National Institutes of Health (NIH) grant awards: does past performance predict future success?

The National Institutes of Health (NIH) funds most of the federal biomedical research within the United States. The investigator-initiated R01 grant is the primary NIH grant mechanism and has increasingly been treated as a measure of faculty value, particularly at medical schools. Success in R01 funding influences recruitment, salary, promotion, tenure, resources, and space allocation. Schools of Medicine/Academic Medical Centers (SOMs/AMCs) compete to increase rankings by increasing numbers of R01 awards, often prioritizing recruitment of Principal Investigators (PIs) with multiple R01s, anticipating long-term return on investment (ROI). These recruits typically require higher salaries, large recruitment packages, and more space, implicitly assuming sustained funding success. There has been little data to test this assumption. The authors analyze NIH R01 awardees to evaluate whether past grant success predicts future success and report that well-funded PIs revert to the NIH-wide mean funding profile within 10–15 years. They argue that balanced recruitment and development of young and existing PIs are likely crucial for long-term institutional success.

Data sources: NIH grant data were exported from NIH RePORTER (https://projectreporter.nih.gov/reporter.cfm) and Palgrave Dataverse (https://doi.org/10.7910/DVN/GECEVA). Scope: All R01s in the category "NIH US Schools of Medicine" for fiscal years (FY) 1999–2015 were analyzed at probe years 2000, 2005, 2010, and 2015. Average R01s (excluding supplements) per year were 14,676 (range 12,870–16,128). For non-R01 mechanisms, U01, P01, P30, and P50 grants were exported (average 1,696/year; range 1,489–1,870). Identification of investigators: NIH Contact Person ID was used to identify discrete investigators; number of R01s per contact PI (excluding supplements) was counted. For multi-PI grants, sole credit was assigned to the contact PI. Funding amounts and normalization: Total cost per award (direct + indirect + administrative/other supplements) was summed. Award values were inflation-adjusted to 2015 dollars using the NIH Biomedical Research and Development Price Index (BRDPI). For visualization, total costs were converted to R01 equivalents: one R01 equivalent defined as up to $375,000 total cost (approximate $250,000 modular direct cost plus simulated 50% F&A). Two equivalents corresponded to $375,001–$750,000, etc. NIH average profiles: For each probe year, NIH-wide average funding profiles were generated directly from NIH RePORTER, bracketing each probe year with the adjacent years to allow error estimation (e.g., profile for 2000 is the average of 1999–2001). Estimating unfunded investigators: Three approaches were used to estimate investigators without an R01. (1) From NIH-published counts of discrete R01-equivalent applications per year, subtract applications for non-R01 grants and the number of funded R01s to estimate unfunded R01 applications; convert to a percentage by dividing by the sum of discrete applications and R01-funded investigators, approximating the total number of investigators seeking R01 funding. Assumption: most investigators do not submit multiple R01s in the same fiscal year. (2) Directly count individuals funded in the probe years who lost all funding during the 15-year analysis (total cohort n = 24,866 discrete investigators). (3) Compare estimates with published NIH unfunded rates (Lauer, 2018). All three methods yielded similar percentages. Cohort analyses and metrics: Investigators in FY2000 were grouped by initial portfolio size (1, 2, 3, 4, or ≥5 R01s). For each cohort, average number of R01s held 5, 10, and 15 years later was computed. Distributional profiles after 5, 10, and 15 years were plotted (ending with 0, 1, 2, 3, 4, or ≥5 R01s). Outperformance relative to the NIH mean was quantified as the proportion of investigators above the NIH average funding profile (area between curves) at each interval. For cohorts where all were below the mean (e.g., 1 R01), negative area was used. Parallel analyses were conducted using total dollars in R01 equivalents (inflation-adjusted) for cohorts starting with 1, 3, 5, 7, or ≥9 R01-equivalents. Sensitivity analysis (retirement bias): A restricted cohort of "continuously funded" investigators (had at least one R01 in both 2000 and 2015) was analyzed to mitigate retirement/departure effects. In this analysis, 5+ R01 starters could not be included due to only four investigators meeting the criterion. Treatment of MPIs and other funding: Multi-PI R01s (introduced 2006; 5% of R01s by 2010, 15% by 2015) were credited to the contact PI only, consistent with NIH and Blue Ridge practices. Funding from foundations, NSF, and DoD was not included due to lack of comparable transparency; these sources contribute fewer biomedical research dollars and are not counted in Blue Ridge rankings.

• NIH-wide distribution of R01 portfolios among funded investigators was stable from 2000–2015: 73.8 ± 2.2% held a single R01 (range 71–78%), 20.8 ± 1.7% held two (18–23%), 4.3 ± 0.5% held three (3–5%), and 0.9 ± 0.2% held four or more. Mean = 1.33 ± 0.03 R01s per R01-funded investigator (n ≈ 14.7k R01s/year).
• Including unfunded investigators (estimated 53–66% between FY2000–2015; consistent with NIH 66% in FY2017), the corrected average is 0.56 ± 0.06 R01s per investigator.
• Cohort outcomes after 15 years (starting in 2000): • 1 R01 starters: average 0.29 R01s in 2015 vs NIH-wide 2015 average 0.52 per investigator; 79% had zero R01s after 15 years. • 2 R01 starters: average 0.53; ~60% had zero R01s after 15 years. • 3 R01 starters: average 0.68. • 4 R01 starters: average 0.69. • ≥5 R01 starters: average 1.1; outperformed NIH average by only 0.58 R01s per PI after 15 years.
• Loss of R01 funding over time: • After 5 years: 20–30% of 2+ R01 holders and 55% of single R01 holders had no R01 funding. • After 10 years: ~40% of 2+ R01 and 70% of single R01 holders had no R01 funding. • After 15 years: ~60% of 2+ R01 and 79% of single R01 holders had no R01 funding.
• Outperformance relative to NIH mean: • 2–4 R01 cohorts: 37% above mean at 5 years; 21% at 10 years; only 5–10% above mean at 15 years. • ≥5 R01 cohort (≈0.2% of investigators): 22% above mean at 15 years.
• Total funding (R01-equivalents, inflation-adjusted to 2015): NIH-wide average in 2015 was 0.89 R01-equivalents per investigator. Portfolio distributions by R01-equivalents changed little over 15 years, with the largest number of investigators holding two equivalents. Cohorts with 1–2 equivalents underperformed the mean after 15 years; 3+ equivalents only slightly exceeded the mean. Only a small subset of highly funded investigators outperformed the total-dollar mean over the long term.
• Continuously funded subset (had ≥1 R01 in both 2000 and 2015): After 15 years, regardless of starting with 1, 2, 3, or 4 R01s, cohorts converged near the NIH-wide average among funded investigators (1.33 R01s/investigator). Only 13–21% of those starting with 2–4 R01s outperformed the mean in 2015.

Findings show that NIH grant success exhibits regression to the mean: regardless of initial R01 portfolio size, investigator funding profiles converge toward the NIH-wide average over 10–15 years. This mirrors long-term performance in domains such as active fund management, where past outperformance does not predict future outperformance. The results indicate institutions are unlikely to sustain above-average returns by recruiting faculty based primarily on past grant success. The analysis also informs debates about concentration of NIH funding among a small subset of investigators: if systemic long-term bias favored the same investigators, highly funded PIs would sustain disproportionate funding. Instead, all cohorts regress at similar rates, offering no evidence of sustained systemic advantage over long horizons. Strategically, institutions should broaden evaluation criteria beyond grant counts to include service, clinical activities, mentorship, education, and public trust when making recruitment and support decisions.

Past NIH grant performance is not a strong predictor of sustained future funding success. Across cohorts defined by initial R01 portfolio size, funding regresses to the NIH-wide mean within 10–15 years, both in grant counts and total (inflation-adjusted) dollars. Very few investigators maintain long-term outperformance relative to the NIH average. Institutions relying heavily on past grant success for recruitment and resource allocation may face poor long-term ROI; balanced strategies that invest in and develop early-career and existing investigators and incorporate broader mission-based criteria are recommended. Future research could examine other institutional types beyond SOMs/AMCs, incorporate demographic analyses (e.g., early-stage and underrepresented investigators), and assess the roles of retirement, alternative funding sources, and collaboration structures (e.g., multi-PI awards) on funding trajectories.

• Institutional scope: Analysis restricted to US Schools of Medicine/Academic Medical Centers; findings may not generalize to other institutional types.
• Demographics: Investigators were not segmented by demographics; prior work suggests early-stage and minority investigators may be at higher risk of funding loss.
• Retirement/departure: Exact retirement status could not be ascertained; retirement may correlate with loss of funding. A sensitivity analysis of investigators with at least one R01 at both start and end still showed regression to the mean, suggesting retirement is not the primary driver.
• Multi-PI treatment: Credit for R01s, including MPIs, was assigned solely to the contact PI; some investigators categorized as unfunded may have been non-contact MPIs.
• Other funders: Foundation, NSF, and DoD awards were not included due to limited transparency and are not part of Blue Ridge rankings; their omission could affect individual funding profiles but likely not NIH-centric trends.
• Continuously funded subset constraints: Very small number of ≥5 R01 starters remained funded at both endpoints (only four), limiting analysis of this extreme cohort.
• Estimation of unfunded investigators relies on assumptions about application behavior (e.g., minimal multiple submissions per year), though triangulation with NIH data yielded similar values.