Clinical prediction model for postoperative ambulatory ability outcomes in patients with trochanteric fractures.

Tomita, Yosuke; Yamamoto, Norio; Inoue, Tomoo; Noda, Tomoyuki; Kawasaki, Keisuke; Ozaki, Toshifumi

INTRODUCTION Regaining independent ambulatory ability is one of the primary goals of treatment in patients with trochanteric fractures. This study aimed to develop and evaluate the discriminative accuracy of a clinical prediction model for ambulatory ability outcomes 3 months after surgery for trochanteric fractures. METHODS This retrospective cohort study included 346 patients treated with intramedullary nailing for trochanteric fractures who had independent ambulatory ability before their injury. Multiple regression models with preoperative and postoperative factors were used to predict ambulatory ability outcomes at 3 months. A clinical prediction model (CPM) was created based on a decision tree developed using a chi-square automatic interaction detector technique. RESULTS Three months after surgery, 263 (76.0%) and 83 (24.0%) patients regained and lost independent ambulatory ability, respectively. Univariate analysis showed that the Barthel index (BI) total score at 2 weeks predicted the ambulatory ability outcome at 3 months with good discriminative accuracy (area under the receiver operating characteristic curve [AUROC]: 0.819; 95% confidence interval [CI]: [0.769, 0.868], cut-off value: 22.5; sensitivity: 69.5%; specificity: 82.3%). Multiple logistic regression analysis showed that preoperative factors (residence before injury, diagnosis of dementia, and serum albumin at admission) and postoperative factors (BI total score at 2 weeks) predicted ambulatory ability outcomes at 3 months (AUROC: 0.710; 95%CI: [0.636, 0.783]; sensitivity: 91.3%; specificity: 41.8%). The CPM with the BI total score at 2 weeks (≤10; 10<, ≤50; >50 points) and dementia status (present; absent) had a moderate discriminative accuracy (AUROC: 0.676; 95%CI: [0.600, 0.752]; sensitivity: 94.7%; specificity: 40.5%). CONCLUSIONS We developed a CPM with moderate accuracy to predict ambulatory ability outcomes in patients 3 months after surgery for trochanteric fractures. Our results demonstrate the importance of the BI score measured soon after surgery and dementia status for the prediction of postoperative ambulation.