Purpose: Sepsis is recognized as a global public health problem, but the proportion due to hospital-acquired infections remains unclear. We aimed to summarize the epidemiological evidence related to the burden of hospitalacquired (HA) and ICU-acquired (ICU-A) sepsis. Methods: We searched MEDLINE, Embase and the Global Index Medicus from 01/2000 to 03/2018. We included studies conducted hospital-wide or in intensive care units (ICUs), including neonatal units (NICUs), with data on the incidence/prevalence of HA and ICU-A sepsis and the proportion of community and hospital/ICU origin. We did random-effects meta-analyses to obtain pooled estimates; inter-study heterogeneity and risk of bias were assessed. Results: Of the 13,239 studies identified, 51 met the inclusion criteria; 22 were from low-and middle-income countries. Twenty-eight studies were conducted in ICUs, 13 in NICUs, and ten hospital-wide. The proportion of HA sepsis among all hospital-treated sepsis cases was 23.6% (95% CI 17-31.8%, range 16-36.4%). In the ICU, 24.4% (95% CI 16.7-34.2%, range 10.3-42.5%) of cases of sepsis with organ dysfunction were acquired during ICU stay and 48.7% (95% CI 38.3-59.3%, range 18.7-69.4%) had a hospital origin. The pooled hospital incidence of HA sepsis with organ dysfunction per 1000 patients was 9.3 (95% CI 7.3-11.9, range 2-20.6)). In the ICU, the pooled incidence of HA sepsis with organ dysfunction per 1000 patients was 56.5 (95% CI 35-90.2, range 9.2-254.4) and it was particularly high in NICUs. Mortality of ICU patients with HA sepsis with organ dysfunction was 52.3% (95% CI 43.4-61.1%, range 30.1-64.6%). There was a significant inter-study heterogeneity. Risk of bias was low to moderate in ICU-based studies and moderate to high in hospital-wide and NICU studies.