Introduction
1 The Mental Health Services-Census Data Integration project combined data from the 2011 Census of Population and Housing with a subset of data from the Medical Benefits Schedule (MBS) and Pharmaceutical Benefits Scheme (PBS). De-identified transaction information from the MBS and PBS was transformed to person-level information. Probabilistic linkage techniques were used to combine this information with person-records from the Census to create the Mental Health Services-Census Integrated Dataset, 2011.
Data
2 Data were produced using the following sources:
- 2011 Census of Population and Housing. The 2011 Census measured the number and key characteristics of people who were in Australia on Census night 9 August 2011. For information about the 2011 Census please refer to Census 2011 Reference and Information and Census Data Quality on the ABS website.
- Medicare Benefits Schedule data. The Department of Human Services collects data on the activity of all persons making claims through the Medicare Benefits Scheme and provides this information to the Department of Health. Information collected includes the type of service provided (MBS item number) and the benefit paid by Medicare for the service. The item numbers and benefits paid by Medicare are based on the Medicare Benefits Schedule which is a listing of the Medicare services subsidised by the Australian Government. The Mental Health Services-Census Integrated Dataset includes those MBS subsidised mental health-related services as defined in Appendix 1;
- Pharmaceutical Benefits Scheme data. The Department of Human Services provides data on prescriptions funded through the Pharmaceutical Benefits Scheme to the Department of Health. The PBS lists all of the medicines available to be dispensed to patients at a Government-subsidised price. The Government is advised by the Pharmaceutical Benefits Advisory Committee (PBAC) regarding which drugs should be listed on the PBS Scheme. The Mental Health Services-Census Integrated Dataset includes those PBS subsidised mental health-related medications as defined in Appendix 2.
Scope
3 The scope of the data is restricted to persons who responded to the 2011 Census of Population and Housing and who accessed subsidised mental health-related items listed on the MBS or PBS datasets in 2011. Data excludes:
- persons whose Census record indicated that they were an overseas visitor on Census night;
- persons who were out of the country on Census night; and
- persons who were not enumerated in the 2011 Census.
4 Data also excludes:
- persons who received services provided by hospital doctors to public patients in public hospitals, or services that qualify for a benefit under the Department of Veterans' Affairs National Treatment Account;
- the Repatriation Pharmaceutical Benefits Scheme which is subsidised by the Department of Veterans’ Affairs;
- persons who were supplied medications or accessed services only through programs that do not use the Medicare processing system; for example, Aboriginal and Torres Strait Islander Health Programmes;
- persons accessing private prescription drugs, over the counter drugs, and drugs that cost less than the co-payment.
5 These exclusions are discussed further in the Data Quality section.
Linkage results
6 At the completion of the linkage process:
- 1,072,284 person-records (69.6%) of the 1,540,836 person-records on the MBS dataset were linked to the 2011 Census;
- 1,669,278 person-records (70.9%) of the 2,354,118 person-records on the PBS dataset were linked to the 2011 Census; and
- 2,279,863 person-records (70.6%) of the 3,226,815 person-records on the combined MBS/PBS dataset were linked to the 2011 Census.
Methodology
Overview of data integration
7 Statistical data integration involves combining information from different administrative and/or statistical sources to provide new datasets for statistical and research purposes [1].
8 Data linking is a key part of statistical data integration and involves the technical process of combining records from different source datasets using variables that are shared between the sources. Data linkage is typically performed on records that represent individual persons, rather than aggregates. Two common methods used to link records are deterministic and probabilistic linkage. Deterministic linkage links person-records on exact matches using a unique identifier (such as a social security number or a created unique identifier such as a linkage key). Probabilistic linkage links person-records on close matches based on the relative likelihood that two records refer to the same person, using a number of linking variables (such as date of birth, sex, geographic area).
9 For further information on data integration see Glossary and the National Statistical Service website – Data Integration.
Integration method
10 The Department of Health provided the ABS with de-identified MBS and de-identified PBS data extracts, while the Department of Human Services extracted and provided the associated de-identified demographic data extract on behalf of the Department of Health. This data was de-identified in that it did not include name, address, Medicare Number or Pharmaceutical Benefits number. ABS then transformed this administrative data from transaction-level to person-level.
11 Data from the 2011 Census, and the transformed MBS and PBS data, were brought together using probabilistic linkage. The variables used to link the MBS and PBS data to the Census were date of birth, sex and Mesh Block. The method involved linking without the use of name and address; this information was destroyed at the end of the 2011 Census processing cycle.
12 The process also placed importance on accuracy and uniqueness. Only records that matched exactly on the linkage variables and were unique matches were retained. In this linkage project, a unique match was defined as instances where a record on the MBS or PBS file had only one matching record on the Census, and that same Census record does not match to any other record on the MBS or PBS file.
13 Before records between datasets are compared, the contents of the linking variables of each dataset need to be as consistent as possible to facilitate comparison. This process is known as standardisation. The standardisation procedure for the Mental Health Services-Census Data Integration project included coding imputed and invalid values on the data to a common missing value. These variables included date of birth, age, sex, Mesh Block, Statistical Area Level 1 (SA1) and postcode.
14 The table below lists the variables used to link in each pass. Each record pair required exact matching of all variables used in the pass in order for a link to be created.
Linking variables used for each pass
| Pass 1 | Pass 2 | Pass 3 | Pass 4 | |
|---|---|---|---|---|
| Sex | Y | Y | Y | Y |
| Date of birth | Y | Y | Y | |
| Age | Y | |||
| Mesh Block | Y | Y | ||
| SA1 | Y | |||
| Postcode | Y |
Representativeness
15 The linkage rates that were achieved for the MBS and PBS datasets were in line with expected results, and were relatively consistent across most sub-populations -- the exceptions were Northern Territory, Remote, Very Remote, and younger adults, which had lower linkage rates.
Linkage accuracy
16 False links can occur during the linkage process because, even when a record pair matches on all linking fields, the records may not actually belong to the same individual. While the methodology is designed to ensure that the majority of links are true some false links will be present within the dataset.
Unlinked records
17 There are three main reasons why records from the MBS and PBS datasets were not linked to a 2011 Census record:
- records belonging to the same individual were present in the MBS or PBS dataset and the 2011 Census but these records failed to be linked because they contained missing or inconsistent information in one or more of the datasets.
- there was no 2011 Census record corresponding to an MBS or PBS record because the person was not counted in the Census.
- there were more than one Census records that agreed on the same linkage variables – only unique matches were retained.
Weighting
18 Some groups of records were more likely to link, or conversely less likely to link, than other groups of records. This resulted in over representation of some groups and under representation of others. Records are more difficult to link when a person has poorly reported, poorly coded, missing or non-applicable values for linking variables. The non-random distribution of links has the potential to cause bias.
19 To compensate for differences in propensity to link, the data were weighted to represent the original MBS and PBS datasets.
20 Weighting is the process of adjusting a sample to infer results for the relevant population. To do this, a 'weight' is allocated to each sample unit - in this case, persons. The weight can be considered an indication of how many people in the relevant population are represented by each person in the sample.
21 For this project, estimates were created by weighting the linked records to represent the original MBS and PBS datasets, using: age group, sex, State/Territory, Remoteness Area, SEIFA, and broad groups for services and medications. For a relatively small number of records some of these variables were imputed for weighting purposes.
Data quality
22 All data collections are subject to sampling and non-sampling error. Non-sampling error may occur in any data collection. Possible sources of non-sampling error include errors in reporting or recording of information, occasional errors in coding and processing data, and errors introduced by the linkage process (discussed above).
23 A small number of geographies (State and Remoteness Area) were imputed, and a very small number of unusual records were removed prior to linkage.
MBS data
24 The Department of Human Services collects data on the activity of all persons making claims through the Medicare Benefits Scheme and provides this information to the Department of Health. Information collected includes the type of service provided (MBS item number) and the benefit paid by Medicare for the service. The item numbers and benefits paid by Medicare are based on the Medicare Benefits Schedule which is a listing of the Medicare services subsidised by the Australian Government.
25 MBS data includes Medicare-subsidised mental health-related services provided by psychiatrists, general practitioners (GPs), psychologists and other allied health professionals—including mental health nurses, occupational therapists, some social workers, and Aboriginal health workers. These services are defined in the Medicare Benefits Schedule (See Appendix 1).
26 Medicare data covers services that are provided out-of-hospital (e.g. in doctors' consulting rooms) as well as in-hospital services provided to private patients whether they are treated in a private or public hospital. The figures do not include services provided to public patients in public hospitals or services that qualify for a benefit under the Department of Veterans Affairs National Treatment Account. The States and Territories are the custodians of public hospital data [2].
PBS data
27 The Department of Human Services provides data on prescriptions funded through the Pharmaceutical Benefits Scheme (PBS) to the Department of Health. The PBS lists all of the medicines available to be dispensed to patients at a Government-subsidised price. The Government is advised by the Pharmaceutical Benefits Advisory Committee (PBAC) regarding which drugs should be listed on the PBS Scheme.
28 PBS data include subsidised prescription medication from the following groups: Antipsychotics, Anxiolytics, Hypnotics and Sedatives, Antidepressants, and Psychostimulants, agents used for ADHD and nootropics (see Appendix 2).
29 The data refer only to prescriptions scripted by registered medical practitioners who are approved to work within the PBS and to paid services processed from claims presented by approved pharmacists who comply with certain conditions. They exclude adjustments made against pharmacists’ claims, any manually paid claims or any benefits paid as a result of retrospective entitlement or refund of patient contributions [3].
30 The PBS data exclude non-subsidised medications, such as private and over-the-counter medications. Under co-payment prescriptions (where the patient co-payment covers the total costs of the prescribed medication) data are available from mid-2012; and therefore not available for 2011 [3].
31 Data does not include the Repatriation Pharmaceutical Benefits Scheme (RPBS) which is subsidised by the Department of Veterans’ Affairs [4].
32 Whilst data on medication dosage was obtained, it has not been considered in analyses in this publication.
Census of Population and Housing
33 The 2011 Census measured the number and key characteristics of people who were in Australia on Census night 9 August 2011. For information about the 2011 Census please refer to Census 2011 Reference and Information and Census Data Quality on the ABS website.
34 Cultural and linguistic diversity can be measured using multiple data items that capture the cultural and linguistic characteristics of a person. For the purpose of this publication, the Australian population has been divided into four groups based on country of birth and language spoken at home:
- Born in Australia, speaks English at home
- Born in Australia, speaks a language other than English at home
- Born overseas, speaks English at home
- Born overseas, speaks a language other than English at home
35 For more information, see Standards for Statistics on Cultural and Language Diversity, 1999 (cat. no. 1289.0).
36 While Indigenous status is a core data item in determining an individual’s cultural and linguistic background, data for Aboriginal and Torres Strait Islander Australians are not presented in this publication. The Aboriginal Health Services Program, funded by the PBS, does not use the Medicare PBS processing system [3]. Medications provided through this program are therefore not captured in the Mental Health Services-Census Integrated Dataset. Most affected are data for Remote and Very Remote areas and data for the Northern Territory.
Geography
37 The Mesh Block information used in the linkage process may not be aligned between the MBS and PBS files, and the Census, for a range of reasons, including:
- differences arising because MBS and PBS mesh block are based on postal address whereas the Census mesh block was based on the usual residential address;
- persons may have changed their address but not updated their Medicare records.
38 Medicare claims data used in this dataset are based on the Mesh Block of the enrolment address of the patient. As clients may receive services in locations other than where they live, these data do not necessarily reflect the location in which services were received. The data therefore reflects geographic information about the patient, rather than where they received each service; for example, the data does not show GP services by state, but rather the GP services provided to patients in each state.
Remoteness areas
39 People living in Remote and Very Remote areas of Australia are under-represented in the data. This may be for a number of reasons including:
- GPs are less likely to charge Medicare in Remote areas [5].
- non-metropolitan hospitals are more likely to admit patients, and people in Remote areas are more likely to attend hospital accident and emergency departments for primary care medical consultations than people from Major Cities [5]. People accessing these hospital services may be public inpatients and therefore not in scope. States and Territories are the custodians for this data and it is not included in the dataset.
- in 2010-11, despite there being more GPs in Remote areas, there were about half the GP services provided per person in Very Remote areas as in Major Cities [6].
- the Aboriginal Health Services Program is funded by the PBS however person-level data is not in the PBS processing system. Data from Remote and Very Remote areas, and the data from the Northern Territory are most affected [3].
- Section 100 of the National Health Act, 1953 allows for the Minister to make special arrangements for the supply of medications to people living in isolated areas. These medications do not appear in the PBS data.
40 The Census also undercounts the number of people living in some areas of Australia more than others. In 2011, the Northern Territory recorded the highest net undercount rate of all states and territories (6.9%) and showed the largest difference in the net undercount rate between its greater capital city and rest of state region (3.7% and 10.9% respectively) [7].
Rounding
41 Estimates presented in this publication have been rounded. Proportions are based on unrounded estimates. Calculations using rounded estimates may differ from those published.
Acknowledgement
42 The ABS acknowledges the continuing support provided by the National Mental Health Commission and the Department of Health for this project. The provision of data by the Department of Health and the Department of Human Services as well as funding from the National Mental Health Commission was essential to enable this important work to be undertaken. The enhancement of mental health statistics through data linkage by the ABS would not be possible without their cooperation and support.
43 The ABS also acknowledges the importance of the information provided freely by individuals in the course of the 2011 Census. Census information provided by individuals to the ABS is treated in the strictest confidence as is required by the Census and Statistics Act (1905). MBS and PBS information provided by the Department of Health and the Department of Human Services to the ABS is treated in the strictest confidence as is required by the National Health Act (1953) and the Health Insurance Act (1973).