Characteristics of road traffic accidents in Kenya
The objectives of this study were: to study Road Traffic Accidents (RTAs) in Kenya and to determine where possible their fundamental characteristics and causal factors related to their occurrence, to develop predictive models for Kenya at the national (macro) level to be used for the monitoring of RTAs and the performance of road safety improvement programnles and lastly to develop predictive models through some selected Kenyan roads at the road level (micro) to assist in the proper understanding of the behaviour of RTAs in relation to road design elements. In order to develop these predictive models, various mathematicalmodels were used. These were: growth curve models, namely the logistic curve model and the logarithmic model; polynomial functions and finite differences techniques; harmonic analysis, generalised linear modelling and statistical methods for testing the fitness of the models developed. Macro level data for Kenya were collected from the Kenya Police records and from the Statistical Abstracts of the Central Bureau of Statistics of Kenya. Micro level data were collected through traffic volume counts, study of road geometry and pavement defects and by specially'coded forms used for extracting data from the Traffic Police RTA records. The data were collected from the two carriageways of the Nairobi-Thika dual carriageway and the Kiganjo-Nanyuki single carriageway. road. The police forms were obtained from the police stations responsible for the roads studied. The data were analysed to provide characteristic patterns and evidence of the mathematical \techniques to be used in model development. Computer facilities were used whenever necessary. The major findings at the macro level were: the logistic model is well suited in predicting the growth of RTAs and related phenomena with time, the logarithmic trend curve is well suited in predicting the growth in the distribution of RTA responsibility and involvement whilst the polynomial function is suited in predicting the trend of RTAs in relation to motorization. The major findings at the micro level were: the polynomial functions are suited in predicting the effects of road factors on RTA rates, the logistic curve is well suited in predicting the growth of RTAsin relation to vehicle flow, harmonic functions are suitable for predicting variations in RTAs and vehicle flow with time of day and the generalised linear model is beneficial when trying to study the effects of traffic and geometrical design elements on RTAs on an interactive basis. It is recommended that there be continuous data collection in the form of an accident data base. Such data will then be used on a continuous basis for model calibration and monitoring of road safety improvement measures.