A pragmatic three-component clinical score for cognitive risk stratification in older adults with multimorbidity and...

Simple, scalable clinical tools are needed to identify older adults with cognitive impairment in low-resource settings, yet whether parsimonious approaches can match complex phenotyping methods remains unclear. This study developed a three-component clinical score and compared its discriminative performance with latent class analysis (LCA)-derived multimorbidity phenotypes.

This cross-sectional study included 504 Bangladeshi community-dwelling adults aged ≥65 years with at least one chronic disease. Frailty was assessed using the Fried phenotype, and multimorbidity was self-reported and coded using International Classification of Diseases, 10th Revision. An additive score (0-5 points) incorporating age ≥80 years, ≥3 chronic conditions, and frailty classified participants into low (0-1), moderate (2-3), or high (4-5) risk. Outcomes included global cognition and cognitive impairment (Mini-Mental State Examination [MMSE] < 25).

The three-component score showed acceptable discrimination for cognitive impairment (area under the curve [AUC] = 0.72) and explained 33% of MMSE variance. LCA-derived phenotypes demonstrated poor discrimination (AUC = 0.44; difference = 0.28, p < .001). A monotonic gradient was observed across risk categories, impairment prevalence increased across risk categories (59%, 83%, and 96%), corresponding to a 12.8-point MMSE difference across the score range. A frailty-augmented LCA (in which frailty was added to the original disease-only LCA) combined with age yielded a modestly higher AUC (0.764), though at substantially greater analytical complexity.

A parsimonious clinical score combining age, multimorbidity, and frailty demonstrated acceptable cross-sectional discrimination for prevalent cognitive impairment and substantially outperformed disease-only multimorbidity phenotyping. Subject to prospective validation, this pragmatic tool may support case-finding and cognitive risk stratification in resource-limited settings.