Program Schedule

Development of an Automated Surveillance System for Identifying Surgical Site Infections and Triaging Clinical Review

Session: Poster Abstract Session: HAI Surveillance and Public Reporting
Friday, October 10, 2014
Room: The Pennsylvania Convention Center: IDExpo Hall BC

Background: Surgical site infections (SSIs) are a common and expensive healthcare-associated infection, costing an estimated $10 billion annually. SSI is also used as a benchmark for healthcare quality. As such, SSI detection and prevention are major targets of infection prevention programs. We sought to improve upon conventional manual detection methods by developing a simple, automated algorithm for SSI detection using administrative data.

Methods: We conducted a case control study among all surgeries performed at the Boston VA HCS during the two-year period from 1/2008- 12/2009. Cases of surgical site infection were matched to controls without surgical site infection. Clinical variables (administrative, microbiologic, pharmacy, radiology) were extracted and compared between the two groups to determine which variables (univariately) or combination of variables best detected SSI. Variables significantly associated with correct detection of SSI were then evaluated in a logistic regression model as independent predictors of accurate SSI detection. Points were assigned to variables based on the magnitude of the odds ratios from logistic regression. 

Results: 70 cases of SSI were matched to 70 non-infected controls. Variables found on multivariable analysis to be significantly associated with SSI identification were ordering of a clinical culture, antibiotics prescriptions within the 30 day post-operative window, ordering of CT or MRI, and use of relevant ICD-9 code for post-operative infection. Among patients who fell into the “very low probability” category (Figure), 98% were correctly identified as having no SSI. Among patients in the “high probability” category, 97.1% were correctly identified as having SSI. The area under the curve for this model using entirely administrative data was 0.87.

Conclusion: We derived an automated surveillance algorithm for SSI detection with excellent operating characteristics. This algorithm could be used to streamline infection control efforts and reduce time required for manual review.



Westyn Branch-Elliman, MD, MMSc, Division of Infectious Diseases, Denver VA Medical Center, Denver, CO; Medicine, University of Colorado, Aurora, CO; Infectious Diseases, VA Boston HCS, West Roxbury, MA, Judith Strymish, MD, Harvard Medical School, Boston, MA; Infectious Disease, VA Boston Healthcare System, West Roxbury, MA, Kamal Itani, MD, Department of Surgery, VA Boston and Boston University School of Medicine, West Roxbury, MA and Kalpana Gupta, MD, MPH, Department of Medicine/Boston University School of Medicine, Boston, MA; VA Boston Health Care System, West Roxbury, MA


W. Branch-Elliman, None

J. Strymish, None

K. Itani, Sanofi: Investigator, Research grant
Merck: Investigator, Research grant

K. Gupta, None

Findings in the abstracts are embargoed until 12:01 a.m. EDT, Oct. 8th with the exception of research findings presented at the IDWeek press conferences.

Sponsoring Societies:

© 2014, All Rights Reserved.

Follow IDWeek