Objective The pathophysiological heterogeneity of heart failure with preserved ejection fraction (HFpEF) makes the conventional ‘one-size-fits-all’ treatment approach difficult. We aimed to develop a stratification methodology to identify distinct subphenotypes of acute HFpEF using the latent class analysis.

Methods We established a prospective, multicentre registry of acute decompensated HFpEF. Primary candidates for latent class analysis were patient data on hospital admission (160 features). The patient subset was categorised based on enrolment period into a derivation cohort (2016–2018; n=623) and a validation cohort (2019–2020; n=472). After excluding features with significant missingness and high degree of correlation, 83 features were finally included in the analysis.

Results The analysis subclassified patients (derivation cohort) into 4 groups: group 1 (n=215, 34.5%), characterised by arrythmia triggering (especially atrial fibrillation) and a lower comorbidity burden; group 2 (n=77, 12.4%), with substantially elevated blood pressure and worse classical HFpEF echocardiographic features; group 3 (n=149, 23.9%), with the highest level of GGT and total bilirubin and frequent previous hospitalisation for HF and group 4 (n=182, 29.2%), with infection-triggered HF hospitalisation, high C reactive protein and worse nutritional status. The primary end point—a composite of all-cause death and HF readmission—significantly differed between the groups (log-rank p<0.001). These findings were consistent in the validation cohort.

Conclusions This study indicated the feasibility of clinical application of the latent class analysis in a highly heterogeneous cohort of patients with acute HFpEF. Patients can be divided into 4 phenotypes with distinct patient characteristics and clinical outcomes.

Trial registration number UMIN000021831.