This paper explores the development of predictive models for geriatric health outcomes using factor scores derived from Bayesian and Maximum Likelihood Estimation (MLE) approaches. Factor scores, representing latent dimensions such as mobility and social participation, are used as predictors to model outcomes like walking difficulty and health impairments. The study compares the performance of various models using AIC, BIC, and R-squared metrics, highlighting the strengths of Bayesian methods, particularly those employing non-conjugate priors like Cauchy-Log Normal. Clustering analysis further validates the differentiation between models, emphasizing the superior predictive accuracy and robustness of Bayesian approaches. These results underscore the utility of advanced statistical techniques in geriatric health studies and provide insights into their practical applications for early risk assessment.