A multivariable model to classify methicillin-resistant staphylococcus aureus infections as health care or community associated

BACKGROUND: Methicillin-resistant Staphylococcus aureus (MRSA) infections are often defined as health care (HA) or community-associated (CA) using common classification schemes involving health care risk factor, infection type, susceptibility pattern, or molecular typing. This investigation compared pulsed-field gel electrophoresis (PFGE) molecular typing results (dichotomized as HA or CA) with our new MRSA infection classification method. The goal was to develop an improved predictive model for PFGE-type based primarily on the other 3 classification variables. METHODS: Methicillin-resistant S. aureus infections reported to the Michigan Department of Community Health from October 2004 to December 2005 were analyzed. Patients' demographics, risk factors, infection information, and susceptibility results were collected for 2151 cases. A subset of 244 MRSA infections with available PFGE results was analyzed. Results of logistic regression are presented using a receiver operating characteristic curve analysis. RESULTS: The multivariable models predicted the PFGE classification as HA or CA (Max-rescaled R = 61%) better than health care risk factor, infection type, or susceptibility pattern alone (max-rescaled R = 21%, 34%, and 46%, respectively). The best model included infection type, susceptibility pattern, age, and hospitalized during infection. CONCLUSIONS: This model provides a simpler, more accurate prediction of HA or CA status, thus enhancing efforts to control MRSA infections. Copyright © 2011 Lippincott Williams &Wilkins.