Several attempts have been made to identify objective quality indicators for causes of death registries (7) – (9). Criteria used include degree of coverage and completeness, the use of an updated and sufficiently detailed coding system (ICD-9 or ICD-10), and that few deaths are registered with unspecific diagnostic codes for their underlying cause.
Degree of coverage refers to the proportion of the population encompassed by a registry. Completeness indicates whether information is actually collected from the persons who are included in the registry. In Norway, both of these are high; on the whole, we assume that we have medical information on more than 98 % of all deaths. The Causes of Death Registry encompasses all residents, irrespective of whether they die in Norway or abroad, and since 2012 non-residents (tourists, labour migrants etc.) who die in Norway are also included. Cross-checks against the National Registry are made, and reminders are sent to the Chief Municipal Medical Officers to collect missing certificates. Each year, death certificates are missing for 500–700 deaths, even after multiple reminders have been sent. Approximately half of these pertain to residents who have died while abroad. For non-residents who die in Norway the degree of coverage cannot be established, since the National Registry cannot be used for verification.
For coding we use the computer program IRIS with the ACME module (10) to allocate an ICD code to the diagnoses on the death certificate and identify the underlying cause of death. In approximately half of all deaths, the underlying cause is selected with the aid of IRIS/ACME. For the remaining deaths, an assessment by a professional coder is required. As noted above, a considerable proportion of the death certificates have an illogical structure, but the underlying cause of death may nevertheless be identified by the ICD-10 coding rules.
An important objection against the data quality of the Causes of Death Registry is the frequent occurrence of unspecified or non-meaningful diagnoses, so-called «garbage codes» (11), as the underlying cause of death. These are often diagnoses that reflect a terminal stage or complication (immediate cause of death), but fail to provide any information on the underlying cause. A typical example is «heart failure», which is reported as the underlying cause of death in approximately 4 % of all death certificates (12). Without any additional information it is impossible to know whether this heart failure is a result of ischaemic heart disease, cardiomyopathy or chronic pulmonary disease, to mention only a few possibilities. Another example is death caused by injuries, without any information on the event that caused the injury.
In two of the quality assessments referred to above (7, 8), the Norwegian Causes of Death Registry was assessed in the second-best group, with a «medium» and «medium-high» quality respectively. In the third (9) we were ranked in the best group, with a total point score of 87.6 out of 100, but below the other Nordic countries. In all these three studies, the extensive use of unspecific codes served to lower our score.
Additional to these criteria is the question of whether the diagnosis on the death certificate actually reflects the real underlying cause of death. The cause of death cannot always be determined, and it cannot always be taken for granted that the idea of a single underlying cause of death conforms to reality. Autopsy is often seen as the gold standard, but autopsies are infrequently undertaken and the selection for an autopsy is not random. Therefore, the results from autopsy studies may not be directly transferable. A Norwegian study showed that autopsy findings changed the underlying cause of death in 61 % of all cases, in 32 % even to another ICD chapter (13). The proportion of deaths in which a medical autopsy was undertaken was low, however, at 4.3 %.
Another approach consists in reviewing the patient records and completing a «renewed death certificate» on the basis of available information. Using this method, the study from Akershus University Hospital (6) found erroneous content in 27 % of the death certificates. Continuous quality assurance at the hospital provided an important corrective at the individual level and reduced the number of unspecific diagnostic codes, but because of the balancing effect of exchanges between different ICD chapters, this generated only minor changes at the group level (14) and thereby also in the statistics on causes of death.
A Swedish study rediscovered the main diagnosis from the last hospitalisation period on the death certificate for 83 % of those who had died in hospital (15). In a French study there was an «acceptable» consistency between the main diagnosis from the last hospitalisation period and the reported cause of death in 88.8 % of deaths that had occurred in hospital and 72.9 % of deaths outside hospital (16). In such studies, however, it must be kept in mind that there is not necessarily any concurrence between the condition for which the patient was treated in hospital and the underlying cause of death.