`Publisher: Institute for Operations Research and the Management Sciences (INFORMS)
`INFORMS is located in Maryland, USA
`
`Management Science
`
`Publication details, including instructions for authors and subscription information:
`http://pubsonline.informs.org
`The Debate on Influencing Doctors' Decisions: Are Drug
`Characteristics the Missing Link?
`
`Sriram Venkataraman, Stefan Stremersch,
`
`To cite this article:
`Sriram Venkataraman, Stefan Stremersch, (2007) The Debate on Influencing Doctors' Decisions: Are Drug Characteristics the
`Missing Link?. Management Science 53(11):1688-1701. http://dx.doi.org/10.1287/mnsc.1070.0718
`
`Full terms and conditions of use: http://pubsonline.informs.org/page/terms-and-conditions
`
`This article may be used only for the purposes of research, teaching, and/or private study. Commercial use
`or systematic downloading (by robots or other automatic processes) is prohibited without explicit Publisher
`approval, unless otherwise noted. For more information, contact permissions@informs.org.
`
`The Publisher does not warrant or guarantee the article’s accuracy, completeness, merchantability, fitness
`for a particular purpose, or non-infringement. Descriptions of, or references to, products or publications, or
`inclusion of an advertisement in this article, neither constitutes nor implies a guarantee, endorsement, or
`support of claims made of that product, publication, or service.
`
`Copyright © 2007, INFORMS
`
`Please scroll down for article—it is on subsequent pages
`
`INFORMS is the largest professional society in the world for professionals in the fields of operations research, management
`science, and analytics.
`For more information on INFORMS, its publications, membership, or meetings visit http://www.informs.org
`
`Opiant Exhibit 2168
`Nalox-1 Pharmaceuticals, LLC v. Opiant Pharmaceuticals, Inc.
`IPR2019-00694
`Page 1
`
`
`
`CORRECTED DECEMBER 3, 2007; SEE LAST PAGE
`MANAGEMENT SCIENCE
`Vol. 53, No. 11, November 2007, pp. 1688–1701
`issn 0025-1909(cid:1) eissn 1526-5501(cid:1) 07(cid:1) 5311(cid:1) 1688
`
`informs ®
`
`doi 10.1287/mnsc.1070.0718
`© 2007 INFORMS
`
`The Debate on Influencing Doctors’ Decisions:
`Are Drug Characteristics the Missing Link?
`Sriram Venkataraman
`Goizueta Business School, Emory University, Atlanta, Georgia 30322, svenka2@emory.edu
`Stefan Stremersch
`School of Economics, Erasmus University Rotterdam, 3000 DR Rotterdam, The Netherlands and
`Fuqua School of Business, Duke University, Durham, North Carolina 27708, stefan.stremersch@duke.edu
`
`Decision making by physicians on patients’ treatment has come under increased public scrutiny. In fact, there
`
`is a fair amount of debate on the effects of marketing actions of pharmaceutical firms toward physicians
`and their impact on physician prescription behavior. While some scholars find a strong and positive influence
`of marketing actions, some find only moderate effects, and others even find negative effects. Debate is also
`mounting on the role of other influencers (such as patient requests) in physician decision making, both on
`prescriptions and sample dispensing. The authors argue that one factor that may tip the balance in this debate
`is the role of drug characteristics, such as a drug’s effectiveness and a drug’s side effects.
`Using a unique data set, they show that marketing efforts—operationalized as detailing and symposium
`meetings of firms to physicians—and patient requests do affect physician decision making differentially across
`brands. Moreover, they find that the responsiveness of physicians’ decision making to marketing efforts and
`patient requests depends upon the drug’s effectiveness and side effects. This paper presents clear guidelines for
`public policy and managerial practice and envisions that the study of the role of drug characteristics, such as
`effectiveness and side effects, may lead to valuable insights in this surging public debate.
`Key words: physician decision making; marketing effort; patient request; drug effectiveness; side effect; drug
`prescription; sampling; sample dispensing; detailing; pharmaceuticals; public policy
`History: Accepted by Linda V. Green, public sector applications; received November 9, 2005. This paper was
`with the authors 6 months for 3 revisions. Published online in Articles in Advance September 14, 2007.
`
`Introduction
`1.
`Decision making by physicians regarding the drugs
`they treat patients with has come under increased
`scrutiny. As pharmaceutical expenses in the United
`States and other developed countries rise sharply
`with aging of
`the population, governments and
`regulators turn their attention to factors that may
`(adversely) affect physician drug decision making.
`Factors that draw particular attention are market-
`ing actions of pharmaceutical firms targeted directly
`at physicians and patient requests for a specific
`drug. “There has been a public outcry, especially in
`America, over the cozy relationship between doctors
`and drug companies. Some practices are illegal, others
`are simply part of the customary trio of food, flat-
`tery, and friendship” (The Economist 2005, p. 9). The
`prosecution of Merck for its marketing actions for the
`drug Vioxx is a very recent, heavily publicized, case in
`point, that regulators take notice (The Wall Street Jour-
`nal 2006).
`Pharmaceutical firms spend a huge and ever-
`increasing budget on detailing visits (sales calls by
`pharmaceutical representatives) and meetings. The
`number of sales representatives in the pharmaceutical
`industry has undergone a six-fold increase in the last
`
`20 years to approximately 100,000 today, and 77% of
`the companies are planning to further expand their
`sales force in 2005 (Hradecky 2004). Detailing (30.6%)
`and sampling (50.6%) to physicians amount to 81% of
`promotion spending by pharmaceutical firms in 2000
`(Rosenthal et al. 2003). In addition, patients increas-
`ingly request a certain brand of drug from the physi-
`cian. In the United States, one in three patients at
`some point has asked about a drug by name (Calabro
`2003). It is a commonly held belief that such patient
`requests are often triggered by direct-to-consumer
`(DTC) advertising, presently at an all-time high of $4
`billion in the United States (Edwards 2005).
`The most important decision of a physician, espe-
`cially if it concerns general practice physicians, is
`which drug to use in treatment of patients. The deci-
`sions physicians make on drug treatment can be wit-
`nessed through observing prescription behavior. They
`can also be observed in sampling behavior, as samples
`are provided together with a prescription (as a finan-
`cial subsidy to the patient), or instead of a prescription
`(as a trial, e.g., when uncertainty about drug-patient
`interaction is high). Sample dispensing by physicians
`is rarely studied. Sampling is an important physician
`decision as well, because sampling may lead to pre-
`
`1688
`
`Downloaded from informs.org by [130.115.156.138] on 31 May 2017, at 03:51 . For personal use only, all rights reserved.
`
`Opiant Exhibit 2168
`Nalox-1 Pharmaceuticals, LLC v. Opiant Pharmaceuticals, Inc.
`IPR2019-00694
`Page 2
`
`
`
`Venkataraman and Stremersch: The Debate on Influencing Doctors’ Decisions
`Management Science 53(11), pp. 1688–1701, © 2007 INFORMS
`
`1689
`
`scribed long-term treatment (Morelli and Koenigs-
`berg 1992), and thus have significant consequences for
`pharmaceutical firms and public health.
`Academic scholars and regulators have turned to
`assessing how both marketing actions of pharmaceu-
`tical firms and patient requests influence physician
`decision making on drug treatment, both prescription
`and sampling behavior. At this point, most research
`has been conducted on how marketing efforts tar-
`geted to physicians affect physicians’ prescription
`behavior. Patient requests as a factor influencing
`physician decision making and sampling as a physi-
`cian decision have received less attention so far.
`Even in the relatively developed research stream
`on marketing efforts and prescription behavior, con-
`troversy has been raised recently. While some studies
`(e.g., Gönül et al. 2001) find that detailing has a pos-
`itive and significant effect on prescriptions written,
`other studies find either a very modest effect (Mizik
`and Jacobson 2004) or no effect at all (Rosenthal et al.
`2003) of detailing on brand prescriptions or sales.
`Recently, Leeflang et al. (2004) posited that the rea-
`son for these incongruent results is that prior models
`may be misspecified, in that they pool the effect of
`marketing expenditures across brands, while brands
`may in fact differ in the extent to which physicians
`are responsive to the marketing expenditures a firm
`makes to promote them through detailing, meetings
`or other promotional instruments. This is also the
`stance we take in the present study.
`This study posits that drug characteristics, such as
`side effects and effectiveness, are a potential source
`for brand-specific differences, if any, in the respon-
`siveness of physicians’ brand prescription behavior
`to marketing efforts by pharmaceutical firms. Our
`insight may contribute to resolving the controversy
`on how marketing efforts of pharmaceutical firms
`affect prescription behavior. We also examine the role
`of these drug characteristics in the effect of other
`“influencers,” such as patient requests, and other
`physician decisions, such as sample dispensing. A
`coherent picture arises from our empirical analysis.
`We find that drug characteristics affect both the influ-
`ence patients (in this study through patient requests)
`as well as the pharmaceutical firms (in this study
`through their marketing efforts targeted to physi-
`cians) exert on physician decision making, both in
`a physician’s prescription and a physician’s sample-
`dispensing decisions. Thus, we underscore the impor-
`tance of including drug characteristics in any study
`of influence by firms and/or patients on any drug
`treatment decision a physician makes. By our knowl-
`edge, this study is the first attempt to test for inter-
`actions between influencers (e.g., detailing by the
`pharmaceutical firm) and drug characteristics (e.g.,
`efficacy) on physician behavior.
`
`For this study, we have composed a unique data
`set that matches three data sources. The first contains
`detailed information on manufacturers’ detailing vis-
`its to physicians, physician attendance at manufactur-
`ers’ meetings, and drug requests of patients for 2,774
`physicians in the United States, as well as the num-
`ber of prescriptions written and samples dispensed
`by each of these physicians on a monthly basis. The
`second and third data sets we composed ourselves.
`These contain data on (1) effectiveness, and (2) side
`effects of each drug in our database.
`The next section discusses the theoretical back-
`ground. Section 3 describes our data set and the
`analysis methodology we use. Section 4 presents our
`results. Section 5 discusses our findings, their implica-
`tions for public policy and management practice, and
`the study’s limitations.
`
`2. Background
`This section first discusses prior research on the
`effects of pharmaceutical firms’ marketing efforts on
`physician prescribing and explores their effects on
`sampling behavior by the physician, which until
`today remained unstudied. Second, we discuss the
`limited prior research on the effects of patient requests
`on physicians’ prescription and sample-dispensing
`behavior. Third, we explore the role that drug char-
`acteristics may play on physician decisions and their
`interactions with firms’ marketing efforts and patient
`requests. Fourth, we discuss any other relevant vari-
`ables that may affect physicians’ prescription and
`sample-dispensing behavior.
`
`2.1. Effects of Pharmaceutical Firms’ Marketing
`Efforts on Physician Prescription and
`Sample-Dispensing Behavior
`One can divide the prior literature regarding the
`effect of pharmaceutical firms’ marketing efforts on
`individual physicians’ prescription behavior into two
`streams, namely, one finding positive effects and one
`finding mixed effects, at best. We discuss each stream
`in turn.
`Gönül et al. (2001) and Manchanda and Chinta-
`gunta (2004) find that marketing efforts by pharma-
`ceutical companies to the physician positively affect
`prescriptions issued by a physician, but there are
`diminishing returns to detailing. Manchanda et al.
`(2004) find that detailing positively affects prescrip-
`tion behavior, but that high-volume physicians, while
`being detailed more, are less responsive to detailing,
`as compared to low-volume physicians. Narayanan
`and Manchanda (2004) find that while detailing influ-
`enced physicians positively in an overwhelming num-
`ber of cases,
`there was significant cross-sectional
`and temporal heterogeneity in physician responsive-
`ness to detailing. Janakiraman et al. (2005) find that
`
`Downloaded from informs.org by [130.115.156.138] on 31 May 2017, at 03:51 . For personal use only, all rights reserved.
`
`Opiant Exhibit 2168
`Nalox-1 Pharmaceuticals, LLC v. Opiant Pharmaceuticals, Inc.
`IPR2019-00694
`Page 3
`
`
`
`1690
`
`Venkataraman and Stremersch: The Debate on Influencing Doctors’ Decisions
`Management Science 53(11), pp. 1688–1701, © 2007 INFORMS
`
`nonpersistent physicians are responsive to both
`detailing and symposium meetings, while persistent
`physicians are only responsive to symposium meet-
`ings. Also, many studies that use aggregate (sales or
`prescription) data find a positive effect of detailing
`on drug sales (e.g., Chintagunta and Desiraju 2005;
`Narayanan et al. 2004, 2005; Neslin 2001; Rizzo 1999).
`According to the prior literature, firms’ market-
`ing efforts may have a positive effect on prescription
`behavior because detailing visits or symposium meet-
`ings provide information to the physician on efficacy
`and side effects of the drug (Gönül et al. 2001). In line
`with a long tradition in economics (e.g., Becker and
`Murphy 1993, Grossman and Shapiro 1984, Leffler
`1981), Narayanan et al. (2005) have argued that firms’
`marketing efforts may actually have both an informa-
`tive role (e.g., reducing cognitive uncertainty) and a
`persuasive role (e.g., inducing positive affect).
`Mizik and Jacobson (2004) find that marketing
`efforts by pharmaceutical companies to the physi-
`cian positively affect new prescriptions issued by
`a physician, but the effect sizes are very modest.
`Their findings cast doubt about a strong and positive
`effect of marketing efforts on physician prescription
`behavior as evidenced in studies using aggregate and
`individual-level data. Parsons and Vanden Abeele
`(1981) find that physician prescription behavior is
`quite unresponsive to marketing efforts by pharma-
`ceutical firms to the physician, and sales calls may
`even have a negative effect. Rosenthal et al. (2003) did
`not find robust and significant effects for detailing at
`the individual brand level.
`To the best of our knowledge, there has been no
`prior research that examines the effect of marketing
`efforts on sample-dispensing behavior by the physi-
`cian. The most useful research for our purposes is
`probably the sparse literature in medicine that exam-
`ines the motives physicians have when dispensing
`free samples to their patients. Motives that have been
`cited are: (1) financial savings for patients; (2) conve-
`nience; (3) initiate therapy immediately; (4) demon-
`strate the appropriate use to patients;
`(5) adjust
`prescribed doses before the full prescription is pur-
`chased; and (6) evaluate early effectiveness or adverse
`effects (Chew et al. 2000, Duffy et al. 2003).
`
`2.2. Effects of Patient Requests on Physician
`Prescription and Sample-Dispensing Behavior
`Most of the research that studies the effects of patient
`requests on physician decision making is driven by
`the growing importance of DTC advertising in the
`United States, mostly after the FDA’s 1997 Draft
`Guidance on DTC broadcast advertisements. DTC
`advertising is an important driver of patient requests
`(Mintzes et al. 2003), and scholars have only studied
`patient requests when triggered by DTC advertising,
`rather than any other reason.
`
`In a study using standardized patients that por-
`trayed major depression, 27% of all patients request-
`ing Paxil also received a prescription for it, 26%
`received an alternative antidepressant, and 47%
`received no antidepressant, while only 3% of patients
`with the same condition were prescribed Paxil if they
`did not explicitly request Paxil (Kravitz et al. 2005).
`Also, in other settings, scholars found a positive rela-
`tionship between patient requests and prescription
`(Kravitz et al. 2003, Lyles 2002, Mintzes et al. 2003)
`and physician referral (Kravitz et al. 2003). This pos-
`itive relationship is driven by patient pressure, and
`research has shown that when physicians do not com-
`ply with patient requests, patients are less satisfied
`with their physician visit (Kravitz et al. 2003).
`Underlying typical studies in this area is the notion
`that patient requests, especially if triggered by DTC
`advertising, are often for mild or trivial ailments
`(Weissman et al. 2004, Wilkes et al. 2000). Kravitz et al.
`(2003) found that subjective health distress predicted
`requests for physician services (referrals and prescrip-
`tions) more powerfully than did an objective count
`of chronic conditions, leading them to conclude that
`“requests may be driven more by anxiety than dis-
`ease burden” (p. 1680). To the best of our knowledge,
`no research exists that examines the effect of patient
`requests on sample dispensing by the physician.
`
`2.3. Moderating Role of Drug Characteristics
`Even though prior research has stated that drug char-
`acteristics may moderate the above effects, their role
`in the effect of firms’ marketing efforts and patients’
`requests on physician decision making remains unex-
`plored (Leeflang et al. 2004). While a drug can
`be characterized among many dimensions, such as
`its approved indications, its dosage, its potency, its
`administration method and frequency,
`its interac-
`tions with food and other drugs, its toxicity, and its
`price, in this first exploratory study we will focus on
`two very salient product characteristics, namely, the
`drug’s effectiveness and the drug’s side effects.
`A drug’s effectiveness is the extent to which the
`drug reduces the likelihood of negative clinical end-
`points. A drug’s side effects are secondary, and usu-
`ally adverse, effects of a drug. For instance, for statins,
`a drug’s effectiveness is the extent to which it reduces
`the likelihood of negative clinical endpoints, such as
`(fatal or nonfatal) myocardial infarction or coronary
`heart disease. The side effects statins may show are
`effects such as gastro-intestinal reactions, headaches,
`and nausea.
`Above, we referenced prior literature that found
`positive informative and persuasive effects of firms’
`marketing efforts on physician decision making. Now
`we explore the extent to which the effects of firms’
`marketing efforts on physician decision making may
`
`Downloaded from informs.org by [130.115.156.138] on 31 May 2017, at 03:51 . For personal use only, all rights reserved.
`
`Opiant Exhibit 2168
`Nalox-1 Pharmaceuticals, LLC v. Opiant Pharmaceuticals, Inc.
`IPR2019-00694
`Page 4
`
`
`
`Venkataraman and Stremersch: The Debate on Influencing Doctors’ Decisions
`Management Science 53(11), pp. 1688–1701, © 2007 INFORMS
`
`1691
`
`depend upon the drug’s effectiveness and side effects
`profile. When the firm promotes a more effective
`drug, as compared to a less effective drug, its abil-
`ity to lower physician uncertainty about the drug and
`increase physicians’ affect toward the drug is higher,
`as there will be stronger scientific evidence to back
`up the marketing effort (Azoulay 2002). The effect of
`the number of side effects on the relationship between
`a firm’s marketing effort and a physician’s decision
`making is more speculative. On the one hand, a drug
`with many side effects creates a high level of physi-
`cian uncertainty (e.g., on the interaction between all
`these side effects), which can be effectively reduced
`by firms’ marketing efforts, while a drug with few
`side effects creates a low level of physician uncer-
`tainty, thus reducing the need for—and the return
`on—uncertainty reduction through firms’ marketing
`efforts (Narayanan et al. 2005). On the other hand,
`it will be harder for firms to persuade physicians to
`treat patients with a drug that has a high number of
`side effects as compared to a drug with a low num-
`ber of side effects. Hence, the total interaction effect of
`side effects and a firm’s marketing efforts is difficult
`to predict ex ante, and hence is worthy of empirical
`investigation.
`As to patient requests, we also referred to prior
`literature that found patient requests to occur more
`often for mild conditions. Thus, we expect that patient
`requests for drugs with many side effects are honored
`by the physician in fewer cases than patient requests
`for drugs with few side effects. The reason is that
`drugs with many side effects may easily do more
`damage to the patient than the damage from the ini-
`tial mild condition (Kravitz et al. 2005). We expect
`that patient requests for drugs with higher effective-
`ness are honored by the physician in more cases than
`patient requests for drugs with lower effectiveness.
`On the one hand, a physician may react more posi-
`tively to an effective drug request as she or he has
`less uncertainty about the drug’s therapeutic value.
`On the other hand, a physician that reacts favorably to
`a patient request for an effective drug is more likely to
`receive favorable feedback afterwards than when he
`reacts favorably to a patient request for an ineffective
`drug. Given this feedback, the physician will increase
`his favorable reaction to patient requests, when it con-
`cerns the effective drug, and will decrease his favor-
`able reaction to patient requests, when it concerns the
`ineffective drug.
`Summarizing, we, a priori, expect the following:
`• Drug effectiveness may strengthen the effects
`of marketing efforts on prescription and sampling
`behavior by the physician.
`• Drug effectiveness may strengthen the effects of
`patient requests on prescription and sampling behav-
`ior by the physician.
`
`• Side effects of a drug may weaken or strengthen
`the effects of marketing efforts on prescription and
`sampling behavior (depending upon information—
`persuasion trade-off).
`• Side effects of a drug may weaken the effects of
`patient requests on prescription and sampling behav-
`ior by the physician.
`
`2.4. Other Variables
`We control for other variables, as well, that may af-
`fect prescription and sampling behavior. First, we
`control for the number of prescriptions and sam-
`ples for competing brands in the prescription model,
`while we control for competitive samples in the sam-
`pling model. Based on Mizik and Jacobson (2004),
`we expect that these effects may be positive or neg-
`ative, without a clear ex ante expectation. They may
`be negative as prescriptions and samples for compet-
`ing brands take away share of the focal brand (brand
`switching). They may also be positive, as increasing
`prescriptions and samples of competing brands can
`be indicative of growth in the drug category of the
`focal brand (category growth).
`Second, we control for the effect of sample dispens-
`ing of the own brand on prescriptions. This effect may
`be positive or negative, dependent upon the reason
`why the physician dispenses a sample (see above).
`Narayanan and Manchanda (2006) argue that a physi-
`cian may dispense a sample, as she or he is uncer-
`tain about a patient’s response to the focal drug. This
`would imply a negative contemporaneous effect of
`own samples on own prescriptions, as the sample
`comes at the expense of a prescription. On the other
`hand, Narayanan and Manchanda (2006) also argue
`that a physician may financially subsidize low-income
`or low-coverage patients through sample dispensing,
`in which case a drug prescription usually comes with
`a free sample. This would imply a positive contem-
`poraneous effect.
`Third, we control for carry-over effects, allowing
`these effects to interact with drug effectiveness and
`side effects. Physician persistence is an often observed
`phenomenon, driven by habit persistence and feed-
`back of patients (Janakiraman et al. 2005). We expect
`physician persistence to be more positive the more
`effective the drug is, as this will increase positive feed-
`back of patients to the physician. On the other hand,
`the more side effects the drug has, the more nega-
`tive feedback the physician will receive from patients,
`which in turn will lower physician persistence.
`
`3. Data and Analysis
`3.1. Data
`The data sets used for the empirical analysis in this
`study include (a) physician-level panel data, (b) drug-
`approval database, and (c) clinical trial reports. The
`
`Downloaded from informs.org by [130.115.156.138] on 31 May 2017, at 03:51 . For personal use only, all rights reserved.
`
`Opiant Exhibit 2168
`Nalox-1 Pharmaceuticals, LLC v. Opiant Pharmaceuticals, Inc.
`IPR2019-00694
`Page 5
`
`
`
`1692
`
`Venkataraman and Stremersch: The Debate on Influencing Doctors’ Decisions
`Management Science 53(11), pp. 1688–1701, © 2007 INFORMS
`
`Table 1
`
`Descriptive Statistics and Correlation Table
`
`Descriptive statistics
`
`Variable
`
`Mean
`
`Std. dev. Min. Max.
`
`Prescriptions Meeting
`
`Detailing
`
`Prescriptions
`Meeting
`Detailing
`Patient request
`Competitive prescription
`Samples
`Competitive samples
`
`1.23
`0.02
`0.73
`0.07
`3.69
`0.15
`0.46
`
`1.86
`0.17
`0.97
`0.48
`3.94
`0.50
`0.99
`
`0
`0
`0
`0
`0
`0
`0
`
`51
`18
`13
`41
`80
`19
`24
`
`1
`0(cid:1)05
`−0(cid:1)00
`0(cid:1)05
`0(cid:1)36
`0(cid:1)20
`0(cid:1)02
`
`1
`0(cid:1)04
`0
`0(cid:1)01
`0(cid:1)03
`
`
`00(cid:1)06
`
`(cid:1)00
`
`(cid:1)01
`
`(cid:1)03
`
`
`
`0(cid:1)17
`
`1
`−0(cid:1)02
`−0(cid:1)03
`−0(cid:1)04
`0
`
`Correlation table
`
`Patient
`request
`
`Competitive
`prescriptions
`
`Samples
`
`Competitive
`samples
`
`1
`−0(cid:1)03
`−0(cid:1)03
`0(cid:1)24
`
`(cid:1)02
`
`1
`
`1
`0
`
`1
`
`physician-level monthly panel data1 span two years
`(January 2002–December 2003) and come from a
`large firm that specializes in pharmaceutical mar-
`keting. Due to confidentiality agreements, we can-
`not reveal the data source. The data sets contain
`information on three therapeutic categories, namely,
`(1) statins, (2) gastrointestinal and coagulation drugs,
`and (3) erectile dysfunction (ED). The panel
`is a
`representative sample of physicians balanced across
`geographic regions, specialties, and prescription vol-
`umes. Monthly brand-specific physician-level vari-
`ables include total prescriptions written, total samples
`dispensed,
`total number of details,
`total number
`of meetings attended, and total number of patient
`requests. These data are collected directly from the
`physician office through an electronic database that
`collects prescription and detailing-call
`information.
`Unlike previously researched databases, our database
`has information on samples dispensed by the physi-
`cian,
`facilitating a more complete understanding
`of physician behavior across two key variables—
`prescriptions written and samples dispensed. We cali-
`brate our empirical model on the four most prescribed
`brands in each category. The shares of the focal brands
`are 85% in Category 1, 78% in Category 2, and 88%
`in Category 3.
`Our measures for drug characteristics, effectiveness,
`and side effects were constructed as follows. We
`obtained the number of side effects from the drug-
`approval database from the FDA that
`includes
`not only a history of drug-application filing dates,
`approval dates, and drug-innovation classifications,
`but also a list of side effects that is periodically
`updated when new indications and/or side effects are
`announced.
`
`1 Note that our physician-level database includes measures of mar-
`keting efforts and prescription data directly at the physician level.
`Due to institutional factors like availability of generics, insurance
`coverage, retail distribution, etc., data collected at the pharmacy
`might not accurately reflect actual physician behavior. Because we
`have access to direct measures of physician-level variables, we can
`get a more accurate picture of effects of marketing activities on
`physician behavior.
`
`We obtained drug effectiveness from a meta-
`analysis of clinical trial reports (source: National Insti-
`tute for Health and Clinical Excellence). This meta-
`analysis provides a standardized measurement of
`effectiveness, namely, a standardized Z-score mea-
`sure of the overall effectiveness of a brand relative to
`a placebo. Because these are standardized, the rela-
`tive effectiveness of brands can be compared directly.
`The measurements are explained in full detail in the
`online appendix (provided in the e-companion).2
`Table 1 provides the descriptive statistics and Pear-
`son correlations for the variables of interest. Table 1
`reflects variance in both the dependent variables of
`interest, i.e., prescriptions written (RX) and samples
`dispensed. The database includes, at the monthly
`level, all prescriptions within the examined drug cat-
`egories by a panel of 2,774 physicians. In all, we have
`39,880 observations.3 From Table 1, we also observe
`that the correlations among the independent variables
`are small, hence attenuating multicollinearity prob-
`lems in the analysis. No physician prescribes the same
`brand to all his or her patients.
`
`3.2. Analysis
`This section describes the empirical model. We begin
`by specifying the econometric model and end this sec-
`tion with a discussion on the estimation procedure.
`3.2.1. Model. To estimate the effects of market-
`ing activities on two physician decision variables—
`(a) prescriptions and (b) samples dispensed—we
`describe our estimated econometric model below.
`Note that the model we specify, given the intricacies
`of the available data, is a descriptive model that does
`not allow normative claims (Franses 2005).
`3.2.1.1. Dependent Variables. These include the
`total number of prescriptions (to new and previously
`diagnosed patients) written and the total number of
`
`2 An electronic companion to this paper is available as part of the
`online version that can be found at http://mansci.journal.informs.
`org/.
`is an unbalanced panel as we do not observe all
`3 Our panel
`physicians in the panel for the complete data window, which is
`24 months.
`
`Downloaded from informs.org by [130.115.156.138] on 31 May 2017, at 03:51 . For personal use only, all rights reserved.
`
`Opiant Exhibit 2168
`Nalox-1 Pharmaceuticals, LLC v. Opiant Pharmaceuticals, Inc.
`IPR2019-00694
`Page 6
`
`
`
`Venkataraman and Stremersch: The Debate on Influencing Doctors’ Decisions
`Management Science 53(11), pp. 1688–1701, © 2007 INFORMS
`
`1693
`
`samples dispensed of brand j at time t by physician p.4
`These are denoted as RXjpt and Samplesjpt, respectively.
`Independent Variables. As
`stipulated
`3.2.1.2.
`above, we study the effect of drug manufacturers’
`marketing efforts (through detailing and meetings)
`and patient requests on physician prescription and
`sample-dispensing decisions. Detailing effort by the
`manufacturer for brand j at time t to physician p,
`denoted by Detjpt, is measured as the total number
`of detailing calls made by the sales force for brand j
`to physician p at time t. In similar spirit, we define
`and denote meetings as the number of meetings
`organized by the manufacturer for brand j at time t
`that were attended by physician p, denoted by
`Meetjpt, and patient requests as the total number of
`patient requests for brand j at time t for physician p,
`denoted by Reqjpt. To accommodate carry-over or
`inertia effects of marketing and nonmarketing efforts
`as shown in Neslin (2001), we include a lagged
`prescriptions term in the conditional mean function.
`Because the markets we study include multiple
`competing brands, we also include prescriptions and
`samples dispensed for competing drugs for all major
`(=4 top brands) brands in the drug category of brand
`j to physician p at time t, and denote these variables
`by CompRxjpt and CompSamplesjpt.
`Our main theoretical interest lies in understand-
`ing how drug characteristics (effectiveness and side
`effects) affect physicians’ responsiveness to pharma-
`ceutical firms’ marketing efforts (detailing and meet-
`ings) and patient requests.
`We include effectiveness of brand j, denoted by
`Effj, which is based on numerous scientific studies
`that compare the effectiveness of brand j against a
`placebo. Consequently, a meta-analysis is conducted
`of an exhaustive set of studies (100+) for each drug
`to generate a meta-analytic Z-score statistic, as com-
`pared to a placebo, yielding our measure, Effj.
`We also include side effects of brand j denoted by
`SEjt, which is measured as the total number of side
`effects listed in the FDA-approved patient labeling for
`brand j at time t. Note that as new side effects surface,
`the drug goes through the label-certification process
`and new side effects are added to the previous list;
`thus, the list of side effects is ti