Data linkage for paediatric trauma and health services research Transport and Road Safety (TARS) Research, School of Aviation Rebecca Mitchell 1,2 1 University of New South Wales, Australia 2 Neuroscience Research Australia Paediatric Injury Prevention and Management Research Forum 1 August 2014
Be committed: approvals and data access can take years
What is data linkage? Linking of two or more data collections using common personal identifiers (eg. name, address, age, gender, MRN) Centre for Health Record Linkage (CHeReL) Creates & maintains record linkages for NSW and ACT Ongoing linkage of personal identifiers from core health databases All projects require data custodian and PHSREC approval CHeReL links personal identifiers & has no access to health information, creates linkage key Data custodian(s) provide data extracts to researcher with the linkage key Researcher has no access to personal identifiers, only health information, and links data extracts using linkage key
CHeReL Master Linkage Key (July 2014)
Eg1: Acute costs for paediatric major trauma in NSW, 2008-09 Aim: Determine if existing funding model provides adequate reimbursement for trauma treatment costs Method: Children aged 15 years Trauma data: from each trauma centre s trauma registry patient demographics; nature and location of incident; main injuries sustained; treatment; injury severity (ISS) Financial data: from financial units at each hospital General ledger costs (e.g. theatre) & indirect costs (e.g. cleaning) Trauma and financial data linked MRN, DOB, admission date Statewide peer-group average costs: from NSW Ministry of Health
Acute costs for paediatric major trauma in NSW, 2008-09 Results: ISS, ICU admission and LOS all predictive of high trauma costs Item Actual cost and LOS NSW state wide AR-DRG average Difference Cost ($) $20,161,777 $18,780,861 +$1,380,916 LOS (days) 9,795 11,417-1,622 Average LOS (days) 2.8 3.3-0.5 Outcomes: Paediatric trauma cost $1.4 million above NSW state wide peer-group average costs Weighting system for trauma costs based on body regions injured/ injury severity
Eg2: Risk factors associated with injury severity for paediatric road trauma in NSW, 2001-2011 Aim: Identify crash and injury risk factors associated with injury severity for non-fatal paediatric road trauma Method: Children aged 16 years Admitted patient data collection: patient demographics; diagnoses, external cause(s); clinical procedures; hospital separation type Police-reports: Crash circumstances; type of vehicle; causal factors; restraint use; helmet use; no information on injury severity Patient and police records linked by CHeReL Calculated injury severity ICISS (minor/ moderate/serious injury)
Risk factors associated with injury severity for paediatric car occupants in NSW, 2001-2011 Results: 2,412 car occupants injured and hospitalised 24.0% minor injuries; 50.3% moderate injuries; 25.7% serious injuries Odds Ratios: minor 1 vs serious injury for key characteristics Lower odds of serious injury Higher odds of serious injury Licence type Learner/provisional Unauthorised Environmental Single vehicle crash Metropolitan area Intersection Curve Daytime Crash mode Vehicle-right angle Vehicle-rear end Vehicle-other angle Speed limit 100-110km/h Restraint Restraint worn 1 Minor injury was reference group. 0 0.5 1 1.5 2 2.5 3 3.5 4
Eg 3: Incidence and burden of childhood injury in Australia, 2002-03 to 2011-12 Aim: examine hospitalised childhood injury & related followup care in Australia by injury severity & factors influencing survival Method: Retrospective epidemiological study Children 16 years National Death Index linked to National Hospital Morbidity Database Required involvement of: 8 Australian states and territories each state has different data linkage application, privacy etc forms (n=50) & application process 9 data custodians 10 HRECs 2 data linkage centres (AIHW and CHeReL)
Key challenges of national data linkage research (1) The clerical admin: Under present conditions it is likely that many studies, although simple in concept, never come to fruition because of the clerical effort required or because of the inaccessibility of the data (Hobbs & McCall, 1970; p376). (2) Keeping investigators/ funders engaged (3) Timeframe & budget: Wait for data custodian & HREC approvals; data linkage; & the data
Key challenges of national data linkage research (4) Hospitalisation data availability across states eg: Data variable NSW SA QLD TAS VIC WA Aboriginal and Torres Strait Islander identifier(s) Involves approval from an appropriate HREC in each state. Age (single year) * * Country of birth # X Date of admission/ separation * * Diagnosis (ICD-10-AM) * Hospital identifier X # * * Postcode of usual residence # X Preferred language X * Procedure (ICD-10-AM) * Separation mode * SLA of residence # X Time of admission/ separation X X X X X Key: will provide this information; # will provide some information; * needs specific approval; X will not provide information.
What would make things easier? Sibthorpe et al (1995), while acknowledging that safeguards were essential, called for...significant streamlining of the approval process so that record linkage studies do not take years to implement (p255). In an ideal world: National application documentation for data linkage studies one size fits all Transparent application & approval process in each state Standard confidentiality agreement for use in all states to use linked health data One HREC to provide approval for national data linkage studies (follow the NHRMC single ethics review eg. clinical trials) Investigate potential of storing & making available linked data extracts from completed studies for future research
Potential of using data linkage for research Admin data; pre-existing time/cost vs prospective studies Retrospective studies; largely population-based Information across the injury continuum Identification of patient comorbidities look back periods Under-reported characteristics eg. Aboriginality Calculation of injury severity ICISS Hospital services & patient care pathways Novel study designs eg. case-cross over studies Vehicle telematics & hospital records Biomarkers & injury (eg Smith et al 2010) Data donors & MS (eg. Ford et al 2012)
Acknowledgments Trauma nurses and data managers in NSW and David Martens from ITIM. Data custodians & data linkage centres Day of Difference Foundation St George Honda Trauma & Critical Care Research Program References: Mitchell R. Cameron C. Bambach M. (2014) Data linkage for injury surveillance and research in Australia: Perils, pitfalls and potential. Australian and New Zealand Journal of Public Health 38 (3) 276-281. Mitchell R. Curtis K. Holland A. Balogh Z. Evans J. Wilson K. (2013) Acute costs and predictors of higher treatment costs for major paediatric trauma in New South Wales, Australia. Journal of Paediatrics and Child Health 49, 557-563. Mitchell R. Curtis K. Chong S. Holland A. Soundappan SVS. Wilson K. Cass D. (2013) Comparative analysis of trends in paediatric trauma outcomes in New South Wales, Australia. Injury 44, 97-103.