Finding coding errors that led to a false sense of security over mortality ratios

Richard has been with CHKS since 2006 and works with clients in the London and South East region.

I was asked if I would be a member of a large multi-site acute trust’s mortality review group. The trust wanted to review mortality indicators, but also look at untoward incidents, coroner’s reports and mortality avoidance. The group includes the deputy medical director, director of public health, associate chief nurse, consultant or senior nurse representation from each Clinical Academic Group, the head of clinical coding and the clinical effectiveness unit manager. The Mortality Review Group also provides assurance to the Clinical Standards Committee that all mortality in the trust is being reviewed at local level and that regular monitoring and any issues arising are reported.

The trust’s Summary Hospital Level Mortality Indicator (SHMI) was one of the lowest in the country. However, I felt that it was actually too good to be a true reflection of mortality and I convinced the group that it would be worth investigating. My feeling was there could be a data quality issue that was making the trust look particularly good.

Many trusts only start to engage with mortality indicators when they are high. My experience is that trusts at the other end of the spectrum also need to be looking carefully at them to see whether there are reasons why the indicators might be flattering.

When we looked at the way the trust was coding we found a significant percentage of attendances were being recorded as emergency admissions when they were actually regular admissions. We benchmarked and compared with other trusts and confirmed this mis-recording of activity and estimated that the SHMI was being reduced by 6/7 points. We took this to the board and further investigation was agreed. This has subsequently found there were changes that needed to be made to the way attendances were coded.