Objectives: To test whether differences in both general and sepsis-specific patient characteristics explain the observed differences in sepsis mortality between countries, using two national critical care (ICU) databases.
Design: Cohort study.
Setting: We analyzed 62 and 164 ICUs in Brazil and England, respectively.
Patients: Twenty-two-thousand four-hundred twenty-six adult ICU admissions from January 2013 to December 2013.
Interventions: None.
Measurements and main results: After harmonizing relevant variables, we merged the first ICU episode of adult medical admissions from Brazil (ORganizational CHaractEeriSTics in cRitical cAre study) and England (Intensive Care National Audit & Research Centre Case Mix Programme). Sepsis-3 definition was used, and the primary outcome was hospital mortality. We used multilevel logistic regression models to evaluate the impact of country (Brazil vs England) on mortality, after adjustment for general (age, sex, comorbidities, functional status, admission source, time to admission) and sepsis-specific (site of infection, organ dysfunction type and number) patient characteristics. Of medical ICU admissions, 13.2% (4,505/34,150) in Brazil and 30.7% (17,921/58,316) in England met the sepsis definition. The Brazil cohort was older, had greater prevalence of severe comorbidities and dependency compared with England. Respiratory was the most common infection site in both countries. The most common organ dysfunction was cardiovascular in Brazil (41.2%) and respiratory in England (85.8%). Crude hospital mortality was similar (Brazil 41.4% vs England 39.3%; odds ratio, 1.12 [0.98-1.30]). After adjusting for general patient characteristics, there was an important change in the point-estimate of the odds ratio (0.88 [0.75-1.02]). However, after adjusting for sepsis-specific patient characteristics, the direction of effect reversed again with Brazil having higher risk-adjusted mortality (odds ratio, 1.22 [1.05-1.43]).
Conclusions: Patients with sepsis admitted to ICUs in Brazil and England have important differences in general and sepsis-specific characteristics, from source of admission to organ dysfunctions. We show that comparing crude mortality from sepsis patients admitted to the ICU between countries, as currently performed, is not reliable and that the adjustment for both general and sepsis-specific patient characteristics is essential for valid international comparisons of mortality amongst sepsis patients admitted to critical care units.