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

PREDICT: Deep Learning Models for forecasting in Rheumatoid Arthritis

Award Number: K01AR085707
ORGANIZATION: NATIONAL INSTITUTE OF ARTHRITIS & MUSCULOSKELETAL & SKIN DISEASES
OPDIV: NIH
AWARD CLASS: DISCRETIONARY
AWARD ACTIVITY TYPE: TRAINING/TRAINEESHIP
PERIOD OF PERFORMANCE START DATE: 09/21/2026
PERIOD OF PERFORMANCE END DATE: 07/31/2031

Group Awards By:

View Award Description

PREDICT: Deep Learning Models for forecasting in Rheumatoid Arthritis - PROJECT SUMMARY This is a resubmission application for a K01 award for Dr. Augusto Garcia-Agundez, a researcher at the University of California, San Francisco (UCSF) focused on leveraging Large Language Models (LLMs) to improve outcomes in rheumatoid arthritis (RA). The training plan includes explainable artificial intelligence (XAI), qualitative research, and human centered design to enhance Dr. Garcia-Agundez's expertise. The motivation for the project is that the Treat-to-Target strategy for RA management suffers from a lack of implementation, and LLMs have the potential to improve it by providing more accurate predictions of disease activity based on clinical characteristics, facilitating personalized treatment to improve outcomes. The goal is to develop and validate an LLM, PREDICT, using electronic health record data to forecast 1-year disease trajectories and improve treatment plans. In Aim 1a, Dr. Garcia-Agundez will utilize extensive data at UCSF and the Rheumatology Informatics System for Effectiveness (RISE) registry, reflecting broad patient populations, to fine-tune PREDICT from the Llama-3.1-70B or similar LLM, using techniques like quantized low-rank adaptation and FlashAttention. Performance will be compared against GPT-4 and a precursor model. In Aim 1b, Dr. Garcia-Agundez will integrate XAI methods to ensure transparency and prediction interpretability in all three models. In Aim 2, qualitative analysis will be performed with 5 focus groups of 5-8 rheumatologists from UCSF and RISE, and semi-structured interviews with up to 20 RA patients from UCSF and RISE. These will assess perceptions of PREDICT’s recommendations and explanations using inductive and deductive techniques to identify themes and sub-themes, providing insights into user acceptability and understanding of PREDICT’s outputs. Feedback will be used to refine the model. Aim 3 will prototype the integration of PREDICT into an existing Epic-based sidecar app for RA at UCSF using principles of human centered design. This will involve conducting co-design workshops with 5-8 rheumatologists and patients each at UCSF and RISE to refine the interface and assess preferences and understanding of PREDICT’s interface. The visual prototype of PREDICT will be iteratively developed with a UCSF design team. Findings from this study will elucidate the applicability of LLMs in RA management, clinician and patient perceptions of their output, and optimized design for clinical use. Dr. Garcia-Agundez has assembled an exceptional mentorship team with expertise in rheumatology, clinical informatics, implementation science, clinical AI, and LLMs, in addition to access to a rich research environment and career development support through programs like the UCSF Clinical and Translational Science Institute K-scholars program. Formal coursework and mentoring will be supplemented with attendance to conferences related to rheumatology, AI, and clinical informatics. Completing the proposed research and career development plan will allow Dr. Garcia- Agundez to gain experience in state-of-the-art computational methods using large datasets to better understand and improve patient outcomes in RA, ultimately leading to independence as a multidisciplinary researcher.


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 = $132,408 )
2026	2026	REGENTS OF THE UNIVERSITY OF CALIFORNIA, SAN FRANCISCO, THE	1855 FOLSOM ST STE 425	SAN FRANCISCO	CA	94103	SAN FRANCISCO	USA	Arthritis, Musculoskeletal and Skin Diseases Research	000	1	9/17/2026	NEW	$132,408
														Subtotal = $132,408

Grand Total All Awards = $132,408

Top

All Categories

About

Search

Reports

Data Submission

Award Information

PREDICT: Deep Learning Models for forecasting in Rheumatoid Arthritis

Award Number: K01AR085707

ORGANIZATION: NATIONAL INSTITUTE OF ARTHRITIS & MUSCULOSKELETAL & SKIN DISEASES

OPDIV: NIH

AWARD CLASS: DISCRETIONARY

AWARD ACTIVITY TYPE: TRAINING/TRAINEESHIP

PERIOD OF PERFORMANCE START DATE: 09/21/2026

PERIOD OF PERFORMANCE END DATE: 07/31/2031

Federal Websites

Department of Health & Human Services

HHS Operating Divisions

HHS Staff Divisions

Download A Document Viewer