Work

Data Resource Profile: Victorian Comprehensive Cancer Centre Data Connect

Year: 2023

Type: article

Abstract: Data Connect has been established as a data linkage resource that brings together Australian General Practice (GP), Metropolitan Hospital and Cancer Registry data sources to enable researchers to address questions in health services research related to the entire patient continuum of care. This resource includes patient-level de-identified data on over 3 million Victorian primary care patients aged over 18 years of age, with GP data covering the period 1996–2021 and hospital data covering the years 2006–21. Data have been linked across health services as this is a requirement for understanding patients’ journey of care, in particular for cancer patients, who interact with many different parts of the health care system, often starting at primary care. The resource includes rich clinical and administrative data from hospital presentations and admissions, GP encounters, prescriptions, pathology test results and observations and cancer diagnoses. Researchers interested in using this resource should contact Data Connect at [https://vcccalliance.org.au/our-work/research-and-translation/data-connect]. Victorian Comprehensive Cancer Centre (VCCC) Data Connect is a single platform for facilitating access to a range of Victorian health data sources, allowing researchers, analysts and policy makers to bring a rigorous, data-driven lens to questions related to health services and patterns of cancer care. It has been developed in Australia since 2017 to connect general practice databases with data from hospitals and clinical cancer registries, and involves nearly 3 million patients spanning the years 1996–2021. It is updated on a yearly basis to incorporate more recent datasets, and is being expanded to include additional health services datasets across the state, including from non-metropolitan hospitals, VCCC Alliance members and clinical registries. In this paper we provide high-level descriptive statistics of the data sources, summarizing representativeness, data quality and overall sizes of patient cohorts. As a specific example we then focus on general cancer statistics, to illustrate the utility of this resource. Cancer patients interact with the health system through using many different health services, with primary care playing a critical role from initial presentation through to survivorship and end-of-life care.1,2 Each health service has its own corresponding data collection systems, so linkage is critical to understand how patients interact with the system across their entire journey of care.3 In Australia, health data linkage is challenging due to different parts of the health system being the responsibility of different jurisdictions at state and federal levels, as well as due to differing data standards and systems.4 Since primary care is usually the first point of contact that patients have with the health system,5 including prior to a diagnosis of cancer, General Practice (GP) data are a key component of research in this area. The intended audience of this paper is researchers interested in data, research and analysis related to health service use. Research studies using these data are already under way, including in the study of times to diagnosis and treatment, pathways of care, early detection of cancer and cancer survivorship. Whereas our particular focus is on cancer, Data Connect contains a wealth of information on other cohorts of patients with broader applications, and the resource is of value for health services researchers in different clinical domains. Linked data resources are becoming increasingly important for health research in many countries.6–8 In the UK, the Clinical Practice Research Datalink (CPRD) is one such resource, covering almost 7% of the UK population, and has been used in over 1000 studies.6 In Australia, state and federal governments have their own linked data infrastructure. Examples include the Centre for Victorian Data Linkage,9 Data Linkage Western Australia,10 the Australian Institute of Health and Welfare11 and also the Australian Bureau of Statistics’ own linkage capabilities.12 However in Australia, primary care is often overlooked in linked data research, despite its critical role in the health system.4,5 For example, in the 2020–21 financial year, over 85% of Australians had at least one contact with a GP service.13 A key source of general practice data is the MedicineInsight database, a representative database covering the entire country.14 Another is the University of Melbourne Patron Data for Decisions Program.4 Few studies have involved linking health services datasets to these types of general practice data sources that contain rich clinical information in addition to billing information.15 Data access is provided by BioGrid Australia,16–18 a not-for-profit organization that provides an online platform for data linkage and management, so data storage and linkage infrastructure are handled by a single entity. Data linkage in Data Connect is carried out through the use of a Unique Subject Identifier (USI), a unique encrypted key for each patient. For hospital data these USI values are generated by processing patient-identifying information through BioGrid’s probabilistic record linkage key software and using the GRHANITETM data linkage software,19 whereas for GP data these are generated at the practice using GRHANITETM. The USI effectively anonymizes patient records as no identifying information is released to researchers, protecting patient privacy and ensuring that data governance, ethics and privacy requirements of data custodians that are met. Data Connect uses the BioGrid federated data infrastructure, which means that it is covered by BioGrid’s data governance model.18 A collaboration agreement among BioGrid member institutions ensures that there is no need to create multiple agreements for different health service sites in order to apply for data access.16 In addition, the human research ethics framework is managed by BioGrid, so that data access, governance and management across sites are covered by a single agreement, removing the need for researchers to approach data custodians individually and ensuring more timely access to data for research. It is important to note that data custodian and ethical approval is required for each research project prior to linked data being made available to researchers through BioGrid. Data linkage allows researchers, analysts and policy makers to capture a more complete picture of patients’ interactions with the health care system.3 Primary care has a critical role through the patient’s continuum of care,1,2 as illustrated in the Figure 1, and linking data from general practice, hospitals and clinical registries allows tracking of patients along this continuum. Stages in the continuum of care and relevant data sources. Patron, University of Melbourne general practice database; VAED, Victorian Admitted Episodes Dataset; VINAH, Victorian Integrated Non-Admitted Health dataset Distribution of patients by age and sex in MedicineInsight vs 2016 Census (left) and Patron vs 2016 Census (right). ABS, Australian Bureau of Statistics; Patron, University of Melbourne general practice database Figure 1 illustrates the relation between points along the cancer continuum of care and corresponding datasets that are available through Data Connect. In this case we provide the example of a patient presenting to general practice with a symptom, and points along their care pathway until diagnosis and treatment. Different events are captured in different electronic health records systems, so linkage is crucial to fully understand patients’ interactions. Questions related to early diagnosis of cancer involve general practice and hospital data, and questions related to survivorship relate to both earlier and later stages. Data Connect currently contains the following datasets: MedicineInsight: General Practice database (Victorian subset): MedicineWise14,20; VAED: Victorian Admitted Episodes Dataset (four hospital networks); VEMD: Victorian Emergency Minimum Dataset (three hospital networks); VINAH: Victorian Non-Admitted Episodes Health dataset (two hospital networks); Patron: University of Melbourne general practice database4; VCR Pilot: A sample of patients from the Victorian Cancer Registry; ACCORD: Australian Comprehensive Cancer Outcomes and Research Database (colorectal and sarcoma subsets); AURORA: AUstralian Registry and biObank of thoRAcic cancers. Additional work is under way to link radiotherapy data, additional cancer registry data and genomics datasets. The datasets contain patient demographic information as well as data related to patient interactions with health services. Table 1 provides a more detailed (though not exhaustive) list of information available. We also highlight certain items that are not available, mainly due to privacy. Further detailed information on each of these is available in the references. Note that the National Prescribing Service ceased operation at the end of 2022 and the primary care database is now managed by the Australian Commission on Safety and Quality in Health Care. Summary of data sources MedicineInsight (General Practice database, Victorian subset) Encounter date Patient residence (at SA3) Patient age Patient sex Health care card status Alcohol status Smoking status ATSI status Requested tests Test results Prescriptions Billing Observations such as weight and height measurements Reason for encounter, including signs and symptoms Conditions Diagnosesa Comorbidities are captured in the reason for encounter, conditions and diagnoses free-text fields Clinical notes Time of encounter Practice location Patient ethnicity Patient country of birth Patron (University of Melbourne general practice database) Encounter date Patient residence (at SA3) Patient age Patient sex Health care card status Alcohol status Smoking status ATSI status Requested tests Test results Prescriptions Billing Observations such as weight and height measurements Reason for encounter, including signs and symptoms Conditions Diagnosesa Comorbidities are captured in the reason for encounter, conditions and diagnoses free-text fields Clinical notes Time of encounter Practice location Patient ethnicity Patient country of birth VAED (Victorian Admitted Episodes dataset) Date of admission Admission site ICD-10 diagnosis Diagnosis date Separation dates Patient age Patient sex CALD indicators Clinical notes VEMD (Victorian Emergency Minimum Dataset) Presentation date Presentation site Patient postcode CALD indicators Clinical notes VINAH (Victorian Integrated Non-Admitted Health dataset) Episode date Episode site Patient postcode Clinical notes Victorian Cancer Registry Pilot (Sample of patients from the Victorian Cancer Registry) Diagnosis date Patient age Patient sex Patient date of birth Patient country of birth Patient residence (at SA2) Cancer diagnosis code (ICD-10-AM) Tumour site Tumour stream SA1 geography Limited stage data AURORA (AUstralian Registry and biObank of thoRAcic cancers) Diagnosis date Treatment date Patient age Patient sex Tumour stage Treatment type and usage Comorbidities Date of death Patient residence Limited symptom data ACCORD Colorectal (Australian Comprehensive Cancer Outcomes and Research Database, colorectal subset) Diagnosis and treatment dates Patient age Patient sex Patient weight Tumour stage Chemotherapy drugs used Dose Radiotherapy fractions Reasons for changes in dose Resections Comorbidities Cancer history Patient residence Limited symptom data ACCORD Sarcoma (Australian Comprehensive Cancer Outcomes and Research Database, sarcoma subset) Patient age Patient sex Cancer history Death date Referring specialty First consultation details Symptom history Pre-treatment imaging status Diagnosis and treatment dates Tumour stage Biopsy details Resection details Surgery details Radiotherapy details Chemotherapy details Limited symptom data MedicineInsight (General Practice database, Victorian subset) Encounter date Patient residence (at SA3) Patient age Patient sex Health care card status Alcohol status Smoking status ATSI status Requested tests Test results Prescriptions Billing Observations such as weight and height measurements Reason for encounter, including signs and symptoms Conditions Diagnosesa Comorbidities are captured in the reason for encounter, conditions and diagnoses free-text fields Clinical notes Time of encounter Practice location Patient ethnicity Patient country of birth Patron (University of Melbourne general practice database) Encounter date Patient residence (at SA3) Patient age Patient sex Health care card status Alcohol status Smoking status ATSI status Requested tests Test results Prescriptions Billing Observations such as weight and height measurements Reason for encounter, including signs and symptoms Conditions Diagnosesa Comorbidities are captured in the reason for encounter, conditions and diagnoses free-text fields Clinical notes Time of encounter Practice location Patient ethnicity Patient country of birth VAED (Victorian Admitted Episodes dataset) Date of admission Admission site ICD-10 diagnosis Diagnosis date Separation dates Patient age Patient sex CALD indicators Clinical notes VEMD (Victorian Emergency Minimum Dataset) Presentation date Presentation site Patient postcode CALD indicators Clinical notes VINAH (Victorian Integrated Non-Admitted Health dataset) Episode date Episode site Patient postcode Clinical notes Victorian Cancer Registry Pilot (Sample of patients from the Victorian Cancer Registry) Diagnosis date Patient age Patient sex Patient date of birth Patient country of birth Patient residence (at SA2) Cancer diagnosis code (ICD-10-AM) Tumour site Tumour stream SA1 geography Limited stage data AURORA (AUstralian Registry and biObank of thoRAcic cancers) Diagnosis date Treatment date Patient age Patient sex Tumour stage Treatment type and usage Comorbidities Date of death Patient residence Limited symptom data ACCORD Colorectal (Australian Comprehensive Cancer Outcomes and Research Database, colorectal subset) Diagnosis and treatment dates Patient age Patient sex Patient weight Tumour stage Chemotherapy drugs used Dose Radiotherapy fractions Reasons for changes in dose Resections Comorbidities Cancer history Patient residence Limited symptom data ACCORD Sarcoma (Australian Comprehensive Cancer Outcomes and Research Database, sarcoma subset) Patient age Patient sex Cancer history Death date Referring specialty First consultation details Symptom history Pre-treatment imaging status Diagnosis and treatment dates Tumour stage Biopsy details Resection details Surgery details Radiotherapy details Chemotherapy details Limited symptom data ATSI, Aboriginal and Torres Strait Islander; CALD, culturally and linguistically diverse; ICD-10, International Classification of Diseases, Tenth Revision; ICD-10-AM, International Classification of Diseases, Tenth Revision, Australian Modification; SA1, Statistical Area Level 1, from the Australian Bureau of Australian Statistical Statistical Area Level from the Australian Bureau of Australian Statistical Statistical Area Level from the Australian Bureau of Australian Statistical are in free-text and by are not on the International Classification of Primary Summary of data sources MedicineInsight (General Practice database, Victorian subset) Encounter date Patient residence (at SA3) Patient age Patient sex Health care card status Alcohol status Smoking status ATSI status Requested tests Test results Prescriptions Billing Observations such as weight and height measurements Reason for encounter, including signs and symptoms Conditions Diagnosesa Comorbidities are captured in the reason for encounter, conditions and diagnoses free-text fields Clinical notes Time of encounter Practice location Patient ethnicity Patient country of birth Patron (University of Melbourne general practice database) Encounter date Patient residence (at SA3) Patient age Patient sex Health care card status Alcohol status Smoking status ATSI status Requested tests Test results Prescriptions Billing Observations such as weight and height measurements Reason for encounter, including signs and symptoms Conditions Diagnosesa Comorbidities are captured in the reason for encounter, conditions and diagnoses free-text fields Clinical notes Time of encounter Practice location Patient ethnicity Patient country of birth VAED (Victorian Admitted Episodes dataset) Date of admission Admission site ICD-10 diagnosis Diagnosis date Separation dates Patient age Patient sex CALD indicators Clinical notes VEMD (Victorian Emergency Minimum Dataset) Presentation date Presentation site Patient postcode CALD indicators Clinical notes VINAH (Victorian Integrated Non-Admitted Health dataset) Episode date Episode site Patient postcode Clinical notes Victorian Cancer Registry Pilot (Sample of patients from the Victorian Cancer Registry) Diagnosis date Patient age Patient sex Patient date of birth Patient country of birth Patient residence (at SA2) Cancer diagnosis code (ICD-10-AM) Tumour site Tumour stream SA1 geography Limited stage data AURORA (AUstralian Registry and biObank of thoRAcic cancers) Diagnosis date Treatment date Patient age Patient sex Tumour stage Treatment type and usage Comorbidities Date of death Patient residence Limited symptom data ACCORD Colorectal (Australian Comprehensive Cancer Outcomes and Research Database, colorectal subset) Diagnosis and treatment dates Patient age Patient sex Patient weight Tumour stage Chemotherapy drugs used Dose Radiotherapy fractions Reasons for changes in dose Resections Comorbidities Cancer history Patient residence Limited symptom data ACCORD Sarcoma (Australian Comprehensive Cancer Outcomes and Research Database, sarcoma subset) Patient age Patient sex Cancer history Death date Referring specialty First consultation details Symptom history Pre-treatment imaging status Diagnosis and treatment dates Tumour stage Biopsy details Resection details Surgery details Radiotherapy details Chemotherapy details Limited symptom data MedicineInsight (General Practice database, Victorian subset) Encounter date Patient residence (at SA3) Patient age Patient sex Health care card status Alcohol status Smoking status ATSI status Requested tests Test results Prescriptions Billing Observations such as weight and height measurements Reason for encounter, including signs and symptoms Conditions Diagnosesa Comorbidities are captured in the reason for encounter, conditions and diagnoses free-text fields Clinical notes Time of encounter Practice location Patient ethnicity Patient country of birth Patron (University of Melbourne general practice database) Encounter date Patient residence (at SA3) Patient age Patient sex Health care card status Alcohol status Smoking status ATSI status Requested tests Test results Prescriptions Billing Observations such as weight and height measurements Reason for encounter, including signs and symptoms Conditions Diagnosesa Comorbidities are captured in the reason for encounter, conditions and diagnoses free-text fields Clinical notes Time of encounter Practice location Patient ethnicity Patient country of birth VAED (Victorian Admitted Episodes dataset) Date of admission Admission site ICD-10 diagnosis Diagnosis date Separation dates Patient age Patient sex CALD indicators Clinical notes VEMD (Victorian Emergency Minimum Dataset) Presentation date Presentation site Patient postcode CALD indicators Clinical notes VINAH (Victorian Integrated Non-Admitted Health dataset) Episode date Episode site Patient postcode Clinical notes Victorian Cancer Registry Pilot (Sample of patients from the Victorian Cancer Registry) Diagnosis date Patient age Patient sex Patient date of birth Patient country of birth Patient residence (at SA2) Cancer diagnosis code (ICD-10-AM) Tumour site Tumour stream SA1 geography Limited stage data AURORA (AUstralian Registry and biObank of thoRAcic cancers) Diagnosis date Treatment date Patient age Patient sex Tumour stage Treatment type and usage Comorbidities Date of death Patient residence Limited symptom data ACCORD Colorectal (Australian Comprehensive Cancer Outcomes and Research Database, colorectal subset) Diagnosis and treatment dates Patient age Patient sex Patient weight Tumour stage Chemotherapy drugs used Dose Radiotherapy fractions Reasons for changes in dose Resections Comorbidities Cancer history Patient residence Limited symptom data ACCORD Sarcoma (Australian Comprehensive Cancer Outcomes and Research Database, sarcoma subset) Patient age Patient sex Cancer history Death date Referring specialty First consultation details Symptom history Pre-treatment imaging status Diagnosis and treatment dates Tumour stage Biopsy details Resection details Surgery details Radiotherapy details Chemotherapy details Limited symptom data ATSI, Aboriginal and Torres Strait Islander; CALD, culturally and linguistically diverse; ICD-10, International Classification of Diseases, Tenth Revision; ICD-10-AM, International Classification of Diseases, Tenth Revision, Australian Modification; SA1, Statistical Area Level 1, from the Australian Bureau of Australian Statistical Statistical Area Level from the Australian Bureau of Australian Statistical Statistical Area Level from the Australian Bureau of Australian Statistical are in free-text and by are not on the International Classification of Primary Table general data quality in the datasets that Data Connect. detailed data quality work for each of the data sources in the references. In for a data quality should first carried out on the specific research as of The statistics should researchers understand of the data quality that Note that both MedicineInsight and Patron are from and Patron also data from As a the data are with and patients are to Patron and A list of data quality Australian Comprehensive Cancer Outcomes and Research Database (colorectal AUstralian Registry and biObank of thoRAcic General Practice database (Victorian Patron, University of Melbourne general practice database; Statistical Area Level Statistical Area Level For VAED, Victorian Admitted Episodes VCR a sample of patients from the Victorian Cancer Registry; Victorian Emergency Minimum Dataset; VINAH, Victorian Integrated Non-Admitted Health A list of data quality Australian Comprehensive Cancer Outcomes and Research Database (colorectal AUstralian Registry and biObank of thoRAcic General Practice database (Victorian Patron, University of Melbourne general practice database; Statistical Area Level Statistical Area Level For VAED, Victorian Admitted Episodes VCR a sample of patients from the Victorian Cancer Registry; Victorian Emergency Minimum Dataset; VINAH, Victorian Integrated Non-Admitted Health Data linkage statistics of data that Data Connect Requested Billing Requested Billing Admitted Requested Billing Requested Billing Admitted Australian Comprehensive Cancer Outcomes and Research Database (colorectal Australian Comprehensive Cancer Outcomes and Research Database AUstralian Registry and biObank of thoRAcic General Practice database (Victorian Patron, University of Melbourne general practice database; VAED, Victorian Admitted Episodes VCR a sample of patients from the Victorian Cancer Registry; Victorian Emergency Minimum Dataset; VINAH, Victorian Integrated Non-Admitted Health Data linkage statistics of data that Data Connect Requested Billing Requested Billing Admitted Requested Billing Requested Billing Admitted Australian Comprehensive Cancer Outcomes and Research Database (colorectal Australian Comprehensive Cancer Outcomes and Research Database AUstralian Registry and biObank of thoRAcic General Practice database (Victorian Patron, University of Melbourne general practice database; VAED, Victorian Admitted Episodes VCR a sample of patients from the Victorian Cancer Registry; Victorian Emergency Minimum Dataset; VINAH, Victorian Integrated Non-Admitted Health detailed statistics of datasets in Data Connect are in Table in Table is for the general practice datasets, covering parts of with in the of In there general in so of general in are For the hospital datasets, no hospitals the are currently available. General Practice database (Victorian Patron, University of Melbourne general practice database; For VAED, Victorian Admitted Episodes a sample of patients from the Victorian Cancer Registry; Victorian Emergency Minimum Dataset; VINAH, Victorian Integrated Non-Admitted Health General Practice database (Victorian Patron, University of Melbourne general practice database; For VAED, Victorian Admitted Episodes a sample of patients from the Victorian Cancer Registry; Victorian Emergency Minimum Dataset; VINAH, Victorian Integrated Non-Admitted Health Figure the age of patient in Patron and MedicineInsight data the broader Victorian in of of age and there are a of patients, This is with other studies that are more to a 18 are not as there is no ethical approval to include in Data Connect. For example, BioGrid under data from we these patients out as there is no ethical approval to include Cancer diagnoses are in registries as well as hospital admission the of different types of in a specific patient in this patients who to one of the hospitals for which data are available through Data Connect. A of patients are with cancer in VAED data, of the The Classification of Diseases, Tenth Revision, Australian have been used to to the specific cancer The results are in Table Cancer types and of patients ICD-10-AM, International Classification of Diseases, Tenth Revision, Australian Cancer types and of patients ICD-10-AM, International Classification of Diseases, Tenth Revision, Australian The GP data is from and The data linkage software GRHANITETM is at each of the Hospital data sources are from the at Metropolitan Melbourne hospitals and are then provided to BioGrid for are many studies currently under way that use of Data Connect. The focus of each involves a different of the continuum of care for cancer include the to diagnosis is an important of cancer data linked between Patron and the AURORA and ACCORD clinical registries, researchers have been for and colorectal Primary care test use and early detection of in the to a diagnosis of cancer, in pathology tests and signs and symptoms the early detection of cancer through to focus on clinical that are to of there is currently a of standards in Australia for from electronic health are of and other which of research In the UK, Health Data Research has an example of such a developed in the is A recent a of for a to in the way that electronic health care data are in Australia, a resource as This provide a way for researchers to that in studies Victorian health and research of Data Connect include the detailed data on over 3 million patients in and its to a range of questions in health services research. It brings together general practice, hospital and clinical registry data and is for the entire patient journey from a patient’s first presentation at a health service through to treatment, survivorship and end-of-life care. clinical information is available test dates of and demographic information such as age, sex and of allowing study of many different questions in the continuum of cancer care. The of types of questions we are in colorectal cancer the utility of these linked datasets. of Data Connect include the The quality and of data across different types and between datasets, which the of certain studies for example, cancers. hospitals in Melbourne are and over general practice sites provide This is over of in are data on and and of cohorts for which general practice data in relation to there are many patients across datasets, for a this is as in the colorectal cancer In with jurisdictions such as the UK, is so this a to studies sample for example the of for specific information access to Data Connect data, contact VCCC Data and Research the Data Connect at This work by Melbourne Health the initial and the overall provided statistics for the hospital and primary care data sources. provided the governance and the entire paper and provided statistics data sources through the including the cancer registries. This work by the and Victorian of Health through the Victorian Comprehensive Cancer Centre The BioGrid Australia for their work in this resource and to this as well as the This research used de-identified patient data from the Patron primary care data from general which has been and is by the of General University of Melbourne The have no of

Source: International Journal of Epidemiology

DOI: 10.1093/ije/dyad148