic diseases, such as cancer, hyper-tension, heart disease, and depressant, increase. Theintroduction of new drugs leads more people to betreated for these diseases. Also, the relatively quickintroduction of new drugs within these disease catego-ries is more likely to induce a drug substitution intreating established patients.

    Overall, our decomposition clearly demonstratesthat the primary driver of increased spending in phar-

    maceuticals is not price. Rather, treatment expansionand substitution effects are the major determinants ofexpenditure growth. Because of the limitation of ourapproach, we are unable to provide the direct evidenceto show the extent of the treatment expansion andsubstitution effects that is induced by pharmaceuticalinnovation. Nevertheless, the decomposition of oursubgroup analysis shows that the adoption of pharma-ceutical innovation is the primary driver for treatmentexpansion and substitution effects. As reported inTable 3, the quantity index is extremely large for thesubgroup of new drugs. Also, Table 3 shows thatquantity indexes are relatively higher for those drugs

    with more pharmaceutical innovation in recent years,such as for cardiovascular diseases, cancer, and mentalhealth. Therefore, our results provide a strong indica-tion that pharmaceutical innovation is the main under-lying cause of increased spending on pharmaceuticals.

    Comparing Costs and Benefits

    To the extent that innovation is the major cause of theexpenditure increase, the next question is: Was theinnovation worth the increased cost?

    To answer this question, we must quantify the ben-efits of new drugs. Compared to costs, benefits aremore difficult to quantify. First, benefits are multiple-dimensional, including at least the following four com-ponents: 1) reductions in mortality; 2) reductions indisability and morbidity; 3) improved quality of life,

    i.e., reduced pain and suffering; and 4) improved laborforce productivity. Second, many potential health ben-efits of prescription drugs accrue in the long run. Forexample, a drug for reducing blood cholesterol mayreduce heart disease prevalence several decades hence.Therefore, the long-run effects of drugs cannot fullybe assessed until many years after such drugs areintroduced.

    In spite of these difficulties, but given the impor-tance of the topic, several studies have quantifiedhealth benefits of prescription drugs [6,23]. In partreflecting lack of data on other outcomes, most studieshave focused on health benefits from reduced mortality

    [23]. To calculate benefits from longevity gains, oneneeds to estimate how much of the increase in lifeexpectancy is attributable to the introduction of newdrugs. Lichtenberg [6] investigated this issue by usingan indirect two-step approach. First, he estimated theaverage effect of change in pharmaceutical innovationon the change in probability of survival to age 65 froma disease-specific panel database. Specifically, he testedthe hypothesis that diseases for which there has rela-tively more pharmaceutical innovation have experi-enced more improvements in health outcomes thanhave outcomes for other diseases for which technologi-cal progress has lagged. Second, he estimated the mean

    effect of change in the probability of survival to 65 forthe whole population on the change in life expectancyat birth using time series data. By multiplying thesetwo estimates, he obtained the mean effect of change inpharmaceutical innovation on the change in lifeexpectancy at birth. His results showed that, between1986 and 2000, mean life expectancy of the entirepopulation in his sample increased by 1.96 years. Thelaunch of new drugs accounted for 0.79 years, 40% ofthe 19862000 increase in longevity, implying themean annual increase in life expectancy from the intro-duction of new drugs for the period 19862000 isabout three weeks (52 0.79/14).

    Hsieh et al. [24] used a disease-specific panel data-base obtained from Taiwan for the period 19852002and a similar approach to estimate the effect of newdrug launches on longevity in that country. They foundthat a 10% increase in the cumulative number of newmolecular entities was associated with an increase inlife expectancy at birth of approximately 0.1%. Theyreported that the cumulative number of new molecularentities increased about 50%, from 548 in 1996 to 821in 2002, which in turn increased life expectancy by0.5%. In 1996, the Taiwanese life expectancies formales and females were 71.89 and 77.77, respectively

    Pharmaceutical Innovation and Drug Expenditure 341

  • 8/3/2019 Adoption of Pharmaceutical Innovation and the Growth of Drug Expenditure

    9/11

    [25]. Thus, the predicted longevity gains resulting fromnew drug launches were 0.36 (71.89 0.005) and0.39 (77.77 0.005) years for males and females,respectively. During 19962002, the actual increase inlife expectancy was 1.33 years for males and1.17 years for females [25]. These results imply thatnew drug launches accounted for about 27% (0.36/1.33) to 33% (0.39/1.17) of the longevity gain inTaiwan. The mean annual increase in life expectancy

    of the Taiwanese population resulting from new druglaunches was about three weeks (52 0.36/6), similarto Lichtenbergs [6] estimates.

    Used the estimated longevity gains from the intro-duction of new drugs, together with the cost impact ofintroducing new drugs into the formulary describedabove, we computed the incremental cost-effectivenessratio for new drugs (Table 4). Since the estimate of theeffectiveness came from another study, the observa-tional period is not exactly consistently with this study.In spite of this difference in timing, the existing esti-mate of health benefit obtained from the period of19962002 serves as a good proxy for the benefit of

    the subsequent year in 2003. As indicated above, theBNHI, a single public payer, spent NT$16,694 millionon new drugs in 2003 (see Table 2). Thus, from theperspective of public insurer, expenditure of new drugsper capita was NT$725total BNHI spending on newdrugs divided by total population in Taiwan, 23million. The mean annual increase in life expectancy ofthe entire male population resulting from the introduc-tion of new drugs was 0.06 years (0.36/6). The ratio ofthese two number is about NT$12,083 (or aboutUS$351 in 2003 dollars), indicating that the cost perlife-year gained resulting from the launch of new drugsis extremely low as compared to most estimates of the

    statistical value of a life-year [7,26,27].Although Taiwan has NHI, only about two-thirds

    of health-care expenditure is financed by the publicsector. Private sector expenditures, either from out-of-pocket payments or private supplemental health insur-ance, account for the remaining one-third of suchexpenditures. Thus, it is necessary to account forprivate payments for new drugs, such as spending onnew drugs not listed in NHI formulary, if one is tomeasure cost from a societal perspective. As shown inTable 2, spending on new drugs accounted for about21% of total pharmaceutical expenditure paid by

    BNHI. Assuming that the private sector also spent thisshare on new drugs and multiplying per capita phar-maceutical expenditure by 21% yields national spend-ing on new drug per capita as a measure of cost onnew drugs from the perspective of society as a whole.In 2003, per capita pharmaceutical expenditure inTaiwan was NT$5980 [28]. Thus, national spendingon new drugs per capita was NT$1256 (5980 0.21).Dividing this number by the longevity gain attributable

    to new drug launches, the cost per life-year gained wasNT$20,933 (or US$608 in 2003 dollars) from a soci-etal perspective.

    As noted by Pauly [29], there are debates on correctmeasurement of drug costs from a societal perspective.Most researchers used average wholesale price (AWP)as the measure of cost from a societal perspective.Nevertheless, this practice is controversial becauseAWP may include the excess profit of the pharmaceu-tical manufactures which is a transfer payment insteadof true cost to the society. Moreover, as mentioned,providers in Taiwan earn the profit margin between thereimbursement and the acquisition prices. Thus, the

    per capita pharmaceutical expenditure may overesti-mate the true cost from a societal perspective. Becauseof the lack of reliable data on the profit margin, theresult reported in Table 4 overestimates the cost onnew drugs from the perspective of society as a whole.

    Similarly, the measurement of benefits reported inTable 4 is subjected to estimate biased. Since we onlyconsidered benefits in the form of mortality reductionsand excluded other benefits of new drugs, such asimproved quality of life and labor force productivity,our cost-effectiveness calculation underestimated thebenefit of including new drugs on the formulary. Bycontrast, we may have overestimated the benefit of

    longevity gains resulting from the launches of newdrugs given that the determinants of mortality reduc-tion are a very complex [30]. Although, as Hsieh et al.[24] acknowledged, they only partially controlled forother determinants of mortality reduction with thefixed-effects and random-effects models, there may stillbe unobservable characteristics not accounted for intheir analysis. In particular, Lichtenberg [6] noted thatother technological advances other than in the phar-maceutical sphere may have co-occurred and were notexplicitly captured by the analysis. Thus, the longevitygain resulting from new drug launches may be overes-

    Table 4 The incremental cost effectiveness ratio of new drug launches in Taiwan

    Perspective ofmeasuringcost

    Cost (spendingon new drug

    per capita)(NT$)

    Effectiveness (averageannual increase in life

    expectancy attributableto new drug launches)

    Cost-effectivenessratio (cost per

    life-yeargained, NT$)

    Public insurer (BNHI) 725 (US$21.06) 0.06 12,083 (US$351)Society as a whole 1,256 (US$36.49) 0.06 20,933 (US$608)

    Note:The exchange rate of US$/NT$ was 34.42 in 2003. Number in parenthesis is cost expressed in terms of US$.BNHI, Bureau of National Health Insurance.

    342 Hsieh and Sloan

  • 8/3/2019 Adoption of Pharmaceutical Innovation and the Growth of Drug Expenditure

    10/11

    timated if the researcher simply uses the cumulativenumber of drug approved with new molecular entitiesas a measure of pharmaceutical innovation.

    To correct the potential bias in quantifying benefits,Lichtenberg [6] took only one-third of the estimatedeffect of new drugs on longevity gain to calculate the

    cost per life- year gained from the launch of new drugs.This correction was based on the evidence that USpharmaceutical R&D expenditure accounts for aboutone-third of national health R&D expenditure. Thus,it is plausible that the remaining two-thirds of esti-mated effect of new drugs on longevity gain may beattributable to other medical innovations.

    Using the same approach by dividing mean annualincrease in life expectancy by three yields a cost perlife-year gained resulting from new drug launches ofUS$1053 from the perspective of public insurer andUS$1824 from the societal perspective. These esti-mates are still far lower than most estimates of the

    value of a statistical life-year. Also, our estimates oncost per life-year gained resulting from the launches ofnew drugs are lower than that estimated by Lichten-berg [6] (US$6750). Health benefits of new druglaunches in Taiwan are almost the same as in othercountries, but the per capita pharmaceutical expendi-ture in Taiwan was lower than that for the USA used inLichtenbergs calculations.

    Conclusion

    This study used two different approaches to investigatethe effect of adopting pharmaceutical innovation on

    the growth of drug expenditure. First, we described thegrowth of public expenditure resulting from the intro-ducing new drugs into the formulary. Our analysisindicates that during 19962003, the single publicinsurer in Taiwan added 399 new drugs to the nationalformulary. In 2003, total expenditure on these 399new drugs was NT$16,694 million, accounting for21% of total public spending in pharmaceuticals.Second, we employed a statistical method to decom-pose the growth of public pharmaceutical expendituresinto three components: 1) quantity (treatment expan-sion); 2) mix (treatment substitution); and 3) pureprice effects. We found that during the period from

    1997 to 2001 the public expenditure on pharmaceuti-cals grew about 57%. The primary determinants ofthis expenditure growth were treatment expansion andtreatment substitution. Prices of pharmaceuticalsdeclined during this period.

    Overall, our analysis provides evidence consistentwith existing literature, namely that price is not theprimary driver of increased spending [2,3,19,20].Rather the introduction of new drugs onto the formu-lary leads to treatment expansion and substitution,which in turn boosts spending. We therefore furthercompared the cost and benefit of introducing new

    drugs into the formulary. Combining an estimatedbenefit from an existing study with our estimate ofcost, we found that cost per life-year gained resultingfrom the introduction of new drugs was US$1053from the perspective of public insurer and US$1824from the societal perspective, far lower than estimates

    of the value of a statistical life-year.Our findings have three implications for publicpolicy. First, price regulation is not an effectiveapproach for controlling pharmaceutical expenditures.Although our study shows that the direct regulation onreimbursement prices did reduce the rise of pricesbelow that which would have otherwise occurred, itdid not reduce overall pharmaceutical expenditure. Bycontrast, there are many side effects of using priceregulation for cost containment, including the launchdelay for new drug and disincentive for innovation[10,16]. Furthermore, in a health-care system in whichproviders are in a position to profit directly from pre-

    scribing drugs, cost containment by cutting regulatedprices often alters profit margins between the reim-bursement and acquisition prices. This in turn createsan incentive to substitute drugs toward products withhigher profit margins. From a clinical view point, theunjustified substitution between drugs may adverselyaffect quality of care, a potential adverse side effect wedid not measure.

    Second, overall adoption of pharmaceutical inno-vations is worth the increased cost of new drugs.Although introducing new drugs on the national for-mulary increased public outlays on pharmaceuticals,adoption of pharmaceutical innovations also in-

    creases longevity. Our conservative estimate suggeststhat cost per life-year gained resulting from theintroduction of new drugs from the societal perspec-tive was US$1,824, only about 1.8% of value ofa statistical life-year at US$100,000 widely used inliterature [7].

    Third, although the adoption of pharmaceuticalinnovation as a whole is cost- effective, all newdrugs arenot equally beneficial relative to their costs, as noted bySkinner et al. [31] that technological change in medi-cine is not always worth it. This study presents anaverage effect and does not imply that every new drug iscost-effective. Our analysis has documented that the

    cost impact of introducing a new drug into the formu-lary is very significant. Facing substantial budgetarypressures, policymakers often postpone the introduc-tion of new drugs or reduce the prices of the new drugsat introduction. Our results suggest that such strategiesare likely to be unwise as an overall strategy. An effec-tive solution for the policy dilemma between control-ling health-care budget and improving the health ofpopulation is a systematic analysis of the costs versusbenefits of each new drug before introduction. Somegovernments throughout the world have used themethod of economic evaluation to systematically

    Pharmaceutical Innovation and Drug Expenditure 343

  • 8/3/2019 Adoption of Pharmaceutical Innovation and the Growth of Drug Expenditure

    11/11

    evaluate the value of new drugs before including theminto the formulary [32]. By contrast, Taiwan has not yetformally implemented this approach. Such evaluationrepresents a promising approach for boosting the valueof public spending still further.

    Source of financial support: No funding was received for thisresearch.

    References

    1 Newhouse JP. Medical care costs: how much welfareloss? J Econ Perspect 1992;6:321.

    2 Dubois RW, Chawla AJ, Neslusan CA, et al. Explain-ing drug spending trends: does perception matchreality? Health Aff 2000;19:2319.

    3 Berndt ER. Pharmaceuticals in U.S. health care: deter-minants of quantity and prices. J Econ Perspect2002;16:4566.

    4 Ess SM, Schneeweiss S, Szucs TD. European health-

    care policies for controlling drug expenditure. Phar-macoeconomics 2003;21:89103.

    5 Murphy K, Topel RH. The economic value of medicalresearch. In: Murphy KM, Topel RH, eds. Measuringthe Gains from Medical Research: An EconomicApproach. Chicago, IL: University of Chicago Press,2003.

    6 Lichtenberg FR. The impact of new drug launches onlongevity: evidence from longitudinal, disease-leveldata from 52 countries, 19822001. Int J Health CareFinance Econ 2005;5:4773.

    7 Cutler DM. Your Money or Your Life: Strong Medi-cine for Americas Health Care System. New York:Oxford University Press, 2004.

    8 Organisation of Economic Co-operation and Devel-opment. OECD Health Data 2006: Statistics and Indi-cators for 30 Countries. Paris: OECD Publishing,2006.

    9 Drummond M, Jonsson B, Rutten F. The role of eco-nomic evaluation in the pricing and reimbursement ofmedicines. Health Policy 1997;40:99215.

    10 Danzon PM, Wang RY, Wang L. The impact of priceregulation on the launch delay of new drugs: evidencefrom twenty-five major markets in the 1990s. HealthEcon 2005;14:26992.

    11 Cheng C, Hsieh CR. Economic analysis of NHI phar-maceutical policies and drug expenditures. SocioeconLaw Inst Rev 2005;35:142 (in Chinese).

    12 Cutler DM, McClellan M. Is technological change inmedicine worth it? Health Aff 2001;20:1129.13 Duggan M. Do new prescription drugs pay for them-

    selves? The case of second-generation antipsychotics. J Health Econ 2005;24:131.

    14 Lichtenberg F. Are the benefits of newer drugs worththeir costs? Evidence from the 1996 MEPS. Health Aff2001;20:24151.

    15 Lu ZJ, Comanor WS. Strategic pricing of new phar-maceuticals. Rev Econ Stat 1998;80:10818.

    16 Danzon PM, Chao LW. Does regulation drives outcompetition in pharmaceutical markets? J Law Econ2000;43:31157.

    17 Elelund M, Persson B. Pharmaceutical pricing in aregulated market. Rev Econ Stat 2003;85:298306.

    18 Danzon PM, Pauly MV. Health insurance and thegrowth in pharmaceutical expenditures. J Law Econ2002;45:587613.

    19 Morgan SG. Quantifying components of drug expen-diture inflation: the British Columbia seniors drug

    benefit plan. Health Serv Res 2002;37:124366.20 Addis A, Magrini N. New approaches to analyzing

    prescription data and to transfer pharmacoepidemio-logical and evidence-based reports to prescribers.Pharmacoepidemiol Drug Saf 2002;11:7216.

    21 Kao YH, Kuo CW, Hong FZ, Kuo SC. ATC classifi-cation and DDD assignment in Taiwan. TechnicalReport. Taipei: Bureau of National Health Insurance,2004 (in Chinese).

    22 WHO Collaborating Center for Drug Statistics Meth-odology. Guidelines for ATC Classification and DDDAssignment (7th ed.). Oslo: Norwegian Institute ofPublic Health, 2004.

    23 Cremieux PY, Jarvinen D, Long G, Merrigan P. Phar-

    maceutical spending and health outcomes. In: SloanFA, Hsieh CR, eds. Pharmaceutical Innovation: Incen-tives, Competition, and Cost-Benefit Analysis inInternational Perspective. New York: CambridgeUniversity Press, 2007.

    24 Hsieh CR, Lo KT, Hong YC, Shih YC. Adoption ofpharmaceutical innovation and health outcomes:empirical evidence from Taiwan. In: Sloan FA, HsiehCR, eds. Pharmaceutical Innovation: Incentives, Com-petition, and Cost-Benefit Analysis in InternationalPerspective. New York: Cambridge University Press,2007.

    25 ROC Department of Health. Vital Statistics. Taipei:DOH, 2006.

    26 Nordhaus WD. The health of nations: the contribu-tion of improved health to living standards. In:Murphy KM, Topel RH, eds. Measuring the Gainsfrom Medical Research: An Economic Approach.Chicago, IL: University of Chicago Press, 2003.

    27 Abelson P. The value of life and health for publicpolicy. Econ Rec 2003;79(Suppl.):S213.

    28 ROC Department of Health. National Health Expen-diture. Taipei: DOH, 2006.

    29 Pauly MV. Measures of costs and benefits for drugs incost effectiveness analysis. In: Sloan FA, Hsieh CR,eds.Pharmaceutical Innovation: Incentives, Competition,and Cost-Benefit Analysis in International Perspective.New York: Cambridge University Press, 2007.

    30 Cutler DM, Deaton AS, Lleras-Muney A. The determi-nants of mortality. J Econ Perspect 2006;20:97120.31 Skinner JS, Staiger DO, Fisher ES. Is technological

    change in medicine always worth it? The case of acutemyocardial infarction. Health Aff-Web Exclusive2006;25:w3447.

    32 Drummond M. Using economic evaluation in reim-bursement decisions for health technologies: lessonsfrom international experience. In: Sloan FA, Hsieh CR,eds. Pharmaceutical Innovation. Incentives, Competi-tion, and Cost-Benefit Analysis in International Per-spective. New York: Cambridge University Press,2007.

    344 Hsieh and Sloan