To evaluate if clinical recovery following midshaft clavicle fracture is associated with nonunion and determine if this has superior predictive value compared to estimation at time of injury.

A prospective study of all patients (≥16 years) who sustained a displaced midshaft clavicle fracture was performed. We assessed patient demographics, injury factors, functional scores and radiographic predictors with a standardized protocol at six-weeks. Conditional-stepwise regression was used to assess which factors independently predicted nonunion at six-months post-injury determined by CT. The nonunion predictor six-week model (NUP6) was compared against a previously validated model based on factors available at time of injury (NUP0-smoking, comminution and fracture displacement).

200 patients completed follow-up at six months. The nonunion rate was 14% (27/200). Of the functional scores, the QuickDASH had the highest accuracy on receiver-operator-characteristic (ROC) curve analysis with a 39.8 threshold, above which was associated with nonunion (Area Under Curve (AUC) 76.8%, p<0.001).

On regression modelling QuickDASH ≥40 (p=0.001), no callus on radiograph (p=0.004) and fracture movement on examination (p=0.001) were significant predictors of nonunion. If none were present the predicted nonunion risk was 3%, found in 40% of the cohort (n=80/200). Conversely if two or more were present, found in 23.5% of the cohort, the predicted nonunion risk was 60%.

The NUP6 model appeared to have superior accuracy when compared to the NUP0 model on ROC curve analysis (AUC 87.3% vs 64.8% respectively).

Delayed assessment at six-weeks following displaced clavicle fracture enables a more accurate prediction of fracture healing.