Accident Analysis and Prevention 43 (2011) 421–428 Contents lists available at ScienceDirect Accident Analysis and Prevention journal homepage: www.elsevier.com/locate/aap Temporal variations in road traffic fatalities in South Africa Anesh Sukhai a,b , Andrew P. Jones a,∗ , Barnaby S. Love a , Robin Haynes a a School of Environmental Sciences, University of East Anglia, Norwich, Norfolk NR4 7TJ, United Kingdom b Crime, Violence and Injury Lead Programme, Medical Research Council, P.O Box 70380, Overport, Durban 4067, South Africa article info Article history: Received 8 June 2009 Received in revised form 23 September 2010 Accepted 27 September 2010 Keywords: South Africa Road traffic fatalities NARX models Multilevel Spatial Temporal abstract The annual road traffic fatality (RTF) burden of 43 deaths per 100 000 inhabitants in South Africa (SA) is disproportionately high in comparison to the world average of 22 per 100 000 population. Recent research revealed strong geographical variations across district councils in the country, as well as a substantial peak in mortality occurring during December. In this study, the factors that explain temporal variations in RTFs in SA are examined. Using weekly data from the period 2002–2006 for the country’s nine provinces, non-linear auto-regression exogenous (NARX) regression models were fitted to explain variations in RTFs and to assess the degree to which the variations between the provinces were associated with the temporal variations in risk factors. Results suggest that a proportion of the variations in weekly RTFs could be explained by factors other than the size of the province population, with both temporal and between-province residual variance remaining after accounting for the modelled risks. Policies directed at reducing the effects of the modifiable risks identified in our study will be important in reducing RTFs in SA. © 2010 Elsevier Ltd. All rights reserved. 1. Introduction The annual road traffic fatality (RTF) burden of 43 deaths per 100 000 population in South Africa (SA) is disproportionately high in comparison to the world average of 22 per 100 000 population (Bradshaw et al., 2003; Murray et al., 2001). A general increasing trend in fatalities has also been reported, with deaths in 2006 being 42% higher than those recorded in 2001 (NDoT, 2007). The large and growing burden of RTFs in low and middle income countries like SA is a particular public health priority; while RTFs are forecast to decrease by about 30% in high income countries over the next 15–20 years, RTFs in low and middle income countries are expected to increase on average by around 80% if current policies and practices continue without novel interventions (Peden et al., 2004). Recent research examining the epidemiology of RTFs in SA using several exposure-based indicators of risk showed strong geo- graphical variations across district councils (Sukhai et al., 2009). Additionally, marked variation was shown throughout the year, with a substantial peak in mortality occurring during December. The reasons for this have not been quantified. Nevertheless the excess of RTFs in December is similar to that shown in recent years for data in Australia (Department of Infrastructure, Transport, Regional Development and Local Government, 2008) whilst fatal crashes in the USA have generally peaked in July (NHTSA, 2008). ∗ Corresponding author. Tel.: +44 1603 593127; fax: +44 1603 591327. E-mail address: a.p.jones@uea.ac.uk (A.P. Jones). Research attempting to explain temporal variations in RTFs is commonly undertaken to understand the long-term effects of inter- ventions such as random breath testing for alcohol (Dunbar et al., 1987), seat belt usage (Houston and Richardson, 2002; Majumdar et al., 2004) and speed limitation (Johansson, 1996), or changes in traffic-related risks such as the role of alcohol or travel-related exposure (Fridstrøm et al., 1995; Ramstedt, 2008). Where temporal variations in RTFs within a year have been studied, the focus has generally been on seasonal influences. Much of this work has been from the northern hemisphere, especially Scandinavian countries (Johansson, 1996; Radun and Radun, 2006) and findings gener- ally point to an excess of traffic fatalities in summer. The summer excess may be associated with higher traffic volumes, or the effect of risk compensation where drivers reduce their speed in poorer winter driving conditions to offset the weather-related hazards (Fridstrøm et al., 1995). Driver fatigue as a result of the heat has also been suggested as a possible contributor to excess summer deaths (Johansson, 1996; Radun and Radun, 2006). The suggestion of seasonal effects in many previous studies is often based on proxies of unmeasured risks and the findings are of limited use in the design of interventions. Furthermore, little is known as to whether the risks identified in high income countries are the same as those in low to middle income countries such as SA. In this research, undertaken to provide new evidence on these issues, weekly data covering the 9 provinces of SA for the period 2002–2006 are used to fit non-linear auto-regression models. The models identify the predictors of temporal variations in RTFs in the country, and examine the degree to which spatial disparities in 0001-4575/$ – see front matter © 2010 Elsevier Ltd. All rights reserved. doi:10.1016/j.aap.2010.09.012