Acute heat illness (AHI) from extreme environmental heat exposure can lead to emergency department (ED) visits, hospitalization, and even death. While the ICD ninth revision codes for AHI have been validated in the U.S., there have been no studies on the validity of ICD-10 codes for AHI in Canada. The objective of this study was to assess the validity of an ICD-10 coding algorithm for ED encounters for AHI. We conducted a retrospective cohort study of children and adults who had ED encounters at two large academic, tertiary hospitals in London, Canada, between May and September 2014-2018. We developed an algorithm of ICD-10 codes for AHI based upon a literature review and clinical expertise. Our "gold-standard" definition of AHI was patient-reported heat exposure and documentation of at least one heat-related complaint. To establish positive predictive value (PPV), we reviewed 62 algorithm-positive records and noted which met our "gold-standard" definition. To calculate negative predictive value (NPV), sensitivity (Sn), and specificity (Sp), we randomly reviewed 964 ED records for associated ICD-10 codes and diagnoses. Two independent reviewers completed blinded data abstraction, with duplicate abstraction in 20% of the sample. Of the 62 algorithm-positive records, mean (SD) age was 38.8 (23.8) years; 37% were female. PPV was 61.3 ± 12.1% (95% CI). Of the 964 randomly selected records, mean (SD) age was 41.7 (26.5) years; 51.1% were female. The NPV was 99.7 ± 0.4%, sensitivity 25.0 ± 42.4%, and specificity 100.0 ± 0.0%. An ICD-10 coding algorithm for AHI had high specificity but was limited in sensitivity. This algorithm can be used to assemble and study cohorts of patients who have had an AHI, but may underestimate the true incidence of AHI presentations in the ED.
Keywords: ICD-10 coding algorithm; acute heat illness; emergency department; retrospective cohort study.