Proposed SA Data Linkage Model

1
Proposed SA Data Linkage Model

• Andrew McAlindon
• Manager Primary Health Care Information Systems
  Modelling

2
Background

• Difficulties in negotiating and obtaining data
  (esp. MBS PBS data)
• Need to renegotiate privacy/ confidentiality
  principles for each new project
• Patient consent is often difficult to obtain for
  retrospective studies.
• Technical difficulties
• Time consuming
• Requires significant expertise
• Appropriate software required

3
Background

• Discussions between DHS and DHA commenced in
  early 2003 regarding the development of a
  Memorandum of Understanding (MOU) for the
  cross-jurisdictional linkage of administrative
  health data in the South Australian Health
  System.
• Both parties agreed in principle to proceed. The
  WA Data Linkage MOU would be used as a template

4
Background

• The MOU provides for
• The creation of a linkage key defining
  person-links between State and Commonwealth-held
  administrative health data
• Procedures for the extraction and supply of
  administrative health data sets covered by the
  Linkage Key.

5
Background

• Separate schedules would be negotiated and
  created for specific administrative data sets,
  eg
• MBS PBS data
• Residential Aged Care Data
• Linked data would be supplied to approved
  researchers as de-identified files for use in
  health services planning and research.

6
Background

• An investigation of local legislative/ privacy
  considerations was instigated
• A draft MOU was created and is currently under
  consideration by both parties.

7
Example
8
Example
9
Philosophy of Approach

• Maximise privacy
• Minimise use of name data
• Allow data custodians full control over the
  dissemination of de-identified data files
• Provide linked data files only to approved
  researchers involved in approved projects
• Provide only the minimal data required and
• Ensure that all copies of named data and linked
  data files are destroyed after use.

10
Model

• Stage 1
• Creation and updating of the Linkage Key using
  probabilistic matching methods (as described
  previously by Dr Bass)
• Stage 2
• Involves an undefined number of separate approved
  projects, each approved by officers of both the
  Commonwealth Department and SA DHS.
• Linked data from the two jurisdictions will be
  supplied directly to approved researchers by the
  appropriate database managers.

11
Model

• For each approved project, a unique set of
  project IDs will be generated and will provide
  the only way of combining the data files from
  different databases and jurisdictions into a
  single linked de-identified file.

12
Model

• Data managers retain oversight of the
  distribution/ use of data, as each project will
  need to be approved before proceeding.
• Linked data files will be provided only to the
  approved researchers working on each approved
  project, and will be destroyed when the work is
  complete.
• A different set of IDs will be used in each
  Project, making it difficult to perform an
  unauthorised merger of linked data.

13
Model

• The Linkage Key will be produced initially and
  updated by a small technical team specialising in
  probabilistic data matching,
• All people involved in the actual linkage and
  therefore requiring access to name-identified
  demographic data used in the linkage process will
  sign confidentiality agreements
• No other personnel will be allowed access to the
  files used in this process.

14
Model

• The linkage personnel will not be permitted to
  have any part in the analyses of the linked data,
  or to have any communication with the
  researchers.
• The Linkage Key will contain no actual data, but
  will contain coded keys to the data sets
  involved.
• Every custodian will supply the approved database
  records, together with a project ID, directly to
  the approved researchers.

15
Model

• The approved researchers will also sign
  confidentiality agreements. They will link the
  data together using the project ID, and will be
  the only people granted access to the
  de-identified linked information.
• They will be specifically forbidden to
  disseminate copies of the data files, and will be
  required to destroy these files on completion of
  the project.

16
Data sets

• Phase One Proposed Datasets
• MBS PBS (HIC)
• ISAAC (DHS)
• Emergency Visits Collection (DHS)
• Domiciliary Care Services (DHS)
• Births, Deaths Marriages (AGD)
• Cancer Registry (DHS)
• Coroners Data
• Childhood Immunisation Register

17
Data sets

• Possible Phase Two Datasets
• Residential Aged Care Data (DHA)
• SA Ambulance Data (SA Ambulance)
• RDNS Visits (RDNS)
• Community Health Information System (DHS)
• Mental Health National Minimum Dataset (DHS)
• Pregnancy Outcomes Collection (DHS)
• National Death Index (AIHW)
• SA Housing Trust
• Aboriginal Housing Authority

18
Progress

• Preliminary meetings between DHA and DHS have
  been held.
• A draft MOU has been prepared, based on the
  Western Australian agreement.
• Local legal and legislative issues are currently
  being investigated. No major pediments, but new
  legislation may be required.

19
Next Steps

• Complete investigation of legal/ legislative
  issues
• Gain support from
• DHS and DHA Executives
• DHS Ethics and Privacy Committee
• Determine model for Data Linkage Unit (contract
  WA Data Linkage Unit or create our own unit?)
• Negotiate schedules for specific data sets

20
References

• A Health Record Linkage Project on Diabetes,
  reported as a model of best practice in the
  Australian and New Zealand Journal of Public
  Health 2002 26 251-5.