Research Domain Criteria in NIMH Grants Characterized Using Large Language Models

Area of Science:

• Mental Health Research
• Neuroscience
• Psychiatry

Background:

• The National Institute of Mental Health (NIMH) has increasingly promoted a transdiagnostic approach to psychiatric research, aligning with the Research Domain Criteria (RDoC) framework.
• The RDoC framework aims to map psychiatric conditions more closely to underlying neurobiology.

Purpose of the Study:

• To analyze trends in NIMH research funding for individual RDoC domains and transdiagnostic investigations over time.
• To assess the differential impact of this funding on publication rates, citation impact, and patent filings.

Main Methods:

• A longitudinal cohort study identified R01, R21, and R03 grants funded by NIMH from 2003 to 2023.
• A large language model was used to categorize grant abstracts according to RDoC domains (negative valence, positive valence, cognition, social, arousal, sensorimotor).
• Regression models analyzed the association between funding domains and outcomes (publications, 5-year citation impact, patents).

Main Results:

• Of 8897 funded projects, a significant portion focused on negative valence (35.3%), cognition (31.3%), and social (18.1%) domains.
• Transdiagnostic perspectives were present in 20.2% of the projects.
• Grants in the positive valence and social domains were associated with fewer publications and lower citation impact.
• Social, cognitive, and transdiagnostic research approaches showed a lower likelihood of resulting in patent filings.

Conclusions:

• NIMH-funded research across different RDoC domains exhibits substantial variation in scientific impact, measured by publications, citations, and patents.
• The study highlights the utility of large language models for large-scale research proposal analysis.
• Findings may inform NIMH resource allocation strategies to optimize scientific return on investment.