17th September, 2008 Rotterdam
In Rotterdam for the initial meeting of our new EU-funded project Avoidable mortality in the European Union: towards better Indicators for the effectiveness of Health Systems (AMIEHS). Jointly led by Johan Mackenbach at Erasmus Medical Academy and us at LSHTM, with partners from France, Germany, Spain, and Estonia, it seeks to understand how the concept termed “amenable mortality” can be used as an indicator of health system performance.
The concept of amenable mortality was developed by Rutstein and colleagues in the 1970s. It was based on the premise that deaths from certain causes, and certain ages, that should not occur in the presence of timely and effective care. Subsequent work has expanded the list of causes of death considered amenable, reflecting advances in health care, and increased the upper age limit for these deaths, reflecting improvements in life expectancy. The concept has also been refined to include differentiation of causes amenable to the health care system and those to public health policy, while specific causes have been partitioned into the proportion to which reductions are attributable to primary, secondary, and tertiary actions.
In recent years, amenable mortality has undergone something of a renaissance. In part this reflects the much greater interest in performance of health systems, stimulated by the 2000 World Health Report, with improved tools being sought avidly by policy-makers seeking to determine whether they are getting value for money. An example was our study showing that deaths from amenable mortality in the USA around the year 2000 had hardly changed at a time when other industrialised countries were experiencing substantial declines.
In its original conceptualisation, amenable mortality included some conditions where medical care could do little to prevent death once the disease process had occurred but where the onset of the disease could be prevented by health promotion activities. This is exemplified by lung cancer, where, it was argued, health professionals could be effective in preventing people smoking or encouraging them to quit. However, assuming they were successful, the deaths that would then be avoided would occur several decades later. Clearly, this is incompatible with the idea that contemporary rates of amenable mortality reflect the current performance of health systems. Hence, only those deaths than can be prevented by contemporary interventions should be included. The definition of contemporary is inevitably somewhat arbitrary but a period of five years has the attraction of consistency with the period used to assess what is popularly considered as “cure” in analyses of cancer survival.
The original list of amenable causes included causes of death that could be prevented entirely by health care and those from which some deaths would be inevitable but the number could be minimised. The former is exemplified by vaccine preventable diseases such as measles; the latter by ischaemic heart disease, where even in the best performing health care system, some deaths will be sudden and unobserved. However, there are also many causes of death not considered to be amenable where, in some circumstances, health care can be life-saving. This is true of many cancers for which a small proportion may be identified early, making possible curative treatment. An example is cancer of the pancreas. This begs the question of what proportion of deaths from a specific cause should be preventable for the cause to be considered amenable. This issue has previously been addressed only implicitly but it is now time to make it explicit. The figure is, again, somewhat arbitrary but we propose that a 50% reduction has the benefit of simplicity.
There are, however, a number of considerations to be taken into account. In some cases, reductions in mortality in this scale will be achievable with a single intervention. The term “magic bullet” recalls the dramatic benefits of penicillin when it was first given to patients with severe staphylococcal infections in the 1940s. More often, health care will prevent deaths through a combination of interventions that were introduced incrementally, perhaps over decades. In these cases it will be necessary to look at changes in death rates over considerable time, introducing the problem of attribution as it is necessary to exclude other explanations for observed changes.
It will be necessary to draw on a variety of sources of evidence. In some cases, there will be randomised controlled trials. However, these are most likely to exist for single interventions; they are much less likely where a combination of interventions is involved. Randomised controlled trials also face the problem of external validity, as they often exclude both children and older people, those with co-morbidities, and historically, women. Hence, it will also be necessary to draw on natural experiments, where it is possible to determine when new treatments were introduced. An example is the introduction of HAART for patients with AIDS, where death rates fell very rapidly. In other cases, even where detailed data are unavailable, it may be possible to infer the impact of health care where there has been wider system change. An example is the political transition in eastern Europe around 1990. The opening of borders to modern pharmaceuticals and ideas of evidence-based medicine made it possible to provide treatment that was previously denied to sufferers from many chronic diseases. Thus, in countries such as Estonia, there was a rapid decline in mortality from stroke, almost certainly as a result of better treatment of hypertension, at a time when such deaths were increasing in neighbouring Russia. It may also be necessary to look at historical evidence. Thus, conditions such as acute appendicitis became amenable to health care once the introduction of asepsis and anaesthesia made intra-peritoneal surgery possible in the late 19th century. Treatment of hypertension has a shorter history but has still been possible since the late 1950s.
In all previous studies, the definition of amenable deaths has had an upper age limit, reflecting the view that “everyone must die of something”. The age limit has increased over time, from 65 to 75, but this creates certain problems. The first is that it is explicitly ageist, as it devalues curative care for those aged over 75. The second is empirical, first because life expectancy in some countries now exceeds this figure but, second, as there is growing evidence that many types of health care are very effective in older people. If, however, the definition of an amenable cause is one where health care can reduce the death rate by 50% or more, then there is no intrinsic reason to have an upper age limit. Yet, while conceptually attractive, this also poses problems of obtaining evidence, first because older patients are often excluded from trials but, second, because the absence of an observed decline in mortality at older ages at a time when an intervention was being introduced may simply mean that this population was not offered treatment.
So far we have not addressed one of the most difficult definitional issues in assessing health system performance, how to define the borders of the health system. The 2000 World Health Report adopted an essentially pragmatic definition as it was necessary to include all of the WHO’s 193 member states, the majority of which had no functioning system f vital registration and certainly no possibility of ascertaining causes of death. As a consequence, it defined the health system extremely broadly. This included a range of inter-sectoral actions. It is, however, difficult t justify holding the health system to account for actions that others must take. For this reason, we propose that the boundaries must be drawn more tightly, to include interventions delivered by those working what is unambiguously the health care system but also those developed by public health agencies, such as immunisations and screening for cancer.
This is a three year project. What I have described above is only the first step, as we then need to show whether changes in amenable mortality actually do correlate with innovations in health care. If we are successful, this should be a valuable contribution to the debate on health system performance.