Clinicians didn't reliably distinguish between different causes of cardiac death using case histories.

摘要

BACKGROUND AND OBJECTIVES: Routine statistics and epidemiologic studies often distinguish between types of cardiac death. Our aim was to assess agreement between doctors on cause of death given identical clinical information, and to assess agreement between a physician panel and the original cause of death as coded on national statistics. METHODS: Clinical information and autopsy reports on 400 cardiac deaths were randomly selected from a defined population in the West Midlands, UK. A panel of eight clinicians was assembled, and batches of 24-25 cases were sent to pairs of these clinicians who, blinded to the certified cause of death, independently of each other assigned underlying cause of death. Physician panel decision was achieved by consensus. Levels of agreement were assessed using the kappa statistic. RESULTS: Reviewers agreed on cause of death in 54% of cases (kappa = 0.34). Consensus decision of reviewers agreed with death certificate diagnosis in 61.5% (kappa = 0.39). Agreement was higher if an autopsy had been performed (kappa = 0.49). CONCLUSION: The process of identifying underlying cause of death is of limited reliability, and therefore, limited accuracy. This has implications for design of epidemiologic studies and clinical trials of cardiovascular disease.