Price controls for prescription drugs are once again a timely policy issue in the United States (US). The latest incarnation revolves around Medicare Part D’s (Part D) reliance on private insurance plans to provide services and negotiate drug prices, which some argue generates higher costs for Medicare and Medicare participants (Frakt et al. 2012). However, research suggests that prescription drug price controls can dampen innovation and adversely affect population health (U.S. Department of Commerce International Trade Administration 2004; Golec and Vernon 2006; Lakdawalla et al. 2009), and access to new therapies is worse in countries with the strictest price controls (Lambrelli and O’Donnell 2011; Tuna et al. 2014). Furthermore, much of the recent rhetoric focuses on the potential short-term savings, with less attention focused on long-term consequences. In this paper, we examine the health and spending implications of government price setting in Part D. We consider hypothetical policies that would impose pricing and formulary designs similar to those currently used by the U.S. Department of Veterans Affairs (VA), which pays some of the lowest prices in the country due to established price controls and the use of a single prescription drug formulary. The results highlight the trade-offs between cost-savings and population health.
Part D is Medicare’s outpatient prescription drug program. Under Part D’s non-interference clause, Medicare cannot interfere in price negotiations (Pitts and Goldberg 2015). Part D was designed to be flexible and market driven so that no single insurance plan, formulary, or pricing system dominates (Coulam et al. 2011). Private insurance firms offer Part D Plans (PDPs) with their own coverage agreements and price negotiations with manufacturers (Jacobson et al. 2007; Kaiser Family Foundation 2016).
Although Part D PDPs can design their own formularies within certain Centers for Medicare & Medicaid Services’ (CMS) constraints, they must follow CMS’ rules for deductibles and out-of-pocket limits (Jacobson et al. 2007). In 2016, nearly 900 PDP options were available nationwide with at least 19 plans available in each region (Kaiser Family Foundation 2013). PDPs compete on coverage and costs to attract enrollment, allowing beneficiaries to select a plan that meets their individual health and financial needs.
In 2014, Part D made up less than 11 percent of the $612.7 billion1 spent by Medicare (Congressional Budget Office 2015). Studies show that Part D encourages use of generics (The Board of Trustees of the Federal Hospital Insurance and Federal Supplementary Insurance Trust Funds 2015), keeps patient premiums low, reduces spending on non-prescription Medicare services, lowers Medicare hospital admissions, and decreases cost-related medication non-adherence and non-persistence (Dismuke and Egede 2013).
Juxtaposed against this apparent success is the concern that drug prices under Part D unfairly burden Medicare participants (Gagnon and Wolfe 2015). In response, policymakers have proposed that Medicare regulate Part D drug prices (Kaiser Family Foundation 2016). A number of studies have estimated that price control policies could save Medicare $16.7–$58.3 billion (Gellad et al. 2008; Frakt et al. 2012; Baker 2013; Gagnon and Wolfe 2015). For example, Frakt et al. (2012) estimated that if Part D were able to reduce prices by 40 percent by adopting VA-style formularies, Medicare would save $16.7 billion. Frakt et al. computed these savings by assuming a 40 percent reduction in 2009 total costs per Part D enrollee times the number of Part D enrollees in 2009 (27 million). Baker (2013) estimated savings at $24.8–$58.3 billion. Baker computed these savings by assuming that Medicare would negotiate prices to levels paid in Canada (20 percent below US prices), at the lower extreme and prices paid by Denmark (65 percent below US prices) at the upper extreme.
We build on the work by Frakt et al. (2012) and Baker (2013) by considering the possible adverse consequences of price controls on Part D prescription drugs, and by examining long-term effects of price controls on innovation and hence on health.
Prior research demonstrates that price controls in the pharmaceutical industry reduce investment in research and development (R&D) (U.S. Department of Commerce International Trade Administration 2004). In 2004, the US Department of Commerce (DOC) found that direct and indirect price regulations in 11 countries decreased pharmaceutical revenues, which was associated with a $8–$12 billion reduction in R&D investments (U.S. Department of Commerce International Trade Administration 2004). The DOC also estimated that pharmaceutical innovation would yield three to four additional new drugs per year in the absence of price controls.
Eger and Mahlich (2014) found that companies with a higher European sales share had lower R&D investment. Vernon (2005) showed that if the US pharmaceutical industry faced price controls such that revenues were similar to the rest of the world, R&D investment would be reduced by 23–33 percent. Abbott and Vernon (2007) found that price controls that reduce expected future pharmaceutical prices by 40–50 percent decreased R&D projects by 30–60 percent.
Reduced pharmaceutical R&D has been linked to reduced innovation, frequently measured as the number and timing of drug launches (Danzon and Ketcham 2004; U.S. Department of Commerce International Trade Administration 2004; Vernon 2005; Giaccotto et al. 2005; Abbott and Vernon 2007; Filson 2012; Grossmann 2013; Cockburn et al. 2014; Eger and Mahlich 2014). Filson (2012) showed that adopting price controls the flow of new drugs by almost 40 percent. And, ultimately, innovation has consequences for population health and social welfare. Murphy and Topel (2003) found that improvements in life expectancy due to medical research added approximately $5.1 trillion per year to US wealth from 1970–1998. In another study, they estimated that increased life expectancy between 1970–2000 added $4.6 trillion per year to national wealth (Murphy and Topel 2006).
Lakdawalla et al. (2009) found that US price controls would reduce life expectancy of Americans aged 55–59 by 2.8 percent by 2060. The authors measured these impacts using The Health Economics Medical Innovation Simulation (THEMIS), a microsimulation model that tracks US individuals aged 55 and older and projects their health status and economic outcomes. THEMIS is also used in this study.
2.1 VA-Style Price Controls
2.1.1 VA-Style Pricing
The VA prescription drug program is often cited as a model for Part D reform (Jacobson et al. 2007; Gellad et al. 2008; Frakt et al. 2012; Gagnon and Wolfe 2015), despite distinct differences between the VA health program and Medicare. The VA serves a significantly smaller and demographically different population. The VA drug benefit features a single national formulary, and Stroupe et al. (2013) found that approximately 54 percent of veterans had both VA and non-VA drug coverage (e.g. private insurance, Medicare, or Medicaid). In contrast, Part D participants rely primarily on Medicare for drug coverage.
The VA program requires pharmaceutical manufacturers to sell at the lower of the Federal Supply Schedule (FSS) price or the Federal Ceiling Price (FCP). Further, the VA gains concessions from pharmaceutical manufacturers by restricting access through the use of a closed formulary (U.S. Department of Veterans Affairs 2015), often awarded to a limited number of pharmaceutical manufacturers (Congressional Budget Office 2005; D’Angelo 2007). This contrasts with Part D’s flexible, market-driven structure with multiple plan options among which beneficiaries can choose (Coulam et al. 2011).
In this paper, we focus on the incentives to innovate in the pharmaceutical industry resulting from hypothetical government price controls imposed on Part D, using VA-style pricing as a proxy. We develop illustrative policy scenarios with varying levels of price control impacts and model long-run consequences using THEMIS.
We simulate the impact of changes in innovation incentives due to price controls in Part D on producer output and the health of current and future generations of patients. THEMIS projects the effects of pharmaceutical regulations on health status and economic outcomes for Americans aged 55 and older (Lakdawalla et al. 2009; Eber et al. 2015). THEMIS has been used in prior studies to examine the innovation impacts of increased clinical evidence generation associated with comparative effectiveness research (Eber et al. 2015), the impact of lifting European price controls on the welfare of current and future generations (Sood et al. 2009), and whether policies that extend data exclusivity would leave the US better off (Goldman et al. 2011). The mechanics of the model have been described in detail previously, and additional details are provided in the Technical Appendix (Lakdawalla et al. 2009; Eber et al. 2015).
THEMIS tracks the evolution of future health and innovation under different policy regimes using its population health and innovation modules. The population health module was built from the Health and Retirement Survey, a nationally representative longitudinal dataset (Poterba et al. 2007). The population module evaluates dynamic transitions that link present health and innovation to future values, and estimates medical and pharmaceutical expenditures and population life-years (Eber et al. 2015). Expenditures are modeled based on data from the Medical Expenditure Panel Study and the Medicare Current Beneficiary Survey. The innovation module links market size for pharmaceutical drugs to future drug introductions via the elasticity of innovation, i.e., the rate at which changes in revenues affect the number of new drugs brought to market in a given period.
Previous research found that a 1 percent increase in pharmaceutical market share leads to a 3–4 percent increase in the annual number of new drugs (Acemoglu and Linn 2004). In THEMIS, we use market size as a proxy for revenue, i.e., we assume that revenues for a broad class of drugs (e.g. anti-diabetics) are proportional to the size of the market for that class (e.g. number of diabetes patients). Thus a reduction in revenue for a class will result in fewer drug innovations in that class. We assume an innovation elasticity of 3.0 for small-molecule drugs based on Acemoglu and Linn (2004). For biologics, we estimate the innovation elasticity at 4.7.
To estimate the innovation elasticity for biologics, we start with the estimate of 3.0 for small molecules, and use the number of small molecule drugs launched in 2013 and the annual expenditures on small molecule drugs in 2010 to calculate the annual revenue needed to generate one additional small molecule drug launch per year (Acemoglu and Linn 2004; Pharmaceutical Research and Manufacturers of America 2012; U.S. Food and Drug Adminstration 2014). We then use the ratio of the expected returns on small molecule drug investments versus biologic drug investments multiplied by the fully capitalized costs of discovering small molecule and biologic drugs to derive the amount of annual revenue needed to generate one additional biologic drug launch per year. We are then able to estimate an elasticity of innovation for biologic drugs of 4.7 using the number of biologic drugs launched annually and the annual expenditures on biologic drugs (Pharmaceutical Research and Manufacturers of America 2012; Serebrov 2014). We assume that changes in net producer revenues affect new drug interactions after an 8-year development lag (DiMasi and Grabowski 2007).
New treatment development has significant consequences for future populations. Novel drug introductions may directly enhance the health of future generations by improving the efficacy of treatments over existing therapies or by broadening the population receiving treatment. For example, improved tolerability of a drug may expand access by making it available to patients who could not tolerate older treatments. As described in the Technical Appendix, we model the benefits of new drug introductions based on empirical evidence on how new drug introductions have historically improved efficacy or broadened access. The model also allows for new drug introductions not providing additional benefits or expanding access to treatment.
Innovative therapies may also influence treatment costs for future generations. While rapid introductions of new brand-name drugs might lead to higher treatment costs, rapid introductions of more effective treatments may reduce total medical costs. In turn, this benefit may be offset by additional health care costs incurred during added life-years.
We report the effects of alternative policy scenarios on innovation and spending, measured as the number of drugs introduced, total medical expenditures, combined impact of access and clinical effects on life expectancy, and discounted values of expenditures and added life expectancy.
We monetize changes in health by multiplying the predicted additional years by the value of a statistical life-year, which captures the value individuals place on reducing mortality risk. The value of a statistical life is often inferred from an individual’s willingness to take, for example, riskier but higher paying jobs. Based on prior studies (Viscusi and Aldy 2003; Aldy and Viscusi 2008; Sood et al. 2009), we assume a value of $200,000 for each statistical life-year.
3.2 Policy Scenarios
Gagnon and Wolfe (2015) found that Medicare could save up to $18.4 billion by imposing price controls on Part D reflecting the prices secured by Medicaid or the VA on brand-name drugs. Baker (2013) estimated that introducing a policy scenario involving price caps to Part D would result in annual savings of $24.8–$58.3 billion to the federal government. Gellad et al. (2008) estimated potential annual savings of $29.1 billion if FSS prices were used by Part D. Frakt et al. (2012) estimated savings of $16.7 billion per year if Medicare obtained the same drug prices as the VA.
In Table 1 we summarize the reported implied impacts on total Medicare spending and total US drug spending (market size). Based on these reported effects, we develop policy scenarios characterized with two key model inputs: the “revenue effect” and the “access effect.”2
In our model, revenue effects measure how a new policy influences pharmaceutical industry revenue, which affects incentives for innovation, which in turn impacts new drug introductions. We simulate how changes in total US pharmaceutical revenue alter the number of new drugs brought to market (innovation) using VA-style pricing as a proxy for a price control policy in Part D.
Access effects measure reduced availability of treatments as a result of stricter formulary designs imposed by hypothetical VA-style pricing in Part D, measured as the percentage of people in six disease categories who get treatment in a particular period.
The model assumes that changes in the access effect impact everyone who has access to treatments specific to a particular condition. In other words, undiagnosed individuals gain nothing from the introduction of a new drug. Individuals who are diagnosed with a condition and treated with an existing drug experience the “clinical effect” of the new drug. Diagnosed but untreated individuals also experience the access effect of the new drug, since a fraction of them may gain access to the new therapy.
3.3 Revenue Effect and Access Effect
3.3.1 Revenue Effect
We focus on the estimates provided by Baker (2013) of $24.8–$58.3 billion in savings to Medicare if price controls were imposed on Part D, a 4–9 percent per-year reduction in Medicare spending, based on 2014 Medicare spending of $612.7 billion (Congressional Budget Office 2015). Since pharmaceutical industry revenues were $383.9 billion in 2014 (IMS Institute for Healthcare Informatics 2015), this translates to a reduction in pharmaceutical revenues of 7–15 percent. Assuming demand for drugs remains constant, Medicare savings implied by Baker (2013) would result in a reduction in Part D drug prices of 37–86 percent.
Since we define the revenue effect as the reduction in annual pharmaceutical industry revenue for 2014 due to the introduced policy change, we obtain the revenue effect by dividing annual cost savings to Medicare by total pharmaceutical industry revenues in 2014, estimated at $383.9 billion (IMS Institute for Healthcare Informatics 2015).
3.3.2 Access Effect
The VA extracts additional discounts by implementing a more restricted formulary (D’Angelo 2007). Thus, we assume Part D would have to similarly restrict access to treatment (Jacobson et al. 2007), which includes potentially excluding expensive novel therapies.
Using data comparing the availability of drugs in Medicare PDPs relative to those available on the VA formulary (Avalere Health 2016), we estimate the difference in coverage between these two systems, then approximate how VA-style pricing in Part D would restrict treatment access. We find that on average, 14 percent of the top 200 Part D drugs are not included on the VA formulary (Avalere Health 2016). Therefore, in our baseline model, we analyze the impacts of price controls in Part D, assuming a 14 percent access effect (“high impact”). In addition, as a conservative alternative, we assume that the access effect is half this average and analyze the impacts of price controls assuming an access effect of 7 percent (“low impact”).
4.1 Low Impact and High Impact Scenarios
We quantify the impact of hypothetical VA-style pricing in Part D on innovation by measuring changes in new drug introductions from 2010–2060 relative to the status quo. We present the results of two scenarios: (1) low impact scenario, in which we assume a 7 percent revenue effect and a 7 percent access effect, and (2) high impact scenario, in which we assume a 15 percent revenue effect and a 14 percent access effect.
Figure 1 shows that, under the low impact scenario, VA-style pricing in Part D is associated with a 6.1 percent reduction in new drug introductions in 2030 and a 11.3 percent reduction in 2060. Under the high impact scenario, the effect on innovation is more substantial, from a 14.9 percent reduction in new drug introductions in 2030 to a 24.9 percent reduction in 2060.
A motivation for VA-style pricing in Part D is to reduce spending on prescription drugs (Frakt et al. 2012). Figure 2 reports the change in lifetime per capita costs of drugs and non-drug care under VA-style pricing relative to the status quo for three cohorts: individuals born in 1971–1975, 1981–1985, and 1991–1995.
For the oldest cohort, we observe a reduction in per capita total costs of 4.7 percent over their lifetimes under the high impact scenario relative to the status quo. For the youngest cohort, we observe a 6.6 percent reduction under the high impact scenario. For all scenarios and cohorts, the reductions in drug costs contribute to a significantly smaller share of total cost savings compared to non-drug costs of care. Relatively larger savings from reduced costs of care are driven by reduced longevity under the price control scenarios.
Figure 3 reports the observed trade-offs between cost savings and life expectancy for the low impact (panel A) and high impact scenarios (panel B). For the oldest cohort, under the low impact scenario, we estimate a lifetime total savings relative to the status quo of $0.1 trillion for the oldest cohort. However, individuals in that cohort lose an average of 0.4 life-years relative to the status quo. For the youngest cohort, under the low impact scenario, we estimate savings of $0.3 trillion, and this group loses 0.9 life-years. In the high impact scenario, for the oldest cohort, we compute lifetime total savings of $0.3 trillion, and for the youngest cohort, savings of $0.5 trillion. The oldest cohort loses 1.2 life-years and the youngest cohort loses nearly 2 life-years under the high impact scenario.
In Figure 4, we summarize the overall effects of VA-style pricing under the low impact (panel A) and high impact scenarios (panel B) relative to the status quo, and present the lifetime impacts for Americans born in 1949–2005. Overall savings are relatively small; $0.1 trillion in savings from lower drug prices and $0.6 trillion in savings from lower medical expenditures (panel A). The trade-off is a loss of health benefits valued at $6.4 trillion and the net effect is a loss to society of $5.7 trillion. In the high impact scenario, savings from lower drug prices total $0.3 trillion and savings from lower medical expenditures are $1.2 trillion. In this scenario, loss in health benefits is $14.8 trillion and net loss to society is over $13.3 trillion.
Our simulations show that VA-style pricing in Part D modestly reduced medical spending via the combined mechanism of increased per-capita, per-annum expenditures (fewer, less effective treatments leading to higher comorbidity burdens) and decreased lifetime medical expenditures (less effective treatment leading to shorter lifespans overall). These societal losses due to reduced lifespans are captured in the “lost health benefits” bar in Figure 4.
4.2 Model Limitations
Simulations such as THEMIS are valuable for assessing potential policy outcomes when data are not available, though they have limitations. Future policies and market conditions may change in unpredictable ways, and retrospective data may not characterize the future.
In order to isolate the effects of the policy in question, we make a number of simplifying assumptions. Our model characterizes the impact of imposing VA-style pricing in Part D as an aggregate impact on pharmaceutical industry sales. To the extent that these price controls disproportionately impact prescription drug innovations used by Part D enrollees, revenue effects may be understated. We also assume price impacts of VA-style pricing in Part D do not affect demand for prescription drugs.
We make a number of assumptions about access effects that allow us to capture the essence of the impact of hypothetical VA-style pricing on access to new drugs in Part D, but may not allow us to capture the nuances of the VA-pricing policy. First, with the data available, we are able to capture generic substitution but not therapeutic substitution. Accounting for therapeutic substitution would require a treatment by treatment assessment of substitutes. By making the assumption that there is no therapeutic substitution the baseline access effect may be high. Our model also assumes the VA-style formulary restricts drug access without taking into account effectiveness or relative clinical value compared to generics or alternatives. Here again, we may overestimate access effects. We assume that access effects persist throughout the simulation, i.e., the access effect does not decay over time. To help us understand the impacts of these assumptions we conduct sensitivity analyses of the access effects, including no access effects (zero percent).
We assume that changes in life expectancy depend only on the policy scenarios. Significant shifts in life expectancy due to other policies may have a large impact on our findings. We mitigate uncertainties in the underlying data and assumptions by presenting all results as differences relative to the status quo – the status quo and alternative policy scenarios are parameterized with the same data.
4.3 Sensitivity Analyses
To test the impact on our results of different access effect assumptions, we estimate total net benefits under access effects from 0–35 percent (see Figure 5). In one sensitivity analysis, we assume that VA-style pricing results in no access effects (zero percent). Even assuming no access effects, we observe a reduction in net benefits of $5.0 trillion in the low impact scenario and $8.7 trillion in the high impact scenario relative to the status quo. As we allow access effects to grow, net benefits to society decrease by $22.0 trillion for the high impact scenario, consistent with expectations, since greater access effects imply greater restrictions on access to treatment.
Another key assumption is the responsiveness of innovation to revenues, or the elasticity of innovation. We assume an elasticity of 3.0 for small-molecule drugs and 4.7 for biologic drugs. As a sensitivity analysis, we compute net benefits under the two price control scenarios with elasticities of innovation ranging from 2.0–5.0 (see Figure 6).
We find that the larger the innovation elasticity, the larger the negative impact on net benefits relative to the status quo. Loss in net benefits decrease when we assume a smaller elasticity. If the status quo innovation elasticity is higher, price controls have a larger negative impact on innovation.
Our analysis relies on total pharmaceutical sales to estimate market size. An alternative approach is to use pharmaceutical sales net of rebates, which would necessarily result in a larger revenue effect, which in turn would generate increasingly larger impacts on the value of health. Relying on total pharmaceutical sales overestimates market size, but provides a more conservative estimate of impacts of price controls.
In the low impact scenario, we estimate that VA-style pricing in Part D would result in a lifetime total loss to society of $5.7 trillion for Americans born 1949–2005. In the high impact scenario, this loss may be as high as $13.3 trillion. Most losses are due to poorer health. Imposing VA-style pricing in Part D decreases life expectancy by nearly 2 years, a significant loss considering that the advent of revascularization surgery for heart patients increased longevity by 1.1 years (Cutler 2007).
Loss in life expectancy is driven by decreased prescription drug innovation and reduced access to current treatments. Under the low impact scenario, our simulations predict a 5.7 percent reduction in new prescription drugs in 2030 and an 11.2 percent reduction in 2060. In the high impact scenario, we predict a 14.9 percent decrease in new drug innovations in 2030, and a 24.9 percent reduction in 2060. These findings should be viewed in light of the model limitations as discussed above.
Our analyses likely represent a lower bound on the loss of benefits to the value of health, as we have not considered the value Part D participants place on choice and flexibility in PDPs. Furthermore, our analysis does not account for the intangible benefits of longevity and higher quality of life – such as avoiding the stress of intensive medical procedures, spending time with grandchildren, or engaging in physical activities. These intangible benefits may not be possible without innovative prescription drugs that reduce side effects.
Overall, we find that the hypothetical VA-style pricing in Part D decrease total lifetime costs (including costs of care and prescription drugs) by $0.7–$1.5 trillion for Americans born 1949–2005. However, these gains come at a cost. As the pace of innovation slows, future generations of older Americans will have lower life expectancy relative to the status quo. Policies that achieve cost savings by imposing price controls and restricting access to treatment options create a significant loss to society. When health benefits are valued appropriately, society experiences a significant loss – $5.7–$13.3 trillion – driven by reduced incentives for the pharmaceutical industry to invest in life-saving and life-extending innovations. The findings suggest that policymakers concerned about drug prices should explore mechanisms that insulate patients from high out-of-pocket costs while preserving the incentive for innovation.
This research was supported by the Pharmaceutical Research and Manufacturers of America. Precision Health Economics (PHE) is a health economics consultancy providing services to the life sciences industry. Gigi Moreno and Jennifer Benner were emloyees of PHE at the time of the study. We would like to thank Brielan Smiechowski for research support.
Technical Appendix: The Health Economics Medical Innovation Simulation (THEMIS) Innovation Module/Global Pharmaceutical Policy Model (GPPM)
A1 General Structure of the Model
This Appendix describes The Health Economics Medical Innovation Simulation (THEMIS); a microsimulation model used to simulate the effect of pharmaceutical regulation on health for the population age 55 and over in the United States (US). THEMIS models and tracks the evolution of future health and innovation under different policy regimes, based on methods developed for its predecessor, the RAND future elderly model (FEM) (Goldman et al. 2004, 2005; Lakdawalla et al. 2005; Shekelle et al. 2005). THEMIS consists of a population health module and an innovation module; the latter entitled the global pharmaceutical policy model (GPPM). Each module is a set of dynamic interactions that link present health and innovation to their future values. For example, next year’s health states depend on today’s health states, and on a set of random health shocks that vary with individuals’ own risk-factors – e.g. demographics, health behaviors, and current disease conditions. The innovation module links this year’s stock of drugs to next year’s, by allowing sales and profits from pharmaceutical sales to affect future innovation. Figure A1 outlines the mechanics of the model.
In a given year, say 2024, sample individuals may have diseases and/or disabilities that put them at risk of contracting new diseases and disabilities, or even dying, in 2025. Moreover, new drugs are introduced in 2024 that reduce some of these risks. We estimate a health transition model to simulate how population health will look in 2025, given the number of new drug introductions and existing health conditions. Finally, mortality will have shrunk the population in 2025, but the sample is “refreshed” by introducing those who were 54 in 2024, and who now “age into” our target population. This forms the set of sample individuals for 2025. The same process is then repeated to obtain the population in 2026, 2027, and all subsequent years, until the final year of the simulation is reached.
Theoretically, the current rate of innovation depends on future sales, which measure the profitability of current research effort. We assume it takes 10 years – from pre-clinical trials to launch – for a pharmaceutical company to introduce a new drug. This is based on the mean duration observed in the data.3 Moreover, we assume that drug companies have rational expectations, in the sense that today’s sales are used as a forecast for sales 10 years in the future. Therefore, the number of new drugs today depends on sales (or market size) 10 years ago. The empirical economics literature provides us with an estimate of how innovation changes in response to changes in market size. For example, this elasticity is applied to the change in market size between 2014 and 2015 to estimate the change in the number of new drugs between 2024 and 2025.
Regulation affects market size. Results from Sood et al. (2008) suggest that price controls leads to a 22.5% reduction in market size. Hence, this is the lever we use to simulate the effect of price controls as well as the effect of introducing co-pay subsidy.
To measure cost and benefits across scenarios, we use life-years and medical expenditures. We translate life-years in dollar terms using a value of a statistical life estimate. For expenditures, we use cost regressions estimated using micro data.
The simulations are stochastic: the arrival of new drugs is random, as is the arrival of new diseases. We discuss how these arrival processes are implemented in the model. Furthermore, we also discuss how weights are used to construct nationally representative estimates.
This document presents details on each of these components. First, we explain how the transition model was estimated using the Health and Retirement Study (HRS) data. Second, we explain how we estimated the clinical effects of new drugs. We then discuss the process of innovation and how it relates to market size. Next we discuss how costs and benefits of alternative policy scenarios were calculated. Finally, we discuss how the stochastic components of the model are implemented and how we used sample weights throughout.
A2 The Health Transition Model
We use the HRS, a nationally representative longitudinal study of the U.S. population age 50+ as our main data source. We use the observed (reported) medical history of respondents to infer incidence rates as a function of prevailing health conditions, age and other socio-demographic characteristics (sex, race, risk factors, such as obesity and smoking). The data from the HRS consists of longitudinal histories of disease incidence, recorded roughly every 2 years, from 1992 to 2002, along with information on baseline disease prevalence in 1992. Since incidence can only be recorded every 2 years, we use a discrete time hazard model.
The estimation of such a model is complicated by three factors. First, the report of conditions is observed at irregular intervals (on average 24 months but varying from 18 to 30), and interview delay appears related to health conditions. Second, the presence of persistent unobserved heterogeneity (frailty) could contaminate the estimation of dynamic pathways or “feedback effects” across diseases. Finally, because the HRS samples from a population of respondents aged 51+, inference is complicated by the fact that spells are left-censored: some respondents are older than 50 at baseline and suffer from health conditions whose age of onset cannot be established.
Since we have a stock sample from the age 55+ population (the HRS starts at age 51, but our model starts at age 55), each respondent goes through an individual-specific series of intervals between interviews. Hence, we have an unbalanced panel over the age range starting from 55 years old. Denote by ji0 the first age at which respondent i is observed and the last age of observation. Hence we observe incidence at ages Record as if the individual has condition m as of age ji. We assume the individual-specific component of the hazard can be decomposed into time-invariant and time-varying pieces. The time-invariant piece is composed of the effect of observed characteristics xi and permanent unobserved characteristics specific to disease m, ηi,m . The time-varying part is the effect of previously diagnosed health conditions (other than the condition m) on the hazard.4 We assume an index of the form Hence, the latent component of the hazard is modeled as
We approximate with an age spline. After several specification checks, a node at age 75 appears to provide the best fit. This simplification is made for computational reasons since the joint estimation with unrestricted age fixed effects for each condition would imply a large number of parameters.
Diagnosis, conditional on being alive, is defined as
We consider functional limitation [measured as activities of daily living (ADL) limitation disability], mortality, and the following seven diseases: heart disease, hypertension, diabetes, cancer, lung disease, stroke, and mental illness. Each of these conditions is an absorbing state. The occurrence of mortality censors observation of diagnosis for other diseases in a current year. Mortality is recorded from exit interviews and closely reflects the life-table probabilities of survival.
A2.1 Interview Delays
As we already mentioned, time between interviews is not exactly 2 years. It can range from 18 months to 30 months. Hence, estimation is complicated by the fact that intervals are different for each respondent. More problematic is that delays in the time of interview appear related to age, serious health conditions and death (Adams et al. 2004). Hence a spurious correlation between elapsed time and incidence would be detected when in fact the correlation is due to delays in interviewing or finding out the status of respondents who will later be reported dead. To adjust hazard rates for this, we follow Adams et al. (2004) and include the logarithm of the number of months between interviews, as a regressor.
A2.2 Unobserved Heterogeneity
The term is a time-varying shock specific to age ji. We assume that this last shock is Type-1 extreme value distributed, and uncorrelated across diseases.5 Unobserved differences ηim are persistent over time and are allowed to be correlated across diseases m=1, …, M. However, to reduce the dimensionality of the heterogeneity distribution for computational reasons, we consider a nested specification. We assume that heterogeneity is perfectly correlated within nests of conditions but imperfectly correlated across nests. In particular, we assume that each of the 7 diseases (heart disease, hypertension, stroke, lung disease, diabetes, cancer and mental illness) have a one-factor term ηim=τmαiC where τm is a disease specific factor-loading for the common individual term αiC. We assume disability and mortality have their own specific heterogeneity term αiD and αiM. Together, we assume that the triplet (αiC, αiD, αiM) has some joint distribution that we will estimate. Hence, this vector is assumed imperfectly correlated. We use a discrete mass-point distribution with 2 points of support for each dimension (Heckman and Singer 1984). This leads to K=8 potential combinations.
A2.3 Likelihood and Initial Condition Problem
The parameters where Fα are the parameters of the discrete distribution, can be estimated by maximum likelihood. Given the extreme value distribution assumption on the time-varying unobservable (a consequence of the proportional hazard assumption), the joint probability of all time-intervals until failure, right-censoring or death conditional on the individual frailty is the product of Type-1 extreme value univariate probabilities. Since these sequences, conditional on unobserved heterogeneity, are also independent across diseases, the joint probability over all disease-specific sequences is simply the product of those probabilities.
For a given respondent with frailty αi=(αiC, αiD, αiM) observed from initial age ji0 to a last age the probability of the observed health history is (omitting the conditioning on covariates for notational simplicity)
We make explicit the conditioning on we have no information on health prior to this age.
To obtain the likelihood of the parameters given the observables, it remains to integrate out unobserved heterogeneity. The complication is that the initial condition in each hazard, is not likely to be independent of the common unobserved heterogeneity term which needs to be integrated out. A solution is to model the conditional probability distribution Implementing this solution amounts to including initial prevalence of each condition at baseline in each hazard. Therefore, this allows for permanent differences in the probability of a diagnosis based on baseline diagnosis on top of additional effects of diagnosis on the subsequent probability of a diagnosis. The likelihood contribution for one respondent’s sequence is therefore given by
where the pk are probabilities for each combination of points of support αk k=1,…,K. The Broyden–Fletcher–Goldfarb–Shanno (BFGS) algorithm is used to maximize the log sum of likelihood contributions in Equation 4 over the admissible parameter space.
A2.4 Clinical Restrictions
Although statistically speaking, all elements of γm for all diseases should be unrestricted, it is likely that some of these estimates will reflect associations rather than causal effects because they help predict future incidence. Although we control for various risk factors, it is likely to that we do not observe some factors which are correlated with other diseases. In medical terms, however, some of these effects might be ruled improbable and we use results from the medical literature to guide restrictions to impose on the elements of the γm.
We use a set of clinical restrictions proposed by Goldman et al. (2005) based on elicitations of medical expert panels. Statistical analysis corroborates the expert advice, since the data fail to reject these restrictions.
A2.5 Descriptive Statistics and Estimation Results
For estimation, we construct an unbalanced panel by pooling all cohorts together. We delete spells if important information is missing (such as the prevalence of health conditions). Hence, in the final sample, a sequence can be terminated because of death, unknown exit from the survey (or non-response to key outcomes), or the end of the panel.
In each hazard, we include a set of baseline characteristics that capture the major risk factors for each condition. We consider education, race, ethnicity, marital status, gender and behaviors such as smoking and obesity. Finally, as discussed previously, we also include a measure of the duration between interviews, in months. The average duration is close to 2 years. Table A2 gives descriptive statistics at first interview.
Estimates of the hazard models are presented in Table A3. Estimates can be interpreted as the effect on the log hazard. To judge the fit of the model we perform a goodness-of-fit exercise. In order to do so, we re-estimate the model on a sub-sample of data and keep part of the sample for evaluating the fit. We randomly select observations from the original HRS cohort with probability 0.5 and simulate outcomes for this cohort starting from observed 1992 outcomes. Table A4 gives the observed frequencies as well as the predicted ones. Predicted and observed frequencies are quite close to each other in 2002.
A2.6 Life-Time Prevalence vs. Current Prevalence of ADL Limitations
The transition model we estimate takes ADL limitations as an absorbing state. In the HRS dataset, we know lifetime prevalence from 1992 to 2004 as well as current prevalence. Hence, we estimate a probit where the dependent variable is “ever had ADL from 1992 to 2004”, and the sample is those with no current ADL in 2004. We use as regressors the whole set of health conditions and demographics used in the model. We then impute life-time prevalence in the U.S. based on the probit model. Table A5 gives the parameter estimates of the probit regression.
A3 Calculating the Health Effects of New Drugs
Prior to the introduction of a new drug, it is useful to think of the population as divided into three subgroups: Those undiagnosed for a particular health condition, those diagnosed with a condition and treated with an existing drug, and those who are diagnosed but untreated. The new drug does not benefit the undiagnosed, but can benefit either or both of the diagnosed groups.
Figure A2 illustrates how each sub-population benefits from a new drug. The treated group may benefit because new drugs are potentially more effective than existing treatments. We will call this the “clinical effect.” A fraction of the previously untreated group may, however, gain access to this treatment. This group will thus experience the full health effect of a new drug (relative to no treatment). We will refer to this as the “access effect”. Finally, the remaining untreated individuals do not gain from the introduction of the drug.6
In such a population, the average effect of a new drug will be a weighted average of the clinical and access effect. To see this, denote by (α0, α1) the fraction of diagnosed individuals who are untreated before and after the introduction of the new drug, respectively, and define ∆α∞=α0–α1. Denote P as the probability of being diagnosed with a new condition or death, and RREXISTING=PNEW/PEXISTING, RRPLACEBO=PNEW/PPLACEBOto be the relative risk (RR) for those replacing existing therapy with the new drug, and those replacing no therapy with the new drug, respectively. Note that the latter corresponds to the risk given by the new drug, divided by the risk given by placebo treatment. As such, the annual RR for the diagnosed population following the introduction of a new drug (relative to the pre-existing situation) is given by
where PU is the baseline health risk among the untreated, PT is the baseline health risk among the treated, and P̅=α0PU+(1–α0)PT is the average baseline risk in the entire diagnosed population.
The first term in square brackets represents the total reduction in risk for the untreated, while the second term represents the total reduction in risk for the treated. Each is weighted by the proportion of the diagnosed population that is untreated and treated, respectively.
For example, suppose a new drug is introduced which potentially reduces mortality for patients diagnosed with cancer. Compared with existing treatment, this drug is 25% more effective. Compared with no treatment, however, it leads to a 50% decrease in mortality risk. Now, suppose 50% of patients diagnosed with cancer take the existing treatment and that both treated and untreated patient face a survival probability of 75%. Finally, assume the introduction of the new drug means that 25% more patients are treated. This leads to a 25% reduction in untreated patients. Hence, 25% of the diagnosed population derives no benefit, 25% enjoy a 50% decrease in mortality risk because they move from no treatment to the new therapy, and the remaining 50% of the population enjoy the 25% improvement over existing treatment. The average RR is then 0.25+0.25*0.5 + 0.5*0.75=0.75 for the diagnosed population.7 Note that ignoring the access effect amounts to imposing ∆α=0. In this case, only the treated will benefit from the new drug. For many diseases, this access effect can be important, particularly when access is relatively low and existing treatments are not easily accessible. The same calculations can be performed for other risks such as the risk of being diagnosed for another health condition.
Below, we provide methodologies for estimating each of the components of Equation 5. The list of new-top selling drugs section is devoted to the construction of a list of drugs for each of the conditions we consider. In subsequent sections, we discuss how the RRS are taken from clinical trials for these drugs, how the access effect (∆α) are estimated from claims data, and how we calculate the remaining parameters (α0, PU, PT) from the HRS data.
For each of the health condition in the transition model, one could in principle consider the whole universe of new drugs and calculate an average effect for each of them. This is likely to be a difficult task. However, top-selling drugs are the most likely to have large effects, and have been in general the most studied and reviewed. This makes the estimated clinical benefits more reliable. For each of our diseases, therefore, we survey the clinical effects for the five “top-selling” drugs in that disease group, and assume conservatively that all other drugs outside the top five have no therapeutic benefits. Therefore, we estimate the effects of drugs in two parts: (i) calculate the probability that a new drug will be a “top-seller,” and (ii) apply the expected therapeutic benefit of a top-selling drug.
A3.1.1 List of New Top-Selling Drugs
We construct a list of new drugs from INGENIX, a large, nationwide, longitudinal claims-based database (1997–2004).8 This data set has drug expenditure information from insurance plan enrollees. We use expenditures as a proxy to identify top-selling drugs. Health conditions in INGENIX are provided at the patient level, which makes it difficult to match drugs to health conditions because patients can take medication for multiple diseases at the same time. For example, it is unclear whether a drug used by a heart disease patient with hypertension is used to treat the heart condition or hypertension. Mapping drugs to Redbook drug class and then to health conditions appears to be a superior strategy (Truven Health Analytics Micromedex Solutions 2016). Hence, we first group drugs by class and then assign each class to at least one particular health condition, based on expert opinion and an extensive web search.9 The result of this class-health condition match is presented in Appendix A1.
We rank drugs for each health condition according to real revenues in the 2nd full year following introduction. (For instance if the drug launches in 1996, we consider revenues in 1998).10 We consider new chemical entities (NCEs) as well as reformulation and recombination drugs, but exclude generics. We define the top five drugs for each health condition as “top-selling drugs”. The name of the drug, Redbook drug class, generic name, and the introduction date are presented in Appendix A2.
The choice of the second full year after introduction is somewhat arbitrary. Ultimately, we are interested in the lifetime revenues of the drug, but this is unobserved. Our analysis suggests that the choice of year does not substantially affect the results. Table A6 displays the pairwise correlations between sales ranks calculated using the first full year after launch, second full year after launch, third full year after launch, and fourth full year after launch. It demonstrates that the correlations between these various rank statistics is extremely high – typically well above 0.90, and in many cases above 0.95.
A3.1.2 Calculating Effects on Health
For each of the top-selling drugs, we survey the medical literature for clinical trials. When the trials do not provide an estimate for the health condition we are interested in, we assume the drug has no effect. Hence, these estimates can be seen as conservative. We do the same when the estimate is not statistically significant. When more than one estimate is available, we use the mean of the effects found.
We searched for the impacts of top-selling drugs on mortality, and on the incidence of all 6 other health conditions under consideration. However, we follow Goldman et al. (2005) in ruling out some causal links, based on expert opinion. For example, we assume that there is a causal link between hypertension and diabetes, but not from hypertension to cancer. We do not investigate the effect of new drugs on recovery or cure rates. Table A7 summarizes results from the survey of the literature for those effects. Appendix A3 gives detail on the calculation of each estimate. These calculations provide the estimates of RREXISTING and RRPLACEBO in eq. (1).
To interpret the table, note that a value of 1.0 means that the new drug has the same impact on health as the relevant alternative, or correspondingly, a value of 0.8 means that the new drug lowers risk by 20%.
A3.1.3 Access Effect
To calculate the access effect, we need to construct an estimate of ∆α which is the decrease in the fraction of untreated individuals following the introduction of a new drug. We estimate this effect using prescription claims data from the INGENIX data. By merging the drug consumption data with data on the introduction date of new drugs (from Appendix A2), we get a panel data set of the number of prescriptions consumed monthly for each class before and after the introduction of the top five drugs (from 1997.1 to 2004.12).11 Our strategy is to compute the effect of a launch on prescriptions relative to the trend in prescriptions for a specific class. The statistical model that implements this strategy explains the logarithm of monthly prescriptions in a class c as
αc are class fixed effects, gc(t) is some class-specific function of calendar time. The variables are indicator functions that take value 1 when a new blockbuster drug has been on the market for k months. Finally ε is some unobserved disturbance with zero mean. We specify linear class-specific time trends: gc(t)=ηct, which was not rejected against a more flexible specification. Hence the total effect of a new drug on prescriptions after 12 months is given by δ12. Table A8 presents estimation results.
New top sellers tend to have a relatively strong effect on the number of prescriptions after 3 months. After 1 year, there is a 24.5% increase in the number of prescriptions. This estimate is statistically significant at the 5% level. Figure A3 presents the estimates of the effect. We use 21% as our estimate of the access effect.
A3.1.4 Access and Incidence Rates by Conditions
Three more estimates are needed to compute average RRs from eq. (1). First, we need to know, for each condition, the fraction of diagnosed individuals not taking existing drugs (α0). Second, we need estimates of the incidence rates for those treated (PT) and untreated (PU). We use the HRS for that purpose. The HRS is a nationally representative longitudinal study of the age 50+ population. It asks about lifetime prevalence of the seven conditions we use as well as the consumption of drugs for those diagnosed with these conditions. It also tracks mortality. Mortality rates from the HRS follow closely figures from life-tables (Adams et al. 2004).12 Most of the differences are attributable to the fact that HRS samples the non-institutionalized population.
To construct estimates of the transition rates (P), we use hazard models estimated on the 1992–2002 waves of the HRS. The hazard models include baseline demographics, prevalence indicators at the previous wave, risk behaviors, and age. Table A9 gives the lifetime prevalence in 2004 of various conditions, the fraction untreated among the diagnosed population and predicted transition rates based on hazard models.
Of the 54.6% of individuals aged 50+ in 2004 with hypertension, only 11% do not take medication for that condition according to the HRS. A somewhat larger fraction with diabetes and heart disease does not take medication (18.3% and 33.8%). Cancer and stroke are the two conditions with the fewest respondents treated with drugs (77.2% and 63.3% are untreated). The next column presents the estimate of ∆α: the reduction in the fraction untreated following introduction of a new drug. Applying the effect calculated of 21% found in the previous section leads to a substantial reduction in the fraction untreated for most diseases. Finally, predicted average incidence rates prior to introduction are similar across groups of treated and untreated patients, if not higher in the treated group. Hence, we explicitly take into account of the fact that benefits for the untreated may be lower than for the treated because their condition is less severe.
A3.1.5 Average Effect of Top-Selling Drugs
We use estimates along with previously estimated parameters to construct the average RRs from eq. (1). Table A10 gives the results for each of the causal link we identified in Table A1.
These estimates are conservative, since they do not include effects on the undiagnosed population. We also assume drugs other than those in the top five list have no effect on health. Moreover, whenever we could not find a clinical effect from a peer-reviewed study we assume that a new drug has no effect.
A3.1.6 Fraction of Top-Selling Drugs (1995–2002)
We use data from the Food and Drug Administration (FDA) to compute how many NCEs – including reformulations and recombinations, in addition to “true” new chemical entities – were introduced over the 1995–2002 period. We compute the fraction of top-seller NCEs as a fraction of all NCEs over the period using the year in which they were approved by the FDA. We map each approved drug to health condition(s) by using the indications listed by the FDA in its annual report. Table A11 presents the results for the fraction of blockbuster NCEs from 1995–2002. The fraction of new top-selling drugs is quite different across diseases ranging from 9% for cancer to 33% for stroke.
A4 GPPM Innovation Module
The innovation module translates changes in market size into new innovation. Acemoglu and Lin (2004) found that, on the margin, a one percent increase in pharmaceutical market share leads to approximately a four percent increase in the annual number of drugs. More recently, David et al. (2009) have estimated that the internal rate of return is 13% for biologics and 7.5% for small molecules. This translates into manufacturers producing one additional biologic drug for every additional $161.3 million of annual potential revenue and one new small molecule drug for every additional $109.4 million of annual potential revenue. From here we can estimate the elasticity of innovation (ε) using the following formulas.
According to the Pharmaceutical Research and Manufacturers of America (PhRMA), total U.S. pharmaceutical sales in 2010 were $184.7 billion (all values reported in 2015$). Similarly, the Congressional Research Service reports that U.S. biologic sales in 2011 were $58.8 billion (Schacht and Thomas 2012). The FDA reported in 2014 that they approved an average of 26 new drugs per year (including both biologics and small molecules) between 2004 and 2013 ( U.S. Food and Drug Administration Center for Drug Evaluation and Research 2014). BioWorld reports that the average number of biologics approved between 2008 and 2013 is 5–6 per year (Serebrov 2013). This approach suggests an elasticity estimate of 4.1% (2.0–4.1%) for small molecules and 5% (3.0–5.0%) for biologics. We implemented an elasticity of 3.0% for small molecules and 4.7% for biologics in the model. DiMasi and Grabowski (2007) report that there is an 8-year lag between the decision to start research on a new drug and the time it arrives on the market. Hence, new drugs at time t depend on changes in market size at t–8. We therefore assume an innovation delay of 8 years in the model.
A5 Costs and Benefits
A5.1 Value of Remaining Life-Years
Denote by na,t the number of people of age a alive in alive in year t. In any given year, the value of discounted life-years ahead can be calculated from the simulation. Using a discount rate ρ and a value of a statistical life year v, this is given by
We use a discount rate of 3% to discount benefits as well as expenditures.
A5.2 Medical and Drug Expenditures
Because the HRS does not have accurate information on total medical expenditures and total drug expenditures, we use the MEPS to construct cost regressions. We regress these expenditures on the same demographics we have in the model as well as age and health condition indicators. Few differences in the definition of variables are observed. We use the sample of age 50+ individuals in MEPS. The regressions are performed separately for male and female as well as for each type of expenditure (drug and medical). The regression takes the form:
Medicare Current Beneficiary Survey (MCBS) provides a better medical cost expenditure estimation for the elderly (age 65+). We mapped the MEPS regression results to MCBS average values for the elderly, by adjusting the constant and age variable coefficients.
Table A12 gives the final results after MEPS regression and mapping MEPS to MCBS:
A6 Stochastic Simulation and Weighting
The horizon for each simulation is 2005–2150 by which time all 2060 new entrants have died. We start the simulation with the HRS 2004 sample. We adjust weights so that they match 2004 population counts provided by the United Nations’ (UN) Population Program. We do this by age in order to smooth out bumps in the age distribution. Each year, the population of age 55 respondents in 2004 is added back adjusting their weights for projected demographic trends. Table A13 presents forecasted growth rates. By using the age 55, 2004 cohort repeatedly and only adjusting for growth, we do not take into account composition effects due to different growth rates across different segment of the population (say obese population).
Since the sample size for each age group is small, weighting introduces a significant amount of variability. To reduce this variability, we use two replicate datasets of the new cohorts and adjust appropriately the weights.
We use an average of 30 replications of the simulation in order to reduce simulation noise. Each replication takes roughly five minutes on a HP DL145 Linux box running on 2 dual core Intel 2.2 GHZ processors with 8 GB of RAM (programmed with OxMetrics). Since we run many scenarios for different parameter values we make use of parallel computing on 2 DL145 machines (total of eight processors) using a message passing interface (MPI). Each processor is assigned a scenario, performs the calculations, and writes the result to file.
A7 Model Validity
The model is structured and calibrated using the best available data, but the most relevant test of validity compares its predictions to actual data. In this spirit, we compared the model’s performance at predicting the population, against widely accepted UN estimates of the elderly population. And, to the greatest extent permitted by the data, we compared the model’s projections for disease growth against actual data on disease prevalence in the U.S. Due to the relative novelty of biologics as compared to small molecule drugs, these validations were done only using the effects of innovation on small molecules.
A7.1 Population Projections
We compared to UN population forecasts to the model’s baseline forecasts for the 60+ population in the U.S. The results are presented in Figure A4. The UN forecasts for the U.S. population project a trend with a slight kink at 2030. The model is able to replicate this kink, and matches the UN forecasts almost exactly after this point. Prior to the kink point, the model’s forecasts are about two to three percent lower than the UN forecasts.
A7.2 Disease and Mortality Projections
In addition to predicting population, the model also predicts disease prevalence. For the U.S. population, we start with the U.S. HRS database, and split this into two. Using one half of the sample, we estimate the GPPM model, and simulate health transitions over a 10-year period, using this subsample. We then compare those estimated health transitions to actual 10-year health transitions in the other half of the HRS sample. Figure A5 compares the model’s predicted prevalence of disease and disability (and the rate of mortality) after 10 years, to the actual HRS prevalence of disease and disability after 10 years. Overall, the model tracks the actual prevalence of disease, disability, and mortality fairly closely, always remaining within one to two percentage points of the actual data. In percent (as opposed to percentage point) terms, the deviations are largest for cancer and stroke, where one and two percentage point differences translate into roughly twelve to thirteen percent deviations. Considering both the sampling variability in the “actual” prevalence rates, and the estimation error in the model parameters, the model performs reasonably well at fitting the data. Formal tests reveal that the confidence intervals around the model estimates always contain the actual prevalence rate, although this conclusion is meaningful only up to the statistical precision of the comparison sample.
Acemoglu, D. and J. Linn (2004) “Market Size in Innovation: Theory and Evidence from the Pharmaceutical Industry,” Quarterly Journal of Economics, 119(3):1049–1090. Google Scholar
Adams, P., M. D. Hurd, D. McFadden, A. Merrill, and T. Ribeiro (2004) “Healthy, Wealthy, and Wise? Tests for Direct Causal Paths between Health and Socioeconomic Status.” In: (David A. Wise, ed.) Perspectives on the Economics of Aging. Chicago and London: University of Chicago Press, pp. 415–518. Google Scholar
David, E., T. Tramontin, and R. Zemmel (2009) “Pharmaceutical R&D: The Road to Positive Returns,” Nature Reviews Drug Discovery, 8(8):609–610. Google Scholar
DiMasi, J. A. and H. G. Grabowski (2007) “The Cost of Biopharmaceutical R&D: Is Biotech Different?” Managerial and Decision Economics, 28(4–5):469–479. Google Scholar
Goldman, D. P., M. Hurd, P. G. Shekelle, S. J. Newberry, C. W. A. Panis, B. Shang, J. Bhattacharya, G. F. Joyce, and D. N. Lakdawalla (2004) Health Status and Medical Treatment of the Future Elderly: Final Report. Santa Monica, CA: RAND Corporation. Available from http://www.rand.org/publications/TR/TR169/.
Goldman, D. P., B. Shang, J. Bhattacharya, A. M. Garber, M. Hurd, G. F. Joyce, D. N. Lakdawalla, C. Panis, and P. G. Shekelle (2005) “Consequences of Health Trends and Medical Innovation for the Future Elderly,” Health Affairs, 24(Suppl 2):W5R5–17. Google Scholar
Heckman, J. and B. Singer (1984) “A Method for Minimizing the Impact of Distributional Assumptions in Econometric Models for Duration Data,” Econometrica, 52(2):271–320. Google Scholar
Lakdawalla, D. N., D. P. Goldman, and B. Shang (2005) “The Health and Cost Consequences of Obesity Among the Future Elderly,” Health Affairs (Millwood), 24(Suppl 2):W5R30–41. Google Scholar
Schacht, W. H. and J. R. Thomas (2012) Follow-on Biologics: The Law and Intellectual Property Issues. Washington, DC, USA: Congressional Research Services. Google Scholar
Serebrov, M. (2013) “Biologic New Molecular Entities Approvals are Nothing to Write Home About.” Google Scholar
Shekelle, P. G., E. Ortiz, S. J. Newberry, M. W. Rich, S. L. Rhodes, R. H. Brook, and D. P. Goldman (2005) “Identifying Potential Health Care Innovations for the Future Elderly,” Health Affairs (Millwood), 24(Suppl 2):W5R67–76. Google Scholar
Sood, N., H. De Vries, I. Gutierrez, D. N. Lakdawalla, and D. P. Goldman (2008) The Effect of Pharmaceutical Regulation on Revenues. Santa Monica, CA: RAND Corporation. Google Scholar
Truven Health Analytics Micromedex Solutions (2016) RedBook Online. Greenwood Village, CO: Truven Health Analytics. Google Scholar
U.S. Food and Drug Administration Center for Drug Evaluation and Research (2014) Novel New Drugs 2013 Summary. Accessed March 25, 2016. Available from http://www.fda.gov/downloads/Drugs/DevelopmentApprovalProcess/DrugInnovation/UCM381803.pdf.
Abbott, T. A. and J. A. Vernon (2007) “The Cost of US Pharmaceutical Price Regulation: A Financial Simulation Model of R&D Decisions,” Managerial and Decision Economics, 28(4–5):293–306. Google Scholar
Acemoglu, D. and J. Linn (2004) “Market Size in Innovation: Theory and Evidence from the Pharmaceutical Industry,” Quarterly Journal of Economics, 119(3):1049–1090. Google Scholar
Aldy, J. E. and W. K. Viscusi (2008) “Adjusting the Value of a Statistical Life for Age and Cohort Effects,” The Review of Economics and Statistics, 90(3):573–581. Google Scholar
Avalere Health (2016) VA National Formulary Covers 16 Percent Fewer Top Prescription Drugs than Medicare Part D Plans 2015 [cited February 10 2016]. Available from http://avalere.com/expertise/managed-care/insights/a-recent-avalere-analysis-found-that-the-va-national-formulary-covers-fewer.
Baker, D. (2013) Reducing Waste with an Efficient Medicare Prescription Drug Benefit. Washington, DC: Center for Economic and Policy Research. Google Scholar
Cockburn, I. M., J. O. Lanjouw and M. Schankerman (2014) Patents and the Global Diffusion of New Drugs. London, UK. Google Scholar
Congressional Budget Office (2005) Prices for Brand-Name Drugs Under Selected Federal Programs. Washington, DC. Google Scholar
Congressional Budget Office (2015) March 2015 Medicare Baseline. Washington, DC. Google Scholar
Coulam, R. F., R. D. Feldman and B. E. Dowd (2011) “Competitive Pricing and the Challenge of Cost Control in Medicare,” Journal of Health Politics, Policy, and Law, 36(4):649–689. Google Scholar
Cutler, D. (2007) The Lifetime Costs and Benefits of Medical Technology. Cambridge, MA: The National Bureau of Economic Research. Google Scholar
D’Angelo, Greg (2016) The VA Drug Pricing Model: What Senators Should Know. The Heritage Foundation 2007 [cited April 11 2016]. Available from http://www.heritage.org/research/reports/2007/04/the-va-drug-pricing-model-what-senators-should-know.
Danzon, P. M. and J. D. Ketcham (2004) “Reference Pricing of Pharmaceuticals for Medicare: Evidence from Germany, The Netherlands, and New Zealand,” Forum for Health Economics and Policy, 7(1):1–56. Google Scholar
DiMasi, J. A. and H. G. Grabowski (2007) “The Cost of Biopharmaceutical R&D: Is Biotech Different?,” Managerial and Decision Economics, 28(4–5):469–479. Google Scholar
Dismuke, Clara E. and Leonard E. Egede (2013) “Medicare Part D Prescription Drug Program: Benefits, Unintended Consequences and Impact on Health Disparities,” Journal of General Internal Medicine, 28(7):860–861. Google Scholar
Eber, M. R., D. P. Goldman, D. N. Lakdawalla, T. J. Philipson, D. Pritchard, M. Huesch, N. Summers, M. T. Linthicum, J. Sullivan and R. W. Dubois (2015) “Clinical Evidence Inputs to Comparative Effectiveness Research Could Impact the Development of Novel Treatments,” Journal of Comparative Effectiveness Research, 4(3):203–213. Google Scholar
Eger, S. and J. C. Mahlich (2014) “Pharmaceutical Regulation in Europe and its Impact on Corporate R&D,” Health Economics Review, 4:23. Google Scholar
Federal Reserve Bank of St. Louis (2016) Consumer Price Index for All Urban Consumers: Medical Care. Federal Reserve Bank of St. Louis, 2016 [cited April 26 2016]. Available from https://research.stlouisfed.org/fred2/series/CPIMEDSL.
Filson, D. (2012) “A Markov-perfect Equilibrium Model of the Impacts of Price Controls on the Performance of the Pharmaceutical Industry,” The RAND Journal of Economics, 43(1):110–138. Google Scholar
Frakt, A. B., S. D. Pizer and R. Feldman (2012) “Should Medicare Adopt the Veterans Health Administration Formulary?” Health Economics, 21(5):485–495. Google Scholar
Gagnon, M.A. and S. Wolfe (2015) Mirror, Mirror on the Wall: Medicare Part D Pays Needlessly High Brand-Name Drug Prices Compared with Other OECD Countries and with U.S. Government Programs Ottowa, ON: Carleton University. Google Scholar
Gellad, W. F., S. Schneeweiss, P. Brawarsky, S. Lipsitz and J. S. Haas (2008) “What if the Federal Government Negotiated Pharmaceutical Prices for Seniors? An Estimate of National Savings,” Journal of General Internal Medicine, 23(9):1435–1440. Google Scholar
Giaccotto, C., R. E. Santerre and J. A. Vernon (2005) “Drug Prices and Research and Development Investment Behavior in the Pharmaceutical Industry,” The Journal of Law and Economics, 48(1):195–214. Google Scholar
Goldman, D. P., D. N. Lakdawalla, J. D. Malkin, J. Romley and T. J. Philipson (2011) “The Benefits from Giving Makers of Conventional ‘Small Molecule’ Drugs Longer Exclusivity over Clinical Trial Data,” Health Affairs, 30(1):84–90. Google Scholar
Golec, J. H. and J. A. Vernon (2006) European Pharmaceutical Price Regulation, Firm Profability, and R&D Spending. Cambridge, MA: The National Bureau of Economic Research. Google Scholar
Grossmann, V. (2013) “Do Cost-Sharing and Entry Deregulation Curb Pharmaceutical Innovation?” Journal of Health Economics, 32(5):881–894. Google Scholar
IMS Institute for Healthcare Informatics (2015) Medicines Use and Spending Shifts: A Review of the Use of Medicines in the U.S. in 2014. Pasipanny, NJ. Google Scholar
Jacobson, G. A., S.V. Panangala, and J. Hearne (2007) Pharmaceutical Costs: A Comparison of Departments of Veterans Affairs (VA), Medicaid, and Medicare Policies. U.S. Library of Congress. Congressional Research Service. Google Scholar
Kaiser Family Foundation (2016) The Medicare Part D Prescription Drug Benefit 2013 [cited May 18 2016]. Available from http://kff.org/medicare/fact-sheet/the-medicare-prescription-drug-benefit-fact-sheet/.
Kaiser Family Foundation (2016) Searching for Savings in Medicare Drug Price Negotiations 2016 [cited February 11 2016]. Available from http://files.kff.org/attachment/issue-brief-searching-for-savings-in-medicare-drug-price-negotiations.
Lakdawalla, D. N., D. P. Goldman, P. C. Michaud, N. Sood, R. Lempert, Z. Cong, H. de Vries and I. Gutierrez (2009) “U.S. Pharmaceutical Policy in a Global Marketplace,” Health Affairs, 28(1):138–150. Google Scholar
Lambrelli, D. and O. O’Donnell (2011) “The Impotence of Price Controls: Failed Attempts to Constrain Pharmaceutical Expenditures in Greece,” Health Policy, 101(2):162–171. Google Scholar
Murphy, K. M. and R. H. Topel (2003) “The Economic Value of Medical Research,” In: (K. M. Murphy and R. H. Topel, eds.) Measuring the Gains from Medical Research: An Economic Approach. Chicago: University of Chicago Press, pp. 41–73. Google Scholar
Murphy, K. M. and R. H. Topel (2006) “The Value of Health and Longevity,” Journal of Political Economy, 114(5):871–904. Google Scholar
Pharmaceutical Research and Manufacturers of America (2012) PhRMA 2012 Industry Profile. Washington, DC. Google Scholar
Pitts, P. J. and R. M. Goldberg (2015) Price vs. Value: The Story and the Rest of the Story. New York, NY: Center for Medicine in the Public Interest. Google Scholar
Poterba, J., S. Venti and D. A. Wise (2007) The Decline of Defined Benefit Retirement Plans and Asset Flows. Cambridge, MA. Google Scholar
Serebrov, M. (2014) “Biologic New Molecular Entities Approvals Are Nothing to Write Home About,” BioWorld, January 17. Accessed December 21, 2016. Available from http://www.bioworld.com/content/biologic-new-molecular-entities-approvals-are-nothingwrite-home-about-0.
Sood, N., H. de Vries, I. Gutierrez, D. N. Lakdawalla and D. P. Goldman (2009) “The Effect of Regulation on Pharmaceutical Revenues: Experience in Nineteen Countries,” Health Affairs, 28(1):125–137. Google Scholar
Stroupe, Kevin T., Bridget M. Smith, Timothy P. Hogan, Justin R. St. Andre, Walid F. Gellad, Saul Weiner, Todd A. Lee, Muriel Burk, Francesca Cunningham, John D. Piette, Thea J. Rogers, Zhiping Huo and Frances M. Weaver (2013) “Medication Acquisition across Systems of Care and Patient–Provider Communication among Older Veterans,” American Journal of Health-System Pharmacy, 70(9):804–813. Google Scholar
The Board of Trustees of the Federal Hospital Insurance and Federal Supplementary Insurance Trust Funds (2015) 2015 Annual Report of the Boards of Trustees of the Federal Hospital Insurance and Federal Supplementary Medical Insurance Trust Funds. Washington, DC. Google Scholar
Tuna, E., K. Atikeler and Z. Calýskan (2014) “Classification of Drugs Bringing from Abroad in Turkey According to Their ATC Codes,” Value in Health, 17(7):A788. Google Scholar
U.S. Department of Commerce International Trade Administration (2004) Pharmaceutical Price Controls in OECD Countries: Implications for U.S. Consumers, Pricing, Research and Development, and Innovation. Washington, DC. Google Scholar
U.S. Department of Veterans Affairs (2016) Contractor-The Ins and Outs of Public Law 102-585 2015 [cited February 10 2016]. Available from http://www.fss.va.gov/faqs/publicLaw102585.asp.
U.S. Food and Drug Adminstration (2014) Novel New Drugs 2013 Summary. Silverspring, MD: Center for Drug and Evaluation Research. Google Scholar
Vernon, J. A. 2005. “Examining the Link Between Price Regulation and Pharmaceutical R&D Investment,” Health Economics, 14(1):1–16. Google Scholar
Viscusi, W. K., and J. E. Aldy (2003) “The Value of a Statistical Life: A Critical Review of Market Estimates Throughout the World,” Journal of Risk and Uncertainty, 27(1):5–76. Google Scholar
An additional scenario parameter is the effect of the policy on production costs (e.g., costs of clinical trials), which we assume are not impacted by price controls in Part D since VA-style pricing is highly unlikely to change production costs.
This method accounts for differences in disease severity between untreated and treated individuals. However, note that our approach presumes that the effects of drugs can be appropriately captured by computing the average reduction in risk. An alternative would be to explicitly assign different reductions in risk to different people — e.g., those with different disease severity levels. While the latter approach would be more general, data limitations preclude its implementation: we do not have a panel of data on treatment status, which makes it ultimately impossible for us to separately model disease dynamics for the treated and untreated population.