Normative data for the Montreal Cognitive Assessment 8.1 in community-dwelling Spanish population

Abstract

The clinical utility of the Montreal Cognitive Assessment (MoCA) is often limited by normative data derived from linear models, which do not adequately capture non-linear age-related decline and score distribution issues like ceiling effects. This study aimed to generate methodologically rigorous normative data for the Spanish MoCA version 8.1 and its Memory Index Score (MIS) using an advanced regression approach. A total of 2,034 community-dwelling adults (18–89 years; 56.3% women) were assessed. Bayesian beta generalized additive models (GAMs) were used to model the effects of age, sex, and education. Normative data were generated as percentiles from the model’s standardized residuals. To facilitate direct clinical application, a user-friendly automated calculator was also developed. Non-linear GAMs demonstrated superior fit and predictive accuracy versus linear models, multiple fractional polynomial (MFP) models, and generalized additive models for location, scale, and shape (GAMLSS). Gains over GAMLSS were small but consistent, and model diagnostics supported the stability of the final solutions. The final models explained a significant amount of variance (Bayesian R² ≈ 0.43 for the MoCA; ≈ 0.26 for the MIS). Age (with a marked nonlinear effect) and education were associated with both outcomes, whereas sex-related effects differed between the MoCA and MIS models. This study provides updated and rigorous normative data for the Spanish MoCA 8.1 and MIS. By accurately modelling demographic effects, the proposed norms and accompanying calculator may support more precise normative interpretation in Spanish adults; however, external clinimetric validation is required before advocating diagnostic use or broad generalization.