Validation of algorithms using International Classification of Diseases for the identification of herpes zoster episodes requiring hospitalization in Quebec, Canada.

OBJECTIVE: We determined secular changes in the incidence of hospitalizations due to herpes zoster (HZh) and assessed the validity of HZ International Classification of Diseases (ICD) code algorithms for identifying HZh in a region of Quebec, Canada. METHODS: We performed a validation study as part of a retrospective cohort study of adult HZ patients hospitalized at Centre Hospitalier Universitaire de Sherbrooke during 2000-2017. Cases were identified using ICD codes from an inpatient administrative database. HZ cases identified by ICD-9 (053.xx) and ICD-10 (B02.x) codes were chart-confirmed, and performance characteristics of ICD code algorithms were calculated (positive predictive value [PPV] and sensitivity). RESULTS: Overall, 1314 hospitalizations with HZ diagnosis (HZh) with or without complications were identified during 2000-2017. Among the hospitalizations, 526 (44.4%) were due to active HZ disease or a complication related to a recent or previous HZ episode. These hospitalizations were due to active disease at the time of admission (340/526, 64.6%), HZ that developed during hospitalization (120/526, 22.8%), or a complication directly related to a recent or previous HZ episode (66/526, 12.6%). PPV was significantly higher when HZ was the primary diagnosis (276/310, 89%, 95% confidence interval [CI]: 85-92%) than when HZ was a secondary diagnosis (254/928, 27%, 95% CI: 25-30%) (p < 0.0001), and the PPV of a first secondary diagnosis (84/140, 60.0%, 95% CI: 51.3-68.2%) was higher than that of other secondary diagnoses (203/794, 25.6%, 95% CI: 22.6-28.8%) (p < 0.0001). An algorithm combining ICD codes and antiviral usage demonstrated the best sensitivity (86.3%, 95% CI: 83.1-89.1%) and PPV to identify HZh (100%, 95% CI: 99.2-100%). Poisson regression revealed no significant changes in HZh over time (incidence rate ratio: 0.98, 95% CI: 0.92-1.04%; p = 0.5). CONCLUSION: HZh incidence was stable over time. Prescription of antivirals might be a useful addition to ICD codes to identify HZh cases from administrative databases.