Introduction The clinical presentation of sepsis is heterogeneous and largely depends on the primary site of infection. Currently, factors associated with sepsis outcomes do not differentiate between infection sites. The objective of this investigation was to identify variables associated with risk of in-hospital mortality or intensive care unit (ICU) admission, according to infection sites.

Methods This was a secondary analysis of a multicentre prospective cohort of ED patients ≥18 years old from three university hospitals in Medellín, Colombia. Multivariable logistic regression models were performed to estimate the association of factors with in-hospital mortality or ICU admission according to five infection sites: urinary tract infection (UTI), community-acquired pneumonia (CAP), intra-abdominal infection, sepsis without evident source (primary) and other sites.

Results The infection sites of the 1947 patients included were: UTI (n=586), CAP (n=585), intra-abdominal infection (n=213), primary (n=224) and other sites (n=339). In the multivariable model, the factors associated with in-hospital mortality or ICU admission varied by infection site: respiratory rate (RR), systolic blood pressure (SBP) and lactate for UTI; heart rate (HR), RR and temperature <38°C for CAP; Glasgow Coma Scale (GCS), lactate and age <65 for intra-abdominal infection; SBP, GCS, lactate and temperature <38°C for primary and RR, GCS and temperature <38°C for other.

Conclusions Our results suggest that the diagnosis and prognosis of sepsis in emergency care should consider different clinical criteria, based on site of infection. Given the heterogeneity and interindividual variability of sepsis, a more individualised approach could help to direct treatment, monitor response and facilitate initial clinical decisions.