Mentoring Patient-Oriented Research Leveraging Bioinformatics to Study CV Risk in Rheumatic Disease - PROJECT SUMMARY/ABSTRACT Rheumatoid arthritis (RA) is the most common autoimmune inflammatory arthritis affecting 1% of the population and confers 1.5-2-fold excess risk of atherosclerotic cardiovascular disease (ASCVD) compared to the general population. Identifying the true at-risk population for ASCVD in RA for primary prevention remains a challenge despite efforts to tailor existing population based CV risk estimators. While a subset of RA patients at high ASCVD risk can be identified based on traditional CV risk factors, e.g., hypertension, hyperlipidemia, from clinical studies, we have observed evidence of substantial ASCVD in patients categorized with lower risk. Prior studies have observed a higher prevalence of high sensitivity cardiac troponin (hs-cTn) among RA patients compared to the general population. A marker of myocardial injury clinically used in the diagnosis of myocardial infarction (MI), hs-cTn is a promising biomarker for CV risk stratification in RA. Aim 1 will recruit RA patients, age>40 with ≥1 CV risk factor for hs-cTn testing and concurrently screen for coronary artery calcium (CAC). We hypothesize that including a step measuring hs-cTn in a subgroup of RA patients could improve identification of patients who would benefit from primary prevention with statin therapy, defined as having detectable CAC, compared to identifying patients based on ASCVD risk estimators alone. In Aim 2, the study will revisit the established association between RA and increased risk of CVD to understand potential biases in artificial intelligence (AI) algorithms developed using electronic health record (EHR) data to identify patient populations. Over the past decade, EHR data have become an alternate source of clinical data generally with a more diverse population than cohort studies assembled through recruitment, enabling studies of associations across populations. However, similar to other types of observational data, great care is needed to understand and correct potential biases. Thus, in Aim 2, we seek to rigorously evaluate for bias in existing algorithms for RA and MI by developing algorithms specifically for the self-reported Black population. We will then compare the performance of the algorithms trained with the Black population with the general EHR population for accurately identifying RA and MI in the Black population. Notably, these algorithms have also been applied to support recruiting efforts for RA patient-oriented research (POR) studies. We will use recently published guiding principles from the Agency for Healthcare Research and Quality, and the NIH. This proposal will support the candidate, a mid-career physician scientist, in creating unique training opportunities at the intersection of rheumatology and cardiology, and clinical research with bioinformatics. The candidate is located at a vibrant academic hospital and medical school campus with numerous training opportunities. Together with the studies outlined above, the proposal will serve as a foundation to mentor the next generation of clinical investigators studying rheumatology and musculoskeletal conditions, and concurrently provide training in state-of-the-art bioinformatics methods highly relevant to POR and clinical epidemiologic studies.
