Revisiting moral hazard in the market for prescription drugs: Evidence from Finland

Transcription

1 Revisiting moral hazard in the market for prescription drugs: Evidence from Finland Ismo Linnosmaa a, Marisa Miraldo b, Matteo Galizzi c, Joni Hokkanen a a CHESS National Institute for Health and Welfare b Imperial College London c London School of Economics 08/07/2013 Ismo Linnosmaa / CHESS 1

2 Outline of the presentation I. Background and objectives II. Moral hazard in the market for prescription drugs III. Previous literature IV. Finnish pharmaceutical market V. Data VI. Empirical model VII. Results VIII. Next and conclusions 08/07/2013 Ismo Linnosmaa / CHESS 2

3 I Background and objectives A typical pharmaceutical market consists of a branded product and generic products Policy makers consider generic substitution (GS) as an important mechanism to contain expenditure growth Direct cost savings realize, if generic pharmaceuticals were used instead of branded products Indirect cost savings may arise, if substitution leads to price competition in the pharmaceutical market Generic substitution policy obligates pharmacists to substitute (or propose a substitution) a prescribed product with the cheapest available generic product in the market 08/07/2013 Ismo Linnosmaa / CHESS 3

4 I Background and objectives Despite policy initiatives, there are various incentives in health care that may distort decision-making and reduce savings Relative profits that pharmacies earn from the branded and generic products may affect generic substitution at pharmacies (Brekke et al., 2013) Health insurance may affect patients and physicians willingness to use and prescribe generic products (e.g. Lundin, 2000) Detailing (advertising targeted to physicians) may influence the physicians willingness to prescribe generic products Focus here on health insurance and the physician s choice between generic and branded products 08/07/2013 Ismo Linnosmaa / CHESS 4

5 I Background and objectives Ex post moral hazard (Pauly,1968): health insurance reduces out-of-pocket payments and increases the consumption of health care beyond the efficient level Ex post moral hazard in the market for prescription drugs (Hellerstein, 1998): If the physician is sufficiently concerned for the welfare of her patients, health insurance increases the physician s propensity to prescribe the branded version of a drug Besides Hellerstein (1989) and Lundin (2000, JHE), empirical work in this area is limited We examine the existence of moral hazard using data from the Finnish pharmaceutical market 08/07/2013 Ismo Linnosmaa / CHESS 5

6 II Moral hazard in the prescription drug market Hellerstein (1998, RJE): A physician chooses between the branded, b, and generic, g, versions of a drug acts as a double agent for the patient and insurance Random utility of the physician from product t = b,g is where U t = v t - c t = y + γ 1 [q t - (1-r)p t ] - γ 2 rp t - c t y = physician labor income (independent of t, ie. no P4P) q t = quality of product t p t = price of product t r in [0,1] = insurance coverage; (1-r) = coinsurance c t is the cost of prescribing version t of the drug 08/07/2013 Ismo Linnosmaa / CHESS 6

7 II Moral hazard in the prescription drug market Physician prescribes the generic version of the drug if v g c g > v b c b (or c b > v + c g ) Assume c t are iid with density and distribution functions f(c) and F(c). Then Pr( t g) [1 F( v cg )] f ( cg ) dcg where Δv = v b - v g = γ 1 Δq - γ 1 (1 - r)δp - γ 2 rδp and Δq = q b - q g and Δp = p b - p g are quality and price differences between the branded and generic drugs It can be assumed that Δq 0 08/07/2013 Ismo Linnosmaa / CHESS 7

8 II Moral hazard in the prescription drug market How does the probability to prescribe the generic product responds to a change in the insurance coverage? Let p b > p g. Then Pr( t g) r p f ( v c ) f ( c ) dc 1 2 g g g The effect may be positive or negative depending the values of γ 1 and γ 2 Proposition If γ 1 > γ 2, there is moral hazard in insurance (and dpr(t = b) > 0, if dr > 0) Empirical work: γ 1 > γ 2 or perhaps γ 1 γ 2? 08/07/2013 Ismo Linnosmaa / CHESS 8

9 The effect of dr = r 1 -r 0 > 0, when p > 0. c b c b = (v b -v g ) 1 +c g Pr(t = g) (v b -v g ) 1 c b = (v b v g ) 0 + c g (v b -v g ) 0 c g If γ 1 > γ 2, then d(v b -v g ) = p(γ 1 -γ 2 )dr > 0, and the probability of prescribing the generic (branded) version of the drug decreases (increases) 08/07/2013 Ismo Linnosmaa / CHESS 9

10 III Previous literature Hellerstein (1998, RJE) examined the impact of insurance on the probability that physicians prescribe generic versions of a drug used 1989 National Ambulatory Medical Care Survey data on patient visits to 1223 office-based physicians found that unobserved physician-specific factors are important in influencing the prescription choice no evidence of moral hazard (has no access to price variables) 08/07/2013 Ismo Linnosmaa / CHESS 10

11 III Previous literature Lundin (2000, JHE): examined prescription and consumption habits and ex post moral hazard in the market for prescription drugs used data on seven drugs for the years 1992 and 1993, collected from two pharmacies in Sweden. found evidence on ex post moral hazard in insurance, ie. γ 1 > γ 2 08/07/2013 Ismo Linnosmaa / CHESS 11

12 IV Finnish pharmaceutical market Wholesale and retail prices: Wholesale prices of reimbursed pharmaceuticals (77% of all outpatient drugs) are regulated by price caps Retail prices = wholesale price + profit margin + VAT Profit margins of pharmacies are regulated by law. No (retail) price competition between pharmacies 08/07/2013 Ismo Linnosmaa / CHESS 12

13 IV Finnish pharmaceutical market Insurance reimbursement in social insurance based on three classes: Basic refund category: 35% of the retail price Lower special refund category: 65% of the retail price Higher special refund category: 100% of the retail price (nominal 3 cost per prescription for patients) Upper limit of out-of-pocket expenses is 700,92, after which no expenses for patients (nominal 1,5 cost per prescription for patients) 08/07/2013 Ismo Linnosmaa / CHESS 13

14 IV Generic substitution in Finland GS policy in April 2003 Pharmacists obliged to propose the substitution of the prescribed product with the cheapest, or close to the cheapest, substitutable product Substitutable products must contain the same active substance in equal amounts have the same form be biologically equivalent. Substitution can be forbidden by the physician or the patient 08/07/2013 Ismo Linnosmaa / CHESS 14

15 IV Price corridor in GS Products A, B, C and D. If the prescribed pharmaceutical is B, the pharmacist must offer A or D to the patient. If the patient requires the cheapest, A will be dispensed. price p B p C p A + 2, if p A < 40 p A + 3, if p A > 40 p D p A p A 08/07/2013 Ismo Linnosmaa / CHESS 15

16 IV Reference pricing (RP) in Finland Generic RP was implemented in April 2009 Clustering of pharmaceuticals based on the classification of substitutable products created in GS Insurance reimbursement based on the reference price, which is the lowest, or close to the lowest, price in a group of substitutable products Reference price p m = lowest price in a group + 1,5/ 2 (observe the change from the GS policy) If p p m, then the insurance reimburses rp and the patient pays (1-r)p If p > p m, the insurance reimburses rp m and the patient pays (1-r)p m + (p-p m ) 08/07/2013 Ismo Linnosmaa / CHESS 16

17 V Data Data from KELA, Finnish social insurance institution Data on dispensed pharmaceutical prescriptions from Finnish pharmacies in 2005 ATCs: C10: Lipid modifying agents N06A: Antidepressants B03XA01 (Epoetin) and L03AA03 (Filgrastim): Biological drugs N = prescriptions 08/07/2013 Ismo Linnosmaa / CHESS 17

18 V Data Data contains information on The pharmaceutical product: atc, producer, name, strength, package size, ddds and nro of packages bought, prescribed product (it substitution at the pharmacy), total cost and insurance reimbursement The patient: age, gender, income, year of death, hospital district, social security benefits etc. The physician: insurance code and field of speciality 08/07/2013 Ismo Linnosmaa / CHESS 18

19 V Data exclusions Additional data on the versions, ie. branded or generic, of products from Finnish Agency of Medicines, FIMEA Best available information on products in C10 antidepressants and biological drugs excluded Computation of alternative prices ready for simvastatins at the moment other drugs than simvastatins in C10 excluded Currently, we focus on simvastatin prescriptions not subject to generic substitution at the pharmacies N = /07/2013 Ismo Linnosmaa / CHESS 19

20 V Price difference Δp Price difference Δp = p b p g not observed in data Price of the prescribed product observable Price of the alternative product not observable, but needs to be estimated Principles: Alternative price to a generic product is the (weighted average) price of the branded product with the same strength and package size Alternative price to a branded product is the (weighted average) price of the mostly prescribed generic product with the same strength and package size (Lundin, 2000) 08/07/2013 Ismo Linnosmaa / CHESS 20

21 V Price difference p: Examples For example, Zocor 10MG 98 tablets were prescribed at Average price per package: 31.5 Average out-of-pocket price per package: 16.2 (51.2%) Average insurance expenditure per package: 15.3 (48.8%) Physicians could have prescribed Simvastatin Ratiopharm 10MG 98 tablets at prices (Weighted) average price per package: 14.0 Average out-of-pocket price per package: 7.2 (51.4%) Average insurance expenditure per package: 6.8 (48.6%) 08/07/2013 Ismo Linnosmaa / CHESS 21

22 V Price difference Δp: Examples Simvastatin Actavis 20MG 98 tablets were prescribed at Average price per package: 13.7 Average out-of-pocket price per package: 9.9 (72.3 %) Average insurance expenditure per package: 3.8 (27.7%) If physicians prescribed Zocor 20MG 98 tablets, the price and cost sharing would have been (Weighted) average price per package: 49.2 Average out-of-pocket price per package: 35.7 (72.6%) Average insurance expenditure per package: 13.5 (27.4%) 08/07/2013 Ismo Linnosmaa / CHESS 22

23 V Descriptive statistics Variable Definition Mean Std. Dev. Min Max Genericv D = 1, if generic version 0,82 0, Pricediff Price difference Δp 23,87 10,218-93,38 100,39 PricediffPAT Share of Δp paid by patient 14,72 7,078-68,09 75,97 PricediffINS Reimbursed share of Δp 9,15 6,720-38,64 70,96 Gender D = 1, if female 0,54 0, Age2005 Age at ,2 10, YoD2005 D = 1, if dies at ,006 0, Tincome Taxable income, , , ,30 Remark: In 22 prescriptions Δp < 0 08/07/2013 Ismo Linnosmaa / CHESS 23

24 VI Empirical model Outcome of interest: physician j s prescription of generic version of the drug (simvastatin) to patient i Model (Lundin, 2000) where Prob(y ji = g) = F(α + γ 1 (1-r i ) p + γ 2 r i p + ζx i ) x i = (gender i, age2005 i,yod2005 i,income i ) Model specifications Logit model Probit model The model is estimated with pooled data with no patientor physician-specific effects (so far) 08/07/2013 Ismo Linnosmaa / CHESS 24

25 VII Results from the logit model Dep. variable: Genericv Model 1 Model 2 Variables Coef Std. Err Coef Std. Err PricediffPAT 0.141*** *** PricediffINS *** *** Gender *** Age *** YoD *** Income *** Constant 0.469*** *** N LR: model significance *** *** Pseudo R Wald: γ 1 = γ *** *** *** p < /07/2013 Ismo Linnosmaa / CHESS 25

26 VII Results from the probit model Dep. variable: Genericv Model 1 Model 2 Variables Coef Std. Err Coef Std. Err PricediffPAT 0.081*** *** PricediffINS *** *** Gender *** Age *** YoD *** Income *** Constant 0.328*** *** N LR: model significance *** *** Pseudo R Wald: γ 1 = γ *** *** *** p < In absolute value, parameters are smaller than in the logit model 08/07/2013 Ismo Linnosmaa / CHESS 26

27 VIII Tentative conclusions Cost allocation between patients and insurance affects the prescription decisions of the physicians The generic (branded) version is prescribed more likely The higher is the share of the price difference p paid by the patient (insurance) Empirical support for moral hazard in insurance: γ 1 > γ 2 and physicians weigh the patients benefits from health insurance more than insurance expenditure Results similar to Lundin (2000) for the year /07/2013 Ismo Linnosmaa / CHESS 27

28 VIII Next More data more years ( available) more drugs More econometric modeling physician-specific effects patient-specific effects drug-specific effects 08/07/2013 Ismo Linnosmaa / CHESS 28

29 Thanks for your attention! 08/07/2013 Ismo Linnosmaa / CHESS 29