Burden of Disease Research Unit

Are the cause of death statistics reliable?

Statistics on the underlying causes of death are essential to assist policy makers in devising policies and implementing interventions to promote health; as well as interventions to prevent, treat or ameliorate disease. South Africa has made great strides in the production of cause of death statistics. Prior to 1994 the record of the number of deaths and the cause of deaths was of limited usefulness because the deaths of many South Africans were not recorded. This was mainly because they lived in the so-called independent homelands or in poorly serviced rural areas where the infrastructure to register deaths was not in place.

Major efforts by the South African government to improve death registration has involved, amongst others, the introduction of a new death notification form as well as the establishment of provincial task teams to identify and resolve logistical problems. It is estimated that over 90% of the adult (aged 15 years and older) deaths are now registered but the proportion is lower for children.

When a person dies, a medical practitioner is required to complete details about the medical condition that resulted in the death (immediate cause) and also provide details about any conditions that may have led to this (underlying cause) on the death notification form. In some areas where it is not possible for a medical practitioner to complete the death notification, a traditional headman may be authorized to do so.

The death notification form is submitted to the Department of Home Affairs to register the event of the death, and at a later stage, these are given to Statistics South Africa to code the information and identify the underlying cause (the medical condition that initiated the sequence that ultimately resulted in death). Statistics South Africa code the causes according to the International Classification of Diseases (ICD-10). They then collate and report the cause of death statistics based on the notifications of death. The production of the cause of death statistics for 1997-2003 reflects that a major milestone has been reached with recent effort by Statistics South Africa to catch up the backlog in the processing of these statistics.

Research into the quality of the information collected on the death certificates has revealed that there is a need to improve the quality of the information collected through this process. In particular:

1. Injuries account for about 12% of all deaths. However, in the majority of these cases only the nature of the injury (eg. head injury) and not the underlying cause of an injury death (assault, suicide, or motor vehicle accident) is specified on the death notification. Such deaths require an inquest to establish the cause of death and there is currently no mechanism to link the finding of the inquest to the cause of death statistics.
2. Many of the AIDS deaths appear to have been certified as due to the indicator condition without reference to HIV or AIDS. This is likely to be due to several factors. Doctors may for reasons of confidentiality decline to list HIV or AIDS on the death certificate, or there are situations where the doctor is not aware of the HIV status of the deceased.
3. There is still a high proportion of deaths that are classified as ill-defined signs and symptoms. In developed countries the proportion of ill defined deaths is usually well below 5%. In contrast, in South Africa in 1990, ill defined deaths accounted for 22%. This has declined to around 12% in 2001, however, further improvement is required.
4. There is a high proportion of ill-defined cardiovascular causes of death such as heart failure, cardiac arrythmias, cardiac arrest etc. Together these account for approximately 25% of cardiovascular deaths. In the 1996 and 2000/1 sample cause of death data approximately 3% of all deaths were recorded as heart failure and approximately 1% as cardiac arrest.

The degree to which cause of death coding involves any interpretation of the details provided on the death notification will affect the statistics produced. When a strict mechanistic coding practice is a adopted (as is the new practice with Statistics South Africa) then AIDS cases certified with euphemisms such as 'acquired immuno-supression' will be classified as 'unspecified immunodeficiencies' and will form part of "diseases of blood". If the immediate cause is specified, then this will be identified as the underlying cause, in favour of a less well specified condition.

There is clearly a need to strive to improve the details of causes of death completed on the death notification forms and to overcome the barriers surrounding social stigma. so that the quality of the cause of death statistics can be enhanced. It is also clear that further analysis of the cause of death data and rigorous comparison with data from other sources are important in order to derive coherent estimates of trends in cause specific mortality. This, however, should not detract from the enormous strides made thus far in improving the production of vital statistics.

Further information can be found at Stats SA to distil mortality trends by Pali Lehohle 3 Feb 2005

Authors: Dr Rosana Norman, Prof Debbie Bradshaw, Michelle Schneider, Desiree Pieterse and Dr Pam Groenewald
Updated 16 February 2005