A Risk-Prediction Platform for Acute Kidney Injury and 30-Day Readmission After Colorectal Surgery.

INTRODUCTION: Few known risk factors for certain surgical complications are prospectively analyzed to ascertain their influence on outcomes. Health systems can use integrated machine-learning-derived algorithms to provide information regarding patients' risk status in real time and pair this data with interventions to improve outcomes. The purpose of this work was to evaluate whether real-time knowledge of patients' calculated risk status paired with a stratified intervention was associated with a reduction in acute kidney injury and 30-d readmission following colorectal surgery. METHODS: Unblinded, retrospective study, evaluating the impact of an electronic health record-integrated and autonomous algorithm-based clinical decision support tool (KelaHealth, San Francisco, California) on acute kidney injury and 30-d readmission following colorectal surgery at a single academic medical center between January 1, 2020, and December 31, 2020, relative to a propensity-matched historical cohort (2014-2018) prior to algorithm integration (January 11, 2019). RESULTS: 3617 patients underwent colorectal surgery during the control period and 665 underwent surgery during the treatment period; 1437 historical control patients were matched to 479 risk-based patients for the study. Utilization of the risk-based management platform was associated with a 2.5% decrease in the rate of acute kidney injury (11.3% to 8.8%) and 3.1% decrease in rate of readmissions (12% to 8.9%). CONCLUSIONS: In this study, we found significant reductions in postoperative acute kidney injury (AKI) and unplanned readmissions after the implementation of an algorithm based clinical decision support tool that risk-stratified populations and offered stratified interventions. This opens up an opportunity for further investigation in translating similar risk platform approaches across surgical specialties.