Does health care save lives? Commentators such as McKeown and Illich, writing in the 1960s, argued that it played very little part, and might even be harmful. However they were writing about a period when health care had relatively little to offer compared to today. Since then, several writers have described often quite substantial improvements in death rates from conditions for which effective interventions have been introduced. But the debate continues, with some arguing that health care is making an increasingly important impact on overall levels of health while others contend that it is in the realm of broader policies, such as education, transport and housing, that we should look to for future advances in health. Inevitably this is to a considerable extent a false dichotomy. Both are important. But how much does health care contribute to population health? One way in which this question has been addressed previously is to look at deaths that should not occur in the presence of effective and timely health care, so-called ‘avoidable’ mortality. However much of this work was undertaken in the 1980s and early 1990s, since when health care has advanced considerably. In addition, ‘avoidable’ deaths were often limited to those before, for example, the age of 65, a figure that seems inappropriately low in the light of life expectancies that are now about 80 in many countries. In this review we have traced the evolution of the concept of ‘avoidable’ mortality from its inception in the 1970s. We have undertaken a detailed methodological critique of this concept, examining questions of attribution, issues relating to comparisons over time and place, and relationships with other indicators of health service provision. To help future researchers we have produced a comprehensive, annotated review of the work that has been undertaken worldwide so far. We note that ‘avoidable’ mortality was never intended to be more than an indicator of potential weaknesses in health care that can then be investigated in more depth. We describe examples of where this approach has been successful, drawing attention to problems that might otherwise have been missed. In contrast, many of the critics of ‘avoidable’ mortality, or more specifically, mortality amenable to health care (amenable mortality), have asked that it do something it was not intended to do, to be a definitive source of evidence of differences in effectiveness of health care. Thus, it is not unexpected that studies seeking to link amenable mortality with health care resources have failed to do so, especially when undertaken within countries, although it is notable that where gross differences exist, as between western and eastern Europe, the gap in amenable mortality is especially high. For these reasons, it seems justifiable to continue and extend the extensive body of research that has already been undertaken to look at ‘avoidable’ mortality, updating the list of conditions included to reflect the changing scope of health care and extending the age limit to reflect increasing expectation of life. However it must be recognised that the concept of ‘avoidable’ mortality does have important limitations, relating to comparability of data, attribution of causes, and coverage of the range of health outcomes. Comparisons of health system performance are now firmly on the international policy agenda, especially since the publication of the 2000 World Health Report. Incorporation of concept of mortality amenable to medical care into the methodology used to generate the rankings of health systems in that report would be an advance on the current situation. We show how, within Europe, this would lead to different rankings from those based on overall disability adjusted life expectancy, which is used in the current rankings. However, any approach based on aggregate data would not address one of the major criticisms of such comparisons, that they do not indicate what needs to be done when faced with evidence of sub-optimal performance. This requires a more detailed analysis of the specific issues facing health systems. Here we propose a new method, in which analyses of amenable mortality identify areas of potential concern that are then examined in more detail by studying the processes and outcomes of care for tracer conditions, selected on the basis of their ability to assess a wide range of health system components. The second part of the review applies the refined method of amenable mortality analysis to patterns of mortality in the countries of the European Union over the past two decades. This shows that deaths that could be prevented by timely and effective care were still relatively common in many countries in 1980. Reductions in these deaths contributed substantially to the overall change in life expectancy between birth and age 75 during the 1980s. The largest contribution was from falling infant mortality but in some countries reductions in deaths among the middle aged was equally or even more important. These countries were Denmark, The Netherlands, the United Kingdom, France (for men) and Sweden (for women). In contrast, during the 1990s, reductions in amenable mortality made a somewhat smaller contribution to improved life expectancy, especially in the northern European countries. However they remained important in southern Europe, especially in Portugal and Greece, where the initial death rates had been higher. These findings provide clear evidence that improvements in access to effective health care have had a measurable impact in many countries during the 1980s and 1990s, in particular through reductions in infant mortality and in deaths among the middle aged and elderly, especially women. However the scale of impact has, to a considerable extent, reflected the starting point. Thus, those countries where infant mortality was relatively high at the beginning of the 1980s, and which had the greatest scope for improvement, such as Greece and Portugal, unsurprisingly saw the greatest reductions in amenable mortality in infancy. In contrast, in countries with infant mortality rates that had already reached very low rates by the beginning of the 1990s, such as Sweden, the scope for further improvement was small. Similarly, the scope for improvement in amenable deaths in adulthood was greatest in those countries where initial rates were highest. The corollary of this is that as rates fall in all countries, the extent of variation decreases. As a consequence, it seems likely that, in the 21st century, the ability to compare health system performance using mortality data at the aggregate level is likely to be limited, simply because the differences will be relatively small. This does not, however, mean that there is not scope for analyses that use amenable mortality rates to screen for potential problems that can then be explored in more depth. It also does not exclude the use of amenable mortality to gain new insights into inequalities in access to care within populations.