AcademyHealth has published a detailed analysis of the 150 research and training awards that the Agency for Healthcare Research and Quality (AHRQ) did not continue in fiscal year 2026. The paper documents what happened and examines whether it was done in the manner federal regulation requires. This post asks a different question: not whether the non-continuations were lawful, but what they suggest about the criteria used to discontinue research that had already been invested in.
The question is worth asking because the non-continuations were accompanied by a rationale. The letters state that AHRQ is realigning its portfolio toward a set of agency priorities. The priorities page the letters cite goes further, distinguishing research it favors from research it does not. It supports work focused on scientifically valid, measurable outcomes and solution-oriented approaches to disparities, and it disfavors work it characterizes as methodologically weak or as invoking poorly measured factors without defined variables. Two implicit standards are embedded there. One is a standard of scientific rigor. The other is a standard for when disparities research is legitimate. Both can be checked against the awards themselves, because we have the abstracts and public health relevance statements for the 150 awards that were not continued and for the 36 that received fiscal year 2026 funding.
When comparing these two groups of grants, it is clear that these standards fail. The way they fail is instructive in understanding how AHRQ views research
Consider rigor first. If the operative criterion were methodological strength, the awards that survived should be stronger by design than the awards that were cut. The reverse is closer to the truth. Among the 150 non-continued awards, roughly two in five describe a randomized or quasi-experimental design in their own text, including cluster-randomized trials, stepped-wedge designs, and studies exploiting randomized payment models. Because a randomized trial splits people into groups purely by chance, the groups end up similar in every way, including in variables that researchers did not think to measure. That means that if one group performs significantly better, we can be confident that the treatment caused it. Among the 36 awards that received funding, a smaller share do. The non-continued portfolio includes some of the strongest study designs the field produces. Whatever separated the two groups, a preference for rigor does not describe it, and a claim that these cuts served rigorous science cannot be reconciled with the cancelation of the most rigorous trials, such as a forty-hospital cluster-randomized trial.
The disparities standard is equally illuminating. There is a real distinction between disparities research that names a mechanism and tests an intervention and disparities research that documents a gap and attributes it to diffuse causes. AHRQ’s priorities page endorses the former and disfavors the latter. If that distinction were the operative screen, we would expect the non-continued awards to fall on the disfavored side--description without intervention, structural attribution without measurable variables.
They do not. Among the non-continued awards, the large majority that engage disparities at all do so in exactly the terms the priorities page says it favors. Nearly all of them pair the disparity with an intervention, an implementation strategy, or a trial. They are, by the agency’s own stated standard, the good kind. What distinguishes the non-continued awards from the funded ones is not the character of their disparities work but the mere presence of disparities language. It appears in a large share of the awards that were cut and a small share of the awards that were kept. The favored-versus-disfavored distinction the page draws does not track the outcomes. The presence-versus-absence of the vocabulary does.
This is where precision matters, and where it warrants being careful about what the evidence supports. These are patterns in the composition of two sets of awards. They are consistent with a screen that operated on the surface vocabulary of a project rather than on its methods or its design, sweeping in the solution-oriented, intervention-based disparities research the agency’s own standard protects. They are consistent with that, but they do not prove it. The agency’s deliberations are not publicly available, and the classification here rests on the language of the awards, which is a proxy for their content and not a substitute for reading each one. The finding is that the stated standards do not fit the observed pattern, not that any particular alternative standard was applied. AHRQ published a distinction between rigorous and weak science, and between solution-oriented and merely descriptive disparities work. It then did not continue a body of research that sits, on the evidence available, largely on the favored side of both lines. An agency is entitled to change its priorities. It is harder to defend the proposition that these particular awards failed the particular standards the agency articulated, because on the awards’ own terms, many of them meet those standards well.
The ambiguity in the letters’ language does real work here. The phrase about measurable outcomes and solution-oriented disparities research can be read as an inclusive statement of what AHRQ supports or as a set of conditions research must satisfy. Read inclusively, it describes most of the non-continued portfolio, which then cannot have been cut for failing it. Read as a filter, it is a filter the non-continued awards largely pass. The letters do not say which reading governs, and the pattern does not resolve the ambiguity so much as expose it. Under either reading, the stated criterion does not explain the outcome.
None of this establishes intent, and this post does not claim to. What it claims is narrower. When an agency offers reasons for a consequential action, those reasons can be checked against the action, and here they do not hold. The awards that were not continued are not, on the evidence, methodologically weak awards or ideologically framed awards the stated rationales guard against. That leaves an open question about what criterion actually operated, and it is a question worth pressing, in oversight and in the field’s own accounting of what was lost.
The companion white paper documents the “how” of these non-continuations and the regulatory requirements they engage. This is an accompanying question of what the pattern suggests, offered in that spirit--not as an accusation, but as an observation that the reasons given do not fit the record, and that the discrepancy deserves an answer.
This post was written by AcademyHealth’s Chief Programs & Science Officer, Elizabeth Cope.
Comments
Nothing yet. Say the first thing.
Sign in to join the conversation.