Market Forces and Unsponsored Research in Academic Health Centers FREE

Despite continued increases in the research budget of the National Institutes of Health (NIH), there is concern that academic health centers (AHCs) in highly competitive health care markets may experience lower research productivity and less success in obtaining research grants.^1- 2 One possible explanation for this phenomenon that has not been explored previously is that increasing local competition leads to reductions in internal sources of funding.^3- 6 Clinical research may be particularly dependent on such sources of support.^7- 9 As clinical revenues are threatened, some fear that AHCs will be less able in the future to provide the same level of internal support for research as they have in the past.¹⁰

In this article, we refer to all research supported by AHCs from internal sources as unsponsored research; that is, lacking sponsorship from external sources. There exist at least 2 types of unsponsored research.^8- 9 One type is institutionally funded research—activities funded by institutions or group practices, using surplus revenues from clinical income, endowment, or income on institutional reserves. A second type refers to activities by investigators at AHCs who, without institutional funding, devote their own resources to conduct unsponsored studies. These resources may be drawn from discretionary funds that they have accumulated from various sources, including their own clinical income or unexpended grant funds, or from working additional uncompensated hours. We refer to research supported in this manner as faculty supported.

It is widely assumed that unsponsored research plays an important role in the investigative process by developing nascent ideas into more fully developed proposals that are capable of external support. In some cases, unsponsored research may lead to published articles or patented products. Little is known about this activity in AHCs. Much of our knowledge comes from studies that have examined the acknowledgments sections of published literature in clinical journals.^7- 9 In some specialties, such as pathology, unsponsored efforts are reported to represent as much as 84% of research leading to publication.¹¹ At 1 institution, it was estimated that more than 50% of all protocols are unsponsored.¹² Studies of research funding based solely on publication rates may be flawed because they miss reports of activities that are not published yet lead to future extramural support or make other important contributions to biomedical science. We found only 1 study in which data were gathered directly from faculty on this topic. Beaty et al¹³ surveyed members of departments of medicine in 1982 and found that 25% had no funding for research and 7% relied solely on intramural support.

We surveyed a sample of more than 2000 medical school faculty to investigate 4 research questions: (1) Do market characteristics influence the magnitude of institutionally funded biomedical research? (2) How common is involvement in unsponsored research? (3) What faculty characteristics predict participation in unsponsored research? (4) What are the outcomes or benefits of faculty-supported research?

Subjects were drawn from a multistage, stratified random sample, similar to previous work.^14- 15 In the first stage, clinical and nonclinical departments were sampled using lists from the Association of American Medical Colleges. At research-intensive institutions (defined as the 50 medical schools with the most funding from the NIH in 1995), we selected all departments of medicine or internal medicine if not previously selected.

In the second stage, a random sample of department members was selected. We excluded faculty who did not have at least the rank of instructor, assistant, associate, or full professor and who did not have a published article in the MEDLINE database during the 5 years prior to the study.

The survey was administered by mail. Of 4058 faculty identified in 117 medical schools (4 of 121 schools refused to participate), we excluded 254 persons after screening who were deceased or too ill to respond, taking sabbatical leave, or determined not to be faculty. Of the 3804 eligible faculty, 2336 responded, yielding a response rate of 62.2%.

Variables Measuring Unsponsored Research Activities. To estimate the magnitude of institutionally funded research, we asked principal investigators (PIs) to report their total funded research budget (direct costs) and the percentage of funding derived from institutional and other sources. The instrument was cognitively tested and respondents seemed reasonably capable of providing accurate financial data on the amounts and sources of the direct costs of their research. For some analyses, we classified PIs by the proportion of their research budget derived from institutional sources (<25%, 25%-49%, 50%-74%, and 75%-100%). To count the number of faculty performing institutionally funded research, we asked all faculty members whether they had received research funding from either their dean's office, their department chair, their faculty group practice, or another administrative unit. All PIs who reported institutional funding and all respondents who answered yes to any of these 4 questions were counted in this category.

To identify subjects who performed faculty-supported research, we asked, "In the last year, have you devoted any of your time to research activities that were not anticipated or planned at the time your research funding was received (yes/no)?" This question was worded specifically to avoid capturing research activities that occurred within the scope of funded projects. To measure the magnitude of faculty-supported research, we asked, "On average, about how many hours a week do you spend on these activities?" In addition, we inquired about their sources of support for such activities, including personal clinical income, currently funded projects, discretionary funds, or extra hours worked. Finally, we asked about the purpose of these activities ("extension of supported work," "necessary to complete supported work," "work on a completely new problem or . . . area," or "work in preparation for an application for funding") and the outcomes ("pilot study that led to winning an extramural grant," "patent application or patent," or "paper in a peer reviewed journal").

Other Variables. Characteristics of the faculty included sex, PI status during the last year (yes/no), professional age (number of years since they received their highest earned degree, with young faculty defined as <10 years), and the proportion of professional time devoted to research. Clinical researchers were defined as those who performed research involving human subjects. Industry funding was determined from the question on funding sources. Medical schools were categorized as public or private and as research intensive or less research intensive.

The competitiveness of local markets was characterized using a classification tool developed by the University HealthSystem Consortium that places health care markets into 4 stages ranging from least competitive (stage 1) to most competitive (stage 4).¹⁶ The tool uses 9 indicators, including the number of large health maintenance organizations (HMOs), HMO penetration, hospital occupancy rates, and others. Although market staging may fail to identify certain competitive activities of hospitals, such as development of horizontal and vertical integrative strategies,¹⁷ it has been used previously as an indicator of HMO activity in a local market.¹ For these analyses, stages 1 and 2 were collapsed.

All responses were weighted to represent national estimates. Analyses were performed using SUDAAN, a statistical software package that incorporates weights from complex survey designs and corrects the SEs.¹⁸

We hypothesized that as markets became more competitive, institutional resources would be squeezed and the institutional subsidy would decrease. To test this, within each market stage we calculated the proportion of direct-cost research budgets represented by institutional funds. Standard errors applicable to the quotients of random variables were calculated.¹⁹

Analysis of faculty characteristics associated with performing institutionally funded research used the individual (weighted) respondent as the unit of analysis. Bivariate differences were tested using the χ² statistic and all variables were entered into a logistic regression. For respondents with any institutional funding, least squares regression was used to model the amount of funding received.

The characteristics associated with faculty-supported research were analyzed in the same manner as described herein, with 2 variations. To test whether institutionally funded research substitutes for faculty-supported research, we included an independent variable indicating receipt of institutional funding. For those who performed faculty-supported research, we estimated a linear regression on the number of hours per week committed to this activity.

Characteristics of the study sample are presented in Table 1. Approximately 80% of the sample were male, 20% were young faculty, 56% were PIs, and about one quarter had received industry funding. A majority of respondents were from public institutions and nearly half were from research-intensive schools. Despite screening efforts, 151 persons were found to have spent none of their time on research activities and therefore were excluded from further analyses.

We compared our data with national figures to assess the representativeness of our sample.²⁰ Our sample (weighted) included slightly more men (80.0% vs 75.3%), more full professors (40.7% vs 26.3%), and fewer faculty with medical degrees (62.4% vs 68.4%) (P<.001 for all). These differences probably reflect our screening process, which selected for faculty engaged in research.

Using weighted results, we estimate that medical school research faculty received $375 million in institutional funding for the direct costs of their research in 1996. This represented 4.4% of an estimated $8.5 billion in total direct costs for research by medical school faculty. Table 2 indicates that nearly half of faculty members with institutional funding received 25% or more of their research funding from institutional sources, and they accounted for more than 60% of institutional amounts.

Figure 1 provides comparisons of mean institutional support as a percentage of the total direct research budgets by market stage. In stage 1 and 2 markets, 6.1% of research funds were derived from institutional sources compared with 4.1% in stage 3 and 2.5% in stage 4 (P = .02, comparing stages 1 and 2 with stage 4). Nearly identical results were found among research-intensive schools, but among non–research-intensive schools, the direction of findings was reversed (but not significant). Faculty from public and private institutions in the least competitive markets also reported lower percentages of institutional funding, but the differences were not significant.

Table 3 describes characteristics of faculty engaged in unsponsored research. Overall, 43% of faculty received institutional funding and 55% engaged in faculty-supported research. Nearly three quarters of respondents (72%) performed at least 1 type of unsponsored research (data not shown). Young faculty were more likely to receive institutional support than senior faculty (53% vs 42%; P<.001), as were faculty located in less research-intensive universities (48% vs 41%; P = .03). Characteristics associated with faculty-supported research activities included clinical research (60% vs 51% for nonclinical research; P = .004), PIs (74% vs 27% for non-PIs; P<.001), industry funding (60% vs 52% for those with none; P = .02), and institutional funding (59% vs 51% for those with no institutional funding; P = .002). Faculty who devoted more than one third of their time to research overall were about twice as likely to engage in faculty-supported research as those who devoted less than one third of their time to research (P<.001).

Results of the logistic regressions predicting faculty involvement in unsponsored research are presented in Table 4. Young faculty were more likely to receive institutional support than senior faculty (adjusted odds ratio [OR], 1.4; 95% confidence interval [CI], 1.1-1.9;P = .004). Clinical researchers (OR, 1.6; 95% CI, 1.1-2.3; P = .02), PIs (OR, 4.3; 95% CI, 2.8-7.0; P<.001), and faculty with high levels of research effort (OR, 6.2; 95% CI, 4.0-9.5; P<.001) were more likely than others to perform faculty-supported research.

Local markets were associated with participation in faculty-supported (but not institutionally funded) research. Faculty residing in stage 4 areas were 1.9 times as likely to engage in faculty-supported research as those residing in stages 1 and 2 (P = .05). Further testing indicated an interaction between type of researcher and market stage. We therefore constructed separate models, one for clinical and one for nonclinical researchers. In this analysis, the effect of market stage was significant only for clinical researchers (P = .004; data not shown). All other results for clinical and nonclinical researchers were similar.

We also examined the magnitude of unsponsored research. Among faculty who received institutional funding, the average amount was $41,910 (SE, $4160). Based on multiple regressions, men received more institutional backing on average than women (P = .04), but no other characteristics were significantly associated. Respondents who performed faculty-supported research committed an average of about 11 hours per week (SE, 0.56) to these activities. Faculty with institutional support (P = .02) or who spent most of their time on research (P<.001) spent more hours than others on these activities.

Sources of support, purposes, and outcomes of faculty-supported research are summarized in Table 5. Compared with scientists in nonclinical fields, clinical researchers were more likely to support their research with clinical income and by working extra hours, while nonclinical researchers relied more on funding from current projects (P<.01 for all). About half of all researchers relied on discretionary funds for a portion of their support. Both nonclinical and clinical researchers spent the same number of hours per week on faculty-supported research (P = .13). The intended purpose of the research did not vary by type of researcher, with more than three quarters reporting that the work was performed as an extension of currently supported projects or as seed work for future funding. About half of respondents said that the research was necessary for supported work and a similar number said that they were working on new topics. Close to four fifths of respondents reported that their activities led to publications, more than half claimed success in using the work as a pilot study, and about 12% to 13% reported that the work led to a patent application.

We surveyed more than 2000 medical school faculty to gather information on unsponsored research activities in AHCs. Nearly three quarters of medical school faculty performed unsponsored research in 1996. Unlike previous studies on this topic, we assessed the effects of health care market competition on these activities and found that the proportion of direct-cost research budgets supported by institutional dollars was smaller in more competitive markets, especially among the most research-intensive institutions. This study also documents significant contributions of faculty-supported research leading to published articles, grants, and patent applications.

It is commonly suspected that many research activities occurring in AHCs are dependent on income from clinical sources. For example, faculty practice plans contributed as much as $2.4 billion to academic purposes in 1992 and 1993, of which $816 million went toward research.³ Concerns that competitive pressures affect clinical income and, therefore, transferable dollars were heightened by a study showing that during the past decade, increases in NIH awards were slowest in markets with the greatest enrollment in managed care.² In the current study, the proportion of research funds supported by institutional sources was greatest in the least competitive markets. This effect was most apparent for schools with the largest NIH budgets. Perhaps one reason why a market effect did not occur among schools with small research enterprises is that a decision to make even slight increases in institutional support would have a proportionally greater impact on total faculty research budgets. Another possible explanation is that administrators of research-intensive schools believe they can scale back on institutional support of research in the face of declining clinical revenue without grave consequences for the school's overall research effort. Smaller enterprises, on the other hand, may think that external support is too unstable and that their cadre of researchers is essential to their identity as academic institutions and, therefore, dig deeper to support the extant efforts.

Our finding that faculty in competitive markets are more likely to engage in faculty-supported research is both puzzling and troubling. Originally, we expected that increased competitive pressures on clinical income would leave less time for faculty-supported work. However, as the market becomes more competitive, faculty may find themselves working extra hours to do what may be covered by "protected time" in less competitive markets. Because academic output is essential to promotion, these faculty may do whatever it takes, within limits, to achieve career goals.

This study was the first to document the degree of institutional support for nonclinical researchers, which may have been underappreciated in the past. Furthermore, we report on the sources of support, purposes, and outcomes of faculty-supported research. While most faculty accomplish these projects by working extra hours, it is worth noting that clinical researchers also rely on clinical income. If managed care squeezes out margins on clinical incomes, this source of support could disappear, although our survey did not measure this potential effect directly.

Our study has several limitations. First, we have probably underestimated the amounts of institutional support. For example, we counted only explicit institutional support, not in-kind funds (eg, funded research staff, contributions to salary, laboratory start-up). Also, we did not measure contributions to infrastructure, including unfunded indirect costs or matching costs, which may be even larger than unsponsored direct costs. However, we may have overestimated unsponsored amounts by including unexpended grant funds, which originally came from external sources. Second, at the time of the study, we were not fully aware of the extent of institutionally funded research activity, and so, to avoid an overly long survey instrument, we collected data on purpose and outcomes only for faculty-supported research. Knowing this information for projects based on institutional funding would have resulted in a richer understanding of these phenomena. Third, our method of distinguishing between clinical and nonclinical researchers was inexact and could have led to overlap. Finally, our results are based on cross-sectional, not longitudinal, data. Effects occurring in the most competitive markets may be due more to adaptive behavior in response to environmental change than to the static pressures of greater competition.

Our research provides empirical evidence of the extent and contribution of unsponsored research in AHCs. It suggests that market pressures may affect the availability of support for these activities and, thus, begins to answer questions raised by others about the mechanism by which health care competition affects academic research. Further explorations of these findings, including debates about the social desirability of relying on clinical income to perform biomedical research, would be useful to extend our work.