Multidimensional prognostic index domain phenotypes identified using latent class analysis and mortality in the EURopean study of Older Subjects with Atrial Fibrillation (EUROSAF).

Background: The Multidimensional Prognostic Index (MPI), an established tool to predict adverse outcomes, classifies frailty using an aggregate-weighted tripartite scoring system based on eight domains (low-, moderate-, or severe-risk). However, this approach may fail to capture specific patient phenotypes that can be characterised by separate MPI domains and for whom health outcome risk also differs.

Objective: We sought to identify latent patient phenotypes based on MPI domain data and to determine their association with mortality in older patients with atrial fibrillation (AF).

Methods: Using data from the EURopean study of Older Subjects with Atrial Fibrillation (EUROSAF), we used latent class analysis (LCA) to identify phenotypes using individual MPI domains and Cox regression models to examine their association with 12-month mortality.

Results: Four MPI domain phenotypes were identified in N=2,019 AF patients (mean (SD) age=82.9 (7.5) years; 57% females): phenotype 1 (relatively fit, few comorbidities; n=672, 33%), phenotype 2 (functionally impaired, polypharmacy, comorbidities; n=685, 34%), phenotype 3 (multidimensional frailty, comorbidities; n=161, 8%), and phenotype 4 (relatively fit, polypharmacy, comorbidities; n=501, 25%). Compared to phenotype 1, 12-month mortality was higher in phenotype 3 (adjusted hazard ratio (aHR):4.68, 95%CI=3.41-6.43), phenotype 2 (aHR:1.98, 95%CI=1.53-2.57), and phenotype 4 (aHR:1.44, 95%CI=1.07-1.94).

Conclusion: In a cohort of older AF patients, LCA identified four MPI domain phenotypes with different risk of mortality. Pending confirmatory studies, the identified sub-groups might allow more targeted interventions to improve outcomes in this population.