A risk prediction model for postoperative recovery of closed calcaneal fracture: a retrospective study

Abstract

Objective To explore the risk factors for postoperative recovery of closed calcaneal fracture and develop a prediction model. Methods We retrospectively enrolled patients with closed calcaneal fracture from January 1, 2017 to December 31, 2020. Patients treated from 2017 to 2019 were selected as a training cohort and those in 2020 as a validation cohort. The outcome variable was the postoperative recovery evaluated by the Creighton-Nebraska calcaneal fracture scoring system. Multivariate logistic regression analysis was used to screen the risk factors of postoperative recovery. A risk prediction model was constructed in the training cohort and the corresponding nomogram was drawn. The model was validated internally using bootstrapping and externally by calculating the performance in the validation cohort. Results A total of 659 patients with closed calcaneal fracture met the inclusion and exclusion criteria, which were divided into the training cohort (n = 509) and the validation cohort (n = 150). 540 cases (81.9%) patients recovered well after calcaneal fracture surgery. According to multivariate logistic regression analysis, female (OR = 2.525, 95% CI 1.283–4.969), > 60 years (OR = 6.644, 95% CI 1.243–35.522), surgery within 8–14 days after fracture (OR = 2.172, 95% CI 1.259–3.745), postoperative infection (OR = 4.613, 95% CI 1.382–15.393), and weight-bearing time longer than 3 months after surgery (4–6 months, OR = 2.885, 95% CI 1.696–4.907; 7–12 months, OR = 3.030, 95% CI 1.212–7.578; > 12 months, OR = 15.589, 95% CI 3.244–74.912) were independent risk factors for postoperative recovery of calcaneal fractures. The C-indices were 0.750(95% CI 0.692–0.808) in the training cohort and 0.688(95% CI 0.571–0.804) in the external validation cohort, and the C-index of internal validation was 0.715. The Hosmer–Lemeshow test showed good fitting of the model (all P > 0.05), which was consistent with the results of the calibration plots. Decision Curve Analysis indicated that the clinical effectiveness was the best when the threshold probability was between 0.10 and 0.45. Conclusions Patients with female, > 60 years, surgery within 8–14 days after fracture, postoperative infection, and weight-bearing time longer than 3 months after surgery are more likely to have poor postoperative recovery. The risk prediction of fracture patients through this model might be translated into clinical guidance and application. Trial registration This study was registered on the Chinese Clinical Trial Registry (Registration number: ChiCTR-EPR-15005878).