Abstract
INTRODUCTION: This study aimed to develop and validate a frailty prediction model incorporating thyroid hormone levels in older adults with normal thyroid function. METHODS: A total of 767 participants aged ≥ 65 years (566 males, 73.8%) were enrolled between December 2019 and December 2023. Participants were randomly assigned to training (n = 546) and internal validation (n = 235) sets; an external validation cohort was recruited between December 2023 and December 2024. Frailty was assessed using the FRAIL scale. Predictors were selected through LASSO and logistic regression analyses, and a nomogram was constructed for clinical application. RESULTS: A total of 767 patients were included, with a median age of 81 years. Of these, 205 individuals (26.7%) were classified into the frailty group. The restricted cubic spline plot showed a negative correlation between the FT3/FT4 and frailty after adjusting for multiple confounding factors. Five predictors, age, polypharmacy, MNA.SF score, grip, and FT3/FT4 were incorporated to establish the model. The model performed well achieving AUC values of 0.92 and 0.80 in the development and external validation datasets, respectively. The Calibration curves and DCA results suggested that the model possesses satisfactory predictive performance. CONCLUSIONS: We developed a clinically applicable model to predict frailty in older adults with normal thyroid function, highlighting the relevance of thyroid hormone status in frailty risk assessment.