Abstract
BACKGROUND: Orthopedists and other clinicians assess recovery potential of hip fracture patients at 2 months post-fracture for care planning. It is unclear if examining physical performance (e.g., balance, gait speed, chair stand) during this follow-up visit can identify individuals at a risk of poor functional recovery, especially mobility, beyond available information from medical records and self-report. METHODS: Data came from 162 patients with hip fracture enrolled in the Baltimore Hip Studies-7th cohort. Predictors of mobility status (ability to walk 1 block at 12 months post-fracture) were the Short Physical Performance Battery (SPPB) comprising balance, walking and chair rise tasks at 2 months; baseline medical chart information (sex, age, American Society of Anesthesiologist physical status rating, type of fracture and surgery, and comorbidities); and self-reported information about the physical function (ability to walk 10 feet and 1 block at pre-fracture and at 2 months post-fracture). Prediction models of 12-month mobility status were built using two methods: (1) logistic regression with least absolute shrinkage and selection operator (LASSO) regularization, and (2) classification and regression trees (CART). Area under ROC curves (AUROC) assessed discrimination. RESULTS: The participants had a median age of 82 years, and 49.3% (n = 80) were men. Two-month SPPB and gait speed were selected as predictors of 12-month mobility by both methods. Compared with an analytic model with medical chart and self-reported information, the model that additionally included physical performance measures had significantly better discrimination for 12-month mobility (AUROC 0.82 vs. 0.88, p = 0.004). CONCLUSION: Assessing SPPB and gait speed at 2 months after a hip fracture in addition to information from medical records and self-report significantly improves prediction of 12-month mobility. This finding has important implications in providing tailored clinical care to patients at a greater risk of being functionally dependent who would not otherwise be identified using regularly measured clinical markers.