Topic choice contributes to the lower rate of NIH awards to African-American/black scientists

Abstract

Despite efforts to promote diversity in the biomedical workforce, there remains a lower rate of funding of National Institutes of Health R01 applications submitted by African-American/black (AA/B) scientists relative to white scientists. To identify underlying causes of this funding gap, we analyzed six stages of the application process from 2011 to 2015 and found that disparate outcomes arise at three of the six: decision to discuss, impact score assignment, and a previously unstudied stage, topic choice. Notably, AA/B applicants tend to propose research on topics with lower award rates. These topics include research at the community and population level, as opposed to more fundamental and mechanistic investigations; the latter tend to have higher award rates. Topic choice alone accounts for over 20% of the funding gap after controlling for multiple variables, including the applicant's prior achievements. Our findings can be used to inform interventions designed to close the funding gap.

Fig. 1. Funding gap between AA/B and WH scientists at each stage of the R01 application and review process.

Arrows on the left indicate the number of AA/B and WH R01 applicants in FY 2011–2015. The total number of applicants with a reported race/ethnicity is 45,998. Rocket charts depict the number of applications that were submitted, discussed, and funded per applicant. Comparative rates of discussion, funding of discussed applications, and overall funding rates are presented on the top right (**P < 0.01).

Fig. 2. Effect of impact score on discretionary funding and resubmission rates.

(A) The distribution of percentile scores for funded and unfunded Types 1 and 2 R01 applications submitted by AA/B (red bars) and WH scientists (blue bars). (B) Resubmission rates by impact score range for unfunded, unsolicited Type 1 R01 applications (FY 2011–2015) from AA/B and WH applicants and (C) AA/B and WH applicants by career stage. ND indicates applications that were not discussed and therefore not scored. All pairwise comparisons between resubmission rates for AA/B and WH applicants within each impact score range in (B) and (C) are not statistically significant (P > 0.07).

Fig. 3. Distribution of applications from AA/B scientists across topics.

(A) Red bars show the percent of applications from AA/B scientists in each topic cluster, ranked from highest to lowest. Clusters were initially defined based on content similarity; thus, clusters that are numerically close also tend to have relatively similar content. Of all applications from AA/B scientists, 37.5% belong to the first eight clusters; at the other end of the distribution, eight clusters contain no applications from AA/B scientists. Because of space constraints, every other cluster number is reported on the x axis; cluster numbers for the first and last eight clusters are highlighted on the graph. (B) Number of applications in (orange bars) and (C) award rate for (blue bars) each topic cluster, ranked by percentage of applications from AA/B scientists in each cluster [i.e., same ranking as in (A)]. The dashed red line represents the overall R01 award rate (16.3%). In the 25 clusters with a significantly above average award rate (see table S6), the number of applications from AA/B scientists was too small to determine how they fared relative to applications from WH scientists.

Fig. 4. Topics most and least commonly proposed by AA/B scientists.

(A) Topic clusters with the highest percentage of applications from AA/B scientists. (B) Topic clusters with no applications from AA/B scientists. Word clouds are placed in a clockwise orientation relative to the order shown in Fig. 3A. Cluster numbers are presented alongside overall award rate (cluster number/award rate). (C) Distribution of applications and awards for AA/B scientists across topics in the NIH portfolio. Each node in the network represents a topic cluster, and related topic clusters are grouped inside blue borders and labeled (rectangles). Node size correlates with the number of applications from AA/B scientists, and nodes are heat mapped by the number of funded applications from AA/B scientists in each cluster. GI, gastrointestinal.