Abstract P413: Latent Class Trajectory Modeling Identifies Distinct Clinical Phenotypes in a Cohort of Patients With Elevated Troponin Levels

Introduction: Clinicians widely use cardiac troponins I and T, sensitive and specific biomarkers of myocardial damage, as diagnostic criteria for myocardial infarction (MI). A common approach, dichotomizing based on whether troponin levels meet thresholds for MI, may miss meaningful heterogeneity in degree and pattern of troponin elevation, with wider clinical implications. Hypothesis: Using latent class trajectory modeling (LCTM) with patients’ patterns of troponin elevations will yield trajectories that represent clinically and prognostically distinct groups of patients. Methods: We used Northwestern Medicine’s Enterprise Data Warehouse to identify patients with at least 3 troponin measurements of which at least one was elevated during their inpatient stay. We built a LCTM to capture troponin trajectories from 13,432 patients with 65,162 measurements since 2000, using Bayesian Information Criterion to select the final model (# of classes varied from 3-7). We then described the clinical and prognostic factors associated with trajectory class membership. Results: LCTM identified 3 troponin trajectories: stable low (77.3%), moderate rise (9.5%), and severe rise-fall (13.2%). The average posterior probability of assignment was >0.95 for all classes. The low group had the most comorbidities vs moderate/severe groups, including cancer (21.9% vs 13.9%/11.3%), heart failure (12.6% vs 8.9%/7.5%), and COPD (15.0% vs 11.8%/10.1%). The moderate/severe rising trajectories had a greater rate of clinically-diagnosed MI than the stable group (34.5%/44.1% vs 5.6%), with relatively more ST-elevation MI in the severe group (19.8% vs 5.3%) and Non-ST-elevation MI in the moderate group (29.3% vs 24.3%). One-year mortality was greatest in the low group vs moderate/severe groups (25.5% vs 21.4%/18.8%). Conclusions: In a cohort of hospitalized patients with elevated troponins, LCTM identified 3 distinct, clinically interpretable troponin trajectories which were associated with differing clinical phenotypes.