BURDEN AND COST OF DISEASE
The concept of the burden of disease allows governments to set priorities in policy formulation and program delivery. One useful burden of disease measure is the Disability Adjusted Life Year (DALY) measure. This combines, for a particular disease, estimates of loss of years of life due to premature mortality with estimates of loss of healthy years of life due to associated disability. Diseases or conditions which cause more severe disability are given a higher weighting, as their impact on a healthy life is greater. For example, severe depressive episodes are given a higher weight than mild depressive episodes as the former are more disabling for the sufferer. For diseases, such as cancer, which follow a relatively predictable disease process through medically identifiable stages, a range of disability weights are used to measure the burden of the disease through its progress. As this is a population level measure, the weighting factors applied are based on the average disability experienced for a disease or injury, and may not accurately reflect the experience of particular individuals with that disease or injury.
Measuring the burden of a specific risk factor involves a second step; as its burden is accumulated across the range of diseases related to that risk factor. For example, poor diet has been linked with a range of conditions including certain cancers, circulatory diseases and diabetes. Hence, the burden of a specific risk factor is measured by estimating what proportion of these related diseases can be attributed to that risk factor. These attributable proportions are called aetiological fractions (AF), or attributable risks. The estimation of AFs is most difficult where a number of risk factors combine to contribute to a disease, or where the relationship between the risk factor and the disease is not well understood. For instance, estimating the contribution poor diet makes to certain cancers is more complex than estimating the contribution that smoking makes to lung cancer.
Government policy is necessarily concerned with the cost to the Australian health care system of specific diseases and risk factors, as well as their burden on the population. Hence, another important measurement issue is the estimation of the direct health care costs attributable to a specific disease. This estimate needs to incorporate measures such as the prevalence of the disease, including its distribution in terms of age and sex, and the method and cost of treatments. Treatment of any condition can be confounded by the presence of other conditions and, as older people often have a number of conditions, cost of treatment for a specific disease may vary according to age. Direct health care cost estimates do not represent costs for the whole health system as they exclude several areas such as community health services and public health programs. They also do not include the economic impact of absenteeism, lost productivity, burden on family members, etc.
Almost all population surveys rely on collecting self-reported health data from members of the public. The data may relate to the individual's own health status or be provided on behalf of another person in the household, e.g. a parent or guardian may be asked to provide health information about a child. Population surveys collect information from people with all types of health experiences, not only those in contact with health services or with a particular disease or condition. Hence, the reported experiences largely reflect the health experiences of the population from which they are drawn. To maximise the potential of this ability to represent the whole population, these surveys generally collect a set of relevant socio-demographic variables enabling health inequalities across groups to be investigated.
One particular strength of population surveys is their ability to collect data on behaviours and attitudes which may impact on health decisions and future health status. Measures of a range of factors known to influence health can be used to provide valuable advance warning of health issues. For example, increases in physical inactivity and increases in poor dietary habits could be used to forecast increases in the prevalence of heart disease and other linked diseases, say in 15-20 years time. If physical inactivity and poor dietary behaviours are increasing more for some groups than others, those groups may be expected to be affected to a greater degree in the future. Information about the lifestyle behaviours of various population groups can therefore influence health programs and initiatives targeted at those groups, and indicate the success of these.
Ideally, self reported health data based on recall of diagnoses or descriptions of symptoms should be verified by checking medical records. However, it is generally not feasible to undertake the physical examinations required to produce medically verified data as part of the collection process (e.g. bone scans are needed to identify osteoporosis in older women). Similarly, self reported data on health related behaviours are not always reliable. For example, in general, people under-report certain behaviours, such as alcohol consumption. Further, people often report symptoms which can be difficult to code to specific diseases, and they cannot report diseases or conditions that are underlying and have not yet been diagnosed by a health professional. Added to this are issues surrounding how well people are able to recall their disease histories or lifestyle changes. Apart from a number of conditions such as asthma, self reported data on physical symptoms cannot be used to estimate accurately the number of people in the population with an underlying condition. A range of possible measures providing more exact physical data might include height and weight to produce data on obesity (based on measured body mass index), blood pressure to provide data on hypertension, and a series of tests that can be undertaken on blood samples.
Albeit based on self-reported data, most population surveys focus on objective health status and related risk factors. However, they can also be used to measure perceptions about health and wellbeing. The simplest way to do this is to ask people to assess their own level of health in a single question. As discussed earlier, responses will depend on an individual's physical and mental state of being, and on external influences such as their social environment. People may compare themselves with friends or, more broadly, with people in their community. This kind of complex interaction between the individual and external factors cannot be measured, even when more than one question is asked to gain insight into general health and wellbeing. However, several ABS surveys have used modules of questions designed to provide a general measure of physical and mental wellbeing and to give an indication of the level of disability attributed to physical and mental problems.
Even where population surveys are conducted at frequent intervals, they do not provide information which can be used to attribute cause and effect. For example, while the ABS's National Health Survey collects information on cardiovascular disease, nutrition, smoking and alcohol consumption, it is not possible to use this data to determine causal relationships between these factors, as the data relate to the same point in time. A complicating factor is that, once diagnosed with a disease, people will often change their lifestyle to a healthier one, so individuals with chronic disease may well be subjecting themselves to fewer lifestyle risk behaviours than people who are healthier, producing an apparent anomaly in the data.
As discussed earlier, it is important to recognise that a snapshot of an individual's health status recorded at one point in time may not reflect the social or economic circumstances they have experienced over their lifetime. While longer term health-related information can be explored more effectively through longitudinal surveys, these are costly and present logistical and sample maintenance difficulties. Another way of adding a time dimension to health data would be to link survey respondent records (with informed consent) to external datasets such as the Medical Benefit Scheme (MBS) or the Cancer Register. Linkage of survey data with administrative datasets such as the MBS would support analysis of, for example, the influence of socioeconomic factors on inequalities in the use of health services. ABS is currently examining the feasibility of data linkage between NHS records and MBS, within the constraints of ABS and other legislation. The project will progress only if an acceptably high proportion of respondents agree to the proposal.
Some collections are based on the administrative records of health services, for example, hospitals provide administrative data about episodes of hospital care. Other collections are registry based, where a central repository of information is notified when certain diseases are diagnosed by health professionals. Disease registries, such as the Cancer Register at AIHW, record the incidence of a specific disease, or new cases of diseases and, over time, will provide an approximate estimate of prevalence of the disease. In addition, essential health information is provided by the system of vital registrations, including registration of deaths, from which causes of death data is compiled by the ABS. As administrative data is based on diagnoses by health professionals, it is generally reliable. However, usually only restricted demographic data is collected so that these datasets have a limited capacity to reveal, for example, differences in the disease profile or service utilisation of population groups. Also, administrative sources focus specifically on people who have made contact with the health care system, a subset of those with health problems or health issues.
To address some of the shortcomings of larger administrative and registry based datasets, there is a substantial body of information collected through health studies conducted by researchers within universities, hospitals and other organisations. Also, with the introduction of the International Statistical Classification of Diseases and Related Health Problems - 10th revision (ICD-10) and the retention of all causes of death named on the death certificate, researchers now have available a wealth of data for exploring disease associations.