Risk factors for methicillin-resistant Staphylococcus aureus colonisation or infection in intensive care units and their reliability for predicting MRSA on ICU admission.

Predicting methicillin-resistant Staphylococcus aureus (MRSA) in intensive care units (ICUs) avoids inappropriate antimicrobial empirical treatment and enhances infection control. We describe risk factors for colonisation/infection related to MRSA (MRSA-C/I) in critically ill patients once in the ICU and on ICU admission, and search for an easy-to-use predictive model for MRSA colonisation/infection on ICU admission. This multicentre cohort study included 69,894 patients admitted consecutively (stay>24h) in April-June in the five-year period 2006-2010 from 147 Spanish ICUs participating in the National Surveillance Study of Nosocomial Infections in ICUs (ENVIN-HELICS). Data from all patients included were used to identify risk factors for MRSA-C/I during ICU stays, from admission to discharge, using uni- and multivariable analysis (Poisson regression) to check that the sample to be used to develop the predictive models was representative of standard critical care population. To identify risk factors for MRSA-C/I on ICU admission and to develop prediction models, multivariable logistic regression analysis were then performed only on those admitted in 2010 (n=16950, 2/3 for analysis and 1/3 for subsequent validation). We found that, in the period 2006-2010, 1046 patients were MRSA-C/I. Independent risk factors for MRSA-C/I in ICU were: age>65, trauma or medical patient, high APACHE-II score, admitted from a long-term care facility, urinary catheter, previous antibiotic treatment and skin-soft tissue or post-surgical superficial skin infections. Colonisation with several different MDRs significantly increased the risk of MRSA-C/I. Risk factors on ICU admission were: male gender, trauma critical patient, urgent surgery, admitted from other ICUs, hospital ward or long-term facility, immunosuppression and skin-soft tissue infection. Although the best model to identify carriers of MRSA had a good discrimination (AUC-ROC, 0.77; 95% CI, 0.72-0.82), sensitivity was 67% and specificity 76.5%. Including more complex variables did not improve prediction capability. Our conclusion is that clinical-demographic risk factors for colonisation/infection related to MRSA should not be used to accurately identify patients who would benefit from empirical anti-MRSA treatment or from specific preventive measures. Independent risk factors for MRSA colonisation/infection during ICU stay and on ICU admission are described. The latter should be considered in future studies for MRSA prediction.