Construction of a classification model for dementia among Brazilian adults aged 50 and over

Background: Dementia is a multifactorial and debilitating condition marked by cognitive decline and behavioral changes that compromise independence and daily activities. This condition is a growing challenge in Brazil, and early identification of associated factors can guide preventive strategies and health policies. Objectives: To build a dementia classification model for middle-aged and older adults Brazilians combining variable selection and multivariable analysis, using low-cost variables, including variables potentially modifiable and non-modifiable sociodemographic variables. Methods: Observational study employed a cross-sectional design and a classification modeling approach to estimate probable dementia and analyze the odds of dementia, using data from the Brazilian Longitudinal Study of Aging, involving 9,412 participants. Dementia was determined based on neuropsychological assessment and informant-based cognitive function. Analyses were performed with Random Forest (RF) and multivariable Logistic Regression (LR). Results: The prevalence of dementia was 9.6%. The highest odds of dementia were observed in illiterate individuals (Odds Ratio (OR) = 7.42; 95% Confidence Interval (CI): 4.04-13.62), individuals aged 90 years or older (OR = 11.00; 95% CI: 5.05-23.95), low weight (OR = 2.11; 95% CI: 1.12-3.97), low handgrip strength (OR = 2.50; 95% CI: 1.09-5.76), self-reported black skin color (OR = 1.47; 95% CI: 1.07-2.00), physical inactivity (OR = 1.61; 95% CI: 1.25-2.08), self-reported hearing loss (OR = 1.65; 95% CI: 1.16-2.37), and presence of depressive symptoms (OR = 1.72; 95% CI: 1.36-2.16). In contrast, higher education (OR = 0.44; 95% CI: 0.21-0.94), greater life satisfaction (OR = 0.72; 95% CI: 0.52-0.99), and being employed (OR = 0.78; 95% CI: 0.61-1.00) were protective factors. The RF model outperformed LR, achieving an area under the ROC curve of 0.776 (95% CI: 0.740-0.811), with sensitivity of 0.708, specificity of 0.702, precision of 0.201, Precision-Recall Area Under the Curve (PR-AUC) of 0.261 (95% CI: 0.217-0.319), F1-score of 0.311, G-means of 0.705, and accuracy of 0.703. Conclusion: The findings reinforce the multidimensional nature of dementia and the importance of accessible factors for supporting screening/triage and prioritization in primary care. Strengthening public policies focused on promoting brain health can contribute significantly to the efficient allocation of resources in primary care and dementia prevention in Brazil.