In September, 2010, the National Research Council (NRC) released A Data-Based Assessment of Research-Doctorate Programs in the United States1 (referred to here as the Assessment), a report describing an extensive database of data and rankings from more than 5,000 doctoral programs, 982 of which were in the biomedical sciences. As part of its support for this project, the National Institutes of Health (NIH) asked the NRC to examine data on the biomedical sciences programs to see if they could shed light on specific questions about research training and support, many of which were highlighted in Investing In the Future, National Institute of General Medical Sciences Strategic Plan for Biomedical and Behavioral Research Training2.
Given its substantial investment in doctoral research training, NIH was particularly interested in the following questions:
1) In fields such as biochemistry, where programs are housed in both medical schools and in arts and sciences faculties, are there apparent differences in median time to degree and completion rates?
2) What correlations exist between student median time to degree and completion rates and other characteristics of the programs, e.g.,
a) What is the correlation between students’ median time to degree and the publication rates of faculty in their program?
b) What is the correlation between GRE scores and student median time to degree and completion rates?
c) Do programs that offer additional student activities, such as writing workshops, career seminars, etc., have longer times to degree, on average?
3) What are the correlations between the diversity of a program’s faculty and the diversity of its students, both with regard to underrepresented minorities and women?
4) A large number of programs in the biomedical sciences classified themselves as “Integrated biological science” programs and span the biomedical sciences. Are these programs different in observed characteristics from the programs in which students specialize in a specific area from the outset of doctoral study?
NIH also encouraged the panel to discuss other relevant issues.
1 National Research Council, 2011. A Data-Based Assessment of Research-Doctorate Programs in the United States. Washington, DC: The National Academies Press. The report and accompanying data table can be found at www.nap.edu/rdp. A corrected data table was published on April 29, 2011.
2Investing In the Future, National Institute of General Medical Sciences Strategic Plan for Biomedical and Behavioral Research Training 2011.Bethesda, MD. National Institutes of Health.
TABLE S-1 Fields in the Biomedical Sciences in the Assessment of Research-Doctorate Programs and Number of Programs Included in Each Field
|Field Name||Number of Programs|
|Biochemistry, Biophysics, and Structural Biology||157|
|Biomedical Engineering and Bioengineering||74|
|Cell and Developmental Biology||120|
|Genetics and Genomics||66|
|Immunology and Infectious Disease||68|
|Integrated Biological and Biomedical Sciences||113|
|Neuroscience and Neurobiology||93|
|Pharmacology, Toxicology, and Environmental Health||117|
At the outset, it is important for the reader to understand the sources and some of the limitations of the data used to produce the correlations and other descriptions in this report. The committee authoring the Assessment identified several sources of errors in the data that could not be eliminated, including classification errors and data collection errors (see Box 2-1). The omission of field-specific measures, such as books, patents, and articles presented at refereed conferences in some science and engineering fields, means that the data do not capture the full scope of a program’s research productivity. Once the data were released, institutions and others identified additional problems, which led to the release of a corrected data table in April, 2011. In addition to data from the Assessment, data on training grants and training slots were collected from the NIH website.
The panel created pairwise correlations for a dozen characteristics of biomedical science programs (variables)3 of interest to NIH:
|Average Publications per Faculty Member||Average GRE Scores|
|Average Citations per Publication||Percent of Non-Asian Minority Students|
3 Definitions of these and other relevant variables used in the Assessment are found in Appendix C.
|Percent of Faculty with Grants||Percent of Female Students|
|Percent of Non-Asian Minority Faculty||Average Ph.Ds per Year, 2002-2006|
|Percent of Female Faculty||Average Cohort Completion Rate|
|Awards per Faculty Member||Median Time to Degree|
The correlations provide insights into the relationships between characteristics that can be explored further. The panel focused its attention on correlation coefficients greater than or equal to 0.34 (highlighted in the report) because they are nontrivial and they may display, in the panel’s view, important relationships between program characteristics. When important correlations are found, further analyses will be required, adjusting for potential confounding variables, to better understand the causal relationships. Such adjustments are beyond the scope of this brief report.
RESPONSES TO QUESTIONS IN THE STATEMENT OF TASK
1) Comparison of Median Time to Degree and Completion Rates in Programs Housed in both Medical Schools and Arts and Sciences Schools
The panel was unable to shed much light on the differences between programs in the same field housed in medical schools and in arts and sciences schools, because the data that the institutions provided for the Assessment were not specific enough to draw these distinctions among individual programs. We did conduct an email inquiry of institutions with medical schools, asking where their biomedical science programs were located administratively, but not enough information was obtained, and too many ambiguities existed, to provide reliable comparisons.
2a) Correlation of Median Time to Degree or Completion Rates with Faculty Research Productivity
The panel found correlations greater than or equal to 0.3 in six fields between the average student time to degree and various measures of faculty research productivity: publications per faculty member, citations per publication, and the percent of faculty with grants. Where appreciable correlations exist, greater faculty research productivity is associated with longer times to degree. We found weaker relationships between the average cohort completion rate and faculty research productivity, with the exception of physiology.
4 Correlations of 0.295 and higher were rounded to 0.3.
TABLE S-2 Fields with Correlations ≥ 0.3 Between Median Time to Degree or Completion Rates and Faculty Research Productivity
|Field||Correlation ≥ 0.3|
|Biomedical Engineering and Bioengineering||Median Time to Degree with Average Cits/Pubs|
|Genetics and Genomics||Median Time to Degree with Average Cits/Pubs|
|Immunology and Infectious Disease||Median Time to Degree with Average Cits/Pubs|
|Microbiology||Median Time to degree with % of Faculty w/ Grants|
|Nutrition||Median Time to degree with Average Pubs/Faculty|
|Physiology||Median Time to Degree with Average Cits/Pubs|
|Physiology||Median Time to degree with % of Faculty w/ Grants|
|Physiology||Completion Rate with % of Faculty w/ Grants|
2b) Correlation of Median Time to Degree or Completion Rates with GRE Scores and Average Number of Ph.D.’s
GRE General Test scores do not have correlations greater than or equal to 0.3 with median time to degree in any fields except microbiology and nutrition, where students with higher GRE scores have longer times to degree. The correlations between completion rates and both average GRE scores and average number of Ph.D.’s are uniformly low, and in several fields are negative (Table 3-2). The exception is physiology.
TABLE S-3 Fields with Correlations ≥ 0.3 Between Median Time to Degree or Completion Rates and Average GRE Scores or Average Number of Ph.D.’s
|Field||Correlation ≥ 0.3|
|Biomedical Engineering and Bioengineering||Median Time to Degree with Average Number of Ph.D.’s|
|Microbiology||Median Time to Degree with Average GRE Scores|
|Nutrition||Median Time to Degree with Average Number of Ph.D.’s|
|Nutrition||Median Time to Degree with Average GRE Scores|
|Physiology||Completion Rate with Average Number of Ph.D.’s|
2c) Correlation of Median Time to Degree with Student Activities
The panel did not conduct an analysis of the possible correlations between median time to degree and student activities such as writing workshops and career seminars. Preliminary examination of the overall data on student activities made it clear that these types of activities are offered in most doctoral programs, so correlations with other variables like median time to degree would be small.
3) Correlation of Faculty Diversity with Student Diversity
The correlations on diversity demonstrate a strong relationship between underrepresented minority (URM) faculty and URM students in six of the eleven biomedical science fields:
TABLE S-4 Fields with Correlations ≥ 0.3 Between Percent of Underrepresented Minority (URM) Faculty and Percent of URM Students
|Field||Correlation of % URM Faculty with % URM Students|
|Biochemistry, Biophysics, and Structural Biology||0.489|
|Integrated Biological and Biomedical Sciences||0.529|
|Pharmacology, Toxicology, and Environmental Health||0.370|
Potential factors associated with increased URM student enrollment are explored in Chapter 5.
With regard to gender, the panel found no meaningful correlation between the percent of female faculty in a program and the percent of female students; the correlations are below 0.3 in every biomedical science field. The highest correlation (0.288) is in nutrition, where over 50 percent of the faculty and over 75 percent of the students are female.
4) Comparison of Programs in Integrated Biological and Biomedical Sciences with Other Fields
The panel took a close look at the programs in the field of integrated biological and biomedical sciences. We wanted to use this diverse field to identify the programs in which students typically spend one year sampling research in different laboratories and then choose an area of specialization. However, the responding institutions provided data for individual fields, even when those fields were part of an umbrella program.
Using data from the Assessment survey of doctoral programs, the panel examined the ratio of the number of students who enrolled to the number of students who received offers of admission to see if the integrated biological and biomedical science programs were more popular (as indicated by a higher enrolled-to-offered ratio), but did not find any evidence of this.
In addition to the specific questions outlined in the statement of task, the panel used the data from the Assessment to explore in a preliminary way several related topics: the relationship of and completion rates to student funding; potential factors associated with increased URM
enrollment; doctoral student experience and related characteristics in neuroscience and neurobiology; and the number and location of postdoctoral fellows.
Median Time to Degree, Funding, and Completion Rates
Median time to degree is relatively constant across programs: medians range between 4.88 and 5.73 years for all biomedical science fields. In almost all programs, more than 90 percent of students are fully funded in the first two years, about one-quarter with an institutional fellowship and the rest through either a traineeship or research assistantship (Table 4-1). By the third year, almost all students are funded through some combination of research assistantships and traineeships. Since funding for the biomedical sciences comes primarily from NIH, the agency can use its influence to encourage program practices in the biomedical sciences in a way that is not available for other fields in science and engineering.
As might be expected, a shorter median time to degree correlates with a higher completion rate; in at least six fields the correlation coefficient is < -0.3.
TABLE S-5 Correlations Between Median Time to Degree and Average Completion Rate by Field
|Field||Median Time to Degree (years)||Average Completion Rate (%)||Correlation|
|Biochemistry, Biophysics, and Structural Biology||5.63||45.9||-0.375|
|Biomedical Engineering and Bioengineering||5.06||46.3||-0.134|
|Cell and Developmental Biology||5.66||50.1||-0.383|
|Genetics and Genomics||5.73||41.6||-0.451|
|Immunology and Infectious Disease||5.36||56.2||-0.071|
|Integrated Biological and Biomedical Sciences||5.62||47.4||-0.362|
|Neuroscience and Neurobiology||5.68||46.2||-0.464|
|Pharmacology, Toxicology, and Environmental Health||5.21||56.1||-0.260|
A Deeper Analysis of Underrepresented Minorities
The Role of Training Grants
Associating NIH training grants with the university to which each program belongs, the panel investigated two questions about the relationship of training grant awards to underrepresented minority (URM) students and to international students.
Do institutions with heavy dependence on training grants recruit more students who are from underrepresented minority groups (URMs) than schools with less dependence?
The panel found that institutions with a large number of training grants do have more minority graduate students, but the programs are larger, and the correlation between the number of training grants and the percentage of minority students is 0.00.
Do institutions with heavy dependence on training grants enroll fewer international students?
The same holds true with regard to international students. In fact, the correlation between the number of training grants and the percentage of international students is slightly negative (-0.240). Since international students cannot be supported on NIH training grants, this correlation is not surprising. Thus, having more training grants does not appear to increase the fraction of minority students or international students.
A Statistical Approach to Factors Associated with URM Enrollment
Simple correlations cannot tell the whole story, and the panel also developed a statistical model that relates enrollment by URM’s to other program characteristics, in order to better understand how to expand URM enrollment and graduation from PhD programs. The model involved answering the three questions below.
How many URM graduates are expected per year across all programs?
Of the approximately 4,700 new Ph.D.’s per year in the biomedical sciences in 2002-2006, roughly 550 (11.7 percent) were URM graduates (Figure 5-4). Based on these numbers, only 17 percent of the biomedical science programs are expected to graduate more than one URM student per year, and only three percent of programs are expected to graduate two or more.
What factors predict higher URM enrollment in a Ph.D. program?
The panel attempted to predict the expected enrollment rate of URM students as a function of three factors (although other variables such as completion rate or percent of first year students with full financial support could also be used):
- number of URM faculty;
- research productivity as measured by the 5th percentile of the NRC “research productivity” ranking; and
- biomedical science field.
As would be expected, the fraction of URM faculty is a very strong predictor of URM student enrollment; overall, an increase in URM faculty members from 10 to 20 percent is associated with an increase in the fraction of URM enrolled students from 10 to 40 percent (a factor of 3). Faculty research productivity is not a strong predictor of URM PhD student enrollment, once the number of Ph.D. students is taken into account.
Having controlled for these factors, to what extent does URM enrollment cluster within universities, and which universities exceed URM enrollment expectations?
Since many universities have a large number of Ph.D. programs in the Assessment, the panel also investigated whether there are unmeasured characteristics of each university that attract URM Ph.D. students beyond the predictors considered above, i.e., whether URM students cluster in biomedical science programs at a given university due to a random “university effect” that is common to all the programs within that institution. Table 5-3 lists the 10 universities with the highest values, which indicate the attractiveness of the university’s programs to URM students beyond what is predicted by the field, percent of URM faculty, and research productivity.
Profile of Neuroscience and Neurobiology
The panel also looked at data from the Assessment’s survey administered to students admitted to candidacy in neuroscience and neurobiology and compared these results to other science and engineering fields included in the student survey (chemical engineering and physics). Although we do not have similar data for the other biomedical fields, we found that 95 percent of the neuroscience and neurobiology students were somewhat or very satisfied with their training program, and (along with chemical engineering) they reported the highest levels of student productivity in research presentations and publications. Neuroscience and neurobiology students were more likely to have their academic progress assessed by program faculty than students in the other surveyed fields, and 86 percent of the programs collected data on students’ postgraduation employment. As in the other biomedical science fields, the percent of female faculty in neuroscience and neurobiology (26 percent) did not correlate with the percent of female students (52 percent), but it did correlate with shorter times to degree (-0.346).
Not surprisingly, most faculty members in every biomedical science field have spent time as postdoctoral scholars, with older faculty having a smaller percentage of people with postdoctoral experience. About 90 percent of the faculty who received their Ph.D.’s in the 1990s, e.g., have held postdoctorates, except for those in biomedical engineering and nutrition. Postdoctorates are concentrated in the largest programs, and they are also concentrated in the programs that are in the top two quartiles for research productivity (Table 7-4). The largest numbers of postdoctorates are being trained in, and presumably are contributing to, the most productive research environments.