Physician Professionalism and Changes in Physician-Industry Relationships From 2004 to 2009

Background  One tenet of medical professionalism is managing conflicts of interest related to physician-industry relationships (PIRs). Since 2004 much has been done at the institutional, state, and national levels to limit PIRs. This study estimates the nature, extent, consequences, and changes in PIRs nationally.

Methods  We performed a national survey of a stratified random sample of 2938 primary care physicians (internal medicine, family practice, and pediatrics) and specialists (cardiology, general surgery, psychiatry, and anesthesiology). A total of 1891 physicians completed the survey, yielding an overall response rate of 64.4%. The main outcome measure was prevalence of several types of PIRs and comparison with PIRs in 2004.

Results  Overall, 83.8% of all respondents reported some type of relationship with industry during the previous year. Approximately two-thirds (63.8%) received drug samples, 70.6% food and beverages, 18.3% reimbursements, and 14.1% payments for professional services. Since 2004 the percentage of each of these benefits has decreased significantly. Higher rates of PIRs are significantly and inversely associated with low levels of Medicare spending.

Conclusion  Among a random sample of physicians, the prevalence of self-reported PIRs in 2009 was 83.8%, which was lower than in 2004.

The medical profession has embraced the importance of placing patient welfare ahead of financial benefits to physicians in clinical decision making. This principle was codified in the Charter on Medical Professionalism in 2002.^1,2 Adopted by more than 100 professional groups worldwide, the Charter on Medical Professionalism is the most comprehensive statement on medical professionalism, to our knowledge. One of the 10 responsibilities of the Charter on Medical Professionalism is maintaining patient trust by managing conflicts of interest. This responsibility requires physicians to manage financial conflicts created by their ownership of health care facilities and/or relationships with pharmaceutical companies, device manufacturers, and the like.

Most practicing physicians in the United States have numerous relationships with industry. In 2004, more than 80% of physicians reported that they received food and beverages in their workplaces and 78% received drug samples.³ More than one-third (35%) were reimbursed by companies for costs associated with professional meetings or continuing medical education (CME), and more than one-quarter (28%) received payments for consulting, speaking, or enrolling patients in clinical trials.

During the last 6 years a number of efforts have aimed to reduce, and in some cases eliminate, certain types of physician-industry relationships (PIRs). These activities include new policies of medical schools and teaching hospitals that forbid physicians from accepting samples and receiving food and beverages in offices, limiting company representatives' access to physicians' offices and clinical areas, and banning faculty participation in speaker bureaus and other forms of promotional activity.^4,5 A number of organizations have driven these changes including, but not limited to, the Institute of Medicine, the Prescription Project, the Association of American Medical Colleges, the Josiah Macy Junior Foundation, and the American Medical Student Association.^5-13 In addition, scores of academics have pushed for reform in this area.^8,9,14-19 Finally, Senator Chuck Grassely (a Republican from Iowa) spearheaded several highly visible investigations of industry relationships involving researchers at more than 2 dozen medical schools and universities (Paul Thacker, oral communication, December 10, 2010). His efforts led to the passage of the Physician Payment Sunshine Act²⁰ to create a national disclosure system of PIRs.

Given the amount of activity in this area, we hypothesized that PIRs are likely to have decreased during the last 5 years. To empirically test this hypothesis and to get updated national estimates of the frequency of PIRs, we repeated the 2004 professionalism survey in 2009.

The institutional review boards at the Massachusetts General Hospital and the University of Massachusetts approved this study. The survey items used in this article were identical to those used in 2004. These items were designed and tested based on information from focus groups, cognitive interviews, and a review of the literature. In addition, before the 2009 administration, the items were subjected to expert review and additional pretesting. However, the survey items were not changed as a result.

Using the 2008 American Medical Association Masterfile, we identified all US physicians in primary care (internal medicine, family practice, and pediatrics) and 4 non–primary care specialties (cardiology, general surgery, psychiatry, and anesthesiology). From this list we excluded all doctors of osteopathy, resident physicians, physicians in federally owned hospitals, those with no address listed, those who requested not to be contacted, and those who were retired. We then randomly selected 500 physicians within each of the 7 specialties, yielding a total sample of 3500. To identify individuals who completed the survey and to avoid unnecessary survey costs, we assigned each of the 3500 sampled physicians a random identification number, which was printed on a sticker and attached to the back of the survey booklet. We used this number to identify and follow up on nonrespondents.

The final survey instrument was a self-administered, 8-page questionnaire. The initial survey packet was sent by priority mail in May 2009 and included a cover letter, fact sheet, questionnaire, postage-paid return envelope, and $20 incentive. Professional interviewers from the Center for Survey Research made telephone calls to all nonrespondents to solicit participation. Of the 3500 sampled physicians, 562 were ineligible because they were deceased, out of the country, practicing a nonsampled specialty, on leave, or not currently providing patient care. Of the remaining 2938 eligible physicians, 1891 completed the survey, yielding an overall response rate of 64.4% and response rates by physician specialty as follows: cardiology, 50.6%; anesthesiology, 64.6%; family practice, 67.5%; surgery, 65.1%; internal medicine, 60.8%; psychiatry, 64.0%; and pediatrics, 72.7%. Having 1891 completed surveys allows us to estimate within 3 percentage points around a mean of 50% with a 95% confidence interval using a 2-tailed P < .05.

The 2004 and 2009 National Surveys on Medical Professionalism¹ contained identical survey items related to PIRs, used the same survey administration methods, and had similar response rates and identical sampling methods. These similar approaches were designed to allow us to conduct 5-year comparative analyses on PIRs. The 2004 and 2009 surveys differed in 2 ways. First, the 2004 survey did not include psychiatrists, who were included in our 2009 sample. As a result, we excluded psychiatrists in any comparisons to our findings from the 2004 study. Second, in 2004 we used separately returned postcards with identification numbers to track respondents, whereas in 2009 we used labels with identification numbers attached to the back page of the survey instrument.

We measured the prevalence of PIRs by asking, “Which of the following have you received in the last year from drug, device, or other medically related companies?” The categories asked about were as follows: “food and/or beverages in your workplace,” “free drug samples,” “honoraria for speaking,” “payment for consulting services,” “payment for service on a scientific advisory board or board of directors,” “payment in excess of costs for enrolling patients in industry-sponsored trials,” “costs of travel, time, meals, lodging, or other personal expenses for attending meetings,” “gifts that you receive as a result of prescribing practices,” “free tickets to cultural and sporting events,” and “free or subsidized admission to meetings or conferences for which CME credits are awarded.” The response categories were yes and no for each. As in our previous work, PIRs were defined as receiving drug samples, gifts (free food or tickets), reimbursements (personal expenses to attend meetings or payment to attend meetings that awarded CME credits), and payments (honoraria for speaking, consulting, being on scientific advisory boards, and enrolling patients in studies) from industry.

In 2004 and 2009 we asked about the number of times each month that physicians met with industry representatives. The survey item read, “In an average month, how many times do you meet with representatives from drug, device, or other medically related companies?” We treated the response as a continuous variable.

Finally, we measured the extent to which physicians reported prescribing brand-name drugs when equivalent generics are available. The survey item read, “In the last year how often have you prescribed a brand-name drug when an equivalent generic was available because the patient asked for the brand- name drug specifically?” The response categories were never, rarely, sometimes, and often.

Independent variables measured the personal, professional, and practice characteristics of physicians. Personal characteristics of physicians included sex (male/female) and underrepresented minority status. Consistent with other research,²¹ nonunderrepresented race/ethnicity categories were white (non-Hispanic) and Asian, whereas underrepresented minorities were all other groups (African American, Hispanic, Native American, or other). Professional characteristics included specialty, number of years in practice, income, and foreign medical graduate status.

Practice characteristics included medical practice organization (hospital clinic, medical school, staff-model health maintenance organization, group practice, solo or 2-person practice, or other) and primary payment mechanism (fee for service, partial capitation, full capitation, salary, or other). To proxy the relative costs within the practice environment, we used a measure of regional Medicare spending created for the Dartmouth Atlas of Healthcare.²² This measure includes Medicare expenditures for beneficiaries during their last 6 months of life to control for differences in case mix across areas. Respondents were assigned to high, medium, or low regional spending groups based on the zip code in which they practice, similar to other published research.²²

We examined univariate and bivariate relationships in the data. To test for significant differences between groups, we used 2-sided t tests (continuous variables) or χ² tests (categorical variables) as appropriate. We constructed a multivariable model based on the bivariate analysis. Because of high correlations among several predictor variables (ie, sex and specialty), we chose to include only 1 of the independent variables when it had a correlation of 0.2 or greater with another independent variable. We chose which variable to include and which to exclude based on what made sense conceptually and what was done in our previous research. Specifically, we dropped sex and income because these variables were highly correlated with specialty for the multivariate models. We fit separate multivariable logistic regression models to evaluate the association of the outcomes with the independent variables—race/ethnicity (nonunderrepresented or underrepresented), specialty (anesthesiology, cardiology, family practice, general surgery, internal medicine, pediatrics, or psychiatry), foreign medical graduate (yes or no), years in practice (<10, 10-19, 20-29, or ≥30), practice organization (hospital/clinic, university/medical school, group practice, solo or 2-person practice, or other), and tertiles of regional spending (low, medium, or high). From these models, we obtained adjusted percentages and standard errors.

Given the size of the sampling frame, we assume that there was little or no overlap among the physicians who responded to the 2004 and 2009 surveys. We treated these samples as independent and used a z test to compare the estimates of the proportion of physicians accepting benefits obtained from the 2 time points.

We were further interested in assessing whether the frequency of meetings with industry representatives had changed over time. Because this variable was not normally distributed, we used a nonparametric test (Wilcoxon rank sum test) to evaluate whether the distribution of this variable differed significantly in the 2 samples (overall and within specialty).

We were interested in whether physicians who had any relationship with industry were also more likely to prescribe brand- name drugs when equivalent generics were available. Because this outcome variable (prescribing brand-name drugs) had 4 levels (never, rarely, sometimes, and often), we used a multinomial logistic model to evaluate this association, adjusting for all the variables specified herein. From this model, we obtained adjusted percentages and standard errors.

All analyses except the nonparametric tests used weights to account for differential sampling rates and nonresponse by specialty. We conducted all analyses in SAS statistical software, version 9.2 (SAS Institute Inc, Cary, North Carolina), and SUDAAN statistical software, version 10.0.1 (RTI International, Research Triangle Park, North Carolina). We considered all 2-sided P values to be statistically significant at the level of P < .05.

Table 1 gives the characteristics of the survey respondents. On the basis of weighted data, 67.2% of respondents were men and 10.4% were underrepresented minorities. In terms of primary practice location, 22.0% were in solo or 2-person practices, 40.4% in group practices (>2 persons), 18.8% in hospitals or clinics, 3.7% in staff-model health maintenance organizations, and 5.5% in university faculty practice plans or medical school. Approximately one-quarter (20.9%) practiced in low-cost regions, 54.9% in medium-cost regions, and 24.2% in high-cost regions.

Table 2 gives the frequency of the various types of PIRs. Approximately two-thirds of physicians (63.8%) received drug samples in the last year. Slightly more (70.8%) received gifts from industry, primarily food and beverages in their offices. A total of 18.3% reported receiving reimbursements for meetings or free or subsidized admission to CME meetings. A total of 14.1% received payments for professional services to pharmaceutical companies. Of these, payments for speaking engagements were most frequent (8.6% of all respondents), followed by serving as a consultant (6.7%), service on a company advisory board (4.6%), and payments in excess of costs for enrolling patients in clinical trials (1.2%). Overall, 83.8% of all respondents reported some type of relationship with industry during the previous year.

Table 2 also gives the changes in the frequency of PIRs in 2004 and 2009. On every measure the percentage of physicians with industry relationships was significantly lower (P < .001 overall) in 2009 than in 2004. For example, the percentage of physicians accepting samples decreased from 78% in 2004 to 64% in 2009 (P < .001). Likewise, the percentages of physicians accepting gifts (83% vs 70.8%, P < .001), reimbursements (35% vs 18.3%, P < .001), and payments (28% vs 14.0%, P < .001) decreased. Overall, the percentage of physicians with any type of industry relationship decreased significantly from 94% to 84% (P < .001).

We also compared the number of meetings between physicians and drug representatives in an average month overall and by physician specialty in 2004 and 2009. Since 2004 the median number of meetings between physicians and industry representatives decreased from 3 in 2004 to 2 in 2009 (P < .001). The median number of meetings decreased significantly among pediatricians (4 to 3, P = .004), internal medicine specialists (5 to 2, P = .002), family practice physicians (8 to 5, P = .008), and cardiologists (6 to 4, P = .05). No differences in the frequency of meetings were shown for anesthesiologists (1 per month in 2004 and 2009) or general surgeons (2 per month in 2004 and 2009); however, the median number of meetings in these specialties was low in both years.

Table 3 gives the results of the multivariate analyses and presents regression-adjusted percentages. After adjusting for personal, professional, and practice characteristics, physician specialty was significantly associated with having PIRs. For example, cardiologists were more likely to have any PIRs than psychiatrists (92.8% vs 79.8%, P = .004). The frequency of PIRs by specialty differed significantly for every type of PIR.

In addition, primary practice organization was significantly related to the frequency of PIRs. For example, physicians in solo or 2-person practices and group practices were significantly more likely than those in hospital and medical school settings to receive samples, reimbursements, and gifts. However, physicians in medical schools were most likely to receive payments from industry.

Tertiles of costs, which measure relative Medicare expenditures in the region in which physicians practice, were also significantly associated with the frequency of several types of PIRs. For example, the frequency of accepting samples among physicians in low-spending areas was significantly less than those in medium- or high-spending areas (56.2% low vs 66.9% medium and 63.4% high, P = .001) after adjusting for the effects of other variables. Similar significant results were shown for gifts (58.0% low-cost, 75.7% medium-cost, and 70.7% high-cost regions, P < .001) and payments (8.8% low-cost, 15.2% medium-cost, and 15.6% high-cost regions, P = .02).

The Figure shows the relationship between PIRs and self-reported prescribing behavior. After adjusting for personal, professional, and practice characteristics, 23.2% of those with at least 1 PIR compared with 35.5% of physicians without PIRs reported that in the last year they had never prescribed a brand-name drug when a equivalent generic was available (not shown in the Figure). However, those with PIRs were significantly more likely than those without PIRs to do so rarely (37.9% vs 34.7%), sometimes (34.5% vs 29.1%), and often (4.4% vs 0.07%) (overall P < .001).

Compared with 2004, PIRs have decreased significantly overall and regarding specific forms of PIRs (samples, gifts, reimbursements, and payments). These significant decreases have likely resulted from a number of secular events, including increased public attention by the press and professional organizations to the propriety of PIRs and new policies by medical schools and hospitals banning certain types of PIRs, such as drug samples and industry-sponsored meals and participation in speaker bureaus. Another contributor may be increased public reporting of PIRs by drug companies, medical schools, states, and the federal government, prompting physicians to engage in PIRs less frequently than in the past. Another possibility is that given the global financial crisis, companies have cut back on their marketing and other expenses. Several complex and potentially interacting factors probably underlie this decrease.

From a policy perspective, the question is what effect, if any, does this reduction currently have or will it have in the future on the costs of medical care? Although our study was not designed to address this issue, our findings raise several possibilities requiring further investigation. For example, our data show that PIRs are significantly less frequent in the lowest-cost geographic areas. However, lower-cost regions are not necessarily associated with higher poverty rates among local residents.²³ Nonetheless, it is possible that companies market less aggressively in regions where many patients are impoverished and covered by Medicaid, which strictly restricts its formulary and use of brand-name products.

Our study also shows an increased propensity to prescribing more expensive brand-name drugs when less expensive generic drugs are available among physicians with PIRs compared with those without. These findings suggest that PIRs are associated with higher health care costs. Future studies should more fully and directly explore these relationships.

Our findings on the reductions in the number of meetings per month between physicians and drug company representatives have implications for drug companies as well. Overall, we found the number of meetings decreased by 33% (3 per month in 2004 to 2 per month in 2009, P < .001). The decrease is most prevalent among family practice and internal medicine physicians, who reduced the number of monthly meetings by 3 per month. Several possibilities might explain these decreases, including increased clinical pressures, especially among primary care physicians, which leave less time for physicians to meet with company representatives; institutional policies that tightly restrict drug representatives from clinical areas; or the increased restriction of gifts and other inducements, which has made physicians less willing to meet with representatives. Regardless of the explanation, our data signal the decrease of the predominant industry marketing strategy of giving inducements to physicians in exchange for their time and attention.

With that said, this study shows that PIRs remain common in medicine today. Given that the vast majority of physicians have PIRs and that research²³ has shown that patients want to know about these relationships, our data support the idea of a comprehensive national disclosure system of PIRs. This is especially true because reliance on institutional-based (hospital or medical school) disclosure systems will not capture most of the PIRs among physicians in noninstitutional practices, such as solo and 2-person practices. For these physicians, creative solutions that build on physician professionalism may be necessary.

Our study has several limitations. First, because we relied on voluntary disclosure and because PIRs tend to be viewed as negative by the press and in some professional settings, our results should be viewed as lower- bound estimates of the actual frequency of relationships with industry. This may also explain some portion of the overall decrease in the frequency of PIRs since 2004. Second, although our response rate was excellent for a physician survey, nonresponse bias could exist. Third, our study was not designed to address the issue of the appropriateness of PIRs—that is, whether it is appropriate for physicians to accept certain types of PIRs, such as gifts or meals. Fourth, to reduce survey costs, we changed the mechanism for tracking respondents from a separate postcard in our 2004 survey to a label with the respondent identification number affixed directly to the back of the survey in 2009. Although we cannot be certain, we suspect this change had minimal effects, given that we achieved significantly higher response rates in 2009 than we did in 2004, we had almost no missing data in both years, and in 2009 only 26 respondents tore off their identification number before returning the survey. Fifth, because our response rate for cardiologists was lower than the rest of the specialties, readers should consider this factor when interpreting the cardiology-related results. Finally, the overall effect of the demonstrated decrease in PIRs is unknown. For example, drug companies may be focusing PIRs on high-volume prescribers and target them directly, thus increasing the overall effect of PIRs on prescribing behavior while decreasing the number of pharmacies with PIRs.

Overall, this study shows that PIRs have been decreasing in the United States—at least for the last 5 years. These data clearly show that physician behavior, at least with respect to managing conflicts of interest, is mutable in a relatively short period. However, given that 83.8% of physicians have PIRs, it is clear that industry still has substantial financial links with the nation's physicians. These findings support the ongoing need for a national system of disclosure of PIRs.

Correspondence: Eric G. Campbell, PhD, Mongan Institute for Health Policy, 50 Staniford St, Ninth Floor, Boston, MA 02114 (ecampbell@partners.org).

Accepted for Publication: June 6, 2010.

Author Contributions: Drs Campbell and DesRoches had full access to all the data in the study and take responsibility for the integrity of the data and the accuracy of the data analysis. Study concept and design: Campbell and DesRoches. Acquisition of data: Campbell, Iezzoni, Bolcic-Jankovic, and Miralles. Analysis and interpretation of data: Campbell, Rao, DesRoches, Iezzoni, Vogeli, and Bolcic-Jankovic. Drafting of the manuscript: Campbell, Rao, and DesRoches. Critical revision of the manuscript for important intellectual content: Campbell, DesRoches, Iezzoni, Vogeli, Bolcic-Jankovic, and Miralles. Statistical analysis: Rao. Obtained funding: Campbell. Administrative, technical, and material support: Campbell, DesRoches, Bolcic-Jankovic, and Miralles. Study supervision: Campbell. Survey design: DesRoches.

Financial Disclosure: None reported.

Funding/Support: This study, the second in a series, was supported by a grant from the Institute on Medicine as a Profession.

Role of the Sponsor: The president of the Institute on Medicine as a Profession, David J. Rothman, PhD, provided feedback on the design and conduct of the study, as part of our expert panel, but had no role in the collection, management, analysis, and interpretation of the data or in the preparation, review, or approval of the manuscript.

This article was corrected for typographical errors on 11/8/2010.