Predictive model for knee traumatic arthritis secondary to tibial plateau fracture surgery

Objective Establish a predictive model for investigating the risk factors of Knee traumatic arthritis (KTA) secondary to tibial plateau fracture surgery. Methods 258 consecutive patients with tibial plateau fracture between December 2009 and April 2015 were collected, 117 cases were male, 141 cases were female. All the patients were divided into the training cohort (2009-2012) and the validation cohort (2013-2015). In the training cohort, univariate and multivariate logistic regression was used to analyze the risk factors for traumatic arthritis; then Nomogram predicting model was built based on logistic regression model. In the validation cohort, goodness of fit test and ROC curve were used for testing calibration and discrimination power respectively. Results All the patients were divided into 151 cases for the training cohort and 107 cases for the validation cohort. In the training cohort (151 patients), 11 patients were diagnosed as KTA with the incidence of 7.3%. The results showed that body mass index, type of fracture, osteoporosis and poor reduction were risk factors of postoperative KTA. In the validation cohort (107 patients), 9 patients were confirmed as KTA, with the incidence of 8.4%. Goodness of fit test showed Pearson χ2 was 321.64, which indicated good calibration. The area under the receiver operating characteristic curve (ROC) was 0.7551, which indicated moderate discrimination. Conclusion High body mass index, unstable fractures, osteoporosis and non-anatomic reduction are independent risk factors for KTA secondary to open reduction internal fixation of tibial plateau fractures. The predictive model has good calibration and moderate discrimination power in predicting KTA secondary to tibial plateau fractures. Key words: Tibial fractures; Osteoarthritis, knee; Forecasting; Calibration