Impact of comorbidity patterns on mortality and length of stay in hospitalized patients with atrial fibrillation: a cohort study.

Abstract

Latent Class Analysis (LCA) is an unsupervised clustering and analytical approach to identify subgroups of people with similar characteristics within a heterogenous population. We examined patterns of comorbidities in people with atrial fibrillation (AF) admitted to hospital using LCA and their relationship with 12-month mortality and length of hospitalisation. We conducted a retrospective cohort study using hospital data from Flinders Medical Centre, a major tertiary public hospital in Southern Adelaide (South Australia), covering a period of 10 years (2009-2018). We explored the patterns of comorbidities using LCA and used Cox regression and logistic regression models to examine their association with 12-month mortality and length of hospitalisation. Three phenotypes were identified using LCA in 12 555 AF patients: phenotype 1 (lower comorbidity burden; n = 7689, 61%), phenotype 2 (higher comorbidity burden; n = 4120; 33%), and phenotype 3 (cerebrovascular, hypertensive disease, nervous system and non-specific abnormalities; n = 746, 6%). The hazard of death was significantly higher in phenotype 2 (adjusted hazard ratio (aHR) = 2.25, 95% CI = 2.01-2.50) and phenotype 3 (aHR = 1.69, 95%CI = 1.38-2.08) compared to phenotype 1. The odds of being hospitalized for ≥10 days (vs. <10 days) were significantly higher in phenotype 2 (adjusted odds ratio [aOR] = 8.53, 95%CI = 7.70-9.44) and phenotype 3 (aOR = 4.23, 95%CI = 3.56-5.04) compared to phenotype 1. In this large cohort study in AF patients, LCA identified three comorbidity phenotypes with distinct associations with 12-month all-cause mortality and length of hospitalisation. Our findings suggest that phenotyping is valuable in identifying high-risk group of patients that may benefit from targeted intervention. Key messages What is already known on this topic? Previous studies have examined the impact of individual comorbidities in people with atrial fibrillation (AF), but there is limited data on how different combinations of comorbidity patterns occur in people with AF and their impact on health outcomes. What this study adds? Latent class analysis identified three comorbidity phenotypes with distinct associations with 12-month all-cause mortality and length of hospitalisation in people with AF. How this study might affect research, practice, or policy? Risk-stratified care management may help improve health outcomes of AF patients.