The introduction of new pharmaceuticals and treatment methods during the late 20th century, and new and more effective ways of organizing care, mean that health care now contributes demonstrably to improved population health.1,2 However, as national health spending has tended to outstrip economic growth in most OECD countries in recent years, reaching 9% of GDP on average,3 it is important to quantify what this additional spending has achieved. The World Health Report 2000 identified health gain as a fundamental goal of health systems4 but was unable to assess the specific contribution of health care. It also stimulated an increasing interest in performance of health systems with improved tools being sought avidly by policy-makers seeking to determine whether they are getting value for money in health care systems.

In the early 1970s Rutstein introduced the concept of mortality amenable to medical care. His working group selected over 90 conditions as “sentinel health events” from which disease, disability or death “should not occur in the presence of timely and effective care”.5 In 1977 and 1980, revisions of this list were undertaken.6,7 They have formed the basis for practically all subsequent studies on amenable mortality. Charlton was the first to apply the concept at the population level in England and Wales in 1974-78, also introducing the terms “avoidable deaths” and “conditions amenable to medical intervention”.8 He narrowed the concept by excluding deaths that were not directly linked to medical care, e.g. deaths avoided by policies on tobacco control. The concept was developed further within the Health Services Research Program of the European Community in the 1980s. This collaborative action resulted in a European Community atlas of avoidable mortality in which the work of Charlton and colleagues was extended and the boundaries of health services were interpreted as encompassing primary care, hospital care and collective health services.9 In 2001 Tobias and Jackson produced an updated list of conditions derived from an expert consensus exercise in which the relative avoidability of death was distributed according to primary, secondary and tertiary actions.10

Although the concept underpinning amenable mortality has been widely accepted since its first introduction by Rutstein, it has also been criticized,11,12 mostly on the basis that the relation between health care interventions and amenable mortality has not been tested empirically. Critics argue that for amenable mortality to be established as a valid indicator of the outcome of health care, direct analyses of its effect on mortality are needed. This has been addressed in the existing literature by two means; first, in time series analyses, based on the implicit assumption that innovations in health care affect the trend in mortality and, second, geographically, by comparing mortality in regions with different levels of health care. Time series analyses show faster decline in mortality in recent decades from most of the conditions identified as amenable compared to non-amenable.13–15 This has been used to support the argument that at least part of the overall mortality decline is due to improvements in health care. Two studies of geographical variation also found mortality differences for amenable conditions correlated with availability of corresponding health care interventions among countries16 and regions.8Table 1 presents a summary of the progressive refinement of conditions amenable to health care.

Previous research applying amenable mortality to health system comparisons have differed in their choice of indicators (e.g. Rutstein, Charlton or Holland, sometimes with adaptations). They have also varied in how they define “health care”. Finally, different upper age limits are used to define which deaths are considered amenable, which is justified because medical conditions may be amenable up to different ages, but which may also introduce a bias if different studies use different age limits. It has been noted that the lack of explicit criteria for selecting indicators of amenable mortality is an important shortcoming of the existing literature.12,14 The knowledge on indicators of amenable mortality is insufficient to believe that amenable mortality can serve as an indicator for the quality of health care because hardly any validation is done to reveal an empirical relation between the quality of health care and mortality outcome.

In the following method section we describe a systematic approach to select indicators of amenable mortality including an explicit list of initial selection criteria. We define health care as primary care, hospital care and personalized public health services (e.g. immunization and screening). The result section shows 14 causes of death that fulfilled the initial criteria, gives two examples how these causes of death were tested in an empirical validation analysis and presents the results of a Delphi procedure.

The AMIEHS project (Amenable Mortality in the European Union: towards better Indicators for the Effectiveness of Health Systems; http://amiehs.lshtm.ac.uk) has six aims, corresponding to the six levels in the project structure in Figure 1. The outcome, a list of indicators of amenable mortality, should provide insights into quality of health care.

Figure 1.

Structure of the AMIEHS project.

(0.37MB).

Figure 2 shows the stepwise selection of causes of death based on the criteria outlined above. Of 644 three-digit ICD codes, 160 codes had 100 or more deaths in 2000 in the UK and, of those, 54 exhibited a decline in age-standardized mortality of 30% or more. The exclusion of 40 codes for different reasons shown in Figure 2 resulted in a preliminary set of 14 conditions and 18 related interventions which are shown in Table 2 and are taken forward to subsequent steps.

Figure 2.

Flowchart illustrating the operationalisation of the desired properties of amenable mortality indicators.

(0.33MB).

Figure 3 shows a graphical representation of the validation analysis for peptic ulcer and renal failure. In most countries mortality from peptic ulcer (Fig. 3A) has decreased throughout the study-period. Only for men in France and for women in the UK there is a positive change in mortality that falls into the expected period of mortality decline, whereas for five countries we do not see this association. According to our criteria this example would lead to the conclusion that there is no sufficient evidence that population level mortality from peptic ulcer has been responsive to the introduction of Cimetidine. Mortality trends from renal failure (Fig. 3B) are very diverse across the countries. We find increasing and decreasing trends, as well as trends that peak in the 1980s. Renal failure could not be studied in Estonia because of incomparable ICD coding. In the UK and Spain mortality declines in the expected period for both men and women. In Sweden we see this coincidence only for men. In three countries we do not see this match and therefore there is no sufficient evidence that population level mortality from renal failure has been responsive to the introduction of cyclosporin. The results of this mortality trend analysis for these two conditions serve as examples for the similar analysis of all 14 conditions with similar results: overall we have not found clear evidence for an association between innovations in health care and changes in mortality.

Figure 3.

A) Estimated mortality trends from peptic ulcer and expected periods of mortality decline based on the introduction of cimetidine (upper panel men; lower panel women). B) Estimated mortality trends from renal failure and expected periods of mortality decline based on the introduction of cyclosporin (upper panel man; lower panel women).

(0.45MB).

Table 3 shows the results of the Delphi procedure. Only for three conditions consensus could be reached. These were those where all scores (except single outliers) were within a three point range.

The present study is based on a rigorous process of constructing an up-to-date list of indicators of amenable mortality. We demonstrate how we got from all 644 three-digit ICD-9 codes to a preselection of 14 causes of death using explicit selection criteria. With two example conditions we illustrate the subsequent empirical validation analysis which shows that the association between medical innovations and changes in mortality is modest. Our Delphi exercise shows consensus among experts on three amenable conditions. This study uses empirical evidence to validate possible amenable mortality indicators, whereas in the past such indicators have often been selected by expert opinion only. The findings from this project help to assess and compare the performance of the health systems.

Some limitations need to be addressed: we focus on innovations introduced after 1970. Some major innovations like the introduction of antibiotics have taken place before and are therefore not considered. We also focus on medical innovations. Yet improvements in population health due to health care reflect three factors: innovations, improvement in the quality with which they are applied and expanded coverage. At the level of international comparisons over time it is not possible to take account of the other factors given the absence of adequate data. Third, we are looking at the effect of individual interventions, rather as if they are “magic bullets”, although in reality many health gains are the consequence of incremental improvements. Fourth, in seeking evidence of the effect on mortality of specific innovations, we were limited as few trials have mortality as an end point, instead using intermediate measures. Fifth, we were constrained by the paucity of high quality observational studies attributing health gains to particular innovations. Sixth, although there is no alternative to using ICD codes at the three digit level, they are not designed for the purpose we are using them and the potential benefits of a therapeutic innovation may be distributed among many codes, such as those related to safer anesthesia. Finally, we adopt a narrow definition of the health system, excluding intersectoral actions. This is inevitable if using this method. A broader definition of the health care system may define more causes of death as potentially amenable. Not as a limitation of the amenable mortality approach, but as a factor that complicates the interpretation of observed trends in mortality, we want to mention that many factors other than innovations in health care influence mortality. Given the difficulty in precisely measuring these factors, it will remain a challenge to interpret mortality changes on the population level as a consequence of the particular change in health care. Socioeconomic differences in the trends of mortality and in the diffusion of medical innovation may hide socioeconomic differences in amenability. This could not be taken into account in this study. A recent study suggests that socioeconomic differences in amenable mortality do not reflect differences in access or quality of care.24

In the AMIEHS project the concept of amenable mortality is addressed in a completely different manner from previous studies by starting from all causes of death and working down to a limited number. Secondly, our selection of indicators is linked directly to changes in health care practice, addressing a previous criticism of the concept of amenable mortality. The indicators we identified help to highlight shortcomings in health care, but mortality rates from amenable causes must be used with caution as indicators of health care effectiveness in international comparisons. Elsewhere we suggest two possible alternative measures of amenable mortality25 and more in-depth analyses will be needed to identify the specific failures in health care systems. For policy makers the indicators identified in this study will offer the opportunity to put extra attention into specific causes of death. For researchers these indicators are hotspots giving rise to new investigations in the future.

R. Hoffmann and I. Plug are work package leaders in the project and drafted the manuscript. B. Khoshaba contributed to the analysis and to drafting the manuscript. M. McKee participated in the design of the study, performed analysis and contributed to the manuscript. J.P. Mackenbach participated in the design of the study and contributed to the manuscript. All authors read and approved the final manuscript.

The AMIEHS project -Amenable mortality in the European Union: towards better indicators for the effectiveness of health systems- has received funding from the European Union through the Executive Agency for Health and Consumers (EAHC) (agreement number 2007106).

None.