Prevalence of motor vehicle crashes involving drowsy drivers, United States, 1999–2008
Highlights
► Information about driver drowsiness often is missing in crash data. ► The proportion of crashes involving drowsy drivers likely has been underestimated. ► Multiple imputation was used to address missing data in a representative sample. ► An estimated 16.5% of fatal crashes in the United States involve a drowsy driver. ► An estimated 7.0% of crashes in the United States involve a drowsy driver.
Introduction
Operator drowsiness, sleepiness, or fatigue (hereafter referred to as drowsiness) has been documented as a causal or contributing factor in aviation, maritime, and trucking accidents (National Transportation Safety Board, 1990, National Transportation Safety Board, 1991, National Transportation Safety Board, 1994). However, estimates of the proportion of motor vehicle crashes that involve drowsy drivers vary widely depending upon sources of data and analytical methods used.
Knipling and Wang (1994) analyzed data from years 1989 through 1993 and reported that 0.9% of all police-reported crashes and 3.6% of fatal crashes in the United States involved a driver coded as drowsy. The authors cited several other studies of large crash databases that reported results in this range, but also noted that studies of crash databases were likely to underestimate the prevalence of crashes that involved drowsy drivers due to data limitations. Wang et al. (1996) analyzed a national sample of crashes that occurred in 1995 in which a passenger vehicle was towed and estimated that 2.6% of these crashes involved a drowsy driver. The authors noted that the role of drowsiness was unknown in 46% of crashes and that their estimate of the proportion of crashes involving drowsy drivers was likely conservative.
An Australian study classified crashes as drowsiness-related if drowsiness was cited by the police or if the crash involved departure from the roadway in the absence of other causes or contributing factors suggestive of attentive driving, and estimated that 6% of all reported crashes and 15% of fatal crashes in 1992 in New South Wales involved a drowsy driver (Fell, 1994). Knipling and Wang (1995) used a similar method to refine their earlier estimates (1994), and estimated that 1.2%–1.6% of all reported crashes involved a drowsy driver. Masten et al. (2006) used data from a sample of crashes in North Carolina to develop a statistical model to classify crash-involved drivers as drowsy or not drowsy, applied this model to national data on fatal crashes, and estimated that 15–33% of drivers involved in fatal crashes nationwide from 2001 through 2003 were drowsy.
In a study in which 109 vehicles were equipped with cameras and other data collection equipment for a period of 12–13 months, Klauer et al. (2006) reviewed pre-event video and estimated that 22% of crashes and near crashes in the study population involved moderate to severe drowsiness. However, the majority of outcomes in the study population were near crashes and unreported minor crashes (Dingus et al., 2006); the extent to which these results may be generalized to more severe crashes is unknown.
The aim of the current study was to improve upon past estimates of the proportion of crashes that involve a drowsy driver, overall and relation to crash severity, using multiple imputation to address missing data on driver drowsiness.
Section snippets
Data
The data used here was obtained from the National Highway Traffic Safety Administration (NHTSA)’s National Automotive Sampling System (NASS) Crashworthiness Data System (CDS) (2009), a sample of police-reported crashes in which a passenger vehicle (car, light truck, utility vehicle, or van) was towed. Data reported in NASS CDS is compiled by teams of investigators who obtain information from police crash reports, medical records, crash reconstructions, and interviews with drivers and passengers
Crash-level estimates
An estimated 7.0% of all crashes (95% confidence interval [CI]: 4.6%, 9.3%), 13.1% of non-fatal crashes in which at least one person was admitted to a hospital (95% CI: 8.8%, 17.3%), and 16.5% of fatal crashes (95% CI: 12.5%, 20.6%) involved a drowsy driver (Fig. 1).
Crashes that occurred between 11:00 PM and 6:59 AM were nearly five times as likely to have involved a drowsy driver as were crashes that occurred between 7:00 AM and 10:59 PM (Prevalence ratio: 4.8, 95% CI: 3.2, 7.3) (Fig. 2). Crashes
Principal finding
This study used a multiple imputation method to estimate the proportion of crashes that involved a drowsy driver. The estimate that 7.0% of all crashes involved a drowsy driver is 79% greater than the estimate derived from the original non-imputed data (3.9%); the estimate that 16.5% of fatal crashes involved a drowsy driver is 358% greater than the estimate derived from the original data (3.6%).
Interpretation
The data analyzed for this study records drowsiness as an assessment of a driver's attention
Conclusion
The proportion of crashes that involve a drowsy driver is likely greater than existing crash databases reflect, due to the possibility that some drivers whose pre-crash state of attention was unknown may have been drowsy. The results of this study suggest that without taking these drivers into account, a study would underestimate the prevalence of all crashes involving drowsy drivers by about 80% and may underestimate the prevalence of fatal crashes involving drowsy drivers by over 350%.
Acknowledgement
The author would like to thank Peter Cummings and Robert Scopatz for their helpful suggestions and feedback on this study.
References (27)
- et al.
A new framework for managing and analyzing multiply imputed data in Stata
The Stata Journal
(2008) Methods for estimating adjusted risk ratios
The Stata Journal
(2009)- et al.
Air bags and passenger fatality according to passenger age and restraint use
Epidemiology
(2002) - et al.
The 100-Car Naturalistic Driving Study. Phase II – Results of the 100-Car Field Experiment. DOT HS 810 593.
(2006) - Fatality Analysis Reporting System, 1999–2008 [Data files] National Highway Traffic Safety Administration, Washington,...
Safety Update: Problem Definition and Countermeasure Summary: Fatigue
(1994)- et al.
A clinical and polygraphic study of episodic phenomena during sleep
- et al.
The Impact of Driver Inattention on Near-Crash/Crash Risk: An analysis using the 100-Car Naturalistic Driving Study data, DOT HS 810 594
(2006) - et al.
Revised estimates of the U.S. drowsy driver crash problem size based on General Estimates System case reviews
- et al.
Crashes and Fatalities Related to Driver Drowsiness/Fatigue
(1994)