My impression of the transport industry is that safety issues are very much under-developed, and that there is little understanding of, or indeed interest in, the scientific knowledge that could be used for reducing accidents and costs substantially. Therefore, I have put together some advice, with some appropriate references (there is much more). The possible savings are dependent upon how bad the situation is to start with, and how many of the changes are implemented, but even for a very good company, there should be at least a 10 percent reduction in accidents to be had if the whole package is used.
As part of my work for Lisa Dorn, a review on accident factors which transport companies can influence has been written and can be downloaded.
af Wåhlberg, A. E. (2007). How managers can influence professional drivers' accident rate. Download pdf-file
Recruitment
Do you use some kind of specialized test to hire safe drivers, or, worse, do you pay some recruitment agency to do this job for you? Well, don't, because there is no test that is worth the money today (see for example the review of accident predictors by Lester, 1991). You might as well just employ middle-aged, experienced (Spratling, 1961; Blom, Pokorny & van Leeuwen, 1987) drivers without a criminal record (West, 1998), and especially without traffic violations on record (Peck, 1993). Those variables have a predictive power that is about the same as any existing test.
However, a multivariate statistical analysis of your existing employees' data may add some other features that can increase the odds of hiring safe drivers. For example, in some countries there might be some predictive power to the number of years you have had a driver's licence.
For an example of the extravagant claims made by driver management companies, and the reality behind, you can read this paper;
af Wåhlberg, A. E. (2007). Making bad look good; the scientific claims of Interactive Driving Systems. Download pdf-file
Surveillance
Unfortunately, people behave differently when no-one is looking, especially the boss. This is very much so for drivers. If you install some kind of surveillance system in your vehicles, it will probably be cost-effective within a year or two (as found for police drivers by Larson et al, 1980). There are two or three important factors to consider when buying or developing such a system; what does it measure, how automatic is the system, and what can the data be used for? Today, there is a wealth of information which can be had from the engine computer, so the trick is to find a variable which is predictive of accidents and fairly easy to use. Speed and acceleration variables would seem to be the best choices today (read more about this in my papers on the Current projects page). To reduce false alarms (identifying safe drivers as unsafe) and in general making the system fairly automatic (and therefore cheap) in operation, it must be calibrated for the company in question, preferably with different levels of deviant behavior leading to different actions from the company (direct feedback from computer, feedback from boss, training and disciplinary actions).
Accident data and characteristics
Do you know what categories of accidents are most common in your company? Have you made a statistical analysis of this? If the answers are yes/no, then you are most probably wrong on the first, because people just cannot make very reliable estimates of relative numbers. In general, the first thing you need is a very thorough system of accident reporting and a database with all relevant information (see my papers on accident characteristics on the Current projects page). Given such information, it is possible to calculate the optimal driver requirements, effects of weather (do you send out a warning to your drivers when there is bad weather up ahead?), the most dangerous manouvers (feedback to training) and shortcomings of the traffic environment (feedback to authorities).
Health, sleep and work schedules
In the area of physical fitness versus work there is not much to be said concerning its relation to accidents, because there is very little research done here. However, it can be stated that bus driving has been found to be among the most stressful occupations there are, with very high-turnover, sick-leave and rates of work-related sickness (Kompier et al, 1990; Kompier & Di Martino, 1995). However, sleep problems (mainly sleep apnoea) has been rather extensively investigated, and it would seem to be no doubt that there is indeed an association between these and accidents.
References
Blom, D. H., Pokorny, M. L., & van Leeuwen, P. (1987). The role of age and experience in bus drivers' accidents. International Journal of Epidemiology, 16, 35-43.
Kompier, M. A., & Di Martino, V. (1995). Review of bus drivers' occupational stress and stress prevention. Stress Medicine, 11, 253-262.
Kompier, M., Mulders, H., Meijman, T., Boersma, M., Groen, G., & Bullinga, R. (1990). Absence behavior, turnover and disability: A study among city bus drivers in the Netherlands. Work Stress, 4, 83-89.
Larson, L. D., Schnelle, J. F., Kirchner, Jr., R., Carr, A. F., Domash, M., & Risley, T. R. (1980). Reduction of police vehicles accidents through mechanically aided supervision. Journal of Applied Behaviour Analysis, 13, 571-581.
Lester, J. (1991). Individual differences in accident liability: review of the literature. Research Report 306. Crowthorne: Transport and Road Research Laboratory.
Peck, R. C. (1993). The identification of multiple accident correlates in high risk drivers with specific emphasis on the role of age, experience and prior traffic violation frequency. Alcohol, Drugs and Driving, 9, 145-166.
Spratling, F. H. (1961). Accidents among older London Transport drivers; an analysis. British Transport Review.
West, R. (1998). Cross-cultural generalisability of relationship between anti-social motivation and traffic accident risk. TRL Report 294. Crowthorne: Transport Research Laboratory.
