Project Summary Providers across multiple specialties face challenges in determining the clinical significance of a positive antinuclear antibody (ANA). While a positive ANA is highly sensitive for autoimmune disease, it is non-specific with up to 20% of the general population having a positive ANA without having autoimmune disease. Risk models to aide clinicians in stratifying positive ANA patients do not currently exist. By identifying high-risk patients, providers could properly triage patients for prompt treatment to reduce autoimmune disease-related morbidity and mortality. Our long-term goal is to build risk models in the electronic health record (EHR) for autoimmune diseases that improve outcomes. The overall objective of this proposal is to identify positive ANA patients who are at high risk for developing autoimmune disease to facilitate appropriate triage to rheumatology for earlier diagnosis and treatment. Our institution with expertise in biostatistics, biomedical informatics, and implementation science has demonstrated success in building and testing robust EHR risk models. Building upon this well-established infrastructure, we hypothesize that we can use available EHR data to identify positive ANA patients that are high risk for autoimmune disease. We hypothesize that using tailored risk assessments in real- time in the EHR can reduce time to autoimmune disease diagnosis and treatment. We will test these hypotheses with the following specific aims: (1) Refine and validate features available in the EHR to distinguish positive ANA patients who develop autoimmune disease from positive ANA patients who do not develop autoimmune disease and (2) Conduct an adaptive, randomized, pragmatic evaluation of an autoimmune disease risk model in the EHR to risk- stratify patients with a positive ANA. For Aim 1, we will validate a risk model for autoimmune disease in positive ANA patients using EHR data with logistic regression and machine learning methods. For Aim 2, we will deploy a risk model for autoimmune disease in real-time in the EHR. We will randomize positive ANA patients to either have a risk score from the model displayed and acted upon vs. not having a risk score displayed or usual care. We will assess if having this EHR system change with a risk score calculated and shared with the ordering provider compared to usual care affects time to autoimmune disease diagnosis and treatment. Our proposal is innovative in that it not only builds a predictive risk model for autoimmune disease but also deploys and assesses if the model impacts patient outcomes. For expected outcomes, we anticipate deploying an EHR risk model that identifies positive ANA patients at high risk for developing autoimmune disease and decreases time to diagnosis and treatment.