The epidemiology of severe sepsis in Sweden : Methodological aspects on the use of ICD coding in national registries

Sammanfattning: Severe sepsis is characterized by acute organ dysfunction caused by an infection. Despite advanced treatment with antibiotics and organ support the mortality remains high. The epidemiological research on severe sepsis has expanded over the years but has led to conflicting results. In this thesis, we present evidence that some of these conflicting results may be explained by inappropriate methods used in epidemiological studies on severe sepsis. The definition of severe sepsis relies on consensus criteria including a number of clinical signs, parameters and laboratory findings. To facilitate large nation-wide studies on the incidence of severe sepsis in already existing administrative datasets, several investigators have attempted to mirror the clinical criteria by using combinations of the International Classification of Diseases (ICD) codes for infection and organ dysfunction. In paper I of this thesis, however, we found that three different ICD code abstraction strategies applied on the same dataset (the Swedish National Patient Register) generated three almost separate cohorts of patients. Furthermore, in paper II, where we followed intensive care unit (ICU)-treated patients with severe sepsis according to clinical consensus criteria to discharge, we observed that most patients did not meet any ICD code abstraction strategies for severe sepsis when reviewing their ICD codes registered at discharge from the hospital. In conclusion, the ICD code abstraction strategies previously used in the epidemiological research on sepsis seem to be inaccurate in the Swedish setting.  Sepsis may also result in poor long-term outcomes, and contribute to an increased risk of late mortality. However, the actual causes of late mortality in sepsis remain unclear. In paper III, we investigated causes of death over 1 year after treatment of sepsis in the ICU. In this study, a matched control group consisting of ICU-treated patients without sepsis was included. The most common causes of late mortality in both the sepsis group and the control group were heart diseases and cancer. The sepsis group, however, had a significantly increased incidence of infectious-related deaths compared with the control group, even over 1 year after the initial ICU event.