Biases in grant proposal success rates, funding rates and award sizes affect the geographical distribution of funding for biomedical research

Abstract

The ability of the United States to most efficiently make breakthroughs on the biology, diagnosis and treatment of human diseases requires that physicians and scientists in each state have equal access to federal research grants and grant dollars. However, despite legislative and administrative efforts to ensure equal access, the majority of funding for biomedical research is concentrated in a minority of states. To gain insight into the causes of such disparity, funding metrics were examined for all NIH research project grants (RPGs) from 2004 to 2013. State-by-state differences in per application success rates, per investigator funding rates, and average award size each contributed significantly to vast disparities (greater than 100-fold range) in per capita RPG funding to individual states. To the extent tested, there was no significant association overall between scientific productivity and per capita funding, suggesting that the unbalanced allocation of funding is unrelated to the quality of scientists in each state. These findings reveal key sources of bias in, and new insight into the accuracy of, the funding process. They also support evidence-based recommendations for how the NIH could better utilize the scientific talent and capacity that is present throughout the United States.

Keywords: Biomedical research; Federal funding; Science policy.

Figure 1. Interquartile analyses of data sets.

Values from Table S1 were binned by per capita funding quartile. Box and whisker plots (median and quartile ranges) and mean values (+) are for: (A) per capita funding; (B) average award size; (C) per investigator funding rate; (D) per application success rate; and (E) scientific productivity. Overall probability values (inset) are from Kruskal-Wallis Test; adjusted p values below X axis are for significant interquartile differences by Dunn’s Post Test. Values are means of 2004–2013 data except productivity, which are sum of 2011–2013 publications citing 2011 awards normalized to funding in that year.

Figure 2. Factors affecting disparities in funding.

Plots show linear regressions of state per capita RPG funding as a function of: (A) average award size; (B) per investigator funding rate; (C) per application success rate; and (D) scientific productivity. Values are as described in Fig. 1 and numerical values by state can be found in Data S1. Regression analyses of log-transformed data are in Supplemental Information 1.