Background: Traditional surveillance for surgical site infections (SSI) after coronary artery bypass surgery (CABG) using National Health Safety Network (NHSN) methodology requires patient records to be reviewed at least twice, once at 30 days post-operation, and a year later. To minimize resources spent, electronic surveillance is appealing. While there is no single electronic data point sensitive enough to capture all SSIs, there may be a group of data points which identify post-surgical patients as likely infected. A software program that integrates patient information from data sources throughout a medical institution may facilitate surveillance and save resources if it can be configured to tag patients meeting these criteria as necessitating review.
Objective: Identify a group of characteristics which can be used to electronically screen patients for SSI and to compare this to gold standard 100% chart review.
Methods: A retrospective chart review was done using NHSN definitions for patients who had CABG surgery between 10/01/08 and 9/30/09 (n=422) in order to identify which patients had an SSI (n=22; rate=5.21%). Using an internet based software program, Theradoc (Salt Lake City, UT) we queried the software-tagged data points (alerts) for the 22 SSI patients looking for related characteristics which the software could recognize. These same alerts for the non-infected patients were then reviewed. Finally, a probabilistic analysis was conducted to identify which alerts would provide an electronic screening tool that would be 100% sensitive.
Results: Patients with a CABG SSI were discovered to most frequently have: been readmitted; a relevant culture; leukocytosis ; or a targeted drug. These same characteristics were less frequent for non-infected patients. Using a combination of alerts (relevant cultures and specific target drugs) gives a sensitivity and specificity of 100% and 42%. The positive predictive value improves if only relevant culture is used (64%), however sensitivity drops to 91%. Even with reviewing only charts of patients electronically tagged as having either a relevant culture, a targeted drug, or both, 100% of SSIs can be identified whilst reducing the number of charts reviewed by 77% (from 822 to 191) a year. This reduction would save approximately 52 hours and more than $1700 a year (based on chart review by an Infection Preventionist with an annual salary of $68,170 and a 5 minute chart review time for non-infected patients).
Conclusions: Electronic screening using a commercially available software program can reliably identify patients with SSI after CABG. Electronic screening saves time and resources and shows promise for the future of electronic surveillance in infection prevention. Our findings need to be validated to assure they are generalizable.