Improving Australia’s dementia data for national action

Introduction

Almost everyone will be affected by dementia in some way over the course of their lives. According to a 2023 AIHW national survey of dementia awareness:

• 2 in 3 people had a family member or friend living with dementia
• 1 in 4 had cared for a family member or friend living with dementia (AIHW 2024).

Current estimates suggest that over 400,000 Australians are living with dementia; this number is expected to double in the next 35 years (AIHW 2023), meaning that dementia will be a part of the everyday lives of millions of Australians for years to come.

The commitment to action on dementia has been made. The Commonwealth is working with state and territory governments to develop a new, 10-year National Dementia Action Plan with clear and measurable objectives to improve outcomes for people living with dementia, their families and carers, including improving dementia data, maximising the impact of dementia research, and promoting innovation. It is anticipated that the Plan will be in place in 2024.

As dementia impacts more Australians, better data will increase our capacity to not only understand how many people need support but also to plan, resource and evaluate the services, initiatives and programs required to support them.

Better dementia data can lead to better lives, both for people living with dementia and, for their formal and informal carers. The needs of informal carers, including for respite, information, peer groups and income support, are urgent. But so too are the needs of the formal workforce who need to be supported with dementia-specific training, dementia-friendly working environments, and improved working conditions.

The final report of the Royal Commission into Aged Care Quality and Safety identified dementia care as an area for immediate attention (Royal Commission into Aged Care Quality and Safety 2021). Yet without robust data on how many people are living with dementia – and needing and accessing care – implementing the report’s recommendations is challenging.

In September 2023, the AIHW released the National Dementia Data Improvement Plan 2023–33, outlining and prioritising data improvement activities to be undertaken over the next 10 years. A framework is now in place; the next step is to ensure that the plan’s implementation remains a priority among the broader strategies in the Australian health, aged care and disability data landscape. The AIHW’s National Centre for Monitoring Dementia (NCMD) is well positioned to take the lead on this work, but it cannot succeed alone. Improving Australia’s dementia data will require long-term planning and commitment from multiple stakeholders at national, jurisdictional and local levels, across government, academic, research, not-for-profit, and corporate institutions. 

What does this article include?

This article is a follow-up to Chapter 8 in Australia's health 2020: data insights ‘Dementia data in Australia – understanding gaps and opportunities’. It presents a ‘report card’ on progress made with national data sources currently available to monitor dementia. It also describes what the NCMD is doing to work towards reducing remaining gaps and opportunities, including:

• working to understand the limitations of administrative data and advocating for changes to improve quality and consistency
• expanding the capabilities of individual administrative data sources by integrating them and creating enduring linked data assets
• evaluating alternative data sources and adding them to linked data systems where possible.

The aim of the AIHW’s work is to understand who is living with dementia, where they are living, who is supporting them, what services they engage with and, what outcomes they experience, in order to evaluate:

• which services are helpful
• where there is unmet need
• whether specific types of people or communities are more heavily affected than others.

While not in scope for this article, preventing dementia and increasing dementia literacy are integral to reducing the impact of dementia. Increased awareness of what dementia is and where to go for support may increase the likelihood of people’s being not only diagnosed at an appropriate time, but also able to access the support they need. Better data are needed, too, on dementia risk factors to inform targeted public health messaging and risk reduction strategies in the community (Anstey et al. 2020; Long et al. 2023). Further, while the focus of this article is on improving data about people living with dementia, better data are also needed to understand the experience of formal and informal carers of people living with dementia.

Dementia prevalence and incidence data remain elusive

To appropriately plan, provide and fund services to support people living with dementia it is necessary to know both:

• prevalence – the number of people with a condition at a given time.
• incidence – the number of people newly identified with a condition within a defined time period.

Prevalence allows appropriate planning for the overall number of services that will be required into the future, while incidence is important to inform diagnostic care pathways. Age-specific incidence is also an excellent indicator of whether public health prevention measures are working to delay dementia onset.

Measuring incidence and prevalence for dementia is particularly challenging as it is a complex syndrome, not a specific disease (see Box DEM.1) and the diagnostic process is not straightforward (for more information, see Clinical guidelines for dementia).

All existing prevalence and incidence estimates have limitations

Current estimates for dementia prevalence and incidence in Australia, include reliance on data that are out of date, out of context, or self-reported.

Limitations of current prevalence data

The AIHW’s national dementia prevalence rates are modelled from combining international prevalence rates with those for a single small-scale Australian study on younger onset dementia in Eastern Sydney (Withall et al. 2014) (for more information, see Dementia in Australia). This method is not ideal and risks underestimating or overestimating the true number of people living with dementia in Australia. This method also does not account for the impact and changes in risk and protective factors, and treatment and care pathways for dementia over time. Further, it does not account for over-representation of dementia in specific social or cultural groups.

Estimates for dementia prevalence can also be obtained from national data collected by:

• the Australian Bureau of Statistics (ABS), including from the Census of Population and Housing (the Census) conducted every 5 years
• the Survey of Disability, Ageing and Carers (SDAC) conducted every 3 to 4 years on a representative sample from the community and people in cared-accommodation, such as residential aged care facilities and hospitals (ABS 2021).

However, to date these estimates have been substantially lower than expected, compared with international prevalence rates (WHO 2021). Perhaps this is because they rely on self-report or proxy responses. Perhaps it is because the questions are new (dementia data were first collected in the SDAC in 2015 and in the Census in 2021) and are asked in the context of other competing health conditions.

Where possible, the NCMD collaborates with the ABS with the long-term goal of improving data collected to measure national dementia prevalence.

Limitations of incidence data

One development in measuring dementia incidence has been the establishment of the Australian Dementia Network (ADNeT) Clinical Quality Registry, which registers new cases of dementia at participating memory clinics and dementia diagnostic services. While still small in scale and non-mandatory, this resource has the potential to offer a measure of the number of people seeking a dementia diagnosis. It is, however, unlikely to reflect the overall incidence of dementia in Australia until data collection is expanded to be more nationally representative and include a wider range of clinical diagnostic pathways.

In the past, epidemiological studies of representative sample populations have been undertaken to estimate dementia incidence and/or prevalence at a national level. These have been more successful internationally in smaller geographical areas like the Netherlands (Ott et al. 1998) and the United Kingdom (Matthews et al. 2013) than in Australia.

Data collection for such studies is complex because people living with dementia are more likely to drop out of the long-term studies required to capture incidence data, and are often spread across community and institutional settings. Large-scale epidemiological studies, while highly valued for their robust findings, are also known to be costly, resource-intensive, and subject to bias and quality issues (Brayne and Davis 2012; Brayne and Moffitt 2022).

Data gaps begin in primary health care

The AIHW’s 2023 Dementia Awareness Survey (AIHW 2024) found that 90% of the people who said they would seek help for dementia symptoms would go to their general practitioner (GP). Data collected during GP visits would be an ideal source for capturing dementia incidence in Australia, if not for the fact that there are:

• known barriers to GP diagnoses (Casey et al. 2020, Lang et al. 2017)
• no standardised or systematic ways for GPs to document the diagnostic process even when they do undertake it.

This means that there are no national data sources for GP visits that contain information on dementia diagnoses, diagnosis dates, treatments, or care. The same is true for any visits to specialists to whom GPs may refer patients to, such as neurologists or geriatricians.

Pilot project to evaluate quality of dementia data from Primary Health Networks

This is a critical data gap that the NCMD, and the AIHW more broadly, have been working to rectify. A pilot project has been undertaken to evaluate the quality of dementia data that could be collected via Primary Health Networks (PHNs), as well as to test the governance, data transformation and data flow arrangements with PHNs providing the data. The overall aim is to develop a National Primary Health Care Data Collection that would become a principal source for identifying people diagnosed with dementia.

Routinely collected data cover some dementia data gaps

Without robust prevalence and incidence data sources, dementia researchers have had to become ‘data detectives,’ tracking traces of lives lived with dementia through administrative data designed for other purposes.

The AIHW’s current national prevalence estimate (based on a combination of prevalence rates from international studies and a small-scale Australian study) is that at least 400,000 people are living with dementia in Australia in 2024. Many of these people, but not all, will interact with health, disability or aged care services. During these interactions, dementia may be noted in their records, such as a reason for hospitalisation, or a health condition affecting their care needs when seeking aged care or disability support. But it may also go unnoticed or unreported.

The dementia data report card presented in Table DEM.1 outlines the major routinely or regularly collected data sources currently used to monitor dementia in Australia, including an indicator of whether these sources have improved, stayed stable, or deteriorated since first reported on in 2020.

Different data sources represent different people living with dementia

One of the main problems with single source administrative data is that they are not necessarily representative. 

Not everyone living with dementia will interact with health, disability or aged care services. Further, dementia may not always be recorded for people who do. Studies in Australia and internationally have shown that pharmaceutical, hospitals, aged care assessment, residential aged care and deaths data all underestimate the number of people living with dementia (Cations et al. 2020; Gao 2018; Solomon et al. 2014; Stokes et al. 2020; Welberry et al. 2020; Xu et al. 2022).

Underestimates or biases with single source data

Both the number and type of people are underestimated in single source data:

• Some data sources, such as prescriptions data, are more likely to include a younger population living with dementia (for example, under 75 years) earlier in their symptom progression, because dementia medications are indicated at this age and stage rather than later.
• Other data sources, like assessments for entry into residential aged care, reflect an older population living with dementia (for example, 75 years and over), when the condition has progressed to requiring substantial care.

Using only one data source risks reaching wrong or biased conclusions about all people living with dementia in Australia because not everyone has been represented (Welberry et al. 2020).

National data linkage holds promise for better dementia data

When multiple administrative data sources are combined, and all the different service records that contain information on dementia for a person are linked, more people living with dementia in Australia can be identified. This is a way to overcome some of the bias inherent in single data sources and improve data on the overall number and type of people living with dementia who are represented (Wilkinson et al. 2018).

Putting together a puzzle – an apt analogy for dementia data detective work

Dementia data detective work using linked data can be likened to putting together pieces of a puzzle – where different data sources are collated to create one big picture that represents as many people living with dementia as possible. The size of the puzzle piece contributed by each data source may be bigger or smaller depending on the underlying population of interest:

• Figure DEM.1 presents an AIHW linked data example where, in a population aged 65 and over living in the community, PBS prescriptions data identify more people living with dementia (58% of the people identified as living with dementia in the community) than in a population of the same age living in permanent residential aged care (32% of the people identified as living with dementia in permanent residential aged care).
• Conversely, ACFI data play a tiny role in identifying people living with dementia in a community population (around 1% of all people identified as living with dementia in the community) but a massive role in identifying people living with dementia in residential aged care (88% of all people identified as living with dementia in permanent residential aged care).

Then, there is always a missing puzzle piece – representing people who have not engaged with the services offered by the included data sources (shown in black with question marks in Figure DEM.1). The size of this piece also varies depending on the sources included, but published literature suggests that it often represents around 20% of people living with dementia (Waller et al. 2017, Welberry et al. 2020).

Figure DEM.1: Example of dementia identification using linked data

Notes

1. Each data source identifies a different proportion of people living with dementia, and this varies depending on the underlying population characteristics.
2. For each population, percentages will not sum to 100% as dementia may be identified in multiple data sources for the same person.

Source: AIHW analysis of all people in the National Health Data Hub linked data asset using services in 2020–2021.

At times it can be important to maximise the capability of linked data assets to identify people living with dementia by using all available data sources. This may be the case when using data linkage for modelling purposes:

• If inputs to the model only represent some of the people living with dementia (for example, only people identified via residential aged care assessments), outputs will be biased to the characteristics of these people (for example, people aged 75 and over with more advanced dementia and higher support needs).

However, there are circumstances where it might be of interest to isolate only a specific group of people within a linked data asset (for example, only people using dementia-specific medications) and then follow their service use and outcomes over time. An illustration of the questions that can be answered with this type of analysis is presented in Figure DEM.2.

Figure DEM.2: Example research questions that can be answered using linked health and aged care data

As shown in Figure DEM.2, there are a myriad of opportunities for data linkage projects that extend well beyond providing an alternative for prevalence measurement.

Investment in linkage paves the way for improved data

If the road to improved dementia data relies heavily on linked data, there is still some work to do to make the ride a little smoother. For instance:

• ethics and governance required for data linkage projects can hinder the timely use of data.
• the process of linking data is time and resource intensive – combined with the expenses of secure data environments, this can result in prohibitive oncosts.

Current initiatives for data linkage

Acknowledging these roadblocks, the AIHW is currently focused on more enduring approaches to data linkage, and is a key player in initiatives such as the National Disability Data Asset (NDDA) and the Australian National Data Integration Infrastructure (ANDII).

The AIHW is also investing in a data linkage system to benefit the wider research community. A key component of this is the National Health Data Hub (NHDH) – an extension and expansion of what was the National Integrated Health Services Information data asset – that will be interoperable with the ANDII and available to external researchers.

Streamlining ethics and governance processes for the NHDH is a priority for the AIHW. As our data linkage matures and grows, methods and processes are continually being refined and becoming more familiar. Over time, stakeholders and data custodians are gaining confidence as they see their data being appropriately used to produce outputs of national importance, including reports on:

• younger onset dementia
• geographical variation in health service use for people living with dementia
• transitions to residential aged care after hospital for people living with dementia.

For more information on the NDDA and the ANDII, see Australia’s health data landscape in Australia’s health 2024: data insights.

Consistency in data collections supports long-term health policy goals

Whether used in linkages or as stand-alone data sources, almost all of the administrative data used to identify people living with dementia in Australia are collected for other reasons – most often in the course of providing essential health and aged care services. This means that the information collected may not always meet data requirements and may change over time.

• One example of this is the collection of information about people living in residential aged care. Until recently, data on health conditions were collected during aged care funding assessments for people living in permanent residential aged care, via the Aged Care Funding Instrument (ACFI). In October 2022, the ACFI was replaced by the Australian National Aged Care Classification (AN-ACC) to shift the funding focus from a person’s health conditions to their functional capacity as a result of those conditions. While a person’s cognition is assessed under AN-ACC, their health conditions, including their dementia status, are no longer recorded.

Not collecting the health conditions of people living in permanent residential aged care has wide-reaching consequences. For dementia data, the number of people living with dementia in this population is now unknown. There is now also one less data source for identifying dementia in linked data assets, which lowers the validity of analyses of outcomes for people living with dementia in permanent residential aged care. In the data example in Figure DEM.1, 23% of the population living with dementia in permanent residential aged care would not have been identified without the ACFI.

Dementia is not the only health condition disproportionately affecting permanent aged care residents. This new data gap will compromise monitoring and surveillance of the health of this vulnerable population in Australia.

While aged care funding assessment data are not collected specifically for monitoring health conditions, additional information collected alongside assessments would provide high coverage (as everyone who enters permanent residential aged care is assessed). The AIHW is therefore exploring avenues with the Department of Health and Aged Care to maximise this opportunity for best practice data collection.

Specifications solidify ongoing data requirements

Consistent and consolidated data specifications underlie the quality, reliability and usefulness of all dementia data. Efforts must be made to ensure that data collections measure and report the same concepts in the same ways, ideally using the same (or equivalent) names, labels and values. For this to occur, agreement about what each data item means is required among organisations that collect and report data relating to people living with dementia. This may sound simple in theory but can be challenging in practice.

There has been a lot of recent movement in this direction:

• For dementia-specific data, a Dementia National Best Practice Data Set was endorsed in October 2022, recommending how basic items related to dementia could be standardised for dementia data collections.
• The next phase is to support a broad range of organisations – including primary health, hospital, and aged care service providers – in taking these recommendations and then monitoring their use in practice, incorporating refinements as agreed and necessary.

For aged care data, the AIHW and the Department of Health and Aged Care have partnered to deliver:

• an Aged Care Data and Digital Strategy to drive system-wide improvement
• associated minimum and best practice data sets that set the standards for what and how data should be collected to improve aged care data quality in stages.

When endorsed and implemented in routine data collection, these specifications will facilitate the provision of consistent information about the number, characteristics and service use of people with dementia across different data sources.

Enforcing the collection of specific data items is difficult without any mandating authority; it requires co-ordination and commitment across sectors. Work between government and industry organisations to support data entry standards such as FHIR (Fast Healthcare Interoperability Resources) is a promising development towards ensuring that the digital architecture is in place to facilitate the exchange of quality and consistent health care data more broadly. Having such standards in operation should support future sharing of dementia data.

Data linkage capacity can still be strengthened

There are ways to maximise opportunities afforded by linked data assets. Some require additions to existing linkages; others involve improving the quality of existing data.

Adding dementia-specific data sources

One way to increase the number of people identified as living with dementia in linked data is to include dementia-specific data sources in the assets themselves. The AIHW has been working on a pilot project to include data in the NHDH from people living with dementia who are registered with either:

• the Australian Commonwealth Government funded national behavioural support service programs of Dementia Support Australia (approximately 90,000 people) or the ADNet Clinical Quality Registry (approximately 2,500 people).

When these data are incorporated, we will not only have a means to assess the validity of dementia identification via routine data sources, but also have a richer understanding of the treatment and support pathways available to people living with dementia, and the impact these have on overall outcomes (such as entering permanent residential aged care).

Targeting data collection and linkage in priority populations

One of the largest dementia data gaps is among people in special interest populations who are historically under-represented in routine data, such as:

• Aboriginal and Torres Strait Islander (First Nations) people
• people from culturally and linguistically diverse (CALD) backgrounds
• people from the lesbian, gay, bisexual, trans, queer, and intersex (LGBTQI+) community.

Dementia prevalence and caregiver strain can be higher in these populations and there can be cultural barriers to accessing support, both for people living with dementia and their carers (Gilbert et al. 2022; Lo Giudice et al. 2016; Radford et al. 2019).

In First Nations, CALD and LGBTQI+ populations, monitoring prevalence and incidence at a national level can be challenging in terms of measuring both the numerator (how many people are living with dementia), and the denominator (the total population at risk).

There are known issues when reporting First Nations data, including sensitivity around disclosure of Indigenous status and large proportions of missing values for this field in administrative data collections. LGBTQI+ people have been invisible in Australian data collections until recently, and Australia still has a long way to go in accurately reflecting this community. Australia is also lagging in capturing data on CALD populations, currently relying on rudimentary proxies such as country of birth and language spoken at home. There is still work to do in the field of ethnicity identification to improve Australia’s dementia data for CALD community groups (FECCA 2015; Low et al. 2019).

In many cultures, including those of Australia’s First Nations people, dementia is thought of differently than it is in a clinical or service delivery context. Methods of data collection historically used in government and research settings in Australia may not be appropriate in some cultural contexts. There is much to be learned from other cultures about supporting people living with dementia, so it is imperative to take the time and find the ways to listen (Antoniades et al. 2022; Bradley et al. 2020).

Engaging with cultural groups that have been systematically marginalised in Australian history requires sensitivity, a willingness on both sides to consult and collaborate, and real commitment to the principles of co-ownership and co-design.

Meanwhile, maximising data linkage opportunities to include data sources that are working to improve their collection of information on First Nations people, people from CALD backgrounds and LGBTQI+ people, such as the Census and the ABS’s Person Level Integrated Data Asset is one way to increase our understanding of how dementia affects these groups and their engagement with health and welfare services.

These are not the only population groups that would benefit from improved data collection and linkage either – other priority populations include veterans, people with disability, people with childhood dementia and homeless people.

Standardising dementia identification in routine data

Another avenue to improve and standardise data on people living with dementia is to include dementia flags in administrative data – a specific field that indicates whether a person is known to be living with dementia, either represented as:

• a simple yes or no, or,
• (preferably for measuring incidence and prevalence) a date of first identification or diagnosis, if that data became available.

This flag could be added at the point of data collection and might then be beneficial for people living with dementia when they engage with services – for example, ensuring that their dementia status is known during a hospital admission so that clinicians can better manage their experience. Alternatively, it could be added retrospectively when deidentified data collections are prepared for use.

Currently, there are no specific dementia flags in routine data sources and linkages. Their inclusion could ensure that people living with dementia receive the care they need, and/or that the same people are being identified as living with dementia for all dementia monitoring and research projects. While there would be many hurdles to overcome to implement dementia flags, a similar recommendation was made for disability data as a result of the recent Disability Royal Commission, and the Commission’s final report discusses merits and barriers to implementation.

Undertaking studies that validate linkage-derived prevalence and incidence

Even with regular supply of high-quality data, data linkage assets such as the NHDH will still under-represent the population living with dementia in Australia, as illustrated by the missing puzzle piece in Figure DEM.1 (Chow et al. 2022). This is because, at present, they include only the population that engage with the specific services included in the linkage. 

An Australian validation study showed that linked data can capture 77% of the population expected to be living with dementia according to global estimates (Welberry et al. 2020), noting that capture rates vary according to age, as some age groups will be more likely to interact with specific services than others. 

The AIHW, in collaboration with The University of Queensland, is currently undertaking a similar validation of dementia identified via linked data in Australia, as opposed to the gold standard of clinical diagnosis in a study cohort. The aim is to use findings from this study:

• to improve dementia incidence estimates
• better understand where the weaknesses lie in ascertaining dementia cases via data linkages. Understanding this is important as analyses of health and service use of people living with dementia will be inaccurate if dementia cases have been missed and wrongly allocated to non-dementia comparison groups.

Given known issues with the under-identification of dementia using self-reported data (McGrath et al. 2021), the AIHW is also currently using linked data to examine the capture of self-reported dementia in the 2021 Census compared with dementia records in other national administrative and survey data. 

Being able to link and validate these data sources against each other brings us ever closer to true estimates of dementia prevalence and incidence in Australia. It also shines a light on data sources where dementia ascertainment may not be as robust. This allows us either to apply caution when interpreting findings from analyses that use these sources, or to invest resources in activities that improve quality and reduce bias – one of which may be regular validation.

Reduce, reuse, recycle can apply to data collection

Minimising waste is important in terms of data collection too, especially in health and aged care service delivery settings where resources are often stretched. Expanding existing data to enable reporting of dementia – with the strategic view to ‘collect once, use often’ – is another improvement opportunity for dementia data.

• For example, residential aged care quality indicators have been collected since 1 July 2019 and have expanded in scope over time; they now include 11 indicators across clinical, health, and wellbeing domains. Although these are reported at aggregate levels – and the data are used as input to the star ratings that provide information on providers’ quality of care at a facility level – the data are collected by assessing each individual in care. If there were a consistent collection of these records at the individual level, and it were possible to access these de-identified data along with the dementia status of the individual, this would open a whole new suite of possibilities for understanding the residential aged care experience of people living with dementia in Australia. The potential for individual-level aged care data is explored further in Chapter 8 of Australia’s welfare 2023: data insights – ‘Measuring quality in aged care: what is known now and what data are coming’.

This is not an isolated example of where innovations in the collection, distribution and configuration of data already being collected could hold potential for dementia monitoring and reporting. Similar opportunities also lie in greater use and application of artificial intelligence to analyse big data such as those generated through clinical care, as shown in recent examples using natural language processing to extract dementia data from electronic medical records (Ienca et al. 2018; Maclagan et al. 2023; Oh et al. 2023).

Dementia action needs data-informed decisions

Dementia is a serious and growing health and aged care issue in Australia. It has a substantial impact on the health and quality of life of people living with the condition, as well as on their family and friends and the wider community. 

Implementing some of the dementia data improvement activities outlined in this article should assist in reaching the ambitious goals of the National Dementia Data Improvement Plan 2023–33. Embracing future data and technological advancements will also be essential, both to ensure that data are future proofed, and to minimise the impact of data collection on the people dedicated to caring and supporting people living with dementia. 

Reducing the impact of dementia relies on co-ordinated, ongoing efforts by governments and local communities to act as needed to improve outcomes for people living with dementia and their carers. This action must be informed by a strong evidence base, the foundation of which is improved dementia data.

Further reading

Related topic summaries

Glossary   Data tables More articles from Australia’s health 2024 data insights