Background: Genetic algorithm (GA)–based scoring has been proposed as a data-driven approach for health risk stratification . However, performance estimates may be inflated when preprocessing, optimization, and evaluation are not strictly separated within a prespecified validation framework.