Latent Class Analysis: A Data-Driven Approach to Uncovering Cognitive Decline Severity Patterns in Population Health Surveillance (Motivated by the BRFSS 2015–2020 Study on Subjective Cognitive Decline by Snead et al.)

Latent class analysis is increasingly recognized as a powerful approach for identifying unobserved subgroups within heterogeneous populations, particularly in the context of public health surveillance. This report applies latent class analysis to data from the Behavioral Risk Factor Surveillance System (2015–2020) to examine patterns of subjective cognitive decline in U.S. adults aged 45 and older. Using responses to a structured five-item cognitive function module, three distinct and interpretable subgroups—Mild, Moderate, and Severe cognitive decline—were identified. Each class exhibited unique symptom profiles and socio-demographic associations, with greater severity linked to lower income, lower education levels, and poorer general health. The optimal three-class model was supported by multiple fit indices, including AIC, BIC, entropy, and the Lo-Mendell-Rubin likelihood ratio test. Conditional response probability tables and profile plots further validated the internal structure of the classes. Additional visualizations illustrated the distribution of socioeconomic covariates across latent classes. This analysis underscores the value of latent class analysis for enhancing measurement precision, revealing hidden population structures, and informing targeted interventions. Future research should focus on integrating LCA with causal frameworks, machine learning techniques, and cross-cohort validation to improve its utility in real-world settings. The approach demonstrated here provides a replicable model for analyzing complex survey data in aging and cognitive health research.