Who are we? Big data in primary care How should it be governed? What should it be governed for? Consumer Clinical medical Clinical nurse Clinical allied health Managerial Policy Decision maker Others Why are we workshopping? Recognise that GP/PC EHR data will be used beyond the clinician-patient relationship. This secondary use has implications for clinicians and managers as data creators. Data quality management, provenance and governance processes and structures must be transparent and acceptable to consumers, GPs and their representative organisations. Program When? What? Who? 2.15 2.20pm Introductions & Setting the scene Teng Liaw 2.20 2.40pm Current initiatives & governance models Established data collection perspective Helena Britt Medicare Local/PHN perspective Chris Pearce Local health neighbourhood perspective Teng Liaw Linked national collections now possible Dougie Boyle 2.45 3.05pm Small groups: levels custodianship and stewardship Small groups 3.05 3.45pm Bringing it together All groups Definitions of key terms Curation: Manage and promote the use of data from their point of creation, to ensure they are fit for contemporary purpose, and available for discovery and use Information Ecosystem: A network that is continuously sharing information, optimising decisions, communicating results and generating new insights for businesses Setting the scene Governance principles and scope Teng Liaw Data Quality Management: Define DQ standards, data collection strategies and assessment of collected data using DQ indicators Data Governance: specifies who holds the decision rights and accountability for an organisation s decisions about its data assets. Information Governance: Ensures necessary safeguards for, and appropriate use of, patient and personal information. Data custodian: person/group responsible for decisions on providing access to and determining uses of data on behalf of the organisation and stakeholders Data Stewardship: Attends to and takes the past into account to influence the future, stretching from data planning to sampling, from data archive to use and reuse. This includes the care of data and information infrastructure, and involves data definitions, data requirements and quality assurance as well as user feedback, redesign and data exchange. 1
Data Clinical Corporate Integrated governance Data quality management (DQM) Information governance (FFP) Quality: cost-effectiveness Safety: patients and clinicians Accountable Medical Homes/Orgns Strategy, resources, metrics Scope: big data challenges collect data that can become useful and actionable Interoperability of devices, applications and services integrate sensing, clinical and service design to ensure actionable data. linkages of health with social/environmental data extract meaningful knowledge: policy & practice efficient databases, tools and interfaces to allow access to appropriate data at point of care. OECD 2011 Some existing Australian programs to collect and use primary care data Program Objective Denominator Governance Tools Other AIHW Specific non-routinely collected practice level data (not from EHR) National sample Program level AIHW in-house tools. www.aihw.gov.au/ data/ BEACH Specifically collected encounter (patient) level data (not from EHR) Sample of 1000 GPs at a time from Program level national database In-house BEACH tools. www.sydney.edu.a u/medicine/fmrc/ab out/index.php Current initiatives & governance models Improvement Foundation Electronic Practice Based Research Network (epbrn) MAGNET (Note: POLAR is a portal to MAGNET) Medicine Insight Collection of practice level EHR data Collection and linkage of patient level EHR data from primary and secondary care datasets in Integrated Health Neighbourhood (IHN) Medicare Local based collection of practice level EHR data with links to other services Collection of patient level EHR data National sample IHN = hospital, ambulatory care, community health and GP services Medicare Local (ML) Note: MLs now superseded by Primary Health Networks (PHN) National sample (Target: 500 practices) PEN-CAT extraction www.improve.org. Program level tool / Canning au GRHANITE Program level. extraction & linkage www.cphce.unsw. Local Health District tool. edu.au/research- (LHD) and UNSW ethics SQL / XML. streams/primary- committees. SAS / SPSS. health-care- informatics Planned: ML/LHD joint governance. Semantic Web tools. Data governance at Was using PENwww.med.monash. program level. CAT extraction tool, edu.au/generalpractice/magnet/ but now using Multiple ethics approvals. GRHANITE Data governance at www.nps.org.au/a program level. GRHANITE bout-us/what-wedo/medicineinsight extraction tool RACGP ethics. Governance: The action, manner, or power of governing :principles of good governance. (www.thefreedictionary.com/governance). National data collection perspective Helena Britt Includes: The ethical, legal and social issues around any data big or small And pertains to its collection, management, and use. BEACH: cross sectional, encounter-based, not longitudinal patient based, continuous, paper based. Running continuously since April 1998, now in Year 18. Began as University of Sydney-AIHW collaboration (1998-2010) ran under the AIHW Act In 2010 agreement dismantled, University of Sydney continues alone. and adopts much the same rules. 2
AIHW Act approved by both AIHW and University Ethics committees, data undiscoverable (cannot be called to court, does not exist) all staff sign confidentiality agreements: breach can result in jail and fine. limitations are set on release of line data and on reporting, so cannot possibly allow identification of individual GP, practice or patient. Since 2010 GPs: OPT IN system, when approached through random sampling Patients OPT IN: supplied Patient information sheet which informs, custodian, purpose, uses, funders, with contact details for ethical concerns to Ethics Committee University of Sydney Ethics: Sydney University Human Ethics Committee, five weekly applications of sub-studied not covered by the broad approval, and annual reports to Committee. Ask ourselves: Yes, its approved but is it ETHICAL in the broadest sense? BEACH Advisory Board: includes RACGP, AMA, ACRRM, CCHF, and rep from each of the organisations supporting BEACH (3 meetings per year). Still have confidentiality agreements Data management Select only those data elements that we need to collect- (unethical to collect data for its own sake), Safe data transfer with de-identificaiton of both clinician/practice and patient. Different PEOPLE can access identifying information (e.g. who is the GP or practice), from those who have access to the clinical data, in which identify practice/gp/ patinets only by project identified by project ID Locked secure data storage, encryption at data transfer. Third party patient matching when needed (e.g, CHeReL) Quality control checks: pregnant males, prostate problems in females; children on HRT etc. Validity checks: what are you trying to represent? Do these data represent what you SAY they are representing? Reliability checks against other data sources: e.g. prevalence estimates of diagnosed selected diseases in the database x age-sex, compared with national data sources. Keep specific record of cleaning methods applied for provision to the end user. Make sure you cleaning rules are uniformly applied in all cases Making the data meaningful and useable for an end-user If data files: stripped of D0B, addresses, etc. Care with how many geographic variables you provide (e.g. GP practice PHN, SEIFA, LHD, size could well identify a practice in some less populous areas of Australia. Provide them with the files with a standardised structure: Data dictionary with formal definitions of each data field, and copies of full code sets used, using international standards whenever possible Written detailed methods of every step of data collection, data manipulation and data management processes; full description of limitations. Note: we gave up line data provision: on phone 8 hours per day The public good AND to what extent are you responsible for correct use and interpretation of the data you provide to others? (examples) Medicare Local / PHN perspective Danielle Mazza on behalf of Chris Pearce POLAR: Population Level Analysis and Reporting Population Level Aggregation & Reporting Service run by MEGPN (and presumably the PHN in the future) delivering information to practices. 50 practices and expanding, 1 million patients Extensive cover of Eastern Melbourne Covered by Practice agreements 18 3
MAGNET Research collaboration between MEGPN and Monash National advisory group Jointly managed Ethics approved Multiple projects in place 20 Data Heirarchy/Capability Support Clinical Interventions Clinical Governance Population Based Decision Support Policy and Strategy Research Administration Local health neighbourhood perspective Teng Liaw 21 GPs + hospital + CH => Health Neighbourhood Opt out consent Secure Data Repository Doubly encrypted Decrypted - hashed identifiers. Records linked and processed A Network of Health Neighbourhoods Feedback to services to improve DQ 4
epbrn governance UNSW and SWSLHD HREC approved and monitor epbrn: geographic scope Steering Committee includes ML/PHN, LHD, Clinician, Consumer Participating GPs sign MoU: roles & responsibilities Standard operating procedures: access, privacy, confidentiality, security, risk management, incident management, etc DQ Assessment framework & protocols Top-down & Bottom-up model? Health neighbourhood = unit of governance PHN/LHD = network of neighbourhoods State (e.g. BHI) = network of PHN/LHD? National (e.g. AIHW) = network of PHN/LHD? integrated governance Scope: OECD challenges Some of the challenges to overcome National data collections now possible Dougie Boyle National data means working with over 7,000 private businesses (practices) There are many GP system vendors, database schemas can changing regularly, very limited standardisation, multiple terminologies Real privacy concerns and limited expertise to deal with such concerns at the practice level Minimal technical experience in the community / practices Aggregate v s de-identified individual-level data extractions What has changed that makes this workshop important? Impact of different levels of infrastructure distribution and governance New tools mean linked data and national data collection are a technical feasibility including implementing consent mechanisms Data strategies for PHN s are in-flux nobody wants 31 separate organisations going their own way Increasing scales of data collection raises appropriate concern about how such data can be governed and the interests of stakeholders represented There is limited experience in Quality Assurance and Terminology standardisation with such data how do we harness this? The GOOD One-stop shop National QA Terminology Standards Cheaper Flexibility Stakeholder Control NATIONAL LEVEL STATE LEVEL PHN LEVEL The BAD Big Brother Bureaucratic Legislative barriers Costly Poor QA No Standards The UGLY: Where s the middle ground? 5
Small groups 2.45 3.05pm Facilitators Levesque, Furler, Boyle, Britt, Liaw The Question for Today How do we set the balance in data management and governance if we are to have any hope of having a national, quality assured and standardised primary care resource? 1. Where should data custodianship lie? Can we have a single store of data or do we need the data managed in a more distributed model? (can you standardise and QA distributed data effectively?) 2. Decisions about who governs data and what can be done with it can be independent of where data sits. What would a workable governance model look like that may be acceptable to stakeholders (from patients federal government)? Bringing it together 6