Latent Class Analysis for the Identification of Phenotypes Associated with Increased Risk in Atrial Fibrillation Patients: The COOL-AF Registry.

Abstract

Patients with atrial fibrillation (AF) often have clinical complexity phenotypes. Latent class analysis (LCA) is based on the concept of modeling of both observed and unobserved (latent) variables. We hypothesized that LCA can help in identification of AF patient groups with different risk profiles and identify patients who benefit most from the Atrial fibrillation Better Care (ABC) pathway.We studied non-valvular AF patients in the prospective multicenter COOL-AF registry. The outcomes were all-cause death, ischemic stroke/systemic embolism (SSE), major bleeding, and heart failure. Components of CHA₂DS₂-VASc score, HAS-BLED score, and ABC pathway were recorded.A total of 3,405 patients were studied. We identified 3 LCA groups from 42 variables: LCA class 1 (n = 1,238), LCA class 2 (n = 1,790), and LCA class 3 (n = 377). Overall, the incidence rates of composite outcomes, death, SSE, major bleeding, and heart failure were 8.69, 4.21, 1.51, 2.27, and 2.84 per 100 person-years, respectively. When compared to LCA class 1, hazard ratios (HR) of composite outcome of LCA classes 3 and 2 were 3.86 (3.06-4.86) and 2.31 (1.91-2.79), respectively. ABC pathway compliance was associated with better outcomes in LCA classes 2 and 3 with the HR of 0.63 (0.51-0.76) and 0.57 (0.39-0.84), but not in LCA class 1.LCA can identify patients who are at risk of developing adverse clinical outcomes. The implementation of holistic management based on the ABC pathway was associated with a reduction in the composite outcomes as well as the individual outcomes.