AN INVESTIGATION OF THE FACTORS ASSOCIATED WITH THE NON-USE OF A SEAT BELT THROUGH THE ANALYSIS OF LINKED WESTERN AUSTRALIAN CRASH, DEATH AND HOSPITALISATION DATA Peter Palamara 1 , Jennifer Oxley 2 , Jim Langford 1 , Claire Thompson 3 , Ann-Marie Chapman 4 1 C-MARC, Curtin University of Technology 2 MUARC, Monash University, 3 WA. Office of Road Safety, 4 Data Linkage Branch, Health Department of Western Australia email: p.palamara@curtin.edu.au, web: www.c-marc.curtin.edu.au INTRODUCTION Seat-belts are one of the most successful occupant protection countermeasures available to reduce the likelihood of death and non-fatal injury in the event of a crash 1 . In Australia, the prevalence of seat-belt use by adult vehicle occupants is estimated to exceed 90% 2 . Yet despite the effectiveness of seat belts and the comparatively high wearing rate in Australia, it is estimated that around 26% of occupants killed 3 and 20% of those seriously injured 4 are unbelted. Understanding the demographics and behavioural characteristics of those who fail to wear a seat-belt is central to the development of appropriate and effective countermeasures to increase seat belt use. To this end, one aim of the current research was to investigate the usefulness of linked crash, death and hospitalisation data to further understand the characteristics of adult vehicle occupants involved in a serious injury crash in Western Australia who fail to wear a seat-belt. METHODS Road crashes reported to Western Australia police between 2001 and 2006 were considered for analysis. The crash records were then restricted to light motor vehicle occupants aged 16+ years with a link to a death (within 30 days of the crash) or hospitalisation (within 60 days of the crash) record. Persons involved in a serious injury crash were selected for analysis because death and hospitalisation records provide additional information about occupants not found in police records (eg., Indigenous status, place of residence, country of birth) and confirm the age and gender of the occupant. The probabilistic linkage methods resulted in police crash reports being linked with records for n=679 deaths and n=7,439 hospitalisations for the period (see Table 1). Table 1: Identified HMDS, death and police crash records and final linked record numbers; Western Australia 2001-2006 Hospital Morbidity Data System and Death Records ECodes V40-V49 Police Recorded Crashes Crash involved occupants meeting relevant inclusion criteria Deaths N=582 Hospitalisations N=9,613 Deaths N=706 Hospitalisations N=11,049 Final linked records for crashes 2001-2006 Deaths n=679 96% of police recorded crash deaths Hospitalisations n=7,439 77.3% HMDS admissions Only linked records for deaths (n=597 deaths) and hospitalisations (n=5,657) where seat-belt use was recorded as ‘worn’ or ‘not worn’ were analysed using logistic regression to find the best predictors of non-use of a seat-belt from a range of occupant demographic and crash factors. The variables were entered to the modeling equation in a single step and manually adjusted (removed) until the ‘best fit’ with the data was obtained based on the Hosmer and Lemeshow chi-square statistic and associated p-value. RESULTS AND DISCUSSION The (abbreviated) results of the multivariate logistic regression modeling of the non-use of a seat-belt by all occupants seriously injured (death or hospitalisation) are presented in Table 2. After adjustment for both the location of the crash and the type of vehicle in which the occupant was injured, five occupant characteristics were found to be significantly associated with an occupant being unbelted. First, occupants of Indigenous background had six times the odds of non-Indigenous persons of being unbelted (OR=6.03; 95% CI: 4.58-7.94). Age was also found to be significantly associated with an increase in the odds of being unbelted, with occupants aged 26-40 years (OR=2.58; 95% CI: 1.98-3.35) and 16-25 years (OR=2.47; 95% CI: 1.92-3.16) having more than double the odds of occupants aged 41+ years of being unbelted. The odds of an injured passenger being unbelted were 2.5 times that of an injured driver (OR=2.47; 95% CI: 2.04-3.0), while male occupants (OR=2.33; 95% CI: 1.89-2.86) and occupants who resided in rural Western Australia (OR=1.64; 95% CI: 1.24-2.17) had 2.3 times and 1.6 times increased odds of being unbelted (compared with females and those who resided in metropolitan Perth). Table 2: Logistic regression modeling of non-use of a seat-belt by all vehicle occupants in a police recorded crash with a linked death or hospital admission record; Western Australia, 2001-2006 OR 95% CI Sig. Sex Female * Male 2.33 1.89-2.86 P<.001 Age 41+ years * 26-40 years 16-25 years 2.58 2.47 1.98-3.16 1.92-3.11 P<.001 P<.001 Residential Location Metropolitan Perth * Rural WA. 1.64 1.24-2.17 P<.001 Indigenous Status Non-Indigenous * Indigenous 6.03 4.58-7.94 P<.001 Occupant Type Driver * Passenger 2.47 2.04-3.0 P<.001 Hosmer & Lemeshow X 2 =7.54, df=8, p=0.479 *Reference group; OR adjusted for the location of the crash and the type of vehicle in which the occupant was injured Due to the strong association of Indigenous status with non-use of a seat belt, the data was subsequently stratified 10 -13 November 2009, Sydney, New South Wales 2009 Australasian Road Safety Research, Policing and Education Conference © 2009 Peter Palamara et al Non peer-reviewed extended abstract 519