An Analysis of the Department of Health’s Data Relating to State-Managed Adult Mental Health Services from 2013 to 2017


Building patient journeys with the data

Mental illness can require long-term care with relapsing and remitting episodes. As a consequence, the ability to analyse how people access different mental health services over time is important. Such a view provides some indication of whether the service mix is appropriate to needs, and if the services can be accessed efficiently and effectively. For this reason the analysis focused on people’s use of mental health services over time.

As noted in the Caveats and exclusions section above, the availability of suitable data effected which services were included in the analysis. We analysed 3 State-managed care settings that provide mental health services, these were:

  • inpatient services, where people received care as an inpatient
  • community treatment services, which include mental health services provided by CMHTs
  • emergency department services.

Inpatient and community treatment services accounted for 90% of the MHC’s total expenditure on mental health services in 2017-18. The care provided in EDs is not considered specialised mental health care, and is not funded by the MHC. However, it was included in our analysis because it is frequently an access point for many people seeking mental health care.

To create a systematic view of people’s pathways, we obtained data for all people who accessed State-managed mental health care. These people accessed at least 1 of the 3 care settings at least once in the 5 year period from 1 January 2013 to 31 December 2017. Care may have included a single presentation to an ED, a telephone call from a community treatment service, an extended admission to hospital for a mental health concern or a combination of activities across services.

We focused on people who accessed these services with a primary and secondary mental health diagnosis. While the data included events that were drug and alcohol related, this was not the focus of the audit.

The way the DoH collects and stores hospital or inpatient data posed some challenges for our analysis. Existing inpatient data is activity-based, structured around units of care[1] provided, and recorded when an event called a ‘separation’ occurs. A separation occurs when a person is discharged from a hospital or when there is a change in a person’s primary reason for care while in hospital. The challenge was connecting the individual separations when they overlapped or were consecutive, to build a stay, while maintaining the detailed information about each separation. There were 4.8 million care events that took place from 2013 to 2017.

An inpatient stay involving multiple separations can occur when a person stays in the same hospital, but the primary reason for care changes. For example, a person may be admitted to hospital after a deliberate poisoning associated with a suicide attempt. The primary reason for care on admission is to stabilise the person, treating the complications resulting from the poisoning. Once medically stable, the primary reason for care becomes management of the mental health condition. This change in the primary reason for care is captured in the system by the creation of 2 separations, but for the individual is 1 episode of care.

A stay involving multiple separations can also occur when a person is discharged from a hospital, transferred and admitted to another hospital before being discharged and sent home. This is reported as 2 separations. There were 117,615 inpatient stays from 2013 to 2017.

Another challenge was to capture information about people who were still in hospital and not yet discharged. As a consequence, inpatient data had to be extracted from 2 different data sources. This was particularly important because without doing this we would not have been able to capture all the people who were undergoing an extended stay in hospital and had not yet been discharged.

Steps to complete the data analysis

Our approach to the analysis, which we repeated a number of times, involved the following steps. We:

  1. Identified the cohort

Using the criteria we set in consultation with the DoH, the DoH extracted data relating to any person who had 1 or more mental health related ED presentations, inpatient events or contacts with a community treatment service (we refer to these as service events) within the 5 year period from 1 January 2013 to 31 December 2017.

  1. Extracted raw data

The DoH extracted all inpatient, emergency and community treatment service events for those people identified as having at least one mental health related service event. This data was extracted from 4 sources:

  • the Hospital Morbidity Data Collection for inpatient events where patients have been discharged from public and State-funded private hospitals
  • the Mental Health Information Data Collection for mental health related inpatient events where patients were still in a psychiatric hospital or mental health inpatient ward when the data was extracted
  • the Emergency Department Data Collection for ED events
  • the Mental Health Information System for community treatment service contacts.
  1. Ensured anonymity and linked data

The DoH Data Linkage Branch created a unique, encrypted identifier for each individual  and ensured obvious identifying information such as name and address was removed. This was necessary because there is no unique identifier for each person across the multiple data sources, and it was essential to ensure that we could not identify individuals.

  1. The DoH provided files to the OAG

The DoH transferred a number of password-protected data files to us by using a secure portal/electronic file transfer. The files were downloaded to and stored on a standalone SQL server with restricted access. On this standalone server we created a platform where the data from the four data sources could be merged. Having the identifier allowed us to merge events for individuals from the multiple sources.

  1. Tested files

We conducted validation tests to ensure that the files were complete and accurate. For example, the data was reviewed to ensure it covered the complete period, all the requested data fields were included and event information was accurately reflected. Any issues identified were resolved, in some instances requiring a new data extract. As an example, in a file provided that was intended to report patients who were still in hospital, we found several people who had been discharged. This meant the DoH had to review and refine the selection criteria they used to extract the data and provide a new extract. There were 7 iterations in total, resulting from refining the selection criteria and ensuring the most current data was available for our analysis. Validation tests were re-run on receipt of new data files to ensure accuracy and completeness.

  1. Prepared and ‘cleaned’ the data

Each version of the data provided by the DoH was loaded onto the secure platform we created. A set of processes was run on the data to identify and correct certain integrity issues. For example, we identified events for some people that occurred after death. We excluded these events from our analysis. We also made changes to the data to recognise services provided to public patients in private hospitals where a public private partnership existed. In addition, using reference documents provided by the MHC, and through consultation with the HSPs, we defined the ward types for mental health units, being acute, subacute and hospital in the home.

  1. Organised event data into episodes of care

Discrete mental health inpatient events were merged to create stays. These stays reflected how long a person spent in inpatient care, with a break of no more than 24 hours. These stays were then brought together with ED and community treatment events to create people’s episodes of care. In our analysis, episodes of care covered care delivered to a person from any of the 3 mental health care settings we looked at, where there was not more than 7 days between events. Joining these episodes of care together allows pathways of care[2] to be identified and analysed.

  1. Quality control process

A sample of patient data, that excluded obvious identifying information from the separate systems was used to manually construct an expected timeline of events. This was then compared to the automatically transformed data to verify that the process was completed as expected.

  1. Analysed the episodes of care

Once the data was organised into episodes of care we analysed it to:

  • establish how many people entered care by year
  • uncover the combinations of service types in which people accessed care
  • ascertain the amount of care people accessed in each service
  • determine the patterns by which people accessed inpatient, community treatment and ED mental health care
  • determine the number of people who accessed care each year by service type
  • recalculate the 28-day readmission rate and the 7-day follow-up with a focus on people rather than events.

[1] The Department of Health refers to inpatient units of care as ‘episodes of care’. The terminology “episode of care” has a specific meaning within the health sector, which is different and should not be confused with the OAG’s reference to episodes of care, which we use to refer to periods of consecutive care across inpatient, outpatient, and emergency settings.

[2] Appendix 2, Access to State Managed Mental Health Services. Office of the Auditor General report number 4 2019

Back to Top