Administrative data provides a wealth of information that can be used to monitor dementia in Australia. Currently, a diagnosis of dementia can be identified in hospital admitted patient care, mortality, aged care and Pharmaceutical Benefits Scheme (PBS) data. However, these sources tend to be best at identifying people with later stages of dementia, as disease progression leads to more frequent contact with various parts of the health and aged care systems.
Primary and secondary care data provide the best opportunity to identify people living in the community who are in the early stages of dementia. However, the lack of explicit identification of a dementia diagnosis in primary and secondary care data is currently a key data gap for dementia monitoring in Australia. There are, however, several Medicare-subsidised services that are commonly accessed by practitioners and their patients in the course of diagnosing dementia in its earlier stages. These include services such as the practitioner appointments themselves, diagnostic imaging and pathology, and team care arrangements.
This feasibility study aims to test whether, at a given point in time, the presence of dementia can be identified from Medicare Benefits Schedule (MBS) item claims that reflect the steps taken by medical practitioners in the diagnosis of dementia. This analysis was undertaken on the National Integrated Health Services Information Analysis Asset (NIHSI-AA), which links together the MBS, PBS, hospitals and aged care data needed to identify a cohort of people with early dementia. Two techniques were tested for identifying early dementia: a decision tree and logistic regression analysis.
Results and future directions
When deployed, both models enumerated around 30,000 possible cases of early dementia. Around 25,000 of these were true cases of early dementia, as far as can be identified by the PBS data. Overall, the models were able to capture approximately 80% of people who went on to receive a dementia-specific medication prescription within 2 years of their use of these Medicare service items.
The models developed and tested in this study have shown that geriatric services and brain scans are highly associated with early dementia, both individually and in combination as a service ‘pathway’. More niche services, such as those associated with consultant psychiatrists, have strong association for individuals accessing them but do not appear to be a common feature of the general early dementia diagnosis and management pathway.
While these initial results demonstrate that MBS items can feasibly be used to bring estimates of dementia prevalence in Australia closer to the true prevalence, these models should continue to be improved to enable more complete and accurate measurement of dementia. Further exploratory work into the characteristics of misclassified individuals will likely yield improvements to this detection method. This may include:
- exploring further individuals who are misclassified due to data gaps (such as missing hospital information, alternative services streams such as Aboriginal Health Service use)
- assessing similarity between dementia diagnosis pathways and other conditions
- identifying individuals who did not access the PBS during the period of time assessed but are captured in other data sets in later years.
This refinement, alongside alternative identification approaches such as machine-learning techniques, may allow for better identification of individuals with early dementia and therefore more accurately assess the service needs and burden of disease.
- Dementia monitoring and diagnosis in Australia.
- Medicare claims can help fill the primary–secondary care data gap
- Aims of this report
- Identifying people with early dementia
- Developing an algorithm to predict early dementia
- Early dementia group characteristics
- Which services did patients with and without early dementia use?
- Predicting early dementia from service items
- Which service items are predictive of dementia?
- Which model performed better?
- What does this all mean for improving dementia monitoring?
- Future directions
Appendix A: Methodology
End matter: Acknowledgments; Abbreviations; Symbols; References; List of tables; List of figures