Improving prediction of adverse surgical outcomes with machine learning algorithm

Editor's Note

In this study, a new machine learning Complexity Score algorithm outperformed three commonly used risk scores in predicting postoperative morbidity, 30-day readmission, 90-day readmission, and postoperative surper-use.

Study patients underwent colectomy, abdominal aortic aneurysm repair, coronary artery bypass grafting, total hip or knee replacement, or lung resection.

The Complexity Score had a good to very good predictive ability for all adverse outcomes and had the highest accuracy in predicting perioperative morbidity−performing better than the Charlson Comorbidity Index and Elixhauser Comorbidity Index, and performing similar to the Centers for Medicare & Medicaid Service’s Hierarchical Condition Category.

Similarly, the Complexity Score outperformed each of the three comorbidity indices in predicting 90-day readmission, 30-day readmission, and postoperative super-use.