Fig 1.
Length of time awardees remain in NIH applicant pool after after the first R01-e Award.
A Kaplan–Meier approach was used to measure the length of time investigators in each cohort remained in the NIH R01-e applicant pool after receiving their first R01-e awards. Y-axis: percent of investigators in each cohort who received an additional RPG award and remain in applicant pool. Investigators who do not remain in pool are considered to have ‘dropped out’. X-axis: years since receiving first R01-e award. Blue line: NIAID awardees. Orange line: other-NIH awardees. Solid red line: dropout slope between 4 and 5 years. Half of the NIAID cohort dropped out by 15 years after the first R01-e award (i.e. half-life 15 years); half of the other-NIH cohort dropped out by 10 years, or 50% sooner than the NIAID cohort.
Table 1.
PI SCORECARD: Grant submission behaviors and grant quality indices.
Table 2.
Number of ENI per cohort year and percent of total cohort.
Table 3.
ENI demographic characteristics.
Table 4.
ENI background and index institution characteristics.
Table 5.
Institution ENI density.
Fig 2.
ENI index award grant activity types.
Table 6.
Index award characteristics.
Table 7.
ENI funding outcomes per cohort year.
Table 8.
Percentage of ENI Who remained in R01-e applicant pool 5 or More years after index award.
Table 9.
Funding outcomes according to ENI demographic characteristics.
Table 10.
Funding outcome according to PI background and index institution.
Fig 3.
ENI funding outcomes by index institution ENI density.
Index institution density tertiles: 1 = 14 to 37 ENI per institution; 2 = 6 to 13 ENI per institution; 3 = 1 to 5 ENI per institution. ENI from institutions in tertiles 1 and 2 were more likely to be funded (70% and 65%, respectively) than ENI from institutions in tertile 3 (47%) (p < 0.0001). The highest ENI funding rate (70%) was in the 1st tertile, which included just 21 institutions.
Table 11.
Funding outcomes according to PI SCORECARD factors.
Fig 4.
ENI index award scores by cohort year.
Index scores varied from cohort year to cohort year, a result of normal NIAID R01 payline changes from year to year. Unexpectedly, index award scores of funded and unfunded ENI were statistically different in cohort years 2003, 2005, 2006, 2007, and 2009. In 2004, 2008 and 2010, index scores were not statistically different.
Fig 5.
Applications scored and triaged from funded and unfunded ENI.
Figure includes 10,228 applications from 1,322 ENI who submitted applications after the IFY. (Index applications are excluded.) Funded ENI submitted 8,026 applications, of which 39% were triaged; unfunded ENI submitted 2,202 applications, of which 63% were triaged. The number of applications from both groups increased between 2003 and 2010, while the cohort was still growing (black dotted line), but more rapidly from funded ENI. Funded ENI consistently had fewer of their applications triaged, even in the early years, suggesting they had an early advantage in grant writing ability.
Fig 6.
Average PI ANNUAL scores, FY 2011 –FY 2016.
The PI ANNUAL Score each year is the average of individual PI ANNUAL scores of ENI who submitted scored applications. Average PI ANNUAL Scores were markedly different between the funded and unfunded ENI, with funded ENI having PI ANNUAL Scores about 10 percentile points lower each year between 2011 and 2016. (Welch Two Sample t-tests: in all years p-values < 0.0001.).
Table 12.
Univariate regression of independent variables on ENI funding success.
Table 13.
Multivariate regression of independent variables on ENI funding success.
Fig 7.
Variable importance in prediction of ENI funding success.
RFVI analysis ranked all independent variables included in multivariate modeling in order of importance in predicting ENI funding success. The strongest predictor was the percent of the PI’s applications triaged (QUALITY), followed by the PI’s: average applications per year (FREQUENCY); percent of renewal applications (RENEW); average application score (SCORE); percent of resubmissions (RESUB) and applications to multiple NIH ICs (REACH); and the index institution ENI density. Having a K award, a renewal index award and a US index institution, were the least important predictors. All of the other variables had approximately equal predictive strength.