Predicting Delayed In-Hospital Recovery of Physical Function After Total Knee Arthroplasty.

Patients scheduled for inpatient rehabilitation or nursing home care were excluded, creating a homogeneous sample of patients who were likely to return home. This limits the generalizability of the findings to patients with more severe disability or social support issues. The model failed to perform well despite a large sample size, likely because the included predictors (physical fitness measures) did not capture the full complexity of recovery, which includes social and medical factors not measured here. Missing data for 15 patients regarding the primary outcome required multiple imputation, which can introduce uncertainty, although sensitivity analysis showed no large differences. The study was retrospective, relying on data collected during usual care, which may have inconsistencies in assessment timing or method compared to a controlled trial setting.

Do not use the finding that walking aid use and mobility scores are associated with recovery time to assume this model can reliably predict individual patient outcomes. The model's poor discrimination and high rate of misclassification mean it is not ready for clinical use.