Data Integration Project Research Articles

Implementing data governance in financial systems: Strategies for ensuring compliance and security in multi-source data integration projects

This study explores the complexities of implementing data governance within financial systems, focusing on strategies to ensure compliance and security in multi-source data integration projects. The primary aim was to identify challenges and establish a comprehensive framework to enhance data integrity, confidentiality, and regulatory adherence in the financial sector. Using a mixed-methods approach, the research included an extensive literature review, case studies, and expert interviews to gain insights into current practices, challenges, and emerging trends in data governance. Key findings reveal that while multi-source data integration offers significant opportunities for enhanced decision-making, it also presents substantial risks, including data breaches and regulatory non-compliance. The study underscores the pivotal role of data governance frameworks in mitigating these risks through the establishment of clear data policies, the implementation of robust cybersecurity measures, and the promotion of a compliance-focused organizational culture. Additionally, the analysis highlights the necessity for harmonized global data privacy laws to address the diverse regulatory environments that financial institutions operate within. The research concludes that effective data governance is crucial for financial institutions to safeguard their data assets, ensure regulatory compliance, and maintain stakeholder trust. Recommendations include the adoption of advanced technologies like artificial intelligence and blockchain to enhance data governance capabilities, the provision of ongoing employee training to foster a culture of compliance, and the implementation of dynamic compliance monitoring systems to keep pace with evolving regulatory demands. By embracing these strategies, financial institutions can develop a resilient data governance framework that not only ensures compliance and security but also drives operational excellence, competitive advantage, and sustained trust in the increasingly digital financial landscape. Keywords: Data Governance, Financial Systems, Multi-Source Data Integration, Regulatory Compliance, Cybersecurity, Data Privacy.

Development of governance for an integrated public data (GIPD) framework: illustrative use of GIPD in South Korea

PurposeUpon the realization of the need for guideline in cross-organizational data integration, in an exploratory manner, this study developed a public data governance framework, specifically, the governance for integrated public data (GIPD) framework and identified the influential factors of its successful implementation. This framework was then subjected to an analysis of a real data integration case in the South Korean public sector to test its efficacy.Design/methodology/approachTo develop the GIPD framework, the authors conducted an extensive meta study, focus group interviews and the analytic hierarchy process involving field experts. Further, the authors performed topic modeling on documents from Korean research and development data integration projects, and compared the extracted factors to those of the GIPD to illustrate the latter's usefulness in a real case.FindingsLegislation, policy goals and strategies, operation organization, decision-making council, financial support size and objective, system development and operation, data integration, data generation, system/data standardization and master data management were derived as the 10 important factors in implementing the GIPD framework. The illustrative case of Korea revealed that decision-making council, financial support size and objective, legislation, data generation and data integration were insufficient.Research limitations/implicationsAlthough this study reveals important findings, it has a few limitations. First, the potential factors for data governance might vary depending on the attribute of the “interviewee” (such as their career or experience period) and the goal and area of GIPD framework building. Second, the inherent limitation of topic modeling in determining topics from groups of extracted keywords means that topics may be interpreted in various ways, depending on the perspective of the expert.Practical implicationsThis study is highly significant in that it provides a starting point for discussions on the issue of data integration among public institutions. Therefore, although this study examined public data governance based on R&D data, it will contribute to providing a sufficient guideline for any type of inter-institutional data governance framework, what to discuss and how to discuss between institutions.Originality/valueThe findings are expected to provide a roadmap to formulate practical guidelines on inter-institutional data cooperation and a diagnostic matrix to improve the existing data governance system, especially in the public sector, from the existing practice of empirical analysis using a mixed methodology approach.

Demographic, health, and prognostic characteristics of Australians with liver cancer: a cohort study of linked data in New South Wales for informing cancer control

BackgroundAustralian age-standardized incidence and death rates for liver cancer are lower than world averages, but increasing as in other economically advanced western countries. World Health Organization emphasizes the need to address sociodemographic disparities in cancer risk. A more detailed sociodemographic risk profiling was undertaken for liver cancer in New South Wales (NSW) by diagnostic stage, than possible with NSW Cancer Registry (NSWCR) alone, by incorporating linked data from the Australian Bureau of Statistics (ABS). The purpose was to inform targeting and monitoring of cancer services.MethodsThe ABS manages the Multi-Agency Data Integration Project (MADIP) which includes a wide range of health, educational, welfare, census, and employment data. These data were linked at person level to NSWCR liver cancer registrations for the period post 2016 census to December 2018. De-identified data were analyzed. Sex-specific age-adjusted odds ratios (95%CIs) of liver cancer were derived using logistic regression by age, country of birth, residential remoteness, proficiency in spoken English, household income, employment status, occupation type, educational attainment, sole person household, joblessness, socioeconomic status, disability status, multimorbidity, and other health-related factors, including GP consultations. These data complement the less detailed sociodemographic data available from the NSWCR, with alignment of numerators and population denominators for accurate risk assessment.ResultsResults indicate liver cancer disproportionately affects population members already experiencing excess social and health disadvantage. Examples where 95% confidence intervals of odds ratios of liver cancer were elevated included having poor English-speaking proficiency, limited education, housing authority tenancy, living in sole-person households, having disabilities, multiple medicated conditions, and being carers of people with a disability. Also, odds of liver cancer were higher in more remote regions outside major cities, and in males, with higher odds of more advanced cancer stages (degrees of spread) at diagnosis in more remote regions.ConclusionsLinked data enabled more detailed risk profiling than previously possible. This will support the targeting of cancer services and benchmarking.

National estimates of occupation-related inequalities in all-cause mortality using linked Census-mortality data

ObjectiveThis report aims to provide national estimates of occupation-related inequalities in all-cause mortality for Australian residents aged 25–64 years. MethodData came from the 2016 Census linked to Deaths Registrations, available via the Multi-Agency Data Integration Project. Using negative binomial regression, we estimated age-adjusted relative and absolute inequalities in all-cause mortality rates in the 13 months following Census according to occupation, defined using the Australian and New Zealand Standard Classification of Occupations (eight major groups), using managers as the reference group. ResultsAmong 10.8M people, there were 20,987 deaths. Age-adjusted mortality rates were lowest among managers and professionals and were generally highest for manual occupations, for example, among men, relative risks (RR) for labourers ranged across age groups from 1.44 (95% CI 1.19–1.75, age 54–64) to 2.99 (1.93–4.65, age 25–34); among women, the RR for machine operators and drivers were 3.95 (1.39–11.21 in age 25–24 and 2.73 (1.66–4.49) in age 45–54, but there was relatively little variation by occupation in women aged 35–44 and 55–64. Around one in five deaths (23% for men, 17% for women) were associated with being in an occupation other than manager. ConclusionThese findings highlight that there is benefit in documenting national mortality inequalities according to occupation in addition to other measures of socioeconomic position. They provide further insights into socioeconomic inequalities in mortality. Implications for public healthMethods that aim to reduce mortality for those in manual occupations, particularly among young men, will reduce inequalities and improve population health.

Changes in general practice use and costs with COVID-19 and telehealth initiatives: analysis of Australian whole-population linked data.

In response to the COVID-19 pandemic, general practice in Australia underwent a rapid transition, including the roll-out of population-wide telehealth, with uncertain impacts on GP use and costs. To describe how use and costs of GP services changed in 2020 - following the COVID-19 pandemic and introduction of telehealth - compared with 2019, and how this varied across population subgroups. Linked-data analysis of whole-population data for Australia. Multi-Agency Data Integration Project data for ∼19 million individuals from the 2016 census were linked to Medicare data for 2019-2020. Regression models were used to compare age- and sex-adjusted GP use and out-of-pocket costs over time, overall, and by sociodemographic characteristics. Of the population, 85.5% visited a GP in Q2-Q4 2020, compared with 89.5% in the same period of 2019. The mean number of face-to-face GP services per quarter declined, while telehealth services increased; overall use of GP services in Q4 2020 was similar to, or higher than, that of Q4 2019 for most groups. The proportion of total GP services by telehealth stabilised at 23.5% in Q4 2020. However, individuals aged 3-14 years, ≥70 years, and those with limited English proficiency used fewer GP services in 2020 compared with 2019, with a lower proportion by telehealth, compared with the rest of the population. Mean out-of-pocket costs per service were lower across all subgroups in 2020 compared with 2019. The introduction of widespread telehealth maintained the use of GP services during the COVID-19 pandemic and minimised out-of-pocket costs, but not for all population subgroups.

Progressivity of out-of-pocket costs for Medicare-subsidised services and medicines in Australia.

ObjectivesIn line with affordability and equity principles, Medicare—Australia’s universal public health insurance system—has measures to limit out-of-pocket costs (OOPC), especially among lower income households. We examined the distribution of OOPC for Medicare-subsidised out-of-hospital services and prescription medicines, for Census households, according to their ability to pay. MethodsWe used 2016 Australian Census data linked to Medicare claims to obtain OOPC for out-of-hospital services and medicines in each household in 2017-18. We derived household disposable income by combining income information from the Census linked to income tax and social security data. All data were available from the Multi-Agency Data Integration Project, enabled through a partnership of various government agencies. We quantified OOPC as a proportion of equivalised household disposable income and calculated Kakwani indices (K) to measure progressivity. We also used linked National Health Survey data to analyse costs separately by chronic conditions. ResultsWe analysed 85% (n=6,830,365) of all Census private households. Overall, OOPC as a percentage of equivalised household disposable income decreased from 1.16% (out-of-hospital services) and 1.35% (prescription medicines) in the poorest decile to 0.63% and 0.34% in the richest decile, respectively. The regressive trend was less pronounced for out-of-hospital services (K = -0.06), with percentage OOPC relatively stable between the 2nd and 9th income deciles; while percentage OOPC decreased steeply with increasing income for medicines (K = -0.24). (Chronic conditions results will be presented—embargoed at time of submission) ConclusionOOPC for out-of-hospital Medicare services were mildly regressive while those for prescription medicines were distinctly regressive. Actions to reduce inequity in OOPC for medicines, such as reducing the co-payments for low income households should be considered.

Factors related to under‐treatment of secondary cardiovascular risk, including primary healthcare: Australian National Health Survey linked data analysis

ObjectiveTo inform national evidence gaps on cardiovascular disease (CVD) preventive medication use and factors relating to under‐treatment ‐ including primary healthcare engagement ‐ among CVD survivors in Australia. MethodsData from 884 participants with self‐reported CVD from the 2014–15 National Health Survey were linked to primary care and pharmaceutical dispensing data for 2016 through the Multi‐Agency Data Integration Project. Logistic regression quantified the relation of combined blood pressure‐ and lipid‐lowering medication use to participant characteristics. ResultsOverall, 94.8% had visited a general practitioner (GP) and 40.0% were on both blood pressure‐ and lipid‐lowering medications. Medication use was least likely in: women versus men (OR=0.49[95%CI:0.37‐0.65]), younger participants (e.g. 45–64y versus 65–85y: OR=0.58[0.42–0.79])and current versus never‐smokers (OR=0.73[0.44–1.20]). Treatment was more likely in those with ≥9 versus ≤4 conditions (OR=2.15[1.39–3.31]), with ≥11 versus 0–2 GP visits/year (OR=2.62[1.53–4.48]) and with individual CVD risk factors (e.g. high blood pressure OR=3.13 [2.34–4.19]) versus without); the latter even accounting for GP service‐use frequency. ConclusionsYounger people, smokers, those with infrequent GP visits or without CVD risk factors were the least likely to be on medication. Implications for public healthSubstantial under‐treatment, even among those using GP services, indicates opportunities to prevent further CVD events in primary care.

Towards Harmonized Data Quality in the Medical Informatics Initiative - Current State and Future Directions.

Health data from hospital information systems are valuable sources for medical research but have known issues in terms of data quality. In a nationwide data integration project in Germany, health care data from all participating university hospitals are being pooled and refined in local centers. As there is currently no overarching agreement on how to deal with errors and implausibilities, meetings were held to discuss the current status and the need to develop consensual measures at the organizational and technical levels. This paper analyzes the discovered similarities and differences. The result shows that although data quality checks are carried out at all sites, there is a lack of both centrally coordinated data quality indicators and a formalization of plausibility rules as well as a repository for automatic querying of the rules, for example in ETL processes.

Data Integration Project in Robologee

Robologee is a sub-unit of PT. Bangun Satya Wacana is part of Kompas Gramedia which is focused in Education section for ages 7 to 12 years. Robologee is a diversification of the existing sub-units in PT. Bangun Satya Wacana. Robologee has branches located at Gramedia World so it is expected that it will have an impact on Gramedia traffic. Currently, Robologee is transforming in order to integrate data that will be stored in the cloud by Amazon Web Service.The goal of this project is that data can be accessed by various users and stored in one platform. In the analysis of the digital transformation project, 15 respondents have been determined who are parents as external customers. Based on the indicators used in DMM. It was found that Robologee's current condition is at the Advancing level. Based on the Roadmap this project is implemented for 1 year and consists of four stages. In the Budgeting analysis, Robologee has payback period of 1.7 years with an IRR of 7.512% greater than the expected return of 5% by the company. Then the NVP is in a positive number, so this project is feasible to implement.

Using Census, Social Security and Tax Data from the Multi-Agency Data Integration Project (MADIP) to Impute the Complete Australian Income Distribution

Using Census, Social Security and Tax Data from the Multi-Agency Data Integration Project (MADIP) to Impute the Complete Australian Income Distribution

Evaluation of Data Integration Plans based on Graph Data

In data management, new data storage models and technologies are gradually adopted, including graph-based data models that have been enabled by technologies such as RDF, GraphQL and LPG. These are particularly fit when dealing with the coexistence of multitude and heterogeneous data sources that require an integration architecture in a data-centric organization due to intrinsic data connectedness. Two prominent approaches are distinguished in our study: (1) embracing a unifying graph format, where, following a series of transformations, all data sources are lifted to a common format in a consolidated graph repository, (2) creating a mediator tier where legacy data sources hold their data but are accessed through a virtual graph integration layer. Each approach comes with a specific way to query data and a plethora of query languages were created for this purpose. In the context of a data integration project we investigated two approaches to data retrieval and data modeling that have gained momentum during recent years: RDF graphs holding data lifted from non-graph data sources and GraphQL acting as a proxy for retrieving virtually connected data from heterogeneous sources. The paper reports on comparative experiments with technological instantiations of the two approaches, which can inform further IT strategies in an institutional project advocating a migration from app-centricity towards data connectedness.

Demonstrating the Utility of New Linked National Health Survey and Administrative Data: Secondary CVD Prevention in The Community

IntroductionCardiovascular events are largely preventable though access to timely quality primary health care and use of guideline-recommended medication. However, around half of Australians with cardiovascular disease (CVD) are not receiving best practice treatment. This study aims to identify factors associated with under treatment, using National Health Survey (NHS) linked for the first time to administrative health data.
 Objectives and ApproachParticipants with self-reported CVD in the NHS 2014-15 were included in the study, with data linked to Medicare (MBS) and pharmaceutical (PBS) data by the Australian Bureau of Statistics through the Multi-Agency Data Integration Project (MADIP). Use of primary and specialist ambulatory care and blood pressure- and lipid-lowering medications and their relation to sociodemographic and health characteristics were quantified using logistic and Poisson regression analyses.
 Results1100 participants with self-reported CVD were available for analysis, with linkage rates for NHS data to a Person Linkage Spine of 95%. We will present our findings from adjusted regression models (incomplete at time of abstract submission).
 Conclusion / ImplicationsThe nationally representative linked data developed under this project provides a unique opportunity to quantify and identify points to improve access to best practice CVD care, with the ultimate aim of preventing secondary CVD events in the population. Findings will also inform optimal use of MADIP data by the research community in order to answer questions of national importance and provide robust evidence to drive improvement in health and health care.

Defining Populations Within Linked Administrative Data

This presentation will outline methods developed by the ABS to define populations within linked administrative data. The methods developed use a combination of both direct and indirect signals (referred to as ‘signs of life’) to infer whether individuals are in a particular population at a given point in time. Data from multiple government sources are used.
 IntroductionLinked administrative datasets hold significant potential to unlock new insights and understanding about different populations of interest. A key challenge is to create a research dataset that is properly representative of the population of interest at a given point in time. Without a representative population to serve as the basis of analysis, research outcomes are harder to interpret and compare against those from other populations. This has flow on consequences for any research findings derived from linked admin data.
 Objectives and ApproachThis project sought to develop methods to define a representative Australian population from the Multi Agency Data Integration Project (MADIP) data asset. The core of the asset is created through a three-way linkage between Australian Medicare, Social Security and Taxation datasets. Together, these datasets have very high coverage of the Australian population and enable high quality linkage of other datasets into the asset. A ‘signs of life’ approach was used that sought to distinguish a representative population at a given point in time from the MADIP asset.
 ResultsAn experimental representative population has been developed from the linkage spine that closely approximates the national age distribution and other breakdowns of the ABS’s Estimated Resident Population.
 Conclusion / ImplicationsThis work demonstrates one approach that can be used to derive useful analytical populations from linked datasets with overly exhaustive scopes.

Using Newly Linked Data to Assess Equity of Out-Of-Pocket Healthcare Costs in Australia

IntroductionDescribing out-of-pocket (OOP) healthcare costs in relation to ability to pay requires multiple linked data sources not previously available. Current estimates of the progressivity of OOP healthcare costs in Australia are based on self-report surveys. Using newly linked Census to administrative income and medical claims data, we aimed to quantify, for the first time, the progressivity of OOP costs for government-subsidised out-of-hospital healthcare in Australia. Objectives and ApproachWe used Australian Census 2011 linked to Personal Income Tax (PIT), Medicare Benefits Schedule (MBS) and Pharmaceutical Benefits Scheme (PBS) data compiled through the Multi-Agency Data Integration Project (MADIP). Personal disposable income was estimated using a combination of PIT data and Census self-reported income, and aggregated across the household to estimate equivalised household income. We estimated annual MBS (out-of-hospital only) and PBS OOP costs as a proportion of equivalised household income, and assessed progressivity by reporting this for each income decile and computing a Kakwani Index. ResultsWe will present findings on progressivity overall, and separately by age, sex and location (incomplete at time of abstract submission). Conclusion / ImplicationsOur study will present one measure regarding the equity of healthcare costs, and help to identify vulnerable or at-risk groups. These findings may inform policy changes on equity in the financing of healthcare. Newly linked data from the MADIP can be used to relate healthcare costs to ability to pay.

Using Linked Whole-Of-Population Data to Estimate Education-Related Inequalities in Cause-Specific Mortality in Australia

IntroductionOfficial Australian estimates of socioeconomic inequalities in cause-specific mortality have been based on area-level socioeconomic measures. Using area-level measures is known to underestimate inequalities.
 Objectives and ApproachUsing recently released census linked to mortality data, we estimate education-related inequalities in cause-specific mortality for Australia. We used 2016 Australian Census and Death Registration data (2016-17) linked via a Person Linkage Spine (linkage rates: 92% and 97%, respectively) from the Multi-Agency Data Integration Project (MADIP). Education, from the Census, was categorised as low (no secondary school graduation or other qualification), intermediate (secondary graduation with/without other non-tertiary qualifications) and high (tertiary qualification). Cause of death was coded according to the underlying cause of death using the ICD-10. We used negative binomial regression to estimate relative rates (RR) for cause-specific mortality at ages 25-84 years, in the 12-months following Census, comparing low vs high education, separately by sex and 20-year age group, adjusting for age.
 Results80,317 deaths occurred among 13,856,202 people. For those aged 25-44 years, relative inequalities were large for causes related to injury and smaller for lesspreventable deaths (e.g. for men, suicide RR=5.6, 95%CI: 4.1-7.5 and brain cancer RR=1.3, 0.6-3.1). For those aged 45-64, inequalities were large for causes related to health behaviours and amenable to medical intervention, e.g. lung cancer (men RR= 6.4, 4.7-8.8) and ischaemic heart disease (women RR=5.0, 3.2-7.7), and were small for less preventable causes e.g. brain cancer (women RR=0.9, 0.6-1.3). Patterns among those aged 65-84years were similar to those aged 45-64 years.
 Conclusion / ImplicationsIn Australia, inequalities in mortality are substantial. Our findings highlight the health burden from inequalities, opportunities for prevention and provide insights on targets to effectively reduce them.

Higher Education Completion Rates for Domestic Undergraduate Students Who Access Mental Health Services Using A Multi-Agency Matched Population Cohort

By linking Education, Health, and Welfare data in the Multi-Agency Data Integration Project (MADIP), our analysis looked at the impact of poor mental health on the likelihood of completing an undergraduate degree in Australia.
 IntroductionCompletion of a bachelor degree is important to both the student and the government, as it provides lifelong benefits and prevents investment loss. Previous research has reported conflicting findings regarding whether students with mental ill health are less likely to complete a degree, with an estimated 25 per cent of young adult university students experiencing mental ill-health each year.
 Objectives and ApproachOur research analysed national mental health service use and related pharmaceutical prescriptions linked with education data to determine the extent and effect of known mental health conditions on undergraduate student six-year completion rates. We followed a de-identified cohort of 120,000 students who commenced an undergraduate degree for the first time in 2011 for six years. Summary statistics and a binomial logit was used on a matched sample to confirm significance.
 ResultsWe found that students with a known mental health condition had a significantly lower six-year completion rate (58 per cent) than those students with no known mental health condition (71 per cent). By simulating a randomised control trial controlling for a wide range of demographics, we showed that these results held and that completion rates worsened with increasing severity of mental health conditions, as measured by usage of psychiatric services.
 ConclusionIntegrated data assets such as MADIP help us better understand the interaction between student success and mental health conditions which in turn will help us improve policy and better evaluate programs.

Person Spine Linkage Methodology and Maintenance

IntroductionAs there is no single unique identifier for Australians across government datasets, commonwealth and state data must be brought together using data integration methods. The ABS has developed the Person Linkage Spine (the 'Spine') as an enduring piece of data linking infrastructure that is the core of the Multi-Agency Data Integration Project (MADIP) data asset, and is used to combine person-centred datasets.
 Objectives and ApproachThe Spine was created using a staged deterministic linkage process combining three core datasets: Medicare Consumer Directory, Social Security and Related Information, and Personal Income Tax. The Spine aims to broadly cover all people resident in Australia at any point during the reference period of 2006 to 2019. The Spine is used in combination with anonymised name and address history information from the core datasets to facilitate linkage of datasets to the Spine from any time period within its scope.
 The Spine was recently updated with new data from the core datasets in a 'Spine Maintenance' process, a key part of which involves the use of graph based methods to identify spine elements requiring updating in the new version. Spine Maintenance was conducted in such a way as to ensure consistency between Spine versions, enabling linkages to be conducted and retained across different versions of the Spine, whilst still allowing for improvements to the Spine based on updated information. This talk will focus on the methodological challenges associated with creating, using and maintaining the Person Linkage Spine.
 ResultsThe Spine has been successfully utilised as the core of the MADIP data asset, enabling valuable longitudinal person centred analysis.
 Conclusion / ImplicationsThe Spine has created a foundational linking infrastructure which is a sound basis for extending across time (longitudinally) and expanding to include datasets from across a broad range of sectors and jurisdictions.

Higher Education Completion Rates for Aboriginal and Torres Strait Islander Students Using A Multi-Agency Matched Population Cohort

Using the Multi-Agency Data Integration Project (MADIP), which combines health, tax, welfare and demographic data with student data, our analysis looked at the relationship between income support and Aboriginal and Torres Strait Islander university completion rates.
 IntroductionDomestic undergraduate university completion rates of Aboriginal and Torres Strait Islander students are significantly lower (40%) than non-Indigenous students (66%). Few prior studies have used population level matched data from multiple agencies to analyse the determinants of Australian Indigenous completion rates.
 Objectives and ApproachWe aimed to quantify the major determinants driving the completion rates of Indigenous students in Australian undergraduate university courses compared with domestic non-Indigenous students. We used the Higher Education Information Management System (HEIMS) linked to the MADIP creating approximately 555, 000 records. A Random forest tree was constructed to determine the most important indicators for outcome of interest which were then used for matching and statistical analysis. Summary statistics and a binomial logit was used on the matched sample to confirm significance.
 ResultsWe found that Indigenous students are more likely to start university belonging to around three equity groups such as having a lower socio-economic status background, older commencement age and being the first member of their family to attend university. However, Indigenous status remains a significant contributor to lower completion rates after controlling for a wide range of equity groups. One factor that has a positive influence on Indigenous university completion rates is access to study assistance. Completion rates for Indigenous students who were not members of other equity groups on income support was 70 per cent compared to 57 per cent for similar students on no income support.
 Conclusion / ImplicationsThese linked datasets provide the opportunity to better evaluate the drivers of completion rates of Australian Indigenous students to inform and evaluate policy reforms.

Learnings from Multi-Source Enduring Linked Data Assets (MELDAS)

IntroductionDemand continues to grow for accessible and large scale linked data assets to answer complex cross-sector, and cross-jurisdiction research questions. To meet this demand, a number of Multi-source, Enduring Linked Data Assets (MELDAs) have emerged including the National Integrated Health Service Infrastructure (NIHSI), National Disability Data Asset (NDDA) and Multi-Agency Data Integration Project (MADIP). Using these MELDAs has proven much more efficient than project-specific linkages, and provides consistent national data assets for multiple uses. However, the development of these assets raises new challenges, including complex data models, governance, and access arrangements, and new approaches to analysis.
 Objectives and ApproachThrough developing the NIHSI Analytical Asset in collaboration with state/territory and Federal Government partners, the AIHW has identified challenges in traditional linkage approaches, which require innovative approaches to ensure high quality linkage. As AIHW commences scoping on new MELDAs, we are taking lessons from building the NIHSI and applying them to future design.
 ResultsAIHW’s development of MELDAs across jurisdictions and portfolios provides new learnings on how to address advanced real world data integration issues. This review will focus on lessons learnt at the Australian Institute of Health and Welfare (AIHW) working with new data sharing arrangements, applications of technologies and innovative approaches to streamline MELDA processes.
 Conclusion / ImplicationsThe learnings from the AIHW development of MELDAs will assist others developing enduring assets to establish effective sharing arrangements, governance and technical solutions to ensure efficient management. These learnings will save time and resources, and prompt further discussion on a gold standard for building MELDAs moving forwards.

Creating Additional Census Income Topics from Linked Administrative Data

IntroductionThe Census is the largest statistical collection undertaken by the Australian Bureau of Statistics (ABS), with its data critical to informing the planning and delivery of Government and community services. While the Census measure of income supports a wide range of analysis, demand exists for additional income topics to complement and extend the range of socio economic analysis that can be undertaken. The ABS has recently developed three experimental income topics for the 2016 Census using linked administrative data: main source of income; main source of government payments; and previous financial year income.
 Objectives and ApproachThis research utilised administrative data integrated by the ABS for the Multi-Agency Data Integration Project (MADIP). Taxation data from the Australian Taxation Office (ATO) and social security data from the Department of Social Services (DSS) were used in conjunction with the 2016 Census data to derive the additional topics.
 ResultsOverall, the three measures compare relatively closely to similar measures from the ABS Survey of Income and Housing (SIH). The ‘Main source of income’ and ‘Main source of government payments’ measures exhibit similar distributions to those from the SIH. The ‘Previous financial year income’ measures compare more closely with Census and SIH at the higher end of the income distribution, with some differences apparent at the lower end of the income distribution.
 Conclusion / ImplicationsThis work demonstrates the potential to supplement and enhance existing Census topics with linked administrative data. Further research, development and consultation with data users and the Australian community is needed.