O N THE
MIT Sloan School of Management
POPI WP #60-01
G. K. Raju & Charles L. Cooney. June 2001
Much of the prevailing connectionist machine learning research in chemical engineering assumes a one-way passive relationship between the learner and the application domain. In this paper, we investigate a two-way active relationship between learner and domain. We argue that this is useful and even necessary if the prevailing research is to be successfully applied to real world problems involving sparse and strongly biased data. A process development case study is used to illustrate the impact of data quality and quantity and to compare the performance of active learning against conventional passive learning. In doing so, we highlight the need to assess data quality and demonstrate the improvement in the rate of learning that results from active learning.
POPI WP #59-01
Sarah C. Stallings, Robert H. Rubin, Thomas J. Allen, Charles L. Cooney,
Anthony J. Sinskey & Stan N. Finkelstein. May 2001
Scientific and technological advances have led to many important innovations in drug discovery and development. Some have allowed rapid gene sequencing consequently making genetic information about cellular mechanisms of disease available at an unprecedented rate. Others are allowing access to increasingly sophisticated data on the disposition of that genetic information. We have analyzed how several of these advances have been integrated into the process of drug discovery and development. Our studies suggest that 1) technological advances are rooted in basic research that may take years to mature into effective strategies for developing new therapies, 2) new technologies are facilitating more confident decisions during drug development, 3) genomics is changing the profile of successful products by suggesting the possibility of more effective medicines with fewer side effects, and 4) to the extent that policies encouraging innovation and access to new drugs are in conflict with health cost containment efforts, decision-makers need to find a solution that provides for both priorities.
POPI WP #58-01
Sarah C. Stallings, Robert H. Rubin, Thomas J. Allen, Charles M. Cooney, Stan N. Finkelstein & Anthony J. Sinskey. February 2001
The burden of disease and the provision of health care are fundamental concerns worldwide. The pharmaceuticals that result from innovation are uncommonly valuable tools for health care. The scientific and technological advances behind today's medicines develop over sometimes long periods of time in a compounding manner. History has demonstrated that if basic scientific research is supported and innovation is encouraged, technology generates more technology, and during the lag between technological advances and clinical successes, benefits to biomedical science accrue continually.
Advances in science and technology are providing opportunities to develop drugs that address three major problems with pharmaceutical use in clinical medicine:
Now biomedical capabilities are providing information and tools for the creation of useful models of complex biological systems. The increasing knowledge and relevant models will lead to new therapies for unmet medical needs and safer and more effective medicines in two ways: first by expanding the understanding of disease, and second by decreasing the uncertainty around component decisions in drug development. Effective application of advances in science and technology to discovering and developing new drugs will be necessary to capture these opportunities for the benefit of public health.
POPI WP #57-00
Stan N. Finkelstein, Iain M. Cockburn, Howard L. Bailit, Jason Verner, Kenan E. Haver & Ernst R. Berndt. November 2000
OBJECTIVES : To examine the cost impact among parents of children with asthma who suffer lost productivity or absenteeism from work due to exacerbations or lapses in the control of the children’s asthma symptoms.
DESIGN AND SETTING
MAIN OUTCOME MEASURES
POPI WP #56-00
Ernst R. Berndt. September 2000
Particularly since 1994, growth in utilization rather than growth in prices has been the primary driver of increased prescription pharmaceutical expenditures in the US. But what are the drivers of utilization growth?
In this paper, I focus on four factors that have increased utilization, even as cost containment efforts have flourished: (i) "the importance of being unimportant"; (ii) increased third party prescription drug insurance coverage; (iii) the introduction of successful new products; and (iv) aggressive technology transfer and marketing efforts by pharmaceutical firms. I also consider what roles these four factors are likely to play in the future.
POPI WP #55-00
Ernst R. Berndt, Anupa Bir, Susan H. Busch, Richard G. Frank & Sharon-Lise T. Normand. January 2000
A common perception is that increases in mental health spending over the last decade reflect increasing costs of treatment. But are increased mental health expenditures due to price increases, quantity (volume) increases, or some combination? This question raises important measurement issues.
Recently researchers and health care policy analysts have renewed professional interest in constructing price indexes for episodes of treatment for an illness, rather than focusing measurements on changing medical care input prices. By pricing episodes of care, the overall cost impacts of changing treatment modalities can be taken into account, incorporating therapeutic substitution among inputs, technological progress, and effects of cost containment.
POPI WP #54-99
Ernst R. Berndt, Robert S. Pindyck & Pierre Azoulay. March 1999
We examine the role of network effects in the demand for pharmaceuticals at both the brand level and for a therapeutic class of drugs. These effects emerge when use of a drug by others conveys information about its efficacy and safety to patients and physicians. This can lead to herd behavior where a particular drug -- not necessarily the most efficacious or safest -- can come to dominate the market despite the availability of close substitutes, and can also affect the rate of market diffusion. Using data for H2-antagonist anti-ulcer drugs, we examine two aspects of these effects. First, we use hedonic price procedures to estimate how the aggregate usage of a drug affects brand valuation. Second, we estimate discrete-time diffusion models at both the industry and brand levels to measure the impact on rates of diffusion and market saturation.
POPI WP #53-99
Iain Cockburn, Rebecca Henderson & Scott Stern. April 1999
There is considerable evidence that there are significant and persistent "fixed effects" across firms. Empirical work in industrial organization routinely controls for firm fixed effects. These are usually statistically significant, often accounting for a substantial fraction of the variation of the independent variable, and in many contexts controlling for fixed effects has quite dramatic effects on estimated coefficients. A number of interpretations of this phenomenon have been offered.
One interpretation focuses on the role of fixed effects in controlling for a number of standard measurement problems. The productivity literature, for example, faces substantial difficulties in computing capital stocks from accounting data, dealing with firm-specific input and output prices, controlling for unobservable factors of production such as managerial effort and so on. Studies of R&D performance face the very serious problem that even within tightly defined industries firms may differ substantively in the range of technologies in which they invest, and hence in the technological opportunity sets that they face. To the extent that these kinds of unobserved heterogeneity persist within the firm over time, and are correlated with observed variables in the model, then failing to control for firm effects can have quite serious consequences. One view, therefore, is that firm dummies are essentially nuisance parameters which are important to estimate but are of little economic interest in and of themselves.
POPI WP #52-99
Ernst R. Berndt, Howard L. Bailit, Martin B. Keller, Jason C. Verner & Stan N. Finkelstein. (June 1999), Revised July/August 2000.
We examine the diagnosed prevalence of mental disorders among data processing employees at a large US employer. We compare average objective productivity performance of employees diagnosed with various mental disorders to random set of employees, we document their differential health care service utilization and expenditures, and we examine patterns of comorbidities among the mental disorders and with other somatic illnesses. We find that employees diagnosed with anxiety have extensive comorbidities with other mental disorders and with somatic diseases, and that the health care utilization and expenditures claims implications of these comorbidities are substantial.
POPI WP #51-99
Pierre Azoulay, April 1999
This paper investigates how different sources of information influence the diffusion of pharmaceutical innovation. In prescription drug markets, both advertising and scientific information stemming from clinical trials can affect physicians' prescription choices. Using a novel index of clinical research output, I find that both marketing and scientific evidence directly influence the diffusion process in the anti-ulcer drug market, with marketing having a more pronounced impact. I also find strong evidence that clinical outputs are important drivers of firms' marketing efforts, affecting sales indirectly. Taken together, the direct and indirect effect of science on demand imply strong private incentives for clinical research. I conclude that product market competition in the pharmaceutical industry is shaped both by advertising rivalries and scientific rivalries. Moreover, drug advertising performs an important informative function.
POPI WP #50-99
Iain Cockburn, Rebecca Henderson & Scott Stern, December 1998
This paper presents empirical evidence that the intensity of research workers' incentives for the distinct tasks of basic and applied research are positively associated with each other. We relate this finding to the prediction of the theoretical literature that when effort is multi-dimensional, firms will "balance" the provision of incentives; when incentives are strong along one dimension, firms will set high-powered incentives for effort along other dimensions which compete for the worker's effort and attention. We test for this effect in the context of pharmaceutical research using detailed data on individual research programs financed by private firms. Consistent with the complementarity hypothesis, we find strong evidence that firms who provide strong promotion-based incentives individuals to invest in fundamental or "basic" research also provide more intense incentives for success in applied research through the capital budgeting process. The intensity of these "bonus" incentives is weaker in firms who use a more centralized research budgeting process. We interpret this latter finding as providing support for theories which emphasize substitutability between contractible and non-contractible signals of effort.
POPI WP #49-99
Paul Ellickson, Scott Stern & Manuel Trajtenberg, December 1998
The main goal of this paper is to develop an empirical framework for evaluating the patient welfare benefits arising the pharmaceutical innovation. Extending previous studies of the welfare benefits from innovation, this paper unpacks the separate choices made by physicians and patients in pharmaceutical decision-making and develops an estimable econometric model which reflects these choices. Our proposed estimator for patient welfare depends on (a) whether patients comply with the prescriptions they receive from physicians and (b) the motives of physicians in their prescription behavior. By focusing on compliance behavior, the proposed welfare measure reflects a specific economic choice made by patients. We review evidence that the rate of noncompliance ranges up to 70%, suggesting an important gulf between physician prescription behavior and realized patient welfare. Since physicians act as imperfect but interested agents for their patients, the welfare analysis based on compliance must account for the nonrandom selection of patients into drugs by their physicians. The key contribution of this paper resides in integrating the choices made by both physicians and patients into a unified theoretical framework and suggesting how the parameters of such a model can be estimated from data.
POPI WP #48-99
Thomas J. Allen, Ralph Katz, Frank Basa & Stephen White, 1999
This questionnaire survey study of 57 drug development teams examines how the locus of decision influence between project leaders and departmental managers affects the performance of drug development projects at the US R&D laboratories of eight large, multinational pharmaceutical firms. Project performance is higher when team members perceive functional managers to have greater influence over go/no go decisions during early and late phases. Project performance is also higher when project leaders have greater influence over clinical decisions during later phases. The technical knowledge of the project leader is related to project performance in a complex fashion. Technically knowledgeable project leaders are more effective during late phase projects. During early phases, project leaders who are rated as having greater technical knowledge head lower performance teams.
Across firms in the sample, team members do not feel that teams have autonomy to carry out their mission, organizations do not have clear criteria for assessing team performance, and individual team member rewards are not linked to their performance as a project team member. These characteristics of the organizational milieu are also correlated with leadership structure. Dual leader firms appear to provide a better environment for project teams.
The study examines a series of relationships among locus of decision making influence, project leader characteristics, organizational support for teams, project phase, leadership structure, and project performance. It concludes with guidelines for structuring organizations.
POPI WP #47-99
Paul M. Healy & Christopher D. Howe, January 1999
This paper develops a simulation model for a pharmaceutical firm to provide evidence on the tradeoff between objectively and relevance for several methods of reporting R&D outlays. The findings indicate that a simple capitalization rule, analogous to the successful efforts method of capitalizing oil and gas exploration costs, provides a significantly higher relation between accounting information and economic values than immediate expensing of R&D outlays or capitalizing the full cost of outlays. More importantly, our findings show the relevance of successful efforts data persists even when there is widespread earnings management.
POPI WP #46-98
Iain M. Cockburn, Ernst R. Berndt, Howard L. Bailit & Stan N. Finkelstein, November 1998
OBJECTIVE: We examined the effects on work productivity of treatment with antihistamines in the retrospective study using linked health claims data and daily work output records for a sample of clerical workers.
METHOD: Among nearly 6000 individuals employed as claims processors at a large US insurance company, 682 individuals filled at least one prescription for an oral antihistamine during the period January 1993 to July 1995. We estimate a multiple regression model to explain variation in work output. We consider separately the productivity impacts of "sedating" and "nonsedating" antihistamines, in addition to characteristics of the employees, their jobs, and day-of-week and month factors.
RESULTS: No significant impact on productivity is visible in the period immediately preceding the day on which patients obtained the drug. During the period after obtaining the drug, differences up to 13% in productivity are found between patients taking sedating and nonsedating antihistamines.
INTERPRETATION: The size of the observed effect suggests a substantial indirect economic cost from use of certain medical treatments. These kinds of costs have until now been largely overlooked, because work productivity has been difficult to measure objectively.
POPI WP #45-98
Scott Stern & Manuel Trajtenberg, October 1998
This paper studies the consequences of physician authority on pharmaceutical prescribing. Physicians engage in a costly process of "matching" patients to the drug which most suits their particular conditions and characteristics. The relative efficiency of this matching process results from the diagnostic skill of the physician along with the investments made by the doctor in learning about different drugs. While the underlying level of physician skill or knowledge cannot be observed, differences among physicians in terms of these attributes are reflected in their prescribing behavior. We provide evidence for two major findings regarding the exercise of physician authority in this context. First, there is substantial variation in the degree to which physician prescribing is concentrated (i.e., some physicians prescribe a more diverse portfolio of drugs than others). Second, this concentration is correlated with observable drug characteristics. In particular, concentrated prescribers tend to prescribe drugs with high levels of advertising, low prices, and high (lagged) market shares. Our empirical results provide evidence for the importance of both physician effort and diagnostic ability in the prescribing process. In particular, physicians who differentiate among their patients more finely are more likely to have less concentrated prescribing portfolios and to be less sensitive to information sources which promote the use of drugs for the "average" patient.
POPI WP #44-98
Ernst R. Berndt, David M. Cutler, Richard G. Frank, Zvi Griliches, Joseph P. Newhouse & Jack E. Triplett, September 1998
We review in considerable detail the conceptual and measurement issues that underly construction of medical care price indexes in the U.S., particularly the medical care consumer price indexes (MCPIs) and medical-related producer price indexes (MPPIs). We outline salient features of the medical care marketplace, including the impacts of insurance, moral hazard, principal-agent relationships, technological progress and organizational changes. Since observed data are unlikely to correspond with efficient outcomes, we discuss implications of the failure of transactions data in this market to reveal reliable marginal valuations, and the consequent need to augment traditional transactions data with information based on cost-effectiveness and outcomes studies.
We describe procedures currently used by the BLS in constructing MCPIs and MPPIs, including recent revisions, and then consider alternative notions of medical care output pricing that involve the price or cost of an episode of treatment, rather than prices of fixed bundles of inputs. We outline features of a proposed new experimental price index -- a medical care expenditure price index -- that is more suitable for evaluation and analyses of medical care cost changes, than are the current MCPIs and MPPIs. We conclude by suggesting future research and measurement issues that are most likely to be fruitful.
POPI WP #43-98
Ernst R. Berndt, Lorrin M. Koran, Stan N. Finkelstein, Alan J. Gelenberg, Ivan M. Miller, George Trapp & Martin B. Keller, Revised 2000.
Most young adults pursue post high-school education in the ten to fifteen years immediately following high school graduation. By foregoing current income and incurring the tuition, board and textbook costs of higher education, individuals (and their financial sponsors) invest in the future. These higher education costs can be envisaged as one form of human capital formation and accumulation. Investment payoffs to the human capital accumulation of young adults include increases in expected productivity an earnings later in life.
The ability of young adults to successfully accumulate human capital depends in part on health status during the critical years in which most post-high school education and training typically occur, i.e., ages 18-30. Early age of onset illnesses and disorders associated with significant reduced physical, social and cognitive functioning are particularly burdensome. The costs of such disorders include not only direct and indirect medical costs, but also the reduced functioning and resulting impaired ability of the young to accumulate human capital, i.e., the costs of "lost" human capital. Detection and effective treatment of early onset disorders involving significant potential lost human capital could have very substantial economic benefits.
POPI WP #42-97
Ernst R. Berndt, Stan N. Finkelstein, Paul E. Greenberg, Alison Keith & Howard Bailit, May 1997
Health status can affect work performance in numerous ways. Economic analysis often focuses on long-term workplace outcomes that result from ill health, including changes in earnings, employment status, job choice, and promotion patterns. The medical and pharmacoeconomics literature also contain numerous examples of much more immediate consequences of illness in the workplace, in terms of excess absenteeism from work as well as impairment while at work.
Although illnesses can lead to numerous types of adverse effects with respect to labor market outcomes, many of the longer-term consequences are not borne directly by any particular business firm or organization but rather by society as a whole. As a result, there may be little incentive in many instances for a particular company to intervene on behalf of individuals in addressing these specific workplace concerns. In contrast, a substantial portion of the more immediate adverse implications of illness are routinely felt by employers.
POPI WP #41-97
Stewart C. Myers & Christopher D. Howe, April 1997
This report describes a financial simulation model of pharmaceutical research and development. It illustrates the major applications of the model, including calculation of the net present value of investment in R&D, of the cost of developing a successful drug, and of the changes in the cost of capital from drug discovery to the end of the drug's commercial life. It also presents estimates of the cost of capital based on analysis of the risk characteristics of pharmaceutical and biotech stocks.
The financial model connects the tools of modern finance to a detailed description of the costs, risks and returns of pharmaceutical R&D. We offer the model as a prototype for financial analysis of R&D investment and as a way of understanding the financial structure and performance of the pharmaceutical industry.
The calculations in this report generally reflect one set of base-case assumptions. Although the assumptions do not match any particular company's situation, we believe they are representative of conditions facing major, research-intensive pharmaceutical companies in the 1980s and early 1990s.
POPI Working Paper Abstracts: # 1-92 thru 20-94
POPI Working Paper Abstracts: # 21-94 thru 40-97
POPI Working Paper Abstracts: # 41-97 thru 60-01
POPI Working Paper Abstracts: # 61-01 thru 64-03
For more information, comments, or suggestions contact
Sloan School of