Seventy studies were included (n=25,488 patients, range 52 to 3,413). The mean Newcastle-Ottawa score was 6.7.
Inappropriate empirical antibiotic treatment was associated with significantly higher mortality in the unadjusted and adjusted comparisons (adjusted OR 2.05, 95% CI 1.69 to 2.49, I2=79.7%; 48 studies). Significant benefit of appropriate empirical treatment was maintained in most subgroup and sensitivity analyses. In analyses adjusted for background conditions and sepsis severity, the pooled odds ratio was 1.6 (95% CI 1.37 to 1.86, I2=46.3%; 26 studies) and the number needed to treat to prevent one fatal outcome was 10 patients (95% CI 8 to 15).
Effect modifiers included study design, time of mortality assessment, reporting methods of the multivariable models and covariates used for adjustment. Septic shock was the only clinical variable found to influence results (was associated with higher ORs).