Key oral hypofunction components for geriatric syndromes: Machine learning approach.

PURPOSE: Oral hypofunction (OHF) is defined as a multidimensional impairment of oral function that may contribute to the development of geriatric syndromes. This cross-sectional study aimed to assess the relative importance of seven OHF components in relation to physical frailty, sarcopenia, and nutritional impairment, using machine learning methods that are robust to multicollinearity.

METHODS

Data were pooled from two observational studies involving community-dwelling adults aged ≥65 years. Light Gradient Boosting Machine (LightGBM) algorithm was used to calculate the SHapley Additive exPlanation (SHAP) values, reflecting the relative contribution of each OHF component in the identification of three outcomes: physical frailty, sarcopenia, and nutritional impairment. Incremental contributions were further evaluated using integrated discrimination improvement (IDI) and C-statistics, weighted by the SHAP magnitudes.

RESULTS

A total of 1,133 participants (mean age 74.4 years) were included. Low tongue pressure had the highest composite SHAP value among the three outcomes (5.20), followed by low masticatory performance, occlusal force, and articulatory oral motor skills. The IDI analysis confirmed that these four components provided incremental value in identifying frailty and sarcopenia. A simplified "OHF-4" model consisting of these four components alone demonstrated comparable discriminatory performance to the full seven-component model across all outcomes based on the C-statistics comparisons.

CONCLUSIONS

Four OHF components-low tongue pressure, masticatory performance, occlusal force, and articulatory oral motor skills -were the most informative indicators of geriatric syndromes. Targeting these specific functions may aid the development of effective oral function management strategies to mitigate the risk of geriatric syndromes.