Patterns of Frailty in Newly Diagnosed Older Patients With Nonvalvular Atrial Fibrillation Initiating Oral Anticoagulation

Abstract

Background

Frailty is a significant predictor of death in patients with atrial fibrillation (AF), with the frailty index (FI) acting as an effective severity classification tool. However, even in patients with a similar FI, the underlying clinical profiles can differ substantially. As the severity classification relies solely on the number of deficits without considering their interaction, distinct clinical subgroups with differing prognoses and care needs may remain unrecognized within the same frailty category. We aimed to identify novel phenotypes based on the deficit patterns in older AF patients.

Methods

Using data from a comprehensive claims database in Shizuoka (2012–2018), we extracted patients aged ≥ 65 years with AF and frailty who initiated oral anticoagulants. Latent class analysis (LCA) was conducted for each frailty status using 34 variables incorporated in the electronic FI (eFI), which is determined through a coding-based algorithm. We performed multivariable Cox proportional hazards to evaluate the associations between the latent classes and all-cause death within each frailty status.

Results

Among 11 533 patients (mean age: 79.3 ± 8.03 years; women: N = 5359 [46.5%]) categorized as mildly (eFI > 0.12–0.24; N = 3967), moderately (> 0.24–0.36; N = 4385) and severely frail (> 0.36–0.60; N = 3181), LCA identified three to four classes within each category: mildly frail, Class 1: high prevalence of hypotension (N = 326), Class 2: high prevalence of heart failure (N = 1404), Class 3: high prevalence of polypharmacy (N = 2237); moderately frail, Class 1: high prevalence of hypotension (N = 966), Class 2: high prevalence of heart failure (N = 1521), Class 3: high prevalence of polypharmacy (N = 1598), Class 4: high prevalence of mobility problems (N = 300); and severely frail, Class 1: high prevalence of hypotension (N = 1378), Class 2: high prevalence of heart failure (N = 1198), Class 3: high prevalence of mobility problems (N = 605). After multivariable adjustment, the other classes exhibited lower mortality risks than in the class characterized by high prevalence of mobility problems in the moderately (HR [95% CI]; Class 1: 0.59 [0.45–0.79], p < 0.001; Class 2: 0.71 [0.55–0.93], p = 0.013; Class 3: 0.68 [0.52–0.88], p = 0.003) and severely frail (Class 1: 0.89 [0.74–1.07], p = 0.22; Class 2: 0.77 [0.63–0.94], p = 0.010), whereas there was no difference among the classes in the mildly frail (Class 1 vs. Class 3: 0.97 [0.67–1.40], p = 0.86; Class 2 vs. Class 3, 0.94 [0.76–1.16], p = 0.57).

Conclusions

The LCA, focused on the deficit patterns incorporated in the eFI, identified phenotypes, each representing distinct clinical outcomes. The classification expands the utility of eFI in clinical practice.