http://www.globalforumhealth.org/pages/index.asp  forum #3    June 1999

C. D. Mathers,  Australian Institute of Health and Welfare
GPO Box 570, Canberra, ACT 2601, Australia
Ph: +(61)62441138, Fax: +(61)62441166
Email: colin.mathers@aihw.gov.au

Summary

In the last two decades, considerable international effort has been put into the development of summary measures of population health that integrate information of mortality and non-fatal health outcomes and international policy interest in such indicators is increasing. Two major classes of summary measures have been developed: health expectancies and health gaps (of which the DALY or disability-adjusted life year is the best known).

An important goal in constructing summary measures is to identify the relative magnitude or burden of different health problems. This paper reviews and compares health expectancy and DALY approaches to the estimation of burden of disease.

Health expectancies have been used to measure the burden of diseases and injuries by estimating the potential gain in health expectancy resulting from the elimination of each specific disease or injury. Potential gains in health-adjusted life expectancy at birth through disease elimination have been calculated for Australia using data from a national population survey of disability and handicap. These estimates are compared in this paper with preliminary estimates of DALYs for disease groups derived from a national burden of disease study currently underway in Australia. These comparisons are used to illuminate some of the advantages and disadvantages of the two approaches.

It is concluded that summary measures of health gaps such as the DALY provide a better approach to measuring the health burden associated with specific causes than do health expectancies. DALYs provide a straightforward partitioning of total burden by an exhaustive set of disease and injury categories, and are additive across disease categories, whereas potential gains in health expectancy are not additive. In addition, DALYs provide a more sensitive measure of changes in burden than gains in health expectancies through disease elimination. The major challenges are to improve the way that comorbidities are dealt with using DALYs and to improve population level data on health states associated with diseases to enable better estimation of both DALYs and health expectancies.

1. Introduction

In the last two decades, considerable international effort has gone into the development of summary measures of population health that integrate information of mortality and non-fatal health outcomes. The interest in such measures stems from their usefulness to compare the health of populations, to monitor trends over time, to measure the total magnitude of health problems using a common currency (and one that can also be used for cost-effectiveness analyses), to ensure non-fatal health outcomes receive appropriate attention and to measure the population-wide benefits of health interventions.

Two major types of summary measure have been developed: health expectancies (disability-free life expectancy (DFLE), active life expectancy etc) and health gaps (disability-adjusted life years, healthy life years etc). Health expectancies have been widely adopted for monitoring progress in health at national and international level. Disability-adjusted life years (DALYs) have been used to estimate the global burden of disease to guide World Bank investment policies for health, to inform global priorities for health research and have been taken up by WHO to measure the global burden of disease in the year 2000.

An important goal in constructing summary measures, and one that may explain the increasing attention to summary measures, is to identify the relative magnitude or burden of different health problems, including diseases, injuries and risk factors. This paper reviews and compares health expectancy and health gaps approaches to the estimation of burden of disease.

2. Health gaps

Health gaps extend the notion of mortality gaps (eg. potential years of life lost to age 75) to include time lived in states worse than ideal health. The most widely known of these is the disability-adjusted life year or DALY (Murray and Lopez 1996). The DALY extends the concept of potential years of life lost due to premature death (PYLL) to include equivalent years of "healthy" life lost by virtue of being in states of ill-health.

DALYs for a disease or health condition are calculated as the sum of the years of life lost due to premature mortality (YLL) due to the condition and the "years lived with disability" (YLD) for incident cases of the condition. Each death or incident case of disease is categorically assigned to one underlying cause using the conventions of the International Classification of Diseases. In addition, the Global Burden of Disease project used an attributable fractions approach to estimate the burden attributable to risk factors or diseases such as diabetes that act as risk factors.

The DALY was explicitly designed to provide disease and injury specific estimates of burden that are additive across disease categories. Figure 1 illustrates how the DALY methodology produces additive disease-specific estimates. It shows draft estimates of the burden of disease for 10 leading broad disease groups for Australia in 1996. These estimates are derived from individual calculations for 175 disease and injury categories involving around 500 disease stages, severity levels or sequelae. This project is being carried out by the Australian Institute of Health and Welfare (AIHW) in close collaboration with the Victorian Burden of Disease Study, a state-level project headedby Dr Theo Vos.

 missing graphic

Figure 1. Provisional estimates of DALYS (YLL and YLD) for some major disease groups, Australia, 1996. DALYs are calculated using a 3% discount rate and uniform age weights.

Source: Mathers (1999a)

3. Health expectancies

Health expectancy is a generic term for population indicators that estimate the average time (in years) that a person could expect to live in a defined state of health. Examples include disability-free life expectancy (DFLE), active life expectancy and disability-adjusted life expectancy. The International Network on Health Expectancy (REVES) has promoted and developed the concept and methods and it is now widely used at national level and by OECD to report on population health (Mathers & Robine 1993, OECD 1998). WHO has recommended the use of health expectancies as summary indicators of population health (de Bruin and Picavet 1996). The recent Jakarta Declaration on Health Promotion into the 21st Century (WHO 1997) identified the ultimate goal of health promotion as to "increase health expectancy, and to narrow the gap in health expectancy between countries and groups".

Despite the effort that has gone into development of health expectancies, there are still very few countries for which they can be compared. Most health expectancies are linked to a particular health status measurement instrument (usually focussed on self-perceived health, long-term illness, chronic disability or long-term handicap). DFLE estimates are commonly based on the prevalence of long-term disability (typically defined as disability which has lasted or is expected to last 6 months or more).

Most health expectancies are also characterised by the use of dichotomous health states (eg. with disability, disability-free). Years of life lived with disability are given an implicit value of zero (equivalent to the valuation of death) for disability above a certain threshold, below this threshold the valuation is 1. The summary indicator is thus not sensitive to changes in the severity distribution of disability, and the overall DFLE value for a population is largely determined by the prevalence of the milder levels of disability and comparability between populations or over time is highly sensitive to the performance of the disability survey instrument in classifying people around the threshold (Mathers 1997).

4. Estimating burden using health expectancies

Health expectancy calculations usually start with population data on disability or handicap in order to estimate expectations of years lived in various health states. Attempts have been made to relate health expectancies back to disease and risk factor causes using data from population disability surveys linking disability to its disease or injury causes. This cannot be done in a categorical manner as with DALYs because all diseases and injury contribute to the risk of death or disability at each age, and the health expectancy is a weighted sum of these risks across all ages in a lifetable population.

One possible approach is to estimate the contribution of each specific disease or injury by comparing the current health expectancy with the value that would be expected with complete elimination of the disease or injury. The validity of such an estimate depends on the validity of the model used to predict the effect of disease elimination on mortality and disability risks. The first authors to use this approach modeled the effects of disease elimination in terms of separate estimates of the effects of mortality reduction and disability reduction (Colvez and Blanchet 1983). Mathers (1992, 1997b, 1999b) used self-report data on the main condition causing disability to estimate the combined effects on disability and mortality of disease elimination. An alternate approach, which takes better account of comorbidity, is to use multivariate modelling approach to estimate the proportion of disability prevalence associated with each of a number of chronic diseases (Nusselder et al. 1996, Nusselder 1998). Wolfson (1996) has calculated attribute-deleted health expectancies (deleting types of disabilities rather than causes).

An alternative approach is to assess the effect of small changes in the disease or injury rather than model complete elimination. The results are expressed as the elasticity of the health expectancy with respect to changes in the disease or injury, or as a numerical approximation of the partial derivative of the health expectancy (Hill et al. 1996, Mathers 1999b).

Mathers (1997, 1999b) has estimated potential gains in health-adjusted life expectancy (HALE) due to disease elimination using 1993 Australian disability survey data. Figure 2 compares these estimates at chapter level of ICD-9 for males and females with the corresponding total DALY estimates (undiscounted) from the Australian Burden of Disease study. To maximise comparability with the prevalence-based HALE estimates, the DALYs plotted in Figure 2 are based on prevalent YLDs rather than incident YLDs for 1996.

There is a high correlation between DALYs and HALE gain across chapters for males and females. The main outliers are mental disorders and respiratory disorders for which DALY estimates give a substantially higher ranking. The discrepancy for mental disorders probably arises mainly from the very substantial under-reporting of disability due to mental disorders in the Australian population survey, not only for psychoses but also for anxiety disorders, affective disorders and substance abuse. It is also possible that the National Survey of Mental Health and Wellbeing (ABS 1998), on which estimates of YLD for anxiety disorders are based, over estimates the number of cases of these conditions.

The respiratory disorders category includes chronic conditions such as asthma and COPD, but also upper and lower respiratory tract infections. The latter do not result in long-term disability that would be captured in the population survey. The burden of asthma is also likely to be under-reported in the disability survey.

 missing graphic

Figure 2. Comparison of DALYs and gains in health-adjusted life expectancy (HALE) at birth, by ICD-9 chapter and sex, Australia. DALYs are undiscounted and use prevalence YLDs.

 5. Discussion

The most significant methdological differences between the DALY and HALE gain approaches to estimating burden of disease are the following:

• HALE gain is usually derived from self-reports of disability whereas the DALY is based on analysis of epidemiological data and expert assessment of the incidence, duration and disability severity distribution for each disease or injury.

• HALE gain is usually based on prevalent disability whereas DALYs are incidence-based.

• Weights used to construct HALEs relate to generic disability or handicap severity classes whereas DALY weights are usually derived for health states linked to specific conditions.

One of the major problems with the HALE approach relate to the first of these points: there are substantial practical problems in mapping disability to diseases and injury using population disability survey data. Issues in using self-report data on causes of disability have been discussed in some detail by Murray and Lopez (1996) and Mathers (1997a).

In addition, health expectancies are quite inelastic to disease elimination in low mortality countries. For example, the complete elimination of cancer in Australia would result in an additional 2.1 healthy years of life for Australian females. This represents a less than 3% increase on the current value of 74.7 years for health-adjusted life expectancy at birth. In contrast, health gaps such as DALYs provide a natural denominator for the magnitude of gains through disease reduction, namely the current health gap for the disease.

The third, and perhaps most important problem with health expectancy gains as measures of burden of disease is that the gains for specific diseases are not additive. Elimination of two or more diseases simultaneously results in a larger gain than the sum of the gains from elimination of each disease on its own. It is possible to make health gaps additive, so that the total health gap is the sum of a set of exhaustive disease categories, but this cannot easily be achieved with health expectancy gains.

The disease elimination method for calculating the contribution of diseases and injuries has some advantages. It is conceptually clearer, and the meaning of the estimates (extra years of life) is readily understood by the average person. In addition, it readily allows multi-causality to be taken into account and it is consistent with economic approaches to evaluating the benefits of health interventions (Murray, Salomon and Mathers 1999). However, when balanced against the problems with attribution of disability to causes and the lack of additivity of health expectancy gains, and the inelasticity of health expectancies in low mortality countries, we must conclude that health gaps measures such as the DALY are much easier and more intuitive to use for the purpose of measuring the burden of disease.

6. Conclusions

In principal, summary measures of health gaps such as the DALY provide a better approach to measuring the burden of disease for specific causes than do health expectancy methods based on cause elimination or estimation of elasticities:

they provide a straightforward partitioning of the total burden by an exhaustive set of disease and injury categories, such as that provided by the ICD;

they are additive across disease categories;

they provide a more sensitive measure of changes in burden than gains in health expectancies through disease elimination.

The main practical problem with the DALY approach is in the measurement of the disability or health state distribution associated with incident cases of disease, for which there is generally not yet adequate population-based data. The issue of multi-causation or comorbidity also requires further methodological development. The many issues associated with obtaining valid and reliable health state valuations are common to both the health gaps and health expectancy approaches.

If complete data on the distribution and valuations of health states associated with incident cases of disease or injury were available for populations, then HALEs and DALYs could be calculated in a manner consistent with each other. They could then be used in a complementary manner to describe the health of a population and the contribution of specific conditions and health determinants to the burden of disease.

Acknowledgements

The author gratefully acknowledges discussions with Chris Murray and Josh Salomon (WHO) on summary measures of health, and Theo Vos (Victorian Department of Human Services) for methodological advice on DALY methods and sharing analyses of YLD for mental disorders, injuries and other conditions in Australia.

see also at Skyaid:  
Years of Life Lost   Cost Effectiveness,   International Health,   Injury