Severe sepsis in the ICU is often missing in hospital discharge codes

BACKGROUND: Different International Classification of Diseases (ICD)-based code abstraction strategies have been used when studying the epidemiology of severe sepsis. The aim of this study was to compare three previously used ICD code abstraction strategies to the American College of Chest Physicians/Society of Critical Care Medicine (ACCP/SCCM) consensus criteria for severe sepsis, in a setting of intensive care patients. METHODS: All patients (≥ 18 years of age) with severe sepsis according to the ACCP/SCCM criteria registered in the Swedish Intensive Care Registry (2005-2009) were included in the study. Using the Swedish National Patient Register, we investigated whether these patients fulfilled an ICD code compilation for severe sepsis at hospital discharge. RESULTS: Overall, 9271 patients with severe sepsis were registered in the Swedish Intensive Care Registry. A majority of these patients (55.4%) were discharged from the hospital with ICD codes that did not correspond to any of the ICD code compilations. A minority of patients (10.3%) were discharged with ICD codes corresponding to all three code abstraction strategies applied. Overall, the proportion of patients discharged with ICD codes corresponding to the criteria of Angus et al. was 15.1%, to the criteria of Flaatten was 39.8%, and to the criteria of Martin et al. was 16.0%. CONCLUSIONS: A majority of patients with severe sepsis according to the ACCP/SCCM criteria were not discharged with ICD codes corresponding to the ICD code abstraction strategies; thus, the abstraction strategies did not identify the correct patients.