Survey
* Your assessment is very important for improving the workof artificial intelligence, which forms the content of this project
* Your assessment is very important for improving the workof artificial intelligence, which forms the content of this project
Just What the Patient Ordered? Direct-to-Consumer Advertising and the Demand for Pharmaceutical Products Marta Wosińska∗ Harvard Business School Working Paper No. 03-058 Abstract Three out of four patients who ask their physician for a drug receive it. Do patients really wield so much influence? While empirical analysis of patient-level prescription claims reveals that advertising affects the choice probability, there are two caveats. First, promotions aimed directly at physicians affect prescription choice much more than promotions aimed at consumers, suggesting that advertising affects treatment probability thereby benefiting all brands. Second, advertising affects demand only for drugs that have preferred status with the patient’s insurer. The high ratio of fulfilled drug requests is driven less by patient’s influence than physician’s existing preference for these drugs. Keywords: advertising, detailing, spillovers, cost containment, prescription drugs, pharmaceutical products JEL Classification: I11, I18, M3 ∗ This paper is a revised version of Chapter 1 of my dissertation. I would like to thank my advisors Tülin Erdem, Paul Gertler, Benjamin Hermalin and Aviv Nevo for their suggestions and support. I would also like to acknowledge Pradeep Chintagunta, Paul Ruud, Kenneth Train and Dick Wittink for their helpful comments. Blue Shield of California made data available for this research. The Agency for Healthcare Research and Quality funded this project through a grant number R03 HS11600. The UC Berkeley Center for Health Research and the UC Berkeley Institute for Business and Economic Research provided additional financial support. Author’s contact information: Harvard Business School, Boston MA 02163, phone: (617) 495-6548, email: [email protected]. 1 Introduction Only five years ago, ads for prescription drugs were rare, but now some prescription drugs have advertising budgets that top familiar brands such as Pepsi, Budweiser, Dell or Nike.1 This is a significant departure from the pharmaceutical manufacturers’ traditional focus on physicians. While direct-to-consumer advertising (DTCA) still comprises a smaller share of the promotional budget than physician-directed promotions, it is the fastest growing promotional expenditure for pharmaceutical firms — advertising expenditures for all therapeutic categories rose from a meager $55 million in 1991 to $2.5 billion in 2000. There is a large public debate whether pharmaceutical manufacturers employ DTCA to indirectly influence physician drug choices. This debate is amplified by the double-digit growth in prescription drug expenditures.2 Physician organizations and managed care companies are concerned that DTCA induces patients, often price-insensitive as a result of insurance, to pressure their physicians to prescribe the advertised drugs. Yet, besides this supposed effect on treatment choice, DTCA may encourage patients to seek treatment and, therefore, increase the total market demand for a therapeutic class. Other potential effects may include increased purchase frequency by means of greater therapy compliance. Identifying these effects is important to the policy debate. If DTCA does not affect physician’s treatment decision, but only increases the number of patients receiving medically justified treatment, then DTCA might in fact be socially beneficial (in addition to being profitable through its effect on total demand). Empirical evidence on the role DTCA plays in the promotional mix is scant (Berndt 2001). What we know about the effect of drug advertising primarily comes from consumer and physician surveys. Perhaps the most widely cited set of surveys are those conducted yearly by Prevention magazine. The 1997 survey, for example, finds that 29% of consumers who saw a drug ad talked to their physician about it and asked for the drug to be prescribed to them. Doctors, in turn, honored 73% of those consumer requests. These findings have led to statements that “the current wave of direct-to-consumer advertising is putting patients, not [the doctor], in the diagnostic driver’s seat” (Maguire 1999). But these conclusions conflict with the findings of a concurrent study by Rosenthal, Berndt, Donahue, Epstein, and Frank (forthcoming). Rosenthal and her colleagues estimate category and brand advertising elasticities for several therapeutic categories. Their estimates of advertising elasticities are positive and significant on the category level but 1 2 Source: Competitive Media Reporting, Index of Leading National Advertisers, 2000. Source: 2000 National Health Accounts statistics complied by the Health Care Financing Administration. 1 not on the brand level. The discrepancy between their study and the common interpretation of Prevention magazine findings renders further study. This paper complements the current empirical evidence on the role DTCA has on demand for pharmaceutical products. I address several questions: does DTCA influence which drugs are prescribed by physicians? What is the role of DTCA vis-à-vis physician-directed marketing (so called detailing)? Does DTCA undermine the price sensitivity mechanisms used by health plans for cost containment? In particular, does DTCA have an impact on demand for drugs insurance companies try to contain? I present a framework for analyzing primary and secondary demand effects of both detailing and DTCA and develop testable hypotheses for secondary demand analysis. Specifically, if DTCA and detailing play a different role in primary and secondary demand expansion, it will be reflected in differing marginal effects on brand choice. I then address the above questions by estimating the probability a physician will prescribe a drug given its characteristics, one of which is its level of advertising. The empirical analysis uses a sample of prescription claims for a large health insurer, Blue Shield of California, over the period 1996-1999. These patient-level data are combined with monthly, brand-level DTCA data from Competitive Media Reporting. The person-to-person detailing expenditures (sales calls to physicians) and free sampling data, also monthly and brandlevel, were compiled by IMS Health. The data are for one therapeutic category — cholesterollowering drugs. This therapeutic category makes up a significant share of prescription expenditures and lacks over-the-counter substitutes thus making the choice solely the physician’s. I assume that the physician is the primary decision-maker and that his or her behavior is influenced by objectives of two principals: the patient and the insurance company. As a result, we may observe price sensitivity in the physician’s prescribing behavior. We may also observe DTCA having a discernable effect on prescription patterns. Results suggest that DTCA does have a significant positive influence on probability of choice — a $1 million increase in advertising increases market share by 0.5%. However, this effect is found only for drugs that have preferential status with the insurer (are listed on the formulary). Possibly physicians suggest that advertised formulary drugs be used when patients inquire about ads for drugs that are not listed on the studied formulary. The estimated marginal impact of detailing is significantly larger than the marginal impact of consumer advertising (on the order of five times). This concurs with an argument that DTCA affects total market size substantially more than detailing can — detailing can shift patients from non-drug therapies or increase testing 2 for a particular ailment, but it cannot bring untreated consumers into the office. In addition, it appears that DTCA has only a short-run effect on choice, while the effects of detailing wear out slowly over the course of a year. These results, coupled with the strong state dependence exhibited in the data, present a compelling picture of the pharmaceutical marketing strategy. Given that more than one drug is likely to work for a given patient, pharmaceutical firms aim to have patients initiate treatment with their brands. DTCA plays two roles — it generates foot traffic into doctors’ offices and it has some effect on prescription choice. However, once in the physician’s office, personal sales calls to physicians and free sampling that have the most significant effect on what drugs are chosen. This paper is structured as follows. Section 3.1 reviews relevant literature. Section 3 provides background information on marketing of pharmaceuticals, price sensitivity mechanisms employed by health insurers and the studied therapeutic category. Section 4 develops the conceptual framework for the primary and secondary demand effects of advertising and for the physician’s choice decision. It also describes the econometric method. Section 5 describes the data used in the study, while section 6 presents estimation results. A summary and policy implications conclude the paper in Section 7. 2 Existing Research Several literature strains are relevant to this study. The discussion of the primary and secondary advertising effects relate to a literature on so called predatory and cooperative advertising. Because one of the purposes of this paper is to determine whether DTCA undermines mechanisms used by insurers to induce price sensitivity in physician behavior, the relevant price sensitivity literature in markets other than health care is described briefly. I also address the literature on marketing of prescription medicines. In that context, research has focused on personal sales to physicians, while the impact of DTCA on demand is still poorly understood (Berndt 2001). Using terminology introduced by Friedman (1983), advertising can be predatory (increasing secondary demand, i.e. market share) or cooperative (increasing total market size).3 If firms are symmetric and advertising is perfectly cooperative, expenditures by one firm increase the market size in a way that equally benefits all the firms in the industry. This inability to fully 3 The term cooperative advertising is somewhat unfortunate as it implies coordination of actions across firms. Therefore beyond this section, I revert to using the primary (category expansion) and secondary (market share) terminology. 3 capture the returns to advertising is mitigated when a firm has a disproportionate market share (see Piga (1998) for a model with a technological asymmetry and Wosińska (2002) for a model with vertically and horizontally differentiated products). Discerning between market size and share effects of advertising has been of particular interest to researchers of the tobacco industry. Roberts and Samuelson (1988) provide one such early analysis. They report measures of output and advertising in the high-tar and low-tar markets and find that larger market shares have the largest advertising shares. They find a strong cooperative effect although they cannot reject the hypothesis that it has no effect on the market share. In a different context, Slade (1995) studies feature advertising4 in the saltine cracker market. Slade obtains estimates suggesting that a brand’s sales are increasing in own advertising and decreasing in rival-brand advertising. Advertising is thus predatory, although it has an overall positive effect on total market size. There are two views on how advertising increases own-brand sales. The first view is that advertising is informative. Many markets are characterized by imperfect consumer information, which leads to inefficiencies which advertising can help resolve. When a firm advertises, consumers receive direct information about the existence of a product or its imperfectly observable brand attributes (Nelson 1974). The firm’s demand becomes more elastic (which necessarily implies increased price sensitivity). The second view is that advertising is persuasive: it alters consumers’ tastes and creates spurious product differentiation and brand loyalty. The increase in differentiation results in lowered price elasticity and may generate barriers to entry for other brands (Comanor and Wilson 1979). A review of empirical marketing literature by Kaul and Wittink (1995) shows conflicting results — studies looking at non-price advertising usually find lowered price sensitivity. This presumed regularity, however, may be driven by a compositional bias that arises from ignoring heterogeneity (Kaul and Wittink 1995, Erdem, Keane, and Sun 2001).5 Ackerberg (2001) takes a different approach to distinguish between informative and prestige effects of advertising. Instead of relying on the interaction term of price and advertising, he considers the effect of advertising on new and experienced users. He argues that advertisements that give consumers product information should primarily affect consumers who have never tried the brand, whereas advertisements that create prestige or image effects should affect both inexperienced and 4 Feature advertisements are local newspaper inserts displaying in-store promotions. These feature ads are distinct from national media advertising studied in Roberts and Samuelson (1988) and this study. 5 More specifically, increased advertising can change the composition of consumers who buy the brand. Using a homogeneous choice model can bias price sensitivity results either way depending on whether less or more price sensitive consumers are drawn in when the brand advertises more. This paper accounts for consumer heterogeneity in price sensitivity thereby avoiding this potential bias. 4 experienced users. Empirical work on the competitive effects of pharmaceutical advertising to physicians has focused mainly on the relation between advertising and new product introductions.6 The results have been mixed — advertising has been found to have either a positive or negative impact on entry.7 The literature on the effects of physician detailing on firm-specific demand has, until recently, been much more limited. Berndt, Bui, Reily, and Urban (1994) attempt to distinguish between “industry-expanding” and “rivalrous” marketing efforts. They find that the impact of total marketing on the expansion of overall industry sales declines with the number of products in the market. King (1996) finds that own marketing reduces firms’ own price elasticities of demand. He also finds that total industry marketing reduces the degree of product differentiation and raises the cost of entry into the market. Rizzo (1999) similarly finds that both the stocks and flows of detailing expenditures decrease the price elasticity through the development of greater brand loyalty. Recently, several researchers have been able to tap into physician-level detailing data from pharmaceutical manufacturers. Manchanda, Chintagunta, and Gertzis (2000) find that detailing increases the number of prescriptions written by a physician. They also find diminishing returns to detailing for a majority of physicians in their sample. Gönül, Carter, Petrova, and Srinivasan (2001) find similar diminishing returns to detailing and find that detailing and free sampling increase price sensitivity (where price is the average retail price for the drug). A paper by Rosenthal, Berndt, Donahue, Epstein, and Frank (forthcoming) estimates category and brand advertising elasticities for several therapeutic categories including the cholesterollowering drug class studied in this paper. Their estimates of DTCA elasticities are positive and significant on the category level but not on brand level. They conclude that DTCA might positively affect class sales, but possibly not within-class product shares. Other empirical evidence comes from numerous surveys documenting the impact of consumer advertising on consumer and physician behavior. Gönül, Carter, and Wind (2000) find that consumers who have an ongoing need for health care (because of small children or a chronic condition) value drug advertising more highly. In turn, physicians with more years in practice, who see more patients, or have more exposure to pharmaceutical marketing are more accepting of DTCA. Wilkes, Bell, and Kravitz (2000) report that awareness of advertising was strongly associated with having been diagnosed 6 These papers use the term ”advertising” to indicate personal selling to physicians. Older papers (Vernon (1981), Telser (1975), and Leffler (1980)) find insignificant or positive impact of advertising on entry. Hurwitz and Caves (1988) find that advertising helped preserve incumbents’ market shares. Most recently, Ellison and Ellison (2000) provide evidence that advertising-based entry deterrence appears to be a profitable strategy in medium size pharmaceutical markets only. 7 5 with the advertised condition. In fact, 19% of their respondents who saw an ad actually asked the physician for the drug. The yearly Prevention magazine phone surveys that began in 1997 have consistently found that close to 30% of consumers who saw a drug ad talked to their physician about it and asked for the drug to be prescribed to them. Coupled with a finding that doctors honored roughly three quarters of those consumer requests, the survey estimates that close to 7.5 million consumers in the United States purchase prescription medications specifically because they have been exposed to advertising. Market research by Scott-Levin has found that doctor visits for heavily advertised drugs rose, on average, 11% between January and September of 1998, compared to a 2% increase in total office visits (Scott-Levin 1998). 3 Background 3.1 Marketing of prescription medicines Pharmaceutical marketing budgets reach 30 percent of sales. These marketing expenditures are largely driven by the large cost of maintaining and utilizing a significant sales force. In 1999, two manufacturers (Pfizer and GlaxoSmithKline) had over 8,000 full-time sales representatives each, and five others had over 4,000.8 IMS Health (2000) reports that physician-targeted advertising (in medical journals) and face-to-face selling (detailing) accounted for $2.7 billion in the first six months of 2000. This implies a 12% rise over first-half 1999 levels. Pharmaceutical companies’ investment in DTCA reached $1.3 billion in 2000, a 44.5% increase over the comparable sixmonths of the prior year. In addition to these direct promotion expenditures, pharmaceutical manufacturers sponsor conferences, training retreats, and provide physicians with substantial amounts of free samples. The retail value of pharmaceutical sampling distributed to office-based physicians by manufacturers’ sales representatives totaled $3.9 billion for January though June 2000. DTCA is a relatively recent phenomenon — it rose from $55 million in 1991 to $2.5 billion in 2000. It is the growth of managed care that may have inadvertently made advertising to consumers more appealing to pharmaceutical manufacturers. First, managed-care companies began to gain more influence over the physician prescribing process through various price sensitivity mechanisms described in Section 3.2. Second, the low patient out-of-pocket prescription costs that are characteristic of managed care introduced insurance-based moral hazard. If patients do 8 Source: “Relentless Rise in Role of Reps and Big Launches,” Financial Times, April 30, 2001. 6 not bear the cost of their drugs, then perhaps by extension, their physicians cannot be expected to appropriately balance benefits against costs — particularly if benefits are made to seem large through advertising. The rapid growth of DTCA may not have been possible without a strong signal from the Food and Drug Administration (FDA), the regulatory body for the pharmaceutical industry, that occurred with the August 1997 clarification of broadcast regulations. Until 1997, FDA regulations for pharmaceutical advertising required a summary of contraindications, side effects, effectiveness, and a “fair balance” of risks and benefits. The required “brief summary” often meant the inclusion of one to three additional pages of advertising space; pages which were rarely read or understood by consumers (Wilkes, Bell, and Kravitz 2000). Moreover, following these rules for broadcast media (radio and TV) was unrealistic.9 This changed in August of 1997 when the FDA allowed their use without the “brief summary.” These new ads can mention the brand name and treated conditions as long as they include limited in-broadcast disclosure of major side effects and mention a supporting national print ad campaign and a web site where viewers could turn for more information. The presentation and language in which the risks are communicated must be “reasonably comparable” to benefits and the ad cannot omit important information. The ad must make a reference to a health-care provider (the familiar “Talk to your doctor about drug X”) and provide a telephone contact number or an Internet site to which the patient can turn for more information.10 DTCA primarily involves products that treat chronic or episodic conditions rather than those that are acute. The advertised conditions also tend to have widespread incidence. DTCA is heavily concentrated among relatively few drugs — about 50 in year 2000. The most heavily advertised therapeutic categories are anti-arthritis medications, cholesterol reducers, anti-depressants, antiulcerants, anti-histamines, and anti-asthma drugs. While DTCA still comprises a smaller share of the promotional budget than detailing, it is by far the fastest growing promotional expenditure for pharmaceutical firms and, by 1996, had surpassed professional journal advertising. Within a particular therapeutic category, the newer drugs tend to be the ones that advertise. There 9 Requirements to supply the “brief summary” apply only to ads mentioning both the indication and brand name. Ads that urge consumers with a particular health concern to see their doctors (e.g., Rogaine’s ads were addressed to consumers concerned about hair loss) have not been subject to restrictions. Similarly, ads promoting certain brands without specifying the indication (“ask your doctor whether brand X is right for you”) have faced no constraints. 10 The ad copy need not be submitted to the FDA for approval. However, the FDA’s Division of Drug Marketing, Advertising and Communications can request that a campaign not meeting the requirements be discontinued or even remedied. 7 are no instances of DTCA of brands that face same-molecule generic competition, although two brands (Claritin and Prilosec) advertised heavily close to patent expiration presumably because the launch of their successors (Clarinex and Nexium respectively) was nearing. 3.2 Pharmacy benefit cost-containment methods On the other side of the table, insurers have been focusing on containing health care expenditures. The pharmacy benefit has recently become a primary target of cost-containment methods since it is the fastest growing health expenditure, with yearly growth rates reaching 20%. Two primary cost-containment methods involve the formulary and cost-sharing requirements. Insurers develop formularies, lists of preferred prescription drugs, to guide physicians in prescription writing.11 Formulary drugs are selected according to therapeutic value, general side effects profile, and cost. If successfully implemented, formularies can induce price sensitivity in physicians. To aid compliance with formulary recommendations, insurers may monitor prescription patterns of physicians. Those physicians whose prescribing patterns tend to deviate from the formulary are contacted and reminded of alternative, more cost-effective treatments. In certain cases, prior authorization or additional supporting documentation are required to prescribe an off-formulary drug. Some insurers also use financial incentives to influence physician prescribing. They may reward formulary compliance, generic substitution, meeting predetermined drug costs per patient, etc. Conversations with industry experts indicate that formulary compliance varies among organizations and tends to be higher in staff Health Maintenance Organizations (HMOs), which hire their physicians and pharmacists, whereas it can be a mere afterthought in more loosely organized Preferred Provider Organizations (PPOs) or Point-of-Service (POS) plans. In particular, physicians who treat PPO and POS patients deal with multiple insurers and, therefore, several formularies. One presumes multiple plans make it more challenging for physicians to remember the formulary placement of a given drug (whether it is listed on the formulary) for a given patient. Cost-sharing requirements mean consumers pay a portion of the cost of each prescription. Their objective is to raise consumer sensitivity to the real cost of their prescription and thus indirectly affect prescription choice. There are two forms of cost sharing: copayments and coinsurance. With copayments, consumers pay an established fixed-dollar amount each time they obtain a prescription. The simplest copayment structure is a uniform copayment (e.g., $5) for all 11 Many third-party payers outsource management of drug benefits to pharmacy benefit managers (PBMs), but the mechanisms that can be utilized for cost-containment remain the same. 8 prescriptions. Differential brand/generic copayments specify lower amounts for generic drugs and higher amounts for brand name drugs. It is now common for plans to include an additional third tier of copayment to differentiate the formulary and non-formulary drugs (the data used here has this kind of three-tier structure). The third tier copayment is often substantially higher than the copayment for branded, formulary-based drugs. As such, three-tier copayments can indirectly support the implementation of the formulary. Finally, under coinsurance, consumers pay a percentage of the cost of each prescription dispensed. That percentage typically does not vary by the type of drug dispensed (brand name, generic, or formulary). Coinsurance has been a traditional feature of indemnity-type insurance programs and is often combined with a deductible. 3.3 Cholesterol-lowering drugs This paper focuses on the effects of DTCA of drugs that treat hyperlipidemia (high cholesterol). In 2000, this therapeutic category ranked first in total sales ($9 billion), second in the number of dispensed prescriptions (96 million total prescriptions, of which 77 million were refills), and is among the most heavily advertised to consumers. This category lends itself well to the study of physicians’ therapy decision because no over-the-counter medications are available to treat cholesterol. Medical research has associated elevated serum cholesterol levels with coronary heart disease, the most common cause of death in the United States. The American Heart Association Science Advisory and Coordinating Committee has developed treatment guidelines for patients with lipid abnormalities. The recommendations relevant for this paper were issued in June 1993, almost three years prior to the studied period. Based on these guidelines, an estimated 13 million adults needed lipid-lowering drugs to meet recommended goals for low-density lipoprotein (LDL) levels (Sempos and et al. 1993).12 A certain level of cholesterol is necessary for the production of hormones and Vitamin D. Serum cholesterol levels are determined by the amounts of dietary cholesterol intake and by the cholesterol produced in the liver; thus diet and exercise alone may not be sufficient to lower serum cholesterol levels if there is significant overproduction of cholesterol. Various drug treatments attempt to block particular steps in the body’s synthesis of cholesterol. Drugs that target the earlier stages of synthesis can more effectively lower the production of cholesterol.13 There are three components that make up total cholesterol: low-density lipoprotein 12 On May 15, 2001, new cholesterol treatment guidelines were released. These guidelines suggest that 36 million Americans should have their hyperlipidemia treated with drug therapy. 13 Source: personal communication with Gabriela G. Loots, Ph.D., Lawrence Berkeley Laboratories. 9 (LDL or “bad” cholesterol), high-density lipoprotein (HDL or “good” cholesterol) and triglycerides (TG). Several subcategories of drugs treat hyperlipidemia: HMG-CoA reductase inhibitors (often called statins), bile acid sequestrants, fibric acid derivatives, and niacin. Statins are the newest and the most commonly prescribed lipid-lowering agents. They are generally effective and are supported by favorable outcome studies. There are no clinically appreciable differences in the safety profiles across these drugs and they have similar side-effect profiles. By 1998, six statins were available: Baycol, Lescol, Lipitor, Mevacor, Pravachol, and Zocor, all under patent protection.14 According to a review by the Harvard Heart Letter (May 1998), statins, as a group, decrease total and LDL cholesterol levels. Although all statins decrease triglyceride levels, not all are labelled by the FDA for this use. All statins have a minimal effect in raising HDL levels, but, again, not all are labelled for this indication. At least two studies that compared all the statins (except Baycol) have shown Lipitor to be the most effective in reducing LDL levels. However, unlike the more extensively studied Pravachol and Zocor, Lipitor has not been proven to reduce total morbidity and mortality. Furthermore, the Harvard Heart Letter review (1998) suggests that even if Lipitor truly is the most “powerful,” that does not necessarily make it the best choice for every patient. People with mildly or moderately elevated cholesterol might just as easily reach their target levels with any of the statins. Table 1 summarizes basic information about these statin drugs. In contrast, non-statin drugs are older and, in general, less effective in lowering LDL levels. Niacin is the oldest lipid-lowering agent and, until the recent introduction of extended-release tablets, it has been plagued by low compliance rates because of side effects. Fibric acid derivatives are used primarily to treat hypertriglyceridemia. They, however, are often associated with gastrointestinal intolerance and other side effects. A similar side effect profile affects bile acid sequestrant drugs. Given that these drugs have been on the market longer than the statins, both generic and branded versions of these drugs exist. Combination regimens (often a statin with a non-statin) can be considered for use in patients who fail to meet target cholesterol levels with one drug (but is recommended only for patients that are compliant with their current therapy). 14 On August 8, 2001, Baycol was removed from the market after, in combination with some drugs used to lower triglyceride levels, it was associated with some over 30 deaths in the United States. 10 4 Conceptual framework 4.1 Primary and secondary demand implications The data available for this study are conditional on choosing drug therapy. As a result, only market share effects can be estimated. However, the presence of two marketing instruments, DTCA and detailing, presents an opportunity to gauge both market share and market size effects. This subsection lays out the intuition for such comparison. Profit maximization implies that a firm with two marketing instruments, m1 and m2 , will allocate expenditures to equate the marginal products of the two instruments: ∂q ∂q = . ∂m1 ∂m2 (1) Both derivatives can be decomposed into a category expansion effect and a market share effect. If the firm’s demand q can be represented as q = sQ, where s is the firm’s market share (or probability of choice) and Q is the total category sales (purchase decision), then: ∂q ∂s ∂Q = Q+s . ∂m ∂m ∂m (2) This expansion is a reflection of the primary and secondary demand effects of advertising discussed in Section 3.1. In the context of DTCA, there are two reasons why advertising may increase category sales beyond own brand sales. First, many of the advertising signals contained in drug ads are category specific because the information provided in prescription drug ads is constrained by FDA regulations. If a manufacturer wants to run a product-claim ad that mentions the brand name and its indication, the ad must present accurate communication of product indications, claims, and risks, many of which are shared across brands.15 This effect may be magnified by the agency distinction further discussed in the next section — the physician, and not the patient, is the primary decision-maker. If advertising increases the probability that the patient will make an office visit, the impact of advertising by a particular brand will automatically spill over across the category. The simple decomposition in Equation 2 has implications that map well to those of models that explicitly account for competition between differentiated firms (e.g.,Piga (1998), Wosińska 15 Wosińska (2002) finds evidence of such informational spillover across brands in therapy compliance — advertising of Zocor improves average compliance of Lipitor patients and vice versa. 11 (2002)). If a firm’s market share approaches 100%, then ∂q ∂m approaches ∂Q 16 ∂m . In other words, a dominant firm is concerned with growing the total market because it can appropriate the effects of its advertising. On the other hand, as market share approaches zero, ∂q ∂m approaches ∂s ∂m Q. The small firm cares about the ability of the marketing instrument to influence market share. Equation 2 along with the equilibrium condition specified in Equation 1 provide insights into the primary and secondary demand effects of the two marketing variables. In particular, in equilibrium the following must hold: µ Q ∂s ∂s − ∂(doc) ∂(dtc) ¶ µ =s ∂Q ∂Q − ∂(dtc) ∂(doc) ¶ (3) where doc is doctor-directed marketing (detailing), and dtc is direct-to-consumer advertising. Casual observation suggests that detailing can only affect the treatment of patients already in the physician’s office, but its ability to induce patients to schedule office visits is limited if at all existent. We would then expect the difference on the right hand side of the equation to be positive. The research by Rosenthal and colleagues (forthcoming) is consistent with this story. ∂Q dtc ∂Q doc ∂(dtc) Q > ∂(doc) Q . ∂Q ∂Q implies that ∂(dtc) > ∂(doc) . They find that advertising elasticity is greater than detailing elasticity, i.e. Because mean detailing is significantly higher than mean DTCA, this This discussion has the following testable implications: • Hypothesis 1 : If DTCA affects secondary demand, then ∂s ∂(dtc) > 0, where s is the market share or the probability of choice for the advertised drug. • Hypothesis 2 : If detailing is primarily responsible for increasing market share while DTCA drives category expansion, then the marginal impact of detailing on market share is greater than the impact of DTCA: ∂s ∂(doc) > ∂s ∂(dtc) . The third hypothesis that DTCA undermines the formulary is developed next alongside the utility specification. 4.2 Physician utility specification Typically, the choice, payment, and consumption of a product fall to the consumer. However, in the pharmaceutical market, three distinct parties are involved: patients who consume the product, physicians who decide on the treatment, and insurers who are the primary payers. The 16 ∂s ∂m Presumably, when market share is very high, the ability of marketing to increase that high share is low, i.e. is small. 12 physician’s primary objective is, presumably, to maximize the health outcome for his patient as a function of patient characteristics and perceived and real drug attributes. The physician rarely knows what drug his patient will best respond to until she actually tries it. This underscores the importance of the patient’s drug treatment history and potentially of the initial choice of therapy. Physicians also form beliefs about drug characteristics from the medical literature, their colleagues and pharmaceutical sales people and the experience of other patients they have treated. If there is a conflict with physician’s objective, the patient and the insurance company attempt to modify the physician’s behavior. Patients may hassle their physicians to prescribe a particular drug they consider the best choice. Most relevant for this study, they may pressure their physicians to choose an advertised drug over a non-advertised drug. In addition, physicians do not incur expenditures for the prescribed drug and therefore any observed price sensitivity is a result of influence by the patient and the insurance company. Physicians are likely to act in ways to minimize these hassles. Insurers can choose from a set of incentive mechanisms for implementing the formulary described in Section 3.2, such as prior authorization or supporting documentation requirements. Alternatively, insurers may tie the copayment structure to the formulary status of a drug, and therefore attempt to induce price sensitivity through the patient. However, the consumer’s full price sensitivity may not be completely internalized by the physician. The above discussion can be summarized in the following indirect utility U that physician j derives from prescribing medication m to patient i on occasion t:17 Uijmt = αijm + βij1 DOCmt + βij2 HISTimt + βij3 CPimt + + βij4 DT Cmt + βij5 OF Fmt + βij6 DT Cmt OF Fmt + (4) + βij7 DOCmt OF Fmt + ²ijmt where the ij subscript combination on the β coefficients corresponds to a particular physicianpatient pair. The parameter αijm is an alternative-specific constant that captures unobserved, time-invariant characteristics, and presumably the individual-specific quality of the drug. The term DOCmt is the aggregate measure of physician sales calls of brand m during time period t, which captures the change in perceived quality, thus the expected sign on βij1 is positive. I initially conduct my analysis using advertising level expenditures and later present results for the wear-out rate of advertising (depreciation). The term HISTimt is a drug-therapy history construct 17 The subscripts on the independent variables reflect variation sources for the data used in this study. See Section 5 for more detail. 13 for patient i at time t. I use a simple history variable indicating whether the last prescription the patient received was for the same drug. One would expect strong positive state dependence in this market, i.e. βij2 > 0. The variable CPimt is patient i’s copayment for medication m prescribed on occasion t. As described in Section 3.2, the copayment may be a function of the retail pharmacy price (coinsurance) or a function of formulary status of a drug (copayment). In the latter case, the patient price does not vary with the price of individual drugs, but rather with their formulary/generic status. The data used in this study have the latter structure. The hassle cost constructs are as follows: DT Cmt is DTCA of brand m during the time period t and is constructed similarly as detailing, first as current level, and then as a discounted sum of current and past advertising. If patients exposed to DTCA induce pressure on physicians to prescribe the advertised drugs, then the βij4 coefficient will be positive (Hypothesis 1 ).18 Hypothesis 2 can be restated as βij1 > βij4 — the marginal impact of DTCA on choice is lower than the marginal impact of detailing. The variable OF Fmt is an indicator variable equal to 1 if drug m is off the formulary at time t. If formulary placement matters, then βij6 will be negative. The price and formulary effects may not be identified separately if copayment structures are a function of formulary status, unless there is sufficient variation in patient prices over time and/or across patients. Price sensitivity is discretized in the health insurance context — a patient faces at most three different prices with rare discrete increases over time. Because prices are based on the formulary status, an interaction term of DTCA with the off-formulary constant (DT Cmt OF Fmt ) best captures the effect od DTCA on price sensitivity.19 Theory provides little guidance on what effect advertising may have on price sensitivity. As discussed in Section 3.1, advertising can increase or decrease price sensitivity depending on whether the information conveyed in the ad is “soft” (i.e. image-building) or “hard.” Direct-to-consumer drug advertisements aim to build a preference for a particular drug but they also provide disease-specific and drug-specific information that may allow consumers to determine their “match” with the drug. However, there is a speculation in the managed care circles that advertising to consumers is aimed at getting around the formulary.20 In other words, DTCA can potentially be a way to increase sales of 18 This coefficient will also capture the direct effect consumer advertising has on the physician because of his or her own exposure to it. 19 No generics advertise therefore a DT CA ∗ generic interaction is meaningless. 20 This sentiment is best captured by a statement from Philip R. Alper, a UCSF physician: “What could be simpler than doing an end run around the cost-containment efforts of health managers and physicians by going directly to patients to create demand that will be difficult to refuse?” (excerpt from a letter to JAMA Editor). 14 drugs without trying to get on the formulary through pricing or superior quality. The inability to differentiate across the formulary status may also be the reason why Rosenthal et al. did not find a market share effect of DTCA. This strategic effect presents the following hypothesis: • Hypothesis 3 : if DTCA has the ability to undermine the formulary then βij4 + βij6 > 0.21 Physician-directed marketing, such as detailing, can similarly have a positive or negative impact on price sensitivity depending on the type of information it provides. King (1996) and Rizzo (1999) find that detailing lowers price elasticity of demand, while Gönül, Carter, Petrova, and Srinivasan (2001) find that it increases price sensitivity.22 Section 6 discusses how the interaction term between detailing and off-formulary status may be confounded by the aggregation of the detailing data. 4.3 Econometric method Assume that a physician will choose an alternative that maximizes the utility specified in Equation 5. If the ²ijmt stochastic error term is distributed extreme value, the probability of choice for alternative k by physician j for patient i from a choice set J at time t is: exp(xijkt βij ) m∈Jt exp(xijmt βij ) P robijt (k) = P (5) where xijmt = [1, DOCmt , HISTimt , ..., OF Fmt DOCmt ] and βij = [α, βij1 , ..., βij7 ]0 . A general model allows for tastes to vary in the population.23 In other words, each decision maker’s preference coefficient, βij , can be represented as the average preference in the population and the individual’s deviation from that mean: βij = β̃ij + β. The individual-specific coefficients are not known, but the individual deviations can be integrated out.24 Let f (β|θ) be the density with which tastes vary in the population, where θ are the parameters of this multivariate distribution such as the mean and standard deviation of tastes in the population. The choice probability that physician j will choose drug k (out of a set of J drugs at time t) for patient i is modified to: Z exp(xijkt βij ) P P robijt (k) = f (β|θ)dβ. (6) m∈Jt exp(xijmt βij ) 21 The sign of βij6 is indicative of the effect of DTCA on price sensitivity. Note that Hypothesis 3 will be satisfied for all positive and some negative values of βij6 . 22 These two findings are not necessarily contradictory. If detailing sufficiently increases the equilibrium quantity, an increase in price sensitivity will translate into a decrease in price elasticity. 23 The discussion of mixed logit borrows heavily from Revelt and Train (1998). 24 If there is no taste heterogeneity, the model collapses to the standard multinomial logit model. 15 The probability of each physician making a sequence of prescription decisions for each patient, conditional on βij ,25 is then Sij (βij ) = Y exp(xijkt βij ) . m∈Jt exp(xijmt βij ) P t (7) Since the values of βij are not known, the actual probability is the integral over the sequence Sij (βij ) over all values of βij : Z P robij (θ) = Sij (βij )f (β|θ)dβ. The log likelihood function is LL(θ) = P ij (8) ln(P robij (θ)) but since P robij (θ) does not have a closed form solution, it has to be approximated numerically through simulation (see e.g., Hajivassiliou (1993), and Hajivassiliou and Ruud (1994)). For a given value of the parameters θ, a value of βij is drawn from its distribution. Using this draw of βij , the probability of a sequence of decisions, Sij (βij ), is calculated.26 This process is repeated for many draws to calculate the simulated probability of a sequence of choices made by physician j for consumer i. The average of the resulting Sij (βij ) is calculated and taken as the approximate choice probability P robij (θ): Sij (βij ) = 1 R X r|θ Sij (βij ). (9) r=1,...,R r|θ where R is the number of draws of βij and βij is the rth draw from f (β|θ). We can then estimate parameters that maximize the simulated log-likelihood function defined as SLL(θ) = P ij ln(SPij (θ)). 5 Data The sample consists of 11,520 new prescriptions for 4,728 cholesterol patients enrolled in individual, family and employer group plans with Blue Shield of California (BSC) between May 1996 and September 1999.27 The observed prescription claims are a result of a series of decisions by the physician and the patient. The physician prescribes a drug. The patient chooses whether to fill the prescription, what type of pharmacy to use (mail-in or retail) and finally whether to pick up the prescription. The physician’s decision presumably depends on the patient’s fill decision, a 25 Note that the individual deviations do not have a time subscript. By fixing the random coefficient to be constant for each patient, patient-specific time correlation is generated. 26 I calculate the sequence of probabilities for each patient, rather than each patient-physician since there are few cases where multiple physicians treat one patient’s high cholesterol under the same plan. 27 I exclude refills because the patient’s decision to refill does not involve a choice between all brands. 16 hypothesis reflected by the presence of patient’s copayment in physician’s utility function. I also assume that the physician’s choice of drug does not depend on the patient’s choice of pharmacy. This is reasonable because all drugs are available in retail and mail-in pharmacies, and the ranking of prices is the same. This assumption allows me to use the retail pharmacy copayments only in analyzing the physician’s choice of drug.28 In its PPO plans, BSC utilizes few direct methods of formulary implementation. Physicians receive a quarterly newsletter from BSC Pharmacy and Therapeutics Committee with updated formulary information. Instead, BSC relies on the physician’s sensitivity to the patient’s price, and links copayments with the formulary status of drugs. BSC’s PPO patients face at most three prices independent of drug category. The lowest copayment (so called first tier) is for generic drugs, a higher one (second tier) for branded drugs on the formulary, and the highest one (third tier) for off-formulary branded drugs.29 One can construct the prices (copayments) faced by the patient for non-chosen alternatives using all the prescription claims (also non-cholesterol claims) submitted by the patient and his or her immediate family in a given month, since they all face the same copayments at a given time. First, using family-level data in a given month, one can identify unique copayments for the three drug groups. In the second stage, one can identify the runs of missing copayments for family time series and fill in these missing values with the end values for the run. Using this method, I am able to complete 60% of family-months for generics, 70% for brand name formulary-based drugs, but only 30% for non-formulary drugs (this percentage is higher for patients treated with cholesterollowering drugs — see Table 3). This results in a sample of 4,728 hyperlipidemic patients (with 11,520 prescription decisions), for whom these two steps generate complete benefits data. See Table 2 for an illustration of the imputation process. Some basic comparisons across the full sample and the sub-sample are presented in Table 3, as well as Figures 1 and 2. The construction method for the sample results in a greater share of patients with short treatment histories — 50% of sub-sample patients have filled only one cholesterol prescription during the 5/96-9/99 period. If patients were driven by advertising to try the cholesterol therapy, then the results may overestimate the impact of DTCA on choice. This bias is not necessarily a problem since Hypothesis 2 relies on a gap between the effectiveness of 28 Mail pharmacy copayments are higher per prescription, but lower per daily dosage. The most commonly used BSC mail copayment is a three-month supply (instead of one month) for double the price to the patient. 29 In 1997, BSC began phasing in this three-tier cost-sharing structure. Until then, the copayments differed only based on the generic status. 17 DTCA and detailing. An overestimation of the DTCA effect makes Hypothesis 2 more likely to be rejected. In addition, the sub-sample patients have a preference for off-formulary drugs. If these sub-sample patients are less price sensitive, we would expect a greater impact of advertising for off-formulary drugs and therefore more likely to find the strategic effect of advertising described in the previous section (Hypothesis 3 ). The choice set constructed for this study also requires some explanation. I consider the choice of drug, rather than the choice of a drug-specific dosage. While the choice of strength is an important part of the physician’s decision, the price sensitivity mechanisms (formulary and copayments) do not vary across dosages and advertising is not dosage-specific. Instead, the dosage considerations will be a factor of the severity of the patient’s hyperlipidemia. However, I do break the Zocor alternative into two separate alternatives because the 5 mg dose of Zocor was being taken on and off the formulary, while the Zocor dosages of 10 mg and higher were on the BSC formulary throughout the studied period (see Table 4 for more information).30 Besides the six statins, there were a total of 21 various non-statin cholesterol-lowering drugs utilized during the study period. Some of these drugs were used less than a dozen times, while several had utilization rates comparable to those of less utilized statins. In a cross-sectional analysis, it would be advisable to exclude the rarely utilized drugs, however in a panel setting, it is not obvious what to do with patients that have switched between these and more heavily utilized drugs. I create three price-based alternatives that incorporate these drugs: generic non-statins, brand name formulary non-statins, and brand name non-formulary non-statins. This results in a choice set of ten alternatives: seven statins and three non-statins. Table 4 provides more specific information on the formulary status for each of these alternatives. The prescription data are matched with monthly, brand-level DTCA expenditures compiled by Competitive Media Reporting, a marketing information firm. These data are broken down into ten media categories such as magazine, newspaper, and various forms of TV, radio and outdoor advertising. I use total dollar expenditures without disentangling different media impacts. This appears to be a reasonable assumption because the potentially more effective TV advertising is also more expensive. DTCA is largely national in nature, but the actual exposure may exhibit regional variation because of heterogeneity, for example, in TV viewing and magazine reading habits. Because advertising is not observed on an individual level, the DTCA coefficients reflect 30 No statin tablets are scored. While some higher dosage tablets can be cut in half accurately with a pill splitter, the 10 mg tablet of Zocor cannot be halved easily or consistently (Crouch 2001). 18 the patient’s exposure to advertising and her responsiveness to the ads, which she has seen. The separation of these two effects, exposure and responsiveness, is not important to the paper, however lack of individual-level advertising data lowers the dimensions that allow for identification of the advertising coefficients. Figure 3 indicates that only three statins, Pravachol, Zocor and Lipitor, have made substantive DTCA investments. While not all manufacturers have invested in drug advertising, those who have were experimenting with the appropriate levels (a fact confirmed by industry insiders). This experimentation resulted in the significant time-series variation observed in Figure 3.31 Brand-level monthly office promotions from IMS Health are used to control for marketing aimed directly at physicians. I use two distinct variables: the cost of sales calls to physicians and number of free samples left for physicians’ use.32 These data do not display the level of time variation seen with DTCA. In the short run, variation is driven by reallocation of detailing efforts across drugs in the company’s portfolio. In the long run, the firm can adjust the number of its trained sales representatives. In contrast to DTCA, using aggregate direct-to-physician marketing is problematic because pharmaceutical companies can tailor their detailing to the formulary of a particular insurer. If a physician treats many patients from one plan and a particular drug is not on that plan’s formulary, a salesperson may spend more time discussing that drug and may be more likely to provide free samples. If that is the case, then direct-to-physician spending measures will be overstated for formulary drugs and understated for non-formulary drugs. This effect can, however, be captured by brand fixed effects and the interaction term of detailing with the off-formulary indicator.33 6 Results Before turning to estimation of choice probabilities, it may be instructive to describe the trends in the prescribing patterns. The utilization of cholesterol-lowering drugs among BSC patients has been increasing rapidly — the number of prescriptions dispensed each month to BSC’s PPO patients increased by 300% between early 1996 and late 1999 while the total PPO enrollment 31 This experimentation aids identification of DTCA coefficients. However, this also suggests that firms may not be in equilibrium. Therefore, I do not attempt to calculate the magnitude of market size expansion (see Hypothesis 2 ). 32 Using free sampling in units, rather than dollars is preferable. The latter is the retail value of the drugs, which is higher than the actual marginal cost of these samples. 33 Using disaggregate detailing data would pose a possibly more significant econometric problem: physicians who are more responsive to detailing are likely to also allow salespeople to spend more time with them. 19 doubled. Figures 1 and 2 present market shares (calculated as a percentage of total prescriptions) for the sub-sample. These indicate that Lipitor has largely driven the above-mentioned utilization growth of cholesterol-reducing drugs. Also note the increased Pravachol market share in the subsample. Mevacor has had the most significant drop in the market share, largely driven by the fact that Merck, the manufacturer of Mevacor, has been phasing out that brand and increasingly promoting its other brand, Zocor. Table 5 shows the incidence of switching behavior among patients. It is apparent that a very high percentage of patients stay with only one therapy — a pattern that is confirmed in later analysis. 6.1 Consumer promotions only I initially restrict my analysis to those promotional activities aimed directly at consumers. These results are compared later with those that incorporate the physician-specific promotional methods. DTCA level is used as an explanatory variable. The robustness discussion in Section 6.2 presents results with more flexible models (lags or discounted cumulative sum of past advertising) and argues that the conclusions do not change qualitatively. Model 1 is the baseline mixed logit specification in which I allow for normally distributed random coefficients on price. The estimates on variables other than the copayment are similar to those in the multinomial logit, but the addition of the random copayment coefficient significantly improves the model’s explanatory power as suggested by the likelihood ratio test. DTCA is found to have a significant positive impact on choice probability, however this varies by formulary status.34 In particular, the effect of DTCA is estimated to be almost three times as large for formulary drugs than for those drugs that are not listed on the BSC formulary. One possibility for this differential effect is that the advertising strategies for drugs that happen to be not listed on the BSC formulary are not as effective as for the drugs that are listed. What is more likely the case, the physician will suggest an advertised formulary drug (Lescol, Lipitor or Zocor) when a patient asks for the off-formulary Pravachol or low dose Zocor. Thus it appears that the informational role of advertising outweighs the differentiating effect of advertising. The log-likelihood ratio test cannot reject the hypothesis that such a model is equivalent to one where only one price sensitivity interaction (DT Cmt OF Fmt ) is included. The baseline Model 1 also presents interesting results with respect to the price variable (co34 A similar model with random coefficients on DTCA yields estimates of standard deviations that are not statistically significant from zero. The log likelihood ratio test reject such a model is an improvement over the one with copayment and brand random coefficients. 20 payment). While the mean of price does not significantly change from that estimated with a multinomial logit (results not presented), the estimate of the standard deviation is surprising. A standard deviation of 0.04, with a mean of -0.005, implies that 45% of patients have positive price coefficients. This might be driven by the fact that the distribution of random coefficients is restricted to be symmetric around the mean. Alternatively, this pattern may be a result of highly restricted covariance structure — correlation across brands is only captured through the price random coefficient. When individual-specific quality is controlled for by incorporating brand random coefficients (Models 2-6), the estimate of standard deviation on price decreases substantially. The covariance structure can be made even more flexible by allowing for correlation in errors between particular options. I allow for correlation between statin drugs, Merck products (Zocor and Mevacor) and the various Zocor dosages. The addition of these parameters (not displayed) only affects the estimates of the standard deviation for brand-specific heterogeneity distributions. The remaining models in Table 6 incorporate a mixed logit specification that allows for random coefficients on alternative specific constants. Identification of such a model requires the scale of the model to be set; i.e., one of the alternative specific constants needs to be normalized, just as in the multinomial logit. In addition, there is the issue of normalizing one of the variance parameters. A recent working paper by Ben-Akiva, Bolduc, and Walker (2001) shows that this normalization is not arbitrary. They recommend normalizing the lowest variance parameter.35 Applying this strategy to my data reveals that Lipitor is the alternative with the lowest variance. This is problematic because it is not in the choice set at all times. The next lowest variance is that of generic non-statins. Therefore instead of excluding the generic constant (which is equivalent to setting the mean and variance to zero), I fix its mean to zero and its standard deviation to one. The addition of random brand coefficients allows me to account for heterogeneity in preferences over the brand specific quality and to obtain a cleaner estimate of the level of state dependence.36 The estimated coefficient on the treatment history variable (whether the patient last used the same drug) in Model 1 is overstated because the specification does not properly control for consumer heterogeneity in this specification, and thus confounds it with state dependence.37 With the addition of brand random coefficients, the marginal effect of having used the 35 Following the recommendations of that paper, estimating a model with all variances specified can identify the lowest variance parameter. This model “will pseudo-converge to a point that reflects the true covariance structure of the model. The heteroskedastic term with minimum estimated variance in the unidentified model is the minimum variance alternative” (Ben-Akiva, Bolduc, and Walker 2001). 36 More specifically, the random coefficients on alternative-specific constants properly account for heterogeneity only if the random coefficients are fixed across choice situations for each person, which is the approach I take. 37 See Heckman (1981) for an excellent exposition of the problem. 21 drug before decreases as the random coefficients on brand dummies now incorporate the time invariant, patient-specific mean utilities. The addition of the nine brand random effect parameters increases the log likelihood by 308, which is a significant improvement of the model as suggested by the log likelihood ratio test. Models 3 and 4 consider the possibility that the effect of advertising is modified by state dependence. I construct a new user indicator variable, equal to 1 for the first choice situation for each patient in the panel and interact it with DTCA.38 The net effect of advertising for new users is more than twice as large than for subsequent decisions (0.023+0.032=0.055 versus 0.023). The effect of DTCA of non-formulary drugs in the first prescription decision is 0.023+0.0320.024=0.031. This effect is significant and larger than for the formulary-based drugs in subsequent decisions. However, the effect of advertising on prescription choice for the subsequent decisions for non-formulary drugs is virtually zero (0.023-0.024=-0.001). Model 5 attempts to explore further the interaction of advertising, formulary status, and first choice situation. The resulting estimates indicate that the coefficient on the interaction term between DTCA and new user status in Model 3 was primarily driven by users of nonformulary drugs. Specifically, the estimate of the marginal effect for new users of formulary drugs is insignificant from zero, while the analogous estimate for new users for non-formulary drugs is highly significant and positive. The net effect of advertising in the first choice decision for off-formulary drug is then 0.051, while the net effect for subsequent decisions is negative. It may be puzzling, that an increase in DTCA by an off-formulary drug decreases the probability that it will be chosen by experienced users. This does not necessarily mean that the makers of drugs not listed on BSC’s formulary are not behaving optimally because Pravachol and the 5 mg Zocor are listed on formularies of many other insurance plans. For example, in 2001 Pravachol was listed on 24 formularies among a sample of 37 California health plans.39 From a firm’s perspective, the negative impact of the formulary is mitigated across plans.40 38 The new user variable is incorrectly overstated for many patients at the beginning of the panel, therefore I exclude the first four months from the analysis. This variable may still be overstated for patients who have been treated for cholesterol under a different insurance plan or who have discontinued therapy before 1996 and then returned to it after April 1996. 39 Source: Citizens for the Right to Know web site (http://ca.mcodrugs.com). 40 There is some evidence that own-brand advertising can lower compliance with prescribed therapy (Wosińska 2002) because of side effect warnings in the ads. A negative effect on choice observed only among experienced users is consistent with this side-effects story. 22 6.2 Consumer and physician promotions Now I consider the addition of detailing — the element missing from most other analyses of the impact of DTCA on demand. Table 8 presents some general statistics on detailing and free sampling. It is apparent that the three major brands that advertise heavily also promote heavily to physicians. However, some differences among the top brands exist. Namely, Pravachol has relied more heavily on DTCA, while Zocor or Lipitor have emphasized detailing and free sampling. Models 5 and 6 in Table 6 are the result of incorporating promotional activities directed at physicians. The addition of detailing expenditures in Model 5 lowers the estimate of the main effect of DTCA. The decrease in DTCA estimates across Models 4 and 5 translates into a 38% drop in the marginal impact on market share (choice probability). The 0.03 estimate implies that a $1 million increase in additional DTCA increases the market share of a formulary drug by 0.5%. Therefore Hypothesis 1 is confirmed — DTCA does influence the probability that an advertised drug will be prescribed. However, Hypothesis 3 remains unconfirmed — there is no evidence that advertising increases the prescription probability for drugs not listed on the formulary. Hypothesis 2 is also confirmed — DTCA appears to have significantly less impact on choice probabilities than physician advertising does. The main effect of detailing (DOC) in Model 6 is five times as high as the main DTCA effect (both variables have the same unit of $1 million).41 Note also that the addition of lagged free sampling activity in Model 6 virtually eliminates the significance of new user effects. This suggests the possibility that these “new” users might have been exposed to these drugs through free samples, effectively making them into experienced customers. In addition, the interaction terms of off-formulary constant with detailing and also with lagged free sampling are both significant and positive. In contrast to DTCA, detailing and free sampling of off-formulary drugs is more effective than that for formulary drugs. This may be, however, an artifact of the aggregation of the detailing and free sampling data, because pharmaceutical companies can tailor their detailing to the formulary of a particular insurer. If a particular drug is not on the formulary, a salesperson may spend more time discussing that drug and may be more likely to provide free samples, which would be reflected in a positive marginal effect of detailing and free sampling for off-formulary drugs. The increased emphasis on non41 Assuming that the effect detailing has on increasing the market size is zero, one could calculate the market size derivative with respect to DTCA. However, this would not be entirely appropriate in this context because the estimates obtained here are for a particular insurer. While this insurer is quite representative in terms of the price sensitivity mechanisms used and its ability to implement them, there is no reason to expect that even in equilibrium, this equality would hold exactly. 23 formulary drugs has been confirmed in conversations with industry experts.42 It also suggests that these two forms of promotions are powerful in their ability to undermine the formulary. This contrasts with the negative sign of the interaction term of off-formulary constant with DTCA, which indicates that DTCA is less effective for non-formulary drugs. Because the data are lacking to estimate a category expansion effect, I next consider alternative explanations for the found gap between the marginal effect of DTCA and detailing on choice. Let us first consider several possibilities that may bias the detailing estimate upward. First, if there are decreasing returns to detailing, then the true marginal effect of detailing may be lower. In fact, other studies find decreasing returns to detailing as in Manchanda, Chintagunta, and Gertzis (2000) or Gönül, Carter, Petrova, and Srinivasan (2001). I test for decreasing returns to detailing and find that the quadratic detailing term is negative and significant. However, this does not significantly affect the gap between detailing and consumer advertising — the marginal effect evaluated at most values of detailing remains at least four times as large. Second, the effects of past advertising are unlikely to dissipate within a month. Adding lags of DTCA and detailing reveals an interesting pattern: only current DTCA appears to affect choice probability, but the effects of detailing are long-lasting. In fact, the impact of current detailing is not estimated precisely and is rendered statistically insignificant. The lagged detailing expenditure estimates are on the order of 0.1, which is at least triple the estimate on current DTCA. This pattern is confirmed by conversations with physicians — physicians will consider putting several patients on a drug and then wait how well these patients respond before making any adjustments in general prescribing patterns. This pattern is also reflected in an alternative specification that includes discounted sum of detailing and DTCA. The discounted sum of DTCA (DDT Ct ) is defined as follows: DDT Ct = n X i δdtc DT Ct−i . i=0 The discounted sum of detailing is defined analogously.43 I estimate a separate discount parameter for DTCA and detailing. In that specification, I cannot reject that δdtc is zero. In turn, the detailing discount factor δdoc is estimated to be 0.97, which is consistent with the discount factor used in Gönül, Carter, Petrova, and Srinivasan (2001). This very low estimate of detailing wear42 Another possibility is that the detailing efforts for the drugs on the BSC formulary are of lower quality. Because the detailing data are available starting from May 1995, the number of months used in the discounted sum equals 12. This is equivalent to restricting the discount factor beyond one year to be zero. This restriction is no needed in the case of DTCA because full history is available. However, for consistency, only 12 months are used with the DTCA construct as well. 43 24 out rate should be put in light of the results uncovered with lagged detailing expenditure — the wear-out pattern indicated there is clearly not exponential since lagged detailing appears to matter more than current detailing. Nevertheless, the patterns of advertising and detailing effectiveness described above suggest that the same dollar expenditure on both marketing channels has varying ability to affect choice. 7 Conclusions and Implications Coupled with double-digit growth in prescription drug utilization, DTCA has caught the attention of many in the health sector and at various levels of government. A heated debate about the merits of such advertising has ensued. A core concern is that patients, the target audience for such ads, do not make brand choice decisions, which are the legislated domain of physicians. Neither is this audience responsible for the full cost of the prescription because of insurance. If patients pressure their physicians for the advertised drugs, inappropriate prescribing may result — patients may be treated for conditions that could be controlled with life-style changes or they may be prescribed brands that are not cost effective. Proponents of DTCA argue the opposite — DTCA increases treatment rates for underdiagnosed conditions and thus lowers total health care costs. This paper empirically investigates the role DTCA has on demand for prescription drugs. In particular, is DTCA effective in inducing patients to request advertised products from physicians? Furthermore, is DTCA effective in increasing market share of those drugs that managed care organizations try to control with such cost-containment methods as the formulary? Alternatively, is this marketing tool used because DTCA induces patients to seek treatment while physicians decide on treatment method? There are two primary findings: • DTCA increases the choice probability but only for drugs that are listed on the formulary. In other words, there is no evidence for the strategic effect to undermine the formulary that has been suggested by industry players (see Footnote 20). • The marginal impact of DTCA on prescription choice is significantly lower than the marginal impact of detailing. This differential impact on the choice component suggests that the primary role of DTCA lies in market size expansion. The market size expansion argument is supported by a steep rise in the total number of pre25 scriptions filled by Blue Shield patients. While the enrollment doubled, the number of cholesterol prescriptions rose 300% between early 1996 and late 1999, a time too short to find a significant cohort effect. Furthermore, the gap in the marginal effectiveness of DTCA and detailing is robust to specification and to sample selection arguments. In particular, decreasing returns to detailing and wear-out rates of promotional activity do not qualitatively affect the gap between DTCA and detailing effectiveness. The selection issues resulting from sub-sample construction and inclusion of only filled prescriptions may lead to overestimated impact of DTCA, which underscores the discrepancy in the marginal effect of these two marketing instruments. However, even if the effectiveness in of detailing and DTCA in driving market shares differ, Equation 3 may not hold and thus the DTCA market expansion argument may not follow. However, the consistency between this paper and Rosenthal et al. suggest that Equation 3 is a reasonable approximation. The differential role of DTCA and detailing has implications for firm strategy. If the primary role of DTCA is to increase the total market size, then the incentives to advertise to consumers will vary across firms. In particular, firms that best control the choice component (either through superior quality, pricing or detailing) will have the greatest incentive to spend on DTCA because they will appropriate a large share of these new patients. In general, the market leaders have advertised to consumers. Yet not all firms appear to have immediately realized the potential spillovers resulting from advertising to consumers. The case of cholesterol-lowering drugs is particularly instructive. The makers of Lipitor did not advertise this drug for well over a year after introduction. Instead, they focused on persuading physicians. At the same time, Bristol Myers heavily advertised its Pravachol to consumers. The total number of patients treated for cholesterol increased significantly, yet Pravachol’s advertising had little impact on its market share. Instead, Lipitor appropriated the vast majority of these new patients. These findings also have implications for prescription drug cost containment efforts. Currently the formulary is the primary tool for influencing physician choice. However, it is designed to affect brand choice and not the decision to prescribe. Given that DTCA primarily affects the latter decision, managed care organizations are left without a good tool to control such behavior. Despite the finding that DTCA is not effective for non-formulary drugs, excluding all the drugs from a formulary will not necessarily have the desired effect — if a physician wants the patient to walk out with a prescription, he will choose one of the non-formulary drugs. Raising copayments across the board might affect the purchase decision. In particular, differentiating the three-tier copayment structure across categories making, for example, the brand copayment higher only for 26 lifestyle drugs may give the desired containment of treatment rates. It is also important for health plans to realize that promotional variables other than the prominent drug advertising highly influence prescription choice. Free samples are of particular interest. The vast majority of patients do not switch brands if the prescribed therapy yields satisfactory results, therefore the initial choice of the drug is crucial. Given the observed strong brand loyalty, it is not surprising that pharmaceutical manufacturers shower physicians with free samples. Physicians often accept detailing visits because they do appreciate free samples they can later distribute to their patients. But if patients respond well to the free samples, they are highly likely to continue with that brand. This would suggest that starter packs might provide the extra incentive to choose formulary drugs or generics, if available, at the time of diagnosis. DTCA, on the other hand, is given too much credit for influencing what drugs are utilized most heavily. 27 References Ackerberg, D. A. (2001): “Empirically Distinguishing Informative and Prestige Effects of Advertising,” RAND Journal of Economics, 32(2), 316–333. Ben-Akiva, M., D. Bolduc, and J. Walker (2001): “Specification, Identification and Estimation of the Logit Kernel (or Continuous Mixed Logit) Model,” Unpublished manuscript. Berndt, E. (2001): “The U.S. Pharmaceutical Industry: Why Significant Growth in Times of Cost Containment?,” Health Affairs, 20(2), 110–114. Berndt, E., L. Bui, D. Reily, and G. Urban (1994): “The Roles of Marketing, Product Quality and Price Competition in the Growth and Composition of the U.S. Anti-Ulcer Drug Industry,” NBER Working Paper 4904. Comanor, W. S., and T. A. Wilson (1979): “The Effect of Advertising on Competition: A Survey,” Journal of Economic Literature, 17(2), 473–476. Crouch, M. (2001): “Effective Use of Statins to Prevent Coronary Hearth Disease,” American Family Physician, (http://www.aafp.org/afp/20010115/309.html). Ellison, G., and S. F. Ellison (2000): “Strategic Entry Deterrence and the Behavior of Pharmaceutical Incumbents Prior to Patent Expiration,” Unpublished manuscript, Massachusetts Institute of Technology, (http://web.mit.edu/gellison/www/drugs20.pdf). Erdem, T., M. Keane, and B. Sun (2001): “Advertising and Consumer Price Sensitivity in Experience Good Markets,” UC Berkeley working paper. Friedman, J. (1983): “Advertising and Oligopolistic Equilibrium,” The Bell Journal of Economics, 14, 464–473. Gönül, F., F. Carter, E. Petrova, and K. Srinivasan (2001): “Promotion of Prescription Drugs and Its Impact on Physicians’ Choice Behavior,” Journal of Marketing, 65(3), 79–90. Gönül, F., F. Carter, and J. Wind (2000): “What Kind of Patients and Physicians Value Direct-to-Consumer Advertising of Prescription Drugs?,” Health Care Management Science, 3, 215–226. 28 Hajivassiliou, V. (1993): “Simulation Estimation Methods for Limited Dependent Variable Models,” in Handbook of Statistics, ed. by G. Maddala, C. Rao, and H. Vinod, vol. 11, pp. 519–543. Amsterdam: Elsevier Science Publishers. Hajivassiliou, V., and P. Ruud (1994): “Classical Estimation Methods for LDV Models Using Simulation,” in Handbook of Econometrics, ed. by R. Engle, and D. McFadden, vol. 4, pp. 519–543. Amsterdam: Elsevier Science Publishers. Harvard Heart Letter (May 1998): “Statins Revisited,” . Heckman, J. (1981): “Heterogeneity and State Dependence,” in Studies of Labor Market, ed. by S. Rosen. University of Chicago Press. Hurwitz, M., and R. Caves (1988): “Persuasion or Information? Promotion and the Shares of Brand Name and Generic Pharmaceuticals,” Journal of Law and Economics, 31(2), 299–320. IMS Health (2000): “Product Samples Account for Nearly Half of Total Promotional Investment,” press release, (http://www.usimshealth.com). Kaul, A., and D. R. Wittink (1995): “Empirical Generalizations About the Impact of Advertising on Price Sensitivity and Price,” Marketing Science, 14(3), G151–G160. King, C. (1996): “Marketing, Product Differentiation, and Competition in the Pharmaceutical Industry,” Program on the Pharmaceutical Industry Working Paper No. 39-97, MIT Sloan School of Management. Leffler, K. (1980): “Persuasion or Information? The Economics on Prescription Drug Advertising,” Journal of Law and Economics, 24, 45. Maguire, P. (1999): “How Direct-to-Consumer Advertising Is Putting the Squeeze on Physicians,” Observer (newsletter of the American College of Physicians - American Society of Internal Medicine). Manchanda, P., P. Chintagunta, and S. Gertzis (2000): “Responsiveness of Physician Prescription Behavior to Salesforce Effort: An Individual Level Analysis,” University of Chicago working paper. Nelson, P. (1974): “Advertising as Information,” Journal of Political Economy, 82, 729–753. 29 Piga, C. (1998): “A Dynamic Model of Advertising and Product Differentiation,” Review of Industrial Organization, 13(5), 509–522. Revelt, D., and K. Train (1998): “Mixed Logit with Repeated Choices,” Review of Economics and Statistics, 80(4), 647–657. Rizzo, J. (1999): “Advertising and Competition in the Ethical Pharmaceutical Industry: The Case of Antihypertensive Drugs,” Journal of Law and Economics, 42(1), 89–116. Roberts, M. J., and L. Samuelson (1988): “An Empirical Analysis of Dynamic Non-price Competition in an Oligopolistic Industry,” Rand Journal of Economics, 19, 200–220. Rosenthal, M. B., E. R. Berndt, J. M. Donahue, A. Epstein, and R. G. Frank (forthcoming): Frontiers in Health Policy Researchchap. Demand Effects of Recent Changes in Prescription Drug Promotion. MIT Press. Scott-Levin (1998): “Patient Visits Up for DTC Conditions,” press release. Sempos, C., and et al. (1993): “Prevalence of High Blood Cholesterol among US Adults. An Update Based on Guidelines from the Second Report of the National Cholesterol Education Program Adult Treatment Panel,” Journal of the American Medical Association, 269, 3009– 3014. Slade, M. (1995): “Product Rivalry with Multiple Strategic Weapons: An Analysis of Price and Advertising Competition,” Journal of Economics and Management Strategy, 4, 445–476. Telser, L. e. (1975): “The Theory of Supply with Applications to the Ethical Pharmaceutical Industry,” Journal of Law and Economics, 18, 449. Vernon, J. (1981): “Concentration, Promotion, and Market Share Stability in the Pharmaceutical Industry,” Journal of Industrial Economics, 19(3), 246. Wilkes, M., R. Bell, and R. Kravitz (2000): “Direct-to-Consumer Prescription Drug Advertising: Trends, Impact and Implications,” Health Affairs, 19(2), 110–128. Wosińska, M. (2002): “The Economics of Prescription Drug Advertising,” Ph.D. thesis, University of California at Berkeley. 30 Table 1: Characteristics of statins Brand name Molecule name Marketed by Baycol Lescol Lipitor Mevacor Pravachol Zocor cerivastatin fluvastatin atorvastatin lovastatin pravastatin simvastatin Bayer Novartis Pfizer Merck Bristol-Myers Merck Available since 1/1998 4/1994 1/1997 9/1987 11/1991 2/1992 Approved by FDA to* ↓LDL, ↓TG ↓LDL ↓LDL, ↓TG, ↑HDL ↓LDL ↓LDL ↓LDL, ↓TG, ↑HDL ↓= decrease, ↑= increase * Note that lack of approval by the FDA for a particular indication does not preclude the drug from being prescribed for that problem. Table 2: Imputation of copayments Each patient faces at most 3 prices: pgeneric ≤ pf ormulary ≤ pnon−f ormulary . Patients that belong to the same family, face the same copayments. Step 1 - Using all prescriptions for the patient’s family, extract copayments for each tier. Date Generic copay Brand copay Non-formulary copay 5/97 8/97 9/97 10/97 11/97 12/97 1/98 2/98 3/98 4/98 5/98 . 25 . 5 25 45 . . 45 5 . . . . . 10 . . 10 25 . . . 45 . 25 . 10 . 50 . 25 . Step 2 - Fill in time series for each group. Date Generic copay Brand copay Non-formulary copay 5/97 8/97 9/97 10/97 11/97 12/97 1/98 2/98 3/98 4/98 5/98 . 25 . 5 25 45 5 25 45 5 25 45 . 25 45 10 25 45 10 25 45 10 25 45 10 25 . 10 25 50 . 25 . This patient is included in the estimation sample as long as no new prescriptions were filled in 5/97,11/97, 3/98 and 5/98. 31 Table 3: Summary statistics Full sample Variable Generic copay Brand formulary copay Brand non-formulary copay New user Last Rx was same drug Number of months in panel Year96 Year97 Year98 Year99 Number obs Mean St dev Min Max 51423 68890 40190 102680 102680 102680 102680 102680 102680 102680 8.02 17.23 26.72 0.28 0.34 30.15 0.11 0.25 0.32 0.30 2.65 6.87 13.23 0.45 0.27 14.93 0.32 0.43 0.46 0.45 0 0 0 0 0 1 0 0 0 0 45 100 135 1 1 48 1 1 1 1 Number obs Mean St dev Min Max 11520 11520 11520 11520 11520 11520 11520 11520 11520 11520 7.08 14.03 24.69 0.35 0.31 34.45 0.15 0.21 0.30 0.32 2.84 6.70 13.09 0.47 0.28 13.66 0.35 0.21 0.45 0.47 0 0 0 0 0 1 0 0 0 0 45 100 135 1 1 48 1 1 1 1 Sub-sample Variable Generic copay Brand formulary copay Brand non-formulary copay New user Last Rx was same drug Number of months in panel Year96 Year97 Year98 Year99 Table 4: Choice set description Choice Baycol Brand-name formulary non-statin Brand-name non-formulary non-statin Generic non-statin Lescol Lipitor Mevacor Pravachol Zocor 10 mg + Zocor 5 mg Zocor 5 mg Zocor 5 mg 32 Formulary status Added 10/99 ON OFF ON Added 4/96 Added 4/97 Removed 3/96 OFF ON ON in 1996 OFF in 1/97 - 4/98 ON in 4/98 Figure 4A: Blue Shield market shares: Full Sample 100% Contribution to total market 80% 60% 40% 20% 0% 5 7 9 11 1 3 5 7 9 11 1 3 5 7 9 11 1 3 5 7 9 1996 1996 1996 1996 1997 1997 1997 1997 1997 1997 1998 1998 1998 1998 1998 1998 1999 1999 1999 1999 1999 Baycol Lescol Lipitor Mevacor Pravachol Zocor Non-statin Figure 1: Market shares of cholesterol drugs (full sample) Figure 4B: Blue Shield market shares: Sub-sample 100% Contribution to total market 80% 60% 40% 20% 0% 5 7 9 11 1 3 5 7 9 11 1 3 5 7 9 11 1 3 5 7 9 1996 1996 1996 1996 1997 1997 1997 1997 1997 1997 1998 1998 1998 1998 1998 1998 1999 1999 1999 1999 1999 Baycol Lescol Lipitor Mevacor Pravachol Zocor Non-statin Figure 2: Market shares of cholesterol drugs (subsample) 33 30,000.00 Lipitor introduced Baycol introduced Advertising expenditure (in $000s) 25,000.00 20,000.00 15,000.00 10,000.00 5,000.00 0.00 1 4 1996 1996 7 1996 10 1996 1 1997 4 1997 7 10 1997 1997 Lescol Lipitor 1 1998 4 1998 Pravachol 7 1998 10 1 1998 1999 4 1999 7 1999 10 1999 Zocor Figure 3: Direct-to-consumer advertising of statins (1996-1999) (Source: Competitive Media Reporting) Table 5: Switching matrix Current alternative Lagged alternative Bayc Lesc Lip Mev Prav Zoc10 Zoc5 Gen Form Nonf Baycol 70 2 10 2 2 5 0 6 0 1 Lescol 3 255 40 3 19 16 2 8 4 2 Lipitor 6 20 1679 19 65 54 1 99 36 20 Mevacor 2 9 51 495 8 37 0 10 4 2 Pravachol 7 19 133 4 1024 62 4 57 21 4 Zocor 10mg+ 5 9 89 19 40 1071 9 38 20 11 Zocor 5mg 0 1 1 0 2 14 45 1 3 1 Generic 5 8 112 8 58 48 0 869 25 11 Formulary non-statin 0 2 34 2 17 18 1 22 310 8 Non-formulary non-statin 0 3 14 1 12 6 0 7 4 148 34 Table 6: Mixed logit estimation results Variable Last used same brand Patient’s copayment M M SD Off formulary M Current DTCA M DTC*OFF M Newuser*DTCA M Newuser* DTCA*OFF Current detailing Detailing*OFF M Lag free sampling FREE*OFF Log-likelihood Model 1 Model 2 Model 3 Model 4 Model 5 Model 6 3.072 *** (-0.014) -0.005 (0.003) 0.040 *** (0.005) -1.157 *** (0.137) 0.027 *** (0.007) -0.017 * (0.010) 2.228 *** (0.019) -0.009 ** (0.004) 0.013 (0.010) -1.919 *** (0.194) 0.039 *** (0.009) -0.023 * (0.013) 2.249 *** (0.019) -0.008 ** (0.004) 0.013 (0.010) -1.883 *** (0.194) 0.023 ** (0.011) -0.024 * (0.013) 0.032 *** (0.010) 2.248 *** (0.020) -0.008 ** (0.004) 0.014 (0.009) -1.902 *** (0.193) 0.046 *** (0.012) -0.073 *** (0.017) -0.015 (0.014) 0.093 *** (0.019) 2.226 *** (0.020) -0.010 ** (0.004) 0.017 * (0.009) -2.642 *** (0.342) 0.030 ** (0.012) -0.076 *** (0.017) -0.016 (0.014) 0.098 *** (0.019) 0.185 *** (0.050) 0.199 *** (0.064) 2.253 *** 0.018) -0.007 * 0.004) 0.005 (0.012) -2.745 *** (0.351) 0.029 ** (0.013) -0.052 *** (0.019) -0.016 (0.018) 0.041 (0.026) 0.152 *** (0.050) 0.187 *** (0.065) 0.038 ** (0.017) 0.171 *** (0.034) -14,115 -13,807 -13,802 -13,790 -13,753 -13,735 M M M M * significant at 10%; ** significant at 5%; *** significant at 1% M=estimated mean, SD=estimated standard deviation of heterogeneity distribution Dependent variable: chosen alternative Sample: 5/96-9/99, 4728 patients, 11520 observations Standard errors are below the estimates 300 random draws are used for mixed logit simulations Brand constant estimates are displayed in Table 7 35 Table 7: Mixed logit estimation results — continued Variable Baycol Lescol Lipitor Mevacor Pravachol Zocor 10mg+ Zocor 5mg Generic non-statin Brand non-statin Brand non-formulary non-statin Log-likelihood M SD M SD M SD M SD M SD M SD M SD M SD M SD M SD Model 1 Model 2 Model 3 Model 4 Model 5 Model 6 0.643 *** (0.162) -1.224 (0.793) 3.090 *** (0.508) -1.057 *** (0.188) 1.632 *** (0.157) 1.901 *** (0.069) 1.001 *** (0.134) 1.076 *** (0.252) 1.697 *** (0.147) 2.434 *** (0.218) 1.901 *** (0.146) 0.159 (0.115) 1.667 *** (0.129) -7.033 *** (0.983) 4.168 *** (0.485) set to 0 set to 1 -1.470 *** (0.235) 1.637 *** (0.186) 0.026 (0.325) 1.549 *** (0.180) -1.255 (0.793) 3.081 *** (0.507) -1.019 *** (0.185) 1.591 *** (0.156) 1.888 *** (0.069) 0.972 *** (0.135) 1.082 *** (0.250) 1.647 *** (0.145) 2.395 *** (0.218) 1.874 *** (0.146) 0.150 (0.115) 1.656 *** (0.129) -7.066 *** (0.996) 4.173 *** 0.493) set to 0 set to 1 -1.445 *** (0.234) 1.612 *** (0.186) 0.011 (0.325) 1.528 *** (0.181) -1.122 (0.776) 3.003 *** (0.500) -1.031 *** (0.187) 1.605 *** (0.157) 1.903 *** (0.070) 1.007 *** (0.135) 1.082 *** (0.251) 1.667 *** (0.146) 2.376 *** (0.219) 1.915 *** (0.149) 0.186 (0.114) 1.632 *** (0.128) -6.961 *** (0.995) 4.135 *** (0.493) set to 0 set to 1 -1.454 *** (0.236) 1.622 *** (0.187) 0.025 (0.325) 1.534 *** (0.180) -1.512 * (0.827) 3.093 *** (0.512) -1.494 *** (0.237) 1.658 *** (0.159) 1.033 *** (0.257) 1.076 *** (0.133) 1.542 *** (0.367) 1.799 *** (0.153) 1.896 *** (0.305) 1.895 *** (0.152) -0.322 * (0.184) 1.613 *** (0.131) -7.491 *** (1.003) 4.140 *** (0.486) set to 0 set to 1 -1.544 *** (0.236) 1.624 *** (0.185) 0.753 * (0.433) 1.333 *** (0.183) -1.584 * (0.828) 3.070 *** (0.510) -1.422 *** (0.231) 1.568 *** (0.158) 1.012 *** (0.257) 0.939 *** (0.141) 1.548 *** (0.368) 1.754 *** (0.151) 1.647 *** (0.310) 1.996 *** (0.154) -0.357 ** (0.183) 1.597 *** (0.130) -6.962 *** (0.969) 3.816 *** (0.476) set to 0 set to 1 -1.602 *** (0.239) 1.602 *** (0.186) 0.731 (0.448) 1.386 *** (0.191) -13,807 -13,802 -13,790 -13,753 -13,735 -0.308 *** (0.060) 1.538 *** (0.055) 0.980 *** (0.146) 1.889 *** (0.147) 0.528 *** (0.059) -1.465 *** (0.106) baseline -0.597 *** (0.065) 0.034 (0.160) -14,115 * significant at 10%; ** significant at 5%; *** significant at 1% M=estimated mean, SD=estimated standard deviation of heterogeneity distribution Dependent variable: chosen alternative Sample: 5/96-9/99, 4728 patients, 11520 observations Standard errors are below the estimates 300 random draws are used for mixed logit simulations 36 Table 8: Promotional expenditures in cholesterol-drug category (1996-1999) Brand Baycol Lescol Lipitor Mevacor Pravachol Zocor Generic non-statin Brand formulary non-statin Brand non-formulary non-statin DTCA* Free sampling** Detailing*** 0 290 3,332 0 5,959 3,473 0 0 0 1,746 1,505 5,870 708 3,159 4,234 0 611 650 2,540 2,301 4,896 434 3,421 3,225 <1 556 800 * Monthly average DTCA in millions of dollars ** Monthly average of tablets distributed as free samples (in millions of tablets) *** Monthly average detailing expenditures in millions of dollars Source: Competitive Media Reporting and IMS Health 37