Predicting Mobility Assistive Device Needs in Elders Using EHR and Healthcare Utilization Data
CT.Wu, CY.Chiang, JY. Wang, Sanford PC.Hsu, FT.Wang, YC.Chen. Gerontechnology 25(s)
Full text PDF 
( Download count: 1)
AbstractPURPOSE: The rapid growth of the aging population has markedly increased the demand for mobility assistive devices (MADs) [1], while conventional labor-intensive assessments have become increasingly unsustainable under workforce shortages [1]. This study aimed to develop a machine learning-based prediction model using electronic health records (EHR) and healthcare utilization data to support a clinical decision support system (CDSS) for the early identification of MAD needs in older adults and to facilitate proactive aging-in-place interventions. METHOD: A retrospective analysis was conducted using real-world data from Taipei Veterans General Hospital and the Veterans Affairs Council (2016–2024), including 5,986 adults aged ≥65 (Figure 1). SMOTEENN was applied to address class imbalance [2]. Seven machine-learning models were evaluated using 5-fold cross-validation, and SHAP was used for model interpretability [3]. RESULTS AND DISCUSSION: Gradient boosting models outperformed traditional classifiers, with LightGBM achieving the best performance (AUROC = 0.959, accuracy = 91.0%, specificity = 95.7%). SHAP analysis identified ICD-10 diagnoses as the most important predictors, followed by residential location and medical specialty (Figure 2). These findings demonstrate the real-world feasibility of integrating MAD prediction into a hospital CDSS to support timely, accessible, and resource-efficient aging-in-place care while reducing caregiver burden.Keywords: Mobility Assistive Devices; Electronic Health Records; Clinical Decision Support System
CT.Wu, CY.Chiang, JY. Wang, Sanford PC.Hsu, FT.Wang, YC.Chen. Gerontechnology 25(s) (2026). Predicting Mobility Assistive Device Needs in Elders Using EHR and Healthcare Utilization Data. Gerontechnology, 25(2), 1-10
https://doi.org/10.4017/gt.2026.25.2.1375.3