Abstract
BACKGROUND: Our objective was to develop and validate a predictive model for non-union following a subtrochanteric fracture of the femur. METHODS: Following institutional board approval, 316 consecutive patients presenting to our institution (84 non-unions) who fulfilled the inclusion criteria were retrospectively identified. To identify potential unadjusted associations with progression to non-union, simple logistic regression models were used, followed by a revised adjusted model of multiple logistic regression. RESULTS: Having established the risk factors for non-union, the coefficients were used to produce a risk score for predicting non-union. To identify the high-risk patients in the early post-operative period, self-dynamisation was excluded. The revised scoring system was the sum of the following: diabetes (6); deep wound infection (35); simple or severe comminution (13); presence of an atypical fracture (14); lateral cortex gap size ≥5 mm (11), varus malreduction (5-10 degrees) (9); varus malreduction (>10 degrees) (20). On the ROC (receiver operating characteristic) curve, the area under the curve (0.790) demonstrated very good discriminatory capability of the scoring system, with good calibration (Hosmer-Lemeshow test; p = 0.291). Moreover, 5-fold cross validation confirmed good fit of the model and internal validity (accuracy 0.806; Kappa 0.416). The cut-point determined by Youden's formula was calculated as 18. CONCLUSION: This study demonstrates that the risk of non-union can be reliably estimated in patients presenting with a subtrochanteric fracture, from the immediate post-operative period. The resulting non-union risk score can be used not only to identify the high-risk patients early, offering them appropriate consultation and in some cases surgical intervention, but also informs surgeons of the modifiable surgery related factors that contribute to this risk.