Still funded, no longer counted: how NIH's 2025 award reviews changed what the government counts as minority health research

2026-10-02

AI Summary

This study addresses the misclassification and undercounting of minority health grants following the 2025 NIH removal of DEI terminology. By applying natural language processing and large-scale text mining to 37,000 NIH-funded projects, validated through statistical regression and blinded review comparisons, this work introduces the novel concepts of “lexical targeting” and “demographic targeting” to quantify how textual editing distorts the veracity of scientific statistics. The analysis demonstrates that eliminating race-related keywords caused 98% of projects to lose their minority-health designation, with only 15.7% retaining their original classification. However, blinded review revealed that 84% (42/50) of these projects substantively still met the definitional criteria. These findings expose a distortion effect wherein policy-driven lexical sanitization renders legitimate scientific contributions effectively invisible.

Abstract

Funders know their portfolios through software that classifies award text. The US National Institutes of Health (NIH) reports its spending in more than 300 categories mined this way, and work on classification and indicators treats the text as the applicant's to write. In 2025 NIH required "DEI language" removed from awards not supporting DEI activities, and so policed the words it also counts. We followed population names through 37,790 continuing awards, checking them against practice records. Names were informative: in new awards with trial baselines, a title naming Black populations predicted a 60-percentage-point higher enrolled share. Text features predicted which names survived, and practice records added little. NIH's minority health category followed the names: of continuing projects carrying it, 99.4% with unchanged summaries kept it, against 15.7% of those whose summaries no longer named a racial or ethnic population, while the projects kept their funding. Blinded reviewers judged 42 of a random 50 such losses to have met its definition in FY2024. Comparing both summaries of 70 name losses, they found aims concerning the population recast in 49. Read alone, 63 FY2025 summaries no longer met the definition. Awards naming sexual and gender minorities were recorded as terminated 37.8 points more often after adjustment for listed terms, institute and activity. We call the mechanism word targeting, its boundary population targeting, and its product uncounted science: funded research the count no longer records. The count followed the edited text, and the edited record cannot say whether the research changed with it.

Problem

• minority health research
• research classification
• funding policy
• DEI language
• uncounted science

Innovation

• word targeting
• text classification
• bibliometrics
• uncounted science
• policy evaluation