Objective: To develop and validate a partially automated algorithm to identify surgical site infections (SSIs) using commonly available electronic data to reduce manual chart review.
Design: Retrospective cohort study of patients undergoing specific surgical procedures over a 4-year period from 2007 through 2010 (algorithm development cohort) or over a 3-month period from January 2011 through March 2011 (algorithm validation cohort).
Setting: A single academic safety-net hospital in a major metropolitan area.
Patients: Patients undergoing at least 1 included surgical procedure during the study period.
Methods: Procedures were identified in the National Healthcare Safety Network; SSIs were identified by manual chart review. Commonly available electronic data, including microbiologic, laboratory, and administrative data, were identified via a clinical data warehouse. Algorithms using combinations of these electronic variables were constructed and assessed for their ability to identify SSIs and reduce chart review.
Results: The most efficient algorithm identified in the development cohort combined microbiologic data with postoperative procedure and diagnosis codes. This algorithm resulted in 100% sensitivity and 85% specificity. Time savings from the algorithm was almost 600 person-hours of chart review. The algorithm demonstrated similar sensitivity on application to the validation cohort.
Conclusions: A partially automated algorithm to identify potential SSIs was highly sensitive and dramatically reduced the amount of manual chart review required of infection control personnel during SSI surveillance.