Does the English NHS really have a problem dealing with cancer – or are the relatively high English death rates a statistical illusion?
The Royal College of Pathologists is currently pressing the UK government to implement the reforms to death certification prompted by the murderous activities of Dr Harold Shipman. These are intended to introduce a more rigorous approach to the recording of the causes of death. They may, however, have the unintended consequence of making the NHS appear to perform more poorly than other developed countries in the treatment of some diseases.
Comparisons of mortality data are the bedrock of modern epidemiology. Great conclusions are drawn about the risks of life – in different locations, occupations, ethnic groups – which are then used to inform policy and practice in health care and public health. International league tables are constructed and used to praise or chastise national health systems for their performance relative to each other. Death seems the most unequivocal available indicator of population health: you are either 1 or 0. Causes of death can be categorized by reference to the extensive and detailed International Classification (ICD) scheme.
In practice, however, such assumptions about mortality records are not well-founded: indeed the bedrock comes to look, if not like quicksand, like rather flaky shale. Death recording is a complex social process that constructs the eventual output in ways that can introduce systematic biases before comparisons are made. Having introduced these, the output then feeds back and accentuates the biases through the effects on those responsible for defining the cause of death. Mortality is a social and organizational construct and cannot be validly interpreted unless those processes are understood. This applies with particular force to comparisons between national health systems working within very different cultures and with different institutional structures for the recording of deaths.
Together with Richard Keller, we carried out a study of death registration in England, France and the USA, which was published (in French) in 2010 (Enregistrer les morts, identifier les surmortalités, Presses d l’EHESP). Although the outcomes of these systems are treated as equivalent for epidemiological purposes, they are actually compiled through quite different social processes. In England, death certificates are public documents, normally completed by the last physician to see the patient alive and lodged with a civil official before being returned weekly to a national agency, which codes the causes for statistical purposes. After processing, death certificates are filed in public archives which anyone can access. In France, death certificates come in two parts: one certifies the fact of death in an open document publicly accessible at local government offices; the other certifies the cause in a confidential document, sealed by the completing physician, who need not have seen the patient before death. Each part is returned to a different national agency. The system was designed to maintain maximum confidentiality about the cause of any individual’s death. In the US, each state – and in some states, each county – establishes its own process within fairly broad federal guidance. In Wisconsin, studied by Keller, for example, the process is co-ordinated by funeral directors, with physicians playing a secondary role in adding data on the cause of death. The certificates are excluded from the state’s freedom of information law, although they can be viewed by anyone who can show a legitimate interest. Summary information is extracted from the certificates at state level and forwarded to national level: in about 15 per cent of cases, this review results in the information being referred back for further checking. All of these systems are gradually migrating to electronic forms consistent with their basic principles.
The differences between these systems are deeply rooted in each country’s history of relations between the state and the medical profession. They introduce a potential for biases that is poorly understood. In England, for example, the public nature of death certificates makes physicians clearly accountable for what they write, although the Office of National Statistics concede that the quality of recording varies greatly. Nevertheless, ICD coding is now done automatically in about 80% of cases, using software that embeds the categories and rules for their application. In France, however, there has been a long struggle to get physicians to record any detail when certifying cause of death, and continuing concern about their tendency to enter rather general comments rather than drilling down in ways that allow full scope for ICD coding. This is compounded by relatively low autopsy rates: around 1.5 per cent, compared with 20 per cent in England and 13 per cent in the USA. The emphasis on confidentiality means that there is little accountability for the entries on the certificate. In the USA, the diversity of physicians who may complete certificates and their marginal role, except where medico-legal issues arise, again means that the potential for error is significant. Keller gives the example of cancer, where physicians in New Jersey can attribute death from an overdose of painkillers to the cancer for which they were prescribed, as a natural cause: in Wisconsin, physicians should classify this as a potential suicide and refer the case for medico-legal investigation. This introduces obvious potential for New Jersey cancer death rates to be inflated relative to those in Wisconsin.
Remaining with cancer, the English NHS is often criticised for its apparently poor performance relative to other European health systems. However, there is great international variation in the extent to which doctors tell patients and their families that they have cancer, particularly where the prognosis is poor. England tends to be at the extreme in this respect, with a very strong ethic of openness between doctors and patients. It is certainly worth investigating the possibility that the relatively low English ranking is an artefact of the interaction between the cultures of the medical profession and the recording systems used in different European countries. To the extent that England introduces more rigorous processes of certification, these differences are likely to increase.
Numbers seem to take on a magic of their own when they are big enough and are detached from the processes that create them. If differences in recording are introducing systematic biases into the data, these need to be understood if we are to draw correct conclusions from comparative studies. Otherwise we may divert a lot of energy into solving the wrong problem.