Differences in Impact of Definitional Elements on Mortality Precludes International Comparisons of Sepsis Epidemiology—A Cohort Study Illustrating the Need for Standardized Reporting*

Objectives: Sepsis generates significant global acute illness burden. The international variations in sepsis epidemiology (illness burden) have implications for region specific health policy. We hypothesised that there have been changes over time in the sepsis definitional elements (infection and organ dysfunction), and these may have impacted on hospital mortality. Design: Cohort study. Setting: We evaluated a high quality, nationally representative, clinical ICU database including data from 181 adult ICUs in England. Patients: Nine hundred sixty-seven thousand five hundred thirty-two consecutive adult ICU admissions from January 2000 to December 2012. Interventions: None. Measurements and Main Results: To address the proposed hypothesis, we evaluated a high quality, nationally representative, clinical, ICU database of 967,532 consecutive admissions to 181 adult ICUs in England, from January 2000 to December 2012, to identify sepsis cases in a robust and reproducible way. Multinomial logistic regression was used to report unadjusted trends in sepsis definitional elements and in mortality risk categories based on organ dysfunction combinations. We generated logistic regression models and assessed statistical interactions with acute hospital mortality as outcome and cohort characteristics, sepsis definitional elements, and mortality risk categories as covariates. Finally, we calculated postestimation statistics to illustrate the magnitude of clinically meaningful improvements in sepsis outcomes over the study period. Over the study period, there were 248,864 sepsis admissions (25.7%). Sepsis mortality varied by infection sources (19.1% for genitourinary to 43.0% for respiratory; p < 0.001), by number of organ dysfunctions (18.5% for 1 to 69.9% for 5; p < 0.001), and organ dysfunction combinations (18.5% for risk category 1 to 58.0% for risk category 4). The rate of improvement in adjusted hospital mortality was significant (odds ratio, 0.939 [0.934–0.945] per year; p < 0.001), but showed different secular trends in improvement between infection sources. Conclusions: Within a sepsis cohort, we illustrate case-mix heterogeneity using definitional elements (infection source and organ dysfunction). In the context of improving outcomes, we illustrate differential secular trends in impact of these variables on adjusted mortality and propose this as a valid reason for international variations in sepsis epidemiology. Our article highlights the need to determine standardized reporting elements for optimal comparisons of international sepsis epidemiology.