PDF A model of the role of free drug samples in physicians ...

Market Lett (2009) 20:15?29 DOI 10.1007/s11002-008-9045-2

A model of the role of free drug samples in physicians' prescription decisions

Kissan Joseph & Murali Mantrala

Published online: 7 June 2008 # Springer Science + Business Media, LLC 2008

Abstract Pharmaceutical companies distribute billions of dollars worth of free prescription drug samples every year giving rise to several concerns about their role in physicians' prescription decisions including: (1) making physicians view drugs with samples "gifted" to them more favorably than those without samples even though this is not justified on medical grounds; (2) inducing physicians to prescribe the sampled drug simply to garner the goodwill of their patients "happy to go home with free samples"; and (3) overly influencing less-experienced doctors. Interestingly, however, these popular concerns are not supported by findings of extant empirical studies in the medical literature. A review of these findings leads the authors to propose that rather than undermining objective prescription decision-making, free drug samples assist physicians find the best patient-drug match in settings characterized by diagnostic uncertainty. Based on a model of a competitive therapeutic category consisting of two differentiated brands and a physician who is focused on successfully treating his/her patients while minimizing associated costs, the authors show that samples can facilitate a prescription trial to resolve the attendant diagnostic uncertainty. The proposed model of the role of samples yields insights consistent with the findings in the empirical medical literature and also offers several implications for a firm's optimal allocation of free samples across physicians with varying levels of experience.

Keywords Pharmaceutical marketing . Samples . Promotion

Kissan Joseph gratefully acknowledges financial support from the General Research Fund of the University of Kansas. K. Joseph (*) School of Business, University of Kansas, 345 H Summerfield Hall, 1300 Sunnyside Avenue, Lawrence, KS 66045-7585, USA e-mail: kjoseph@ku.edu

M. Mantrala College of Business, University of Missouri-Columbia, Columbia, MO 65211, USA e-mail: mantralam@missouri.edu

16

Market Lett (2009) 20:15?29

1 Introduction

1.1 Motivation and research objectives

Pharmaceutical companies give away billions of dollars worth of free prescription drug samples every year amid much controversy about their impact on physicians' prescription decisions (Rabin 2007). Indeed, samples have been termed the "soul of selling in the prescription drug industry" (Liebman 1997). IMS Health reports that the retail value of samples distributed in 2003 exceeded $16 billion and that drug samples typically constitute the single largest item in the promotional budget of pharmaceutical firms. A typical firm may distribute as many as 10 million units of physician samples in a given year (Friedman 1995).

Critics of free drug sampling charge that this practice influences physicians to perceive drugs with samples more favorably than those without samples. In particular, "physicians don't realize the extent to which their medical judgment is influenced by their acceptance of the samples" (Rabin 2007). A second concern is that samples induce doctors to prescribe less-than-optimum drugs to simply garner patients' goodwill, i.e., "patients like going home with free samples and doctors are happy to oblige" (Rabin 2007). A third concern is that samples are focused on inducing relatively junior, inexperienced physicians (e.g., residents) to adopt newer, more expensive branded products despite the availability of cheaper, but equally effective older products (see, for example, Chew et al. 2000). Again, the implicit assumption is that samples are able to more strongly influence beliefs about drug efficacy among residents. However, these concerns overlook several key empirical research findings in the extant medical literature.

1. Samples influence physician choice even after controlling for physician perceptions with respect to efficacy and tolerability of drugs, factors that are in and of themselves important (Ubel et al. 2003). However, there are few differences in perceptions about the efficacy and tolerability of the sampled drugs between physicians who employ samples and their counterparts who choose not to employ samples (Ubel et al. 2003). This finding is not consistent with the charge that sample users develop more favorable perceptions of the sampled drug than is justified on medical grounds.

2. The intensity of sampling varies substantially across therapeutic categories (Backer et al. 2000). Some categories such as asthma and allergy medications, anti-infective agents, and analgesics are characterized by high levels of sample dispensation (18.7%, 16.9%, and 15.3% of all patient encounters, respectively) whereas other categories like lipid lowering therapies, anti-diabetic and thyroid medications, and topical dermatologic agents, have lower levels of sample dispensation (1.2%, 3.7%, and 4.9% of all patient encounters, respectively). This begs the question: if doctors are simply seeking goodwill, why is sample dispensation intensity not more uniform across therapeutic categories?

3. In a study conducted by Boltri et al. (2002) on physician choice between a branded drug and an older generic drug for hypertension, two periods were examined: one in which samples were available and another in which samples were restricted. In their study, the period of sample availability preceded the

Market Lett (2009) 20:15?29

17

period in which samples were restricted. During the period in which samples were available for the branded drug, experienced physicians (faculty) prescribed the branded drug in 65% of hypertension cases while inexperienced physicians (residents) prescribed it for 61% of hypertension cases. During the period in which no samples for the branded drug were available, faculty chose to prescribe it in 53% of patient encounters characterized by a diagnosis of hypertension while residents chose to prescribe the branded drug in only 28% of their patient encounters characterized by a diagnosis of hypertension. These findings reveal that inexperienced physicians do indeed demonstrate a much more pronounced incremental response to sampling by a brand; however, their baseline preference, as evidenced by their choice of that brand in a non-sampled environment, is substantially lower. This finding clearly calls for a much more nuanced understanding of physician response to samples, rather than simply assuming that samples alter perceptions of drug efficacy.

In this paper, we propose that the above medical research findings imply that rather than undermining objective prescription decision-making, free drug samples assist physicians find the best "match" between idiosyncratic patient characteristics and the available set of drugs in settings characterized by diagnostic uncertainty.1 In other words, we posit that a physician's diagnostic technology is inherently noisy and the treatment goal is to match the patient to the drug that is best suited for him or her.

Our singular focus on uncertainty to understand the role of free samples in physicians' prescription decision-making is consistent with its central role in extant models of pharmaceutical demand (e.g., Crawford and Shum 2005) and the mounting evidence that the probability of a successful patient-drug match even in the case of so-called established drugs is much lower than generally assumed. For example, (2005) reports that even blockbuster drugs, i.e., those with peak sales of $1 billion and above and prescribed for the general population use, are typically effective in only 40?60 percent of the patient population (see also Carey 2008). Indeed, a senior Glaxo executive has admitted recently that "..most prescription drugs do not work on most people who take them" (Connor 2003). This may be startling news to the lay public but physicians have long recognized that patients can respond very differently to the same medication. More specifically, the Personalized Medicine Coalition has assembled data which show that the percentages of the patient population for which any particular drug is ineffective ranges from 10?30% in the case of hypertension drugs to 40?70% in the case of asthma drugs (Personalized Medicine Coalition, November 2006.)

Our research considers a competitive therapeutic category comprised of two differentiated brands and a physician who is focused on successfully treating his/her patients while minimizing associated costs. We employ the idea that samples can facilitate a prescription trial to resolve the attendant diagnostic uncertainty. Our model delineates how the role of samples varies across therapeutic categories and

1 It is also commonly, but erroneously, believed that samples are provided to assist indigent patients. In a study based on survey data collected in a retrospective manner and reported in the American Journal of Public Health, Cutrona et al. (2008) find that neither insurance status nor income were predictors of the receipt of drug samples. In fact, poor and uninsured Americans are less likely than wealthy or insured Americans to receive free drug samples.

18

Market Lett (2009) 20:15?29

differing levels of physician skill (experience). In addition, in those instances where one brand is the sole provider of samples, we also delineate how the provision of samples can help such a brand increase its share. Thus, the proposed model of the role of samples explains the aforementioned empirical findings and offers several implications for optimal sample plans when a brand is able to exclusively furnish samples.

1.1.1 Positioning with respect to extant research in marketing

Surprisingly, despite the widespread use of pharmaceutical samples, there is very little research in the marketing literature that attempts to explicitly understand the micro-level processes through which samples influence physician behavior. Indeed, until recently, there was very little research on the role of free samples in general. This deficiency has been corrected for in consumer packaged goods products by the recent work of Bawa and Shoemaker (2004). However, their analysis is unsuitable for our purposes because our context is markedly different. Many of the features found for frequently purchased goods, such as purchase acceleration and category expansion, are not applicable in our context. Moreover, in the case of pharmaceutical samples, the sample is not provided to the final user but to an intermediate decisionmaker. This necessitates a different analysis altogether.

There is, of course, a substantial body of published and emerging empirical work that has successfully documented the market response to pharmaceutical samples. G?n?l et al. (2001) find, for example, that the provision of samples influences physician choice, albeit at a diminishing rate. Moreover, in their empirical work, they find that activities like detailing and samples fulfill an informative role rather than a persuasive role. Manchanda and Chintagunta (2004) find a positive main effect for samples but a negative interaction effect with detailing. The negative interaction effect is replicated in Manchanda et al. (2004). Finally, Mizik and Jacobson (2004) find that free drug samples have a positive impact; however, the magnitude of this effect is modest.

The rest of the paper is organized in the following manner. We first develop our model and allied assumptions. We then present the analysis of our model. Next, we delineate promotion implications stemming from our analytical model. We conclude by highlighting our contributions, implications, and directions for future research.

2 Model and assumptions

2.1 Basic model structure

We begin our model development by describing our assumptions, A1?A5. A1: Distribution of Patient Health States

Patients suffer from a particular health affliction (e.g., allergies, hyperlipidemia, hypertension, Type 2 diabetes, erectile dysfunction, etc) and seek treatment to improve their condition. Patients vary in their "health state", i.e., the manner in which they are afflicted by a particular disease and, consequently, their most suitable

Market Lett (2009) 20:15?29

19

course of treatment. For example, Type 2 diabetes patients typically vary with respect to their glycated hemoglobin levels, fasting plasma glucose (FPG) levels, body weight, proneness to side effects, and underlying complications. Depending on their overall health profile, some Type 2 patients may be suitably treated with generic Glucophage (Metformin) whereas others may be better matched to a branded anti-diabetic drug (e.g., Eli Lilly's ACTOS or GlaxoSmithKline's Avandia).] Of course, some patients may be equally cured by either drug.

In general, patients' health state is likely to be a multi-dimensional construct. However, in our model, we reduce it to a single dimension, . The relationship between and the effectiveness of the two drugs is as follows (please see Fig. 1). Patients drawn from the left side of the health state continuum are effectively treated only by Brand A. In contrast, patients drawn from the right side of the health state continuum are effectively treated only by Brand B. In addition, we include a zone wherein patients are effectively treated by either drug.2 In straightforward fashion, we denote this as the indifference zone, and it extends from D-d to D+d. Obviously, the promotional goal of both brands is to acquire patients from the indifference zone since all other patients are "captive" to one or the other brand. A2: Physician Diagnostic Technology: Noisy

We posit that physicians are endowed with a diagnostic technology that is noisy. We formalize this notion in the following manner: when a physician observes a patient with true health state , the diagnostic technology is such that s(he) views the patient's health state as arising from the uniform distribution, [ - , + ]. The parameter is a measure of noise in the diagnostic technology and is likely to vary by such factors as therapeutic category and physician skill. For example, conditions like hyperlipidemia are identified by laboratory tests that reveal the best course of therapy (e.g., statin or ezetimibe). In contrast, tests associated with allergies are less informative and some experimentation is required to reveal the best course of therapy. Finally, physician skill is also likely to reduce the noise in diagnosis. A3: Cost of Uncertainty to the Patient

We now specify the cost of diagnostic uncertainty to the patient. When the physician is uncertain about which drug will effect cure, s(he) cannot write a script for a full course of therapy. Rather, the physician is forced to conduct a trial. This, in turn, implies that the patient has to make a special trip to the pharmacy just to purchase a starter dose of the drug recommended by the physician. In this event, we assume that the patient bears a trip cost t. This cost may be thought of as the pecuniary (e.g., cost of transportation) and non-pecuniary costs (e.g., cost of time away from work) of making this special trip to the pharmacy. Now, although trip costs are borne by the patient, we include it in the doctor's calculus because trip costs adversely impact patient satisfaction. This, in turn, impacts patient retention. In this context, the benefit of the sample is that it can eliminate the trip cost to purchase the starter dose. The physician can simply provide a sample to conduct the trial, and the patient is spared a visit to the pharmacy.

2 In our conceptualization, assessment of the "effectiveness" of a drug includes its efficacy in treating the underlying condition as well as accounting for side-effects.

 ................
................

In order to avoid copyright disputes, this page is only a partial summary.

Google Online Preview   Download