An improved predictive model for postoperative pulmonary complications after free flap reconstructions in the head and neck.

Abstract

Commonly used predictive models for postoperative pulmonary complications (PPCs) do not perform when applied to head and neck cases. A head and neck-specific risk prediction tool is needed. Data on 794 free flap head and neck surgery cases at a single center were abstracted from the electronic medical record. Each case was reviewed for the development of PPCs. A predictive model was developed and was then compared to existing predictive models for PPCs. The least absolute shrinkage and selection operator procedure identified age, alcohol use, history of congestive heart failure, preoperative packed cell volume, preoperative oxygen saturation, and preoperative metabolic equivalents as predictors of PPCs in the head and neck population. The model demonstrated an area under the receiving operating characteristic curve of 0.75 (0.69-0.80) with moderately good calibration. Comparisons to the performance of existing models demonstrate superior performance. The model for the development of PPCs developed in this article displays superior performance to existing models.