Cognitive decline is a major concern in aging populations, necessitating models that are both predictive and interpretable. While Ordinary Least Squares (OLS) regression has long been employed to quantify cognitive risk factors such as age, education, depression, and chronic disease burden, its outputs often lack accessible explanation for end-users. Meanwhile, Large Language Models (LLMs) such as DeepSeek have demonstrated strong capabilities in generating natural language explanations but often operate without grounding in statistical inference. In this study, we propose a hybrid framework that integrates OLS regression with LLM-based narrative generation to enhance both statistical rigor and personalized interpretability. Using Wave 4 data from the Health and Retirement Study (HRS), we first identify key predictors of cognitive performance and then apply structured prompt engineering to generate LLM explanations tailored to individual participants. Our results show that LLM outputs informed by regression coefficients provide more medically coherent and actionable interpretations than generic prompts. This hybrid methodology offers a scalable approach for precision aging, interpretable risk assessment, and human-centered AI in health research.