217 Meta-analysis female sex and mortality in SAB 10 text. We evaluated statistical heterogeneity with the Cochran Q and I2 statistics. To explore potential sources of heterogeneity, we performed meta-analyses on subsets of studies to determine if variation in factors such as mortality time point (eg, 30day vs 90-day mortality), bacterial groups (eg, MSSA only, MRSA only, both MSSA and MRSA), or geographic location between studies could be contributing. Statistical analyses were performed with RStudio version 2022.02.0 (R Project for Statistical Computing). Publication bias was assessed using funnel plots with the Egger test [18] when ten or more studies were included in the analysis. We used the Evidencebased Practice Center (EPC) model from the US Agency for Healthcare Research and Quality (AHRQ) to grade overall strength of evidence [19]. A full description of the EPC approach is detailed in eAppendix 3 in Supplement 1. Results We screened the title and abstract of 5339 studies, and 4778 were deemed irrelevant (Figure 1). A full-text assessment was performed on 561 studies, and 472 of these were excluded. We included 89 studies in the analysis, with a total of 132 582 patients (50 258 female [37.9%], 82 324 male [62.1%]) (Table) [3-8,14,20-101]. All data on mortality by sex were from observational studies: 88 of 89 cohort studies and one post hoc analysis of a randomized clinical trial. Mortality was most frequently assessed at 28 to 30 days (54 of 89 studies [61%]). The majority of studies were conducted in Europe (36 [40%]), Asia (24 [28%]) and North America (20 [22%]). The majority of studies were published after 2010 (68 [76%]). Thirty-two studies (36%) were rated as having low risk of bias, and 57 studies (64%) as having high risk of bias (detailed quality assessment of each study in eTable 1 in Supplement 1). Mortality by sex Unadjusted mortality data was available from 81 studies (109 828 patients) and revealed an increased mortality risk in female compared with male patients (pooled OR, 1.12; 95% CI, 1.06-1.18) (Figure 2). Moderate heterogeneity was observed in this analysis (Q = 130.17; P < .001; I2 = 37%). An influence analysis revealed that exclusion of any single study did not significantly alter the findings from the overall cohort (eAppendix 4 in Supplement 1). A sensitivity analysis with only studies that had an OR that was either reported or could be directly calculated (ie, excluding 14 studies in which RR or HR were reported) similarly did not change the overall findings (eFigure 1 in Supplement 1). Exclusion of single-center studies did not change the overall findings. No funnel plot asymmetry was found (eFigure 2 in Supplement 1).
RkJQdWJsaXNoZXIy MTk4NDMw