After the Freeman-Tukey double arcsine transformation,20 MRSA prevalence was pooled by either a fixed- or random-effects model, depending on the overall heterogeneity among studies (fixed if P > .1; random if P ≤ .1). The Wilson method was used to calculate the 95% CI of overall prevalence. 21 Heterogeneity among studies was estimated using Cochrane Q statistics (P < .1 considered significant) and quantified with the I2 statistic. 22 and 23 The subgroup analysis was performed according to continent, study design, and occupation. A sensitivity analysis was conducted to determine whether our assumptions or decisions with regard to studies selected for inclusion had a major effect by omitting each study (1 at a time). Publication bias was evaluated by visual inspection of Begg's funnel plot and tested by Begg's and Egger's tests (P < .1 considered significant). 24 and 25 All statistical analyses were conducted using STATA version 13.0 (StataCorp LP, College Station, Tex).