Background Motor Vehicle Crashes (MVCs) are a major cause of morbidity and mortality worldwide. This thesis explores the potential use of large databases of primary care medical records to investigate the epidemiology of MVCs in the United Kingdom and to supplement the data available from national statistics, which are believed to understate both the number of crashes, and the number of injuries which occur as a result. Methods Details of all individuals enrolled in The Health Improvement Network (THIN) database whose primary care records indicated involvement in a MVC were used to calculate a series of summary measures describing the burden and consequences of MVCs. These were compared with data available from police accident reports and from hospital admissions. Data from THIN were used to conduct a series of studies of the impact of health and healthcare-related factors on the risk of involvement in MVCs. Specifically: a case-control study of the impact of modifiable lifestyle factors on the risk of MVC; case-crossover and self-controlled case-series studies of the effect of exposure to prescribed medications on the risk of MVC; a case-control study investigating the impact of disordered sleep on the risk of MVC; a case-control study of the risk of involvement in MVC among individuals with diabetes relative to the general population; and; a cohort study assessing whether there is evidence to suggest that involvement in a MVC may indicate the presence of undiagnosed disease which may impair driving performance. Results The socio-demographic characteristics of individuals involved in MVCs recorded in THIN differ markedly from those recorded in police accident reports and hospital admissions data. There was no evidence of consistent trends in MVC incidence over time in the three data sources. Differences in data collection methodology and the severity and scope of crashes recorded may account for these variations. Evidence was found of an association between having a high Body Mass Index and involvement in MVCs, and between past (but not current) smoking and involvement in MVCs, however the recording of data on lifestyle-related exposures such as smoking and alcohol consumption in the age-groups most likely to be involved in MVCs was poor, complicating interpretation of these results. Current exposure to benzodiazepines and preparations containing opioid analgesics was found to increase the risk of involvement in MVCs, as was longer-term use of non-benzodiazepine hypnotics, selective serotonin reuptake inhibitors and antihistamines. No increased risk of MVC was observed with exposure to beta-blockers or tricyclic antidepressants. Individuals reporting insomnia or snoring to their primary care practitioner were found to be at increased risk of MVC, as were individuals with diagnosed sleep apnoea. This association was independent the use of sedative or antidepressant medications. Individuals with diabetes were not found to be at an increased risk of MVC compared with the general population, and there was no difference in risk between those receiving different forms of treatment. Involvement in a MVC was associated with an increased risk of being diagnosed with cardiac disease in the two years following the crash. Conclusions Current sources of data about MVCs in the UK use different data collection methodologies, none of which is likely to accurately describe the overall burden of MVCs in the population. Primary care data remain a useful resource for those wishing to study the epidemiology of MVCs, but care must be taken to ensure that the uses to which the data are put are appropriate. Studies investigating lifestyle-related exposures are unlikely to produce reliable results as primary care recording of such factors is poor in the age-groups most likely to be involved in MVCs. Primary care data are more useful when studying the time course of pharmacological effects, or the effects of diagnosed illness, and can successfully detect previously observed associations. Primary care data is currently of little use in the study of injuries associated with involvement in MVCs as it is rare for both an injury and its proximate cause to be recorded. The investigation of methods by which this problem might be resolved is an important avenue for future research.