Applicability of two surgical-site infection risk indices to risk of sepsis in surgical patients

OBJECTIVE: To compare the ability of the Study of the Efficacy of Nosocomia l Infection Control (SENIC) and the National Nosocomial Infection Surveilla nce (NNIS) indices to predict the development of nosocomial sepsis in subje cts undergoing surgery. DESIGN: 1-year prospective case-control study. SETTING: A tertiary-care center in Spain. PATIENTS: Cases were surgical patients with nosocomial sepsis defined using the criteria of the Consensus Conference on Sepsis, identified by daily pr ospective surveillance. METHODS: Controls were randomly selected from the daily list of surgical in patients. Data were prospectively collected. To determine whether either in dex added explanatory information to the other, two methods were used. The first method involved computing a set of residuals for both variables. Resi duals and primary variables were introduced in logistic regression models. The second method evaluated both indices with the Goodman-Kruskal (G) nonpa rametric coefficient. RESULTS: 99 cases and 97 controls were included. Afte r controlling for confounders, both the SENIC index (P<.001) and the NNIS i ndex (P=.04) showed a significant trend. Residuals of the SENIC index added discriminating ability to the NNIS index, whereas residuals of the NNIS in dex did not improve the predic tion ability of the SENIC index. Similar res ults were yielded by the G statistic: the SENIC index showed higher predict ive power than the NNIS index (G=0.56 vs G-0.41). CONCLUSIONS: Both indices performed about equally well for discriminating r isk of nosocomial sepsis. The SENIC index had a somewhat better ability tha n the NNIS index only when the number of discharge diagnoses (not truly a p redictive factor) were involved in the calculation of the SENIC index.