Introduction The categorical view dominates the traditional diagnostic approach to heart failure (HF) but ignores possible within-class heterogeneity such as individual differences in severity.

Objective To assess if HF should be considered as categorical or dimensional, and to validate a novel scale of severity for clinical HF.

Methods Cross-sectional evaluation of 1115 community participants aged ≥45 years in 2006–2008. We considered items related to troubled breathing and fatigue (4 items), volume overload (6 items) and objective evidence of cardiac structural /functional abnormalities (3 items). Bayesian Information criteria from latent class analysis (LCA) and latent trait analysis (LTA) were used to assess if HF could be considered as categorical or dimensional. BNP values and American College of Cardiology (ACC)/American Heart Association (AHA) stages of HF, classified by experienced clinicians with access to all data, were used to validate the scale.

Results Bayesian Information criteria suggested a 3-class solution for the LCA and a 2-factor solution for the LTA, with the best result being the last one. The first factor was associated with the items on troubled breathing/fatigue and cardiac abnormalities; the second factor was associated with the items about volume overload. The prevalence of BNP≥30 pg/ml, BNP≥100 pg/ml and stage C/D of clinical HF was significantly higher in the group of individuals with high scores for all factors than in that of individuals with low scores.

Conclusions The use of latent models applied to HF provided evidence for considering HF as dimensional rather than categorical as traditionally considered.