## Shifting sands

In civil engineering, no building can be sounder than the foundation on which it rests.  A similar comment applies to statistical analysis, which is obviously limited by the quality of the underlying data.  This is an issue for mortality projections, too, since these are ideally based on a historical record of good-quality data.

In the case of all-cause mortality rates, the quality of deaths data in the developed world is high.  In England and Wales, for example, the nationwide registration system means that the majority of deaths are recorded within a week of actual occurrence.  Without being flippant, there is never any real doubt as to whether someone is dead or not.  Projections of all-cause mortality rates therefore rest on a sound foundation.

What about mortality projections based on data disaggregated by cause of death?  I have written before about numerous issues surrounding data quality for cause of death, including the accuracy of categorisation and the impact of changes in coding guidelines for certifying doctors.  However, I recently came across yet another source of trouble: changes in how the information on death certificates is mapped to a cause of death.

The problem is illustrated by an exercise carried out by the ONS in England and Wales.  The deaths occurring in 2009 were taken and coded according to two different interpretations of ICD 10, the international classification of death currently in use in England & Wales.  The idea was to test the impact of the change in mapping death-certificate data to cause-of-death codes.  The following quotation is quite illuminating:

"Diseases of the Circulatory System showed a 5 per cent decrease between ICD–10 v2001.2 and ICD-10 v2010.  Within this chapter cardiomyopathy (I42), heart failure (I50), and cere[b]rovascular diseases (I60–I69) are affected by the selection rule wording changes that mean these conditions will not be picked up as the underlying cause unless other specific terms are also mentioned on the death certificate. The majority of these deaths are now coded to Respiratory Diseases’ (Chapter X) – in particular pneumonia (J18)."

Source: Office for National Statistics, Results from the ICD–10 v2010 bridge coding study.

How reliable can a projection methodology be where the leading cause of death in retirement ("Diseases of the Circulatory System") can fall 5% purely due to coding methodology?  Furthermore, the change resulted in deaths moving between two broad cause categories which are usually kept separate: circulatory diseases and respiratory diseases.  These changes are all the more remarkable because we are dealing with the same deaths (England & Wales, 2009) and the same classification system (ICD 10), so the only thing which has changed is the mapping process from death certificate to cause code.

Mortality forecasting is a complex business, and any projection process is vulnerable to uncertainties in the data.  Methodologies that forecast from a reliable and unambiguous starting point offer the best chance of achieving a credible result.

### RECENT POSTS

A short and simple administrative announcement ... In common with ... Read more
The upcoming EU General Data Protection Regulation places focus on ... Read more
Assume we have a random variable, $$X$$, with expected value ... Read more
Stephen Richards is the Managing Director of Longevitas