Statistical accountability: who pays for the excess?

Author: Tom King

The blue circle is the global

symbol for diabetes, introduced

by the International Diabetes

Federation.

Diabetes is an unusual disease because it affects so many people in the UK: more than 3 million adults. This means that the typical concerns about minimal clinically significant difference tend to not apply - high prevalence means any significant difference can be a substantive concern.

The cost of diabetes can also be very high because poor management over a number of years causes complications. These are expensive as they involve renal failure, retinopathy and amputations. Diabetes is also associated with obesity – resistance to insulin is acquired due to excess sugar in the diet, so the prevalence is rising.

All of this means that diabetes is an economic and public health problem on a grand scale but its clinical aspects are more indirect. This has led the Public Accounts Committee of the House of Commons to consider the efficiency of the money being spent on ensuring appropriate care and good outcomes for people with diabetes. Insulin resistance is more common than insulin underproduction but as the latter occurs later in life and for different reasons, a cost comparison is not trivial.

Diabetes can reduce life expectancy by as much as 20 years: the ultimate accountability of the NHS is mortality. Another calculation shows that annually there are 24,000 deaths of people with diabetes above the number that would be expected from a comparable population without diabetes. This is known as the ‘excess mortality’ and this methodology has been used to attribute deaths to poor care in Mid Staffs Hospital Trust and to the practice of Dr Harold Shipman. This excess represents around 3-4% of mortality in the UK making it important, despite the principal care for diabetes being glycaemic control and avoidance of complications.

In the case of Harold Shipman’s practice, much analysis was done which demonstrated the age groups he was targeting. It was also possible to identify a suspicious peak in deaths in the afternoon, when he habitually did his home visits. This information allowed confidence about how many people he had killed (nearly 200) and the group they were likely to be a part of but not identification of individual victims. It was also possible to see that this practice was paused at one point when he thought the authorities had rumbled him.

The inquiry into failings at Mid Staffordshire Foundation Hospitals Trust is still ongoing. Once again this is partly due to lack of specificity in the statistics: the excess mortality shows around 500 more deaths than would have been expected over a three year period. Some specific concerns about certain types of surgery were already known and poor resolution of these problems was certainly a failure but it does not account for much of the problem. Other deaths are most obviously down to poor care, processes and infections but this leads to a dispute concerning what is known as case/mix. Severity of illness, including comorbity, are an important factor in measuring the effectiveness of healthcare and this is not easy to measure and may not be well recorded. Arguments about the data led to warnings not being interpreted as indicative of a problem and there are still concerns that those in the health system do not have the capacity to engage critically with the statistics. Relatives hoping for accountability for the death of a loved one are left without any justice by the statistics.

On diabetes, those questioned by the Committee hearing did not see it simply either, quibbling over whether excess deaths could be regarded as ‘avoidable’ as others have described them. Sir David Nicholson, head of the NHS, promised to come up with a formal position but as it is early diagnosis and good management which have been found to be effective, all of which he is responsible for, it is likely to be most of the excess which he should have avoided. Another possibility would be that excess deaths were a responsibility of a much lower level of the system, beyond the control of the chief executive, although this would still represent weak leadership.

There is a technical point around the inevitability of diabetes (especially type 1) and the consequences of it. This is why the National Service Framework requires 9 care processes to be taking place annually (monitoring, education, advice and checks for complications). Although the report of the NAO went into this coverage in some detail, data were limited and impact is slow so that the discussion was inconsequential. While death may be an important consideration, the central focus of the NHS should be health but care is the product of complex interplay within a system which is hard to pin down to a simple number. Sadly data on delivery of the care processes has only been collected relatively recently and is still piecemeal.

Rather than being an example of statistical accountability, excess mortality demonstrates the need for appropriate statistics – politicians are not experts and need good information which gets to the root of the issue. Data should be sufficient for the accountability: for a long term condition deaths are far removed in time from the initial poor care and regional structure may have a much longer and complex legacy. There also needs to be agreement on the denominator for accountability: the NHS prioritises mortality but diabetes is a chronic condition where complications are the expensive issue to be avoided by the care process.

It may be that expectations of statistics are too great: an epidemiological method does not set out to prove the mechanism of a particular case but to estimate risks in aggregate. Often statistics give more questions than answers but good statistics give better questions, and in the public domain when the users are not experts, whether politicians or media, accountability depends on good information. In the case of diabetes, the Committee promised another hearing in a few years’ time – we can only hope that better data will lead to fewer excuses.