Time-updated motor function and fall in subacute orthopedic rehabilitation: Cox model with splines

Objective : To examine whether the association between time-updated motor function and fall risk is non-linear among older orthopedic inpatients undergoing subacute rehabilitation. Design : Retrospective cohort study. Setting and Participants : We included 1,111 orthopedic inpatients aged ≥65 years admitted to a Japanese subacute rehabilitation ward. Methods : Motor and cognitive status were assessed at admission and approximately every 2 weeks using the Functional Independence Measure (FIM) motor and cognitive scores. The dataset carried forward the most recent FIM values until the next assessment. The outcome was the first fall identified from the incident-reporting system. Cox proportional hazards models were adjusted for age, sex, orthopedic diagnosis, and time-updated FIM cognitive score. Non-linearity was assessed by comparing linear and restricted cubic splines models using the likelihood ratio test. As a supplementary exploratory analysis, model-based hazard ratios were visualized across representative motor-function anchor points and dichotomized cognitive status. Results : The non-linear model fit better than the linear model (p for non-linear=0.020), showing an inverted U-shaped association with peak risk at intermediate motor scores. The estimated hazard increased from lower motor scores to intermediate motor scores and then declined at higher motor scores. Conclusions and Implications : Time-updated FIM motor scores showed a non-linear association with first-fall risk. This finding should be regarded as hypothesis-generating and requires external validation before being used to guide targeted fall-prevention strategies.