In the last 20 years, many individuals and organizations have expressed concern over the current status and future prospects of research that involves the use of living human beings as research subjects—often referred to in the literature as patient-oriented research (POR).1,2,3 A number of barriers are thought to exist that may have a negative influence on academics' willingness or ability to conduct POR.
The first barrier is the structure of the promotion and reward system in academic medicine. Essentially, promotion and tenure for research faculty depend on the extent to which they publish their results in peer-reviewed journals.4,5 It is believed that faculty have a reduced incentive to conduct POR because such studies take longer to initiate, implement, and publish, and the potential publications that result from these studies are fewer than they would be for non-clinical research.1 A second barrier is the limited supply of research funding for POR. It is generally believed that the amount of funding available for POR from governmental sources, such as the National Institutes of Health (NIH), is significantly smaller than is the amount available for non-POR research.1,2,3
In addition to conducting research, faculty are expected to teach and, if they are physicians, to provide patient care. Over the years, medical schools have become heavily dependent on revenues from the patient care activities of physician faculty—who are thought to make up a substantial portion of the POR workforce.2 Currently, half of all medical school revenues are derived from clinical services (including revenues from faculty practice plans, network affiliations, and other medical organizations, and payments from hospitals to support medical school activities).6 This dependence may translate into increased pressure on physician—faculty to spend more time in patient care and less time in other activities such as research.7
These hurdles faced by POR researchers may be exacerbated by the level of competition in the local health care market—manifested by the spread of managed care. Studies have found a reduced flow of managed care patients into academic health centers (AHCs)8 and a reluctance on the part of some managed care companies to cover patient care costs incurred as part of clinical research.9 Other studies have found slower growth in awards to clinical departments from the NIH, fewer institutional dollars to support research, and lower publication rates among clinical researchers in medical schools in the most competitive markets.10,11,12 Interestingly, these studies did not find similar relationships between market competition and non-clinical research. It is believed that the non-clinical researchers in medical schools may be less vulnerable to the negative effects of market forces than are those conducting POR.
To understand the importance of these potential barriers to conducting POR, we conducted a national survey of medical school faculty. This paper addresses three research questions using data from this survey:
1. How have the amounts of POR and non-clinical research conducted by medical school faculty changed in recent years?
2. What is the relationship between the level of local market competition and changes in the amounts of POR and non-clinical research faculty have conducted in recent years?
3. What is the relationship between the amounts of patient care faculty provide and changes in the amounts of POR and non-clinical research medical school faculty have conducted? In particular, are there differences in the relationship between patient care duties and faculty participation in POR compared with non-clinical research?
The participants in the study were drawn from a multi-stage, stratified, random sample, as described in our previous work.13 In the first stage, using lists from the Association of American Medical Colleges, we identified all academic departments in all U.S. medical schools (excluding two-year schools and schools in Puerto Rico). We also identified the 50 medical schools that received the most funding from the NIH in 1995.
In the second stage, at each institution we randomly selected at least two clinical and at least two non-clinical departments. For the 50 most research-intensive medical schools, we included the departments of medicine or internal medicine, because these departments receive the majority of NIH funding.14
For the third stage, using faculty lists provided by the medical school or available on the World Wide Web, we identified all faculty with the rank of instructor or higher. Four medical schools refused to furnish necessary lists. From each of the 117 participating schools, we randomly selected a maximum of 20 faculty from the clinical departments and 12 faculty from the non-clinical departments. To increase the likelihood that members of our clinical sample were involved in research, we eliminated any faculty member in a clinical department who had not published an article in a journal listed in Medline in the five years preceding the study. These procedures resulted in a sample of 4,058 faculty.
The Center for Survey Research at the University of Massachusetts administered the survey in 1996–1997. We excluded 254 faculty who no longer held a faculty appointment, were on sabbatical, were deceased or ill, or were no longer at the sampled institution, yielding a final eligible sample of 3,804 faculty.
Five key variables and measures were assessed on the questionnaire:
* Type of research conducted. Subjects were asked, “During the last year have you conducted research requiring human subjects approval? (yes/no)” Those who answered “yes” were then asked, “Has this research involved the use of living humans as research subjects?” Those who responded “yes” to the second question were considered to conduct some POR.All subjects were asked whether they had conducted any biomedical research in the preceding year that had not required human subjects review. Those who responded “yes” were considered to conduct non-clinical research. The POR and non-clinical research groups are not mutually exclusive, since 496 respondents reported conducting both POR and non-clinical research.
* Changes in POR and non-clinical research. All respondents conducting POR were asked “Has the amount of research that you have done involving living human subjects in the last three years increased, decreased, or stayed the same?” A similar question referring to non-clinical research was asked as well. The resulting variables, “Change in POR” and “Change in non-clinical research,” served as the primary outcome variables for this study.
* Hours of patient care. All subjects were asked “How much time do you spend in a typical work week in direct patient care services: include all time you spend directly related to patient care, patient record keeping, patient-related office work, travel time connected with seeing patients. Please exclude time on call when not actually working.” Subjects were grouped into four categories based on their responses, “None,” “1–12 hours per week,” “12–24 hours per week,” and “25 or more hours per week.” This variable was used primarily as a control variable.
* Market stage. The markets in which medical schools were located were categorized into stages based on a modeling tool developed by the University Health System Consortium.15 The model evaluates the characteristics of local health care markets and categorizes them into one of four stages, ranging from least competitive (stage 1 markets) to most competitive (stage 4 markets). These characteristics include the number of health maintenance organizations (HMOs) with more than 100,000 enrollees, the percentage of total enrollees in the top three HMOs, hospital occupancy rates, average hospital days per 1,000 population, the percentage of specialists who are capitated, the percentage of the Medicare and Medicaid beneficiaries in HMOs, and commercial HMO premiums. Although market staging may fail to identify certain competitive strategies of hospitals, such as horizontal and vertical integration strategies,16 it has been used previously as an indicator of HMO activity in a local market.11 Because stage 1 and stage 2 markets closely resemble one another on a number of dimensions, we grouped these responses together to increase sample sizes for statistical testing. Also, this tool was developed at the time the survey was administered and validated by our previous work in this area.11,12
* Research productivity. We employed two measures of faculty research productivity: the numbers of publications over their careers and the numbers of publications in the preceding three years. On the questionnaire, participants were asked, “Approximately how many articles have you published in refereed journals in your career?” We then asked a similar question, “How many of these were published in the last three years?”
In bivariate analyses of the data, we used the χ2 statistic to test for differences in proportions with the outcome variables (change in POR and change in non-clinical research) having three response categories (decreased, stayed the same, and increased). We used logistic regression to estimate the odds of two combinations of responses to the outcome variables.17 First, we compared those who reported decreases in the amounts of research they had conducted in the previous three years with those who did not report such a decrease (stayed the same and increased). Second, we compared those who reported increases in the amounts of research they had conducted in the previous three years with those who did not report such an increase (stayed the same and decreased). Differences in means were tested using ordinary least-squares regression.
Because of the nested nature of the sample and because we wanted to generalize findings to the population from which the sample was drawn, we conducted weighted analyses to reflect the sampling procedure and to adjust for non-response bias. We computed the weights by calculating the inverse probability of selection within sample strata. Unless otherwise noted, the data presented in this paper reflect the results of the weighted analyses. All analyses were performed using a statistical software package designed to analyze data from complex survey designs and to correct the standard errors that are used in computation of statistical significance.18
Of the sample, 2,336 completed a questionnaire, yielding an overall response rate of 62.2%. The unweighted and weighted characteristics of all respondents are shown in Table 1. Of the respondents to the survey, 20% were women, 62.4% were physicians, and 30% were full professors. In terms of patient care responsibilities, 36.4% had none, 15.6% spent 1–12 hours, 17.4% spent 13–24 hours, and 30.6% spent 25 or more hours per week in patient care. Forty-one percent of the respondents were located in stage 1 and 2 markets, 46.5% in stage 3 markets, and 12.9% in stage 4 markets.
We found that 48.5% of respondents reported conducting some POR. Of these, 15.8% reported decreases in the amounts of POR they had conducted in the three years preceding the study, 48.7% reported no change, and 35.4% reported increases (see Table 2). Fiftynine percent of the respondents had conducted some non-clinical research. Of these, 19.8% reported decreases in the amounts of non-clinical research they had conducted in the three years preceding the study, 55.7% reported no change, and 24.4% reported increases.
Among those conducting POR, 29% of full professors reported increases in the amounts of such research they had done in the previous three years, compared with 47.4% of associate professors and 37.5% of assistant professors (p = .03; see Table 3). In terms of the numbers of hours of patient care provided per week, 9.6% of those with no patient care responsibilities reported decreases in the amounts of clinical research they had conducted, compared with 16.5% of those with 1–12 hours, 11.2% of those with 13–24 hours, and 19.0% of those with 25 or more hours of patient care per week (p = .09). Among clinical researchers in stage 1 and 2 markets (the least competitive), 11.8% reported decreases in the amounts of POR conducted in the previous three years, compared with 29% of those in stage 4 schools (p = .007).
Slightly different results were found for non-clinical research (see Table 3). Unlike the situation with POR, significantly more men than women reported decreases in the amounts of non-clinical research they had conducted in the previous three years (23.2% of men versus 13.5% of women, p = .02). We found no significant difference by academic rank in the proportions of faculty reporting decreases or increases in the amounts of non-clinical research they had conducted in the previous three years. Regarding patient care, 35.5% of those who had spent at least 25 hours per week in patient care reported decreases in the amounts of non-clinical research they had conducted, compared with 14.2% of those with no patient care responsibilities (p < .001). Among respondents in stage 1 and 2 markets, 19.3% said they had decreased the amounts of non-clinical research, compared with 32.4% of those in stage 4 markets (p = .04).
Change in amounts of research. Respondents in stage 4 markets were at least two times more likely to report having decreased their POR and non-clinical research efforts (OR = 2.80 and 2.11, respectively) than were faculty in stage 1 and 2 markets (see Table 4). Among those conducting POR, researchers who had 25 or more hours of patient care per week were significantly more likely to report decreases than were those who had no patient care duties (OR = 2.35). However, among those who had conducted some non-clinical research, those with 3–24 and 25 plus hours of patient care per week were significantly more likely than were those with no patient care duties to have reported a decrease (OR = 2.31 and 3.05, respectively).
The only variable that was significantly associated with an increase in POR was academic rank. Compared with full professors, associate professors were 2.04 times more likely to have reported an increase. However, gender, academic rank, hours of patient care, and market stage were not significantly associated with a greater likelihood of having reported an increase in non-clinical research.
Research productivity. Figure 1 shows the mean numbers of publications the respondents reported having in the three years preceding the study and over their careers controlling for the independent effects of gender, rank, the number of hours of patient care, and market stage. Those who reported decreases in the amounts of POR had been no less productive over their careers or in the previous three years than had those who did not report a decrease. Also, for non-clinical research we found no significant difference in the numbers of career publications or publications in the previous three years when those who reported decreases were compared with those who did not.
Despite ongoing concerns regarding the health of the POR enterprise in U.S. medical schools, we found that 84% of the respondents conducting POR and 80% of the respondents engaged in non-clinical research reported conducting the same amount of research in 1996–1997 as they had conducted in the preceding three years, or more. While we do not know the actual magnitudes of the reported changes, this finding qualitatively suggests that both the clinical and the non-clinical research missions of medical schools and teaching hospitals were likely to be stable, and perhaps even growing, in the mid to late 1990s.
Consistent with previous studies, we found that high levels of market competition were associated with trends in the amounts of clinical research conducted by faculty.10,11 Essentially, we found that faculty conducting POR were significantly more likely to have experienced reductions in the amounts of research if they were located in medical schools in the most competitive health care markets (stage 4). However, unlike the findings of previous studies, our data showed declining research efforts among those conducting non-clinical research in the most competitive markets. While we do not know the causal mechanism at work, ours is the first study to demonstrate a relationship between market competition and the non-clinical research mission of medical schools and teaching hospitals.
Not surprisingly, the average number of hours of patient care duties per week was significantly related to the likelihood that faculty had decreased the amounts of clinical and non-clinical research they had conducted in the preceding three years. We found that those conducting POR who had 25 or more hours of patient care per week were significantly more likely to report they had decreased the amounts of POR they had conducted than were those with no patient care duties. However, having more than 12 hours of patient care per week was significantly related to a decrease in the amount of non-clinical research conducted. Unfortunately, because of the cross-sectional natural of the study, we are unable to establish a causal relationship between patient care duties and changes in the amounts of POR and non-clinical research in the previous three years.
Our examination of research productivity demonstrated that faculty who had decreased their POR and/or non-clinical research efforts in the previous three years had been as productive (as measured by publications) in the previous three years and also over their careers as had those who reported conducting the same amount of research or more. This finding suggests that decreases in the amounts of research conducted are not confined to unproductive faculty.
This study has limitations that may reduce its generalizability or interpretation. First, the study measured only the percentages of faculty who reported that the amounts of research they had conducted in the preceding three years had decreased, remained the same, or increased. We did not measure the extents to which the amounts of research increased or decreased. Second, because we surveyed faculty who were engaged in research, we may have missed those faculty who had abandoned research altogether, thus resulting in an underestimate of the changes in the previous three years. Third, we did not measure changes in the amounts of patient care faculty had provided in recent years. Fourth, our results may have been influenced by response bias, in that faculty are likely to overreport the extents to which they are engaged in socially acceptable behaviors such as publishing and underreport less socially desirable behaviors such as decreasing their research efforts.
Despite these limitations, this study demonstrates the continued vitality of the research missions of medical schools and teaching hospitals. The study also provides additional evidence of the negative relationships that exist between high levels of market competition and patient care on the POR and non-clinical research missions of medical schools and teaching hospitals.
The fact that high levels of patient care and market competition appear to have negative relationships relative to the amounts of clinical and non-clinical research being conducted has implications for individual researchers, medical schools and teaching hospitals, and public policy. Individual researchers in the most competitive markets who have primary interests in conducting clinical and non-clinical research may consider relocating to less competitive markets. This may be especially true for physicians, who, unlike their non-physician colleagues, face the additional responsibility of providing patient care.
At the same time, medical schools and teaching hospitals in highly competitive markets may need to devote additional resources if they want to grow their research missions at rates similar to those of medical schools and teaching hospitals in less competitive markets.
Implications for public policy are difficult to draw given our study design. However, further study into the nature, extent, and consequences of market competition relative to the research conducted in medical schools and teaching hospitals is needed. Such studies may lead to new and revised policies designed to maximize our nation's research investment in the life sciences.