Validation of an ICD-10 coded case definition for the identification of patients diagnosed with sepsis and severe sepsis using administrative data

Abstract

Background: We assessed the validity of existing ICD case definitions used to identify sepsis in administrative data and validated and optimized an existing ICD-10-CA coding algorithm to identify patients diagnosed with sepsis. Methods: Standard systematic review methodology was applied to assess the validity of ICD case definitions for sepsis. The CIHI ICD-10-CA coding algorithm for sepsis was validated and optimized using a randomly selected cohort of ICU and non-ICU patients. Sensitivity (Sn), specificity (Sp), positive predictive value (PPV), and negative predictive value (NPV) were calculated. Results: Twelve studies were identified in the systematic review with a range of diagnostic accuracy reported indicating that sepsis is highly under-coded. We increased the accuracy of the CIHI ICD-10-CA coding algorithm for sepsis (Sn: 71.9%, NPV: 66.6%) and severe sepsis (Sn: 65.1%, NPV: 70.1%) while slightly decreasing Sp and PPV. Conclusions: Sepsis is highly under-coded in administrative data. The new definition has a much higher sensitivity and negative predictive value.