Performance of the National Nosocomial Infections Surveillance System (NNIS) risk index in predicting surgical site infections in an Australian setting

Abstract

BACKGROUND/OBJECTIVES: The VICNISS system was established in 2002 in Victoria, Australia, and receives surgical site infection (SSI) surveillance data from 26 hospitals of more than 100 beds. We evaluated the association between the NNIS risk index (RI) and SSI rates for six surgical procedures. METHODS: SSI surveillance is performed utilizing NNIS definitions and methods. All hospitals performing coronary artery bypass grafts (CABGs) submit data, otherwise hospitals choose surveillance activities from 20 surgical procedure groups. Data for this report were prospectively collected on surgical procedures from November 11, 2002, through September 30, 2004. Correlation was assessed using the Goodman-Kruskal Gamma score. RESULTS: Data were submitted on the following total numbers of procedures; appendectomy (APPY), 545; CABG, 4632; cholecystectomy (CHOL), 1001; colon surgery (COLO), 623; hip arthroplasty (HPRO), 3825; and knee arthroplasty (KPRO), 2416. An ASA score of 3 or more was recorded for the following proportion of patients: APPY, 5.5%; CABG, 97%; CHOL, 13%; COLO, 38.6%; HPRO, 46.4%; and KPRO, 35.5%. The proportion of procedures where the procedure duration exceeded the NNIS duration cut point was: APPY, 16%; CABG, 24%; CHOL, 6%; COLO, 25%; HPRO, 26%; and KPRO, 28%. The proportion of procedures with contaminated or dirty wounds was: APPY, 57%; CABG, 0.04%; CHOL, 8%; COLO, 27%; HPRO, 0.4%; KPRO. 0.1%. Ninety-eight percent (98%) of CABG patients were spread across RI categories 1 and 2. In contrast, joint prosthesis candidates were spread across three categories and patients undergoing APPY and COLO were well spread across all categories. NNIS RI and SSI rate were well correlated for COLO (G=0.48). There was a moderate positive correlation between the NNIS RI category and the SSI for APPY (gamma, G=0.33), CHOL (G=0.18), HPRO (G=0.2), and KPRO (G=0.16). However, for CABG there was a poor association between the NNIS RI and SSI (G=0.02). The correlation between ASA score and SSI rate was very similar to the correlation between NNIS RI and SSI rate. Wound class and SSI rate were well correlated for both APPY and COLO. CONCLUSIONS: The NNIS RI performed well at predicting SSIs for COLO. However, for procedures with a poor spread of patients through the RI categories, the correlation was less optimal. For CABG, the NNIS RI did not adequately stratify the risk of SSI. As the number of procedures in our database increases, we will be able to clarify further the performance of the NNIS RI.