Over the last 10 years, conflicts of interest in clinical care have received unprecedented attention both in the medical literature and in the media. Earlier, conflicts of interest in medical research were the predominant concern1–9; now, equal attention is devoted to the impact of physician–industry relationships on patient care.10–14 This new interest reflects several developments. First, beginning in the 1990s, a substantial body of research emerged to show the power of industry gifts and payments in shaping physicians' beliefs and behaviors.10,12,15–18 Second, the media have relentlessly exposed the potential for physician–industry ties to unduly influence patient care.11,12,14,19 Third, a series of high-profile legal cases cast a spotlight on drug companies' illegal marketing of drugs to physicians, including instances that involved physicians accepting inducements for prescribing company products.11,12,20
As a result, several prominent organizations, concerned about scientific integrity and medical professionalism, began to devise and evaluate potential guidelines for managing clinical conflicts of interest (CCOIs). In January 2006, the Institute on Medicine as a Profession (IMAP) and the American Board of Internal Medicine (ABIM) Foundation, publishing in the Journal of the American Medical Association, urged academic medical centers (AMCs) to lead the profession in eliminating undue industry influence in clinical care.19 In June 2008, the Association of American Medical Colleges (AAMC) released a task force report on managing conflicts of interest in medical education,21 followed by the Institute of Medicine's (IOM) proposal in January 2009.1
The ABIM–IMAP, AAMC, and IOM publications are notable for the extraordinary concordance among their recommendations. All three organizations set forth specific policy proposals for a range of physician–industry interactions, including restrictions on gifts, meals, ghostwriting, and speakers' bureaus; the establishment of central repositories for product samples and scholarships; and full transparency in consulting and research contracts. The differences among their recommendations are minor. For example, ABIM–IMAP recommended that AMCs prohibit speakers' bureaus, whereas the IOM and the AAMC would “strongly discourage” them.
With the ABIM–IMAP, AAMC, and IOM reports, a clear consensus now exists on how AMCs should manage CCOIs. So too, all three organizations' recommendations have been widely discussed in the medical literature and in the press, including an endorsement of the ABIM–IMAP publication by the editorial board of The New York Times.22
Yet, despite this attention, little is known about the policies that medical schools have instituted to address CCOIs. Anecdotal reports suggest that many medical schools have begun to consider new initiatives, but no reliable national data exist on the extent to which U.S. medical schools have implemented policies to eliminate or manage CCOIs. The American Medical Student Association (AMSA) used the ABIM–IMAP recommendations to develop “scorecards,” assigning letter grades to schools' policies.23 This effort has proved effective at mobilizing media attention and, in several cases, promoting change. Methodologically, however, the scorecard project has flaws that limit its use as a database for research. For example, schools that do not respond to AMSA's request for information are given failing grades rather than categorized as nonrespondents. Moreover, it is unclear how AMSA determines whether schools' policies are “unlikely to have a substantial effect on behavior,”23 because the project does not examine their implementation efforts.
To obtain more robust data, we set out to document what policies were in place at medical schools in the United States and how they compared to the ABIM–IMAP, AAMC, and IOM recommendations. Anticipating that more calls for change would follow, we sought to establish a baseline for future research. We also developed and analyzed a database of CCOI policies in an effort to encourage schools to revise and strengthen their guidelines.
We asked deans and compliance officers at all 125 MD-granting U.S. medical schools that existed at the time to participate. We first informed deans of the purpose of the project, noting that a questionnaire would be sent to their institution's compliance office for completion. Compliance officers then received a letter describing the study, a paper copy of the questionnaire, and a password-protected link for online completion of the questionnaire and uploading of relevant policies.
The questionnaire asked about institutions' policies in 11 key areas of CCOI:
- Gifts from industry
- Meals provided by industry
- Vendor provision of product samples
- Vendor access
- COI policies for pharmacy and therapeutics (P&T) committees (formal decision-making bodies, usually composed of physicians and pharmacists, which oversee the purchasing of pharmaceutical products)
- Industry funding for continuing medical education (CME)
- Consulting agreements
- Industry donations for scholarships, fellowships, travel
- Faculty authorship on articles ghostwritten by industry employees
- Faculty participation in industry-organized speakers' bureaus (arrangements whereby companies train and pay physicians to speak on specific topics, typically using slides or other materials prepared by the company)
It provided fixed-choice responses for each area (“yes,” “no,” or “in progress/under consideration”). We asked participants to provide copies of all relevant policy documents. They were not asked to rate the strength or effectiveness of their institutions' policies. The study received approval from the institutional review board of Columbia University. No financial incentives were offered for participation.
Our initial contacts with deans, via mail, and compliance officers, via e-mail, began in October 2007. Nonresponding compliance officers received follow-up e-mails, followed by Federal Express mailings. A final round of follow-ups occurred by telephone. The study closed on December 31, 2008.
Measures and variables
The project covered the 11 key areas of CCOI identified as most salient by the IOM, AAMC, and ABIM–IMAP (Table 1). The questionnaire asked respondents to indicate whether their institutions had formal policies in these areas and prompted respondents to attach copies of all relevant policies.
Policy coding and analysis
We designed a coding system to gauge the strength of schools' CCOI policies, based on the recommendations put forth by ABIM–IMAP, AAMC, and IOM. These advised that institutions (1) prohibit industry gifts, meals, and ghostwriting, (2) prohibit or “strongly discourage” speakers' bureaus, (3) establish central repositories for product samples and industry funds for CME, scholarships, fellowships, and travel, (4) require that members of purchasing committees be free of conflicts of interest, and (5) require full transparency for industry honoraria and consulting contracts. Examples and specific elements of model policies were identified through intensive fieldwork by two authors (S.C. and D.J.R.) at more than a dozen medical schools.20,24,25
For nine policy areas, we developed four codes: 0 (no policy), 1 (permissive), 2 (moderate), and 3 (stringent). The other two policy areas, CME and ghostwriting, were coded on a scale of 0 (no policy), 1 (permissive or moderate), or 3 (stringent) because no appreciable difference between a 2 (moderate) rating and a 1 (permissive) rating existed. The following are examples of code ratings and stringent policies. (See www.imapny.org for the complete codebook; see also this report's online Supplemental Digital Table 1, available at http://links.lww.com/ACADMED/A39.)
- 3—Stringent: No gifts or payments permitted, regardless of dollar amount
- 2—Moderate: Imposes some limits on gifts
- 1—Permissive: Weak or no restrictions on gifts; disclosure may be suggested or required
- 0—No policy
Here is an example of a “stringent” gift policy (The University of Maryland School of Medicine):
School faculty and students may not accept any gift from industry, whether given at the School or at another location.
P&T committee membership:
- 3—Stringent: No conflicted individuals allowed on purchasing committees OR required disclosure and recusal
- 2—Modest restrictions: Conditional disclosure and/or recusal
- 1—Permissive/weak: No restrictions on P&T committee participation
- 0—No policy
Here is an example of a “stringent” P&T policy (Yale Medical Group):
[Yale Medical Group] physicians who are involved in institutional decisions concerning the purchase of or approval of medications or equipment, or the negotiation of other contractual relationships with industry, must not have any financial interest (e.g., equity ownership, compensated positions on advisory boards, a paid consultancy or other forms of compensated relationship) in pharmaceutical companies that might benefit from the institutional decision.
All policy summaries were coded independently by one researcher (L.P.) and a research assistant. Differences in coding scores were resolved through discussion with the principal investigator (S.C.).
A number of schools indicated on their questionnaires that their policies were “in progress.” These included schools with existing policies under reconsideration or revision, as well as those in the process of creating new CCOI policies. For the former, we coded only their current, approved policies. The latter (schools with unapproved, draft policies) received “no policy” codes.
We also assessed policy strength at the institutional level, looking at scores across all 11 CCOI areas. Once all policies were coded, we averaged each school's policy scores to create a measure of overall strength. The 11 policy areas were weighted equally because all have been identified as vital by the AAMC, IOM, and ABIM–IMAP.
We also coded the following institutional characteristics:
- Institution type: Coded as either “1” (public) or “2” (private)
- Hospital ownership: Coded as “3” (owned), “2” (affiliated), or “1” (mixed)
- NIH funding status: Dollar amount of NIH funding in 2007
The NIH figures only include grant and contract funding. However, to our knowledge, no more complete index of medical schools' public research funding exists.
We hypothesized that more research-intensive institutions, as assessed by NIH funding, would be more likely to have stringent policies because of their prominence and the attention given to conflicts of interest by state and federal regulators. We also hypothesized that public institutions would be more likely than private institutions to have strong CCOI policies because of the existence of state laws that address conflicts of interest in public decision making.26 Similarly, we hypothesized that schools that owned their hospitals would have more stringent policies than hospital-affiliated schools, given their vested interest in managing CCOIs that could result in more costly prescribing.26
We used chi-square and one-way ANOVA tests to assess whether these variables (public versus private, hospital ownership versus hospital affiliation, and level of NIH funding) were associated with differences in CCOI policy strength. A P value of .05 was used to test significance. We used SPSS statistical software, version 19 (SPSS Inc., Chicago, Illinois) to conduct our analysis.
We received responses from 77 (62%) of the 125 accredited MD-granting medical schools in the United States that existed at the time of our study. We compared our sample with the overall population to determine whether there was a response bias by institutional characteristics. We found few differences. Forty-nine respondents (64%) from our sample were public institutions compared with 80 (64%) of all AMCs. Forty-eight respondents (62%) owned at least one of their associated hospitals, which is comparable to the 73 (58%) medical schools nationally. Schools in our sample received an average of $100,435,867 in NIH funding (median $52,576,636) in 2007 compared with $84,641,815 (median $51,078,632) for the population.
In assessing medical schools' policies, we first analyzed the extent to which those in our sample reported that they had, or were considering, policies in the 11 areas of CCOI (Table 2). A majority of schools indicated on their questionnaires that they had policies, approved and not under revision, concerning gifts, meals, vendor access, CME, consulting, and honoraria. The rates of “no” responses (policy neither enacted nor in the works) varied from 8% for consulting and honoraria to 49% for ghostwriting. The prevalence of “in progress” responses (existing policy under revision, or new policy under consideration) ranged from 17% to 29% for the 11 CCOI policy areas.
We then analyzed and coded the CCOI policies that participating schools submitted to us, assessing their strength in comparison with the recommendations of ABIM–IMAP, IOM, and AAMC. For this analysis, the “in progress” responses were disaggregated: For schools with existing policies under revision, we coded only their current, approved policies. Institutions with unapproved or draft policies received “no policy” codes. Schools that did not provide copies of their policies were treated as “missing.”
Agreement between coders in assessing policy stringency was excellent: Intraclass correlation coefficients ranged from 0.892 for honoraria to 0.998 for meals, gifts, samples, and travel/scholarships, for an average kappa of 0.979 (815 of 847 [96%]). Collecting, analyzing, and coding the policies, using a uniform set of standards and multiple, trained coders, yielded accurate, high-quality data.
The frequency of stringent policies among medical schools varied considerably (Table 3). The policy areas with the highest rates of stringency were gifts (30%), meals (26%), P&T (22%), and ghostwriting (22%). The lowest rates of stringency were observed for consulting (1%), honoraria (3%), and speakers' bureaus (4%). “No policy” was the most common status for ghostwriting (70%), samples (48%), P&T committees (48%), speakers' bureaus (45%), vendor access (40%), fellowships/scholarships/travel (36%), and meals (29%).
No statistically significant differences were observed between public and private institutions. Nor was hospital ownership correlated with policy strength in any CCOI area. Schools with greater NIH funding were more likely than those with less NIH funding to have formal policies (and, often, more stringent policies) in all areas except P&T committees and speakers' bureaus (Table 4).
Each school's individual policy codes were averaged to create a measure of overall strength, or “policy strength average” (PSA). For example, the institution with the highest overall policy strength had codes of 2 for samples, consulting, and honoraria and codes of 3 for all other policy areas. Averaging these codes produced a PSA score of 2.7.
Figure 1 shows the distribution of PSA scores in our sample. The mean PSA score was 1.2 (median 1.4).
Grouping schools by their PSA scores into four quartiles indicated that about one-quarter of the sample had almost no CCOI policies. The middle two quartiles had PSA scores averaging, respectively, 0.89 (just below “permissive”) and 1.55 (about halfway between “permissive” and “moderate”). The top quartile had a mean PSA score of 2.22, just over “moderate” (Figure 2). Supplemental Digital List 1 (http://links.lww.com/ACADMED/A40) identifies the schools in the top quartile of our 77-school sample and their PSA scores.
Schools' PSA scores were not associated with public/private status or hospital ownership/affiliation. However, greater NIH funding was correlated with higher PSA scores, F(22, 54) = 2.302, P = .007. The mean NIH funding of the top PSA quartile was $314 million in 2007 compared with $7 million for the bottom PSA quartile.
Discussion and Conclusions
Our analysis provides the first comprehensive overview, to our knowledge, of CCOI policies at medical schools. Our findings indicate that, as of December 2008, adoption of CCOI policies by U.S. medical schools was notably incomplete. The absence of policy was the most prevalent finding in 7 of 11 CCOI areas. Even the most frequently regulated areas—gifts and consulting—had “no policy” rates of 25% and 23%, respectively. Ghostwriting was the most neglected policy area: 70% of medical schools in our sample had no explicit policy governing this practice. When they did have policies, however, they were usually stringent: Of the 19 policies we received, 17 (89%) prohibited it, and the remaining 2 discouraged it. Recent evidence of ghostwriting's prevalence in medical research demonstrates the need for more medical schools to prohibit it.27–29
Our findings also indicate that many schools were reluctant to limit physician–industry exchanges that occur primarily off-site and outside of faculty's official duties. Very low adoption of stringent policies (less than 5%) was observed for consulting, honoraria, and speakers' bureaus, which usually take place off-campus. Greater stringency existed for activities that commonly take place at, or directly involve, the institution: gifts (30%), meals (26%), P&T committees (22%), vendor site access (19%), fellowships/scholarships/travel (18%), samples (17%), and CME (16%). This disparity may encourage companies to shift their marketing efforts off-site.
The most crucial question is why so few medical schools had strong, comprehensive policies. Most schools lacked stringent policies in many, if not most, areas, and more than half of the respondents in our sample had PSA scores of 1.0 (“permissive”) or less. Even schools in the top quartile for overall strength had policies that, on average, were closer to “moderate” than “stringent.” The prevalence of “in progress” responses across all policy areas—ranging from 17% to 29%—may indicate changes that are to come. Follow-up research should determine whether stronger policies emerge in the future.
Greater NIH funding was associated with policy stringency. This finding, too, merits further investigation. It may be that academic prominence and visibility prompts institutions to strengthen their policies. NIH funding level is also a proxy for institutional size, and larger organizations may be more likely to adopt formal CCOI policies than their smaller counterparts. A “peer effect” is also possible, in which schools of similar standing tend to track and emulate one another's policies. Anecdotal reports indicate that many medical schools are alert to CCOI policy changes among their reputational cohorts. So, too, NIH oversight in research may promote a more general culture of policy stringency. Whatever the reason, the NIH is in a unique position to exert a positive influence on CCOI policies.
There is reason to be optimistic that change is coming. In an effort to increase transparency and promote more comprehensive management of financial conflicts of interest, the NIH has proposed new conflict-of-interest rules for grantees and their institutions.30 Whereas the proposed changes focus on research-related conflicts of interest, they are sufficiently broad that they would also require disclosure and management of many CCOIs. Specifically, researchers would be required to report to their institutions all financial interests relevant to their research—not just those related to the NIH-funded research—including any consulting fees, travel reimbursements, and honoraria from companies. The institutions would then report to the NIH specific details about all identified conflicts, including their nature and value and a description of how the institutions will manage them. Institutions would also be required to post information about all conflicts in excess of $5,000 on a publicly accessible Web site. Moreover, investigators would undergo financial conflict-of-interest training before engaging in NIH-funded research and every two years thereafter.
The NIH is currently considering comments from the public, and the final rule is expected to be issued in early 2011. Should the proposed changes be enacted, they would dramatically alter how many medical schools manage physician– industry ties. To be sure, the proposals fall short of the recommendations put forth by the ABIM–IMAP, AAMC, and IOM. For example, the NIH would provide “examples of conditions or restrictions”30 that might be part of a researcher's conflict-of-interest management plan, but institutions would retain discretion in determining if and how industry relationships should be reduced or eliminated. Still, the new rules would implement higher minimum standards for the disclosure and management of many physician–industry ties—a critical step forward when so many institutions lack strong policies.
This study has several limitations. First, some institutions may address CCOIs through informal means rather than relying on policies. Second, formal policies may be ignored or not enforced; our study did not evaluate implementation or enforcement practices. Third, our data collection ended on December 31, 2008, and policies enacted after this date are not reflected in our findings. We have continued to track medical schools' CCOI policies, and approximately 21% of all schools have revised their policies since our data collection ended. This is promising but hardly transformational. Our major finding—that few have sufficiently stringent policies—still holds. Our study should thus serve as a useful baseline for evaluating future policies.
Our data point to the ongoing need for medical schools to consistently and comprehensively address the challenges presented by CCOIs. Most schools lacked strong policies to manage conflicts of interest in clinical care. Wider adoption of CCOI policies is crucial in order to eliminate undue industry influence in clinical care and to preserve public trust in the medical profession.
To help promote change, we have made our database of CCOI policies publicly available, in a searchable format, at http://www.imapny.org/conflicts_of_interest/search_policies. It currently houses policies from 104 institutions, and it is regularly updated as we receive new information. Our hope is that deans, compliance officers, faculty, and students will use the database to examine, compare, and implement strong policies to manage this ongoing problem.
The authors are pleased to acknowledge the research assistance of Poonam Pai and Frederica Stahl. Ms. Pai and Ms. Stahl were compensated for their contributions by the Center on Medicine as a Profession, College of Physicians and Surgeons, Columbia University, New York, New York. The authors also benefited from the methodological guidance of Patrick Moynihan, PhD, Harvard University, Boston, Massachusetts, and Ivo Antoniazzi, PhD, Teachers College, Columbia University, New York, New York.
This work was made possible by a grant from the state Attorney General Consumer and Prescriber Education Grant Program which is funded by the multistate settlement of consumer fraud claims regarding the marketing of the prescription drug Neurontin. This research was also funded by the Institute on Medicine as a Profession, which supports the Center on Medicine as a Profession at Columbia University. This work was also funded by a grant from the Institute on Medicine as a Profession (IMAP) to the Center on Medicine as a Profession (CMAP).
Dr. Rothman is president of IMAP and director of CMAP, and he had a direct role in all aspects of the study. He is a board member of IMAP and has received travel reimbursements from IMAP. CMAP, which employs all authors, received a grant from IMAP to support this work.