Reliability of causes of death in persons with Type I diabetes.

Aims/hypothesis: To compare causes of death assessed by a clinical review committee, the information given on death certificates, and ICD-codes provided by the State Documentation Office in deceased persons with Type 1 (insulin-dependent) diabetes mellitus.
Methods: A cohort of 3674 patients were monitored for 10+/-3 (mean +/- SD) years. Vital status and end-stage diabetic complications were documented for 97%; 251 patients had died. Causes of death were assessed by a clinical review committee and compared to the information provided by death certificates and ICD-9 codes.
Results: The review committee defined a leading cause of death in 94% of cases, whereas death certificates were available for 73% and ICD-codes for 79% of patients; 10% of death certificates could not be evaluated due to insufficient information. Diabetes was mentioned on 71% of death certificates, and renal disease in 75% of cases with renal replacement therapy. There was acceptable agreement between the committee, death certificates and ICD-codes only for deaths due to neoplasma, and between the committee and death certificates for deaths due to acute myocardial infarction, cerebrovascular events and accidents. In only one out of four deaths due to hypoglycaemia and in four of seven deaths due to ketoacidosis was this diagnosis mentioned on the death certificate. No death due to hypoglycaemia or ketoacidosis and 41% due to suicide were identifiable by ICD-codes.
Conclusion/interpretation: Reliance on death certificates or ICD-codes as the only sources of information on the cause of specific mortality does not provide data of sufficient reliability for evaluation of clinical outcome in Type I diabetes.