Screening for multidrug-resistant bacteria as a predictive test for subsequent onset of nosocomial infection

Abstract

Objective: To investigate whether carriage of multidrug-resistant bacteria is a risk factor for nosocomial infection and whether detection of carriage is predictive of subsequent onset of nosocomial infection.

Methods: In this observational cohort (study period, June 1998 through October 2002), nasal and rectal swab specimens from 412 consecutive patients admitted to the intensive care unit were tested for carriage of multidrug-resistant bacteria. Concomitantly, the bacteria responsible for any subsequent nosocomial infection, the date of infection, and some of the known clinical risk factors for nosocomial infection were noted. These factors were adjusted for potential confounders, using a Cox model stratified on the propensity score of multidrug-resistant bacteria carriage. The diagnostic characteristics of a carriage test, including the positive and negative diagnostic likelihood ratios, were calculated for all strata of the propensity score.

Results: Forty-two patients were carrying multidrug-resistant bacteria. Nosocomial infection occurred in 95 patients, of whom 16 (38%) were carriers, and 79 (83%) were noncarriers (P=.01). After adjustment for potential confounders, statistical analysis revealed that carriage remained a risk factor for nosocomial infection (relative risk, 2.08 [95% confidence interval {CI}, 1.13-3.81]). Receipt of antibiotic treatment at the time of intensive care unit admission was found to be protective against nosocomial infection. A positive result of test for detection of carriage seemed to be an efficient predictor of subsequent nosocomial infection (positive diagnostic likelihood ratio, 2.05 [95% CI, 1.15-3.66]), although a negative test result was not a predictor of subsequent nosocomial infection (negative likelihood ratio, 0.91 [95% CI, 0.73-1.11]).

Conclusion: Carriage proved to be a risk factor for subsequent nosocomial infection. However, the carriage test was useful as a predictive tool only for patients with a positive test result.