MEDICINEINSIGHT: BIG DATA IN PRIMARY HEALTH CARE Rachel Hayhurst Product Portfolio Manager, Health Informatics NPS MedicineWise
WHAT IS MEDICINEINSIGHT? Established: Federal budget 2011-12 - Post-marketing surveillance and support general practice Large scale general practice data - Longitudinal - Collected from clinical information systems Over 560 practices Whole of practice (all GPs) Patient level data de-identified at source Approximately 3.5 million active patients Granular data = richer insights
EARLY DAYS International models Engaging general practitioners target 7% or 500 practices Data extraction Data quality, cleaning and analysis Practice reports and model for practice visits
ETHOS For public good To provide value to government To provide value to participating practices To be respectful and trusted data custodians To integrate with other QI offerings for maximum impact To fill a gap and not duplicate
DATA GOVERNANCE Guiding principle of data use public good Data governance framework - NPS MedicineWise data custodian - Data security and privacy - Defines data lifecycle: access controls and disposal - Conduct risk assessment / audit and review of processes Comply with legal and ethical frameworks Data governance committee - Independent, external - Mechanisms to ensure correct and proper access to data
CONSENT PROCESS Recruitment of whole practice Explicit consent from practice owners to extract whole of practice data Explicit consent from individual GPs to collect personal information and for their name to appear in customised reports Implied consent from practice patients (with practice information and opt-out process)
HOW DOES IT WORK?
MEDICINEINSIGHT DATA Practice Encounter Provider Patient Diagnosis Prescription Investigation Observations Management activities Patient risk factors Allergies / drug reactions Name; software; extract date; location; type of practice Reason for encounter (text and codes); duration; date Consent; type (doctor/nurse/other); start date; (age; sex; active; yrs in practice for consenting doctors) Demographics, eg year of birth; sex; indigenous status; location; pension; year of death; active status, start date Diagnosis/history (text and codes); onset date Medicine name, reason for prescription (text and codes); product code; strength; frequency; dose; instructions; authority ; private; date; prescription history; etc Pathology and imaging: tests ordered; name of test; date ordered; pathology results; tests name, date of result Observations type eg temperature, pulse rate, height, weight; date of observation Health assessments; management plans; immunisations; PAP smears; referrals; recalls; medication review Eg BP; smoking; alcohol status; BMI; waist circumference Allergy type; reason eg bee sting
PRACTICE REPRESENTATIVENESS State Practice Size MedicineInsight (%) AMPCo (%) NSW 29 38 VIC 23 23 SA 4 7 QLD 20 19 TAS 8 2 WA 12 8 NT 2 1 ACT 2 1 1-2 GPs 18 46 3-5 GPs 38 25 6-8 GPs 24 14 >8 GPs 20 14
PATIENT REPRESENTATIVENESS MedicineInsight (%) National(%) Demographics Female Gender 54 50 1 Aboriginal and/or Torres Strait Islander 1.9 2.5 1 Pension 25 23 2 DVA 0.8 1.0 3 1. ABS National Stat 2013 2. FACHSIA 2011-12 3. DVA statistics 2014
CURRENT REGULAR ACTIVITIES Quarterly drug utilisation reports to DoH Regular quality improvement activities for general practice
INFORMING NATIONAL MEDICINE POLICY Assess uptake of new drugs Evaluate compliance with guidelines/pbs requirements Understand how medicines are used DUSC reports Diabetes Stroke COPD Biological medicines Depression Testosterone Antibiotics
AURA 2016: 1 ST AUSTRALIAN REPORT ON ANTIMICROBIAL USE AND RESISTANCE IN HUMAN HEALTH Patients prescribed systemic antibiotics in the community Measure Category Percentage of patients prescribed systemic antibiotics State New South Wales 33.8 Queensland 30.1 Tasmania 30.4 Victoria 29.0 Australian Capital Territory, 26.3 Northern Territory and Western Australia Rurality Major cities 31.1 Inner regional 28.2 Outer regional, remote and very 29.3 Socioeconomic status (SEIFA quintile) remote 1 2 (most disadvantaged) 31.3 3 4 28.7 5 6 30.5 7 8 30.7 9 10 (most advantaged) 30.5
IDENTIFYING PATIENT COHORTS We use complex algorithms based on diagnosis, reason for prescription, reason for visit, prescription history and tests to identify relevant patient cohorts
MONITOR PRACTICE IMPROVEMENT Able to monitor key risk factor and clinical indicators for individual practices and all practices or practices in location groupings (eg rurality, state, PHN)
IMPACT Percentage of practices achieving a > 3% improvement and > 5% improvement for clinical indicators 6 months after the MedicineInsight visit
IMPACT Percentage of practices achieving a > 3% improvement and > 5% improvement for monitoring and data quality indicators 3 months after the MedicineInsight visit
ACCEPTABILITY OF ACTIVITY Strongly agree/agree Report relevant Report useful Report easy to understand Appropriate meeting 0% 20% 40% 60% 80% 100%
VISUALISATION PORTAL Extend capability of MedicineInsight focus areas of choice Secure web portal Underlying data warehouse: Enable richer insights Minimise duplicates Built with end user in mind Ease of navigation Brings MedicineInsight reports to life Graphical, interactive, drill down functionality Demo contains dummy data, no real practice or PHN names
Hover: display detail View trend over time Compare Drill down Custom age range
Data quality
RECENT AND CURRENT PROJECTS Chronic kidney disease in Type 2 diabetes Renal toxicity (risk management plan compliance) Supporting primary health care policy development (10% sample) WA PHN needs assessment Vaccination coverage in vulnerable subgroups Vitamin D and chronic musculoskeletal conditions Preventing obesity in childhood and adolescence Lung cancer management in primary care Primary and secondary prevention of CVD in adults in Australian general practices
WA participating practices 69 agreed 46 in the database 29 in Perth 1.5% indigenous patients recorded
What we know about WA practices* Top 5 conditions for visit % URTI 2.9 Hypertension 2.8 Depression/ Anxiety 1.7 UTI 0.8 Back pain 0.7 * Based on available MedicineInsight data 1 July 2015; not include screening test, prescriptions
What we know about WA practices* All WA MedicineInsight practices Conditions # patients % % Type 2 Diabetes 12 694 4.87 4.96 Heart failure 1 972 0.76 0.95 Hypertension 40 882 15.68 17.26 Stroke 3 134 1.20 1.35 Transient ischaemic attacks 1 462 0.56 0.61 * Based on available MedicineInsight data 1 July 2015
What we know about WA practices Patients (over 45 years) at high risk of stroke risk factor recording WA practices 12% not recorded
% of active patients PHN DATA EXAMPLE 28 practices (5.3% of all emi practices) 134,499 active patients (3 visits last 2 years) 2.2% Aboriginal and/or Torres Strait Islander (all emi practices 1.9%) 14 12 10 8 6 4 2 0 <5 5-14 15-29 30-44 45-59 60-74 75+ <5 5-14 15-29 30-44 45-59 60-74 75+ Male Age Group Female * Based on available MedicineInsight data 1 Sept 2015
PHN DATA EXAMPLE* PHN All emi Conditions # % % Atrial Fibrillation 3088 2.3 1.8 Cardiovascular Disease 9306 6.9 5.1 T2 Diabetes 8145 6.0 5.0 Heart Failure 1996 1.5 1.0 Hypertension 30259 22.5 17.3 Stroke 2514 1.9 1.4 Transient Ischaemic attacks 1220 0.9 0.6 * Based on available MedicineInsight data 1 Sept 2015
PHN DATA EXAMPLE Absolute CVD risk levels of patients 45+ Heart Foundation algorithms calculate the number of patients at risk of stroke and other CVD events Provide patient list to assist practice to: - Identify those at high risk for review - Identify those that need additional assessment to asses risk Absolute CVD risk levels for patients
DEMONSTRATION PROJECTS MedicineInsight data are a rich source General Practice data is complex and requires an understanding of the source system and data entry Use of multiple fields is required for some insights ie reason for prescription plus past history Linkage to this data would increase insights MedicineInsight data can be used for risk management plan evaluation, thus contributing to improved patient safety
FUTURE OPPORTUNITIES Support for chronic disease management Linkage with other data sets: hospital, registries, mortality, PBS/MBS Case finding e.g. rare diseases Sub-optimal management of vulnerable people Support primary health care policy, incentives and practice models Integrated health care offerings Support public health initiatives - Cancer screening - Vaccination
THANK YOU