Predicting surgical exploration in renal trauma: Assessment and modification of an established nomogram

BACKGROUND: This study aimed to externally validate a previously described nomogram that predicts the need for renal exploration in the trauma setting.

METHODS: The predicted probability of nephrectomy was manually calculated using prospectively collected data from consecutive patients with renal trauma who presented to our institution between May 2001 and January 2010. To assess nomogram performance, receiver operating characteristic curves against the observed exploration rate were generated, and areas under the curve were calculated. Calibration curves were generated to assess performance across the range of predicted probabilities. Logistic regression modeling was used to determine clinical factors predicting exploration in a contemporary setting, and a nomogram was derived and internally validated using bootstrapping.

RESULTS: The established nomogram was applied to the 320 patients who presented during the 9-year period. The global performance of the established nomogram was very high, with an area under the curve of 0.95. However, the model performance was poor for higher predicted probabilities, thus lacking predictive ability in the population where the model has the greatest potential utility. A clinical tool was generated to better predict trauma nephrectomy in our contemporary population, using platelet transfusion within the first 24 hours, blood urea nitrogen, hemoglobin, and heart rate on admission. The global accuracy for the new model was similar to the previous nomogram, but it was significantly better calibrated for patients with higher probabilities of nephrectomy, with good predictive accuracy even in patients with Grade 5 injuries.

CONCLUSION: Older nomogram fails to accurately predict renal exploration in high-grade injuries in the contemporary setting. A new nomogram that more accurately predicts the need for exploration is presented.