ARTIFICIAL INTELLIGENCE FOR PREDICTING FALLS IN GERIATRIC PATIENTS WITH HYPERTENSION
Sergeev V.V.1
1. Academy of Postgraduate Education of the Federal State Budgetary Institution Federal Scientific and Clinical Center of Specialized Types of Medical Care of the Federal Medical and Biological Agency, Moscow
Falls in elderly patients with arterial hypertension (HTN) remain one of the most common and costly problems in geriatrics. Traditional clinical prognostic scales have limited sensitivity due to the nonlinear interaction of multiple factors (orthostatic reactions, polypharmacy, sarcopenia, cognitive impairment). Modern artificial intelligence (AI) and machine learning (ML) methods enable the integration of large datasets from electronic medical records and wearable devices. A review based on studies from 2024–2026 demonstrated that ensemble models (CatBoost, XGBoost, Random Forest) in individual studies (primarily with limited samples) achieved AUC values of up to 0.97. Using SHAP analysis (Shapley Additive Explanations), key predictors were identified: previous falls, comorbidity, sleep disturbances, and a new phenotype—orthostatic hypertension (OHYPER)—associated with frailty. Explainable AI and multi-horizon prediction technologies make it possible to distinguish between acute and chronic risk mechanisms, paving the way for personalized adjustments to antihypertensive therapy and fall prevention.