Injury surveillance in Australia: aims and issues James Harrison Research Centre for Injury Studies Adelaide, South Australia September 2006
Acknowledgements Colleagues at NISU/Research Centre for Injury Studies, Flinders University Other colleagues in Australia, especially Andrew Hayen and Soufianne Boufous at the Injury Risk Management Research Centre, University of NSW
Purpose National injury surveillance especially for primary prevention of injury in the general community also interest in outcomes, costs, injury & other complications of care How much (serious) injury occurs in Australia? What are the injuries? Diagnoses, body parts affected How does injury occur? External causes, etc How is it distributed? By person age, sex, Indigenous status, etc. By place State/territory, remoteness, etc. Over time trends What are its consequences? Survival, hospital utilisation, rehabilitation, quality of life, economic costs, etc
Current sources Cases Deaths (ABS/registrations; NCIS/coroner notified) Universal; unit record (case); ICD-10 diagnosis & external cause; other data; text Hospital admissions (public & private; via States under agreed NMDS) Universal; unit record (episode); ICD-10-AM diagnosis & external cause; other data Ambulatory health service data Patchy; National ED collection in development for service utilisation (Dx? Ext??); regional ED injury surveillance (ICD-based NDSIS); GP/Primary Care Physician rolling sample (ICPC). Self-reported occurrences Population surveys; nb National Health Survey (currently each 3 yrs); ICD-based codes Special-purpose collections Eg. registers (SCI; trauma); ambulance service data; OHS inspectorates; workers’ compensation; compulsory third party road traffic insurance; air safety inspectors Denominators/exposure Population estimates (census; annual estimated resident population) Various estimates relevant to aspects of injury surveillance: eg. workforce (persons; hours); road transport (registered vehicles; distance); sport (participants); etc.
Can do (more or less) Injury mortality Rates, trends, description Issues: (i) mismatch b/w sources; (ii) late deaths; (iii) uncertain reliability & limited detail Responses: (i) & (iii) link ABS & NCIS deaths data; (ii) link death & hospital data; (iii) link deaths to other sources (e.g. OHS fatality reports) Injury hospitalisation Rates, trends, description Issues: (i) cases vs episodes; (ii) uncertain reliability & limited case detail Responses: (i) internal linkage (special study (WA) -> routine); (ii) quality assurance / enhancement study High threat-to-life injury Rates, trends, description Based on deaths + non-fatal hospital cases with ICISS > threshold Issues: (i) linking hospital and deaths data; (ii) ICISS technical issues (method for deriving weights; method for applying weights; comparability of weights over time and between settings) Responses: (i) Special study (WA) -> broader application; (ii) further use & development of ICISS
Can’t do (except as special studies) Surveillance of serious injury i.e. injury presenting threat to well-being, quality of life Issues: Definition of “serious” (i.e. Which cases to include? Specify in terms of diagnosis? Outcome? How to find those not in the deaths or admitted patient collections? Outcome of hospitalised injury Current data tell us vital status & type of destination at the end of an episode in hospital Issues: How useful is information currently in hospital records for assessing QoL, cost, or other dimensions of outcome? In aggregate? At case level? What additional data would enable importantly better outcome assessment? Is it feasible to obtain this information?
Special study: counting cases The problem Records in national hospital data collection refer to episodes (‘separations’), not cases or persons. We want to analyse in terms of cases & persons. Initial solution Use ‘mode of admission’ and/or ‘mode of separation’ to omit classes of records likely to be counted more than once / case. This approach might deal with transfers & type change within a hospital but not with readmissions. Preliminary study Used person-linked data for one state (Western Australia; c 10% of national total) Findings: mode of admission method is better than mode of separation method, but not very reliable. Work in progress Seeking collaboration of all states and territories in a project to: document hospital data person-specific internal record linkage activities done/in progress compare methods & variables used and assess likely effect of differences identify technically feasible opportunities for linkage where not yet done specify a preferred method (technically feasible in all states), seek agreement and apply it. Issues No national person ID and great sensitivity concerning use of proxies for one Status Proposal with jurisdictions for consideration Positive responses so far from three jurisdictions covering >50% population. No refusals so far.
Special study: external cause coding of hospital data The problem Australian hospital admission records for injury include external cause codes, but there is little published evidence on their quality. Work in progress ARC-funded project “Developing and enhancing the quality of national injury- related hospital morbidity data” ( ). Lead investigator: Kirsten McKenzie, Queensland University of Technology / NCCH. Other Investigators: S Walker & G Waller (NCCH), J Harrison & G Henley (Flinders), R McClure (Griffith). Four health departments are partners. Aims (i) better understanding of data; (ii) guide to QA; (iii) guide to ICD-10-AM Stages 1.Analysis of unit record administrative data Apparent completeness and specificity of external cause coding. (Done) 2.Surveys of clinical coders and end-users of data Knowledge, attitudes and perceptions concerning the external cause classification in ICD-10-AM, the quality of its use, barriers to use, potential for improvement. (In field) 3.Examination & re-coding of a sample of records Probability sample of records in four states. (Planning)