Detection and eradication of meticillin-resistant Staphylococcus aureus (MRSA) represents a public health priority worldwide. Our aim was to do a systematic review and meta-analysis of randomised, non-randomised, and observational studies to summarise the available evidence on the effect of MRSA detection by rapid screening tests on hospital-acquired MRSA infections and acquisition rate.

Eligible studies were retrieved from Medline, EmBase, Science Citation Index, and the Cochrane database. We judged as eligible those studies that compared hospitals and wards in which active screening for the detection of MRSA carriers was done at hospital admission by use of a rapid molecular test to those in which active screening was done with culture alone or not at all.

To account for statistical heterogeneity between studies, random-effects models were used. Ten studies (nine interventional studies and one unblinded, cluster-randomised, crossover trial) were reviewed. Meta-analysis was done for studies reporting data on the same outcome.

Primary outcomes included MRSA acquisition rate per 1000 patient-days (four studies); incidence of MRSA bloodstream infections per 1000 patient-days (three studies); and incidence of MRSA surgical-site infections per 100 surgical procedures (five studies). Compared with culture screening, use of rapid screening tests was not associated with a significant decrease in MRSA acquisition rate (risk ratio 0·87, 95% CI 0·61—1·24).

Between wards applying rapid screening tests and those not applying screening, we noted a significantly decreased risk for MRSA bloodstream infections (0·54, 95% CI 0·41—0·71), but not for MRSA surgical-site infections (0·69, 95% CI 0·46—1·01). We conclude that active screening for MRSA is more important than the type of test used. Since important and costly decisions, such as mandatory legislation for MRSA universal screening, are under consideration in many countries worldwide, policy makers should be aware of the limits and the heterogeneity of the available evidence.