Nationwide individual record linkage study showed poor agreement of causes of death and hospital diagnoses at individual level but reasonable agreement at population level

Objectives: To investigate to what extent underlying and multiple causes of death represent end-of-life morbidity in individuals and at population level. Study Design and Setting: Cause of death and national hospital data were individually linked for all deaths at the age of 50-84 years, in 2005 in the Netherlands (n = 86,987). The individual agreement of diseases registered as a diagnosis of discharge in the last 2 years of life and underlying and multiple causes of death recorded was assessed. Cause-of-death ranking was compared with ranking of hospital diagnoses. Results: The percentage of persons with a hospital diagnosis registered as the underlying cause of death was <30% for most diseases, except for cerebrovascular disease, chronic obstructive pulmonary disease and bronchiectasis, acute myocardial infarction (40-60%), and cancers (70-90%). Low Cohen's kappa values confirmed poor individual agreement between hospital diagnoses and underlying and secondary causes of death recorded. At population level, however, frequency rankings of underlying and multiple causes of death agreed reasonably well with frequency ranking of hospital diagnoses (Spearman rho of 0.58-0.60 and 0.61-0.63). Conclusion: Underlying and multiple causes of death poorly represent diseases present at the end of life in individuals but show reasonably well which diseases are most common at population level. (C) 2014 Elsevier Inc. All rights reserved.