Fourteen observational cohort studies were included. Five studies (n=1,245) provided data on oral treatment and 12 studies (n=3,357) provided data on parenteral treatment.
Oral antibiotic treatment was associated with significantly reduced mortality (5 studies; combined RR 0.17, 95% confidence interval: 0.07, 0.44). Statistical heterogeneity was not significant.
For parenteral antibiotics, 8 studies showed a beneficial effect (one significant) and four a harmful effect of treatment. Significant heterogeneity was present among these 12 studies and among the 7 studies that only included patients seen in primary care before admission. In the 5 studies that included data stratified by severity of disease, the proportion of cases classified as severe ranged from 11 to 76%. Statistical heterogeneity was significant (P=0.02). None of the studies showed a statistically significant effect on mortality.
The meta-regression indicated that differences in the proportion of cases treated accounted for 100% of the variance between studies of parenteral antibiotics. There was a significant negative association (P=0.04) between the proportion of patients given parenteral antibiotics before admission and mortality after such treatment.