Analysis agrees with news report findings that suggest greater scrutiny of research pertaining to diversity, equity, and other topics of political interest
Words reflecting diversity language have appeared less frequently in grants awarded by the National Institutes of Health (NIH) since the 2024 US presidential election, with a 25% relative decrease between January 2024 and June 2025, finds a study in the Christmas issue of The BMJ.
News reports have indicated that US federal agencies have recently limited or discouraged use of words related to diversity, health inequities, and other scientific subjects commanding political attention in NIH research grants.
News reports also suggest that researchers may be modifying their language or research topics to align with federal priorities, but this issue has not been rigorously evaluated.
To explore this further, researchers measured changes in diversity language in newly awarded NIH grant abstracts during 2024 and 2025, using a list of words based on existing news reports from the New York Times and PEN America, ranging from abortion, climate change, and disability to gender, racism, and women.
Overall, 17,701 grants were analysed. The rate of words reflecting diversity language decreased sharply between October and November 2024, from 11.11 to 5.42 words per 1,000, a 51% relative decrease. The decrease persisted through 2025, with an overall relative decrease of 25% between January 2024 and June 2025.
In an additional analysis of 1,967 pairs of the same grants in 2024 and 2025, words reflecting diversity language were deleted from grant abstracts at a 10-fold higher rate than other (control) words in 2025.
"This result is consistent with anecdotal evidence that researchers have modified language to prevent grant abstracts from being flagged for governmental review, suggesting a limitation on researchers' ability to freely use specific terms in federal research grants," say the researchers.
These are observational findings, so no firm conclusions can be drawn about cause and effect, and the authors acknowledge that their analysis was limited to eight months, they used lists of words compiled by news outlets, and did not separate words reflecting diversity language into separate domains, which may obscure changes in the relative prevalence of certain subjects over others.
Nevertheless, they say: "Our analysis directionally agrees with findings of news reports, which broadly suggest greater scrutiny of research pertaining to diversity, equity, and other topics of political interest."
They add: "Future work should consider monitoring changes over a longer period and should consider using a semantic or word embedding strategy to capture different subtopics."
