Medical care and public health
K. Jamrozik and M. Hobbs
Assessment of the current contribution of medical care to the health of populations
Features of the population served
What health problems currently exist?
What health services already exist?
How are the existing health services performing?
Special problems in particular populations
Assessment of proposed new services
Assessment of efficacy
Evaluation of the effectiveness of new strategies
To assist us to come safely into this world and comfortably out of it, and during life to protect the well and care for the sick and disabled. (McKeown 1979)
This chapter examines the role of medical care services as a determinant of the level of health experienced in a particular community. The discussion is deliberately limited to personal medical services that are delivered in a one-to-one setting. As well as emergency medical or surgical care, rehabilitation, and palliative measures, such personal medical services can include a range of preventive activities such as health education, immunization, and family planning. In considering the entire range of activities designed to maintain or improve the health of a given population, a distinction is drawn between personal medical care services of the type described and, on one hand, measures for protecting health such as provision of clean water, waste disposal, quarantine, and food standards, and, on the other hand, community-wide health promotion activities such as advertising campaigns or creation of public facilities for exercise. Clearly, proper control of certain health problems, such as infectious diseases, requires a co-ordinated approach containing elements of health protection, health promotion, and personal medical care services. In practice, however, these three different activities are overseen by separate parts of the health service or even by separate governmental departments. It is therefore meaningful to examine the effect on the health of the population of personal medical services, even if this fragmented approach is somewhat remote from the ideal one.
The approach of this chapter is based on assessment of the impact of personal medical care services on health in three stages corresponding to the past, the present, and the future. These provide different challenges because the data that are available or potentially available to examine the impact of personal medical services at each stage are qualitatively very dissimilar. To consider the historical dimension—to answer the question ‘What have personal medical services contributed to the state of health that we enjoy at present?’—is not only to reflect on what has been achieved, but also to identify those areas where humankind is reaching ‘the flat of the curve’ and further gains, in terms of further reductions at least in mortality, are likely to be small or expensive or both. With regard to the present and the future, to those charged with the planning and delivery of health services, identification of which of a number of possible initiatives might be truly useful in terms of improving health or reducing disease is at least as important as being able to demonstrate that programmes already in place are effective. Some appear to pay an obvious and direct dividend in terms of benefit to health, but it is less easy to be certain whether others justify the corresponding investments.
How best to measure health is a matter of central importance to this chapter but this is not a simple task. Instead, the second part of this chapter is built around five key questions that should inform the planning and delivery of health services to whole communities—’Who lives there?’, ‘What health problems do they have?’, ‘What health services already exist?’, ‘How are the existing health services performing?’, and ‘Are there any special features of the population or situation that need to be taken into account?’ Consideration of the overall gains in health already made provides an opportunity to review the basic tools and a number of approaches to the measurement at least of ill health. The evaluation of the effects of single, readily identifiable components within complex systems of medical care is also discussed as this is a key practical issue. This section therefore includes a series of case studies of the evaluation of medical care using the various approaches and types of data previously identified. Several of the examples are drawn from Western Australia, but the principles they illustrate have wide application.
The third part of the chapter is devoted to the assessment of proposals for new services. This task is slowly becoming easier as it is accepted that new health services and strategies should be introduced only after rigorous assessment via randomized trials, first under tightly controlled conditions, and later under conditions that resemble everyday practice more closely. Debate continues, however, as to the applicability of the randomized trial to situations where the outcome of interest is rare and the validity of using observational data to chart the diffusion and impact of new methods in health care. The final section of the chapter is a brief discussion of what the future for the evaluation of medical care might hold, including areas that require further developmental work before a comprehensive overview can be obtained of the effect of medical care on health.
Western industrialized nations invested heavily in health protection long before medical care services began to develop into the highly organized and elaborate systems that we see today. While it is true that hospitals and various kinds of individual practitioners have been providing personal medical care services in these countries for many years, quarantine regulations date back at least to the fifteenth century (Burnet and White 1972) and the state became embroiled in the campaign to provide basic water supplies and sanitation at the beginning of the nineteenth century. In the United Kingdom, governments had provided hospitals for the poor but state involvement in the provision of universal personal medical care services is usually dated to the inception of systematic antenatal care in the early 1900s.
There have been many achievements since that time. Immunization has led to the eradication of smallpox worldwide (Fenner 1980) and, in the developed countries, the virtual control of poliomyelitis, diphtheria, tetanus, and measles. Substantial progress has been made towards reduction in morbidity from rubella and the congenital abnormalities it causes. Through the exigencies of two World Wars and several other major conflicts, substantial improvements have been made in the management of injury, and in restorative surgery and rehabilitation. In the field of reproduction, large advances have occurred in the control of fertility and in obstetric and neonatal care; it is now possible to contemplate radical programmes for population control as part of public health. Modern anaesthesia and parenteral therapy have greatly increased the safety of emergency surgery and have extended the scope of restorative surgery. New drugs have led to the successful treatment of many bacterial conditions, although multiple-drug resistance is now a growing problem. Developments in pharmacology have also contributed significantly to improved prognosis in surgery for injuries and emergencies. Certain malignant tumours of childhood and testicular cancer are now curable and substantial relief, if not cure, can be provided for many of the more common cancers of adults and for psychoses, epilepsy, diabetes, ischaemic heart disease, and some forms of arthritis. New prosthetic devices have opened the door to surgery for conditions previously associated with major disability and handicap. Advances in methods for diagnosis have supported many of these developments in treatment, and the theory and practice of screening asymptomatic individuals for risk factors or early evidence of disease have steadily become more sophisticated.
In parallel with technical advances there have been major changes in the philosophy of the provision of health care and in the structure of health systems. In most developed countries the concept of universal access to essential medical care, regardless of means, has been accepted. Even though systems of payment differ, most health care is heavily subsidized by governments and the principle of charity no longer applies (Kohn and White 1976; Maxwell 1981; Roemer and Roemer 1981). To be effective, medical care must be efficacious and access to it must be unimpeded.
Hospital and professional structures have changed to reflect increases in specialized knowledge in medicine. This has led to improvement in technical care but at the expense of holistic care. Conversely, previous shortcomings in relation to the care of the chronically sick are now becoming recognized and are slowly being redressed. For example, geriatric medicine, pioneered in the United Kingdom, has greatly increased insights into the care of the disabled aged (Brocklehurst 1975), which is timely, given the rapid expansion of the elderly population. New concepts for multidisciplinary services have been developed and put into practice in both hospital and community settings, although some remain to be properly evaluated. The hospice movement has drawn attention to the relief of pain and discomfort as a legitimate primary objective in health care, and palliative medicine has emerged as a clearly defined area of specialized care (Hillier 1988). Not only has the scope of medical care been broadened, but access to medical services has been greatly increased.
The relationship between declining mortality in the nineteenth and early twentieth centuries and medical care measures has been extensively reviewed by McKeown (1965, 1979). He provides convincing evidence that improvements in health as measured by falling mortality rates during that time had little to do with personal medical care. This chapter is therefore restricted to the examination of the effects of medical care on health during the last 80 years, the era of ‘scientific medicine’.
That the advances in medical care described above have made an important contribution to health in the second half of the twentieth century would appear to be self-evident to many people. Despite this common view, some writers have seriously questioned the benefits of medicine. The most radical views have been expressed by Illich (1975), who suggested that apart from the direct iatrogenic consequences of some medical treatments, the overall approach of the medical and allied professions to health issues—defined as health ‘problems’—and the ways they are thought about and discussed have broader negative repercussions for both individual patients and society at large. These views are probably shared by only a small minority, but other observers, including McKeown (1979) and Cochrane (1971), have also drawn attention to the fallacies inherent in some beliefs about the contribution of medical care to health. McKeown argued that much of the fall in mortality during the nineteenth century was due to social changes and, therefore, that any further improvements in health seen now should not automatically be attributed to modern medical care. Rose (1985, 1992) made the point that the health of individuals significantly reflects the norms of lifestyle and behaviour of the population in which they live.
Concerned because of escalating cost of health care, and in particular because of the unrestrained introduction of new medical technologies, many other writers have since used McKeown’s arguments in relation to mortality to question the value of modern medicine, even though McKeown (1979) took a much broader view of how the benefits of medical care should be assessed. Since the 1970s, the focus of debate has shifted considerably, so that the concern is now about the marginal cost-effectiveness and cost-benefit of different aspects of medical care. Moreover, 30 years on, the assertion that there is little or no relationship between levels of mortality and the introduction of various medical interventions must be seriously questioned.
Cochrane was more concerned to show that the benefit of much medical care activity is unproven and should be carefully reappraised, partly to ensure that unsafe procedures are discontinued and partly to ensure that resources for medical care are employed in the most effective and efficient ways. In a widely quoted aphorism he asserted: ‘It is surely a great criticism of our profession that we have not organised a critical summary, by specialty or subspecialty, updated periodically, of all randomised controlled trials’ (Cochrane 1979). This challenge has now been accepted and a worldwide network, the Cochrane Collaboration (Chalmers and Haynes 1994), has been established to undertake systematic reviews of the evidence relating to the benefits of all aspects of medical care.
Writers such as Illich, McKeown, Cochrane, and Rose have broadened the focus from the direct benefits of health care to the indirect effects, adverse effects as well as beneficial effects, and social effects in groups as well as biological effects in individual patients.
Health services that are at all effective can have profound effects at the population level. For example, slowing population growth by successful control of fertility may lead to major improvements in the social environment in favour of health. In contrast, declining mortality without an associated reduction in fertility can lead to rapid growth in a population, which, by overtaxing the social and physical environments, is eventually counterproductive as far as health is concerned. Concomitant falls in fertility and mortality, while stabilizing population growth, will also produce the type of demographic transition observed in industrial countries and this, in turn, will produce entirely new patterns of morbidity with relative increases in the dependent aged and reductions in the dependent young.
As far as direct adverse effects are concerned, the rapid development of new methods of treatment has produced many dramatic and salutatory lessons that must caution us against unqualified enthusiasm for the presumed benefits of modern medicine. While thalidomide is perhaps the landmark example, there are unresolved questions pertaining to many other drugs including oral contraceptives, hormone replacement therapy, and calcium-channel blockers, drugs to which many people worldwide are or have been exposed, and to procedures such as electroconvulsive therapy and laparoscopic surgery. Questions about the balance of risks and benefits apply equally to established programmes and to new interventions.
Assessment of the current contribution of medical care to the health of populations
Evaluation of existing systems of medical care is one of five key questions that should inform the planning and delivery of health services to populations, be they a settled and affluent community, refugees inhabiting a makeshift border camp, the population of a school, a gaol or a cruise ship, or a military formation on active duty. The full set of questions is: ‘Who will use the services?’, ‘What health problems do they have?’, ‘What health services already exist?’, ‘How are the existing health services performing?’, and ‘Are there any special features of the population or situation that need to be taken into account?’ These questions are considered here in this order because it is impossible to judge whether health services are performing well and appropriately without knowing the size and nature of the population they serve, the level and pattern of need, and the resources that are available. At a very broad level, such as a developing versus an industrialized country, a given set of general principles for the provision and evaluation of medical care is likely to have wide application, but the wise practitioner will always take care to gather and use intelligence on special features of the local population that should also be taken into account.
Features of the population served
While it seems obvious that the age, sex, and ethnic structure of a population, along with its geographical and occupational distribution, are likely to be fundamental determinants of its level and pattern of need for, and use of, health services, these basic data are too often ignored or surprisingly difficult to find. National census data are frequently available at least to the level of local government area or county, but health services, be they general practices, infant welfare clinics, or hospitals, are not necessarily confined by these administrative boundaries. The principal risk here is not that ill-defined denominators complicate accurate measurement of activity and need in the health system, but that unmet need in subgroups of the population defined by one of the demographic characteristics already mentioned is overlooked. Aggregating data to larger geographic units for which the populations are well described can overcome the former, but is no substitute for careful collection and analysis of more local information (Rissel et al. 1996). An example here is the problem of osteomalacia (rickets) in the relatively small South Asian population of Glasgow, Scotland, recognition of which prompted a successful campaign of vitamin D supplementation (Goel et al. 1981). At the same time, needs assessment based upon measurement of the gap between current experience of (ill) health in a community and some predetermined target level may be a simplistic basis on which to set priorities for the allocation of health resources (Mooney 1998).
What health problems currently exist?
Any rational approach to the assessment of the impact of medical care on health requires that we have basic knowledge about the incidence, prevalence, and natural history of the problems that collectively comprise ill health. Unfortunately, our understanding of these aspects of ill health is often rudimentary. Another of the difficulties of measuring the benefits of medical care lies in the fact that most readily available data relating to health or medical services were collected for other purposes. The types of data generally available for monitoring health and the performance of health services are reviewed elsewhere in this book. Here the main classes of data in relation to their possible use in assessing the contribution of medical care to health are reviewed briefly.
Mortality data have the great advantage that they are collected in a relatively uniform way in different countries and are available for the whole period in which effective medical care has evolved. Such data are therefore of central importance in making broad comparisons of health status between countries and between different periods since 1900. Mortality data can be used to derive other useful indices such as life expectancy and potential years of life lost, but they do have some shortcomings as indicators of the state of health of populations. To begin with, they are subject to variations caused through changes in diagnostic practice and in coding and classification procedures so that, at the level of individual diseases, it is often not possible to provide valid trends over time or to make comparisons between countries. Mortality data also do not directly reflect the need for medical care associated with conditions, such as mental illness, which cause significant morbidity but are infrequent causes of death. Nevertheless, trends in mortality may be a useful surrogate for changes in morbidity as well as demonstrating actual gains in years of life.
The use of mortality data to plan and assess the impact of changes in medical care is subject to special difficulties in the area of confounding. As the general economic situation of a country improves allowing progressively more money per head of population to be invested in health services, so it is likely that the average standard of living, including housing and nutrition, also improves. It then becomes impossible, post factum, to ascribe any improvement in overall mortality rates to increases in total expenditure on medical care because changes in the other likely determinants of general health usually have not been monitored and because their actual effects on health and mortality are indirect and very difficult to measure.
The potential importance of confounding is exemplified by comparisons of trends in perinatal mortality and in mortality rates from ischaemic heart disease in developed countries over the period since the Second World War, for which there is strong evidence for contributions from both social change and medical intervention. The impact of medical care on perinatal mortality and ischaemic heart disease is further discussed below.
Compilation of registers of certain diseases tends to be reviewed as a research function by those with responsibility for delivery of services and as a service activity by agencies that fund biomedical research. Thus funding of registers is a perennial problem facing those who can see their potential, but only one of several challenges that registers present. Other recurring issues include achieving and maintaining complete ascertainment, in order that selection bias does not affect cases that are registered; the trade-off between the volume of data to be collected for each registered case and the completeness of the data; and the need for consistent and accurate application of definitions for each item of information that is recorded. Collection of information about patients for purposes not immediately related to the medical care that they require is subject to increasing and increasingly bureaucratic scrutiny. The involvement of multiple institutions and hence of their ethics committees poses a real threat to the continued use of name-identified population-based registers for the study of particular health problems, and identification of inefficiencies in the health system and of hazards in occupational and general environments as well as in health-care institutions (Jamrozik and Kolybaba 1999).
Once established, registers have many uses, several of which are of direct relevance to the planning and evaluation of medical care. For example, the incidence, severity, and outcome of a particular condition may reflect all of preventive, curative, and rehabilitative programmes. In any case, these parameters are needed for informed planning of health services. Registers may also serve as a source of cases for controlled trials or for detailed studies of the management of a condition or of its causes. It would be wasteful to develop a register corresponding to every aspect of medical care, but priorities for monitoring health problems by this method can be defined. Thus registers are especially useful for monitoring problems that are preventable, to help to ensure that knowledge already available is applied appropriately, problems that are extremely cost-intensive, since marginal increases in the efficiency of care for such conditions should result in savings that are relatively large in absolute terms, and problems that are important causes of disability if not mortality, since they contribute to a significant and ongoing need for care.
Cancer registries exemplify many of these issues, with a legal duty on medical practitioners to notify newly diagnosed cases of malignancies other than non-melanocytic skin cancer now established in many jurisdictions. In a well-developed health system, ascertainment and objective confirmation of relevant cases are usually easy to achieve. In contrast, many registers do not routinely collect or validate information on the stage of the cancer at presentation, despite this being a key datum in clinical decision-making about the management of patients. Because they are very difficult to collect, even fewer registries are able to provide data on the outcome of individual cases, thus complicating routine population-wide evaluation of oncological care.
Intensive community studies such as those in Framingham (Dawber et al. 1951; Dawber 1980), Busselton (Curnow et al. 1969) and North Karelia (Puska et al. 1985), which measure the incidence and prevalence of specific health problems, provide a basis for detailed epidemiological studies of various forms of health-care intervention. However, most studies of this type are necessarily restricted to a narrow range of causes of ill health in small communities and are therefore of limited use in assessing the impact of medical care on a wider front. In addition, as exemplified by the Nurses’ Health Studies (Colditz et al. 1997), the populations and settings in which they are conducted may not be typical of national populations and conditions, adding to the difficulty of extrapolating results concerning medical care, even if it is valid to extrapolate results regarding risk factors and aetiological relationships.
General health surveys
More general sample surveys of illness and disability in the community such as the health component of the General Household Survey in the United Kingdom, the National Health Interview Survey and associated National Health and Nutrition Examination Survey in the United States, and the National Health Surveys in Australia, have provided insight into levels of treated and untreated sickness and disability in the community as well as into illness behaviour. However, the types of questions that can be asked in surveys using non-medical interviewers are limited, with the result that the information obtained may be too general for the purpose of relating the effects of medical care to the health of the community. Changes in the survey instruments over time and variations in response according to characteristics such as age, ethnic group, and economic and educational status, may make interpretation of trends and geographical differences difficult. The problems of using various measures of health status in the United States for this purpose have been extensively reviewed by Wilson and Drury (1984).
Health service utilization data
The universality, historical continuity, and comparability across populations of morbidity data are all limited. Compared with mortality data, the quality of morbidity data is suspect as it is dependent firstly on the quality of medical records, and, secondly, the records themselves tend to be coded and processed at several locations, whereas in most countries the coding of mortality data is relatively centralized and subject to greater quality control. In addition, although hospital morbidity data lend themselves reasonably well to coding using the International Classification of Diseases, which was originally intended for use in classifying deaths, other systems have had to be developed for data collected in general practice, for example, or to distinguish between impairment, disability, and handicap. Translation from one of these systems to another is not always straightforward. Even so, morbidity data have the potential to fill in some of the gaps in mortality data.
The most commonly used measures of morbidity such as hospital statistical collections or data generated from general practice are in fact measures of service utilization rather than measures of the level of illness in the community. Data of this kind may obviously be affected by factors that are quite independent of true need, such as changes in clinical practice, availability of resources, and general factors that affect access to medical care of all types. While measures of utilization are important if they can indicate the extent of diffusion of interventions shown to be efficacious (or not efficacious) in randomized controlled trials, most hospital morbidity data systems record only that contact with the hospital occurred and the nature of the problem, a smaller number record which operative procedures, if any, were performed, and only a minority record information on other aspects of investigation or management of the patient’s condition. Nevertheless, an appropriately designed and maintained hospital morbidity data system may be invaluable for monitoring the incidence of common and important problems such as major injuries and infections, and non-fatal acute myocardial infarction, as well as the extent of uptake of cost-intensive procedures such as organ transplants.
Morbidity data frequently measure episodes of care or occasions of consultation rather than people treated. Record linkage offers one means of overcoming this, providing person-based statistics relating to hospital use, while cross-linkage with mortality data allows for the study of survival following admission to hospital for a selected condition or surgical procedure (Acheson 1967; Gelding et al. 1987; Goldacre 1987). Opportunities for this type of analysis are extremely limited, with record linkage being extensively developed only in Canada, Scotland, and Oxford, and some capability available in Sweden, Denmark, Finland, and Western Australia. Even so, the use of morbidity data as one source of information for disease registers, and as a starting point for ad hoc studies for planning and assessing the effect of medical care, should be kept in mind. In Perth, Australia, for example, linked morbidity and mortality data have been used retrospectively to study trends in incidence and survival in acute coronary disease, parameters which reveal much about the medical care received both by people at risk of developing ischaemic heart disease and by those who have developed symptoms (Hobbs et al. 1984; Martin et al. 1984, 1987a, b).
Data from general practice
The technical problems relating to consistent capture and coding of data from general practice are even greater than those relating to hospital morbidity statistics. It is difficult to recruit general practitioners to continuing studies, and compliance with collection of data is likely to be low. Most studies of illness or treatment using general practice records have therefore been based on selected practices that may not be fully representative. It follows that data from general practice of most value derive from special broad-based but time-limited investigations which have specific goals such as the study of adverse effects of oral contraceptives undertaken by the Royal College of General Practitioners (1974).
Nevertheless data from general practice have other important potential uses in planning, assessing, and improving medical care. Given appropriate records and, in particular, the possibility of defining clearly the population served by a given practice (as is the case in the British system of capitation where, theoretically, all people are registered and only with one practice), it is at least possible to measure the proportion of the target group which has received a particular service. This is an important step on the way to determining the impact on health of that medical care activity. For example, Mak and Straton (1993) have described the establishment and operation of a cervical cytology register based in the single general practice in Fitzroy Crossing in the remote North-West of Australia. The setting is unusual because of the large distances involved—two-thirds of women screened in the first year of operation lived more than 20 km from the township—and because 80 per cent of the population served are Aboriginal Australians, many of whom do not read English. Initial contact and recall for screening of many of the women on the register is therefore verbal and opportunistic. Despite this, 54 per cent of women aged 15 to 69 years in the population were screened in the first year of operation of the register, indicating that the programme was likely to meet its target of screening all eligible women biennially. More importantly, one-quarter of women screened in the first year had never had a Pap smear taken previously. In addition to monitoring coverage of the target group, the quality of smears collected is reviewed regularly. As Mak and Straton (1993) note, the incidence and mortality from cervical cancer is significantly higher in Aboriginal women than in the remainder of the Australian population. Their study shows how responses to such problems can be designed, implemented, and evaluated using simple data collected in general practice.
What health services already exist?
Existing health services can be classified according to target group and function. Thus, there are community-wide services intended to protect and promote health, and personal medical services concerned with prevention, cure, restoration, and palliation. Personal medical services can be further classified according to level, that is, into primary, secondary, and tertiary services. To fulfil its function each component of the health system requires appropriate numbers and types of personnel plus the buildings, equipment, transport, and information and communication systems that together constitute the infrastructure. What constitutes an ‘appropriate’ level of provision of human and other resources presupposes a clear and shared view of an optimal state of health and of the kinds of services needed to achieve and maintain that state. Universal provision of adequate food, water, and sanitation is an agreed goal but one that is yet to be realized in much of the world. In contrast, in the industrialized countries, health protection systems concerned with the food and water supply and with sanitation are frequently taken for granted and their importance only rediscovered when human error, war, natural disaster, or union action leads to interruption of services. Nevertheless, there is more agreement about minimum goals for environmental health than about the optimal numbers of doctors or hospital beds per head of population. The sources of disagreement include the fact that demand for health services is probably boundless and the tensions between philosophical arguments about access, equity, and basic human rights, on the one hand, and vested bureaucratic and professional interests in rationing services or numbers of practitioners, on the other. Given finite resources in the health system, it is legitimate that advocates of new or expanded services are challenged to demonstrate the efficacy and cost-utility or marginal cost-effectiveness of the changes they are proposing. This principle is already well-established in several countries in relation to subsidized provision of pharmaceuticals and is likely to be applied more widely in the health system in the future.
How are the existing health services performing?
Evaluation of the contribution to health of medical services already in place is one of the key issues facing those charged with meeting the role of medicine defined by McKeown. Traditionally, most aspects of medical care have been evaluated and funded on the basis of measures of activity rather than beginning with an assessment of need, making an initial investment, and then determining whether the health of the community in which the service was delivered has improved to the extent expected. The assumption that medical institutions that are busy must be doing good is now being questioned. Several developments have contributed to this change in outlook. Firstly, it has been realized that successful completion of a course of treatment does not necessarily mean a return to the premorbid state and that it is necessary to check that the latter has in fact occurred. This has led to greater emphasis on quantifiable outcomes of health care, and it may be that the increasing attention attracted by controlled clinical trials, with their need to define endpoints unambiguously, has also influenced this trend. A second development, also related to the increasing influence of clinical trials as well as to the writings of Cochrane (1971, 1979) and others, is greater scrutiny of the actual diagnostic and therapeutic strategies employed in medical care and of the extent to which they are supported by good scientific evidence rather than empirical observations or deductions from ‘first principles’. When coupled with questions about value for money, such scrutiny becomes searching indeed. A third factor is probably the rise of epidemiology and, in parallel with this enlarging body of theory and practice, a burgeoning capability to store, process, and analyse data. Previously, the simple measures of what health services actually did were all that it was possible to collect and it was not easy to assess, for example, the impact on the health of the local community of particular units within a large hospital. Finally, a prolonged period during which the governments of many countries were committed to ‘economic rationalism’ meant that public services of all kinds came under intense scrutiny in attempts to improve their efficiency.
It is inevitable that the performance of health services will be subject to ever-closer examination because they constitute such a large part of social expenditure and their cost rises faster than the cost of living. The disciplines of epidemiology and health economics have now developed sophisticated tools for evaluating medical care, but the comprehensive sets of data on health and outcomes of health services that are required for an exhaustive examination are not routinely collected in any community. Many questions might be answered if registers of certain diseases were available but special studies may still have to be undertaken to evaluate other particular aspects of medical care.
Even so, in order to evaluate medical services presently operating, an appropriate conceptual framework which embraces all of the activities of health care and which describes the desired outcomes in relation to each of these must first be established. Secondly, the technical problems involved in measuring the extent to which these benefits are achieved must be considered. This leads to a review of the data that are required and that are currently used to assess the performance of health services because, where the most appropriate resources are not available, evaluation of need for and performance of medical care is often attempted by pressing into service data that are collected routinely but intended for other purposes. This section then concludes with a series of case studies of the evaluation of medical care. The examples are drawn from across the full range of activities in medical care and have been deliberately selected to illustrate the use of the different types of data discussed above. The common thread that links them is the need to evaluate for a whole population the impact of medical care for a health problem which is either common or is associated with serious consequences.
A conceptual framework for the evaluation of medical care
This chapter has followed McKeown’s (1979) classification of the functions of personal health-care services into those that contribute to the prevention, cure, or mitigation of disease and subdivided the latter into measures that aim to restore function and those that are solely palliative in intent. Clearly each of these main categories can be further subdivided down to the level of separate and distinct clinical activities. An illustrative schema is depicted in Table 1. This is by no means exhaustive but it serves to illustrate the complexity of the medical task on the one hand, and the equally complex task of determining suitable measures of outcome for the various components of medical care, on the other. When discussing the contribution of medical care to health it is important to specify both which of the above functions is in question and the nature of the expected benefit. The latter may be a reduction in the incidence or prevalence of specific health problems, reduction in overall mortality or premature loss of life, reduced suffering, or improved quality of life. Quite different measures are required to assess the impact of a preventive programme (measured by the absence of disease), of acute care, as in the management of injuries (not only reduction of mortality but also absence of residual disability), or the management of chronic illness (prolongation of life in some instances and also minimization of pain and restriction of function). To define all of the activities that form part of medical care and their potential contribution to the health of populations is beyond the scope and purpose of this discussion. Nevertheless it should be noted that under models for funding health services such as those which separate owners, funders, purchasers, and providers, this exhaustive approach will be required if there is to be a meaningful attempt to relate funding to specified outcomes. For the present, we will draw upon the framework illustrated in Table 1 for examples that illustrate the range of benefits expected from health-care systems in industrialized countries.
Table 1 A framework for the evaluation of medical care
Problems in assessing the performance of medical care services
Apart from the establishment of an appropriate conceptual framework for considering the benefits of health services, the most fundamental problems are difficulties in defining and measuring need for medical care and in devising suitable measures of its effects.
Firstly, it is obvious that benefits of health care can be determined only if the size and nature of the health problem in question can be clearly defined. In other words, a baseline must be established against which changes in frequency or severity of disease can be measured. Much has been written about the need for more satisfactory and comprehensive measures of health status that could be used to define specific needs for medical care and to determine the impact of various types of care at the population level (Fingerhut et al. 1980; Murnaghan 1981; Hunt et al. 1986; Bergner and Rothman 1987). The volume of this literature is in itself testimony to the difficulties involved and to the fact that answers to the problem remain elusive.
Secondly, while it is possible to examine outcomes of disease in biological terms, it is difficult to measure its personal and social impact. Biological states that can be measured and classified as disease do not necessarily correspond to a subjective experience of illness or to a socially acknowledged sick role (Hunt et al. 1986). As discussed by Butler and Vaile (1984), there are several aspects to social outcomes. For example, are benefits considered in terms of the productive capacity of the individual, the impact of disability on families, or the opportunity costs of expending resources on this type of care rather than on some other service which might yield greater benefit overall to the community? Although the case for particular programmes of preventive and acute care of proven effectiveness might be regarded as beyond doubt, they may not be assets if the opportunity costs are high.
An important issue in assessing health outcomes is the assessment of personal benefit—the extent to which the change produced by medical care is valued by the subject. Here it is essential to recognize the difference between extension of life and its quality. The latter is necessarily a matter of subjective judgement, but whose judgement?
The need for measures to help us ensure that medical care results in better quality of life, if not necessarily a longer life, has been championed by Katz (1987), himself a pioneer of measures of physical function (or activities of daily living) that may be used in assessing outcomes of rehabilitation. The importance of describing function, as opposed to describing disease, was also stressed by a Scientific Group of the World Health Organization (WHO 1984) who advocated that ‘loss of autonomy should be used broadly as an endpoint in epidemiological studies of the relationships among specific diseases, impairments and handicaps’.
In a review of progress in the assessment of quality of life, Spitzer (1987) underlined the difference between generally accepted measures of health status, which may distinguish between the healthy and ostensibly healthy, and measures of quality of life, which are used in situations of established illness and disability. He conceded that some overlap in these approaches may be necessary in the transition zone between the ostensibly healthy and the definitely ill, and also noted the need for a hierarchy of measures extending from the conceptual level, to the operational level, and finally to targets for specific clinical interventions. At the higher level, he suggested that, to be of value, any generic instrument must cover the domains of physical and mental function, emotional and mental state, the burden of symptoms and perception of health, and sense of well being. At the clinical level, there is the additional need for ‘hypothesis-determined functional measures for clinical evaluation’, specific for particular diseases and perhaps applicable to treatments or to different areas of activity, whether these be preventive, therapeutic, or restorative in nature. For example, Fletcher et al. (1987) stressed the need to include appropriate measures of quality of life in randomized controlled trials of drugs used in the treatment of cardiovascular diseases.
Trying to measure the effects on quality of life of many different medical problems and interventions is another challenge. Use of a single method for all situations stands to provide comparability at the cost of insensitivity to subtle effects on quality of life of particular diseases or treatments, while the converse applies to the numerous problem-specific scales. Some widely used generic instruments such as the SF-36 (Short Form Health Survey with 36 items) address several different dimensions of function and well being but do not yield a single measure of utility, thereby complicating the weighting of years of life gained by their quality or the level of ongoing disability.
Another difficulty is that medical care is usually provided by generic services. The objectives of such services are not explicitly defined in terms of prevention, cure, or care and it is therefore difficult to identify, let alone quantify, the components of medical care that are devoted to each of these functions, or, as health economists would wish, to apportion costs to them. Problems are also encountered in moving from a consideration of the benefits of specific treatments for specific diseases to the aggregate benefit of medical services applied to the whole population. Thus it is beyond debate that many diseases have been reduced or eliminated through immunization, that obstetric care has minimized the likelihood of maternal and perinatal mortality or morbidity, that surgery for injuries and particular abdominal emergencies may be life-saving, and that drugs developed in the past 80 years offer the prospect of a cure or substantial relief of symptoms in many illnesses. Yet, the lack of valid measures of outcome and difficulty in controlling for confounding make a formal and overall evaluation of established medical care services impossible. Proxy measures of outcome are therefore frequently used. Arising from the classic paper of Donabedian (1966), measures of ‘structure’, such as numbers of beds per head of population, and of ‘process’, such as numbers of admissions to hospital, have become widely accepted as alternatives to measures of outcome in assessment of the performance of health services.
How measures of structure and process relate to unmet need or to the true contribution of medical care to health is not immediately clear. Ultimately, using such variables to assess medical care is likely to yield misleading conclusions if some nexus between them and the effectiveness of care is not established. But this has rarely been done, and many of the current norms and standards for health services have been determined mainly by historical precedent rather than through explicit rational processes. Uncritical acceptance of measures of structure and process for the assessment of health services carries dangers because of the complacency that it may engender in removing the apparent need for true measures of outcome. Yet, as the preceding discussion indicates, ‘outcome’ itself is a complex notion extending beyond the simple died–survived dichotomy to encompass concepts such as disability and quality of life. How to combine multiple measures of outcome for a single condition, what weight to give to each, and whether the weights are, or should be, different for different populations are all very active areas of research and publication.
Assessing the present performance of medical care services
In evaluating the performance of health services already in place, there is clearly a need to go beyond measures of structure and process to examine direct indicators of health and ill health. However, in so far as they reflect levels of health, many of the statistical systems mentioned above always have some element of lag, making an up-to-the-minute assessment impossible, and it is a truism that data collected today on survival and quality of life in the long term must reflect not current patterns of investigation and treatment but those that were in vogue at the time that the patients first became ill. While cross-sectional surveys of the population and registers of specific diseases have the potential to provide data that are more up to date and can be very useful in exploring fully the levels, patterns, and short-term impact of medical activity, including the extent of application of new treatments, it may be years before the outcome of medical care for particular problems is seen clearly. Evaluation of screening for colorectal cancer is a case in point (Mandel et al. 1993).
Mooney et al. (1986) suggested that, rather than assess the performance and impact of the health system as a whole, or at the other extreme, the benefit from treatment of individual diseases, it is possible to identify and examine specific functions or ‘programmes’ performed by particular sectors of the health-care system as illustrated in Table 1. These may be further divided into subprogrammes each with its own explicit goals relating to particular problems of certain client groups. For example, immunization services might constitute a ‘programme’ with various subprogrammes concerned with delivery of vaccines for particular infectious agents to defined subgroups of the population. The assumption is made that all of the individual medical care activities undertaken by a given programme contribute, either directly or indirectly, to the achievement of its goals—in this case, minimizing morbidity and mortality from vaccine-preventable diseases, and this assumption obviates the requirement to identify the contribution of each individual activity to reaching the goals.
In contrast, Kessner et al. (1973) advocated detailed study of the performance of health services for selected sentinel or tracer conditions and Rutstein et al. (1976) developed an equivalent approach concerning certain avoidable causes of death. Thus, one way to evaluate a specific programme may be to assess carefully the extent to which the goals of one of its constituent subprogrammes are being achieved. The inference here is that if, for example, the incidence of measles in a population is consistently very low and the sporadic cases seen are almost entirely attributable to infections imported from outside, one might conclude not only that the effort to prevent measles is working well, but also that the whole programme to control vaccine-preventable diseases is effective, at least in childhood. It may also be concluded that a wider range of preventive services for infants and children are probably reaching their targets. Clearly the extent to which it is possible to extrapolate in such a fashion will be a matter of judgement informed by local knowledge, but the converse situation is illustrative. If, despite investment of significant financial and human resources, the incidence of measles failed to fall or to stay low in a particular community, one would legitimately have concerns not only about measles and other vaccine-preventable diseases, but also about whether all kinds of preventive services were reaching infants and young children.
In keeping with this approach of starting with a specific service and making judicious inferences about wider programmes, examples are given below of ways in which mortality data and data on morbidity from hospitals, general practices, special registers of diseases and cross-sectional surveys can be used to evaluate particular aspects of medical care.
Data currently used to assess the benefits of medical care
Global approaches based on mortality from all causes combined The work of McKeown (1965, 1979) has demonstrated that the major proportion of the decrease in mortality rates in the United Kingdom since the early nineteenth century was probably due to improvements in nutrition, water supply, and housing, with relatively minor contributions from specific medical interventions, including immunization. On the basis of earlier trends, McKeown suggested that at least some part of the continuing decline in mortality following the development of effective medical care would still have occurred without the advent of scientific medicine. The fact that mortality rates in middle-aged and elderly men demonstrated little improvement during most of the twentieth century, even after effective medical measures became available, has also been cited as evidence that medical care has contributed little to the decline in mortality.
However, as suggested by Levine et al. (1983), further sharp declines in mortality rates in males and females of all ages that have occurred recently in some countries indicate a need to review this interpretation. Firstly, it must be noted that mortality in females in several countries has continued to improve throughout most of the twentieth century. For example, since 1900, expectation of life in Australian females has increased at all ages but the trend has accelerated in the past 40 years. Expectation of life at birth also increased in Australian males, albeit at a slower rate than in females. Between 1932–1934 and 1974, as in other developed countries, there was no improvement in male life expectancy from the age of 60 years. Since then, expectancy of life at all ages in males has increased at a rate comparable with that in females (Australian Bureau of Statistics 1998a,b). Figure 1 and Figure 2 summarize the percentage decline in the risk of dying between selected ages in Australia since the 1930s, that is, since a period generally recognized to predate the introduction of effective medical care.
Fig. 1 Change in mortality in Australia since the 1930s (males). (Data from Australian Bureau of Statistics 1998b).
Fig. 2 Change in mortality in Australia since the 1930s (females). (Data from Australian Bureau of Statistics 1998b).
Evidence from consideration of major causes of death separately Analysis of subgroups of causes of death also provides evidence of the impact of medical care. Morris (1967) has shown that mortality in men in England and Wales would have declined continuously if there had not been increases in deaths from lung cancer and ischaemic heart disease of such proportions as to swamp an underlying downward trend in mortality from other causes. While certain writers have suggested that the apparent increases in lung cancer and ischaemic heart disease were due to statistical artefact (Stehbens 1987), the subsequent marked decline in mortality from ischaemic heart disease in some countries (Epstein and Pisa 1979; Thom et al. 1985; Uemura and Pisa 1985; Dobson 1987), and downturn of mortality from lung cancer in others (Doll and Peto 1981), attest to the epidemic nature of these diseases. Although the fall in mortality from ischaemic heart disease in countries such as Australia, Finland, New Zealand, and the United States, and the more recent appearance of the same phenomenon in most countries in Western Europe (Epstein and Pisa 1979; Uemura and Pisa 1985), are not necessarily related to medical care, by the same argument, the earlier intractability in male mortality rates cannot be used as evidence against benefits from medical care.
Data from Western Australia in Table 2 and Table 3 and Fig. 3(a), show that, apart from recent improvements in mortality from ischaemic heart disease, impressive reductions in mortality rates have occurred in the past 50 years across most broad categories of disease as well as across all age and sex groups. With the exception of deaths due to malignant neoplasms, the mortality rate fell in virtually all major categories of disease subsumed in the chapters of the International Classification of Diseases. In both males and females, the greatest proportion of this improvement by far was attributable to falls in mortality from circulatory diseases (Table 2 and Table 3). However, reductions in mortality rates from injuries, perinatal disorders, genitourinary diseases, disorders of the digestive system, and infectious diseases each made substantial contributions to the decline in males, while in females, perinatal disorders, injuries, respiratory diseases, disorders of the digestive system, and genitourinary diseases contributed much of the remaining improvement. Less frequent causes of death which showed marked relative improvements include diseases of the skin and subcutaneous tissues in both sexes, congenital and musculoskeletal disorders in males, and infectious diseases, disorders of pregnancy, labour and puerperium, endocrine disorders, and diseases of the blood and blood-forming tissues in females.
Table 2 Change in mortality in Western Australia 1953–1957 to 1993–1997 (males)
Table 3 Change in mortality in Western Australia 1953–1957 to 1993–1997 (females)
Fig. 3 (a) Change in mortality in Western Australia 1953–1957 to 1993–1997 (all causes). (b) Change in mortality in Western Australia 1953–1957 to 1993–1997 (respiratory causes). (c) Change in mortality in Western Australia 1953–1957 to 1993–1997 (digestive disorders). (d) Change in mortality in Western Australia 1953–1957 to 1993–1997 (genitourinary diseases). (e) Change in mortality in Western Australia 1953–1957 to 1993–1997 (injuries, poisoning, and violence).
Substantial though these changes may be, trends in overall age-standardized mortality rates obscure the complexity of different rates of change in specific age groups or specific diseases. More detailed examination of the latter is helpful in understanding some of the multiple factors that may be contributing to general improvements in mortality. For example, modest changes in age-standardized total mortality rates for respiratory diseases conceal larger improvements in people under the age of 45 years and rising rates or mixed trends at older ages (Fig. 3(b)). One explanation for these findings is that mortality from acute respiratory infections has declined at younger ages, due at least partly to the use of antibiotics, while mortality due to chronic, smoking-related, and industrial respiratory diseases in older age groups has proved less tractable.
Similarly, with regard to mortality from disorders of the digestive system, the contributions of marked improvements in people aged under 25 have been accompanied by smaller decreases at later ages due to rising mortality from cirrhosis of the liver. Discounting the latter, it is clear that mortality from digestive disorders has fallen at all ages (Fig. 3(c)). As several authors, including McKeown (1979) and Beeson (1980), have suggested, it is likely that this is due principally to advances in medical and surgical care. Further improvements are possible with the recognition by Marshall and Warren (1984) that peptic ulcer is largely an infectious disease which can be treated simply with antibiotic therapy, thereby probably reducing the need for elective surgery and the risks of perforation or life-threatening haemorrhage. Mortality from genitourinary diseases also decreased at all ages (Fig. 3(d)), but presumably for different reasons: under 15 years of age because of a decline in acute nephritis, in adults from 15 to 64 years of age due to improvements in the management of chronic renal failure, and in males aged 65 years and over because of lower mortality associated with obstructive uropathy.
Trends in mortality from injuries (Fig. 3(e)) are more difficult to interpret. Because of the disproportionately heavy impact of vehicular crashes and suicide in younger adults, and to a lesser extent, the middle-aged, improvement in mortality from injuries was greater in children under the age of 15 years and in people over 64 years of age. Reduced mortality from injuries in children may be due to a number of improvements in prevention. Conversely, lower rates in the elderly could reflect improvements in medical care for conditions such as fractures of the proximal femur due to falls, which constitute the largest single cause of admission to hospital for injury in this age group. Since 1971, the age-standardized rates of admission for these fractures in Western Australia have remained unchanged, but in people over the age of 55 years, case fatality at 1 month has fallen from 7.1 to 3.6 per cent in females and from 16.6 to 6.6 per cent in males (Hockey and Hobbs 1990).
In virtually all of the examples discussed above, a range of non-medical as well as medical factors could have contributed to lower mortality rates. Clearly there is no way of quantifying these various effects. Nor is it necessarily true that the changes are replicated elsewhere. Nevertheless, as Kleinman (1982) suggested, there may still be lessons to learn from closer scrutiny of mortality rates for the past 50 years.
Contribution of medical care to reductions in specific causes of death An alternative to relating trends in total mortality or in mortality from major causes of death to medical care is to examine the possible benefits of specific treatments for specific diseases on a case-by-case basis and then to sum the total effects. For example, Reid and Evans (1970) assessed the possible effects of changes in drug therapy on mortality rates of selected non-communicable diseases in the United Kingdom, while Mackenbach and Looman (1988) documented the relationship between changes in the rate of decline in mortality from infectious diseases from 1911 to 1978 and the introduction of antibiotics in the years following the Second World War.
While this approach has not so far been applied over a wide range of diseases, Beeson (1980) used subjective judgement to compare the efficacy of medical treatments for a large range of diseases in 1975 with the impact of measures in use in 1927. Emphasizing also the increased scope and safety of surgery and improved care of severe injuries, including burns, that result from antibacterial drugs and other technical improvements, he concluded that ‘substantial advances have been made along the whole frontier of medical treatment’ and that ‘a patient today is likely to be treated more effectively, to be returned to normal activity and to have a better chance of survival than fifty years ago’. This qualitative assessment falls far short of what we require to know about the effects of medical care on health, but it is an approach that might be furthered by the combination of meta-analysis of the effects of specific medical interventions with good epidemiological data on the incidence, prevalence, and the natural history of the corresponding diseases and on the treatment received by representative samples of relevant patients.
A further example of assessment of specific areas of medical care is that provided by the Reports of Confidential Inquiries into Maternal and Perinatal Deaths in the United Kingdom, which have attempted to distinguish between inevitable deaths and those associated with possibly preventable factors (Department of Health 1998a,b). Rutstein et al. (1976) adopted a similar approach for broader causes of death, on the basis that the frequency of selected, potentially preventable causes of death could be used as a measure of the performance of medical services. With the advice of an expert panel, these authors constructed a list of conditions where, in general, death should not occur if appropriate medical care is given. The original intention was to use the list for the purposes of medical audit. It was recognized, however, that death rates based on aggregated causes of death potentially preventable by medical intervention could be used to monitor the overall performance of the health-care system. Clearly not all causes of deaths will be preventable, and while some authors continue to refer to those that are as ‘preventable’ deaths, others have preferred the term ‘amenable’ deaths. Charlton et al. (1983) in the United Kingdom developed this concept further and examined variations among Area Health Authorities in mortality from 14 conditions selected from Rutstein’s original list and representing 6.5 per cent of all deaths. The list included some conditions such as appendicitis and cholecystitis, where prevention of death confers an all-of-life benefit, and others, such as hypertensive disease and stroke, where intervention may lead only to death being deferred.
This work provided the stimulus for a number of further studies of mortality amenable to medical care, culminating with the publication of atlases of preventable mortality for countries of the European Community (Holland 1988, 1993). These studies have been of two general kinds: those that have attempted to relate changes in amenable mortality over time to the introduction and diffusion of medical interventions; and those that have sought to find a relationship between regional variation in amenable mortality and the provision of health services. Studies in the former category are summarized in Table 4.
Table 4 Average annual decline in amenable and non-amenable mortality in selected countries
The first major exploration of long-term trends in amenable mortality was undertaken by Charlton and Velez (1986) who undertook a comparative study of contemporaneous trends in the United States, the United Kingdom, France, Italy, Japan, and Sweden. Similar studies were undertaken by Pokilanen and Eskola (1986) in Finland, Mackenbach et al. (1988, 1990) in The Netherlands, and Gil and Rathwell (1989) in Spain. More recently the same method has been applied to the European Community as a whole (Carstairs 1993), separately to Ireland (Barry 1992) and Scotland (Carstairs 1993), and independently in New Zealand (Malcolm and Salmond 1993), Quebec (Pampalon 1993), Sweden (Westerling and Smedby 1992), and Greenland (Bjerregaard and Juel 1990). As shown in Table 4, in all of the studies with comparable presentation of data, substantially greater declines have been observed in amenable causes of death compared with remaining causes of death. This was taken by most of the authors to indicate that medical care had contributed to declining mortality, even though the absolute benefits could not be separated from those due to other non-medical factors.
While the relative changes have been considerable, absolute effects on mortality are also an important consideration. In the studies included in Table 4, amenable causes of death initially accounted for about 15 to 20 per cent of total deaths but subsequently fell to about 11 per cent in European countries, a useful but only a modest improvement in terms of mortality from all causes. Moreover, most studies have been restricted to people under the age of 65 years.
Conversely, studies of trends in amenable mortality by Mackenbach et al. (1988, 1990) extended the selection of causes of death (to 35, compared with 14 on Rutstein’s original list) and included deaths in people up to the age of 75 years. Amenable deaths, including deaths in infants, accounted for 42 per cent of all deaths, or 35 per cent if deaths in infants were excluded. More importantly, Mackenbach and coworkers sought to relate variation in the rate of change in particular groups of deaths to the period in which specific and relevant medical interventions were introduced. In general, the findings consistently support a relationship between medical care and lower mortality. The results strengthen considerably the case for an appreciable effect of medical care on declining mortality in developed countries in recent decades.
In contrast to these studies of national trends, those that have attempted to relate regional variation in amenable deaths to intranational variations in health services have been inconclusive. Charlton et al. (1983) in the United Kingdom found wide variation between health areas in causes of amenable mortality that persisted after adjustment for social factors. They suggested that this should be examined further in relation to levels of provision of health services, an approach disputed by Carr-Hill et al. (1987). Mackenbach et al. (1989) also found marked regional variation in amenable mortality in The Netherlands, but the relationship between this and variation in the supply of health services was weak and inconsistent. Similarly, the European Atlas of Preventable Mortality reveals differences between countries that persist after taking into account a number of social indicators associated with variation in deaths from all causes and in non-preventable mortality. These persisting differences are too large in absolute terms to be credibly attributed to differences in standards of medical care.
A number of further caveats are necessary in relation to the interpretation of declining amenable mortality. The first is that, as already intimated, the extent of change is likely to be influenced by the selection of both the particular causes of death and the age range considered. For example, hypertensive diseases and stroke tend to dominate trends in amenable deaths, even though these two conditions have not necessarily shown the greatest relative declines. From data in the first and second editions of the European Community Atlas of ‘Avoidable Death’ (Holland 1988, 1993), total mortality declined by 7.8 per cent, mortality from hypertension and cerebrovascular disease declined by 20.9 per cent, while mortality from all other amenable causes of death declined by over 30 per cent. However, hypertensive diseases and stroke accounted for over 80 per cent of amenable deaths in both periods and contributed nearly three-quarters of the decline in mortality from amenable causes. All non-amenable causes declined by 4 per cent.
The second consideration is that mortality from at least some of the amenable causes of death was falling and may have continued to decline without the benefits of medical care. Conversely, some deaths that were previously not preventable may now have become amenable to medical intervention. Thus the list of causes of amenable mortality will need to be modified over time as effective interventions become available for conditions not previously amenable to medical care. For example, the second edition of the European Community Atlas of ‘Avoidable Death’ (Holland 1993) recognized that medical intervention may have some effect on mortality from both ischaemic heart disease and cancer of the breast. While only a part of mortality from these conditions is likely to be prevented by currently available interventions, in absolute terms these effects could overshadow the gains relating to other, less common, amenable causes of death. The possible savings in deaths from these two conditions are discussed below.
A third question relates to the need to distinguish between deaths that indicate the standard of medical care and those, such as deaths caused by smoking, that reflect broader national health policies concerned with prevention. The example of smoking also highlights the issue of lag time. Even if a totally effective method of helping people to stop smoking were to become universally available today, deaths caused by smoking would continue to occur in former smokers for at least another 20 years (Doll and Peto 1981; Peto et al. 1992).
Finally, one should distinguish between reduced mortality from acute conditions that provide benefits for the whole of life, compared with deferment of death from chronic illnesses which may be associated with changes in the prevalence of people with disability. For example, aggressive resuscitation may save the life of a premature baby but the patient may never function normally. It is thus not only the reduction in mortality that is important; the quality of the years of life gained must also be taken into account.
Trends in mortality from ischaemic heart disease In the studies described above, stroke was regarded as a cause of amenable mortality but ischaemic heart disease was not considered as such until the second edition of the European Community Atlas of ‘Avoidable Death’ (Holland 1993). With hindsight, it can now be appreciated that mortality from ischaemic heart disease is not only potentially amenable to various forms of medical treatment, but also that the magnitude of this benefit is potentially so large as to overshadow the improvements in mortality from other causes. For example, we have shown in Table 2 and Table 3 that, between 1953 and 1993, mortality from circulatory diseases in Western Australia declined by over 50 per cent in both males and females, and accounted for nearly two-thirds of the total improvement in mortality in that period. It is now well documented that mortality from ischaemic heart disease began to fall in the United States, Canada, Australia, and New Zealand from the late 1960s and has since more than halved. A similar decline was not observed in Western Europe until at least a decade later, while mortality from ischaemic heart disease is still rising in some Eastern European countries (Epstein and Pisa 1979; Thom et al. 1985; Uemura and Pisa 1985; Dobson 1987; Tunstall-Pedoe et al. 1999).
The reasons for this fall in mortality from ischaemic heart disease are unclear, but there is evidence to suggest that it has been associated with both a decline in the incidence of acute ischaemic events (Goldberg et al. 1987; Elveback et al. 1981; Martin et al. 1984; Stewart et al. 1984; Tunstall-Pedoe et al. 1999), and improved short- and long-term survival following acute myocardial infarction (Gomez-Marin et al. 1987; Thompson et al. 1988; Hopper et al. 1989; Hammar et al. 1992). The first of these associations suggests that there has been an improvement in the risk factors for ischaemic heart disease. The second observation is consistent with an effect from medical treatment. However, these explanations are not independent, as medical treatment in people with hypertension or subacute ischaemic heart disease could pre-empt or defer the development of acute ischaemic events, while improved survival following myocardial infarction could be due to a change in the natural history of the disease associated with an improvement in risk factors in the general population.
Thus, some authors such as Dwyer and Hetzel (1980), Marmot (1985), and Dobson (1987) have linked the decline in mortality from ischaemic heart disease to changes in lifestyle factors such as diet and smoking or to changes in treatment of hypertension, while others have attributed at least part of the decline to direct treatment of ischaemic heart disease. Goldman and Cook (1984), for instance, suggested that 60 per cent of the decline was due to changes in lifestyle and the remainder to medical care. Beaglehole (1986), using similar methods, estimated that 42 per cent of the decline in ischaemic heart disease in New Zealand could be attributed to medical care.
It is clear from randomized controlled trials that medical and surgical treatment can reduce mortality from ischaemic heart disease. What is uncertain is the impact that such innovations in treatment have had on mortality at the level of whole populations, or the extent to which the recent improvements in mortality from ischaemic heart disease can be attributed to medical care.
Studies based on disease registers
Obstetric and neonatal care The major objectives of modern obstetric and neonatal practice include the reduction of perinatal loss and of the prevalence at birth of major, disabling defects. Whereas 40 years ago both of these problems appeared to be intractable, perinatal mortality has improved dramatically and there are now credible prospects for substantial reductions in the prevalence of birth defects or their consequences. In both instances, a range of interventions, including primary prevention, screening, and medical and surgical treatment are involved. The study of these problems requires the maintenance of registers for the systematic recording of birth defects and to integrate information relating to maternal and infant health during pregnancy, labour, and at least the perinatal period.
Although antenatal care is usually thought of as serving a purely preventive function, in practice it operates as a screening system to identify pregnancies at high risk of poor fetal or maternal outcome in order that the ‘full force’ of the obstetric services can be brought to bear early, before irreversible problems develop. Most obstetric activity is therefore fundamentally curative in intent, in that it seeks to recognize and terminate established disease processes rather than to prevent them from ever starting.
Historically, standards of obstetric and neonatal care, together with levels of reproductive fitness, have been assessed by monitoring maternal and perinatal mortality. Maternal mortality is now so low in most developed countries that it makes a negligible contribution to total mortality in females and is of doubtful value as a measure of outcome in obstetric care (Department of Health 1998b). Through the Confidential Inquiries into Maternal Mortality (Department of Health 1998a, b) it is apparent that most of the deaths that still occur under the general rubric of ‘complications of childbirth, pregnancy, and the puerperium’ are related more to social and maternal factors or to abortion, than to obstetric care per se. There is much evidence to suggest that the virtual elimination of maternal mortality has been due to factors such as improved maternal fitness, stature, and nutrition in early life, together with better education, control of fertility, and, in many countries, more liberal attitudes to termination of pregnancy with a subsequent reduction in illegal abortions. But there is equally strong evidence, again from the Confidential Inquiries into Maternal Mortality, that certain avoidable causes of maternal death have been eliminated through the disciplined application of modern obstetric and anaesthetic techniques.
With the marked decline in maternal mortality, attention has turned to the assessment of preventable factors in perinatal deaths. The relatively high frequency of these compared with maternal deaths dictates the need to focus attention on deaths which are likely, a priori, to be associated with potentially avoidable factors. For example, the Perinatal Mortality Committee in Western Australia has excluded perinatal deaths associated with lethal malformations and concentrated its attention on singleton stillbirths weighing 2000 g or more and neonatal deaths of infants weighing at least 1500 g at birth, and, for multiple births, stillbirths weighing 1500 g or more and neonatal deaths of infants weighing 1000 g or more (Health Department of Western Australia 1987).
In 1986 approximately 20 per cent of perinatal and infant deaths met the above criteria for full investigation. Of these cases 14.8 per cent, representing 3.4 per cent of total perinatal deaths, were associated with potentially avoidable factors. Thus it appears likely that in most developed countries, perinatal mortality, like maternal mortality previously, has declined to something approaching an irremediable minimum.
However, concepts of preventability must clearly be considered in the context of the knowledge and medical practices of the day. Thus it is now evident that a sizeable proportion of cases of anencephaly could be prevented by women taking supplements of folic acid during the periconceptual period (Bower et al. 1998). Furthermore, the birth-weight limits for preventability now used are considerably lower than those used by Baird et al. (1954) in their studies of perinatal mortality in Aberdeen in the 1950s and may need to be reduced still further in the future. This change is emphasized particularly by the accelerated decline in neonatal mortality that has occurred in developed countries in the past two decades in association with the development of neonatal intensive care and, more recently, intrauterine and intrapartum fetal monitoring.
In Western Australia, perinatal mortality fell from 28 per 1000 births in 1970–1975 to 9 per 1000 in 1987–1991 (Stanley and Waddell 1985; Gee 1988, 1992). The improvement occurred in all components of perinatal mortality. The effect was most marked initially in the case of intrapartum stillbirths and neonatal deaths, with subsequent marked improvement in antepartum stillbirths (Table 5). Approximately 20 per cent of the fall in neonatal mortality after 1970 can be attributed to improvement in the distribution of birth weights but, as shown in Table 6, the greatest part is due to improved survival in lower-birth-weight babies, during a period in which there have been marked changes in neonatal intensive care. Similarly, the decline in both antepartum and intrapartum stillbirths has occurred in association with new methods for monitoring intrauterine growth in late pregnancy and fetal distress during labour.
Table 5 Changes in stillbirths, neonatal, and postneonatal infant mortality rates in Western Australia (1970–1991)
Table 6 Changes in birth weight specific neonatal mortality rates per 1000 live births in Western Australia (excluding lethal malformations) from between 1971 and 1974 to 1987 and 1991
A concern relating to the development of neonatal intensive care and consequent survival of babies with very low birth weight who would previously have died, is the potential for increases in long-term morbidity and disability, resulting particularly from cerebral damage. Stanley and Watson (1992) reviewed the incidence of cerebral palsy in Western Australia over the period of falling neonatal mortality described above. They found no increase in the birth-weight-specific incidence of cerebral palsy, but observed an overall increase in cases of approximately 5 per cent due to the combined effects of the relatively high frequency of cerebral palsy in low-birth-weight babies and their improved survival.
In summary, maternal mortality has become obsolete as a measure of obstetric outcome in developed countries, probably due to a combination of general improvements in maternal health and better medical care. Furthermore, perinatal mortality is now also approaching a level where only very large studies will have any chance of detecting additional improvements, and the contribution of improved obstetric and paediatric care to this trend would appear to be beyond question. However, there is some evidence that this reduction in mortality has been associated with an overall increase in the prevalence of long-term disability in children of low birth weight, underlining the necessity for taking into account a number of different measures of outcome, including those relating to adverse as well as beneficial effects.
Birth defects The prospects for the prevention of birth defects have been reviewed by Czeizel et al. (1993). In a comprehensive analysis based on the prevalence of individual birth defects (as recorded in the Hungarian Congenital Abnormality Registry) and the effectiveness of a range of interventions relevant to each, they estimated that 70 per cent of birth defects could be prevented or successfully treated. The figure fell to 60 per cent if congenital dislocation of the hip, which has an unusually high prevalence in Hungary, was excluded. The authors state that their estimates are probably conservative as they do not allow for the potential benefit of improved ultrasonography applied routinely during pregnancy. Conversely, frequent ultrasound examinations during pregnancy do not lead to improved obstetric outcomes (Newnham et al. 1993). Nevertheless, the interventions considered by Czeizel et al. (1993) included avoidance of pregnancy as the result of genetic counselling, avoidance of teratogens (particularly rubella), termination of pregnancy in association with estimation of serum a-fetoprotein, human chorionic gonadotrophin, and oestriol, amniocentesis, and ultrasonography, and medical treatment and early neonatal surgery for remediable conditions detected by neonatal screening. They did not include periconceptional supplementation of the maternal diet with folate.
The convergence of several of the above measures is seen in the prevention of neural tube defects. Clinical trials have now clearly established a substantial protective effect of dietary supplementation with folate in preventing both recurrent and first occurrences of neural tube defect, if folate is given before conception and early in pregnancy (Medical Research Council 1991; Czeizel and Dudas 1992; Lumley et al. 1999); the benefits of routine screening of maternal serum a-fetoprotein during pregnancy have been well established in some communities; and there is evidence of increasing termination of affected fetuses as the result of wider use of ultrasonography in pregnancy and growing community acceptance of this type of intervention. In the United Kingdom, there were dramatic declines in the birth prevalence of both anencephaly and spina bifida between 1970 and 1987 (Fig. 4), during the period in which screening programmes became generally available, although it was not possible from available data to determine the precise proportion of the decline that could be attributed to screening and termination of pregnancy (Wald and Cuckle 1992). The prevalence has stabilized at 8 per 10 000 live and stillbirths during the 1990s, down from 34 per 10 000 two decades previously (Murphy et al. 1996). Similar improvements occurred in South Australia (Chan et al. 1993), where a population-based screening programme has been in operation since 1982. The proportion of women screened increased from 52 per cent in 1983 to 81 per cent in 1986–1991, while the proportion of affected pregnancies detected increased progressively to 90 per cent in 1990–1991. Termination of affected pregnancies also rose from low levels in 1978–1979 to 80 per cent in 1990–1991. In contrast, in the States of Western Australia, Victoria, and Tasmania, where there was no such programme (Bower et al. 1993a, b), there was a slow increase in the termination of affected pregnancies from less than 10 per cent in 1980–1982 to 40 per cent in 1989. Throughout this period, the overall frequency of neural tube defects (including terminations as well as completed pregnancies) in all States remained unchanged at approximately 20 per 10 000, but it has since fallen by about one-third in Western Australia (Bower et al. 1997, 1998).
Fig. 4 Birth prevalence of anencephaly and spina bifida in England and Wales from 1964 to 1968 (data from the Office of Population Censuses and Surveys adjusted for estimated under-reporting). (Source: Wald and Cuckle 1992.)
Control of infectious disease by immunization Modern developments in immunization have seen the global eradication of smallpox and the virtual elimination of poliomyelitis in some populations. Promising results have been achieved with both measles and rubella immunization, although both present problems in relation to vaccine failure and hence optimal vaccination outines. An essential component of global programmes for the control and eradication of communicable disease is the maintenance of information systems which record both levels of immunization in the target populations and the frequency of major disease endpoints, such as death in the case of measles and tetanus, flaccid paralysis in the case of poliomyelitis, birth defects in the case of rubella, and primary hepatoma in the case of hepatitis B.
In reviewing the WHO Expanded Program on Immunization, Kim-Farley (1992) asserted that at current levels of immunization, with the proportion children receiving three doses of diphtheria–pertussis–tetanus (DPT) and polio vaccine estimated as 80 per cent in 1990, 3.2 million deaths from measles and 450 000 cases of paralytic poliomyelitis are prevented annually. Nevertheless, there are an estimated 1.2 million deaths from measles and 120 000 cases of paralytic poliomyelitis that might be saved if full immunization is achieved. Among the targets of the programme were the reduction, by 1995, of 90 per cent of deaths from measles compared with levels prior to the immunization programme, and the global eradication of poliomyelitis by 2000. In contrast to the favourable estimates of coverage by polio and DPT vaccines, only 38 per cent of newborn infants are protected against neonatal tetanus by maternal immunization in countries where this disease is a public health problem. The programme nevertheless had a target for the elimination of neonatal tetanus by 1995. To support achievement of these goals, the programme promoted further improvement in structures for primary care, improved immunization schedules, the introduction of more effective vaccines, and further improvements in surveillance systems.
While worldwide eradication of measles and rubella is not yet a realistic goal, the efficacy of immunization in terms of the major sequelae of infection can be judged. For example, Stanley et al. (1986) demonstrated that the inception of a universal immunization programme for young teenage girls is the most likely explanation for an observed decline in the frequency of congenital rubella syndrome in Western Australia. Vaccination of schoolgirls aged 12 or 13 years began in 1971 (when the proportion immunized was 75 per cent), and acceptance had increased to 88 per cent by 1983. The observed ‘background’ frequency of congenital rubella in Western Australia was approximately 20 per 100 000 live births from 1968 onwards, but rose sharply in association with epidemics of rubella in 1970 and 1974 (Fig. 5). At the latter time, the first girls to have participated in the programme would have been 15 years old. Over the subsequent years, the frequency of congenital rubella syndrome in Western Australia fell to approximately 5 per 100 000 live births (Condon and Bower 1993; Bower et al. 1993a, b) and only a single case was observed in the 3 years from 1995 to 1997 (Bower et al. 1998). In 1990–1991, 95 per cent of women attending the major obstetric hospital in Western Australia demonstrated serological immunity to rubella, with over 97 per cent of Australian-born women seropositive. There is therefore strong presumptive evidence that a specific medical care activity has virtually eliminated one source of lifelong and at times life-threatening disability in the target community. However, the examples provided by the rubella vaccination programme in Western Australia and the measles eradication campaign in the United States (Mitchell and Balfour 1985) are relatively straightforward compared with the problems encountered in assessing the impact of more complex medical care activities where the relationship between the intervention and the outcome is less specific.
Fig. 5 Birth prevalence of congenital rubella syndrome in Western Australia 1968 to 1992. (Sources: Stanley et al. 1986; Bower et al. 1993a.)
Screening mammography Records of breast cancer from a population-based cancer registry can be used to predict the likely impact of screening mammography and therefore can be used to determine whether mammography is having a potentially useful impact on the pattern of breast cancer at diagnosis in that population. Table 7 presents data on breast cancer in Western Australia (total population 1.9 million). In 1989, when 9 per cent of the 584 invasive breast cancers were first detected by mammography (Jamrozik et al. 2000), 36.1 per cent of tumours were stage I at diagnosis, that is, the tumour itself was confined to the breast tissue and did not involve the overlying skin or underlying muscle, there was apparently no involvement of lymph nodes and no distant metastases. Such tumours are potentially curable and are almost always treatable by breast-conserving surgery rather than mastectomy. The average size of all invasive tumours diagnosed in 1989 was 22.7 mm.
Table 7 Potential impact of screening mammography in Western Australia
Universal mammographic screening has been progressively introduced in Australia since 1989 but, under current policies, it is offered only to women aged 50 to 69 years, with some self-referral of other women and selective screening of women of other ages who are at high risk of developing breast cancer. The third line of the table shows that if all women in the target age group were screened and all breast cancers detected by mammography or presenting clinically in that age group were stage I cases, the overall proportion of these tumours in 1994 would have increased to 64.7 per cent and the average size of all tumours would have fallen to 18.3 mm. A more realistic scenario is that mammography might have the same effect on size and stage but that only 70 per cent of eligible women undergo screening. In this case, 52.2 per cent of tumours diagnosed in 1994 would be stage I and the average size of all tumours would be 19.9 mm. In fact, in 1994 55.1 per cent of all cases were stage I at detection, and the average size of tumours had fallen to 20.0 mm (Jamrozik et al. 2000).
Thus, increasing access to and participation in mammographic screening is reflected in an increasing proportion of all malignant breast tumours detected by this method and in a reduction in the average size and clinical stage of the tumours at presentation. In addition the proportion of tumours that are malignant but not yet invasive (in situ ductal carcinomas) should also increase. Progress in reaching such intermediate endpoints has been monitored in Western Australia from data collected by a population-based cancer register covering the community in which mammography is being introduced. Confirmation of the appearance of such trends provides important evidence that the benefits on mortality from breast cancer of mammographic screening demonstrated previously in randomized controlled trials are likely to be shared by the target community. Conversely, failure to see such changes in the data from the register would raise important questions about the implementation of the screening programme.
Uptake of b-blockers, fibrinolytic therapy, and antiplatelet agents in myocardial infarction; changes in case-fatality in patients admitted to coronary care units Over the period since 1980, the single greatest change in the epidemiology of ischaemic heart disease has been the change in the pharmacological management of patients with acute myocardial infarction while they are in hospital. Using a population-based register of major coronary events compiled as part of the WHO MONICA Project (Monitoring Trends and Determinants in Cardiovascular Disease), Thompson et al. (1992) have tracked this revolution in Perth, Western Australia. Furthermore, they have applied the results of previous randomized placebo-controlled trials of three classes of drugs—b-adrenergic blockers, antiplatelet agents, and thrombolytic agents—to their data on the use of such drugs in the population under study to estimate the impact on coronary mortality expected from the earlier experiments.
As may be seen from Fig. 6, the period from 1984 to 1990, inclusive, witnessed major increases in the use of each of the three classes of agent, while the use of calcium-channel blockers, which have not been tested as extensively as the other drugs in relatively unselected cases of suspected acute myocardial infarction, first increased slightly but then fell significantly.
Fig. 6 Trends in the use of b-blockers, calcium-channel antagonists, antiplatelet agents (predominantly aspirin), and streptokinase in the hospital treatment of acute myocardial infarction. Vertical bars represent 95 per cent confidence intervals. (Source: Thompson et al. 1992.)
Thompson et al. (1992) estimate that changes in the use of thrombolytic agents, aspirin, and b-blockers should have reduced early case-fatality after acute myocardial infarction in their population by 7.6 per cent, 17.4 per cent, and 5.2 per cent respectively. Since these agents exert their influences via different and independent mechanisms, one potentially can add these figures and suggest that case-fatality should have improved by 30 per cent. However, this figure will only apply to the proportion of patients who survive long enough after the onset of acute coronary symptoms to receive one or more of the treatments. With approximately two-thirds of coronary fatalities occurring before the patient reaches hospital, and further deaths occurring very soon after arrival at hospital, the impact of the major changes in pharmacological management of acute myocardial infarction will be ‘diluted’ considerably. Even so, this conclusion itself has implications for improving health services further since it throws into perspective the importance of early recognition and early presentation for treatment of patients with acute coronary symptoms.
Studies based on community surveys
Population prevalence of impairments, disabilities, and handicaps Periodic population surveys of disability and handicap are undertaken by the Australian Bureau of Statistics (1982, 1990) using the conceptual framework embodied in the International Classification of Impairments, Disabilities and Handicaps (WHO 1980). Under this classification, a disability is defined as ‘a restriction or lack of ability to perform an activity normally, for example, less than full use of arms and fingers’, while a handicap is defined as ‘the cultural, economic and social consequences of disability, for example, needing assistance getting dressed’. Changes in the prevalence of disability and handicap revealed in the surveys of 1981 and 1988 (which used identical questions) have been reviewed by McCallum (1990). The overall level of reported disability and handicap rose by 31 and 49 per cent, respectively. These changes were more marked in the age group 65 to 84 years, with disabilities increasing by 52 per cent and handicaps by 72 per cent. The principal causes of disability in both surveys were circulatory disease, musculoskeletal diseases, loss of hearing, loss of vision, and mental disorders. The age- and sex-specific changes in prevalence of these disorders are reproduced in Fig. 7. Increases in prevalence of disability related to circulatory disease, loss of hearing, and musculoskeletal diseases occurred in both sexes and in most age groups. The prevalence of loss of vision increased in people over 75 years, but was generally less in people under 75 years. Changes in the prevalence of mental disorders were variable. As McCallum states, it is uncertain whether the overall increases in the prevalence of disability are due to increases in relevant risk factors, or whether they are due to deferment of death.
Fig. 7 Changes in prevalence ratios of disabling conditions in Australia, 1981 to 1988, in people 45 years and over. (Source: McCallum 1990.)
The increase in disability related to circulatory disease is in marked contrast to the dramatic decline in mortality from these diseases in Australia since the late 1960s. It is possible that increased diagnosis of various forms of circulatory disease, and hence greater awareness of such diseases in the respondents, could have contributed to the increase in reported prevalence of disability. However, the observation is also consistent with improved long-term survival in people with cardiovascular disease as noted previously. The extent to which these changes are actually associated with handicap or impaired quality of life is uncertain, but Czarn et al. (1992) have reported that only a minority of survivors of working age have paid employment following acute myocardial infarction. These findings illustrate the need for careful studies to determine the extent to which gains in years of life from medical care are offset by a reduction in the quality of life.
Detection and treatment of hypertension In common with New Zealand, Canada, and the United States, mortality from ischaemic heart disease has been falling in Australia since the late 1960s (National Heart Foundation of Australia 1998). Since a change in the incidence of acute myocardial infarction due to a change in the prevalence or severity of risk factors could be contributing to the downwards trend in mortality, whether detection and management of individuals with avoidable risk factors has improved is of considerable interest. In the Australian setting, the best set of data in which to examine trends in risk factors has been compiled by a private charity, the National Heart Foundation. The Foundation has conducted major surveys of risk factors in residents of the capital city of each of the six states in 1980, 1983 and 1989. These reveal significant decreases in mean blood pressure in both sexes as well as a significant fall in the prevalence of hypertension (Bennett and Magnus 1994). The latter has been accompanied by significant decreases in the prevalence of both previously undocumented hypertension, indicating improved detection, and inadequately treated hypertension, along with potentially important increases in the prevalence of hypertension that is treated and controlled (Table 8). Thus, these surveys not only reveal trends in mean blood pressure and the prevalence of hypertension that would contribute to a continuation of the fall in mortality from ischaemic heart disease but they also provide important data indicating the impact of previous programmes of professional and public education.
Table 8 Trends in the prevalence of hypertension in Australia in people aged 25–64 years (1980–1989)
Studies based on hospital utilization data
Orthopaedic and ophthalmic surgery in the aged Since 1970 greatly improved anaesthesia and the development of prosthetic devices have led to a wide range of surgical procedures for the prevention or correction of disability and handicap. These include improved methods for internal fixation of fractures, arthroplasty, ophthalmic procedures, valvular surgery, arterial grafting, cardiac pacing, and organ transplantation.
Two of the principal objectives in care of the aged that can be readily related to quality of life are the maintenance of mobility and of independence in activities of daily living. Particular conditions that threaten these and are amenable to surgery include fractures of the proximal femur, osteoarthrosis in the hip, and impairment of vision.
Total hip replacement is undertaken in osteoarthrosis, for both relief of pain and restoration of mobility, and following some fractures of the proximal femur. From both clinical and economic points of view the procedure has been judged to be highly effective (Attenborough 1977; Harries and Sledge 1990). As a consequence, demand for total hip replacement continues to increase and far exceeds previous predictions based on earlier studies of the prevalence of osteoarthrosis (Williams et al. 1994). For example in the North West Thames Region of the United Kingdom, rates for hip arthroplasty in people over the age of 65 years increased by approximately 50 per cent between 1978 and 1984 (Rajaratnam et al. 1990), while in Western Australia, rates increased slowly from 1972 to 1986 and then rose steeply by approximately 40 per cent by 1992. They have since stabilized but, with a growing population in late middle-age and old age, this still represents an increase in the absolute numbers of operations being performed (Fig. 8). Some measure of the proportion of elderly people who may ultimately benefit from this procedure can be judged from studies conducted in the United States in 1986 (Sharkness et al. 1993) and in the United Kingdom in 1991 (Williams et al. 1994) of the prevalence of people who have received the operation, as summarized in Table 9. In the former study the prevalence of people having undergone the procedure was 1.1 per cent in those aged 65 to 74 years and 2.4 per cent in those aged 75 years and over. In the study in the United Kingdom the equivalent proportions were 3.5 and 7.5 per cent. Comparative studies of rates for surgical procedures (Holman and Brooks 1987; Rutkow and Starfield 1984; Schacht and Pemberton 1985) generally show that the rates for any given operation in the United States at least equal if not exceed those in the United Kingdom. As rates for hip replacement are still rising, it would appear that a significant proportion of people over the age of 75 years, possibly of the order of 10 per cent, might eventually benefit from this procedure. On this basis, arthroplasty of the hip may be judged to play an important role in maintaining mobility and independence in old age.
Fig. 8 Trends in hip replacement surgery in Western Australia 1972–1998.
Table 9 Prevalence (%) of hip replacement (unilateral or bilateral) in the United States and the United Kingdom
Surgical treatment of cataract provides a further example of an elective procedure in which continuing improvements in techniques and in prostheses have been accompanied by continuing rises in the rate of replacement procedures. Previously, the frequency of cataract operations was tempered by concerns for intraocular complications and problems in some patients with visual adjustment following operation (Jaffe 1978). Such difficulties have been greatly reduced by refinements in surgical technique and by the development, in the early 1980s, of flexible prosthetic lenses that are inserted directly into the lens envelope. Improved postoperative function and the ability to perform the operation under local anaesthesia as a day case have led to a more liberal attitude to cataract surgery, with the majority of procedures in many centres now being performed under local anaesthesia in day hospitals. Thus, having been relatively stable from 1974 to 1980, rates for lens surgery in Western Australia have increased by a factor of almost 4 between 1980 and 1998 with no sign that they will stabilize (Fig. 9). To some extent this increase may have been brought about by earlier intervention, as it is no longer considered necessary to wait for cataracts to ‘mature’ or for vision to deteriorate in both eyes as was frequently the case before the era of lens replacement. At the anecdotal level, impairment of vision sufficient to result in the loss of a driving licence is commonly the motivation for lens surgery.
Fig. 9 Trends in surgery for cataract in Western Australia 1972–1998.
Restoration of mobility and vision in the elderly are clearly matters of considerable importance. While the data shown here fall mainly into the category of process measures, it is of interest to note the general reduction in disability due to loss of vision in people under the age of 65 years observed between 1981 and 1988 in the Australian Surveys of Impairments, Disabilities and Handicaps reported above. In future surveys of this nature, attempts should be made to elicit the cumulative incidence of common restorative procedures to assess their possible impact on the prevalence of the specific disabilities for which they are generally performed.
Palliative care Palliative care may be defined broadly as the alleviation of symptoms of disease without curing the disease, which is undoubtedly the longest standing function of medicine. Indeed much of medical care is still concerned with relief of symptoms of common, incurable diseases or disabilities, whether or not these are associated with life-threatening illness. Thus the term ‘palliation’ may relate to the medical response to conditions that range from highly prevalent ‘nuisance’ disorders, such as headache or allergic rhinitis, to pain or loss of function with less common but more disabling conditions such as rheumatoid arthritis, or pain, loss of function, depression, and anxiety associated with terminal illness such as cancer, and to the distress and disruption to the family associated with these problems. Through detailed studies, medical care can be shown to provide very significant relief from many disorders individually, but it is difficult to find an appropriate index of benefit that can be applied to the whole range of such problems as they affect entire communities.
It is ironic, therefore, that palliative care, as one of the newest specialties in medicine (Hillier 1988), has had to be redefined. This stems from the increase in the prevalence of chronic or intractable illness on the one hand, and, on the other, the inability of generic services, which have an essentially curative focus and philosophy, to deal adequately with the dying. As Higginson (1993) has noted in a review of past changes and future trends in palliative care, modern hospices emerged in response to poor treatment given by conventional providers of care. Thus there is now general acceptance of Doyle’s (1987) definition of palliative care as ‘the total (physical, emotional and spiritual) care of patients with life-threatening disease and their families. The focus is both the quality of the remaining life of the patient and the support of families and friends’.
Higginson (1993) also suggests that while palliative care was concerned formerly with the care of patients with terminal illness, this has now broadened to include the care of those who have life-threatening illnesses but are not imminently dying. This change in emphasis has important implications for the selection of patients for palliative care services. When the concern previously was for patients with terminal illness, the majority of patients were those dying from cancer. If services are to be concerned with patients from the point of diagnosis, as proposed by the WHO Expert Committee (1990), the spectrum of conditions accepted for palliative care services will be much broadened and the volume of patients supported by such services at any point in time, dramatically increased. The Expert Committee recognized that this would require significant redeployment of resources. The potential scale of such a development inevitably raises one of the long-standing issues surrounding palliative care—whether this extended role should be accepted by palliative care services as they now exist, or whether the palliative care function should be ‘reintegrated’ into generic services through, for example, the establishment of palliative care support teams (Higginson 1993) and appropriate modifications of the curricula of schools of medicine and nursing.
Studies of the cost of medical care at the end of life, principally in the United States, have been reviewed by Scitovsky and Capron (1986). These authors concluded that medical care in the last year of life accounts for 20 to 30 per cent of total expenditure on health care. The question then arises as to whether this expenditure represents unjustified use of technology and other costly measures. One argument for the further development of hospice and palliative care services is that they provide cost-effective alternatives to conventional care. However, in reviewing hospice care in the United States, Torrens (1985) noted that such services are frequently costly because of the intensive support that they provide. He concluded that evidence that hospice care is cost-effective compared with conventional care was equivocal. Conversely, Higginson (1993), when reviewing experience in the United Kingdom, concluded that ‘hospice and specialist home care services are at least as effective as and probably more acceptable and efficient than conventional care’. It is noted, however, that this conclusion is based on relatively small studies, which do not allow us to determine the true extent of community benefit provided by palliative care services.
These contrary conclusions point, perhaps, as much to the difficulty of assessing the benefits of palliative care, and lack of evidence based on controlled trials, as to its true value as an activity of medical care. There are clearly major conceptual, technical, and ethical difficulties in measuring both the physical and social dimensions of the effects of chronic and terminal diseases and the changes that might occur as the result of medical intervention. The importance of dealing with this problem is now well recognized and much progress has been made, as discussed for example, in the proceedings of the Portugal Conference on Measuring Quality of Life in Clinical and Epidemiological Research (Katz 1987; Spitzer 1987).
Special problems in particular populations
Insightful provision and planning of health services (or any other services) for a whole community requires a thorough knowledge of that community. A straightforward area of difficulty concerns languages and literacy. Access to medical care and to information about health will be greatly impeded if a sizeable proportion of the community speaks a different language from the dominant one or is unable to read signs and printed information. Addressing differences in spoken and written language requires professional interpreters and that staff understand and appreciate that use of interpreters of convenience, such as members of a patient’s family or of the hospital’s support staff, may actually compound problems of communication through such individuals being perceived as inappropriate by the patient.
Challenges relating to differences in culture can be both more subtle and more far-reaching than those related to language. Apart from dietary rules, the most obvious examples of differences in cultural norms influencing medical care are those relating to the roles of the two sexes and rules regarding the sex of the practitioner whom each may consult during illness or for preventive care. However, all of the major events of life—birth, marriage, reproduction, retirement, death, and bereavement—are likely to have attached cultural prescriptions which, if ignored, may add to distress or even lead to services not being used. Refugees may experience the sharpest differences of language and culture, and they tend to make heavy demands on health services because of the combination of a lack of material and social resources, the psychological consequences of an often sudden and unplanned exodus from their homes, and the physical and mental trauma associated with the events that precipitated their flight.
Human characteristics aside, certain populations are dominated by extremes of distance or weather that inevitably shape both provision of and access to health services. The continental United States and Australia are not dissimilar in size, for example, but the one has cities scattered throughout its interior while virtually all sophisticated medical services are located on the coast of the other. Thus, provision of free mammographic screening for breast cancer in Australia has necessitated the development of mobile services that go to the women at risk rather than expect women to come to them.
In addition to patterns of place and person, time is a factor that should be borne in mind when planning health services. Worldwide, the absolute and proportionate numbers of elderly people are growing very rapidly, with enormous potential consequences in terms of need for curative and restorative health services for chronic non-communicable diseases. The impact of HIV disease is yet to be fully felt, tuberculosis is once again a growing problem, and the morbidity and premature mortality associated with smoking will continue to increase for at least another generation (Peto et al. 1992). These trends have timescales longer than the lifetimes of most governments, but the inevitability of their impact on health and health services obliges planners and administrators to obtain the best available information on both the current position of a given community with regard to these problems and the likely future developments.
Assessment of proposed new services
A systematic approach to the evaluation of new services, both during their development and following their general introduction, holds the promise of avoiding many of the difficulties now faced by those trying to establish whether aspects of routine medical care that are already in place truly are effective. Here Cochrane (1971) was a leading advocate of the need for randomized trials of new medical strategies, while Sackett et al. (1991) have emphasized the difference between efficacy—whether a strategy works under ideal conditions—and effectiveness—whether it works under the conditions of everyday practice.
Assessment of efficacy
Many decisions that need to be made in running health services relate not to the overall impact of the ‘health endeavour’ but to the adoption or discontinuation of single, narrowly-defined strategies for preventing or dealing with particular problems. In this situation, evidence from good clinical experiments is more helpful than any other, not least because it is usually the most valid. This has been recognized by the United States Preventive Services Task Force (1996) whose ‘hierarchy of evidence’, originally developed by its Canadian equivalent, is now widely used. In answering any given question about medical care, greatest weight (‘level I evidence’) is accorded to concordant results from at least two very large trials or to a systematic overview or ‘meta-analysis’ of all relevant randomized controlled trials. Level II evidence corresponds to evidence from a single well-conducted trial of adequate size. Rigorous observational investigations such as case–control or cohort studies can provide level III evidence at best, while case series and clinical experience can only ever contribute level IV evidence.
Randomized controlled trials
Despite randomized controlled trials having become so common and influential, certain of the fundamental precepts on which they are based are still not widely understood. It is therefore worth reiterating some of the key principles relating to such studies here even though there are now many textbooks dedicated entirely to the design, execution, and analysis of these experiments. Randomization itself is the most basic of these principles as, subject to the play of chance, the process of allocating participants to treatments at random should balance known and unanticipated confounding variables across the study groups, leaving exposure to different interventions as the only systematic difference between them. Thus, systematic differences in outcome may logically be ascribed to differences in the way that the participants were managed rather than to some other factor. Less widely understood is that most randomized controlled trials are testing policies, that is, they are designed to answer questions of the kind: ‘All other things being equal, should my policy in a particular situation be to employ strategy A or B?’ The method of the controlled trial can therefore be applied to diagnostic and screening activities as well as to therapeutic strategies, to interventions other than drugs, and to whole communities as well as to individuals. A third key principle is analysing the results of trials on the basis of ‘intention to treat’. In practice, trial protocols almost always make allowance for the clinical imperative, meaning that some participants never receive the treatment to which they were allocated and others begin on the allocated treatment but subsequently cross over to the other treatment or withdraw from the study altogether because the clinical circumstances of the particular individual change. Nevertheless, the least biased assessment of the question of policy that the trial was designed to answer comes from including all recruited patients in the analysis, regardless of their later withdrawal or cross-over, and including them in the study group to which they were originally allocated. This intention-to-treat approach avoids introducing selection bias (since participants who withdraw or cross over are not random subsamples of all participants allocated to particular treatments) and is conservative in that it helps to avoid reaching an incorrect conclusion that one strategy under test is superior to the other.
From this discussion it can easily be seen that properly conducted trials account for all patients who are ever recruited to them. As Sackett et al. (1991) point out, they should also report adverse as well as positive effects of each strategy under test, and distinguish between clinical and statistical significance (since a sufficiently large study will render the smallest of differences in outcome significant at the 5 per cent level). To be useful, reports of trials need also to describe the participants carefully in order that the reader can judge whether the population that took part in the experiment is sufficiently similar to his or her own to be reasonably confident that the local population would experience the same outcome from the same intervention. Finally, the treatment or intervention itself must be described in adequate detail for it to be replicated by someone who wishes to introduce that strategy in his or her own practice.
The medical and public health literature is laden with otherwise well-designed and well-conducted trials that have failed to demonstrate clear differences between the strategies under test. Not infrequently this occurs because ‘miracle cures’ are unlikely to be found in populations that already enjoy sophisticated health services, and therefore any advances in care are likely to be small. However, for common problems such as cardiovascular disease, such small increments in prevention or treatment are potentially very important because they ‘translate’ into many episodes of illness avoided or lives saved. A good trial may detect such a clinically important effect but fail to declare it as statistically significant because the trial is too small and thus lacks statistical power. Common reasons for such situations include overestimation, in the design phase of the study, of both the likely benefits of a given strategy and the rates of recruitment, and greater than anticipated cross-over during the conduct of the trial itself.
The technique of systematic overview, sometimes called meta-analysis, was developed to bring together all randomized controlled trials addressing a given question. It provides a summary figure of what they reveal, in aggregate, regarding two strategies for dealing with the same clinical problem. The statistical validity of the technique has been the subject of some debate, but there is general agreement that it rests on reasonable assumptions when the true clinical effects are small. Like all statistical techniques, meta-analysis needs to be used with care. Overviews based on analyses of data for individual patients participating in relevant trials are generally held to be preferable to those that combine summary data for groups of participants but, ideally, either type of review is followed by one or two major controlled trials that are sufficiently large to provide unambiguous confirmation or refutation of its findings. Failing that, well-conducted overviews are becoming increasingly influential as guides to which strategies for improving health are efficacious, and the number of such overviews appearing in the literature is climbing sharply as the Cochrane Collaboration gathers momentum.
The second International Study of Infarct Survival (ISIS-2) trial of fibrinolytic therapy and aspirin in the management of suspected acute myocardial infarction (ISIS-2 Collaborative Group 1988) is an early example of the sequence of numerous small trials followed by a systematic overview and then a definitive large study. The trial employed a two-by-two factorial design to test the effects of streptokinase and aspirin, separately and in combination, on short-term (35-day) case-fatality after acute myocardial infarction. Systematic overviews of data from the previous smaller studies had indicated that each treatment was effective (Yusuf et al. 1985; Antiplatelet Trialists Collaboration 1988) but, with some 17 000 patients enrolled, ISIS-2 was several times larger than any of the earlier trials and its results were convincingly unambiguous. In turn, ISIS-2 has been followed by a further meta-analysis of the effects of thrombolytic therapy in suspected myocardial infarction (Fibrinolytic Therapy Trialists’ Collaborative Group 1994). The new overview has been very helpful in clarifying answers to important clinical questions about the efficacy of thrombolytic therapy when a patient presents relatively late after the onset of symptoms with consequent delay in the inception of treatment. The appropriate use of aspirin and related agents also continues to be explored via meta-analyses (Anti-Platelet Trialists’ Collaboration 1994).
The technique of systematic overview has now been applied in a wide variety of fields. For example, the Early reast Cancer Trialists’ Collaborative Group (1992) has published an overview of the results of randomized controlled trials of adjuvant treatment for potentially curable breast cancer that began recruiting patients before mid-1985. Four principal modalities of treatment were examined—hormonal therapy with tamoxifen, ovarian ablation, chemotherapy, and immunotherapy—with the smallest analysis, that for ovarian ablation with or without chemotherapy, still including 1817 women. In total the overview covered 75 000 women, 42 per cent of whom had died or experienced a recurrence of breast cancer by the time that the data were analysed. This total was estimated to account for 90 per cent of women ever recruited to a randomized controlled trial in early breast cancer, but it can also be put in perspective by considering that a general medical oncologist working on a mixed service might see a maximum of 100 new patients with breast cancer each year, or 3000 in a working lifetime. Thus, the overview can be equated with the clinical experience that might be accrued in 25 professional working lifetimes. While the treatment of individual patients should always be influenced by the particular clinical circumstances, as a guide to general policy in the management of patients with early breast cancer, the evidence provided by such overviews must be regarded as compelling. As more data and longer follow-up accrue and further cycles of analysis are completed, clear answers to other questions of central importance in the management of breast cancer are being obtained (EBCTCG 1998).
Major overviews of obstetric and neonatal practice have also been completed and published (Chalmers et al. 1985; Enkin et al. 1989).
Increasingly, these efforts to bring together and review all available evidence are being translated into preparation and publication of ‘guidelines’ for clinical practitioners, sometimes accompanied by lay equivalents. For example, in 1995, the National Health and Medical Research Council in Australia released a pair of publications relating to the management of early breast cancer (NHMRC 1995a,b). It is perhaps insufficiently understood, however, that being based on randomized controlled trials, level I and II evidence is derived from patients for whom there was clinical equipoise. Thus, ‘guidelines’ founded on such evidence should not necessarily inform the decisions about cases where there are compelling clinical reasons to adopt or avoid one of the possible strategies available for management of the patient’s problem.
Evaluation of the effectiveness of new strategies
It is one thing to demonstrate, via one or more large randomized controlled trials or via a systematic overview, that a particular strategy for the prevention, investigation, or management of a certain condition is efficacious, and another to show that it is effective under the conditions of everyday practice. There are many reasons why a strategy that works well in a ‘clinical laboratory’ can be a failure in the field. Failure to maintain ‘the cold chain’ leading to worse than expected results from early immunization programmes in tropical developing countries is an example familiar to many. Good practice in clinical trials, therefore, is that a study that successfully demonstrates efficacy is followed by another that is designed to confirm effectiveness. Even then, assuming that the second study is also successful, there remains the challenge of ensuring that the findings become widely known and adopted, and that the anticipated benefits do flow through to the relevant population. On occasion and particularly when a specific therapeutic agent is shown to be harmful, the problem can be solved easily, in this case, by prompt withdrawal of the product from sale. Tracking the adoption of beneficial advances and the extent to which their potential benefits are realized is no less important but frequently is more difficult.
Diffusion of the innovation
There is a large sociological literature on the ways that new ideas and techniques spread through a population, and a smaller but growing number of papers on the diffusion of innovation in medical practice. For some operations and drugs, uptake of the new approach can be monitored via routinely collected statistics such as hospital morbidity data or records of subsidized prescriptions. However, pitfalls also abound in using such systems for this purpose. Official coding systems for surgical procedures are updated infrequently, leading to old rubrics being adapted to accommodate new operations, making trends difficult to follow, and to different parts of the health system solving the same ‘problem’ in different ways, making data from different institutions or regions difficult to compare. For example, because the procedure did not have a unique rubric in the manuals used to code surgical procedures, Mattes et al. (1997) had to search the casebooks of radiology departments in several different hospitals in Western Australia to track trends in the use of thrombolysis in acute peripheral arterial obstruction. While a particular operation may be specific to a certain medical condition, there are frequently multiple indications to prescribe the same pharmaceutical, leading to confusion as to the exact reason or reasons underlying changes in sales figures. A case in point here is the extension of the permitted indications for prescription of H2-receptor blocking agents, known as H2-antagonists, from peptic ulcer only to include gastric reflux. Coupled with the wide indications for the use of antibiotics, this change prevents Australian data on the sales of histamine antagonists being used to monitor the replacement of antisecretory therapy for peptic ulcer by treatment intended to eradicate Helicobacter pylori. Similarly, the multiple indications for use of each of b-adrenergic blockers and calcium-channel antagonists mean that aggregate data on sales of these agents cannot be used to monitor changes in the management of hypertension. In the absence of routinely compiled data of information concerning the indications for using a particular strategy, one is obliged to study trends in the uptake of new, efficacious treatments via serial cross-sectional surveys or special registers.
Impact of the innovation
Data showing increased use of a particular strategy only provide circumstantial evidence that the population is experiencing the benefits of the new approach; the definitive proof comes with evidence that the pattern of the relevant health outcome has changed as far and as fast as one would expect given the extent and speed of diffusion of the innovation and its efficacy or effectiveness as indicated by previous controlled trials. Ideally, this evidence is gathered in the form of data on exposure to the strategy and the eventual outcome in the same individual. Failing this, ecological comparisons of condition-specific use of the approach and an appropriate, and preferably unique, outcome may provide indirect evidence. An example of the latter method is the link between trends in Western Australia in use of thrombolytic therapy and angioplasty for peripheral arterial disease, on the one hand, and falling rates of amputation of the lower limb, on the other (Mattes et al. 1997). Here, the two new treatments have unique codes when they are used for peripheral vascular disease and amputations for vascular insufficiency are readily distinguished in routine hospital morbidity data from those following trauma via inspection of a field denoting the presence or absence of an external cause for the patient’s problem.
Compared with the sophistication of systematic overviews and large multicentre double-blind placebo-controlled trials, using ecological studies to assess the spread and impact of new strategies in medical care is a very imperfect science, although it does have an important role in generating new questions and hypotheses for further controlled studies (Rigg et al. 1999). Demonstrating that medical care is effective as well as efficacious is an essential activity at the very core of accounting for the massive public investment that the health system represents. As initiatives such as the Cochrane Collaboration develop and their influence expands, energy and resources will have to be found to maintain and extend collections of data in which evidence can be sought that new developments in care are being followed by improvements in health.
The impact of personal medical services on the health and survival of individuals seems readily apparent. With modern investigations and treatments, patients are now regularly saved and make very good recoveries from infections, injuries, and a variety of other conditions that were almost uniformly fatal even a few years ago. Surprisingly, it is more difficult to demonstrate conclusively the impact of these medical advances on the health of whole communities. Health is more than the avoidance of death, yet mortality statistics are our principal tool for the assessment of historical trends in disease and its management. Not only are mortality statistics subject to artefacts arising from changes in classification and in coding practice, but trends in them are confounded by general improvements in the overall standard of living. Nevertheless, mortality rates for most developed countries show increasing life expectancy with decreases (and certain increases) in particular causes of death that are so large as to seem unlikely to be due to artefacts.
Whatever the impact on overall health of past developments in medical care, the pressing matters are to evaluate existing health services, to identify and withdraw those that are not effective, and to redeploy the resources that they were consuming to implement new programmes of proven worth. This must be built upon a foundation of detailed knowledge and understanding of the population for which services are being provided, the health problems that that population faces and the resources that one has available for mounting a response to those problems. As many medical strategies have never been subjected to controlled trials, evaluating those presently in use frequently requires judicious inferences to be drawn from routinely collected data on morbidity and mortality, careful analysis of information from special registers, and sometimes special surveys of health problems, disability, and quality of life. This approach is undeniably piecemeal but it is also true that health services tend to evolve via adjustments at the margins rather than by full-scale revolutions.
Looking to the future, there is the promise that the imperfect pragmatism of today will be replaced by much more rigorous science tomorrow. The randomized controlled trial has become widely accepted as the method for establishing the efficacy of particular medical strategies, whether they be the choice between investment in prevention now or treatment later, between expectant observation or immediate and active investigation of some clinical problem, or between surgical or pharmaceutical management of the patient’s complaint or some combination of these and other modalities of treatment. Demonstrated efficacy is not the end of the matter, however, as what may work well under the selected and tightly controlled conditions of a research centre may be a failure in everyday practice. Controlled trials establishing efficacy therefore must be complemented by others confirming effectiveness and, in turn, these must be followed by further studies to check that the new strategies have been taken up and applied appropriately and that the expected benefits, in terms of mortality, morbidity, or quality of life, are being seen. All elements of this sequence need to be put in place if we are to be confident that medical services are indeed improving health.
Acheson, E.D. (1967). Medical record linkage. Nuffield Provincial Hospsitals Trust and Oxford University Press, London.
Antiplatelet Trialists Collaboration (1988). Secondary prevention of vascular disease by prolonged antiplatelet treatment. British Medical Journal, 296, 320–31.
Antiplatelet Trialists’ Collaboration (1994). Collaborative overview of randomised trials of antiplatelet therapy–I: Prevention of death, myocardial infarction, and stroke by prolonged antiplatelet therapy in various categories of patients. British Medical Journal, 308, 81–106.
Attenborough, C.G. (1977). Arthritis in the elderly. In Geriatric orthopaedics (ed. M.B. Devas), p. 77. Academic Press, London.
Australian Bureau of Census and Statistics (1982). Handicapped Australia 1981. Catalogue no. 4343. Australian Bureau of Statistics. Australian Government Publishing Services, Canberra.
Australian Bureau of Census and Statistics (1990). Disability and handicap in Australia 1988. Catalogue no. 4120. Australian Bureau of Statistics. Australian Government Publishing Services, Canberra.
Australian Bureau of Statistics (1998a). The Australian year book. Catalogue number 1301. Australian Bureau of Statistics, Canberra.
Australian Bureau of Statistics (1998b). Life tables. Deaths in Australia. Annual reports from 1978. Catalogue number 3302. Australian Bureau of Statistics, Canberra.
Baird, D., Walker, J., and Thomsen, A.M. (1954). The causes and prevention of still births and first week deaths. Journal of Obstetrics and Gynaecology of the British Empire, 61, 433.
Barry, J. (1992). Avoidable mortality as an index of health care outcome: results from the European Community Atlas of ‘Avoidable Death’. Irish Journal of Medicine, 161, 490–2.
Beaglehole, R. (1986). Medical management and the decline in mortality from coronary heart disease. British Medical Journal, 292, 33.
Beeson, P.B. (1980). Changes in medical therapy during the past half century. Medicine, 49, 79.
Bennett, S.A. and Magnus, P.M. (1994). Trends in cardiovascular risk factors in Australia. Medical Journal of Australia, 161, 519–27.
Bergner, M. and Rothman, R.L. (1987). Health status measures: an overview and guide for selection. Annual Review of Public Health, 8, 191.
Bjerregaard, P. and Juel, K. (1990). Avoidable deaths in Greenland 1968–1985: variations by region and period. Arctic Medical Research, 49, 119–27.
Bower, C., Forbes, R., Rudy, E., Evan, A., and Stanley, E. (1993a). Report of the birth defects register of Western Australia, 1980–92. Health Department of Western Australia, Perth.
Bower, C., Raymond, M., Lumley, J., and Bury, G. (1993b). Trends in neural tube defects, 1980–89. Medical Journal of Australia, 158, 152–4.
Bower, C., Rudy, E., Ryan, A., and Grace, L. (1997). Report of the Birth Defects Registry of Western Australia 1980–1996. King Edward Memorial Hospital, Perth.
Bower, C., Ryan, A., Rudy, E., and Grace, L. (1998). Report of the Birth Defects Registry of Western Australia 1980–1997. King Edward Memorial Hospital, Perth.
Brocklehurst, J.C. (ed.) (1975). Geriatric care in advanced societies. MTP Press, Lancaster.
Burnet, M. and White, D.O. (1972). Natural history of infectious disease. Cambridge University Press.
Butler, J.R. and Vaile, S.B. (1984). Health and health services. Routledge and Kegan Paul, London.
Carr-Hill, R.A., Hardman, G.E., and Russell, I.T. (1987). Variations in avoidable mortality and variations in health care resources. Lancet, i, 789–92.
Carstairs, V. (1993). Avoidable deaths in countries of the European Community and in Scotland. Health Bulletin, 51, 151–7.
Chalmers, I. and Haynes, B. (1994). The Cochrane Collaboration (editorial). British Medical Journal, 309, 969–70.
Chalmers, I., Enkin, M., and Keirse, M.J.N.C. (1985). Effective care in pregnany and childbirth. Oxford University Press.
Chan, A., Robertson, E.E., Haan, E., Keane, R.J., and Ranieri, E. (1993). Prevalence of neural tube defects in South Australia, 1966–91: effectiveness and impact of prenatal diagnosis. British Medical Journal, 307, 703–6.
Charlton, J.R.H. and Velez, R. (1986). Some international comparisons of mortality amenable to medical intervention. British Medical Journal, 292, 295–301.
Charlton, J.R.H., Hartley, R.M., Silver, R., and Holland, W.W. (1983). Geographical variations in mortality from conditions amenable to medical intervention in England and Wales. Lancet, i, 691–6.
Cochrane, A.L. (1971). Efficiency and efficiency: random reflections on health services. Nuffield Provincial Hospitals Trust, London.
Cochrane, A.L. (1979). 1973–1971: a critical review, with particular reference to the medical profession. In Medicines for the year 2000. Office of Health Economics, London.
Colditz, G.A., Manson, J.E., and Hankinson, S.E. (1997). The Nurses’ Health Study: 20-year contribution to the understanding of health among women. Journal of Women’s Health, 6, 49–62.
Condon, R.J. and Bower, C. (1993). Rubella vaccination and congenital rubella syndrome in Western Australia. Medical Journal of Australia, 158, 379–82.
Curnow, D.H., Cullen, K.J., McCall, M.G., Stenhouse, N.S., and Welborn, T.A. (1969). Health and disease in a rural community. Australian Journal of Science, 31, 281.
Czarn, A.O.S., Jamrozik, K., Hobbs, M.S.T., and Thompson, P.L. (1992). Follow-up care after acute myocardial infarction. Medical Journal of Australia, 157, 302–5.
Czeizel, A.E. and Dudas, I. (1992). Prevention of the first occurrence of neural-tube defects by periconceptional vitamin supplementation. New England Journal of Medicine, 327, 1832–5.
Czeizel, A.E., Intody, Z., and Modell, B. (1993). What proportion of congenital abnormalities can be prevented? British Medical Journal, 306, 499–503.
Dawber, T.R. (1980). The Framingham study: the epidemiology of atherosclerotic disease. Harvard University Press, Cambridge.
Dawber, T.R., Gilcin, M.E., and Moore, E.E. (1951). Epidemiological approaches to heart disease: the Framingham study. American Journal of Public Health, 41, 279.
Department of Health (1998a). Why some babies die: report on confidential enquiries into maternal deaths in the United Kingdom, 1994–1996. HMSO, London.
Department of Health (1998b). Confidential enquiries into maternal deaths 1994–1996. HMSO, London.
Dobson, A.J. (1987). Trends in cardiovascular risk factors in Australia, 1986–1983: evidence from prevalence surveys. Community Health Studies, 11, 2.
Doll, R. and Peto, R. (1981). The causes of cancer: quantitative risk of cancer in the United States today. Oxford University Press.
Donabedian, A. (1966). Evaluating the quality of medical care. Millbank Memorial Fund Quarterly, 44, 169.
Doyle, D. (1987). Education and training in palliative care. Journal of Palliative Care, 2, 5–7.
Dwyer, T. and Hetzel, B.S. (1980). A comparison of trends in coronary heart disease in Australia, USA, England and Wales with reference to three major risk factors—hypertension, cigarette smoking and diet. International Journal of Epidemiology, 9, 65.
EBCTCG (Early Breast Cancer Trialists’ Collaborative Group) (1992). Systemic treatment of early breast cancer by hormonal, cytotoxic, or immune therapy. Lancet, 339, 1–15, 71–85.
EBCTCG (Early Breast Cancer Trialists’ Collaborative Group) (1998). Tamoxifen for early breast cancer: an overview of the randomised trials. Lancet, 351, 1451–67.
Elveback, L.R., Connolly, D.C., and Kurland, L. (1981). Coronary heart disease in residents of Rochester, Minnesota. II. Mortality, incidence and survivorship, 1950–1975. Mayo Clinic Proceedings, 56, 665.
Enkin, M., Keirse, M.J.N.C., and Chalmers, I. (1989). A guide to effective care in pregnancy and childbirth. Oxford University Press.
Epstein, E. and Pisa, Z. (1979). International comparisons in ischaemic heart disease mortality. In Proceedings of the conference on the decline in coronary disease mortality (ed. R.J. Havlik and M. Feinleib), NIH Publication No. 79-1610, p. 58. United States Department of Health Education and Welfare, Bethesda, MD.
Fenner, R. (1980). Smallpox and its eradication. In Changing disease patterns and human behaviour (ed. N.E. Stanley and R.A. Joske), p. 215. Academic Press, London.
Fibrinolytic Therapy Trialists’ (FTT) Collaborative Group (1994). Indications for fibrinolytic therapy in suspected acute myocardial infarction: collaborative overview of early mortality and major morbidity results from all randomised trials of more than 1000 patients. Lancet, 343, 311–22.
Fingerhut, L.A., Wilson, R.W., and Feldman, J.J. (1980). Health and disease in the United States. Annual Review of Public Health, 1, 1.
Fletcher, A.E., Hunt, B.M., and Bulpitt, C.J. (1987). Evaluation of quality of life in clinical trials of cardiovascular disease. Journal of Chronic Diseases, 40, 557–66.
Gee, V. (1988). Perinatal statistics in Western Australia. Annual report of the Western Australian Midwives Notification System. Health Department of Western Australia, Perth.
Gee, V. (1992). Perinatal statistics in Western Australia. Annual report of the Western Australian Midwives Notification System. Health Department of Western Australia, Perth.
Gelding, J., Vivian, S.P., Baines, C.J., and Baldwin, J.A. (1987). Analytical methods of time-sequenced linked records. In Textbook of medical record linkage (ed. J.A. Baldwin, E.D. Acheson, and W.J. Graham), p. 55. Oxford University Press.
Gil, L.M. and Rathwell, T. (1989). The effect of health services on mortality: amenable and non-amenable causes in Spain. International Journal of Epidemiology, 18, 652–7.
Goel, K.M., Sweet, E.M., Campbell, S., Attenburrow, A., Logan, R.W., and Arneil, G.C. (1981). Reduced prevalence of rickets in Asian children in Glasgow. Lancet, ii, 405–7.
Goldacre, M.J. (1987). Implications of record linkage for health services management. In Textbook of medical record linkage (ed. J.A. Baldwin, E.D. Acheson, and W.J. Graham), p. 305. Oxford University Press.
Goldberg, R.J., Gore, J.M., Alpert, J.S., and Dalen, J.E. (1986). Recent changes in attack and survival rates of acute myocardial infarction (1975–1981): the Worcester Heart Attack study. Journal of the American Medical Association, 255, 2774.
Goldman, L. and Cook, E.E. (1984). The decline in ischaemic heart disease mortality rates: an analysis of the comparative effects of medical interventions and changes in lifestyle. Annals of Internal Medicine, 101, 825.
Gomez-Marin, O., Folsom, A.R., Kottke, T.E., et al. (1987). Improvement in long-term survival among patients hospitalized with acute myocardial infarction, 1970 to 1980. The Minnesota Heart Survey. New England Journal of Medicine, 316, 1353–9.
Hammar, N., Larsen, F.F., Sandberg, E., Alfredsson, L., and Theorell, T. (1992). Time trends in survival from myocardial infarction in Stockholm County 1976–1984. International Journal of Epidemiology, 21, 1090–6.
Harries, W.H. and Sledge, C.B. (1990). Total hip and total knee replacements (first of two parts). New England Journal of Medicine, 323, 725–8.
Health Department of Western Australia (1987). The seventh annual report of the Perinatal and Infant Mortality Committee of Western Australia and the annual report of the Maternal Mortality Committee for 1986, Statistical Series No. 7. Health Department of Western Australia, Perth.
Higginson, I. (1993). Palliative care: a review of past changes and future trends. Journal of Public Health Medicine, 15, 3–8.
Hillier, R. (1988). Palliative medicine. A new specialty (editorial). British Medical Journal, 297, 874–5.
Hobbs, M.S.T., Hockey, R.L., Martin, C.A., Armstrong, B.K., and Thompson, P.L. (1984). Trends in ischaemic heart disease mortality and morbidity in Perth Statistical Division. Australian and New Zealand Journal of Medicine, 14, 381.
Hockey, R.I. and Hobbs, M.S.T. (1990). Report on femoral neck fractures in Western Australia, 1971–1988. University of Western Australia, Department of Public Health, Perth.
Holland, W.W. (ed.) (1988). European Community atlas of ‘avoidable death’. Commission of the European Communities Health Services Research Series 3. Oxford University Press.
Holland, W.W. (ed.) (1993). European Community atlas of ‘avoidable death’ (2nd edn) Commission of the European Communities Health Services Research Series 6. Oxford University Press.
Holman, C.D.J. and Brooks, B.H. (1987). Surgical procedures in Western Australia. An analysis of surgery type in 1985 and trends in surgical procedure rates 1972–1985. Occasional paper No. 15. Health Department of Western Australia, Perth.
Hopper, J.L., Bhupendra, P., Hunt, D., and Chan, W.W.C. (1989). Improved prognosis since 1969 of myocardial infarction treated in a coronary care unit. British Medical Journal, 299, 892–6.
Hunt, S.J., McEwen, J., and McKenna, S.P. (1986). Measuring health status. Croom Helm, London.
Illich, I. (1975). Medical nemesis: the expropriation of health. Marion Boyars, London.
ISIS-2 Collaborative Study Group (1988). Randomized trial of intravenous streptokinase, oral aspirin, both, or neither among 17 187 cases of suspected acute myocardial infarction. Lancet, ii, 320–31.
Jaffe, N.S. (1978). Cataract surgery—a modern attitude toward a technological explosion. New England Journal of Medicine, 299, 235.
Jamrozik, K. and Kolybaba, M. (1999). Are ethics committees retarding the improvement of healthcare services in Australia? Medical Journal of Australia, 170, 26–8.
Jamrozik, K., Byrne, M.J., Fitzgerald, C., et al. (1993). Breast cancer in Western Australia in 1989: 1. Presentation. Australian and New Zealand Journal of Surgery, 63, 617–23.
Jamrozik, K., Byrne, M.J., Dewar, J.M., et al. (2000). The impact of mammographic screening on breast cancer in Western Australia. Medical Journal of Australia, 172, 203–6.
Katz, S. (1987). The science of quality of life (editorial). Journal of Chronic Diseases, 40, 459– 63.
Kessner, D.M., Kalk, C.E., and Singer, J. (1973). Assessing healing quality: the case for tracers. New England Journal of Medicine, 288, 189.
Kim-Farley, R. and the WHO Expanded Programme on Immunization Team (1992). Global immunization. Annual Review of Public Health, 13, 223–38.
Kleinman, J.C. (1982). The continued vitality of vital statistics. American Journal of Public Health, 72, 125.
Kohn, R. and White, K.L. (ed.) (1976). Health care. An international study. Report of the World Health Organization/International Collaborative Study of Medical Care Utilization, p. 125. Oxford University Press.
Levine, S., Feldman, J.J., and Elinson, J. (1983). Does medical care do any good? In Handbook of health care and the health professions (ed. D. Mechanic), p. 394. Free Press, New York.
Lumley, J., Watson, L., Watson, M., and Bower, C. (1999). Periconceptional supplementation with folate and/or multivitamins to prevent neural tube defects (Cochrane Review). In The Cochrane Library, Issue 1. Update Software, Oxford.
McCallum, J. (1990). Health and the quality of survival in older age. In Australia’s health, 1990. The second biennial report of the Australian Institute of Health. Australian Government Publishing Service, Canberra.
Mackenbach, J.P. and Looman, W.N. (1988). Secular trends of infectious disease mortality in the Netherlands, 1911–1978: quantitative estimates of changes coinciding with the introduction of antibiotics. International Journal of Epidemiology, 17, 618–24.
Mackenbach, J.P., Looman, C.W.N., Kunst, A.E., Habema, J., Dik, E., and van der Maas, P.J. (1988). Post-1950 mortality trends and medical care: gains in life expectancy due to declines in mortality from conditions amenable to medical interventions in the Netherlands. Social Science in Medicine, 27, 889–94.
Mackenbach, J.P, Kunst, A.E., Looman, C.W.N., Habema, J., Dik, E., and van der Maas, R.J. (1989). Regional differences in mortality amenable to medical intervention in The Netherlands: a comparison of four time periods. Journal of Epidemiology and Community Health, 42, 325–32.
Mackenbach, J.R., Bouvier Colle, M.I., and Jougla, E. (1990). Avoidable mortality and health services: a review of aggregate data studies. Journal of Epidemiology and Community Health, 44, 106–11.
McKeown, T. (1965). Medicine in modern society. Allen & Unwin, London.
McKeown, T. (1979). The role of medicine: mirage or nemesis? Nuffield Provincial Hospitals Trust, London.
Mak, D.B. and Straton, J.A.Y. (1993). The Fitzroy Valley Pap Smear Register: cervical screening in a population of Australian Aboriginal women. Medical Journal of Australia, 158, 163–6.
Malcolm, M.S. and Salmond, E. (1993). Trends in amenable mortality in New Zealand 1968 1987. International Journal of Epidemiology, 22, 469–74.
Mandel, J.S., Bond, J.H., Church, T.R., et al. (1993). Reducing mortality from colorectal cancer by screening for fecal occult blood. Minnesota Colon Cancer Control Study. New England Journal of Medicine, 328, 1365–71.
Marmot, M.G. (1985). Interpretation of trends in coronary heart disease mortality. Acta Medica Scandinavica, 701 (Supplement), 58.
Marshall, B.J. and Warren, J.R. (1984). Unidentified curved bacilli in the stomach of patients with gastritis and peptic ulceration. Lancet, i, 1311–15.
Martin, C.A., Hobbs, M.S.T, and Armstrong, B.K. (1984). The fall in mortality from ischaemic heart disease in Australia: has survival after mvocardial infarction improved. Australian and New Zealand Journal of Medicine, 14 (Supplement), 435.
Martin, C.A., Hobbs, M.S.T, and Armstrong, B.K. (1987a). Estimation of myocardial infarction mortality from routinely collected data in Western Australia. Journal of Chronic Diseases, 40, 661.
Martin, C.A., Hobbs, M.S.T, and Armstrong, B.K. (1987b). Identification of non-fatal myocardial infarction from hospital discharge data in Western Australia. Journal of Chronic Diseases, 40, 1111.
Mattes, E., Norman, P.E., and Jamrozik, K. (1997). Falling incidence of amputations for peripheral occlusive arterial disease in Western Australia between 1980 and 1992. European Journal of Vascular and Endovascular Surgery, 13, 14–22.
Maxwell, R.J. (1981). Health and wealth. An international study of health care spending. Lexington Books, Toronto.
Medical Research Council (1991). Vitamin Study Research Group. Prevention of neural tube defects: results of the Medical Research Council Vitamin Study. Lancet, 338, 131–7.
Mitchell, C.D. and Balfour, H.H. (1985). Measles control: so near and yet so far. Progress in Medical Virology, 31, 1.
Mooney, G. (1998). ‘Communitarian claims’ as an ethical basis for allocating health resources. Social Science and Medicine, 47, 1171–80.
Mooney, G.H., Russell, E.M., and Weir, R.D. (1986). Choices for health care. Macmillan, London.
Morris, J.N. (1967). Uses of epidemiology. Churchill Livingstone, Edinburgh.
Murnaghan, J.H. (1981). Health indicators and information systems for the year 2000. Annual Review of Public Health, 2, 299.
Murphy, M., Seagroatt, V., Hey, K., et al. (1996). Neural tube defects 1974–1994. Down but not out. Archives of Disease in Children, Fetal and Neonatal Edition, 75, F133–4.
National Heart Foundation of Australia (1998). Heart facts report—1996. National Heart Foundation, Canberra.
Newnham, J.P., Evans, S.F., Michael, C.A., Stanley, F.J., and Landau, L.I. (1993). Effects of frequent ultrasound during pregnancy: a randomised controlled trial. Lancet, 342, 887–91.
NHMRC (National Health and Medical Research Council) (1995a). Clinical practice guidelines for the management of early breast cancer. Australian Government Publishing Service, Canberra.
NHMRC (National Health and Medical Research Council) (1995b). Early breast cancer: a consumer’s guide. Australian Government Publishing Service, Canberra.
Pampalon, R.T. (1993). Avoidable mortality in Quebec and its regions. Social Science in Medicine, 37, 823–31.
Peto, R., Lopez, A.D., Boreham, J., Thun, M., and Heath, C. Jr (1992). Mortality from tobacco in developed countries: indirect estimation from national vital statistics. Lancet, 339, 1268–78.
Pokolainen, K. and Eskola, J. (1986). The effect of health services on mortality: declines from amenable and non-amenable causes in Finland, 1969–81. Lancet, i, 199–202.
Puska, P., Tuomilehto, J., and Salonen, J.T. (1985). Ten years of the North Karelia project. Acta Medica Scandinavica, 701 (Supplement), 66.
Rajaratnam, G., Black, N.A., and Dalziel, M. (1990). Total hip replacements in the National Health Service: is need being met? Journal of Public Health Medicine, 12, 56–9.
Reid, D.D. and Evans, J.G. (1970). New drugs and changing mortality from non infectious disease in England and Wales. British Medical Bulletin, 3, 191.
Rigg, J.R.A., Jamrozik, K., and Myles, P.S. (1999). Evidence-based methods to improve anaesthesia and intensive care. Current Opinion in Anaesthesiology, 12, 221–7.
Rissel, C., Ward, J., and Sainsbury, P. (1996). An outcomes approach to population health at the local level in NSW: practical problems and potential solutions. Australian Health Review, 19, 23–39.
Roemer, M.T. and Roemer, R.J. (1981). Health care systems and comparative manpower policies. Marcel Dekker, New York.
Rose, G. (1985). Sick individuals and sick populations. International Journal of Epidemiology, 14, 32–8.
Rose, G. (1992). Strategy of preventive medicine. Oxford University Press.
Royal College of General Practitioners (1974). Oral contraceptives and health. Pitman Medical, London.
Rutkow, I.M. and Starfield, B.H. (1984). Surgical decision making and operative rates. Archives of Surgery, 119, 899.
Rutstein, D.D., Berenberg, W., Chalmers, T.C., Child, C.G., Fishman, A.P., and Perrin, E.B. (1976). Measuring the quality of medical care. A clinical method. New England Journal of Medicine, 294, 582.
Sackett, D.I., Haynes, R.B., Guyatt, G.H., and Tugwell, P. (1991). Clinical epidemiology: a basic science for clinical medicine (2nd edn). Little, Brown, Boston, MA.
Schacht, P.J. and Pemberton, A. (1985). What is unnecessary surgery? Who shall decide issues of consumer sovereignty, conflict and self regulation. Social Science and Medicine, 20, 199.
Scitovsky, A.A. and Capron, A.M. (1986). Medical care at the end of life: the interaction of economics and ethics. Annual Review of Public Health, 7, 59.
Sharkness, M., Hamburger, S., Moore, R.M., and Kaczmarek, R.G. (1993). Prevalence of artificial hip implants and use of health services by recipients. Public Health Reports, 108, 70–5.
Spitzer, W.O. (1987). State of science 1986: quality of life and functional status as target variables for research. Journal of Chronic Diseases, 40, 465–71.
Stanley, F.J. and Hartfield, M.J. (1979). Livebirths and perinatal mortility in Western Australia 1976–78. Report of NH and MRC Research Unit in Epidemiology and Preventive Medicine and Department of Health and Medical Services, Western Australia.
Stanley, F.J. and Waddell, V. (1985). Changing patterns of perinatal and infant mortality in Western Australia: implications for prevention. Medical Journal of Australia, 143, 379.
Stanley, F.J. and Watson, L. (1992). Trends in perinatal mortality and cerebral palsy in Western Australia, 1967 to 1985. British Medical Journal, 304, 1658–63.
Stanley, F.J., Sim, M., Wilson, G., and Worthington, S. (1986). The decline in congenital rubella syndrome in Western Australia: an impact of the school girl vaccination programme? American Journal of Public Health, 76, 35.
Stehbens, W.E. (1987). An appraisal of the epidemic rise of coronary heart disease and its decline. Lancet, i, 606.
Stewart, A.W., Beaglehole, R., Fraser, G.E., and Sharpe, D.N. (1984). Trends in survival after myocardial infarction in New Zealand, 1974–81. Lancet, ii, 444.
Thom, T.J., Epstein, F.H., Feldman, J.J., and Leaverton, P.E. (1985). Trends in total morbidity and mortality from heart disease in 26 countries from 1950 to 1978. International Journal of Epidemiology, 14, 510.
Thompson, P.L., Hobbs, M.S.T., and Martin, C.A. (1988). The rise and fall of ischaemic heart disease in Australia. Australian and New Zealand Journal of Medicine, 18, 327.
Thompson, P.L., Parsons, R.W, Jamrozik, K., Hockey, R.L., Hobbs, M.S.T., and Broadhurst, R.J. (1992). Changing patterns of medical treatment in acute myocardial infarction: observations from the Perth MONICA study 1984–1990. Medical Journal of Australia, 157, 87–92.
Torrens, P. (1985). Hospice care: what have we learned? Annual Review of Public Health, 6, 65.
Tunstall-Pedoe, H., Kuulasmaa, K., Mahonen, M., Tolonen, H., Ruokokoski, E., and Amouyel, P. (1999). Contribution of trends in survival and coronary-event rates to changes in coronary heart disease mortality: 10-year results from 37 WHO MONICA project populations. Monitoring trends and determinants in cardiovascular disease. Lancet, 8, 1547–57.
Uemura, K. and Pisa, Z. (1985). Recent trends in cardiovascular disease mortality in 27 industrialized countries. World Health Statistics Quarterly, 38, 142.
US Preventive Services Task Force (1996). Guide to clinical preventive services (2nd edn). Williams and Wilkins, Baltimore, MD.
Wald, N. and Cuckle, H. (1992). Antenatal screening and diagnosis. In Epidemiology and control of neural tube defects (ed. J.M. Elwood, J. Little, and J.H. Elwood), pp. 711–26. Oxford University Press.
Westerling, R. and Smedby, B. (1992). Trends in ‘avoidable’ mortality in Sweden, 1974–1985. Journal of Epidemiology and Community Health, 46, 489–93.
Wilson, R.W. and Drury, T.E. (1984). Interpreting trends in illness and disability: health statistics and health status. Annual Review of Public Health, 5, 83.
WHO (World Health Organization) (1980). International classification of impairments, disabilities, and handicaps. WHO, Geneva.
WHO (World Health Organization) (1984). Uses of epidemiology in aging. Report of a scientific group. Technical report series 706. WHO, Geneva.
WHO (World Health Organization) (1990). Cancer pain relief and palliative care. Report of an expert committee on pain relief. Technical report series 804. WHO, Geneva.
Williams, M.H., Newton, J.N., Frankel, S.J., Braddon, E, Barclay, E., and Gray, J.A.M. (1994). Prevalence of total hip replacement: how much demand has been met? Journal of Epidemiology and Community Health, 48, 188–91.
Yusuf, S., Collins, R., Peto, R., et al. (1985). Intravenous and intracoronary fibrinolytic therapy in acute myocardial infarction: an overview of results on mortality, re-infarction and side effects from 33 randomized controlled trials. European Heart Journal, 6, 556–83.