As good as it gets? Sensitivity and specificity of death certificates for diabetes: The Rancho Bernardo

Purpose Diabetes, reported as the sixth major cause of death in the US, may be an underestimate, as it is often under-reported on the death certificate. This study examines the change over time in accurate reporting and death certificate sensitivity and specificity of diabetes. Methods Death certificates for 3209 decedents from the Rancho Bernardo cohort were obtained. Decedents were aged 30 and older at enrollment (1972–74); mortality data was available through 2003. True diabetes status was based on history obtained at periodic clinic visits and mailed questionnaires during follow up. Diabetes listed as cause of death or reported anywhere on death certificates was abstracted. Sensitivity and specificity were stratified by sex, age at death, year, place, and cause of death, and medication use among diabetic participants; characteristics that contributed to reporting diabetes were analyzed in univariate and multivariate logistic regression models. Results There were 378 decedents with diabetes, of whom 168 had diabetes listed anywhere on the death certificates. Thus the sensitivity and specificity of diabetes reporting anywhere on the death certificate were 34.7% and 98.1% respectively. Death certificates for those with diabetes listed cardiovascular disease (CVD) as the cause of death more often than did death certificates for those without diabetes (p < 0.05). Sensitivity for diabetes reporting was higher for deaths occurring in hospitals, CVD listed as the underlying cause, and among diabetic participants who reported use of medication for diabetes. Overall sensitivity did not vary significantly by age or year of death, or sex. After stratification by decade of death, sensitivity for diabetes reporting for deaths occurring in 1992–2003 was greater when the cause of death was listed as CVD compared to any other cause (48.9% vs. 28.6% respectively, p < 0.05). Conclusion This study showed no overall improvement in diabetes reporting on death certificates. Reporting improved in recent years (1992–2003) for those whose death certificates indicated CVD as cause of death, likely reflecting the recent interest in diabetes as a major CVD risk factor.