Measuring what counts in Aboriginal and Torres Strait Islander care: a review of general practice datasets available for assessing chronic disease care

Introduction

There has been significant success in reducing the fatal burden of disease among Aboriginal and Torres Strait Islander peoples (Australian Institute of Health and Welfare 2022). However, Aboriginal and Torres Strait Islander peoples still experience high rates of chronic disease (Australian Indigenous HealthInfoNet 2024).

There are many sources of population health data reporting on Aboriginal and Torres Strait Islander peoples, including surveys, registries and indicators derived from health system data (Doran et al. 2020). Notably, national longitudinal surveys, such as the Mayi Kuwayu Study (Jones et al. 2018) and the Longitudinal Study of Indigenous Children (Department of Social Services 2020), provide important information about the relationship between culture, socioeconomic circumstances and wellbeing. However, self-reported surveys are inherently limited in their ability to provide granular insights into how health care is delivered for Aboriginal and Torres Strait Islander peoples. Datasets derived from clinical information systems (CISs) have an important role in filling this gap.

Despite significant improvements being made, chronic diseases, such as cardiovascular disease (CVD), chronic kidney disease and diabetes, are still responsible for much of the morbidity and mortality seen among Aboriginal and Torres Strait Islander peoples (Australian Indigenous HealthInfoNet 2024). Most chronic disease is diagnosed and managed in primary care contexts. As such, access to high-quality primary healthcare data is critical for undertaking quality improvement, service planning and generating evidence-based policy regarding chronic disease. Significant gaps still exist in terms of the indicators of chronic disease screening, assessment and management that are validated and reported nationally. For example, CVD is the leading contributor to the burden of chronic disease among Aboriginal and Torres Strait Islander peoples (Australian Indigenous HealthInfoNet 2024), and yet the national Key Performance Indicators provide only two indicators related to CVD, and no indicator of appropriate CVD management (Australian Institute of Health and Welfare 2020).

Large amounts of data are collected about Aboriginal and Torres Strait Islander peoples, but it is often unclear how these data are being used to improve health. In Australia, several large datasets make de-identified primary healthcare data extracted from CISs available for research. These datasets take several forms, including datasets of de-identified patient-level data and centrally held service-level datasets reporting on pre-defined indicators. Patient-level datasets are typically either held centrally by a data custodian (centrally held datasets), or in a distributed fashion by health services and extracted on demand (distributed datasets).

There have long been difficulties accessing healthcare data in Australia (Productivity Commission 2017; Youens et al. 2020). The fragmented nature of Australian primary health networks and constrained funding for data management restricts the utility of primary healthcare data for research. Australian data collection arrangements have been criticised for failing to collect adequate patient-level primary healthcare data (Australian Institute of Health and Welfare 2018), and data are often underutilised (Productivity Commission 2017). CIS data are collected primarily for clinical documentation, not research, necessitating robust validation of derived indicators. There are unique difficulties in using these datasets to report on Aboriginal and Torres Strait Islander health care. Some datasets do not adequately record Aboriginal and Torres Strait Islander status (Australian Institute of Health and Welfare 2010) or appropriate indicators of social and cultural determinants of health. This typically reflects a failure to record such data in a readily extractable format.

In Australia, most systematically collected data do not reflect Aboriginal and Torres Strait Islander worldviews or needs. Collected data overwhelmingly reflect a deficits discourse characterised by ‘5D Data’: data that focuses on difference, disparity, dysfunction, disadvantage and deprivation (Walter 2016). The Maiam nayri Wingara Indigenous Data Sovereignty Collective have produced principles addressing this issue (Maiam nayri Wingara Indigenous Data Sovereignty Collective and Australian Indigenous Governance Institute 2018). It is held that Aboriginal and Torres Strait Islander peoples have a right to Indigenous Data Sovereignty (ID-SOV); that is, a right for ‘Indigenous peoples to exercise ownership over Indigenous Data’ (Maiam nayri Wingara Indigenous Data Sovereignty Collective and Australian Indigenous Governance Institute 2018). ID-SOV is enacted through Indigenous Data Governance (ID-GOV), a right to ‘decide what, how and why Indigenous Data are collected, accessed and used’ (Maiam nayri Wingara Indigenous Data Sovereignty Collective and Australian Indigenous Governance Institute 2018). ID-GOV requires that the data ecosystem is controlled by Aboriginal and Torres Strait Islander peoples, that data respect individual and collective interests, are contextualised and disaggregated, are accessible, and empower self-determination (Maiam nayri Wingara Indigenous Data Sovereignty Collective and Australian Indigenous Governance Institute 2018). The extent to which these principles are applied in primary healthcare datasets is unclear.

Prior reviews of national primary healthcare datasets have not focused on Aboriginal and Torres Strait Islander peoples (Canaway et al. 2022). Reviews of Aboriginal and Torres Strait Islander health data have not focused on primary care datasets extracted from CISs (Doran et al. 2020). Here, we assess all key national primary healthcare datasets for their potential to provide insights into chronic disease care among Aboriginal and Torres Strait Islander peoples, and for their utility in directing quality improvement and service planning. A non-peer-reviewed unpublished report of this study was initially produced for the Australian Government Department of Health (McBride Kelly et al. 2021).

Methods

Results

Seven datasets were identified (D1–D7), and all data providers consented to interview. Following their consideration of our preliminary review, D6 asked to be excluded from this study. D7 did not respond to our interview request, but later provided detailed feedback on the preliminary report. Of the six datasets analysed, four provide centrally held de-identified patient-level data (D3, D4, D5 and D7), one provides distributed de-identified patient-level data (D2) and one provides service-level indicators (D1). As D2 is not a data custodian, some criteria could not be assessed, as this would require interrogation of the underlying patient records. Table 1 provides a summary of the assessment of the datasets, and Table 2 provides a summary of the indicators available from each dataset reporting on CVD.

Discussion

We sought to review all national primary care datasets in Australia that could be used to provide insights into chronic disease among Aboriginal and Torres Strait Islander peoples (McBride Kelly et al. 2021). All reviewed datasets had the potential to provide such insights. However, there is significant scope to improve the quality and usability of this information, both at the level of the data collected in primary health care and dataset governance arrangements.

The key driver for the quality of data held in primary healthcare datasets is what and how information is entered and stored in the CIS. CISs are primarily a tool for clinical documentation, not research. The lack of standardisation of CISs in Australia, absence of national standards for the representation of medical information (Youens et al. 2020; Canaway et al. 2022) and difficulties extracting information from ‘free text’ fields all contribute to data quality issues. Fortunately, Aboriginal and Torres Strait Islander health services often record chronic disease data in a readily extractable format (Davis et al. 2015).

Some datasets reported low rates of completion of Aboriginal and Torres Strait Islander status data, reflecting a failure of clinicians to record this information in a readily extractable manner. Aboriginal and Torres Strait Islander people are frequently under-identified in health records (Australian Institute of Health and Welfare 2010), and such misclassification is non-randomly distributed (Briffa et al. 2010; Thompson et al. 2012). Analysis of datasets with low rates of Aboriginal and Torres Strait Islander status completeness may be susceptible to selection bias.

Important contextual data, including social and cultural determinants of health, patient-reported outcome measures and patient-reported experience measures, are not routinely recorded by clinicians in CISs. Most health services, apart from some specialised Aboriginal and Torres Strait Islander-specific organisations, do not routinely collect data related to social and cultural determinants of health. This is reflected in our finding that data related to social and cultural determinants of health are sparse among the datasets reviewed. Capturing this information would require significant changes to the data collected by clinicians, and changes to CISs, so that these data can be stored and extracted appropriately. Structured approaches exist for the collection of these data, but they are incompletely implemented and not currently linked to CIS data. Templates have been developed for the Aboriginal and Torres Strait Islander Peoples Health Assessment (Medicare Benefits Schedule item 715) that facilitate the collection of data related to social and cultural determinants of health (Royal Australian College of General Practitioners 2020). Similarly, the Mayi Kuwayu Study developed a detailed survey including items on culture, health and wellbeing, sociodemographic factors, and family connections, and has sought consent from participants to link responses to morbidity and mortality datasets (Jones et al. 2018). Further, the Australian Bureau of Statistics collects patient-reported experience measures through the annual Patient Experience Survey (Australian Bureau of Statistics 2022), as do many health services through various surveys. The incorporation of these data into CISs, and integration into primary healthcare datasets, would significantly expand opportunities for patient-centred quality improvement.

Almost all primary health datasets in Australia are not designed specifically for the needs of Aboriginal and Torres Strait Islander peoples. Only two of the datasets reviewed included a significant proportion of Aboriginal and Torres Strait Islander people, captured chronic disease data, and collected data nationally. The primary healthcare provider for 36% of Aboriginal and Torres Strait Islander peoples is an Aboriginal and Torres Strait Islander-specific health service (Australian Bureau of Statistics 2019). Despite this, no dataset was identified that captured a significant number of both mainstream GP clinics and Aboriginal Community Controlled Health Organisations, with all datasets either overrepresenting one or the other.

Multiple datasets would need to be combined to represent the whole healthcare system used by Aboriginal and Torres Strait Islander peoples. However, this is a costly and technically complex process owing to a lack of standardisation in the way that GP data is extracted, stored and represented, with most datasets having their own non-standardised coding systems. Encouragingly, two datasets (D5 and D7) implemented the Observational Medical Outcomes Partnership Common Data Model, which translates each element in the dataset into a standardised vocabulary and data schema (Hripcsak et al. 2015). Further standardisation of primary healthcare datasets would expand opportunities for datasets to be used in combination.

Only one dataset assessed was adherent with ID-SOV principles. The collection of the medical information of Aboriginal and Torres Strait Islander peoples into research datasets could be seen as an example of the violation of ID-SOV (Renehan 2018). This is unsurprising given the focus on a top-down approach of imposing data collection arrangements on health services, a lack of substantial engagement with Aboriginal and Torres Strait Islander peoples (The Centre of Research Excellence in Aboriginal Chronic Disease Knowledge Translation and Exchange 2020), and limited opportunities for data to remain within the control of Aboriginal and Torres Strait Islander peoples. Data providers must act to implement ID-SOV principles, empowering Aboriginal and Torres Strait Islander peoples to exercise self-determination in this domain. Only then will collected data reflect the worldviews and visions of Aboriginal and Torres Strait Islander peoples.

There is significant scope for improving the accessibility and utility of primary healthcare datasets. Although many data custodians provide surveillance reports to health services, these reports are often not disaggregated by Aboriginal and Torres Strait Islander status. As such, the potential for quality improvement related to Aboriginal and Torres Strait Islander patients is limited. Further reporting specific to the needs of Aboriginal and Torres Strait Islander patients, and expanding dataset access would support self-determination. Indeed, healthcare services report these data are of assistance in quality improvement, funding applications and identifying gaps in service delivery (Australian Institute of Health and Welfare 2020).

This study’s key strength is that interviews were undertaken with those responsible for the relevant datasets, hence providing insight into capabilities that may not be publicly available. The interim report was circulated to data providers for their review. A positive outcome of this feedback process was that one data provider welcomed suggestions on how principles of ID-SOV could be applied to their ways of working: including Aboriginal and Torres Strait Islander representation in their governance structure, and Aboriginal and Torres Strait Islander-specific reporting. One limitation of our approach is that we did not obtain access to the underlying data held in the reviewed datasets. It is likely that further dataset capabilities and limitations would become apparent with access to the underlying data, particularly limitations regarding dataset interoperability.

Conclusion

This study is the first review of the capacity of national, CIS-derived primary healthcare datasets to provide insights into the health care of Aboriginal and Torres Strait Islander peoples. It is critical that more accurate, comprehensive, and culturally meaningful Aboriginal and Torres Strait Islander healthcare data are collected. Improvement of these datasets must be led by the principles of ID-SOV, enabling public health policy to be directed by community priorities and self-determination. Key steps include developing national policy for the implementation of ID-SOV principles in these datasets, sponsoring Indigenous traineeships in data science to ensure data remains in Indigenous hands, and developing governance structures that support ongoing health data collection and use by Aboriginal and Torres Strait Islander peoples. Further, validated and appropriate national indicators for chronic disease assessment and management among Aboriginal and Torres Strait Islander peoples should be developed, including indicators for social and cultural determinants of health. This work would be significantly supported by implementing standardised tools for capturing social and cultural determinants of health, patient-reported experience measures, and patient-reported outcome measures in CISs and GP datasets.