Big linked data projects Louisa Jorm Centre for Health Research, University of Western Sydney Centre for Big Data Research in Health, University of NSW (from 10 Nov 2014)
Research focus Policy partnerships Linked whole-of-population administrative data Linked administrative and cohort study data Evaluating natural experiments Multilevel modelling
Researchers and key collaborators Louisa Jorm Alys Havard Deborah Randall Sanja Lujic Michael Falster Danielle Tran Amy Gibson Bich Tran Holger Möller Alastair Leyland Emily Banks Sandra Eades David Preen Christine Roberts Sally Redman Bob Elliott Fiona Blyth Judy Simpson Rebecca Ivers Jim Warren John Lynch Federico Girosi Kathleen Falster
The APHID Study Assessing Preventable Hospitalisation InDicators Funding: NHMRC Partnership Project Policy partners: Australian Commission for Safety and Quality in Health Care, Agency for Clinical Innovation, Bureau of Health Information Research questions: Are "potentially preventable hospitalisations" a good measure of health system performance? How can the performance indicator be improved? Data: 45 and Up Study (n=267,000), Medicare (MBS, PBS), hospital admissions, emergency department presentations, deaths Methods: Multilevel Cox and Poisson regression
The APHID Study Data linkage Assessing Preventable Hospitalisation InDicators 45 & Up Study NSW Admitted Patient Data Collection MBS PBS Emergency Department Data Collection Prospective cohort of 267,091 men and women aged over 45 in NSW. Completed 2006-2008 Questionnaire data Demographics Health status Risk factors Census of all hospital separations in NSW public and private hospitals and day procedure centres. Linked data, 2000-2010 N=1,206,742 records Claims for subsidised medical and diagnostic services in Australia Linked data, 2004-2011 N=46,203,507 records Claims for subsidised pharmaceuticals In Australia Linked data, 2004-2011 N= 35,453,776 records Presentations to 80 EDs (75% 0f NSW presentations) Linked data, 2006-2011 N= 347,602 records + Fact of death to 2012
The APHID Study Assessing Preventable Hospitalisation InDicators 45 & Up Study NSW Admitted Patient Data Collection Medicare Benefits Schedule Pharmaceutical Benefits Scheme Detailed data for 267,153 people...... WHO they are... HOW they have interacted with the primary health system... WHETHER they were admitted to hospital
More geographic variation explained by the supply of GP services in area For which conditions does GP supply The APHID Study have the greatest impact? Assessing Preventable Hospitalisation InDicators GP supply explains the greatest amount of geographic variation in admission rates for influenza and asthma More geographic variation in admission rates
IHOPE: Indigenous Health Outcomes Patient Evaluation Policy partners: Aboriginal Health and Medical Research Council, NSW Ministry of Health Research questions: How do Aboriginal status, socioeconomic status and rurality interact to drive health disparities? In cardiovascular disease, injury, cataract surgery, preventable hospitalisation, otitis media, diabetes mellitus, renal disease Data: Hospital admissions (5.6M people), deaths Methods: Multilevel Cox, Poisson, logistic and negative binomial regression
IHOPE: data linkage Total persons 1 2 3 Hospital admissions (NSW Admitted Patient Data Collection) Fact of death (NSW RBDM) Jul00 to Dec09 Cause of death (ABS) Jul00 to Dec07 4 5 6 7 8. Jul00 to Dec08 18 638 151 separations 5 580 151 persons 433 453 338 826. 5,628,960
IRRs (Incidence rate ratios) IHOPE: Disparities in transport injuries (random intercept model) Aboriginal rate higher Non-Aboriginal rate higher Aboriginal rate higher Non-Aboriginal rate higher
MUMS: Maternal Use of Medications and Safety Funding: NHMRC Project Grant Policy partners: Department of Health Research questions: What are the maternal and neonatal health outcomes of medications used during pregnancy? Specifically: smoking cessation medications, antihypertensives, medications for diabetes, thrombosis, rheumatoid arthritis Data: Perinatal data (~800,000 pregnancies), Medicare (PBS), hospital admissions, emergency department presentations, birth defects, deaths Methods: Interrupted time series analysis (impact of policy changes), logistic and negative binomial regression
Mapping the outcomes of calls to healthdirect Australia Funding: Healthdirect Australia Policy partners: Healthdirect Australia Research questions: To what extent is healthdirect Australia telephone triage advice being followed? What factors influence patient advice-taking and outcomes? Data: healthdirect Australia call data (1.3M records), hospital admissions, emergency department presentations, deaths, 45 and Up Study, Medicare (MBS) Methods: Logistic and Cox regression
Seeding Success Funding: NHMRC Project Grant Policy partners: NSW Kids and Families, NSW Department of Community Services, Aboriginal Health and Medical Research Council Research questions: What are the social, perinatal and early childhood health factors that promote positive early childhood development in Aboriginal children? Do current prevention and early intervention programs work? Data: Australian Early Development Index (~180,000 children), perinatal data, MBS, DOCS, hospital admissions, emergency department presentations, deaths, income assistance Methods: Multilevel linear and logistic regression, propensity matching
AREA-LEVEL CONTEXTUAL DATA (e.g. 2011 Census data on community characteristics) 2009 AEDI COHORT PRENATAL BIRTH EARLY CHILDHOOD SCHOOL Perinatal Data (PDC) Medicare Benefits Schedule (MBS) APDC (mother) Birth registrations (RBDM) Hospital data (APDC) Emergency department data (EDDC) 2009 AEDI Congenital Conditions Register (RoCC) Community Services data (KiDS) Centrelink income assistance data (mothers/fathers) 2012 AEDI COHORT PRENATAL BIRTH EARLY CHILDHOOD SCHOOL Perinatal Data (PDC) Medicare Benefits Schedule (MBS) APDC (mother) Birth registrations (RBDM) Hospital data (APDC) Emergency department data (EDDC) 2012 AEDI Congenital Conditions Register (RoCC) Community Services data (KiDS) Centrelink income assistance data (mothers/fathers) AMIHS funded in 21 areas AMIHS expanded to 50 additional areas Brighter Futures program data collection commenced 2000 2001 2002 2003 2004 2005 2006 2007 2008 2009 2010 2011 2012
Thank you!