Our work in end of life care suggests coding is being significantly under-recorded. This impacts on risk-adjusted mortality rates and issues relating to length of stay, which in turn has implications for organisations that compile data.

Medway trust in Kent asked CHKS to review its mortality rates as part of its application for foundation trust status. So we undertook an audit of recorded deaths and coding. We found that a number of patients had been placed on the pathway for end of life care but this had not been clinically coded as such.

The trust also had concerns that it was providing this service in an acute setting and wanted to quantify the impact of these patients on its services. It also wanted to know the size of the issue in order to plan the best mode of care for patients, families and staff in an acute setting.

We recommended coding these patients through code Z515, designated for end of life care, supported by the documented care pathway in the patients’ notes. The trust agreed, updating the clinical codes for all deaths in 2007-08. We then refreshed the data and compared it with the previous trend (see first graph). The effect of the review was that 37 per cent (513 out of 1,401) of the deaths were excluded from the risk-adjusted mortality rate (previously only 8 per cent had been classified as end of life care), so the scores now reflect the acute deaths rather than including end of life care.

Extent of the problem

To find out if the problem is widespread we analysed hospital episode statistics data, looking at the proportion of deaths coded as end of life care across general acute hospitals (see second graph). There is a wide variation. The mean is only 4.5 per cent - and even the high end outlier has only 22 per cent of its deaths coded this way. It appears that this element of care is being massively under-recorded. This has various implications. First, commissioners and providers need to get a proper fix on the scale of the issue in order to improve end of life care. Potentially end of life care is under-reported up to eight times what it should be.

Second, most of the methodologies for risk adjustment of mortality (producing a hospital standard mortality rate) will include many deaths that should be excluded, thus distorting the relative positions of providers. The first graph shows the risk-adjusted mortality index, moving from 105 to 68. In other words, inaccurate recording means hospital standard mortality rates are erratic.

Third, given that providers are under pressure to reduce length of stay, these patients need to be separated from length of stay measurements.