Study setting

The state of Queensland is situated in the north east of Australia, where 4.6% of the Queensland population identify as Aboriginal and Torres Strait Islander people, representing 28.7% of the national Indigenous population.8 In Queensland, one-third of Aboriginal and Torres Strait Islander people live in major cities (33%, 66 600), half in regional areas (51%, 109 100), 1 in 14 (7%, 14 300) in remote areas and 1 in 10 (9%, 20 100) in very remote areas.19 As in the rest of Australia, Queensland’s health system is a multifaceted combination of public and private providers.20 Overall management of the public health sector is the responsibility of the state government’s Department of Health and implementation is undertaken through 16 regional hospital and health services (HHS).21 In contrast, the majority of primary care and a considerable proportion of hospital care is provided by private sector providers.20 Aboriginal and Torres Strait Islander health services have complex funding arrangements through federal, state and territory governments, including subsidies provided through the nationally funded Medicare Benefits Scheme (MBS) and Pharmaceutical Benefits Scheme (PBS).20

Study design

This is a retrospective cohort study of people with a first hospitalisation of five CVDs (IHD, stroke, CCF, ARF and RHD) identified from the Queensland Hospital Admitted Patient Data Collection (QHAPDC). The first eligible hospital admission is referred to as the index admission. The study will use linked regional, state and national health and administrative data collections to investigate preventive primary healthcare before the index admission of each disease; acute care after the index admission; and in-hospital outcomes to assess the continuum of care provided against the clinical guidelines and best practice standards for primary and secondary prevention and in-hospital care. Where there are available data to do so, clinical care will be compared with the clinical guidelines current at the time: the Essential Service Standards for Equitable National Cardiovascular Care for Aboriginal and Torres Strait Islander People (ESSENCE)22 and the Australian Guideline for Prevention, Diagnosis and Management of Acute Rheumatic Fever and Rheumatic Heart Disease, Second Edition.23

Study population

The study population will be any person identified in the QHAPDC with a Queensland residential postcode who had a hospital admission between 1 July 2010 and 30 June 2016 and a primary or other diagnosis code of International Classification of Diseases codes (I20–I25), stroke (I61, I63 and I64), CHF (I50), ARF (I00–I02) and RHD (I05–I09).

Data sources

The research questions will be addressed by linking 11 Queensland datasets and three national datasets to the study population (table 1). The resultant linked deidentified dataset will include data from 1 July 2005 to 31 December 2018. Once the study population is identified, a project-specific linkage key will be assigned to each individual in the cohort by Data Linkage Queensland (DLQ), to link with other datasets. Other datasets considered that were not included due to the recentness of the collection were Queensland Non-admitted Patient Data Collection and Queensland Cardiac Outcomes Registry. Datasets that we planned to include but were unable to obtain custodian approval for included:Queensland Medical Laboratory and Sullivan and Nicolaides Pathology.

Data linkage process

DLQ will link the records using personal identifiers, including full name, full residential address, sex, full date of birth and date of death (where applicable) to those in the Master Linkage File for datasets provided securely to DLQ (figure 1). These datasets include Emergency Department Information System (EDIS)/Emergency Data Collection (EDC), Specialist Outpatient Data Collection (SODC), Queensland Death Register (QDR), Rheumatic Heart Disease Register and RHD Enhanced Surveillance Database. For Ferret, Best Practice Primary Healthcare Database (BP), Queensland Laboratory (AusLab), National Hospital Costing Data Collection (NHCDC) and Costing Funding Values (CFV), DLQ will supply the linking variables and the project specific linkage key to the Queensland Health data custodians to extract the approved data. Once extracted, the custodians will remove the identifiers and send the approved variables with the project-specific linkage key back to DLQ via a secure link. DLQ will then transfer the linked datasets to the Secure Unified Research Environment (SURE)24 supported by the Sax Institute for analysis by the project team. For the national datasets, MBS, PBS and National Death Index (NDI), DLQ will provide the linking variables and project specific linkage key to the Australian Institute of Health and Welfare (AIHW) data custodians who will link and extract the approved data and transfer the deidentified dataset to SURE.

• Download figure
• Open in new tab
• Download powerpoint

Figure 1

Flowchart of data linkage process, data custodians and datasets.

Data will be deterministically and probabilistically linked at the unit record level by DLQ25 and AIHW26 27 in accordance with ethical and public health approvals, and privacy considerations. All data-specific activities (ie, storage, quality assurance and analysis) will be done within SURE.24 Data quality reports of the linkage pairing will be provided by DLQ and AIHW.25 28

Cohorts

From the study population, six separate retrospective cohorts will be created each comprised of people with their first-ever hospitalisation for (1) CVD, (2) IHD, (3) CHF, (4) stroke, (5) ARF and (6) RHD. The first-ever CVD cohort (table 2) will include Queensland residents who had their first (index) hospital admission for one of the five CVDs, as indicated in the principal or other diagnosis fields in the QHAPDC, during July 2010 and June 2016. To maximise the likelihood that individuals enter the cohort at their first-ever CVD hospital admission, we will exclude those who had a hospital admission with a CVD principal or other diagnosis in the 5-year period prior to the index admission (referred to as the ‘lookback’ period). Individuals could only enter the study up until 30 June 2016 to ensure all individuals had a minimum of 2 years follow-up data to measure service use, cost and outcomes. The single disease cohorts will be created using the same approach; that is, individuals will enter the cohort at the index hospitalisation for the specific disease (eg, IHD) and will be excluded if they had a hospital admission in the lookback for the same specific disease (eg, IHD).

Statistical power for hypothesis testing

Aboriginal and Torres Strait Islander people and those living in regional and remote areas experience a disproportionately high burden of cardiac disease yet are typically underestimated in national surveys. Analysis of state-wide administrative data provides an opportunity to examine cardiac care and outcomes for underserved and at-risk populations that may otherwise not be possible.29 Investigation of variation in cardiac care and outcomes within the Aboriginal and Torres Strait Islander population is a primary aim of this study, and while the sample size of population subsets are relatively small, we are still adequately powered to detect important differences between groups.

Preliminary analysis indicates there are over 180 000 Queensland residents who were hospitalised for the first time for IHD, CHF, stroke, RHD or ARF during July 2010–June 2016, of whom over 5800 are identified as Aboriginal and Torres Strait Islander people. For the condition-specific cohorts, the cohort size varies from approximately 4200 (first IHD cohort) to 380 (first ARF cohort). Assuming 35% of the Aboriginal and Torres Strait Islander cohort receive an annual health check,10 a minimally important difference of 0.5 days in average length of stay or 1 day in time to CVD readmission, and an alpha of 0.001, we will have >90% power (beta 0.10) to statistically detect differences in these outcomes between Aboriginal and Torres Strait Islander people who received an annual health check and those who did not. For the IHD, CHF and stroke cohorts, but not the RHD and ARF cohorts, we are similarly powered to detect minimally important differences in survival estimates, as well as the proportion hospitalised with CVD within 28 days of index. As this is a whole-of-population cohort, we are unable to alter the cohort size. As such, underpowered analyses (ie, some RHD and ARF analyses) will be considered exploratory, and interpretation and reporting of the results will be cautiously conducted, with due consideration given to the limitations of the data and the plausibility of findings in light of existing published data.

Identifying Aboriginal and Torres Strait Islander people

Indigenous status is collected at point of care in Queensland hospitals using the standard question for Indigenous identification, ‘Are you of Aboriginal or Torres Strait Islander origin?’30 This is typically coded in health datasets as ‘Aboriginal and Torres Strait Islander origin’, ‘Aboriginal but not Torres Strait Islander origin’, ‘Torres Strait Islander but not Aboriginal origin’, ‘neither Aboriginal or Torres Strait Islander’ and ‘not stated or unknown’, as it is in the QHAPDC. In the Queensland healthcare system, Aboriginal and Torres Strait Islander people have the right to self-identify or not, at each healthcare presentation. Incomplete and inconsistent reporting of Indigenous status usually results in an underestimation of the Indigenous population and their use of services, which has an impact on the accurate planning and delivery of services to Aboriginal and Torres Strait Islander people.31

National best practice guidelines for analysis of linked administrative data related to Aboriginal and Torres Strait Islander people published by the AIHW recommends the use of algorithms that draw on multiple data to enhance the completeness and accuracy of Indigenous status information.32 It is recommended that multiple algorithms be explored and outcome measures using each of the algorithms be compared to determine the sensitivity of the approach used.32 We will use three of the recommended algorithms and, for each individual in the study, will draw on Indigenous identification information from all QHAPDC records (2005–2018) (table 3). The QHAPDC has reasonably high levels of accuracy (>80%) for Indigenous status and is considered gold standard among administrative datasets for Indigenous identification data.33 Additionally, we will use a fourth algorithm, using MBS data to enhance the QHAPDC majority-based algorithm.

Preliminary data suggest the three different AIHW Indigenous status algorithms based on QHAPDC data only yielded different proportions of Indigenous women in the QHAPDC, with an absolute difference of 0.8% (n=1468).

Other covariates

Covariates measured in this study include person-level factors (eg, sex and age), service-level factors (eg, HHS and type of facility) and area factors (remoteness of residence and area-level socioeconomic status):

• Sex (male/female) captured in the index hospitalisation record.

• Age, which will be derived from birth date (day (DD)/month (MM)/ year (YYYY)) and index hospitalisation admission date (DD/MM/YYYY), both of which will be captured in the index hospitalisation record.

• Remoteness of residence will be measured using the Australian Remoteness Index of Areas, based on the Statistical Area (SA2) of an individual’s residential address, as captured in the index hospitalisation record. Individuals will be categorised as living in major cities, inner regional, outer regional, remote and very remote areas of Australia.34

• Area-level socioeconomic status will be measured using the Index of Relative Socioeconomic Advantage and Disadvantage, based on the SA2 of an individual’s residential address, as captured in the index hospitalisation record. This 100-point index will be categorised into quintiles, with quintile 1 representing the most disadvantaged and quintile 5 representing the most advantaged.35

• Comorbidities will be identified in the hospitalisation records in the 5 years prior to the index hospitalisation and will be measured as the presence/absence of specific conditions deemed important to the development or prognosis of cardiovascular disease (eg, hypertension and diabetes).

• Comorbidity level, also informed by hospitalisation records from the 5-years prior to the index hospitalisation, is measured using the Elixhauser Comorbidity Index and categorised as ‘no known comorbidity’, ‘one known comorbidity’, ‘two known comorbidities’, ‘three known comorbidities’ and ‘four or more known comorbidities’.36

• HHS area of the facility of the hospital that an individual was first admitted to for their index hospitalisation.

Outcomes

The following CVD outcomes will be examined:

1. Primary prevention—for the 5 years prior to the index hospitalisation, we will describe and quantify the annual rate of attendance to general practitioner (GP)/specialist appointments and completion of the adult health check from items in EDIS, EDC, MBS, SODC, Ferret and BP. PBS items will be used to determine what proportion of the study cohort were receiving medications prescribed for CVD.

2. Secondary prevention—for the 2-years post the index hospitalisation, we will quantify the average time to first disease management (emergency department, filling prescriptions, scheduled appointments and rehabilitation) and the proportion of the study cohort who saw a GP, Aboriginal and/or Torres Strait Islander health practitioner, practice nurse, allied health practitioner and/or specialist, for management of their CVD, using items from EDIS, EDC, PBS, AusLab, MBS, SODC, Ferret and BP. Individuals will be censored or removed from the denominator if they die during the follow-up period; this will be determined from the QDR and NDI. PBS data will be used to measure the proportion of the cohort who are on appropriate cardiac medication in the 30 days, 1-year and 2-years postdischarge from the index hospitalisation.

3. In-hospital therapeutic procedures for acute coronary syndrome (ACS)—for the index ACS admission, we will describe the proportion of the cohort that received diagnostic angiography, cardiac reperfusion and revascularisation within the index admission and within 30 days of the index admission.

4. Survival—for each cohort, we will quantify 30 day, 1-year and 2-year survival (excess mortality) using the QHAPDC, QDR and NDI.

5. Service use—length of stay, readmission rates and post-in-hospital procedure complications will be measured using data from the QHAPDC.

6. Health system expenditure and out-of-hospital costs—using items from NHCDC and CFV, the average cost associated with each hospital admission will be determined. MBS and PBS items will be used to describe out-of-hospital costs for GP/specialist visits and relevant CVD medications.

Multivariable analyses will maximise the impact of the study by allowing us to (1) model variation in subgroups and investigate temporal trends, (2) examine interaction effects between person and system level factors, and (3) adjust for known and measured confounders to estimate causal effects between care receipt and outcomes. Measured and known confounders include the covariates listed previously, namely, age, sex, remoteness of residence, area-level socioeconomic status, pre-existing comorbidity level and HHS area. Due to the nature of routinely collected data, we will not have information or complete information on all potential confounders (eg, smoking status and Body Mass Index) and thus the causal effects derived from this study will be interpreted in the context of these data quality considerations. We will compare the prevalence of these confounders, or risk factors, for Aboriginal and Torres Strait Islander people with other patients to assess their greater risk of adverse outcomes and compare the difference in outcomes for Aboriginal and Torres Strait Islander people with other patients before and after adjustment for these risk factors in multivariable analysis to assess the extent to which adjustment for these risk factors reduces the disparity in clinical outcomes.

Absolute risk measures will be reported for the cohort overall and by population groups. Where appropriate, relative measures may also be reported to compare subgroups (eg, Indigenous vs non-Indigenous) or temporal trends. The reporting of both absolute and relative measures of risk is commonly recommended as gold standard reporting practice as this gives a more comprehensive picture of inequities and the implications of it.37 38 The use of generalised linear models with the assumption of normal, gamma or Poisson distributions will be used for different variable types. Cox models for time-to-event data (eg, survival) and mixed models or generalised estimating equations for longitudinal data will be used to measure change over time, including where temporal effects exist. Variation in care pathways and patient outcomes will be compared by Indigenous status, sex, age group, remoteness of residence, year of index hospitalisation and socioeconomic status.

How this study will contribute to the evidence base

The higher incidence of and mortality from CVD for Aboriginal and Torres Strait Islander than other Australians is well documented, but there has been only limited investigation of access to and effectiveness of diagnostic and clinical care for Aboriginal and Torres Strait Islander people (AIHW, 2018b) and for other underserved populations,39 40 and most Australian analyses of CVD using linked datasets have been conducted in NSW and Western Australia.41 This study will address these knowledge gaps by investigating these issues in detail for five cardiovascular conditions in a population-based study for the entire Queensland population. Additionally, it will examine the impact of adverse clinical care on service use, health system costs and patient survival across population groups to identify service gaps and at-risks groups. An Indigenous-majority cardiovascular advisory group will review study findings and propose priorities for research, practice and policy to improve cardiovascular care and reduce disparities in cardiovascular outcomes for Aboriginal and Torres Strait Islander people.