Article Text

PDF
OP08 Using Linked Administrative Data and Multilevel Modelling to Identify Targets for Interventions to Tackle Health Disparities
  1. L R Jorm1,
  2. D A Randall1,
  3. M O Falster1,
  4. A H Leyland2
  1. 1Centre for Health Research, University of Western Sydney, Penrith, Australia
  2. 2Medical Research Council (MRC)/Chief Scientist Office (CSO) Social and Public Health Sciences Unit, Glasgow, UK

Abstract

Background Linked administrative data for whole populations, combined with multilevel modelling methods, offers a powerful approach to identifying targets for tackling health disparities. Two studies will be presented, examining disparities between Australian Aboriginal and non-Aboriginal people, in very different outcomes: (i) revascularisation and mortality after acute myocardial infarction (AMI); and (ii) rates of serious road traffic injury.

Methods Both studies used whole-of-population hospital inpatient data for the state of New South Wales, Australia, linked with mortality data. For Study (i), hazard ratios for revascularisation following AMI were estimated using multilevel Cox regression and ORs for 30- and 365-day post-admission mortality were estimated using multilevel logistic regression. For Study (ii), adjusted rate ratios were estimated using multilevel Poisson regression models.

Results Study (i) found that Aboriginal patients had a revascularisation rate 37% lower than non-Aboriginal patients of the same age, sex, year of admission and AMI type (adjusted hazard ratio [AHR] 0.63 [95% CI 0.57, 0.70]). This disparity was no longer significant after adjusting for hospital of admission, comorbidities, substance use and private health insurance (AHR 0.96 [0.87, 1.07]). After adjusting for age, sex, year and hospital, Aboriginal patients had a similar 30-day mortality risk to non-Aboriginal patients (AOR 1.07 [0.83, 1.37]) but a higher risk of dying within 365‐days (AOR 1.34 [1.10, 1.63]). Study (ii) found that, overall, Aboriginal people had higher rates of road transport injuries (incidence rate ratio [IRR] 1.18 [95% CI 1.09, 1.28]). However, there was no significant difference when geographic clustering was taken into account (1.00 [0.96-1.04]). This effect was further influenced by mode of transport for the injury, with Aboriginal people having higher rates of pedestrian (IRR 1.96 [1.75, 2.19]) and lower rates of motorcycle (IRR 0.64 [0.59, 0.70]) injuries in all almost all local areas, while there was no systematic pattern across geographic areas for small vehicle injuries (IRR 1.01 [0.94, 1.08]).

Conclusion Use of linked administrative data and multilevel modelling permitted quantification of the contributions of personal, geographic and health system factors to health disparities, and thereby of the potential contributions of primary prevention, primary care, and hospital care to tackling these. It is essential that geography is taken into account in studies of health disparities, especially in countries with significant differences in the distribution of disadvantaged populations, and in the provision of health services, between urban and rural areas.

Statistics from Altmetric.com

Request permissions

If you wish to reuse any or all of this article please use the link below which will take you to the Copyright Clearance Center’s RightsLink service. You will be able to get a quick price and instant permission to reuse the content in many different ways.