Does quality control of death certificates in hospitals have an impact on cause of death statistics?

BACKGROUND The effects of inaccurate death certificates on cause of death statistics are uncertain. Since 2008, Akershus University Hospital has systematically corrected all death certificates. The effects of these corrections on the total cause of death statistics from the hospital were studied. MATERIAL AND METHOD ICD-10 codes for the underlying cause of death on the original and the corrected death certificates issued by Akershus University Hospital were retrieved from the Cause of Death Registry for the period 1 May 2008-31 December 2009, once the Cause of Death Registry had processed the death certificates with the aid of the computer program ACME (Automatic Classification of Medical Entities). RESULTS Altogether 1,001 deaths were investigated (547 men and 454 women). A total of 223 death certificates were corrected. This entailed changing the underlying cause of death in 176 cases. Death certificates for women were corrected most frequently. In 121 cases, the changes entailed a change of disease chapter in ICD-10. The corrections caused a significant reduction in the number of unspecific diagnoses, such as sepsis, cardiac arrest, pneumonia with no further specification, renal failure and fractures without any specific cause. There was a significant exchange of individuals within all the large diagnostic groups, with the exception of cancer. Because of the balancing effect of exchanges within and between the disease chapters, this generated only minor effects on general statistics on causes of death. INTERPRETATION The continuous correction of death certificates in the hospital was important for adjustments at the individual level and as a quality control of cause of death statistics, but had only minor effects on the general statistics from the hospital.