BACKGROUND: Pharmacoepidemiology studies examining serious ventricular arrhythmias (SVA) may ascertain this outcome from data on hospitalizations, but there is limited information on the validity of these diagnoses in administrative databases. An ongoing study of SVA and sudden cardiac death offered the opportunity to perform such a validation. Objectives: Describe the validity of a two-stage process for identifying idiopathic SVA among hospital discharges in the Canadian Saskatchewan Health database.
METHODS: The study population comprised over 86,000 users of domperidone or a proton pump inhibitor by prescription (1990–2005). Hospitalizations with ICD-9 and ICD-10-CA codes specific for ventricular tachycardia or fibrillation as the principal diagnosis were identified. A cardiologist, blinded to the exposure of interest, reviewed electronic patient profiles; cases with terminal illness, cancer, or other prespecified known immediate causes of arrhythmia were excluded. Suspected cases were validated by hospital medical record abstraction. The positive predictive value (PPV) was calculated for the SVA screening algorithm. Validation of noncases was not performed.
RESULTS: A total of 117 patients met the screening definition of SVA; after profile review, 74 were judged as suspected cases for abstraction. Of these, 65 records (88%) were abstracted and 34 were validated as idiopathic SVA (PPV¼52%; 95% CI: 40–65%). An additional 14 cases were confirmed as SVA but did not meet the study criteria for idiopathic SVA. Thus, the PPV of the screening algorithm for all SVA (idiopathic or not) among the validated cases was 74% (95% CI: 63–85%).
CONCLUSIONS: Even when using hospital discharge diagnoses, case validation is useful for identifying SVA and appears essential for identifying idiopathic SVA. Profile review before chart abstraction can increase the efficiency of the validation process.
