NURD 2.0: Prediction of tibial nonunion after intramedullary nail fixation at any time within 3 months after injury.

Introduction Nonunion after fixation of long bones negatively impacts outcomes and requires additional surgery. The ability to predict likelihood of nonunion after tibial shaft fracture would be helpful to clinicians and patients. The goal of this work was to combine three previous models of tibial shaft nonunion at different time points into one overall model that incorporates time as a continuous variable. Methods We conducted a retrospective review at a Level I academic trauma center. The study cohort consisted of patients with tibial shaft fractures treated with nail insertion from 2007 through 2014, excluding patients who did not have contact between bone ends, those who had planned bone grafting for acute bone defects, and those who lacked adequate follow-up. Three previous models were combined: 382 patients at time 0, 323 at 6 weeks, and 240 at 12 weeks. The primary outcome variable was surgery for nonunion. Bivariate and multivariate regression analyses determined which of 42 clinical and radiographic variables were significantly associated with nonunion. Predictive power was evaluated using area under the curve (AUC). Results The original nonunion risk determination (NURD) score was significantly improved through addition of 6- and 12-week radiographic union scores for tibial fractures, infection and complications, smoking status, and need for flaps. Overall, over the course of 12 weeks, the NURD-based model produced an AUC of 0.87 at initial time of fixation that improved to >0.9 at 6 and 12 weeks. Data were used to bin patients into five clinically important risk strata (p < 0.001). Patients in the lowest risk strata had 0% probability of nonunion (0 of 97 patients); in the second lowest risk strata, 4% (three of 73 patients); and in the highest risk strata, 48% (38 of 80 patients). Conclusions We created a NURD 2.0 score that predicts nonunion at various time points during the first 3 months after fracture. The new model is a notable improvement over previous models. A computerized version allows surgeons and patients to use the score when making treatment decisions regarding need for nonunion surgery.