Accuracy of ICD-10 coding for sepsis and organ dysfunctions in children with blood culture-proven sepsis in Switzerland, results from the prospective Swiss Pediatric Sepsis Study

Abstract

Background: The recent resolution on sepsis by the World Health Or-ganization urges member states to better characterize the burden of sep-sis. Sepsis coding based on the International Statistical Classification of Diseases and Related Health Problems (ICD-10) is widely used, how-ever, there is a lack of paediatric data assessing accuracy of ICD-10 sepsis coding. In addition, the recent change in sepsis definitions implies a need for reliable coding of organ dysfunctions. We analysed accuracy of sepsis diagnosis based on ICD-10 codes in a prospective cohort of children with sepsis. Methods: We performed a multicentre, prospective cohort study at ten paediatric hospitals in Switzerland from 01.09.2011 to 31.12.2015, re-cruiting children younger than 17 years with blood culture-proven sepsis. For this analyses, we excluded prematurely born neonates and infants less than 7 days old at sepsis onset. We used 2005 consensus defini-tions - based on prospectively collected clinical and laboratory data - as the reference standard to define sepsis and organ dysfunctions. ICD-10 codes were extracted from mandatory official hospital discharge data re-lated to the sepsis episodes at 8 study sites. Results: Of 679 episodes with blood culture-proven sepsis recruited in the participating hospitals, 455 (67%) were classified as sepsis by ICD-10 codes. 305 (45%) had at least one organ dysfunction according to 2005 consensus criteria and 249 (37%) according to ICD-10 codes; the accuracy of ICD-10 codes for organ dysfunction was 69%, sensitivity 56%, and specificity 79%. In 163 (24%) episodes multi-organ dysfunction syndrome (MODS) was present as per 2005 consensus criteria and in 98 (14%) with ICD-10 codes; the accuracy of ICD-10 codes for MODS was 64%, sensitivity 39%, and specificity 93%. Sensitivity of ICD-10 codes was especially poor for central nervous system (6%), hepatic (7%), respiratory (26%), and hematological dysfunction (30%), and only fair for renal (45%) and cardiac dysfunction (48%). Conclusion: Compared to the reference standard, ICD-10 codes under-estimated incidence of sepsis, organ dysfunctions, and MODS and did not reliably depict organ dysfunctions in children with blood culture-proven sepsis.L6287264032019-08-05 <br />