Drowsy Driving in Fatal Crashes, United States, 2017–2021

Abstract

Introduction

The National Highway Traffic Safety Administration (NHTSA) states that 91,000 police-reported crashes resulting in 50,000 injuries and 800 fatalities annually—amounting to approximately 1%–2% of all crashes, injuries, and deaths—involve drowsy driving; however, the contribution of drowsy driving in motor vehicle crashes is difficult to measure. Although reports by police officers who investigate crashes sometimes indicate that a driver was drowsy, data derived from these reports are widely regarded as substantial underestimates of the true scope of the problem.  

The current study used data derived from in-depth crash investigations conducted for the National Highway Traffic Safety Administration to develop and validate a model to impute driver drowsiness in cases when the driver’s pre-crash alertness or drowsiness could not be ascertained. The model was then used to impute the involvement of drowsiness in all fatal crashes nationwide that involved at least one car, pickup truck, van, minivan, or sport utility vehicle. 

Methodology

The current study used data from in-depth crash investigations to develop a model to impute driver drowsiness when a driver’s pre-crash alertness or drowsiness could not be determined. The model was validated by treating drivers’ drowsiness as “unknown” in a subset of the cases in which it was actually known, using the model to impute drowsiness in those cases, and comparing the percentages actually drowsy versus imputed as drowsy. The model was then used to impute drowsiness for all drivers involved in fatal crashes in the United States in years 2017–2021. The resulting dataset, containing imputed values of drowsiness for all drivers in fatal crashes, was then analyzed to estimate the percentage of fatal crashes that involved driver drowsiness overall and in relation to selected factors. 

Data was drawn from a representative sample of drivers of cars, pickup trucks, vans, minivans, and sport utility vehicles (hereafter collectively referred to as passenger vehicles) that were towed from the scenes of crashes as well as all drivers of passenger vehicles involved in fatal crashes in the United States in years 2017–2021. Records of drivers of passenger vehicles towed from the scenes of crashes were obtained from the NHTSA’s Crash Investigation Sampling System (CISS). The current study included data from 20,292 drivers of passenger vehicles in transport that were towed from the scenes of 11,268 unique crashes included in CISS in years 2017–2021. Data on drivers of passenger vehicles involved in fatal crashes were obtained from the NHTSA’s Fatality Analysis Reporting System (FARS). The current study included data from 208,727 drivers of passenger vehicles involved in 152,174 unique fatal crashes in years 2017–2021; drivers of other types of vehicles (e.g., large trucks, buses, motorcycles, etc.) were not examined.

Key Findings

Results show that an estimated 17.6% of all fatal crashes in years 2017–2021 involved a drowsy driver. Over the 5-year study period, an estimated 29,834 people were killed in crashes that involved drowsy drivers. The percentage of fatal crashes involving drowsy driving remained approximately constant over the study period; however, the annual number of fatal drowsy driving crashes increased significantly over the study period due to a large increase in total annual fatal crashes. 

In percentage terms, the proportion of fatal-crash-involved drivers who were drowsy was greatest among drivers aged 16–20; however, the largest number of drowsy drivers in crashes were aged 21–34. Men were significantly more likely than women to be drowsy, and men accounted for the vast majority of drowsy drivers involved in fatal crashes. Approximately two-thirds of drowsy drivers in fatal crashes had not been drinking alcohol (blood alcohol concentration, BAC = 0.00), however, one-third had non-zero BAC values. Drowsiness was significantly more prevalent among drivers who had been drinking than among those who had not. An estimated 17% of drivers with BAC of 0.01–0.07 and 20% of drivers with BAC ≥0.08 were drowsy, compared with 11% of those who had not been drinking. 

The percentage of drivers who were drowsy was highest among those who crashed on rural collectors and local roads; however, the greatest number of fatal drowsy driving crashes occurred on urban arterials, as the largest proportion of all fatal crashes irrespective of drowsiness occurs on these roads. The greatest number of fatal drowsy driving crashes occurred between 11:00 PM and 2:59 AM; the percentage of drivers who were drowsy was highest among those who crashed between the hours of 3:00 AM and 6:59 AM. 

Studies of data compiled from in-depth crash investigations alone include too few fatal crashes to provide the spatial or temporal resolution needed to track changes or trends over time or to evaluate countermeasures. The current study sought to overcome these limitations by using data from in-depth crash investigations to develop a model that could then be used to impute, or infer, the involvement of drowsiness in other samples of crashes, including fatal crashes. This methodology could be useful in future studies to investigate trends over time in drowsy driving crashes or to evaluate the impact of countermeasures intended to reduce drowsy driving crashes.