Official websites use .gov

A .gov website belongs to an official government organization in the United States.

Secure .gov websites use HTTPS

A lock ( 🔒 ) or https:// means you've safely connected to the .gov website. Share sensitive information only on official, secure websites.

Need Help

AI-driven Multimodal Model for PrEP Candidacy Assessment in Women (AiM-PrEP)

Award Number: R01MH143269
ORGANIZATION: NATIONAL INSTITUTE OF MENTAL HEALTH
OPDIV: NIH
AWARD CLASS: DISCRETIONARY
AWARD ACTIVITY TYPE: SCIENTIFIC/HEALTH RESEARCH (INCLUDES SURVEYS)
PERIOD OF PERFORMANCE START DATE: 09/01/2026
PERIOD OF PERFORMANCE END DATE: 06/30/2031

Group Awards By:

View Award Description

AI-driven Multimodal Model for PrEP Candidacy Assessment in Women (AiM-PrEP) - The South accounts for 52% of new HIV infections annually, however HIV pre-exposure prophylaxis (PrEP) use in the South is lower than any other region in the country. Women account for a disproportionate amount of new HIV infections in the South compared to other regions, yet substantially fewer women use PrEP compared to men. Defining HIV risk among women is complex, as evidenced by the failure of HIV risk prediction models to accurately identify women at elevated risk of HIV developed by our team and others. Most electronic-health record (EHR) derived HIV risk prediction models calculate aggregate risk based on factors encoded by structured data, which does not include important, hidden narratives about risk among women. We hypothesize that using multimodal data (data from unstructured clinical notes and structured EHRderived variables) and artificial intelligence (Al) methods will enhance our ability to accurately quantify the risk of HIV acquisition in women, a key step to identifying PrEP candidates in this population. We also hypothesize that the predictive ability of our HIV risk quantification models in women will be further enhanced by incorporating neighborhood level social drivers of health (SDOH) data from publicly available data sources into the EHR-based model. This model will directly inform the clinically facing PrEP Candidacy Assessment Tool (PCAT), the featured intervention of our study. Our approach will be enhanced by knowledge graphs informed by women who live in the communities of interest adding context to the relationship between model features and HIV risk, as well as enhancing model explainability to end users and patients. In Aim 1, a community informed knowledge graph providing relational context to determinants of HIV risk will be developed through an iterative process by women living with HIV, HIV-focused community advocates, and HIV providers. In Aim 2, we will develop multi-stage algorithms that blend classical NLP and knowledge graph informed large language models to identify (a) relevant unstructured data from clinical notes and (b) important longitudinal and spatially informed SDOH data for inclusion in HIV risk prediction models. The outcome of this aim will be six Al-derived model variations on our baseline model, whose performance will be compared for predictive value for HIV acquisition in women. In Aim 3, we will conduct a pilot stepped-wedge, clustered randomized study of 6 Duke Health primary care clinics to compare the ease of implementation and preliminary effectiveness of two PCAT versions informed by different iterations of the HIV risk prediction mode (structured v. SDOH/unstructured data enhanced) on increasing provider-initiated PrEP discussions among women seeking care in participating clinics. Following the study, we will conduct interviews with participating clinicians to determine perceived intervention feasibility, acceptability and trust in PCAT outputs among providers. The proposed research will inform a Hybrid Type II study to determine the effectiveness of a multi-site implementation of the PCAT in increasing discussions about PrEP, and ultimately PrEP uptake among women in the South.


Issue Date FY	Funding FY	Legal Entity Name	Legal Entity Address	Legal Entity City	Legal Entity State	Legal Entity Zip Code	Legal Entity COUNTY	Legal Entity COUNTRY	Assistance Listing	Award Code	Budget Year	Action Date	Action Type	Action Amount

Issue Date FY: 2026 ( Subtotal = $1,221,740 )
2026	2026	DUKE UNIVERSITY	2200 W MAIN ST	DURHAM	NC	27705	DURHAM	USA	Mental Health Research Grants	000	1	8/19/2026	NEW	$1,221,740
														Subtotal = $1,221,740

Grand Total All Awards = $1,221,740

Top

All Categories

About

Search

Reports

Data Submission

Award Information

AI-driven Multimodal Model for PrEP Candidacy Assessment in Women (AiM-PrEP)

Award Number: R01MH143269

ORGANIZATION: NATIONAL INSTITUTE OF MENTAL HEALTH

OPDIV: NIH

AWARD CLASS: DISCRETIONARY

AWARD ACTIVITY TYPE: SCIENTIFIC/HEALTH RESEARCH (INCLUDES SURVEYS)

PERIOD OF PERFORMANCE START DATE: 09/01/2026

PERIOD OF PERFORMANCE END DATE: 06/30/2031

Federal Websites

Department of Health & Human Services

HHS Operating Divisions

HHS Staff Divisions

Download A Document Viewer