Brazilian researchers build dementia prediction model using Random Forest on 9,412 patients
The Random Forest model achieved 0.776 AUC, identifying key risk factors like illiteracy and age over 90.
Deep Dive
Researchers from Brazil built a dementia classification model using data from 9,412 participants in the ELSI-Brazil study. They used Python, Random Forest (RF), and multivariable logistic regression on low-cost variables. The RF model outperformed regression with a 0.776 AUC, 70.3% accuracy, and identified major risk factors (illiteracy OR=7.42, age 90+ OR=11.00) and protective factors (higher education OR=0.44). The tool aims to help identify vulnerable individuals for public health intervention.
Why It Matters
Provides a low-cost, data-driven tool for early dementia screening and resource allocation in public health systems.