Chapter 3
Improving health policy development
Introduction
3.1
Australian governments rightly place a high priority on the health of
their citizens. As a result Australia delivers some of the highest quality and
best value hospitals and primary care in the world. However, a world-class
healthcare system is an expensive business. In 2013-14, combined government
health-related expenditure was greater than $100 billion per annum.[1]
The Commonwealth alone expended more than $63 billion in that year, the
equivalent of 25 per cent of Australian Government tax revenue.[2]
Over the past decade overall health expenditure has grown significantly above
the inflation rate at 5 per cent in real terms.[3]
3.2
At a time when the government is struggling to effectively manage the
growing health budget, it is clear that new opportunities to evaluate current
practices and deliver more effective and cost-efficient policies and programs should
be vigorously pursued.
3.3
This chapter explores the new opportunity that big data provides to
ensure that our health expenditure is as efficient as possible, and more
importantly to improve the standard of healthcare in Australia.
The traditional approach
3.4
Witnesses told the committee that the current approach to government
health policy evaluation and development lacks a firm evidence-base. For
instance the Centre for Big Data Research in Health cited evidence underpinning
the Medicare Benefits Schedule (MBS) and Pharmaceutical Benefits Scheme (PBS)
to highlight the limitations of the current approach:
Australian governments invest more than $100 billion annually
on healthcare, yet we have a relatively limited understanding of Australia’s
return on this investment. For example, the Medicare Benefits [Schedule]
(Commonwealth spend approximately $21 billion annually) consists of some 6000
items, but fewer than 5% have been assessed for safety, effectiveness and
cost-effectiveness against contemporary evidence. Even when medical treatments
have undergone extensive pre-market evaluation in randomised controlled trials,
like all of the items listed on the Pharmaceutical Benefits Scheme (Commonwealth
spend approximately $9 billion annually), they are most often tested over
relatively limited time frames, even if medicines are used for a lifetime, and
in populations that do not represent the people using them in routine clinical
care.[4]
3.5
In a recent research paper, the Productivity Commission pointed to
potential negative consequences of data holders not permitting the release of
administrative health data:
Concealing data can mean that patients receive ineffective
(or even harmful) care, adverse effects of drugs go undetected, or significant
money is spent on interventions that do not improve health outcomes (rather
than on interventions that do). It can also make it difficult to hold health
care providers to account for their performance.[5]
3.6
Dr Barbara Mintzes, a Senior Lecturer at the University of Sydney's Faculty
of Pharmacy, gave the committee a concrete example of risks associated with the
traditional approach to PBS listings:
When a medicine is first approved for marketing, we know very
little about its safety, especially in the longer term. On average, drugs are
approved based on the experience of around 2,000 people who have used the
medicine for short periods of time. Once on the market, millions of people may
use the same drug. This is what happened with the arthritis drug Vioxx [a
prescription anti-inflammatory which was recalled in 2004]... In its five years
on the market, Vioxx caused up to 140,000 heart attacks in the US.[6]
3.7
At a subsequent hearing, Professor Fiona Stanley who pioneered ground data‑linked
population health research in Western Australia in the 1970s and '80s, explained
how the problems with Vioxx could have been mitigated through the use of data
linkage:
I have one example around a PBS linkage to all the health
outcomes... Vioxx was not picked up for a long time—perhaps for four or five
years—because it caused a common outcome of heart attacks and heart-related
deaths. However, if we had linked our PBS into all our health outcomes, how
many deaths and serious, morbid and costly heart attacks could we have
prevented in that four or five years? Hundreds and hundreds. In my opinion, not
doing this linkage of PBS to health outcomes alone is actually negligent.[7]
3.8
The Australian Health Economics Society (AHES) pointed out that under
current arrangements, certain 'basic questions' cannot be answered:
Australia has been – and still is – lagging behind [the US,
UK, Canada and New Zealand in the access and use of health care administrative
data]. As a consequence, Australia has a poorer health economics and health
services research infrastructure and many basic questions cannot be addressed
(e.g. changes in the out-pocket payments by consumers using Medicare)...[8]
3.9
As a result the AHES submitted that Australia is foregoing 'considerable
benefits in terms of understanding health system which can lead to both greater
efficiency and improved health outcomes.'[9]
The AHES concluded that:
...research within government focuses on short term issues
within electoral cycles and is not driven by an overarching research strategy
that focuses on the key long term questions. As a consequence, key research
questions and policy issues remain unanswered for decades and governments
continue to revisit these issues inefficiently leading to waste.[10]
3.10
SA-NT DataLink highlighted the difficulties faced by state and territory
governments in formulating their health policies due to the inaccessibility of
Commonwealth data:
Lack of timely and affordable access to critical Commonwealth
data (such as MBS, PBS, Centrelink) is a serious obstacle to developing a more
informed understanding of health outcomes and services at the State/Territory
levels.[11]
3.11
Finally, the Centre for Big Data Research in Health argued that, given a
multitude of modern-day pressures, the traditional approach is 'no longer
adequate':
The increasing complexity of healthcare in terms of rapidly
evolving and fragmented service delivery models, the disruptive impacts of new
therapies and technologies, and people living longer with multiple health
conditions means that traditional methods guiding health policy and practice
are no longer adequate.[12]
New opportunities for health policy development
3.12
By contrast, a variety of submitters explained the significant benefits
that could flow to the development of health policy if decision-makers had a
more robust evidence-base.
3.13
The Department of Health provided the committee with a long list of
'significant benefits' which big data can bring to the health system:
- Better information to inform the government’s policy decisions
- A clearer picture of the real experiences of patients as they
engage with the health system
- A better understanding of what works, how well, for what cost,
and in what circumstances
- Earlier detection of trends – both positive and negative
- Earlier detection of anomalous behavior and deviations from
expected results
- A more efficient health system, by supporting the most cost‑effective
treatments, strategies and interventions on broad‑based independent
evidence.[13]
3.14
SA-NT DataLink drew the committee's attention to analysis by the Productivity
Commission that highlighted the critical need for evidence-based policy:
[The Productivity Commission] recognised that the ability to
undertake population based research by linking administrative data held by
government agencies and other bodies is essential to supporting evidenced based
policy. [The Productivity Commission] strongly argued the need for systematic
evidence-based policy to ensure the effectiveness of the massive expenditure in
the areas of health, welfare, education and other areas of Government activity.
[The Productivity Commission] regarded the demonstrable effectiveness of this
expenditure in achieving the planned for outcomes as critical, particularly in
periods where there are very strong budgetary pressures on Government to reduce
expenditure.[14]
3.15
The Council of Academic Public Health Institutions Australia (CAPHIA) also
provided a compelling account of the benefits of linking health datasets to
deliver improved health policies:
The availability and accessibility of linked data collections
is vital in working towards improvements in the health of Australians and in
healthcare delivery. At a population level, data linkage provides a more
complete understanding of health than is otherwise possible utilising
alternative research methods. Providing approved researchers with access to a
range of linked State and Commonwealth health and social data has the potential
for national, state and local comparative effectiveness, clinical trials and
registry research that has thus far been largely untapped, to drive health
policy, redesign, quality improvement and evidence translation in health care.
Additionally, it enables, for example, the rigorous objective evaluation of
health policy for government and key policy professionals; and the ability to
compare trends nationally, to identify programs that deliver value for money
and to avoid wasting resources on those that are not delivering. The result is
better targeted, evidence-based and more cost-effective health policy, services
and interventions for the Australian community.[15]
Linking Commonwealth datasets
3.16
Health policy development at the Commonwealth level was a key area where
submitters argued that significant improvements and efficiencies could be made.
For instance Dr Julian Elliott, a Senior Research Fellow at the Australasian
Cochrane Centre explained that the benefits or transforming existing datasets
into 'evidence-informed policymaking':
So we need to have a capacity in Australia, particularly
public agencies, to use the datasets that are becoming available to really
drive effective decision making, evidence-informed policymaking... It is really
about how we take data and then transform that into evidence-informed,
up-to-date recommendations across the whole of the health sector, whether it is
policymaking at a Commonwealth level or right down to the decision of an
individual clinician. We have a huge opportunity to improve that cycle...[16]
3.17
Dr Elliott elaborated on the benefits of using big data analytics to
effectively evaluate the impact of policies on health outcomes:
...it is really about how we monitor the effect that these
systems then have on the outcomes that we are interested in. Is it actually
changing healthcare practice? Are we getting a return on investment for our
healthcare interventions? Ultimately, what effect is it having on patient
outcomes? Those elements can also be collected and understood within these data
systems.[17]
3.18
Witnesses referred the committee to an array of important Commonwealth
data sources that could be beneficially linked to support evidence-based policy
development as well as providing medical researchers with valuable source data.
Perhaps the most comprehensive list was provided by SA-NT DataLink which
suggested the following:
PBS, MBS, Immigration, Justice, Childcare Benefits, Private
and Public School Education, Higher Education enrolment and academic results,
Aged care, Family Tax Benefits, Employment related data from ATO Personal
Income Tax and Company Tax/ABN GST and ABS.[18]
3.19
Of this list, two datasets, the MBS and the PBS, were virtually
universally recognised by submitters as key Commonwealth data source for
linkage. CAPHIA for instance submitted that:
MBS and PBS data represent two of the most important datasets
in the Commonwealth repository, as they provide information on the uptake of
primary care and specialist medical services, as well as use of medicines in
the community, which are not available through routinely-collected State‑based
data collections. When combined with other data, they can provide a rich source
of information to allow analysis of clinical outcomes, effectiveness of health
policy, cost-effectiveness analyses and access to services across a range of
dimensions, including residential location, socioeconomic status and
Aboriginality.[19]
3.20
The current restrictions on linking MBS and PBS data were highly commented
on during the inquiry. This issue is explored in greater detail in the next chapter.
A related discussion regarding several Commonwealth departments' reticence to
release de-identified data to other agencies is covered in Chapter 5.
3.21
Finally, it is worth mentioning a recent Commonwealth Government
initiative, the Multi-Agency Data Integration Project, which is linking a
series of related Commonwealth datasets. The ABS which is leading the project
submitted:
A key example of these [data custodian] partnerships is the
Multi-Agency Data Integration Project which brings together, for the first
time, Census data with administrative data on health, income, and social
security payments, to establish a foundational, linked data resource. The
project aims to create an enduring integrated data resource that is:
... A comprehensive data source for evidence-based policy
development across areas of broad social and economic concern...[20]
3.22
This initiative is further discussed in Chapter 5.
Linking Commonwealth and state datasets
3.23
Due to the shared responsibility for the development of health policy in
Australia, significant quantities of health data is collected at both the
Commonwealth and state levels. In this regard the PHRN has recognised that:
Australia has a federated health system. The country also has
high quality health data collections which can be used for planning and
research. However, because of the federated system, information about a
person’s lifetime health journey is collected and stored in many places. For
example, the States and Territories collect the birth, hospital and death data
and the Commonwealth collects the childhood immunisation, Medicare Benefits
Scheme (MBS), Pharmaceutical Benefits Scheme (PBS) and aged care data. In order
to compare national trends and to evaluate the effectiveness of health policy
for government and key policy decision makers it is necessary to be able to
link this information together and use it in a timely fashion.[21]
3.24
Professor David Preen from CAPHIA also noted how linking of Commonwealth
and state health datasets can provide a sound evidence-base to government
policy development:
Critically, the Commonwealth and state linked data provides
for really robust, evidence based decision making in health care that can
benefit not only the health system but also, ultimately, health consumers
across the country. Also, we know it can be done effectively because there have
been a lot of precedents over the last 10 years where people have used
Commonwealth and interstate data successfully for research to address a number
of issues that have directly informed government policy, at both a state and
federal level.[22]
3.25
The Centre for Big Data Research in Health spelt out some of the
beneficial health outcomes that could flow from an evaluation based on linking Commonwealth
and state datasets:
Data linkage, across national and state collections provides
a platform for answering questions about access to, and outcomes of, population
and individual health interventions, surveillance of disease and mortality,
health system performance, policy impact and economic analysis. Put simply, it
allows us to identify high-risk and low-value health services and high-risk
population subgroups, and transfer this knowledge into evidence-based policies.[23]
Examples of linked datasets
3.26
The committee received many examples of past, current or potential data
linkage projects which strongly point to the benefits of the technique. One
outstanding example of linked Commonwealth-state datasets was provided by
researchers from the School of Public Health and Community Medicine at UNSW. Dr
Heather Gidding and her colleagues are linking two Commonwealth datasets, the
Australian Childhood Immunisation Register (ACIR) and the National Death Index,
to de‑identified health data from 1.8 million children across New South
Wales and Western Australia to identify specific populations at risk of
preventable diseases:
The ACIR alone is a significant resource, being one of only a
handful of national population based immunisation registers. However, there is
insufficient information on ACIR about each child to identify specific
sub-populations at greatest risk of preventable diseases. Our study brings
together a wealth of routinely collected information about each child to
produce the first population-based estimates of effectiveness for vaccines
continuing to cause morbidity in Australian children, a strategic priority area
in the recently released National Immunisation Strategy. It is also the first
population-based record linkage study in the world to provide estimates of
vaccine impact in an indigenous sub-population, and identify specific maternal,
infant and demographic characteristics of infants with delayed vaccination. We
anticipate our findings will provide assurance of the effectiveness of
vaccination in all Australian children and strengthen the rationale for
improving vaccination timeliness, by quantifying its impact on disease burden.[24]
3.27
The Department of Health is also leading a data linkage trial labelled
the “Better targeting of mental health services” project:
The project will explore apparent disparities in provision of
mental health services and assist in better targeting these services. The
project is being undertaken in collaboration with the Australian Bureau of Statistics
using already linked MBS and PBS data with the 2011 Census of Population and
Housing data. The proposed demonstration project aims to conduct further data
linkage by expanding the dataset and using more sophisticated analytical
techniques to help explain the drivers of these disparities and, if
appropriate, potential targets for policy intervention. A report detailing
findings of the project, including both implications for mental health policy
and implications for public sector data management, will be completed in early
2016.[25]
3.28
The committee understands from the Department of Health that as at 13
April the report is yet to be finalised.
3.29
The CSIRO's submission provided a highly-practical example of a data
linkage project designed to improve the efficiency of our hospital system:
Our Patient Admission Prediction Tool (PAPT) shows how the
use of routinely collected administrative data can be used to make hospitals
more efficient. PAPT uses a model built on historical data to forecast the
number of patients to present at an emergency department within a certain time
and the number that will go on to need admission to the main hospital wards.
Access to reliable public health care is a key foundation to Australia's social
and economic well-being. PAPT is designed to make improvements in resource
allocation efficiency, reduce waiting times, and increase timely access to care
and is now used in a number of Queensland Hospitals and is undergoing its first
trial in a Victorian Hospital...
This important study required linking data across emergency
department and hospital data sets. Sets from member hospitals of The Health Roundtable
were provided, in accordance with their academic policy for use of collected
data for research purposes...[26]
3.30
The CSIRO's submission went on to explain how the PAPT project could be
improved through access to Commonwealth hospital datasets:
Although based on 12.5 million ED records, 11.6 million
inpatient episodes and 46000 hospital deaths, the [PAPT] coverage represents
only 79% of Australian tertiary hospitals and 40% of all Australian Hospitals.
Improving access to hospital datasets held by the Commonwealth for quantitative
analysis can avoid such limitations and ensure important policy decisions are
based on the most comprehensive data available.[27]
3.31
Finally, the Australian Commission on Safety and Quality in Health Care
provide a future linked health data scenario which would enable the efficient
monitoring of actual care and inform safety and quality improvement:
National guidelines for the management of acute coronary
syndrome (ACS) specify that patients admitted to hospital for management of ACS
be discharged on five medications...
However, studies show compliance with this guideline has been
shown to vary across hospitals and hospital types highlighting this as an area
for potential improvement.
Linking admitted national patient datasets...to PBS datasets
using anatomic therapeutical chemical codes would allow analysis of adherence
to national guidelines and variation from best practice, and provide valuable
information for improving care of patients with ACS. Similar analyses could be
conducted to monitor guideline compliance by healthcare facilities for a range
of other conditions including recommended stroke discharge medicines.[28]
Committee view
3.32
The committee is greatly encouraged by the strong interest expressed by
government agencies, data linkage organisations and medical researchers, in the
potential for improving evidence-based health policy development through data
linking. There is clearly a wealth of worthy health policy proposals and
evaluations that will commence once access to de-identified administrative data
is more readily available. The novel insights that will flow from these
projects will not only enable the development of innovative, evidence-based and
more cost-effective health policy, it will also lead to better patient outcomes
and improve the standard of healthcare in Australia.
3.33
The committee is however concerned by aspects of the Health Department's
publicly stated approach to big data. Although the department supports the
government's more open data policy, and also the recognises the significant
potential of big data, it appears to be taking an intentionally slow approach
to utilising data linkage in developing new health policies:
The use of Big Data technologies and analytics will be one of
the focus areas in a broader activity that the Commonwealth Department of
Health has just commenced to develop more comprehensive health analytic
capabilities.
Initially the use of Big Data technologies will supplement
the existing technology in the Department such as the Enterprise Data Warehouse
(which provides for secure storage of health data for use across a range of
health portfolio agencies) and the business intelligence Health capabilities. Later
stages may look at how a broader range of health data can be consolidated to
develop deeper insights into the impact of health policy initiatives.[29]
3.34
This approach appears at best ambivalent, and at worst contrary to the
government's public data policy statement which declares that 'Australian
Government entities will...securely share data between Australian Government
entities to improve efficiencies, and inform policy development and
decision-making...'[30]
3.35
The committee also notes the Department of Health has delayed its report
into the data linkage project to better target mental health services. This further
demonstrates that the department is not adequately prioritising important data
linkage projects.
3.36
With Commonwealth funding of $63 billion per annum at stake and recognising
the importance of improving the health outcomes of all Australians, the
committee urges the Department of Health to significantly increase its focus on
data linkage.
Recommendation 2
3.37
The committee recommends that the Department of Health, as a high
priority, actively explore and then implement measures to advance cost‑effective,
evidence‑based policy development through the use of data linkage.
Recommendation 3
3.38
The committee recommends that relevant government departments should
include information in their annual reports which describes the processes and
projects being undertaken to establish evidence‑based policy based on
data linkage as well as strategies they have adopted to contribute to the
government's pubic data policy.
Navigation: Previous Page | Contents | Next Page