From Manual to Machine Learning for Post-operative Infection Surveillance

Reviewed by Clare Marlin, MS, RN, CIC, CRRN, CCRC

Post-operative surgical infections may impact a patient’s health outcomes and their quality of life, as well as increase the cost of the patient’s care. The American College of Surgeons National Surgical Quality Improvement Program (ACS-NSQIP) is a leader in surveillance of post-operative outcomes through a manual chart review process cited by the authors. As surgical care quality improvement initiatives have continued to innovate, approaches in which the evaluation of a wider scope of electronic health record (EHR) data and the production of increasingly timely results through the use of artificial intelligence (AI) have been evaluated.

In their novel research, the authors describe their engagement of surgical care quality improvement through Automated Surveillance of Postoperative Infection (ASPIN), which examines EHR information via statistical models developed with artificial intelligence. Specifically, supervised machine learning to label inputs and predict outputs to best determine the pre-operative likelihood a patient will develop a post-operative infection.

Within a large healthcare system, the authors evaluated the risk-adjusted outcomes of ASPIN’s application with the manual chart review process of the EHR for all surgeries meeting study criteria within the six-year period of 2013-2019.

ASPIN provides risk-adjusted predicted outcomes for post-operative infections similar to the manual chart review process via ACS-NSQIP. The authors endorse that ASPIN, or a similar model, may serve as a partner, or replacement, for the manual chart review process and provide for more timely data generation. The authors acknowledge model drift may occur and therefore the model applied to the EHR data will need ongoing validation and review of the model’s inputs and resultant outputs.

Reference:

Colborn, K. L., Fei, Y., Henderson, W. G., Zhuang, Y., Dyas, A. R., Matheny, M. E., Stuart, C. M., & Meguid, R. A. (2026). Estimation of risk-adjusted postoperative infection outcomes using interpretable machine learning and Electronic Health Record Data. American Journal of Infection Control, 54(2), 139–144. https://doi.org/10.1016/j.ajic.2025.09.015