Frequency of fatigue-related crashes
Different methods yield different estimates concerning the frequency of fatigue-related crashes. The following sections present estimates based on various methods:
The police crash reports in different countries (e.g. Netherlands, UK, North Carolina USA) indicate a 1-4% incidence of sleep-related crashes of all registered crashes . For example in the Netherlands, the combined primary cause of a crash that is attributed to sleep/illness occurs in about 1% of all registered crashes. It is likely that these police reports greatly underestimate the problem. In most countries, police are not (yet) so alert to fatigue as crash cause. Also, most drivers will be reluctant to admit that they were very tired or had fallen asleep at the time of the crash. In addition, the crash itself would have made most of the symptoms of fatigue disappear. So the 1% figure certainly represents an underestimation.
Questionnaire studies have provided completely different conclusions about fatigue's role in road crashes . Based on these methods, estimates of the percentage of sleep-related crashes vary greatly, but often are in the range of 10-25 percentage points higher than can be concluded from police reports. The higher percentages have been found particularly in studies that have examined lorry crashes and/or fatal crashes. All these estimates are from abroad.
Based on findings from a survey study amongst 4600 male car drivers in England, Maycock  concluded that fatigue played a role in 9-10% of all crashes. This percentage was higher for motorways (20%) than for roads inside urban areas (7%) or for other roads outside urban areas (14%).
Naturalistic observation study
Naturalistic observation of driving behaviour provides the most direct evidence of driver fatigue in real circumstances. A naturalistic driving study unobtrusively registers the actual driving behaviour of drivers who drive their own cars to destinations of their own choosing without an experimenter present. The registration of driving behaviour is continuously done by various instruments over a longer period of time (one year or longer). A naturalistic observation study may link the outward signs of fatigue (such as closed eyes) to real driving behaviour.
The 100-Car Naturalistic Driving Study is an instrumented vehicle study designed to collect a large volume of naturalistic driving data over an extended period of time. The researchers installed instruments and sensors in 100 vehicles that were then driven as ordinary vehicles by ordinary drivers for one year. Drivers were given no special instructions, no experimenter was present, and the data collection system was unobtrusive. In addition, driversí own vehicles were instrumented for 78 out of the 100 vehicles.
The study collected data on 15 police-reported and 67 non-police reported crashes, 761 near-crashes (situations requiring a rapid, severe evasive manoeuvre to avoid a crash) and 8,295 incidents (situations requiring an evasive manoeuvre occurring at less magnitude than a near-crash). In this study, fatigue was judged to be a contributing factor in approximately 12% of crashes, 10% of near-crashes, and 7% of crash-relevant conflicts . Fatigue was measured by an observer rating of drowsiness, measured on a scale from 0 to 100 in increasing severity of drowsiness. The scale was based on the Wierwille and Ellsworth  rating system for driver fatigue. This rating system is based on observable personal characteristics such as facial tone, eye blinks, eye closures, head movements, staring, lack of activity, eye expression etc.
In-depth studies investigate characteristics of crashes to find out whether fatigue may have played a role. In an in-depth study, Horne and Reyner  established that about 20% of crashes on motorways were sleep-related. The injury level of these crashes is quite high since no braking occurs.
In France, Philip et al.  applied the criteria from Horne and Reyner on serious injury crashes in the period 1994-1998. This study looked at single vehicle crashes under good weather and road circumstances on road segments without intersections. They found that about 10% of 68.000 analysed crashes were related to fatigue (as determined by the Horne/Reyner criteria). This is probably an underestimation since collisions with other vehicles that satisfied the Horne/Reyner criteria were not taken into account.
In Germany, a similar in-depth crash study established that about 24% crashes on a German motorway had to do with fatigue .
In Finland, all fatal road accidents are investigated in-depth by multidisciplinary investigation teams. The percentage of fatal accidents involving fatigue or falling asleep between 1995 and 1999 fluctuated between 16-19% (Hantula, 2000 in: ETSC ).
Haworth et al.  estimated sleep or fatigue to be involved in about 20 % of truck-involved fatal crashes. Based on a literature study involving both in-depth and questionnaire studies, Amundsen & Sagberg  found that fatigue was a contributing factor in 15 to 20% of truck crashes.