iassist quarterly - 41

Australian Health

Statistics

by Roger Jones '

Social Science Data Archives.

Australian National University.

Canberra. Australia.

Introduction

The impetus for this paper was a workshop

held earlier this year to identify current

problems associated with Australian health

statistics and determine priorities for a national

health data base. Generally, academic

researchers in Australia have focussed on

aetiology, that is the study of causes of diseases,

and attention will be given to the data used to

develop aetiological hypotheses and, ultimately,

test these hypotheses. By identifying the major

causes of avoidable death, disease or disability

as well as possible with existing data, the gaps

in the data will be identified and priorities for

the collection of new data will be established.

'Paper prepared for presentation at the

IFDO/IASSIST Conference. Amsterdam, May
20 - 23, 1985.

Editors note: Unfortunately the accompanying,
tables were not submitted with this paper. We
have therefore deleted all specific table

references. With apologies to the author.

Data Requirements

Increased attention is being given to the need

for improved occupational health surveillance in

western societies. In Australia, concerns about

potential health effects of exposure to the

herbicide 2,4,5-T and its dioxin contaminant,

Vietnam War service in general, proximity to

atomic testing, past employment in uranium

mines, and exposure to asbestos and lead are

prominent among the issues that have focussed

attention on occupational risks. The risks of

cancer resulting from occupational exposures

have been discussed widely, and the patterns of

accidents, injuries and illnesses associated with

industry and occupation examined.

Personal and lifestyle factors such as diet,

cigarette smoking, alcohol comsumption, stress

and drug use are also now recognised as

important 'risk factors' associated with the

health status of the population. There is now
evidence to support the view that the chronic,

degenerative diseases such as heart disease and

cancer which are the major killers of Australians

substantially result from these socio-behavioural

factors rather than simply being the diseases of

old age.

The main causes of death and hospitalization in

Australia, for people below the age of 45, are

motor vehicle accidents and other accidents,

poisoning and violence, including suicide. For

older people the chronic diseases of ischaemic

heart disease, stroke and other diseases of the

circulatory system and various cancers

predominate.

A number of factors are known to influence, or

to be associated with these causes of death.

The most important behavioural factors, in

terms of the amount of resultant disease, are

almost certainly smoking and alcohol

consumption. The Commonwealth Department

of Health suggested that there were 16,200

Summer 1986



42 - iassist quarterly

deaths in Australia associated with tobacco use

in 1980. equivalent to 110 deaths per 100.000

population. The diseases with which smoking is

associated are clearly indicated in the warning
'

Smoking causes lung cancer, as well as heart

and other lung disease ' which the NH & MRC
(editor's note:) (National Heath and Medical

Research Council) recommended to replace the

less specific
'

Smoking is a health hazard' on

cigarette packets sold in Australia. Almost half

the number of deaths attributed to alcohol are

associated with road traffic accidents, from

which disability and injury also result Diet is

also impUcated in several of the major causes of

alcohol-related deaths.

Descriptive Studies

Considerable difficulties arise when trying to

establish causal links with diseases, such as the

argument over smoking and cancer. Individual

studies are rarely definitive and considerable

time may elapse before sufilcient evidence is

available to assume a causal relationship. Such

studies can also be expensive. Accordingly, less

intensive descriptive studies are undertaken in

the first instance, using readily available data

sources from which crude measures may be

derived in order to generate or provide an

initial check of hypotheses. Fully developed

analytical studies would only follow when the

results of exploratory investigations £ire

sufficiently suggestive and important to wanant

the expense of more rigorous confirmatory

research.

Information for general descriptive studies may
be available from routine collections of health

event data or from prevalence surveys of

population samples.

Routine Health Collections

The routine health reporting systems in

Australia are poorly developed in comparison to

other developed coimtries, particularly at the

national level. Under the federal/slate system

of government that operates in Australia, state

governments are primarily responsible for the

provision of hospital and health services within

their own borders, and for the collection of

health event data associated with these services.

While all states administer similar collections,

substantial variation exists in terms of coverage,

scope and means of collection.

Mortality data play an important role as health

status indicators. Data are compiled in each

state by the Registrars of Births, Deaths and

Marriages and causes of death coding is added

by the Australian Bureau of Statistics (ABS)

from medical certificates. At the present time,

there is no national compilation, although health

ministers have agreed to establish a National

Death Index, subject to appropriate

confidentiality legislation being enacted. To
obtain Australia-wide mortality tapes at present,

with personal identifiers removed, separate tapes

for each state would have to be developed by

the ABS and sent to the state registrars for

release, subject to their approval, and

re-aggregated by the researcher. Clearly this is

a very cumbersome procedure. The main public

source of data at present is the ABS publication

'Causes of Death'.

Hospital morbidity collections based on data

collected on patients leaving hospitals are

compiled by the health authories in each state,

providing limited information on the

demographic background and illnesses of

hospital in-patienls. In all states, data are

collected on separations of public patients from

public hospitals, but private hospitals, psychiatric

hospitals and nursing homes may be included in

some states and not in others, so that

Summer J 986



iassist quarterly - 43

comparisons between slates and aggregation

aCTOss states cannot be accomplished from

published results, although a uniform minimal

data set could be achieved with access to the

computerised records.

Records of primary contact between the public

and health professionals such as private medical

practitioners, community health centres, hospital

clinics and casualty services are totally lacking at

present, at least in any unified form. However,

the introduction of the universal health

insurance scheme. Medicare, in February 1984

has created an opportunity to obtain more
information in this area. Medicare gives

automatic entitlement to a subsidy of 85% of

the schedule fee for medical services provided

by private practitioners, and approximately 100

miUion records per year relating to claims are

stored and interfaced with a Medicare enrolment

file containing personal data. However, as yet,

only the information necessary for payment of

benefit is being collected: name, date of birth,

sex, geographic area, and usual residence, and a

provider's identifier. No data are collected

about either diagnoses or procedures.

With regard to specific diseases, each state now
has a Cancer Registry based on notifications by

hospitals and laboratories of cancer patients and

by Registrars of deaths attributed to cancer. A
proposal to establish a National Cancer Statistics

Gearing House is currently before the NH &
MRC and, if accepted, national figures should

be available within a few years. Instances of

communicable diseases are reported to state

health authorities and collated by the

Department of Health but are of very hmited

research value. Information on work-related

health problems can be obtained from

compensation claims to the stale-based insurance

schemes, but large numbers of workers are not

covered and there would be under-reporting

because neither medical practitioners nor

workers are aware of the significance of work

factors in the aetiology of many chronic

diseases. Police, legal authorities and traffic

authorities collect information which yields

statistics on road traffic accidents, and on

drink-driving offences and drug offences. At
present, these have no research value, but the

Department of Transport is developing a

National Road Traffic Accidents data base based

on police reports of accidents involving fatalities

and casualties.

How can these routine reporting systems be

used to develop and check hypotheses about

causes of disease? The traditional approach is

to produce estimates of the risk of various

health outcomes across various subgroups of the

population under study. Thus, for example, an

indicator of the health risk associated with

particular occupations is given by ratios of the

number of deaths from various causes to the

population at risk and making comparisons

across occupational categories. The population

at risk is usually obtained from census data.

Clearly there are severe limitations with this

t>'pe of unlinked data analysis, particularly in

the lack of ability to apply controls for what

might be relevant covariables. This is limited to

those factors coded in both the census and

health event data, and there are usually very

few in the latter - age, sex, location and

perhaps occupation, often in groups too broad

to be useful.

The value of this type of data is greatly

enhanced when it can be linked directly at

individual level to data on personal

characteristics and lifestyle factors. In this case,

files must include full names, previous surnames

and dates of birth at least Concern about

confidentiality is, however, very high in

Australia and the opportunities for record

linkage are very limited, although some useful

work has been done by linking cancer records

and death records to employment records.

However, identifiable census returns have been

destroyed in Australia since the start of the

century' and proposals that a sample of these be

retained, as has been done recently in England

and Wales, seem unlikely to succeed.

Summer 1986



44 iassist quarterly

Population Surveys

Given the poor state of routine health data

collections, it is fortunate, though perhaps not

unrelated, that Australia shows up reasonably

well in the area of health surveys in the

International Health Data Guide . This is

probably due to the preference of the Australian

Bureau of Statistics for using its limited

resources for population surveys which can

cover a range of topics rather than for more

specific health related collections. In additioa

state health authorities and other agencies have

undertaken a number of surveys relating to

lifestyle factors, particularly cigarette smoking,

alcohol consumption and drug use.

Prevalence surveys of population samples have

the advantages that, for established risk factors,

such as smoking, they can be used to describe

cross-sectional and longitudinal variations which

may relate to health outcomes, and they provide

a basis for evaluating community behaviour -

intervention programmes. For postulated or

possible risk factors, they can be used to

examine relationships with health outcomes and

may provide clues about causal factors in

disease.

The Australian Health Surveys in particular

provide valuable information on the health

status of the population, particularly since recent

changes to the ABS Act now permit the release

of de-identified unit record data. A data file

from the 1977-78 survey has been released, and

the 1983 data file should be available later this

yejir. A further survey is planned for 1986.

The surveys give detailed information on a wide

range of personal characteristics, although the

categories of such important variables as

occupation and birthplace were too broad in the

unit record file for many research purposes.

Categories had been collapsed in order to

ensure non-identifiability of respondents.

particularly those in small subgroups of the

population. For such subgroups, of course,

sample surveys are of Uttle value because of the

small number, if any, of cases interviewed.

Nevertheless the practice of collapsing variables

into standard classifications without giving

sufTicient thought to the potential research uses

needs to be changed.

Interview data on self-reported health status

may also be a suspect, and it would be useful

to have some medical verification of a

subsample of respondents or from pilot tests.

Another shortcoming of these data is the lack

of information on health risk factors such as

smoking and drinking, although 12 items from

the General Health Questionnaire were included.

Several community health studies have been

carried out in Australia in recent years, and

these may provide clues to the relationship

between life-style characteristics and health

status. One of the advantages of such studies is

that they often include clinical checks on the

self-reported health status of respondents. The

main disadvantage is that the sample sizes are

generally too small to allow tests of hypotheses.

The prevalence of hean disease as a major

cause of morbidity and mortality has generated

a number of studies aimed at determining the

associated life-style factors. The best of these

are the two Risk Factor I*revalence Studies

conducted by the National Heart Foundation in

1980 and 1983. Both studies used large national

random samples and included clinical

examinations to obtain height, weight, blood

pressure and blood lipid levels in addition to

interview data on tobacco, alcohol and

medication consumption, diet, physical activity

and psychological stress.

While major programmes designed to infiuence

the smoking behaviour of the community have

been launched throughout Australia, the ABS
has chosen to cease conducting surveys on

smoking and to exclude questions on smoking

Summer 1986



iassist quarterly - 45

from the Health Surveys. The last national

survey on smoking conducted by the ABS was

in 1977. A series of much smaller national

surveys, on about 6000 adult respondents, have

been conducted by the Anti-Cancer Council of

Victoria in 1974, 1976, 1980 and 1983.

However, a large number of school-based

surveys of alcohol, tobacco and drug use have

been undertaken, although as a basis for

national figures these have problems of

comparability. Two national surveys of school

children aged 9-16 years were conducted by the

NH & MRC in 1969 and 1973, and a similar

survey has recently been carried out by the

Australian Cancer Society and the National

Heart Foundation.

Efforts to reduce the number of road traffic

accidents have centred on programmes designed

to reduce alcohol use with random breath

testing being introduced in most states of

Australia and substantial expenditure on

television advertising campaigns. However, the

availabihty of statistical data to evaluate the

effectiveness of these approaches is limited

largely to the mortality and morbidity statistics

and some small studies on knowledge, attitudes

and behaviour relating to drink-driving.

Concerns about invasion of privacy, the lack of

legislation on the preservaton of confidentiality

for some collections (morbidity), and

over-rigorous interpretation of such legislation

in others (census), have restricted the use that

could be made of these data. Population

surveys have been adopted as an alternative, but

these are generally too small for detailed

analytic studies and are thus limited to

monitoring the prevalence of established risk

factors.

The National Health Statistics Workshop held in

February expressed its concern over this lack of

appropriate data and recommended that a

national health statistics agency be established as

part of the newly formed Australian Institute of

Health. This new agency should promote the

development of national collections such the

natural death index, cancer index and morbidity

collections, and ensure that the necessary

legislation is enacted to provide for preservation

of confidentiality. P*riority should be given to

assembling data already available in most cases

at state level into unified national collections, to

the development of record linkage procedures,

and to risk factor surveys of diet, smoking,

alcohol and illegal drugs.

Concluding Comments

As indicated in the above brief review of

Australian health status data, only limited

attention has been given to the needs of

researchers for aetiological analyses. Routine

health collections lack uniformity across state

boundaries and do not include sufficient

information on parents' background or possibly

associated risk factors. Record linkage could

overcome some of these deficiencies but has

generally been resisted by the appropriate

authorties.

This is obviously a large agendum which will

require considerable resources and time to

implement Nevertheless, there is strong support

behind the recommendations and a reasonable

hope that a substantial improvement in

Australian Health Statistics will be achieved, n

Summer 1986