Chapter 5
Issues and barriers for researchers
Introduction
5.1
During the course of this inquiry, the committee was alerted to the many
imperfections in Australia's data linkage system. This chapter examines a
number of the most commonly raised issues and makes a number of recommendations
aimed at progressing public health research in Australia.
5.2
This chapter examines issues that researchers and others may encounter
in the process of attempting to obtain data held by government agencies. In
particular this chapter will consider custodianship issues, the approvals
process, linkage and the desultory departmental approach.
5.3
Australia's health data holdings are currently dispersed. The dispersed
nature of the data means that the issues that follow are not a problem in all
parts of Australia or are necessarily problems for all agencies. They are
however, the issues that have been continuously raised with the committee.
5.4
Dr Merran Smith, Chief Executive of the PHRN conveyed the sentiment of
many of the researchers, musing:
...some of the problems with the Commonwealth that we are
facing at this point in time is that, in my experience over many years, the
Commonwealth government agencies often do not look outside. They are busy doing
what they need to do in Canberra. Perhaps they do not appreciate, or are not in
a position to leverage, the value and benefit that can be obtained from working
with other groups within the nation.[1]
5.5
Certainly there is a view in the research community that 'departments
tend to want to own that information and control it and use it for their
purposes as opposed to a public purpose'.[2]
5.6
Ms Helen Owens, an Assistant Secretary at the Department of the Prime
Minister and Cabinet, told the committee that the Prime Minister moved data
policy into his department to try and improve the coordination of government
data policy.[3]
5.7
As part of that effort, the Department of the Prime Minister and Cabinet
have produced a Public Sector Data Management report and a Public
Data Policy Statement. However, as Professor Kearney, Chair of the PHRN
identified 'it is a long and slow process and a lot more needs to be done'.[4]
Custodianship issues
Accessing data
5.8
The first step in any data linkage project is to determine whether the required
data is available and what permissions are required to access it. That process
can be more difficult than one might anticipate.
5.9
Dr Heather Gidding, a Senior Lecturer in the School of Public Health and
Community Medicine at the University of New South Wales observed that in
undertaking her data linkage project:
It took us a long time to find the data custodian for that
ACIR [Australian Childhood Immunisation Register] data. When we did find
someone who did understand the dataset, they are very hard to access. They are
overworked and hard to get a hold of. It certainly does make the research quite
difficult.[5]
5.10
The frustration and uncertainty about which department holds particular
data was explained by Professor John Glover, Director of the Public Health
Information Development Unit (PHIDU):
One of the difficulties with the datasets we cannot get is...who
owns which data, who the data custodians are. We go to the Department of Human
Services and put a case to them for health data. They have a committee. Health [the
Department of Health] sits on that committee and then basically decides whether
or not you can have the data. The DHS people do not really know anything about
the data and the policy issues; that comes from Health, and that decision is
then made and passed back by email. You then complain and you make some
comment, 'What about so and so, can we do this or that?, and then it goes back
again. The same with PBS.
It is very hard to engage totally with whoever the custodian
really is—in this case it is Health, even though you are dealing with the
Department of Human Services, who would run the tables.[6]
5.11
If researchers can work out which department holds the relevant data,
Dr Smith explained that there is no clear process to access data:
Part of the challenge with accessing the Commonwealth data is
that there is not an explicit process. When you are accessing state data, most
states have online an application form and an explanation for how you need to
go about it. But even now when you go online and try to find how to access
data, there is a little bit on the AIHW site, hardly anything useful on the
Department of Health's site or on the Department of Human Services' site. To
acknowledge the work of my colleagues in the Commonwealth, they certainly did
as a result of this process go through some conversations and streamline
processes, but it is still not really explicit in the way that a researcher who
says, 'Look, I think I want to use this Commonwealth data set' needs. It is not
really clear how they go about the process to get the approval.[7]
5.12
Part of the problem is that Australian Government departments are often unaware
of what data they hold and no list is publicly available.
5.13
This problem was highlighted when Ms Michelle Wilson, a General Manager
at the Department of Human Services was unable to tell the committee what data
the Department of Human Services held.[8]
However, the Department of Human Services were not alone. The committee placed
questions on notice to obtain a list of the health related data sets held by several
health-focussed Australian Government departments and agencies.[9]
Some departments required more than two weeks to compile their lists. The AIHW
provides a list in its annual report.[10]
The lists provided by the departments appear in Appendix 4.
5.14
Adding to the complication facing researchers of discovering what data sets
are held by each department, some departments hold data that are owned by other
departments. Ms Wilson explained:
The Department of Social Services are the data custodians of
[Centrelink] data, so while the Department of Human Services holds [the
Centrelink] data, we are not the custodians of it. It is the same as with the
Medicare MBS and PBS data that we collect in the course of our business: we are
not the custodians of that data; we are the holders of it.[11]
5.15
Insofar as the Department of Human Services holds data on behalf of
other departments, it considers itself to be a 'steward' of those data. Ms
Wilson explained that a data 'steward', is 'somebody who sees the data through
and ensures that it is well looked after and that the guidelines are adhered to'.[12]
5.16
If researchers manage to navigate the challenges in finding the dataset
and the custodian and obtain the relevant permissions, it is possible that the
department may not provide the data in the level of detail required for the
project. In this regard, Professor Glover explained to the committee:
...a major gap in the data we have relates to the Medicare
Benefits Schedule data and the Pharmaceutical Benefits Scheme data. We cannot
get it at what I believe is an appropriate geographic level—a level that will
show the major variations that occur within our cities and within our regional
and rural areas, particularly in variations in health outcomes, health service
youth and social determinants of health.[13]
Under-resourced
5.17
The committee heard from multiple witnesses that the data custodians
they dealt with were not sufficiently resourced to ensure that the data asset
could be maximised. For example Professor Sallie-Anne Pearson, Head of the
Medicines Policy Research Unit, Centre for Big Data Research in Health
acknowledged that:
One of the greatest challenges [custodians] face is that this
is incredibly under-resourced, and designated roles and having people responsible
to do this work would be a very pragmatic and significant advance in getting
some of those current logjams freed.[14]
5.18
Professor Louisa Jorm, Director of the Centre for Big Data Research in
Health echoed those comments:
I would second that. In fact, several data custodians have
said to me over the years: 'Providing data for research is not core business
for our agency.' It is viewed as an add-on... If the leaders of their agencies
are saying, 'This is your core business and we're going to resource it,' then things
will start to move.[15]
5.19
The committee attempted to gauge whether recent shifts in government
priorities may mean that appropriate resources will now be allocated to data
custodians.
5.20
Ms Foster, First Assistant Secretary with the Department of Health explained
to the committee some of the evolving processes within the Department of
Health:
In terms of access to data and better use of data, the
Department of Health has been through a number of different reviews that talked
about better use of data, better valuation, better research to inform policy
development towards better health outcomes. In fact, that was the reason the
division I head up was established. It was to try and make better use of data
and ensure better use of evaluation. Certainly, in my KPIs it would be a fairly
key feature that there be better use of data. In the department's vision...there
is that reference to better health outcomes and one of the key underpinnings of
that is that better use of research and data. So it really is quite widespread
throughout the department. There is a major cultural change that is occurring.[16]
5.21
However that statement was not reflected in the Department of Health's
Answer to a Question on Notice. When asked about the KPIs the Department of
Health uses to evaluate data usage the department provided the following
cryptic response:
The Department facilitates a Data Governance Council that
includes representation from the Department, the AIHW, the Australian Bureau of
Statistics, Department of Human Services and other Health portfolio agencies.
The Council is responsible for ensuring effective policies and governance for
the Department's approaches to data collection, management, interrogation,
sharing, access and release.[17]
5.22
The Australian Bureau of Statistics identified that it has a KPI
entitled 'Use of Microdata Increases'.[18]
The AIHW has measures relating to data tied into its Portfolio Budget
Statements.[19]
However, it appears that neither the Department of Health, the Department of
Human Services or the Department of Social Services have any KPIs relating to
data.[20]
The changing data culture
5.23
Professor Pearson pointed out that data custodians play an important
role in releasing health-related data:
The custodianship of the data is a very important position to
be placed in. I think traditionally custodians have been very risk averse.
There are different interpretations of legislation guidelines, et cetera, about
the nature and the type of data that can be released. I have actually seen a
massive change over the last few years where custodians are now thinking about
arguments as to why data should not be released...rather than thinking, 'We have
to stop this.'[21]
5.24
Professor Pearson also noted:
There clearly is an enormous responsibility on the part of
custodians to protect the identity of individuals, but the challenges that they
face relate to the interpretation of legislation, which can be different with
different individuals. The other great challenge that they face is that
oftentimes this role is done on top of a highly powered, high-level and highly
committed role. It is really about goodwill and trying to serve the research
community.[22]
5.25
Apparently a new culture, with regard to the use and release of
administrative data held by Australian Government departments, is in its infancy.
While change is occurring, the speed at which parts of the public service are
adapting varies. An example of the more cautious approach was exemplified by Ms
Foster of the Department of Health:
...the whole senior executive [are] helping to drive that
cultural change to make data available.
But having said that, while we are moving from that attempt
to make data as available as possible to researchers and for policy development
purposes there is that very key issue of privacy that we are concerned about. I
think that for many years there was that culture, 'We must absolutely protect
this data at all costs.' But, of course, as techniques—computing and
statistical techniques and all sorts of things—get more sophisticated there are
more ways to 'perturb' the data, as I think you referred to it, or to
confidentialise the data so we can actually protect people's privacy and still
be able to make information available for use by researchers.[23]
5.26
Ms Foster continued:
There are very strict guidelines under the National Health
Act, the Health Insurance Act, the privacy guidelines and the Privacy Act. We
also observe those provisions very strictly. Indeed, sometimes those rules can
limit our own potential to use data internally. But we do have to be very
conscious of those privacy rules to ensure that people's information cannot be
identified.[24]
5.27
Ms Foster's answer can be starkly contrasted with the dynamic response
of Mr David Dennis, a Branch Manager with the Department of Social
Services who submitted:
We have endeavoured, especially in recent times, to be at the
vanguard of testing technologies and new techniques and developing new
techniques to allow more data to be released to intelligent sources...
Internally, we have worked with DHS to streamline our approval processes and,
in so doing, have devolved a lot of routine approvals to DHS directly so that
they no longer need to seek recourse to DSS where routine approvals are
involved, so it moves a lot more quickly.
Where we have been successful in this regard is by not
treating every data application equally. In so doing, we have published to
data.gov.au large amounts of information that is now open, albeit in a
perturbed or confidentialised manner. That has allowed us to turn our minds to
what we see as the current challenge...of arranging for researchers to have
access to individual unit record data in a safe and appropriate environment,
because this is what they need to perform the statistical analyses that they
want to perform. We have trialled a number of models to facilitate that...using
innovative products...such as the Secure Unified Research Environment... so that
finally we can put an enduring dataset of DSS data in a safe place that
credentialled individuals can access directly at the unit record level and
where they can withdraw the products of those analyses, and we are assured that
the data is safe and the researchers get the access that they need...[25]
5.28
The Australian Bureau of Statistics also appears to be embracing the new
data culture. The Australian Bureau of Statistics now operates a number of
linked data projects including the Mental Health Services-Census Integrated
Dataset.
5.29
The Mental Health Services-Census Integrated Dataset is a good example
of linking diverse data sets to provide a solid evidence base. It combines 2011
Census data with mental health related items from the MBS and PBS along with
demographic data. The Australian Bureau of Statistics is currently
investigating adding state and territory hospital data to this dataset to
enable 'a more complete picture of the patterns of service usage'.[26]
Approvals processes
5.30
Various approvals to access data and conduct research are necessary to
ensure that the data being released are used for the public benefit and that
appropriate measures are being taken to ensure the privacy and security of
personal information.
Ethics approvals
5.31
To ensure that health information is used appropriately, delegated
legislation provides that researchers must undertake ethics approval if they
wish to use health data.[27]
5.32
Ethics approvals play an important part in upholding high ethical
standards in research. However, as Dr Gidding told the committee, having to satisfy
multiple approvals can add substantial time to a research project:
I am leading a team of researchers in New South Wales and
Western Australia that predominantly have linked up the Australian Childhood
Immunisation Register, the vaccination records of Australian children to
records for children in both WA and New South Wales—1.8 million children... We
had to get nine approvals for our project to go ahead, which also included six
ethics applications that took two years.[28]
5.33
The Australian Commission on Safety and Quality in Health Care (ACSQHC)
endorsed the idea that multiple ethics approvals hindered data linkage:
One of the main issues is...[the] complicated ethics approvals
systems in the various jurisdictions. In Australia, projects using linked data
must be approved by the data linkage unit; the data custodian responsible for
each data set; and one or more Human Research Ethics Committee(s)... In order for
data on safety and quality to be most effective in driving improvement, regular
and timely review is required.[29]
5.34
Dr Nicky Antonius, Acting Assistant Secretary of the Department of
Health and Mr Warren Richter, Head, Chief Information Officer Group, AIHW both told
the committee they would not request an additional ethics approval where the
researcher had already obtained an approval that met the research ethics
standards of the National Health and Medical Research Council (NHMRC). However,
ethics approvals from other jurisdictions were more problematic. This continues
to be a problem even within Australia where Commonwealth departments will not
accept the ethics approval of a State, such as Victoria.[30]
5.35
In his evidence to the committee, the Acting Australian Information
Commissioner Mr Timothy Pilgrim questioned whether ethics approvals were
necessarily the best mode of protecting privacy:
I have heard researchers remark regularly about some of the
challenges around ethics committees, and there may be a good argument for
looking at other mechanisms.[31]
Delays
5.36
Once a researcher has obtained ethics approval, the committee received
evidence that there have often been significant delays in obtaining data from departments.
5.37
In 2009 Professor Fiona Stanley and Emeritus Professor D'Arcy Holman wrote
a letter to the then Secretary of the Department of Health and Ageing,
Ms Jane Halton. The letter notified the Secretary:
Of the 23 research projects [that had applied for
Commonwealth held health data], only nine have received useable Commonwealth
data. All nine have experienced lengthy delays in obtaining data, some as much
as two years from the time of application. Those still to receive data have
been waiting for as long as three years. It would be fair to say that the
affected researchers have found it extremely difficult to engage with the
Australian Department of Health and Ageing in progressing a solution to these
delays and solving attendant issues such as data queries and explanations.[32]
5.38
Similarly when Professor David Preen, Deputy Head of the School of
Population Health at the University of Western Australia, was asked about the
source of delays in cross-jurisdictional linkage, he responded:
The approval processes through the Commonwealth were slow... [In]
a number of cases we had queries or had made submissions but did not hear back
for a period of time, and we would follow up. But it was not clear to whom we
should be following up with... But, there being delays, what we could do to
facilitate that and what else may needed to provide in terms of information, et
cetera, was very unclear. That, in itself, caused further delays. Whether that
was due to inaction or a lack of transparency with the existing processes, I
guess, is debatable. But the end result is a huge delay in provision of these
data.[33]
5.39
It is important to note that there is a large financial cost to
researchers from these delays. In particular, the lengthy delays have led to a significant
wastage of public research funds.[34]
5.40
The degree of wastage was articulated by Emeritus Professor D'Arcy
Holman who estimated that in 2008:
...over 100 medical researchers waiting for [Commonwealth]
data, had $11 million in precious public research funds, mostly from
NHMRC, simply going to waste.[35]
5.41
The problem of significant delays in obtaining data is not limited to
researchers. It also affects government agencies. In its most recent annual
report the AIHW explained the impact of the delays in producing its reports:
This was because we did not gain access to records from the
Medical Benefit Schedule and the Pharmaceutical Benefits Schedule as expected,
which meant that the committee [the AIHW Ethics Committee] did not receive
applications from external researchers for ethical clearance of projects
involving use of these particular data sets.[36]
Linkage
Linkage logjam
5.42
The committee heard that one of the greatest impediments to the widespread
use of linked data is the lack of facilities accredited to link Commonwealth
data.
5.43
Professor Louisa Jorm described the current situation:
...the bottleneck is our limited capacity for data linkage.
Currently only the Australian Bureau of Statistics and the Australian Institute
of Health and Welfare are accredited as integrating authorities for
Commonwealth health data, and yet we have state-of-the-art health data linkage
facilities operating in states and territories, including New South Wales,
South Australia and the Northern Territory, and Western Australia. National
capacity for data linkage would be dramatically boosted if these state and
territory based units were accredited to link Commonwealth as well as
jurisdictional data. In particular, it makes a lot of sense for these units to
service projects that involve linkage of data from a single jurisdiction with
Commonwealth data, rather than these projects adding to the linkage logjam and
linkage queue that currently exists at the Australian Institute of Health and
Welfare.[37]
5.44
Similarly, the Centre for Big Data Research in Health strongly advocated
for State and Territory data linkage units to be accredited by the
Commonwealth:
Australia has world‐renowned
health data linkage facilities yet relatively few of these are accredited to
link and integrate Commonwealth data collections. Health data linkage
activities have increased substantially over the last decade and this activity
will continue to grow in the coming years. The Commonwealth data linkage
facilities based at the ABS and AIHW are not necessarily equipped to manage
this growth and are not always an efficient option for projects involving
linkage of Commonwealth data with collections from one or more States. One
solution is to let others with proven track records do this too.[38]
5.45
According to the National Statistics Service, the only Integrating
Authorities accredited to link Commonwealth health data are the ABS, the AIHW
and the Australian Institute of Family Studies.[39]
5.46
Dr Smith advised the committee that attempts to have State data
linkage units accredited have been unsuccessful:
One of the state linkagers has sought accreditation to be
able to receive Commonwealth data and the view from the Commonwealth was that
they could not accredit it because it was a state agency, so they did not have
jurisdiction.[40]
Cost of linkage for researchers
5.47
The committee heard evidence that data linkage can be a costly barrier
to the development of new knowledge and innovation.
5.48
Some Australian Government departments charge researchers on a cost
recovery basis to access the data for linkage projects.[41]
In some cases, the amount charged is cost-prohibitive. SA-NT Datalink provided
two examples:
For one project with a cohort of about 10,000 individuals and
linking 4 datasets, the SA-NT DataLink cost was estimated at $10,000. The
researchers also wished to link to a Commonwealth dataset for which they quoted
approximately $160,000.
Because of the high Commonwealth costs, the researchers could
not include this data.
Another project with a cohort of about 240 individuals and
linking 4 datasets, the SA-NT DataLink cost was estimated at $8,500. The
researchers also wished to link to another Commonwealth dataset for which they
quoted approximately $40,000.
Again, because of the high Commonwealth data costs, the
researchers could not include this data.[42]
5.49
Then linkage units can also charge researchers depending on the work
performed. Officials at the Australian Institute of Health and Welfare told the
committee that they operate on a cost-recovery basis:
As you may be aware, the Australian Institute of Health and
Welfare receives about 30 per cent of its funding from appropriations, so 70
per cent of our revenue comes from the provision of goods and services to
others. We run our data-integrating authority and data-linkage services on a
cost-recovery basis. That is, essentially, the cost of a salary plus the
overheads associated with running buildings... There are other organisations involved
in the chain. If you have a state-linkage organisation, a Centre for Health
Record Linkage in New South Wales, they are also running on a cost-recovery
basis, so the researchers have to pay the costs that are incurred to deliver
the service they are asking for.[43]
5.50
When the committee asked NACCHO [the National Aboriginal Community
Controlled Health Organisation] what it cost to link data through the AIHW Dr
Robert Starling replied 'it is in the tens of thousands of dollars—it is not 50
bucks here or there.'[44]
5.51
In response to a question on notice, the AIHW advised that the average
cost of a linkage project in 2015 was $4414 inclusive of GST.[45]
5.52
The ABS notified the committee that the average fee of obtaining Custom
Data Information was $1970 but that average cost of using Information
Consultancies carried out by the Health subject matter area was $4645 inclusive
of GST.[46]
5.53
Researchers may also be charged to use laboratories such as the SURE
research facility to analyse data. Mr Wells, the Deputy Chief Executive Officer
of The Sax Institute informed the committee:
On the researcher side, we do charge researchers. But that
charge does not reflect the full cost, because we have a subsidy through NCRIS
funding. So we charge them a charge which is not the full cost of providing it.
But, certainly, it is a very costly facility to run. I am not saying that is a
bad thing, but it is costly. We are constantly looking at enhancing its
efficiency et cetera, but there is very much a cost to operating and providing
this service.[47]
5.54
One strategy that has the potential to bring down the cost of linkage is
the creation of enduring linked datasets.
Enduring data linkage
5.55
At the end of a linkage project there is a legal requirement that
researchers must destroy the links between the datasets as Mr Crettenden,
Assistant Secretary, Department of Health explained:
The Privacy Commissioner's guidelines...say that Medicare...and
the [Department of Health] are only able to link the datasets under very
specific circumstances and that the dataset that is used to create the linkage
must be destroyed within one month of it being created.[48]
5.56
However that creates problems for ongoing research according to Dr
Smith:
If you go to a lot of effort to create a high-quality linkage
and then, after the project is finished, you destroy the links and start doing
the same thing again for the next project, first of all, it is highly
inefficient; it is both expensive and time consuming. Second, it does not lead
to improvements in linkage quality. When we talk about the importance of
enduring linkage, that is why it is important.[49]
5.57
The benefits of being able to maintain a linked dataset is that the data
can be accessed very quickly and it can become a permanent resource that
researchers can use.[50]
5.58
Australia has created enduring datasets before as Professor Preen recalled:
It was a formalised and accepted process for linkages
de-identified unconsented or non-consented Medicare, PBS and state data. In WA
a memorandum of understanding was developed and signed by the Commonwealth and
by the state government to set up what was referred to as the
cross-jurisdictional linkage facility to specifically link the projects after
approval. These data would be linked in unconsented form but they would be
provided in de-identified form for researchers. My understanding is that 25
projects went through, were approved and received data through that process.
That system was functional, and the fact that the MOU was agreed to and signed
means that it should be a possibility, obviously, within the existing legal
framework. That has now been replaced because of some of the issues with the
layers, with integrating authorities.[51]
5.59
Departmental officials indicated that they were currently undertaking an
ambitious Multi-agency Data Integration Project (MADIP). The MADIP aims to
combine 'information from existing surveys, administrative collections and
censuses' to develop 'a more complete picture of the circumstances of
individuals, households and businesses'.[52]
The MADIP 'brings together, for the first time, Census data with administrative
data on health, income, and social security payments'.[53]
5.60
Unlike the 'link and destroy' model described above, MADIP is going to
be an enduring linkage.[54]
5.61
Promisingly, there is a commitment amongst several key government
agencies to move towards greater use of enduring data sets. Ms Gemma Van
Halderen, the General Manager of the Strategic Partnerships and Projects
Division at the Australian Bureau of Statistics told the committee:
...the Commonwealth secretaries and some dep secretaries across
the Australian Bureau of Statistics, Department of Human Services, Department
of Social Services and the Department of Health have been working...since about
2009 on some Commonwealth arrangements for data sharing and data use. The
secretary commissioned a review of those arrangements in 2014. That review was
undertaken by a consultant, Ms Susan Linacre, who used to be a deputy at
the Australian Bureau of Statistics. She recommended to the deputy secretaries
and the secretaries that the Commonwealth should move away from a
link-and-destroy model to a create-and-reuse-and-keep model, so we are
currently moving towards that... There is a very strong appetite within the
Commonwealth agencies and the members of that data integration initiative to
put in place a research create-and-reuse-and-keep model, not a
create-and-destroy model for data linking.[55]
The desultory departmental approach
5.62
As the above evidence reveals, there is no consistent policy between
Australian Government departments in their approach to releasing data to
researchers.
5.63
Part of the reason for the haphazard approach may be attributed to the lack
of coordination of government policy in this area. Ms Owens, Assistant
Secretary, Department of the Prime Minister and Cabinet informed the committee
that her department is responsible for the coordination of this area of policy:
Our role is really the coordination of the policy element. We
would still rely upon our colleagues here at the table to do the actual coordination
within their own thematic areas. We are not actually going to do the doing in
this space...[56]
5.64
The problems identified in this report highlight the failures of some
departments to adapt to a new, more flexible approach to releasing
de-identified data. Those failures must be addressed and rectified before
Australia can make the most of its big data potential.
Release of information between
government departments
5.65
The complexity of the Australian Government's data arrangements were
reinforced when Senator Moore tried to ascertain what data was held by the
Department of Human Services:
Senator MOORE: I am interested to know whether the
human services data includes only Medicare data or whether it has Centrelink
and DVA [Department of Veterans' Affairs] data as well...
Ms Wilson: ...The Department of Social Services are the
data custodians of that data, so while the Department of Human Services holds
that data, we are not the custodians of it. It is the same as with the Medicare
MBS and PBS data that we collect in the course of our business: we are not the
custodians of that data; we are the holders of it.
Senator MOORE: Are you custodians of anything?
Ms Wilson: We are custodians of the data about how
our customers behave in our service delivery systems.[57]
5.66
Dr Nicky Antonius, Acting Assistant Secretary at the Department of
Health also tried to assist the committee explaining:
If DHS [the Department of Human Services] was asked to
divulge data to AIHW [Australian Institute of Health and Welfare] that would
depend what dataset and who is the custodian of that dataset. For MBS, PBS and
Medicare enrolments data it would be the Department of Health but we are the
ones who will pursue the public interest certificates and then share that with
the Department of Human Services to forward it to the requester. If the
requester is asking for social services data then they would be directed to the
Department of Social Services.[58]
5.67
Mr Warren Richter of the AIHW and Ms Michelle Wilson from the Department
of Human Services told the committee that the process of releasing data between
Australian Government department could be very involved:
Mr Richter: We have a schedule agreement that was
signed last week with the department [Department of Human Services]. We have a
public interest certificate which is in the process of being signed within the
department and we have arrangements in place with the Department of Human
Services to receive the Medicare enrolments data, which we need to do the
linkage—to identify the links. As soon as that public interest certificate has
been signed—I do not know if you know, Michelle, but we have your people all on
tap to hit the button—
Ms Wilson: We are looking at the piece of equipment
that is holding the data to make sure it is ready to go. We will wait until the
public interest certificate has been tied up with a bow and signed by
everybody. It is going through the final processes, as is the multi-agency data
integration project. The public interest certificate is also at that final
stage of processing. Each of our departments has to go through its own legal
processes.[59]
5.68
That somewhat confusing state of affairs was compounded by Ms Wilson who
went on to argue:
We [the Department of Human Services] relied on a public
interest certificate and a collaborative process to look at the privacy and
secrecy issues with the Department of Health. Indeed, with the Department of
Social Services data that relates to social security information we also have a
collaborative process with the Department of Social Services. So it is
generally not just up to the Department of Human Services about how the data is
used. In fact we cannot really make decisions ourselves about how that data is
used.[60]
5.69
The Public Sector Data Management Report provides additional
examples of current inefficiencies:
In one case where an important dataset was being linked with
the Census, the negotiations on the MOU [Memorandum of Understanding] took up
to 18 months while the linking only took two weeks.[61]
5.70
And:
An agency reported having up to 11 MOUs with the same
department to access data.[62]
5.71
Accordingly, the Public Sector Data Management Report concluded:
Overall, the lack of data sharing prevents feedback on policy
and hinders the potential of data to improve future service delivery.[63]
Committee view
5.72
The committee is deeply concerned by the lack of transparency regarding
the data held by government departments. If Australia is to maximise the value
of its data resources, researchers need to know what data sets are available
and how to access them.
5.73
During the course of this inquiry it became obvious that some
departments were uncertain about what datasets they held. In some cases, when
asked to provide a list to the committee on notice some departments required two
weeks to list their data holdings.
5.74
It ought to be clear to both researchers and departments what data each
department collects and what area within the department is responsible for data
custodianship. The Public Data Policy Statement, the Public Sector
Data Management Report and a chorus of witnesses all agree that
non-sensitive data should be publicly available as a right and that sensitive
data should be able to be accessed subject to appropriate privacy and security
constraints. For this purpose, a list of data sets held by each department
ought to be publicly available on their websites and on data.gov.au. Departments
also ought to publish a clear statement that identifies exactly how requests
for data are made and how those requests will be dealt with.
Recommendation 6
5.75
The committee recommends that each Australian Government agency develop
and maintain on its website a list of datasets held by the agency along with
the contact details of the data custodian. This list should be updated at least
twice annually.
Recommendation 7
5.76
The committee recommends that all datasets held by the Commonwealth be
listed on www.data.gov.au, identifying which agency is the data custodian.
Recommendation 8
5.77
The committee recommends that each Australian Government agency that is
a data custodian develop and publish on its website guidance for researchers detailing
its process for data requests and approvals.
5.78
In certain departments the current system of approving access to data is
at times chaotic. The evidence demonstrates that seeking multiple ethics
approvals was time consuming without necessarily being of significant
additional benefit.
5.79
The committee is of the view that there ought to be a consistent method
of applying for approval to access sensitive data. The committee notes that
ethics approvals of other states are currently not recognised by Australian
Government departments. Such approvals ought to be nationally recognised to ensure
consistent and systematic application of process. Therefore, the relevant
departments need to work at both the Commonwealth level and with the States and
Territories.
Recommendation 9
5.80
The committee recommends that the government take a whole-of-government
approach to streamlining the ethics approval process and the authorising
environment in consultation with the Privacy Commissioner, privacy advocates,
the NHMRC, data custodians, academics, consumers and the States and
Territories. The government should also work with the States and Territories to
establish a national accreditation system so that ethics approvals from
accredited jurisdictions are recognised by the Commonwealth.
5.81
The evidence received by the committee demonstrates that data custodians
are currently not being adequately resourced. Furthermore, it was clear from
the lack of KPIs in most government departments that the responsibilities of
data custodianship is not a priority for departments despite the considerable
benefits that would flow from a sharper focus.
5.82
The burdens placed on custodians have undoubtedly made them more
hesitant to release data. The relevant departments need to empower the officials
in these positions and give them the resources and confidence to be able to
release datasets where researchers have obtained the approval that demonstrates
their ability to maintain the security and privacy of the data.
5.83
The committee was shocked to discover that precious Commonwealth funded
grants for important medical research projects were going to waste because
Australian Government departments would not provide the data required. The
committee believes that the $11 million figure cited by Emeritus Professor
Holman is highly conservative.
5.84
The committee notes that the Productivity Commission and others have
urged government to make de-identified datasets more readily available to
researchers. Where data has been de-identified, open access ought to be the
default position.
5.85
It is absolutely unacceptable that researchers should have to wait years
to access data to facilitate important research projects in the public
interest. This is a situation that requires immediate attention.
5.86
Therefore the committee makes the following recommendations aimed at
prioritising data access and encouraging appropriate open access.
Recommendation 10
5.87
The committee recommends that relevant government agencies give greater
priority to, and adequately resource, their data custodians.
Recommendation 11
5.88
The committee recommends that relevant government agencies provide
guidance to data custodians to assist them in their decision-making, with a
view to making more de-identified data available on an enduring basis.
Recommendation 12
5.89
The committee recommends that the government adopt the Productivity
Commission's proposed principle that open access to de-identified datasets
should be the default position.
Recommendation 13
5.90
The committee recommends that the government should direct relevant
agencies to release de-identified datasets on an enduring basis as the default
position.
Recommendation 14
5.91
The committee recommends that departments that have data custodianship
responsibilities must establish and publish realistic Key Performance
Indicators for the timely consideration and approval of datasets requests.
These departments must publicly report on their KPIs in their annual reports.
If after 5 years departments continue to delay the release of
datasets, then the committee recommends that the government establish binding
timeframes for processing applications for data. Failure to comply with the timeframe
should trigger appeal rights similar to those found in other information access
regimes.
5.92
Currently, the release of various datasets between Australian Government
departments and agencies at times appears shambolic. The evidence provided to
the committee indicates that several departments, despite the imprimatur of the
new data policy, were very protective of their data and demonstrated great
reticence to release it to other government agencies. Whilst the committee
understands and supports the need to protect privacy concerns and to act in
accordance with legislation, the current practices are inefficient and stifle
innovation.
Recommendation 15
5.93
The committee recommends that Government encourage collaboration on data
linkage projects between government agencies, as well as academia and industry
to provide for evidence-based policy development and facilitate research that
is undertaken in the public interest.
5.94
The committee is confident that increasing the availability of linked
data will make Australia a more attractive research destination that produces
significant public health research and better evidence-based policy proposals.
5.95
To ensure that researchers can be supported into the future,
consideration ought to be given to the linkage units. The committee heard that
some units were almost already at capacity and that there was at the very least
the potential for a 'linkage logjam'.
5.96
Australia has a number of excellent linkage units at the state level,
some of whom would like to be accredited to link Australian Government data. The
government should give the prospect serious consideration.
5.97
The committee also notes that depending on the nature of the project
some researchers may be charged multiple times (potentially at the Commonwealth
and state levels) for accessing various services in order to facilitate that
research. To ensure that the medical research sector is as vibrant as it can
be, the government ought to consider the cumulative costs incurred by
researchers across the linkage project.
Recommendation 16
5.98
The committee recommends that government consider accrediting State data
linkage units to link Commonwealth data with State data collections, subject to
comprehensive privacy and security protocols.
Recommendation 17
5.99
The committee recommends that the Government review the cost of data
access and linkage work undertaken by Commonwealth entities with a view to
facilitating research and innovation in the national interest.
5.100
Australia has the capacity to be a world-leader in this field. For the
benefit of the health of all Australians the committee strongly urges the
government to implement all recommendations to ensure that Australia realises
its big data potential.
Senator Deborah O'Neill
Chair
Navigation: Previous Page | Contents | Next Page