Pitfalls in the use of death certificates for assessing cause of death: a study of tonsil carcinoma patients.

A study was undertaken to determine the accuracy of underlying cause of death as stated on the death certificate for patients treated at Mallinckrodt Institute of Radiology (MIR) for carcinoma of the tonsil. The sample consisted of 110 patients who were treated between 1953 and 1976 and who subsequently died; all patients were seen by a physician within 3 months of their death. Death certificate cause of death was compared with the cause of death as stated in the MIR patient record. Death certificates detected 42.5% of the deaths due to tonsil carcinoma recorded in the hospital record, and detection was found to correlate with time between treatment and death. Of the deaths reported on the death certificate as due to tonsil carcinoma, 91% were confirmed by the hospital record to be correctly reported. Overall agreement in assignment of cause of death between the hospital record and the death certificate was 59%, and was significantly affected by stage of disease at diagnosis. This review underscores the low reliability of death certificates and raises a serious question concerning the use of this information for patients who are registered in clinical trials without additional documentation of tumor extent and cause of death.