Market Structure and Physician Relationships in the Joint Replacement Industry Anna Levine Harvard University May 2010 Abstract This article empirically examines how hospital market structure affects the prevalence and intensity of relationships between orthopedic joint replacement companies and physicians. These effects are identified using variation across cities in the number and size distribution of hospitals serving an area and information about the individual payments the top five orthopedic joint replacement companies made during 2007 to physicians in the United States. In particular, I quantify how geographic scope economies, competition between companies within a hospital, and competition between hospitals in the same geographic market affect the incentives for orthopedic joint replacement companies to develop financial relationships with the physicians of a particular hospital. My estimates suggest one way orthopedic joint replacement companies counterbalance the bargaining power of hospitals in concentrated hospital markets is through increasing their investments in relationships with physicians. In addition, my results suggest geographic scope economies in creating and maintaining these relationships may confer cost advantages to large firms thereby decreasing the ability of new orthopedic joint replacement firms to enter. 1 Introduction In many health care markets, physicians act as a gatekeepers between patients and firms producing new technology. This is particularly true for joint replacement (implant). Many orthopedists have preferences for and relationships with particular medical device companies (Abelson 2005). In an effort to control costs, hospitals are increasing attempting to negotiate more favorable agreements with implant suppliers. However when hospital administrators attempt to limit a physicians choice of implants they often face substantial push back from physicians. I would like to thank Michael Chernew and Scott Stern for helpful comments and suggestions. This research was funded in part by the Ewing Marion Kauffman Foundation and the Robert Wood Johnson Foundation. The contents of this paper are solely the responsibility of the author. 1
In recent years, there has been controversy about how the relationships between product manufactures and physicians affect physicians treatment choice. In addition to possibly influencing the static treatment choice of a physician, the strength of these relationships may also affect a hospital s ability to negotiate favorable prices with suppliers. When physicians have strong allegiances with a particular company a hospital may face push back if they try to exclude or limit the companies products. Also in markets where the costs to device manufacturers to create and maintain these relationships are large, these costs may discourage the entry of new device companies. I this article I document the level of direct financial payments from orthopedic joint replacement companies to physicians and identify the factors affecting a companies payments to physicians of a particular hospital. In particular, I quantify how geographic scope economies, competition between companies within a hospital, competition between hospitals in the same geographic market, as well as how overall hospital and patient market characteristics affect the incentives for medical device companies to develop financial relationships with the physicians of a particular hospital. In order to identify these affects I use variation across cities in the market power of hospitals serving an area as well as the level of financial investment the top 5 orthopedic joint replacement companies made in their relationships with the physicians who admit patients to each hospital in the United States during 2007. I find at that geographic scope economies in creating and maintaining these relationships confer cost advantages to large orthopedic implant firms. My results also show that when one firm in- creases their investment in relationships with the physicians of a hospital this decreases the propensity of other firms to invest in relationships with the physicians of that same hospital. These two empirical facts suggest an explanation for the relatively high concentration of sales in this industry. My estimates also suggest one way orthopedic joint replacement companies counterbalance the bargaining power of hospitals in concentrated hospital markets is through increasing their investments in relationships with physicians. These investments therefore may act to decrease to the bargaining power of hospitals in their negotiations with implant companies thereby increasing these hospitals overall costs. There have a been a number of studies that have examined the effects of medical device companies 2
and pharmaceutical firms relationships with physicians. In particular, von Hippel (1986) and Chatterji et.al (2008) demonstrate the importance of relationships between surgeons and medical device companies for developing new products and for product improvement. Complementary to this work, in this paper I show that in addition to influencing the direction and efficiency of the innovative process these relationships confer cost economies to larger firms may act as entry barriers to new or smaller firms in this industry. In Berndt et.al. they discuss the role marketing to physicians has in influencing the treatment choices of physicians. In this paper, I instead look at how these relationships may also affect hospitals procurement decisions and the ability of hospitals to economize on costs. In Sutton s (1991, 2007) work on endogenous sunk costs, he explains why in industries where firms can affect the perceived product quality of their products through higher expenditures on advertising, we may expect to see higher levels of industry consolidation. He explains how in larger markets where we would typically expect to see more firms entering we instead see fewer firms than expected but larger outlays in advertising. In this context as firms expect these large anticipated outlays on advertising after their entry in effect these advertising costs increase entry costs and industry consolidation. I discuss this argument in more detail later as this reasoning forms the basis for my later empirical analysis. In section 2, I describe the industry and the data used in the analysis. In section 3, I present motivation for the subsequent empirical estimation. In section 4, I present my empirical results. I conclude in section 5. 2 Industry Overview and Data Orthopedic joint replacement is an expensive and increasingly prevalent procedure. According a Department of Justice September 2007 news release, more than 700,000 total hip and knee replacement surgeries were performed in the U.S. and approximately two-thirds of these procedures were subject to Medicare reimbursement. A Department of Justice 2002-2008 summary reports in 2008 hip and knee surgical implants had combined sales of over $6.5 Billion. Furthermore, according to the 2008 3
Stryker 10-K report, the total orthopedic implant market in 2008 had sales of $12.9 billion this market is growing at an estimated rate of 10% per year. Orthopedists and the companies selling orthopedic implants often have close relationships. Physicians are sometimes involved in product design and improvement (Chatterji, et.al. 2008). Additionally there are often highly trained representatives from implant companies with the surgeons while they operate. Physicians are often advised by these company reps while they are operating and these reps are an important resource for surgeons. There has been controversy in recent years about the conflicts of interest created when physicians close relationships with the companies whose products they use. In September 2007 the United States Department of Justice (DOJ) entered into a deferred prosecution agreement with four joint replacement companies (Stryker, Depuy, Zimmer, and Biomet) to resolve criminal allegations against these companies. The allegations claimed that from 2002-2006 these companies entered into agreements with physicians whereby they paid these physicians to do little if any work in return for the physicians exclusively using the company s devices. As part of the deferred prosecution agreement these firms agreed to report on their websites all financial relationships with physicians for 2007 and 2008. Although no charges were filed against Stryker, another joint replacement company, Stryker entered into a non prosecution agreement agreeing to disclose the same information on their website. According to the DOJ these five firms produce 95% of all hip and knee surgical implants. The overall market shares of these top 5 firms in the total orthopedic implant market in 2008 were: Zimmer 25%, Depuy 21%, Stryker 19%, Smith & Nephew 12%, and Biomet 11% (Stryker 2008 10-k). During this period, on each of these 5 firms websites they posted both their direct payments, and in-kind payments (food, airfare, etc) disaggregated by physician. For my analysis I first collected payment information from these firms websites and then matched each physician listed with the hospitals they admit their patients to using a dataset from Health Marketing Science. In addition, for physicians not included in the Health Marketing Science data set, I placed phone calls to each of these physicians offices and inquired about which hospitals they admit their patients. Next, from the 2005 American Hospital Association database on hospitals I was able to obtain 4
information about the location of each hospital as well a variety of other information including the size and the ownership structure of each hospital. Additionally from the 2005 Area Resource File I was able to obtain county level demographics, including the median age and per capita income. Finally I used information from the US News and World report ranking of the top 50 hospitals in orthopedics, as a measure of hospital quality. The final dataset details the payments received by each physician from of the top five orthopedic implant companies during 2007 and 2008 and the hospitals these physicians admit their patients. In addition to payment information, I was able to create measures capturing the level of competition between hospitals in the same county. For example in my later analysis I use the Herfindahl-Hirschman Index of admissions in a county to measure the level of competition in the hospital market. Finally the dataset also includes county level demographic information. 2.1 Overview of Relationships In this section I will further describe the relationships between hospitals, medical device companies and physicians. Understanding the importance of the relationships between the different players in this industry will be important for motivating my later empirical analysis. 2.1.1 Hospitals and Medical Device Companies Orthopedic joint replacement is typically billed through a procedural payment, or diagnosis-related groups (DRG). These payments are negotiated with insurers and do not typically vary with the cost of the specific implant used. Therefore the hospital has a strong incentive to negotiate lower prices with suppliers. Unlike procurement for most medical devices, which is done through group purchasing organizations (GPOs),hospital administrators negotiate directly with suppliers over prices for high cost specialized devices such as orthopedic implants. 1 For these items, also referred to as physician preference items, physicians typically have strong preferences for a particular brand of implants and also have a larger say as to whether or not a hospital contracts with a particular company. In the past 1 GPOs are large companies that negotiate with many suppliers over prices and then in turn offer these low prices to hospitals. 5
most hospitals had contracts with most implant companies, however as cost pressure has risen hospitals increasing are looking to negotiate more aggressively for favorable pricing with implant suppliers (Wilson et.al. 2008). 2.1.2 Physicians and Implant Companies When a hospital tries to exclude a brand of implants there is often push back from the physicians who have been using these devices. Physicians have brand preferences for several reasons. First of all, a physician may think the products of one company are superior to the products of another company. In addition they may have a relationship with a particular company s representative whom they trust and are able to receive assistance from while they operate. In addition, as joint replacement surgeries are complicated having experience using a company s product may improve the quality of the surgery and decrease the amount of time it takes the surgeon to operate. Additionally, physicians may have consulting relationships with a particular device company. These relationships may involve the physician visiting the implant company s plants and advising these company s on their products. In addition, physicians may also be involved in product design and improvement. If a physician has a relationship with a company, and the hospital they admit their patients to makes it harder for the physician to use their preferred implant company s products, this may decrease whatever pecuniary and non-pecuniary benefit they derive from their relationship. 2.1.3 Physicians and Hospitals When a hospital negotiates with implant suppliers they first consult their physicians about their preferences. If a device is not on the approved or preferred list of devices for a hospital there is typically a lengthy process the physician must go through to be able to use the device. This process sometimes involves the physician making an argument to a hospital board as to why the requested device is better for the patient than then preferred device. If a physician s preferred company is excluded or not included in a preferred position the physician must either change the hospital they are admitting their patients to, or begin using a different company s implants. Needless to say when a hospital tries to restrict the use of a physician s preferred 6
implant companies products they often face strong push back from the surgeons. Therefore the costs associated with excluding a device depend not only on the level of product differentiation, but also on the strength of the relationships between the physicians of the hospital and the excluded device company. Increasing a physician s loyalty to a particular company may be a way to influence a hospitals contracting decision. 3 Motivation for Empirical Estimation In this section I will outline a basic economic framework I will use to first test if one of the motivations for implant companies investing in their relationships with physicians is to influence the ability of hospitals to negotiate favorable contracts with suppliers. In addition this framework will allow us to understand the extent to which these payments are a driver of scale economies for device companies and make it more difficult for new implant companies to enter this industry. The basic economic intuition behind the framework and the later empirical analysis is that device companies are competing for the allegiance of the physicians of a hospital. If a company does not have a strong enough relationships with the physicians of a hospital then the hospital may be able to limit the use of the company s devices. Therefore in hospitals where a device company has more to lose from being excluded, those in which there are more joint replacements, we would expect to see larger payments and more companies paying the physicians of these hospitals. In addition, when the hospital market in an area is more consolidated, the threat of being excluded to the device company may be more credible. Therefore if developing relationships with physicians is a way for device companies to increase the chance they will be included as a preferred provider by the hospital we would expect to see even larger payments in more consolidated hospital markets. 3.1 Economic framework In this section I provide a basic economic framework for the decision of a orthopedic implant company to invest in relationships with the physicians of a particular hospital. This framework will motivate later empirical analysis and the interpretation of the estimates from that analysis. 7
We consider the decision of a device company to develop relationships with the physicians of a hospital as a three stage entry game. In the first stage each implant company simultaneously chooses Enter or Don t Enter and incurs a set cost if they choose to enter. In the next stage each firm can incur an additional cost to increase the perceived product quality of their products u. An increase in perceived product quality increases physicians allegiance to particular company s product. Finally in the third stage all the firms who have entered choose quantities and compete (Cournot) for consumers. One of the insights from work on these types of models is that as the size of the product market increases, the expected return from investing in quality increases also. Hence firms increase their investment in the second stage thereby increasing the expected total entry costs by any firm in the first stage, discouraging additional entry (Sutton 1991, 2007). Empirically these models would suggest that as market size increases we would expect to see more investment in product quality by each entrant thereby increasing the expected entry cost of each firm in the first stage. We therefore would expect the market size threshold for entry of each additional firm to increase non-linearly. 3.2 Empirical Predictions In our empirical analysis we will consider the decision of each device company to invest in relationships with the physicians of a particular hospital. The direct payments from a medical device company to physicians of a hospital will serve as an empirical proxies for the presence of these relationships and the size of the payment will proxy for the intensity of the relationship. Investment in relationships with physicians is interpreted as increasing perceived product quality for the physicians of the hospital. The discussion in the previous section suggests that in hospitals with higher demand for orthopedic implant surgeries we would expect each firm that has decided to develop relationships with the physicians in a hospital to invest more in these relationships. This extra investment in product quality increases the expected entry costs of new firms. Therefore while we would expect more firms to be courting the physicians of hospitals where more implant surgeries are performed we would expect implant market size and the number of entrants to increase non linearly. As we do not have data on the number of joint replacement surgeries performed we will use proxies 8
for the demand for implants in a hospital. In particular will use the level of overall admissions and the age and wealth of the population in the surrounding area as proxies for the level of demand for orthopedic implant surgeries. In hospitals where the marginal return on investing in physician relationships increases and the marginal cost remains the same, we would expect firms to invest more in these relationships. Next, I will discuss reasons the marginal return and marginal cost of investing in physician relationships may shift. 3.2.1 Geographic Scope Economies When a firm has developed relationships with other nearby hospitals it may be less expensive to develop a sales force to work with the physicians of a hospital. Companies in this industry, as in many other industries, organize their salesforces at the regional level. Therefore the cost of deploying extra personnel to a particular hospital maybe lower if the company already has a presence in a nearby hospital. In order to take this into account, I will allow the level of a firm s investment in surrounding hospitals to shift the marginal cost of influencing perceived product quality and therefore we would expect payments from these companies that are able to realize cost economies to be larger. 3.2.2 Hospital Market Power When the hospital market in an area is more consolidated, and hospitals have more market power the threat to a device company of being excluded may be larger. Therefore if developing relationships with physicians is a way for device companies to increase the chance they will be included as a preferred provider by the hospital we would expect the marginal return on investing in their relationship with physicians to increase and therefore we would expect to see larger payments by device companies. As companies in the first stage would expect these higher outlays in the second stage it is unclear how hospital market power would affect the number of firms making payments in a particular hospital. In my empirical analysis I will include measures of the level of hospital market power in my regressions. I will use the Herfindahl-Hirschman Index (HHI) of hospital admissions as a proxy for the level of hospital market competition in geographic market. 9
3.2.3 Competition between Device Companies within the same hospital In my estimation I will allow for the investment of other implant companies in relationships with the physicians of a particular hospital to affect the incentives of a company to develop relationships with the physicians of the same hospital. From our previous discussion we would expect larger outlays of firms on their relationships with physicians to discourage other firms from courting the physicians of a hospital as when one company increases their payments to physicians other companies will have to increase their payment to these physicians as well thereby increasing their expected entry costs. 4 Empirical Results 5 Summary Statistics In this section I will first summarize the pattern of orthopedic implant companies direct payments to physicians. Next using the economic framework outlined in the previous section to guide my analysis I show how the geography of hospital market, the level of competition in the hospital market, the actions of rival implant companies, as well as market and hospital characteristics affect implant companies payments to physicians. In my analysis I will restrict attention to hospitals in counties with greater than 100,000 people. In addition as the deferred prosecution agreement was released in September of 2007 I restrict my analysis to payments from 2007 as after firms knew these payment would be public they may have changed their behavior. Table 1 presents a basic description of the hospitals used in the analysis. In this table, payments are summed over all physicians that admit patients to a particular hospital. If a physician admits patients to multiple hospitals the payments they receive are split equally across each hospital. Table 1 shows that on average the physicians of a hospital in my sample received a total of $48,147 in direct payments and $2,280 in kind payments in 2007 (sum across all five companies). Among the 794 hospitals whose physicians receive at least one payment from a company the average yearly direct payments were $216,847, and the average in kind payments totaled $10,127. In addition this table 10
shows that the propensity of each firm to pay physicians in each hospital increases with the 2008 market shares of the each company as reported on the Stryker 10-K report. Recall the market shares were: Zimmer 25%, Depuy 21%, Stryker 19%, Smith & Nephew 12%, and Biomet 11%. Congruently the proportion of hospitals each firm made payments to were: Zimmer 11.8%, Depuy 8.4%, Stryker 5.2%, Smith & Nephew 4.7%, and Biomet 3.9%. Figure 1 shows the aggregate payments from companies to hospitals in each state. Figures 2-6 show these payments adjusted for the population of the state and disaggregated by company. From these maps we see that while population explains some of the variation in payments across states it does not explain all of the variation. In addition we see that while Zimmer makes the most payments overall in some states other companies spend more than Zimmer. Figures 7, 8, and 9 show the payments in three similarly sized cities, Philadelphia, San Antonio, and San Diego respectively. In these maps the only hospitals shown are those whose physicians received at least one direct payment in 2007. The size of the circle in these maps corresponds to the overall size of the payments received by physicians in the hospital and the proportion of the circle of a particular color corresponds to the proportion of total payments made by each company. From these graphs we see that within a hospital there is typically one company that far outspends other companies, however the company making the most payments varies across cities. 6 Regression Results 6.0.4 Aggregate Implant Companies Payments First we examine what influences the number of firms that pay physicians in a particular hospital. In these next two regressions the unit of analysis is a hospital. In these regressions there is only one observation per hospital. The dependent variable in the first regression is the number of firms paying the physicians of the hospital, while in the next regression the dependent variable is the total payments received by the hospital from all five companies combined. The majority of hospitals in our sample have no payments from any joint replacement firms. Figure 10 shows the distribution of the number of firms that make payments to the physicians of each 11
hospital. Physicians in a total of 794 out of 3,580 hospitals in our sample receive payments from at least one joint replacement company. Table 4 show the results of an ordered logit predicting the number of firms that make payments to the physicians of a hospital. Larger and higher quality hospitals in more wealthy areas have more companies paying their physicians. However as we may have expected the increase in market size to induce an additional entrant into a hospital increases non linearly. Table 3 shows a results from a tobit regression of the total payments received by the physicians of the hospital. Again we see larger and higher quality hospitals in more wealthy areas have more companies paying the physicians that admit to them. 6.0.5 Individual Implant Company Payments In the next set of regressions I examine what drives implant companies payments to the physicians of a particular hospital. There are five observations for each hospital, one for each implant company. The dependent variable is either an indicator of whether or not any physician in a hospital received a direct payment from a specific company (there are 5 observations per hospital), or the total dollar amount of payments received by physicians who admit patients to a hospital by a particular company (again there are 5 observations per hospital). Payments are again aggregated over all physicians that admit patients to a hospital. If a physician admits patients to multiple hospitals the payments they receive are split equally across each hospital they admit their patients to. Many hospitals in the sample do not receive payments from any of the companies and therefore this later measure while continuous above zero, is censored from below at 0. Table 2 shows the average payment to the physicians of a hospital by a company in 2007 was $9,618. Conditional on receiving at least one payment from a company the average aggregate payment to the physicians of a hospital by a company was $141,360. Results from a tobit regression where the dependent variable is the size of payments from a company to the physicians of a particular hospital are shown in Table 5. This regression takes into account that payments are censored at zero. 12
6.0.6 Endogeneity We may be concerned that there may be an unobserved characteristic of the hospital that increases the payments of all firms to physicians of a hospital, for example unobserved hospital quality. In this case, in the previous regression the total payments of other firms to physicians of the same hospital would be an endogenous regressor. In order to correct for this endogeneity problem we need a variable that shifts the payments of other firms in a hospital but does not shift the firm whose payments we are trying to predict s payments except through its affect on the investments of other firms. In the next set of regressions we instrument for other companies payments to the same hospital using the payments the other firms made to other hospitals in the same county. Given the evidence of geographic scope economies, we would expect these variables to shift the cost function of the other firms to develop relationships with the physicians of a particular hospital. One rational for the existence of economies in geography is that they indicate the firm already has a sales force deployed in the area. 6.0.7 Instrumental Variables Estimation Using these instruments we re-estimate the previous model and find that while the sign and magnitude of the other regressors remain the similar, the effect of other firm s investment in a hospital changes sign. Therefore if we are predicting a firm s propensity to develop financial relationships with the physicians of a hospital, when other firms make payments to physicians of the hospital this decreases the propensity of the firm to court the same physicians. The coefficient estimates in Table 6 show that payments to a hospital increase with the payments the firm makes to other hospitals in the same county, suggesting the presence of geographic cost complementarities. In this table there are two coefficients that demonstrate this effect. The first coefficient is for a variable which summarizes a companies total direct payments to other hospitals in the same county, and the second variable is a dummy variable that indicates whether the company made any payments to any hospitals other hospitals in the same county. Both of these variables have positive and significant estimated coefficients. 13
In addition, the coefficients on the proxies for the number of orthopedic implant surgeries a hospital performs suggest that payments increase with the demand for implant surgeries. In particular payments increase with the quality and overall size of the hospital as well as with the median age and per capita income of the county. The negative coefficient on the squared hospital admissions shows that firms are more likely to pay more to physicians in larger hospitals, this effect is non linear as we would have expected. In addition, interestingly non profit hospitals tend to receive more money from the joint replacement firms. The coefficient on the HHI of hospital admissions is large and positive suggesting that the more concentrated admissions are in the county, that is the more market power a hospitals have in the patient market, more payments are made from implant companies to physicians. In other words these results suggest that when hospitals have more bargaining power, medical device companies respond by increasing their investments in relationships with physicians. Continuing the analysis we examine the binary decision of a firm to pay physicians of a hospital by running a probit regression using the same instruments and we get similar results as the tobit regressions. These results are shown in 7. 6.1 Summary of Results Several factors affect the intensity with which implant companies invest in their relationships with the physicians of a particular hospital. In particular, companies invest more heavily in these relationships in hospitals where larger number of implant surgeries are likely performed. Additionally, when firms also have strong relationships with the physicians in nearby hospitals they are able to realize cost economies and therefore invest more heavily in their relationships with physicians. These estimates also show that in markets where hospitals have more market power implant companies invest more heavily in their relationships with physicians. This suggests one way orthopedic joint replacement companies counterbalance the bargaining power of hospitals in concentrated hospital markets is through increasing their investments in relationships with physicians. However while we see that in concentrated hospital markets each firm has a higher propensity to invest in their relationships with physicians we do not find evidence that in more concentrated markets the total 14
number of firms that enter a hospital increase or that overall payments to each hospital increase. One explanation for this is that conditional on entry (in stage 2) each firm invests more in their relationships with the physicians of a hospital, thereby increasing the expected entry costs of each firm (in stage 1), and hence although there are stronger incentives to invest in relationships with physicians in these hospitals there aren t more firms investing, rather their are deeper relationships. Finally we see that when one firm increases the depth of their relationships with the physicians of a hospital this decreases the propensity of other firms to invest in relationships with physicians of the same hospital. This result coupled with the presence of economies of scale in geography suggest one of the entry barriers for small firms in this industry: large firms are able to realize cost economies in developing relationships with physicians and once a firm has developed a relationships with a group of physicians it is harder for other firms to court those same physicians. Therefore small firms both face higher costs to develop relationships with physicians because they do not have an established sales force in an area, and also because they must to overcome the influence of other companies who already have relationships with physicians. 7 Conclusions In this article I examine how the relationships orthopedic implant companies have with orthopedic surgeons affect the organization of these firms as well as how hospital market structure impacts the incentives of firms to invest in their relationships with physicians. I have shown geographic scope economies in creating and maintaining these relationships confer cost advantages to large firms. Additionally, when one device company increases their investment in relationships with the physicians of a hospital this decreases the propensity of other vendors to do the same. These two facts together show how relationships confer cost advantages to large firms and may decrease the ability of new implant companies to enter these markets. In addition, I have shown how implant company investments in these relationships are affected by hospital market structure. My estimates suggest one way orthopedic joint replacement companies counterbalance the bargaining power of hospitals in concentrated hospital markets is through 15
increasing their investments in relationships with physicians. These results suggest the relationships of companies with physicians may affect the ability of hospitals to negotiate favorable contracts with suppliers. Together these results suggest when considering the impact of the relationships of medical technology companies with physicians, we need not only consider how these relationships affect the static choice by a physician as to what device they will use for an individual patient but also how these relationships affect the way the industry is organized and how these relationships may affect the split of producer surplus along the supply chain. References Abelson, R. Possible Conflicts for Doctors are Seen on Medical Devices, The New York Times, September 22, 2005. Berndt, E., Bui, E., Reiley, D., Urban, G., The roles of marketing, product quality and price competition in the growth and composition of the US anti-ulcer drug industry In: Bresnahan, T., Gordon, R. (Eds.), The Economics of New Goods. University of Chicago Press, Chicago, 1997. Bresnahan, T. F., and Reiss, P. C., Entry and Competition in Concentrated Markets, Journal of Political Economy, vol. 99(5),(1991) 977-1009. Chatterji, A. K., K. Fabrizio, W. Mitchell, and K. Schulman, Physician-Industry Cooperation In the Medical Device Industry, Health Affairs, vol.27(6), (2008) 1532-1543. Department of Justice Office Report 2002-2008, www.justice.gov/usao/nj/press/press/files/pdffiles/finalw accessed November 2009. Department of Justice Press Release September 27, 2007, newark.fbi.gov/dojpressrel/2007/nk092707.htm, accessed November 2009. 16
Harris, G., B. Carey and J. Roberts, Psychiatrists, Children and Drug Industry s Role, New York Times, May 10, 2007. Sutton, J., 1991. Sunk Costs and Market Structure: Price Competition, Advertising, and the Evolution of Concentration, The MIT Press. Sutton, J., Market Structure: Theory and Evidence, in M. Armstrong and R. Porter, eds., Handbook of Industrial Organization, Vol. III, North Holland, 2007, 2301-2368. Stryker Corporation. 2008 Form 10-K. Filed February 20, 2009, accessed November 2009. US News and World Report Best Orthopedic Hospitals 2008, http://health.usnews.com/health/besthospitals/orthopedics-hospital-rankings/, accessed August 2009. Wilson, N., E. Schneller, K. Montgomery, and K. Bozic, Hip and Knee Implants: Current Trends and Policy Considerations, Health Affairs, vol.27(6), (2008) 1587-1598. Von Hippel, E., Lead Users: A Source of Novel Product Concepts, Management Science, vol.32(7), (1986) 17
Tables and Figures Table 1: Summary Statistics: Hospital An observation in this table is a hospital in a county with greater than 100,000 people. Only payments are from 2007 are used. Data on payments to physicians comes from company websites. Physicians are linked to the hospitals they admit patients to using a 2008 Health Marketing Science data set, and through additional phone calls to physician offices. Hospital information is from the American Hospital Association annual survey 2005, while information on county level demographics is from the Area Resource File 2005. 18
Table 2: Summary Statistics: Total Payments by Each of the Top Five Orthopedic Joint Replacement Companies to Physicians in Each Hospital: 2007 An observation in this table is a company, hospital pair where the hospital is in a county with greater than 100,000 people. Only payments are from 2007 are used. Data on payments to physicians comes from company websites. Physicians are linked to the hospitals they admit patients to using a 2008 Health Marketing Science data set, and through additional phone calls to physician offices. Hospital information is from the American Hospital Association annual survey 2005, while information on county level demographics is from the Area Resource File 2005. Table 3: Tobit Regression: Aggregate Direct Payments by the Top Five Orthopedic Joint Replacement Companies to Each Hospital: 2007 Marginal Effects from a tobit regression are shown. An observation in this regression is a hospital in a county with greater than 100,000 people. Dependent variable is the sum of the direct payments made by the top five orthopedic joint replacement companies to the physicians of a hospital in 2007. Data on payments to physicians comes from company websites. Physicians are linked to the hospitals they admit patients to using a 2008 Health Marketing Science data set, and through additional phone calls to physician offices. Hospital information is from the American Hospital Association annual survey 2005, while information on county level demographics is from the Area Resource File 2005. 19
Table 4: Ordered Logit Regression: Number of the Top Five Orthopedic Joint Replacement Companies that made Direct Payments to Physicians in Each Hospital: 2007 Coefficients from an ordered probit regression are shown. An observation in this regression is a hospital in a county with greater than 100,000 people. Dependent variable is a count (0-5) of the number of companies that made at least one direct payment to any physicians that admits patients to the hospital in 2007. Data on payments to physicians comes from company websites. Physicians are linked to the hospitals they admit patients to using a 2008 Health Marketing Science data set, and through additional phone calls to physician offices. Hospital information is from the American Hospital Association annual survey 2005, while information on county level demographics is from the Area Resource File 2005. Table 5: Tobit Regression: Total Payments by Each of the Top Five Orthopedic Joint Replacement Companies to Physicians in Each Hospital: 2007 Marginal Effects from a tobit regression are shown. An observation in this regression is a company, hospital pair where the hospital is in a county with greater than 100,000 people. Dependent variable is the total direct payment a company makes to the physicians of a hospital in 2007. Data on payments to physicians comes from company websites. Physicians are linked to the hospitals they admit patients to using a 2008 Health Marketing Science data set, and through additional phone calls to physician offices. Hospital information is from the American Hospital Association annual survey 2005, while information on county level demographics is from the Area Resource File 2005. 20
Table 6: Instrumental Variables Tobit Regression: Total Payments by Each of the Top Five Orthopedic Joint Replacement Companies to Physicians in Each Hospital: 2007 Marginal Effects from a IV tobit regression are shown. Then endogenous regressor is the direct payments other firms made to physicians of a hospital, and the instruments used are the total direct payments made by each of the other firms to other hospitals in the same county. An observation in this regression is a company, hospital pair where the hospital is in a county with greater than 100,000 people. Dependent variable is the total direct payment a company makes to the physicians of a hospital in 2007. Data on payments to physicians comes from company websites. Physicians are linked to the hospitals they admit patients to using a 2008 Health Marketing Science data set, and through additional phone calls to physician offices. Hospital information is from the American Hospital Association annual survey 2005, while information on county level demographics is from the Area Resource File 2005. 21
Table 7: Instrumental Variables Probit Regression: Total Payments by Each of the Top Five Orthopedic Joint Replacement Companies to Physicians in Each Hospital: 2007 Marginal Effects from a IV probit regression are shown. Then endogenous regressor is the payments other firms made to physicians of a hospital, and the instruments used are the total payments made by each of the other firms to other hospitals in the same county. An observation in this regression is a company, hospital pair where the hospital is in a county with greater than 100,000 people. Dependent variable is an indicator variable which equals one if the company makes at least one direct payment to any physicians that admits patients a hospital in 2007. Data on payments to physicians comes from company websites. Physicians are linked to the hospitals they admit patients to using a 2008 Health Marketing Science data set, and through additional phone calls to physician offices. Hospital information is from the American Hospital Association annual survey 2005, while information on county level demographics is from the Area Resource File 2005. Figure 1: Total Direct Payments by Top Five Orthopedic Joint Replacement Companies to Physicians by State in 2007 Information on Payments comes from company websites. 22
Figure 2: Per Capita Direct Payments by State 2007 - Zimmer Information on Payments comes from company websites, population information comes from the 2005 Area Resource File. Figure 3: Per Capita Direct Payments by State 2007 - Depuy Information on Payments comes from company websites, population information comes from the 2005 Area Resource File. 23
Figure 4: Per Capita Direct Payments by State 2007 - Stryker Information on Payments comes from company websites, population information comes from the 2005 Area Resource File. Figure 5: Per Capita Direct Payments by State - Smith and Nephew Information on Payments comes from company websites, population information comes from the 2005 Area Resource File. 24
Figure 6: Per Capita Direct Payments by State 2007 - Biomet Information on Payments comes from company websites, population information comes from the 2005 Area Resource File. Figure 7: Direct Payments to Physicians by Hospital, 2007 - Philadelphia The size of the circle is proportional to the size of the total payments to physicians in that hospital. All hospitals in the area where at least one physician that admits patients to that hospital received at least one payment from one of the top five orthopedic joint replacement companies in 2007 is included. Information on Payments comes from company websites, and hospital information comes from 2005 American Hospital Association survey. 25
Figure 8: Direct Payments to Physicians by Hospital, 2007 - San Antonio The size of the circle is proportional to the size of the total payments to physicians in that hospital. All hospitals in the area where at least one physician that admits patients to that hospital received at least one payment from one of the top five orthopedic joint replacement companies in 2007 is included. Information on Payments comes from company websites, and hospital information comes from 2005 American Hospital Association survey. 26
Figure 9: Direct Payments to Physicians by Hospital, 2007 - San Diego The size of the circle is proportional to the size of the total payments to physicians in that hospital. All hospitals in the area where at least one physician that admits patients to that hospital received at least one payment from one of the top five orthopedic joint replacement companies in 2007 is included. Information on Payments comes from company websites, and hospital information comes from 2005 American Hospital Association survey. 27
Figure 10: Number of Orthopedic Joint Replacement Companies making Direct Payments to Physicians by Hospital, 2007 All hospitals in counties with greater than 100,000 people in 2005 are included. Information on Payments comes from company websites, and hospital information comes from 2005 American Hospital Association survey. 28