Measuring Surgical Site Infections in Children: Comparing Clinical, Electronic, and Administrative Data

Background Surgical site infections (SSIs) are an important end point and measure of quality of care. Surgical site infections can be identified using clinical registries, electronic surveillance, and administrative claims data. This study compared measurements of SSIs using these 3 different methods and estimated their implication for health care costs. Study Design Data were obtained from 5,476 surgical patients treated at a single academic children's hospital (January 1, 2010 through August 31, 2014). Surgical site infections within 30 days were identified using a clinical registry in the NSQIP Pediatric, an electronic surveillance method (Nosocomial Infection Marker; MedMined), and billing claims. Infection rates, diagnostic characteristics, and attributable costs were estimated for each of the 3 measures of SSI. Results Surgical site infections were observed in 2.24% of patients per NSQIP Pediatric definitions, 0.99% of patients per the Nosocomial Infection Marker, and 2.34% per billing claims definitions. Using NSQIP Pediatric as the clinical reference, Nosocomial Infection Marker had a sensitivity of 31.7% and positive predictive value of 72.2%, and billing claims had a sensitivity of 48.0% and positive predictive value of 46.1% for detection of an SSI. Nosocomial Infection Marker and billing claims overestimated the costs of SSIs by 108% and 41%, respectively. Conclusions There is poor correlation among SSIs measured using electronic surveillance, administrative claims, and clinically derived measures of SSI in the pediatric surgical population. Although these measures might be more convenient, clinically derived data, such as NSQIP Pediatric, may provide a more appropriate quality metric to estimate the postoperative burden of SSIs in children.