As providers of patient care, physicians bring a unique perspective to the scientific understanding of human disease.1 Researchers first observed a decline in the number of U.S. physicians pursuing research over 30 years ago.2 Since then, others have attributed this decline to a number of factors, including the challenging process of developing a successful research program, the lack of protected time for physicians to conduct research, insufficient mentoring, and inadequate institutional support.3
In response, the U.S. government and private foundations alike established programs to support early-career physicians conducting research, which appear to have a positive influence on physicians’ decisions to become researchers.4 The Doris Duke Charitable Foundation established the Clinical Scientist Development Award (CSDA) in 1998 to support mentored research conducted by early-career physician scientists. The CSDA program requires that awardees dedicate a minimum of 75% of their professional time to research activities. Establishing this protected time is especially important early in an investigator’s career, when developing research potential is critical and clinical responsibilities are often demanding. The CSDA program supports early-career scientists who could go on to receive larger research project grants. A metric of this transition from mentored to independent research funding is the attainment of a National Institutes of Health (NIH) R01 research grant, which is the NIH’s most commonly used grant.5
The CSDA program has awarded peer-reviewed grants annually since 1998, except for a hiatus in 2003–2004. Applicants are nominated by the U.S. academic institution with which they are affiliated, except in 2005–2006, when the competition was open to all qualifying applicants. The CSDA program requires applicants both to hold an MD or equivalent foreign degree and to have attained a faculty appointment not higher than the assistant professor level within five years of applying for the grant. From 2000 to 2002, fellows close to achieving a faculty position were also eligible to apply for fellow-to-faculty awards. Applicants must propose a clinical research project, which is a scientific investigation of the etiology, prevention, diagnosis, or treatment of human disease using human subjects, human populations, or materials of human origin. Currently, the CSDA program supports projects in all disease areas; prior to 2007, however, it supported only four disease areas: cardiovascular disease, cancer, HIV/AIDS, and blood disorders. A panel of experts, most often physician scientists, selects the awardees after a rigorous review process.
Although researchers have reported the characteristics6 and career choices7 of medical students in physician scientist training programs, few have reported the characteristics and impact of private foundation awards to support clinician investigators at the faculty level. Existing analyses of approaches to supporting mostly PhD early-career researchers as faculty or in training, however, suggest that private foundation initiatives contribute to the research productivity and career development of awardees.8,9
A 2006 study by Fanzo and Gallin10 reported early outcomes of CSDA awardees. In the current report, we used more extensive data and asked three questions: (1) What are the characteristics of successful applicants in the CSDA competitions? (2) Is attainment of a CSDA associated with recipients’ transition to independent research funding? and (3) Do CSDA alumni remain in academic medicine with research as their main professional activity?
The Doris Duke Charitable Foundation awarded the first CSDA grants in 1998, and we conducted our analyses in 2011. Therefore, all applicants and awardees from 1998 to 2011 (n = 1,441) were eligible for inclusion in our study (see Figure 1).
Calculation of CSDA success rates, 1998–2011
We defined the number of applications to the CSDA competition from 1998 to 2011 as the number of individuals who submitted proposals reviewed by the expert panels. We calculated the success rate as the number of applications funded divided by the total number reviewed.
Analysis of applicant data, 2006–2011
Unfortunately, the foundation did not systematically collect data on gender, race, ethnicity, and age prior to 2006, so we tested for associations between demographics and CSDA success using the 104 applicants who were awarded grants and the 767 applicants who were not from 2006 to 2011 (see the left side of Figure 1). To maintain privacy, a contractor collected, then deidentified, the applicants’ self-reported data regarding gender, age, race/ethnicity, and medical and other academic degrees. We categorized the deidentified responses regarding racial/ethnic background into the following groups based on the report by Haak and Schnell11—no response, two races, American Indian/Alaska Native, black, Hispanic/Latino, white, Asian, and other. We categorized applicants who indicated East Indian, South East Asian, or Pacific Islander as Asian. For our analysis, we categorized American Indian/Alaska Native, black, and Hispanic/Latino as underrepresented minorities in the biomedical workforce. In our analysis, we did not use the data for those in the mixed race and other categories.
To determine the funding rank of each applicant’s institution, we obtained a list of all the U.S. higher education and research institutes funded by the NIH and their total annual research grant funding for the years 2006 to 2011 from the publicly available NIH Research Portfolio Online Reporting Tools (RePORT) database.12 For each year, we categorized institutions as having funding greater than or equal to the 95th percentile of funding (meaning the 5% of schools with the highest funding levels) or lower than the 95th percentile (the other 95%).
Finally, to determine the association between CSDA applicants’ success and their age, we divided applicants into two groups—younger than 40 years old and 40 years or older.
Analysis of data for 1998–2007 alumni and comparison group
We examined the research trajectory and subsequent grant attainment of CSDA alumni who first received their grant at least four years earlier (awards granted from 1998 to 2007) and completed at least three years of CSDA-supported research as a faculty member (see the right side of Figure 1). This cohort included 120 CSDA alumni.
We also analyzed the corresponding data from a comparison group. The peer review process for each annual competition produced a list of applicants by final score. Grants were awarded to those applicants whose scores exceeded a specific threshold or those who were selected according to program priorities and their ability to meet the requirements of the award. The comparison group included those individuals whose proposals were not funded but had scores within or immediately below the funding threshold. We matched the number of applicants in the comparison group to the number of awardees in each year, except when the complete list of scores included less than twice the number of awards granted. As a result, the overall number of applicants included in the comparison group (n = 105) was slightly less than the number of awardees (n = 120) over the period 1998–2007.
Next, we obtained the number of first-time NIH R01 and K grants received by the 1998–2007 CSDA alumni and comparison group by searching the NIH RePORT database.13 For each alumnus and applicant, we searched the RePORT database for the last name and first name and confirmed his or her institutional affiliation. We recorded all associated grants received from the year he or she applied to the CSDA program to 2011. When a name query returned no results for women, we made no attempt to determine whether her last name had changed because of marriage. A study of K08 and K23 awardees showed that only 13 of 734 women had name changes.14
Finally, in 2011, we administered an electronic survey to the 120 CSDA alumni from 1998 to 2007. We did not survey the applicants in the comparison group.
Survey risks included the public release of identifiable responses resulting from security breaches and the use of survey responses, including contact information, by third parties. To minimize these risks, we collected responses through SurveyMonkey and made the survey accessible through a Web link. We did not collect personal contact data, nor did we provide such data to third parties. We ensured privacy and confidentiality of individuals’ responses by storing the responses on the foundation’s secure servers, which are only accessible to foundation staff. We contacted survey respondents to inform them that certain data would be shared publicly in deidentified form and in aggregate only. We offered respondents the option to opt out by directly contacting us if they did not agree to share their deidentified data for publication.
The survey asked respondents to identify their current main professional activity from six categories based on those used by the American Medical Association—hospital-based practice, office-based practice, medical research, medical teaching, administration, and other.15 The survey also asked respondents to allocate 100% of their current professional time to each of the following categories in 5% increments—research, patient care, teaching, administration, and other.
We calculated measures of central tendency for continuous and ordered discrete variables and evaluated normality with the D’Agostino–Pearson omnibus test using the GraphPad Prism software version 5 (GraphPad Software, San Diego, California). When data were skewed, we used the median as the measure of central tendency. We used chi-square or Fisher exact test contingency analyses (as recommended based on the number of observations in the cells of the contingency table) to test for associations between categorical variables using the GraphPad Prism software. All tests were two-tailed, and we considered associations as significant if the P value was less than .05. We used multiple logistic regression to determine which predictor variables were independently associated with dichotomous outcome variables (Statpages.org).16
Characteristics of successful and unsuccessful applicants
The number of CSDA applications reviewed by the expert panels ranged from 77 to 172 per year in the 1998–2011 period (1,441 total) and doubled in 2006 compared with the previous years, possibly because of increased awareness of the award after the program was restarted in 2005 (see Supplemental Digital Figure 1, https://links.lww.com/ACADMED/A156). The success rate generally declined over the years, because of both an increase in the total number of applications and the variable number of awards granted per year. The number of applications funded varied each year because of fluctuations in the foundation’s endowment, growth, and other commitments. The mean and median success rates for the 1998–2011 competitions were both 13% (186/1,441). For comparison, the success rate for the NIH Mentored Patient-Oriented Research Career Development Award program (K23) from 2000 to 2005 was 41%.11
The demographic data available were limited to the 2006–2011 period. Of the applicants to the 2006–2011 competitions, 40% (352/871) were women, 31% (273/871) held combined MD/PhD degrees, and 59% (515/871) were white, 29% (251/871) Asian, and 9% (76/871) underrepresented minorities. Applicants’ ages ranged from 23 to 55 years, with a median age of 38 (interquartile range = 36–40, n = 867). Applicants’ gender, age, and race/ethnicity distributions were comparable from year to year. The percentage of applicants with MD/PhD degrees, however, grew from 23% (40/172) in 2006 to 38% (49/130) in 2011.
We analyzed the association between CSDA applicants’ success and specific characteristics by comparing successful and unsuccessful applicants (see Table 1). Gender, age, and underrepresented minority status were not significantly associated with funding attainment, although we found a nonsignificant but disconcerting suggestion that the percentage of underrepresented minorities among successful applicants (4%, 4/102) decreased compared with the percentage among unsuccessful applicants (10%, 72/740). A significantly higher percentage of successful applicants (44%, 46/104) than unsuccessful applicants (30%, 227/767) held MD/PhD degrees. Being affiliated with an institution in the top 95th percentile of NIH funding also was associated with funding attainment: 81% of successful applicants (84/104) but only 54% of unsuccessful applicants (415/767) came from these highly funded schools. In our multiple logistic regression, holding a combined MD/PhD degree and being affiliated with an institution in the top 95th percentile of NIH funding remained independent predictors of funding attainment (see Supplemental Digital Table 1, https://links.lww.com/ACADMED/A156).
Transition to independent research funding for 1998–2007 CSDA alumni and comparison group
To determine whether the CSDA program supports an investigator’s transition to independent research funding, we compared alumni who first received their grant at least four years earlier (awards granted from 1998 to 2007) with a comparison group of nonalumni. The 120 alumni who fit our inclusion criteria (see Figure 1) were affiliated with 52 different U.S. institutions at the time of their award, and the 105 nonalumni were affiliated with 44 institutions. The two groups were similar with respect to funding rank of the institution (66% [79/120] of alumni and 70% [74/105] of nonalumni institutions were in the top 95th percentile, P = .46) and other support (39% [47/120] of alumni and 39% [41/105] of nonalumni held NIH K awards, P = .99) (see Supplemental Digital Table 2, https://links.lww.com/ACADMED/A156). The percentage of women was not significantly different between the two groups but tended to be higher in the nonalumni group (31% [37/120] of alumni and 43% [45/104] of nonalumni were women, P = .054).
A greater percentage of CSDA alumni than nonalumni received at least one R01 grant as of 2011 (62% [74/120] of alumni and 42% [44/105] of nonalumni, P = .0045). From year to year, the percentage of alumni and nonalumni who received at least one R01 grant ranged from 45% to 71% and 33% to 50%, respectively. By comparison, 30% (298/1,002) of MD and MD/PhD combined K23 awardees from fiscal year (FY) 2000 to FY 2005 received an R01 grant versus 10% (76/785) of those who applied for a K23 grant in the same period but were not funded.11 However, neither the CSDA nor K23 percentages indicate success rates. Individuals for whom the NIH RePORT database did not indicate an R01 grant could have applied and been unsuccessful or could have never applied.
The average time to first R01 grant attainment for CSDA alumni was 4 ± 2 years (mean ± standard deviation) from the start date of the CSDA and was not statistically different for nonalumni. The association of being awarded a CSDA with receiving an R01 remained strong and significant after we controlled for gender, funding level of the individual’s institution, and federal career development grant support (see Table 2).
Research as the main professional activity for 1998–2007 CSDA alumni
A goal of the CSDA program is to cultivate physicians who are dedicated to research. In our survey of CSDA alumni from 1998 to 2007, we asked whether they remained in research and continued to spend the majority of their time conducting research after their CSDA. Of 120 eligible alumni, 92 (77%) responded to our survey. Of those, 78 (85%) indicated medical research as their main professional activity. According to the American Medical Association, only 2% of U.S. physicians report medical research as their primary activity.15 According to our respondents, the next most common choice for main professional activity was hospital-based practice (12%, 11/92) followed by administration (2%, 2/92) and office-based practice (1%, 1/92).
In the same survey, we also asked CSDA alumni to allocate their current working time to professional activity categories (see Table 3). For CSDA alumni who were 10 or more years from the start date of their grant, their median percent effort toward research activities was 68%. For those who were 4 to 9 years from the start date of their grant, their median percent effort was 75%.
With only 2% of all U.S. physicians reporting research as their main pro fessional activity15 and an estimated 34% to 44% attrition rate from academic medicine for first-time assistant professor physicians at U.S. medical schools,17 we must provide research support for early-career physician scientists. The Doris Duke Charitable Foundation provides this support through the competitive CSDA grant. In this analysis, we attem pted to determine the association between success ful CSDA funding and an investi gator’s transition to independent research funding. Our data showed that the CSDA is highly competitive and has supported early-career scientists who continue on to develop careers in research. Our analyses also showed that the funding rank of an investigator’s institution and his or her doctoral training combined with medical training (MD/PhD degrees) are both significant predictors of success in the CSDA competitions. We also found that success in the CSDA competitions is associated with an investigator’s transition to independent research funding—62% of alumni from 1998 to 2007 received an R01 grant compared with 42% of nonalumni.
Characteristics of successful and unsuccessful applicants
The number of applications to the CSDA competitions has increased over time, but the gender composition of the applicant pool has remained constant. The percentage of female applicants (40%) is comparable to the percentage of female assistant professors in academic medicine and science (43% female in 201218). Similar to studies of K23 awards,11,19 our data also show that the female CSDA applicants from 2006 to 2011 were as successful as the male applicants. An analysis of data that includes earlier competitions (see Supplemental Digital Table 2, https://links.lww.com/ACADMED/A156) comparing the top-ranked funded and unfunded applicants suggests that women might not have been as successful as men. We do not know the reasons for these possible differences.
We found a significant association between holding an MD/PhD degree and receiving a CSDA. Awardees are selected on the basis of their research accomplishments, leadership potential, research plan, and the strength of their research environment. The lower success rates of non-PhD MD applicants compared with their MD/PhD counterparts could reflect differences in grant crafting skills, research training, or accomplishments between the groups, or in factors related to the peer review process.
CSDA applicants affiliated with institutions in the top 95th percentile of NIH funding also were more likely to receive an award. Ginther and colleagues20 observed an increased probability of receiving an NIH research grant for applicants affiliated with institutions with high NIH funding. These higher success rates could reflect differences in research resources, mentors, or professional development opportunities for investigators at such institutions, or in factors related to the peer review process.
Funding organizations are concerned about diversity. For example, the NIH recently set up a Working Group on Diversity in the Biomedical Research Workforce. They created this working group partly in response to a 2011 study that found a decreased association between self-identifying as black and receiving an NIH research award.20 Private philanthropic organizations similarly are starting to analyze awardee data to address these issues.21 The percentage of CSDA applicants, for example, from different racial and ethnic groups is similar to estimates of that of full-time medical faculty,22 but some racial and ethnic groups remain underrepresented as compared with the U.S. population.23 Still, the percentage of underrepresented minorities in the CSDA applicant pool is similar to that of the K23 applicant pool.11 Although the percentage of underrepresented minorities among CSDA applicants is similar to that of other award programs, the difference relative to the percentage in the U.S. population, as well as the nonsignificant but troubling trend of a decrease in the percentage of CSDA awardees who are underrepresented minorities, indicates that we must consider strategies to enhance diversity in career development programs.
Transition to independent research funding
One of our most exciting findings was that the percentage of CSDA alumni who received first-time R01 awards was higher than that for nonalumni. This finding is particularly encouraging because this success of CSDA alumni may indicate an increase in human subject research funding, which traditionally has had a lower likelihood of receiving funding.24 Many factors could explain this finding. The CSDA program could truly have an impact on the research enterprise by providing flexible funds and protected time for research activities at a critical time in the development of early-career scientists. CSDA awardees also could benefit from the networking opportunities with each other, more senior Doris Duke grantees, and the Doris Duke Scientific Advisory Council. Alternatively, either the CSDA peer review process may identify the best scientists who would have succeeded with or without the award, or the recognition that comes with receiving a named award could have an impact on alumni receiving other awards. Similarly, physicians who received K23 awards were more likely than nonawardees to receive R01 grants.11 Both our findings and those of Haak and Schnell11 therefore support the concept that mentored career awards increase the likelihood of an investigator transitioning to independent research funding.
Our findings also suggest that the highest-ranked unsuccessful CSDA applicants were more likely to receive an R01 grant than physician investigators who held K23 awards. This finding could reflect the institutional nomination process prior to the CSDA peer review process, which was in place for most of the competitions from 1998 to 2007. Differences in the number of applications (the K23 program receives two to four times more applicants per year than the CSDA program and is not restricted by nominations) and funding cutoffs between our career development grant program and that of the K23 program also could explain this finding.
Research as physicians’ main professional activity
CSDA awardees are required to dedicate a minimum of 75% of their professional time to research activities during the term of their award. Because a goal of the CSDA program is to support individuals with potential for careers in research, we asked alumni whether research remained their main professional activity. We found that alumni remained in research and continued to dedicate the majority of their time to this professional activity. However, the time that these physicians allocated to research dropped from at least 75% to a median of 68% later in their careers. This decrease in the amount of time they devoted to research activities seems to be balanced by an increase in their patient care and administrative responsibilities and may be the result of a decrease in the amount of protected time they have for research activities after the end of their career development awards.
Our study has several limitations. First, it was observational, which restricts our ability to draw causal inferences. Ideally, we could have used a regression discontinuity design, such as that used to investigate K awards,11 wherein we compared funded and unfunded applicant groups with proposals that had overlapping priority scores. However, the number of individuals with overlapping scores between our alumni (funded) and nonalumni (unfunded) groups was small, which limited our sample size and precluded us from applying this approach.
Another limitation to our study is the fact that the funding mechanism has changed over the years—prefaculty fellows were supported early in the program, the time period of the grant changed from five to three years, the disease focus expanded, and the annual dollar amount increased. We were unable to discern whether and how these differences affected our findings because of limitations in sample size. To limit the potential bias introduced by the award being granted to fellows, however, we restricted our analysis to those awardees who completed at least three years of CSDA-supported research as a faculty member. We also limited our analysis of the demographic and educational predictors of CSDA attainment to data collected from 2006 to 2011 because data for all predictor variables were not available for the prece ding years. Therefore, the variables we tested might not be predictors of outcomes for all years of the CSDA program.
Another potential limitation is the fact that we were not able to separate Pacific Islanders from Asians when we analyzed data by race/ethnicity. The NIH includes Native Hawaiians and other Pacific Islanders as underrepresented minority groups. Our analysis of minorities, therefore, may slightly underestimate the actual number of minorities in the CSDA applicant pool.
Finally, we gathered data, which are not widely or publicly available, regarding a private foundation award to early-career physician scientists. Extracting data from historical records highlighted the importance for private foundations to both establish evaluation parameters and collect the corresponding data in a systematic and robust manner from the inception of a program. Lack of accessibility and completeness of data across all years of the CSDA program limited the scope of our analysis.
We found that the factors associated with success in a CSDA competition included a combined clinical and doctoral research degree and affiliation with a well-funded institution. Also, more CSDA alumni received NIH independent research funding than those who applied but did not receive the award. Therefore, we concluded that this privately funded Doris Duke Charitable Foundation early-career award was associated with physicians establishing independent and recognized research careers. Although we cannot completely attribute the success of alumni to the CSDA program, our findings suggest that supporting early-career physician scientists who have the potential for success in academic research is a valuable investment.
A significant lesson learned in this investigation was that the lack of evaluation plans at the inception of a program and the nonsystematic collection of data will limit the extent of future analyses. When possible, organizations should create prospective data collection plans to aid researchers in evaluating the impact of the program. Questions regarding the CSDA program remain to be answered, such as what the association is between an investigator’s specialty or subspecialty field and his or her successful attainment of early-career awards. These questions should inform future research.
Acknowledgments: We thank Jennifer Petitt at the American Institute of Biological Sciences for her efforts in data collection and the Scientific Advisory Council of the Medical Research Program at the Doris Duke Charitable Foundation for the members’ input into the design of the CSDA program and their contribution to the preparation of this article. We also thank the CSDA alumni and the peer reviewers.
1. Schafer AI. The vanishing physician–scientist? Transl Res. 2010;155:1–2
2. Wyngaarden JB. The clinical investigator as an endangered species. Bull N Y Acad Med. 1981;57:415–426
3. Lipira L, Jeffe DB, Krauss M, et al Evaluation of clinical research training programs using the clinical research appraisal inventory. Clin Transl Sci. 2010;3:243–248
4. Ley TJ, Rosenberg LE. The physician–scientist career pipeline in 2005: Build it, and they will come. JAMA. 2005;294:1343–1351
5. U.S. Department of Health and Human Services, National Institutes of Health. . Types of grant programs. http://grants.nih.gov/grants/funding/funding_program.htm
. Accessed July 17, 2013
6. Bradford WD, Anthony D, Chu CT, Pizzo SV. Career characteristics of graduates of a medical scientist training program, 1970–1990. Acad Med. 1996;71:484–487
7. Brass LF, Akabas MH, Burnley LD, Engman DM, Wiley CA, Andersen OS. Are MD-PhD programs meeting their goals? An analysis of career choices made by graduates of 24 MD-PhD programs. Acad Med. 2010;85:692–701
8. Mavis B, Katz M. Evaluation of a program supporting scholarly productivity for new investigators. Acad Med. 2003;78:757–765
9. Pion GM, Cordray DS. The Burroughs Wellcome Career Award in the Biomedical Sciences: Challenges to and prospects for estimating the causal effects of career development programs. Eval Health Prof. 2008;31:335–369
10. Fanzo JC, Gallin EK The Doris Duke Clinical Scientist Development Award: A Seven-Year Retrospective and Summary. 2006:11 Washington, DC In: Reinhart, GR, ed. Enhancing Philanthropy’s Support of Biomedical Scientists: Proceedings from a Workshop on Evaluation.
11. Haak L, Schnell J National Institutes of Health Individual Mentored Career Development Awards Program. August 29, 2011 Washington, DC National Institutes of Health
12. U.S. Department of Health and Human Services, National Institutes of Health. . Research Portfolio Online Reporting Tools (RePORT). NIH awards by location and organization. http://report.nih.gov/award/index.cfm
. Accessed July 17, 2013
13. U.S. Department of Health and Human Services, National Institutes of Health. . Research Portfolio Online Reporting Tools (RePORT). NIH RePORTER. http://projectreporter.nih.gov/reporter.cfm
. Accessed July 17, 2013
14. Jagsi R, Motomura AR, Griffith KA, Rangarajan S, Ubel PA. Sex differences in attainment of independent funding by career development awardees. Ann Intern Med. 2009;151:804–811
15. Smart DR Physician Characteristics and Distribution in the US, 2011. 2011 Chicago, Ill American Medical Association
16. Pezzullo JC. Logistic regression. http://statpages.org/logistic.html
. Accessed July 17, 2013
17. Alexander H, Lang J. The long-term retention and attrition of U.S. medical school faculty. AAMC Analysis in Brief. June 2008;8(4)
18. Jolliff L, Leadley J, Coakley E, Sloane RA Women in U.S. Academic Medicine and Science: Statistics and Benchmarking Report 2011–2012. 2012 Washington, DC Association of American Medical Colleges
19. Pohlhaus JR, Jiang H, Wagner RM, Schaffer WT, Pinn VW. Sex differences in application, success, and funding rates for NIH extramural programs. Acad Med. 2011;86:759–767
20. Ginther DK, Schaffer WT, Schnell J, et al Race, ethnicity, and NIH research awards. Science. 2011;333:1015–1019
21. Myers ER, Alciati MH, Ahlport KN, Sung NS. Similarities and differences in philanthropic and federal support for medical research in the United States: An analysis of funding by nonprofits in 2006–2008. Acad Med. 2012;87:1574–1581
22. Andriole DA, Jeffe DB, Hageman HL, et al Variables associated with full-time faculty appointment among contemporary U.S. medical school graduates: Implications for academic medicine workforce diversity. Acad Med. 2010;85:1250–1257
23. Humes KR, Jones NA, Ramirez RR. Overview of race and Hispanic origin: 2010. 2010 Census Briefs. 2011 Washington, DC U.S. Census Bureau
24. Ginther DK, Haak LL, Schaffer WT, Kington R. Are race, ethnicity, and medical school affiliation associated with NIH R01 type 1 award probability for physician investigators? Acad Med. 2012;87:1516–1524