Skip to main content

Advertisement

SpringerLink
Log in

Medical providers as double agents in a universal health care system: evidence from generic pharmaceutical adoption in Taiwan

  • Published:
Empirical Economics Aims and scope Submit manuscript

Abstract

This paper investigates medical providers’ generic pharmaceutical adoption in Taiwan, stresses that this decision involves the interests of providers, patients and insurance payer. We examine this prescription behavior using Taiwanese data because patients and physicians did not self-select their insurance plans under a universal health care system. Physicians in Taiwan also respond to strong financial incentives because they are allowed to both prescribe and dispense drugs. The empirical results show that a larger price difference between brand-name and generic drugs increases physicians’ likelihood for prescribing generic prescriptions. However, this effect decreases as the payer’s cost share percentage increases. This study also demonstrates that some physicians prescribed more generic drugs than the others, including the hospital and clinic owners, and the ones practicing in clinics and private institutions.

This is a preview of subscription content, log in via an institution to check access.

Access this article

Log in via an institution

Price excludes VAT (USA)
Tax calculation will be finalised during checkout.

Instant access to the full article PDF.

Institutional subscriptions

Fig. 1
Fig. 2

Similar content being viewed by others

Notes

  1. Pauly (1968) and Hellerstein (1998) have identified this incentive conflicts to be a moral hazard problem, in which either the amount or the price of the health care demanded is greater than the socially optimal level.

  2. Using the data from USA, where physicians are not allowed to both dispense drugs, Carrera et al. (2015) found that medical providers are responsive to the pharmaceutical costs only when the price change is large and universal. Meier and Thomas of The New York Times (2012) reported that insurers pay substantially high markups when physicians dispense drugs in their offices. This practice is common among physicians who treat injured workers, because there are loopholes in some states’ worker compensation insurance that allow physicians to sell the drugs for substantially high markups. Physicians argue that this practice is best for patients because it saves their time, but the resulting cost is borne by the payer, not the patient or physician. Whoriskey and Keating of the The Washington Post (2013) also reported that physicians profit from more-costly treatment when the alternative is equally effective in the USA. This costs Medicare a billion or more of extra payment annually.

  3. Note that \(\gamma _{1}\)+\(\gamma _{2}\) is not restricted or normalized to be 1, because another weight is to be introduced.

  4. Notice that the definition of \(\Delta \) of the \(\Delta P_{k}\) is opposite from the \(\Delta \)’s of \(\Delta q_{k}\) and \(\Delta u_{ik}\), because the price is negative in patient’s utility.

  5. We are grateful for comments from several physicians on this point. Also, \(\Delta q_{k}\) can be interpreted as perceived quality for patients. For example, most patients in Taiwan believe that brand-name drugs are better than their generic alternatives. Given this preference, patient i’s utility from generic prescriptions decreases as the quality difference between brand-name and generic drugs increases.

  6. For example, consider \(\xi _{jzk}\)=\(P_{zk}-P_{zk}^{T}\), where \(P_{zk}\) is the reimbursement price of drug k paid by the NHI to providers, and \(P_{zk}^{T}\) is the acquisition cost paid by the provider to pharmaceutical producers. Because \(P_{zk}^{T}\) was not observed in the data, \(\xi _{jzk}\) account for this profit that varies across providers j. \(\xi _{jzk}\) also proxies for other financial incentives for providers that are not revealed in the data, such as gifts or donations from pharmaceutical firms to hospitals.

  7. For example, consider \(\xi _{jzk}\)=\(P_{zk}-P_{zk}^{T}\) as discussed in footnote 6. Large hospitals are able to negotiate a better price to purchase a z version of drug k, \(P_{zk}^{T}\), than the price negotiated with clinics.

  8. By contrast, Liu et al. (2009) hypothesized that, when market competition increases, generic firms provide a higher discount rate for physicians than their brand-name counterparts do, because their marginal cost is lower. However, because the reimbursement price of a drug decreases as the number of competitors increases, and the marginal cost is unaffected, the generic firms’ profits that can be applied to promote their drugs decreases faster than that of the brand-name firms. From physician’s perspective, consider \(\xi _{jzk}\)=\(P_{zk}-P_{zk}^{T}\) as discussed in footnote 6, the expected profit difference from prescribing a generic drug and a brand-name drug is \(\xi _{j1k}-\xi _{j0k}= (P_{0k}^{T}-P_{1k}^{T})-(P_{0k}-P_{1k})\). Therefore, \(P_{0k}-P_{1k}\) increases as market competition increases because the price of brand-name drugs are affected less than generic drugs, which then reduce physician’s financial incentives of prescribing generic drugs. By contrast, Liu et al. (2009) assumed that, as market competition increases, \(P_{1k}^{T}\) decreases thus \(P_{0k}^{T}-P_{1k}^{T}\) increases.

  9. A fixed-effect probit or logit method can be applied by using physician-specific fixed effects to proxy \(c_{jk}\). However, to obtain a consistent estimate, the number of patients per physician must grow at a rate faster than the number of physicians in the sample (Hellerstein 1998). Therefore, it is not applied in the current study.

  10. We thank Dr. Heng-Sim Lee for sharing these institutional details about hospitals and clinics in Taiwan.

  11. For example, The China Post reported a large hospital chain in Taiwan was found to defraud from the NHI program for more than $100 million dollars and prosecuted in the late September, 2014. The owner of the medical group is suspected of procuring pharmaceutical products at lower prices through a subsidiary company, then sold the medicine to his hospital chain at regulated prices. He leveraged bulk-purchase discounts from suppliers and reaped the profits through the price difference (Chen 2014).

  12. Sometimes patients would ask for certain drugs, but physicians are not obligated to prescribe. For example, each time when the NHIA adjusted the reimbursement price for pharmaceuticals, media would report that some patients being worried that their prescription drugs for chronic disease would be switched from branded drugs to generic ones.

  13. Specifically, glimepiride and acarbose were not included in the 2003 sample; repaglinide and nateglinide were included only in the sample for 2005–2007; and pioglitazone was included only in the sample for 2006–2007. Appendix Table 7 lists these drugs as 0% under Percentage Generics column in certain sample years.

  14. We thank a referee for this point.

  15. Daily dose of each OHA is defined by the WHO Collaborating Centre for Drug Statistics and can be found at http://www.whocc.no/atc_ddd_index. The defined daily doses is important here because drug price were defined as the price required for defined daily doses of a drug, which makes the price comparable among various drugs.

  16. As mentioned in the previous section, brand-name drugs are assumed to be always more expensive than the generic ones. Because the dependent variable is a dummy variable showing whether generic drugs were prescribed, \(\Delta P\) is positively related to the dependent variable. However, this study focuses on the effect of incremental increase in \(\Delta P\), and the effect of government’s share of \(\Delta P\), \(\theta \), on physician’s prescription behavior.

  17. As shown in the Appendix Table 6, cost sharing percentage is calculated by dividing total amount paid to the applicant hospital (T\(\_\)APPL\(\_\)AMT) by total cast per medication order (\(\hbox {T}\_\)AMT), where the variable names in the NHIRD are listed in the brackets. Total cost per order is a sum of drug expense (DRUG\(\_\)AMT), treatment expense (TREAT\(\_\)AMT), diagnostic expense (DIAG\(\_\)AMT), and pharmaceutical service expense (DSVC\(\_\)AMT). While the later two expenses are labor costs for physicians and pharmacists, respectively, and the drug expense follows the copayment scheme listed in Table 2, treatment expense is the likely source that lower the payer’s cost sharing. Because the sample were restricted on the prescribed OHAs, it is assumed that all the expenses are related to the prescribed medications.

  18. Hellerstein (1998) also used a random sample of patients of selected physicians.

  19. For example, experienced physicians and physicians in long-established hospitals probably are more familiar with pharmaceutical suppliers than the other physicians.

  20. For example, as shown in Tables 3 and 4, owners of hospitals and clinics were 24% more likely to prescribe generic drugs than other physicians; however, clinic owners were only 4% more likely to prescribe generic drugs than other physicians in clinics.

  21. The results observed when using only physicians in academic medical centers are similar to the results shown in Table 4 and are not reported in this paper, but are available upon request.

  22. The results observed when using only physicians in academic medical centers show that physicians were 20% less likely to prescribe generic drugs with one more competitor in the ingredient market. This effect was statistically significant with p value < 0.001.

  23. Regarding physicians’ access to generic drugs, O’Malley et al. (2006) documented several policies adopted by Blue Cross Blue Shield Michigan to increase physicians’ prescriptions of generic drugs, including direct communications with physicians, newspaper advertising, financial incentives for physicians to prescribe generic drugs, and distributing free samples of generic drugs. Regarding patients’ access to generic drugs in Taiwan, most patients consider the quality of brand-name drugs to be better than generic drugs, even when their quality is proved similar.

  24. Lien (2008) described how to conducting economic research using the NHIRD in Chinese, using his research as an example.

  25. Registry files that were not used in this study were ignored here.

References

  • Baker LC, Bundorf MK, Kessler DP (2016) The effect of hospital/physician integration on hospital choice. J Health Econ 50:1–8

    Article  Google Scholar 

  • Blomqvist A (1991) The physician as double agent: information asymmetry, health insurance, and medical care. J Health Econ 10:411–432

    Article  Google Scholar 

  • Carrera M, Goldman D, Joyce G, Sood N (2015) Do physicians respond to the costs and cost-sensitivity of their patients? Working paper

  • Chamberlain G (1980) Analysis of covariance with qualitative data. Rev Econ Stud 47:225–238

    Article  Google Scholar 

  • Chandra A, Cutler D, Song Z (2014) Who ordered that? The economics of treatment choices in medical care. Handb Health Econ 2(6):397–425

    Google Scholar 

  • Chen T (2014) Ex-lawmaker probed over NT\({\$}400\) mil. NHI fraud. The China Post. https://chinapost.nownews.com/20140926-50140

  • Chen BK, Gertler PJ, Yang C-Y (2013) Moral hazard and economies of scope in physician ownership of complementary medical services. NBER working paper 19622

  • Cheng T-M (2003) Taiwan’s new national health insurance program: genesis and experience so far. Health Aff 22(3):61–76

    Article  Google Scholar 

  • Chou YJ, Yip WC, Lee CH, Huang N, Sun YP, Chang HJ (2003) Impact of separating drug prescribing and dispensing on provider behaviour: Taiwan’s experience. Health Policy Plan 18(3):316–329

    Article  Google Scholar 

  • Coscelli A (2000) The importance of physicians’ and patients’ preferences in the prescription decision. J Ind Econ 48:349–369

    Article  Google Scholar 

  • Dafny LS (2005) How do hospitals respond to price changes? Am Econ Rev 95(5):1525–1547

    Article  Google Scholar 

  • Dickstein M J (2014) Physician vs. patient incentives in prescription drug choice. Working paper, Stanford University

  • Duggan MG (2000) Hospital ownership and public medical spending. Q J Econ 115:1343–1373

    Article  Google Scholar 

  • Engelberg J, Parsons C A, Tefft N (2013) First, do no harm: financial conflicts in medicine. Working paper

  • Hellerstein J (1998) The importance of physician in the generic verse trade-name prescription decision. RAND J Econ 29:108–136

    Article  Google Scholar 

  • Howard DH, David G, Hockenberry J (2016) Selective hearing: physician-ownership and physicians’ response to new evidence. NBER working paper 22171

  • Hsieh CR (2009) Pharmaceutical policy in Taiwan. In: Eggleston K (ed) Prescribing cultures and pharmaceutical policy in the Asia-Pacific. Brookings Press, Baltimore

    Google Scholar 

  • Iizuka T (2007) Experts’ agency problems: evidence from the prescription drug market in Japan. RAND J Econ 38(3):844–862

    Article  Google Scholar 

  • Iizuka T (2012) Physician agency and adoption of generic pharmaceuticals. Am Econ Rev 102(6):2826–2858

    Article  Google Scholar 

  • Lai HP, Tang MC (2018) Hospital efficiency under global budgeting: evidence from Taiwan. Empir Econ 55(3):937–963

    Article  Google Scholar 

  • Lien HM (2008) Conducting economic analysis using national health insurance data. Taiwan Econ Rev 36(1):115–143 (in Chinese)

    Google Scholar 

  • Liu YM, Yang YHK, Hsieh CR (2009) Financial incentives and physicians’ prescription decisions on the choice between brand-name and generic drugs: evidence from Taiwan. J Health Econ 28:341–349

    Article  Google Scholar 

  • Liu YM, Yang YHK, Hsieh CR (2012) Regulation and competition in the Taiwanese pharmaceutical market under national health insurance. J Health Econ 21:471–483

    Article  Google Scholar 

  • Lu JFR, Hsiao WC (2003) Does universal health insurance make health care unaffordable? Lessons from Taiwan. Health Aff 22(3):77–88

    Article  Google Scholar 

  • Lundin D (2000) Moral hazard in physician prescription behavior. J Health Econ 19:639–662

    Article  Google Scholar 

  • McGuire T (2000) Chapter 09: Physician agency. In: Culyer J, Newhouse JP (eds) Handbook of health economics, vol 1, Elsevier, pp. 461–536

  • Meier B, Thomas K (2012) Insurers pay big markups as physicians dispense drugs. N Y Times. https://nyti.ms/2m4oPxz

  • Ministry of Health and Welfare (2015) Statistics of medical care. National Health Insurance 2014. https://dep.mohw.gov.tw/DOS/lp-1942-113.html (in Chinese). Accessed 20 May 2016

  • Mundlak Y (1978) On the pooling of time series and cross section data. Econometrica 46:69–85

    Article  Google Scholar 

  • O’Malley AJ, Frank RG, Kaddis A, Rothenberg BM, McNeil BJ (2006) Impact of alternative interventions on changes in generic dispensing rates. Health Serv Res 41:1876–1894

    Article  Google Scholar 

  • Pauly MV (1968) The economics of moral hazard: comment. Econ Rev 58:531–537

    Google Scholar 

  • Reiffen D, Ward MR (2005) Generic drug industry dynamics. Rev Econ Stat 87(1):37–49

    Article  Google Scholar 

  • Shiu JL, Chou YH, Tang MC (2018) Generic pharmaceutical adoption with mismeasured physician agency. Working paper

  • Sloan FA, Hsieh CR (2012) Health economics. MIT Press, Cambridge

    Google Scholar 

  • Whoriskey P, Keating D (2013) An effective eye drug is available for \({\$}50\). But many physicians choose a \({\$}2,000\) alternative. The Washington Post. http://wapo.st/1gghjqH?tid=ss_tw&utm_term=.168e783230c2

  • Wooldridge J (2010) Econometric analysis of cross section and panel data. MIT Press, Cambridge

    Google Scholar 

Download references

Author information

Authors and Affiliations

  1. Department of Economics, National Chung Cheng University, 168 University Road, Min-Hsiung, Chiayi, 62102, Taiwan

    Meng-Chi Tang

  2. Chunghwa Post Co., Ltd., Taichung, Taiwan

    Yi-Nong Wu

Authors
  1. Meng-Chi Tang
    View author publications

    You can also search for this author in PubMed Google Scholar

  2. Yi-Nong Wu
    View author publications

    You can also search for this author in PubMed Google Scholar

Corresponding author

Correspondence to Meng-Chi Tang.

Additional information

Publisher's Note

Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.

We thank Charlie Chern, Ya-Ming Liu, Ji-Liang Shiu, and the seminar participants at the iHEA 9th World Congress and Feng Chia University for providing various comments on and suggestions to an early draft of this paper. Ya-Chi Wang, Jei-Ying Wei, and Yu-An Tsao provided research assistance for this paper. We are indebted to Heng-Sim Lee, Head of Pharmacy Division of the Chiayi Branch, Taichung Veterans General Hospital for providing institutional details on various aspects of generic pharmaceutical adoption and the institutional details of hospitals and clinics in Taiwan. The financial support from the National Science Council, Taiwan, is greatly appreciated (NSC101-2410-H-194-018-). This study is based in part on data from the National Health Insurance Research Database provided by the Bureau of National Health Insurance, Department of Health and managed by National Health Research Institutes. The interpretation and conclusions contained herein do not represent those of Bureau of National Health Insurance, Department of Health or National Health Research Institutes. All errors are our own.

Appendix

Appendix

This appendix provides details regarding how the data were assembled using the sampling data from the NHIRD, and details on the OHAs included in the sample. Although these data are entangled and a considerable effort is required to organize the variables, previous studies using this database have rarely provided instructions on how their samples were assembled.Footnote 24 This appendix provides necessary details for researchers interested in using this database for follow-up studies. For more details on the NHIRD, please see http://nhird.nhri.org.tw/en/index.htm.

1.1 A1: Data management

The two basic components of the NHIRD are registration files and claims data. The registration files consist of information that does not vary within a year, including information on drugs (DRUG), medical personnel (PER) and their certification information (DOC), and contracted medical facilities (HOSB).Footnote 25 In particular, the DRUG registry provides necessary information on the drugs used in this study, which is discussed in the following section. The PER registry contains information on physicians, such as age (BIRTHDAY), sex (PRSN\(\_\)SEX), and whether they own a hospital or clinic (WORK\(\_\)STATUS). Physician’s experience is calculated using the certified date of their certification (INIT\(\_\)DATE) in the DOC registry. In addition, the HOSB registry contains hospital information such as length of operation (HOSP\(\_\)OPEN\(\_\)DATE), accreditation level (HOSP\(\_\)CONT\(\_\)TYPE), location (AREA\(\_\)NO\(\_\)H), and whether the hospital is publicly owned (PUBLIC). This information is summarized in the Appendix Table 6.

The NHIRD claims data comprise each claim that hospitals and clinics file for reimbursement. Diagnostic and patient information are recorded in the data on ambulatory care expenditures by visit (CD), which provides the patient’s age (ID\(\_\)BIRTHDAY), sex (ID\(\_\)SEX), and the code of the disease (ACODE\(\_\)ICD9\(\_\)1, ACODE\(\_\)ICD9\(\_\)2, ACODE\(\_\)ICD9\(\_\)3). The disease code was used to identify patients with diabetes in this study. The CD data also consist of related expense of each visit, including the amount that hospitals and clinics filed for reimbursement (T\(\_\)APPL\(\_\)AMT) and the expense (T\(\_\)AMT). The difference between the two values equals the patient’s deductibles (PART\(\_\)AMT). In addition, information on treatments and medications for each visit are recorded in the data on ambulatory care orders (OO). As indicated in Sect. 3, a patient can receive multiple treatments in each visit. Six variables in the CD and OO files were applied as identifiers to merge the two data sets: FEE_YM, APPL_TYPE, HOSP_ID, APPL_DATE, CASE_TYPE, and SEQ_NO. The OO data indicates which data were used in each visit, linking the CD and DRUG data sets. This information is also summarized in Appendix Table 6.

Table 6 Variable sources in the NHIRD
Full size table

1.2 A2: Drug information

The DRUG registry in the NHIRD provides necessary information on the drugs used in this study. In particular, the ingredient of a drug (DRCON\(\_\)NAME) and the identification code of this ingredient (DRUG\(\_\)ITEM) were used to identify the OHAs prescribed in the sample. The identification code of a drug (DRUG\(\_\)ID) was used to identify its version (brand name or generic), whereas the name of the provider of a drug (DRGIST\(\_\)NAME) was used to calculate the number of competitors in the market for a given ingredient. In addition, the price (DRUG\(\_\)PRICE) and quantity (DRCON\(\_\)QTY) of a drug were used to calculate the drug price per defined daily doses (DDD) according to the definition from the WHO Collaborating Centre for Drug Statistics (WHOCC, http://www.whocc.no/atc_ddd_index). This information is also summarized in Appendix Table 6, and the DDD and unit of each drug obtained from the WHOCC is summarized in Appendix Table 7. As described in Footnote 15 in the main text, information on DDD was crucial in this study because drug price was defined as the price required for the DDD of a drug, which enables the price among various drugs to be compared.

Table 7 Types available, number of providers, and number of prescriptions of oral hypoglycemic agents in the Taiwanese market in 2003, 2004–2005 and 2006–2007
Full size table
Table 8 Generic entry and prescription frequency by hospital accreditation level
Full size table

Appendix Table 7 lists the recorded information on the available OHAs in the DRUG registry during the sample period, including the number of providers and number prescribed according to drug and year. Five categories of OHAs with twelve different ingredients are recorded in the DRUG registry, and their ATC codes were obtained from the WHOCC. Because the NHIRD does not provide the information on whether the drug prescribed is a generic or brand-name drug, this study used the drug identifier (DRUG\(\_\)ID) in the DRUG registry to link with the drug query system maintained by the National Health Insurance Administration (NHIA) for this information (http://www.nhi.gov.tw/Query/query1.aspx?menu=18&menu_id=703, in Chinese). In addition, the name of a drug provider (DRGIST\(\_\)NAME) in the DRUG registry was applied to calculate the number of firms that provide drugs with the same ingredient in the market. Some of the provider information was missing in the 2003 and 2007 DRUG registry, and was recovered using the same NHIA website.

As shown in Appendix Table 7, drugs with certain ingredients were widely prescribed in Taiwan, and the prescriptions varied between the brand-name and generic versions. For example, drugs with metformin were prescribed 5,602 times in the data, and 66% of these drugs were generic. These drugs were prescribed frequently because they are often used as the drug for patients with diabetes in the early stage. By contrast, we excluded drugs with ingredients without generic alternatives, or for which the generic version was not prescribed in a given year. For example, glimepiride and acarbose were not included in the 2003 sample because no generic version was available during that year. Although the generic versions of repaglinide and nateglinide were available in the 2004 sample, they were not included in that year because no generic prescription was recorded in the sample. Consequently, glimepiride and acarbose were not included in the 2003 sample; repaglinide and nateglinide were included only in the 2005–2007 sample; and pioglitazone was included only in the 2006–2007 sample. Appendix Table 8 further separates these prescription frequencies by hospital accreditation level.

Rights and permissions

Reprints and permissions

About this article

Check for updates. Verify currency and authenticity via CrossMark

Cite this article

Tang, MC., Wu, YN. Medical providers as double agents in a universal health care system: evidence from generic pharmaceutical adoption in Taiwan. Empir Econ 59, 169–203 (2020). https://doi.org/10.1007/s00181-019-01674-9

Download citation

  • Received:

  • Accepted:

  • Published:

  • Issue Date:

  • DOI: https://doi.org/10.1007/s00181-019-01674-9

Keywords

JEL Classification

Search

Navigation