The effects of sleep deprivation and text messaging on pedestrian safety in university students

Abstract

Study Objectives

This study assesses the impact of sleep deprivation and text messaging on pedestrian injury risk.

Methods

A total of 36 university students engaged in a virtual reality pedestrian environment in two conditions: sleep deprived (no sleep previous night) and normal sleep (normal sleep routine). Sleep was assessed using actigraphy and pedestrian behavior via four outcomes: time to initiate crossing, time before contact with oncoming vehicle, hits/close calls, and looks left/right. During each condition, participants made half the crossings while text messaging. Participants also completed the Useful Field of View test, the Psychomotor Vigilance Test, and Conners’ Continuous Performance Test in both conditions.

Results

While sleep deprived, students crossed significantly closer to oncoming vehicles compared with after normal sleep. While text messaging, crossed closer to vehicles and took longer to initiate crossings. Safety risks were amplified through combined sleep deprivation plus text messaging, leading to more virtual hits and close calls and shorter time before vehicle contact while crossing. Sleep-deprived students demonstrated impairments in functioning on cognitive tests.

Conclusions

University students’ pedestrian behavior was generally riskier, and their cognitive functioning was impaired, when sleep deprived compared with after normal sleep. This effect was exacerbated when distracted by text messaging.

Statement of Significance.

This research demonstrates the risk of sleep deprivation on pedestrian behaviors in university students. Furthermore, risky pedestrian behaviors were exacerbated by distraction when university student pedestrians are both text messaging and sleep deprived. Sleep deprivation was found to affect cognition, and impaired cognition predicted riskier pedestrian behavior when text messaging while sleep deprived.

Introduction

Sleep deprivation is a prominent health problem in the United States. While experts recommend that individuals ages 18–60 years old obtain a minimum of 7 h of sleep each night, about 35% of adults do not adhere to this guideline [1, 2]. In university students, this problem is exaggerated: a study with 1,125 Midwestern US undergraduate students found that 20% of students reported at least one “all-nighter” over the course of a month, 71% slept less than 8 h every night, and overall sleep schedules were highly erratic [3]. Inadequate sleep is linked to a range of negative outcomes, including increased risk taking and impulsivity, reduced reaction time and attention, impaired academic performance and higher rates of obesity, diabetes, stroke, and depression [1, 2, 4–8].

Sleep deprivation is also related to increased risk of injury. This consequence is pervasive, spanning multiple settings including pedestrian and bicycling safety, sports injury, falls risk, and motor vehicle crashes [4, 9, 10]. These outcomes could be the consequence of changes in cognitive function that result from sleep restriction such as increased risk taking and decreased reaction time and attention [11]. In one study of pedestrian safety, for example, 14- and 15-year olds who crossed a virtual reality pedestrian environment street several times the morning after 4 h of sleep experienced more virtual hits and close calls, crossed closer to oncoming vehicles, looked left and right more frequently, and took longer to initiate crossings [10]. However, research investigating the effects of sleep deprivation on pedestrian safety has not incorporated rigorous cognitive measures and is limited to early adolescent populations, who as newly skilled pedestrians may be less capable of compensating for the negative cognitive and behavioral outcomes associated with sleep deprivation [12].

Previous research also demonstrates strong relations between media use and pedestrian injury risk [13, 14]. College students frequently use media [15], and research suggests an increased risk of unsafe crossings and collisions with virtual vehicles among college students who were text messaging or listening to music while crossing the road compared with undistracted pedestrians [16]. Similar results were found in children ages 10–11 when talking on the phone during street crossings [17]. College students frequently use media, which could put them at risk.

The present study used a within subjects experimental crossover design to assess the effects of both sleep deprivation and smartphone distraction on university students’ pedestrian safety. Further, we explored the interaction between both sleep deprivation and smartphone distraction as it related to pedestrian crossing outcomes as well as assessing basic cognitive outcomes relevant to pedestrian behavior. Participants completed cognitive tasks and virtual street-crossing twice, once after normal sleep and once after a night of complete sleep deprivation. We hypothesized that sleep-deprived individuals would have poorer pedestrian outcomes and that this effect would be exacerbated by smartphone use. We also hypothesized that sleep deprivation would significantly impair relevant cognitive functioning.

Methods

Participants

Participants were recruited from the student population at University of Alabama at Birmingham. All students age 19 and over (the age of majority for consenting in Alabama at the time the study was run), both undergraduate and graduate, were eligible. Exclusion criteria were minimal and included only cognitive or physical disabilities that prevented full participation in the experimental protocol (e.g. severe intellectual disability, blindness, and use of a wheelchair).

The study protocol was approved by the Institutional Review Board at the University of Alabama at Birmingham. Informed consent was obtained from participants, and participants were compensated for their time.

General protocol

Following informed consent processes and completing a basic demographic questionnaire, participants completed laboratory visits twice in groups of four, once after obtaining a normal night’s sleep in their own home/apartment/dormitory and the other time following a full and monitored night of sleep deprivation as a foursome in a hospital sleep lab. Visit order was randomized across participant foursomes. During both laboratory visits, participants completed the following procedures, all detailed below, during a 2-h laboratory visit: 3 cognitive tasks (Conners’ Continuous Performance Task, Psychomotor Vigilance Test, and Useful Field of View), a brief demographic self-report survey, 15 undistracted crossings within the virtual reality pedestrian environment, and 15 crossings within a virtual reality pedestrian environment while text-messaging.

In the virtual reality pedestrian environment, the participant was immersed in the virtual environment in which they were able to watch vehicles pass bidirectionally on three consecutive computer monitors and hear environmental and traffic noise through speakers in the room (Figure 1). After deciding it was safe to cross, the participant stepped off of a wooden curb onto a pressure plate connected to the computer and a gender-matched avatar was then activated to cross the street. The walking speed of the avatar was matched to the participants’ average pace recorded prior to the simulated task. Specifically, participants walked the same distance as the road’s crosswalk four times in a long hallway, with the instructions to “walk at a speed you would use to cross the street.” The walking time was recorded by stopwatch and the average of the four walks inputted into the virtual pedestrian environment.

Figure 1.

Virtual reality pedestrian environment.

During each laboratory visit, participants performed practice trials to familiarize themselves with the program and then engaged in 30 virtual street crossings. These crossings were divided into groups of 15 distracted (text-messaging) crossings and 15 crossings without distraction. The crossing conditions were randomized between participants with a testing break from the virtual reality between conditions. Text-messaging conversations were conducted between participants and a research assistant following a standardized script. Topics were benign and not highly emotional (e.g. what are you studying in school, favorite music, and favorite movies). Participants were instructed to check and respond to text messages as quickly as possible. If participants did not check their phone and respond after the text message was sent, they were prompted to check and respond as quickly as possible. Participants generally responded almost immediately upon receiving a text message during the experiment.

To ensure full sleep deprivation, participants reported to the Children’s of Alabama Sleep Disorders Clinic at 10:30 pm the night before their sleep deprived visit. A research assistant observed the group of four participants throughout the night to ensure they stayed awake. Refreshments were provided, but no caffeine, alcohol, or stimulant medication was permitted. The observation room contained a television and magazines for entertainment. Additionally, participants were allowed to bring other entertainment, such as computers, board games, or video game consoles. At 7:30 am, the following morning, participants were escorted a few blocks to their laboratory visit.

During the normal sleep condition, participants were asked to obtain their “normal sleep” and to sleep in their typical location (home, apartment, and dormitory). Participants wore actiwatches the night before the normal sleep condition to confirm that they obtained sleep before the laboratory visit [18, 19]. Experimental appointments following the night of normal sleep were almost always scheduled in the morning, although scheduling complications led to a small number of afternoon appointments. The two visits were scheduled about 2 weeks apart, providing sufficient time for participants randomized to the sleep deprivation condition first to “reset” their typical sleep schedule. Morning caffeine consumption and day-before daytime napping were prohibited on both laboratory visit days.

Measures

Participants completed three standard cognitive tasks at each laboratory visit.

“The Useful Field of View” (UFOV) test [20] is a validated measure evaluating visual attention. The touch screen response personal computer version was used for this study [21]. UFOV is an approximately 15-min computer assessment that consists of a visual stimuli presenting in central vision, peripheral vision, and with white triangles presented throughout the screen as added distractors. For example, an image or a car or truck may appear for a fraction of a second in the center of the screen with the instruction to identify which image was shown. The same car or truck image will then flash in different locations around the outside of the screen; this is the peripheral target localization task in which the participant must identify the screen location where the target appeared. Three subtests are completed: visual processing speed, selective attention, and divided attention. The visual processing speed subtest measures processing speed without other demands or distractions. Selective attention is a measure of participants’ ability to selectively react to certain stimuli when several occur simultaneously. Divided attention is the ability to focus on multiple tasks at once. Based on these three subtests, an overall useful field of view score is produced, and this final score provides a measure of each participant’s overall attention and processing speed, with greater numbers indicating poorer outcomes.

“The Psychomotor Vigilance Test” (PVT) [22] evaluates reaction time and attention. During the 10-min test, participants were instructed to monitor a red box and tap the screen as soon as a yellow stimulus counter appeared on the screen. This action stopped the counter and displayed the reaction time in milliseconds. Participants were instructed to respond as quickly as possible without responding before the stimulus appeared. Three scores are obtained from the PVT. First, mean reaction time is calculated by averaging the length of time between the appearance of the stimulus and the participant tapping the screen. Second, lapse count refers to the number of times the response time was greater than or equal to 500 ms after the stimulus appeared. It is considered a measure of attentional processes. Third, false starts indicate the number of times the participant tapped the screen before or less than 100 ms after the stimulus appeared; this offers a measure of errors of commission.

“The Conners’ Continuous Performance Task” (CPT) [23] evaluates attention and executive control. During this 14-min computer test, participants are instructed to hit the spacebar when any letter except “X” appears. The test produces numerous scores aimed at understanding participant measures of inattentiveness, response time, and sustained attention. Four scores were most relevant to the present study’s hypotheses: Attention-Deficit/Hyperactivity Disorder (ADHD) clinical index, hit reaction time, omissions and detectability. The ADHD clinical index is an overall percentage score produced by indexing participants’ performance in inattentiveness, impulsivity, sustained attention, and vigilance, and it indicates the probability of the respondent having a clinical disorder characterized by attention deficits. Hit reaction time is a score of response speed [24]. Omissions refers to the number of missed targets occurring when a letter other than X appeared on the screen without the participant pressing the spacebar. This score is a measure of overall inattentiveness. Detectability measures participants’ ability to discriminate between target and non-target stimuli.

Pedestrian safety measures

We considered four outcome measures based on behavior in the virtual reality pedestrian environment [25, 26].

• 1) Hits and close calls: Hits were defined as any direct collisions between the virtual pedestrian and a vehicle. Close calls were instances when the pedestrian was within 1 s of being struck by a virtual vehicle. For analysis, hits and close calls were summed and are presented as a count (out of 15 crossings).

• 2) Looks left and right: This variable assessed attention to traffic and was evaluated by summing the number of times participants looked to the left and right while waiting to cross, divided by their average wait time (the time from when traffic started to move until the time when the pedestrian stepped off the curb and into the crosswalk) in seconds. Looks toward traffic were tallied by TrackIR 4 [27]; a commercial head-tracking software that monitored participants’ visual attention to traffic from the left and right.

• 3) Start delay: As a proxy for the cognitive decision-making task of pedestrian behavior, we considered start delay, defined as the time in seconds after a safe gap between oncoming cars appeared and before the participant initiated crossing into it. A pedestrian who decides quickly to enter a gap in traffic is likely to be safer, as traffic will be further away as the pedestrian crosses the street.

• 4) Time to contact: As a second measure of safety during the crossing, we assessed time to contact, which was defined as the smallest temporal gap in seconds between the virtual pedestrian and any oncoming vehicle during the crossing.

Analysis plan

Analyses were conducted in five steps. First, descriptive statistics were examined. Second, Pearson correlations and chi-square tests were used to assess if demographic variables and sleep duration were related to pedestrian safety or cognitive outcomes and therefore needed to be included in the analyses as covariates. Third, to assess our primary hypotheses that both short sleep duration and text messaging would result in riskier pedestrian behaviors, linear mixed effects models employing random intercepts for each participant were conducted comparing main effects of texting and sleep conditions on each of the four pedestrian variables. Each model assumption was examined and determined to be valid. Thus, all data were used in each mixed model without transformation. Post-hoc tests were performed without adjustment due to our small sample and complex modeling scheme. Percent of total variability for each outcome explained by these random effects were estimated. Fourth, paired samples t-tests were conducted comparing the sleep deprived and normal sleep conditions for each of the results from the UFOV, PVT, and CPT. Finally, linear regressions were performed with cognitive outcomes predicting pedestrian behavior for each of the three pedestrian variables in which the sleep deprivation and texting condition was significantly different than another condition (hits and close calls, start delay, and time to contact). To determine which cognitive variables to include in the models, correlations were performed between the cognitive variables and the text messaging and sleep-deprived conditions of each pedestrian variable. Then, linear regression models were computed with the text messaging and sleep-deprivation condition for each of the three pedestrian variables as the dependent variable and cognitive factors with a Pearson correlation coefficient ≥0.23 as the independent variables except for hits and close calls, for which the three significant correlations (Pearson correlation coefficient ≥0.36) were used.

Results

Thirty-six university students (M_age = 22.85, SD_age = 2.76; 25% male; 41.7% Caucasian) participated. Table 1 lists descriptive statistics about the sample and Table 2 lists descriptive data for the four pedestrian crossing outcomes in all four conditions.

Table 1.

Participant demographics N = 36

Characteristics	%
Gender
Female	74.3
Ethnicity
Caucasian	41.2
African American	29.4
Other	29.4
Age (years), mean ± SD	22.96 ± 2.72
Sleep duration-normal sleep (h), mean ± SD	7.05 ± 1.77

Table 2.

Pedestrian outcomes

Condition	Normal sleep and no texting		Normal sleep and texting		Sleep deprived and no texting		Sleep deprived and texting
	Mean	SD	Mean	SD	Mean	SD	Mean	SD
Hits/close calls	0.64	0.93	1.11	1.33	0.83	1.18	1.39	1.79
Start delay	0.88	0.38	1.02	0.50	0.78	0.27	0.92	0.33
Time to contact	4.70	0.74	4.18	0.71	4.42	0.73	4.09	0.88
Looks left/right	0.49	0.22	0.41	0.22	0.51	0.19	0.44	0.19

Residual diagnostics were assessed for each model to examine all model assumptions. Using conditional residuals, no large departures from normality were observed. Because randomization order, age, gender, race, and sleep duration the night of the normal sleep condition were unrelated to the pedestrian safety outcomes, no covariates were included in subsequent analyses.

As hypothesized, there was a significant overall effect for each condition (Table 3). For hits and close calls, post-hoc comparisons indicated that the mean score for the sleep deprivation and text messaging condition (M = 1.39, SD = 1.79) was significantly different than the sleep deprivation and no text messaging (M = 0.83, SD = 1.18) and the normal sleep and no text messaging (M = 0.64, SD = 0.93) conditions, indicating that participants were hit or almost hit by a car more often when both sleep deprived and text messaging than when they were not text messaging while sleep deprived or normally rested. For start delay, the mean score for the normal sleep and text messaging condition (M = 1.02, SD = 0.50) was significantly different from the normal sleep and not text messaging (M = 0.88, SD = 0.38) and sleep deprivation and not text messaging (M = 0.78, SD = 0.27) conditions, indicating that participants waited longer to initiate a crossing after a safe gap appeared between oncoming cars when normally rested and text messaging than when not text messaging in either sleep condition. Additionally, the sleep deprivation and text messaging condition (M = 0.92, SD = 0.33) resulted in significantly longer delay in initiating a crossing after a safe gap appeared than the sleep deprivation and not text messaging condition (M = 0.78, SD = 0.27).

Table 3.

Results of repeated measures ANOVAs for pedestrian variables

	Mean difference						SE mean diff	Overall p	% Variance explained
Condition 1 vs. Condition 2	Normal sleep, no texting Sleep deprived, no texting	Normal sleep, no texting Normal sleep, texting	Normal sleep, no texting Sleep deprived, texting	Normal sleep, texting Sleep deprived, no texting	Normal sleep, texting Sleep deprived, texting	Sleep deprived, no texting Sleep deprived, texting
Hits/close calls	−0.19	−0.47	−0.75**	0.28	−0.28	−0.56*	0.25*	.02	0.37
Start delay	0.1	−0.15*	−0.05	0.25**	0.10	−0.14*	0.07**	<.01	0.46
Time to contact	0.27*	0.52**	0.61**	−0.25	0.09	0.34*	0.13**	<.01	0.51
Looks left/right	−0.02	0.08*	0.05	−0.10*	−0.03	0.07	0.04*	0.03	0.39

* p < .05.

** p < .01.

For time to contact, participants crossed with significantly greater temporal distance from cars in the normal sleep and no text messaging condition (M = 4.70, SD = 0.74) as compared with the other three conditions. Further, participants crossed significantly closer to cars when sleep deprived and text messaging (M = 4.09, SD = 0.88) than when sleep deprived and not text messaging (M = 4.42, SD = 0.73). For looks left and right, participants looked left and right significantly less frequently when normally rested and text messaging (M = 0.41, SD = 0.22) than when normally rested and not text messaging (M = 0.49, SD = 0.22) and when sleep deprived and not text messaging (M = 0.51, SD = 0.19).

To assess for differences on the cognitive tasks between conditions, paired samples t-tests were conducted (Table 4). Results for the UFOV demonstrated significantly poorer scores for selective attention (t(26) = 4.28, p < .01), divided attention (t(26) = 2.41, p = .02) and total UFOV (t(26) = 3.91, p < .01) when sleep deprived compared with after normal sleep. There was no significant difference between the groups in visual processing speed. On the PVT, participants had a significantly slower reaction time (t(34) = 2.52, p < .05) and greater lapse count (t(34) = 3.68, p < .01) when sleep deprived. The difference in false starts was not significant. Finally, for the CPT, participants had a significantly higher ADHD Clinical Index (t(31) = 4.75, p < .01), increased omissions (t(27) = 3.65, p < .01), and a slower hit reaction time (t(31) = 2.87, p < .01) when sleep deprived. There was no significant difference between the groups in detectability.

Table 4.

Paired samples t-tests for cognitive tasks

Variable	M (SD) (sleep-deprived condition)	M (SD) (normal sleep condition)	df	t
UFOV
Total score	229.23 (±135.75)	134.57 (±62.18)	26	3.91**
Visual processing speed	19.42 (±8.49)	17.33 (±2.27)	27	1.24
Divided attention	57.07 (±64.99)	24.98 (±20.83)	26	2.41*
Selective attention	152.74 (±83.7)	92.36 (±55.36)	26	4.28**
PVT
Mean reaction time (ms)	497.14 (±501.42)	288.84 (±41.17)	34	2.52*
Lapse count	9.63 (±11.63)	2.43 (±2.44)	34	3.68**
False starts	4.20 (±3.42)	3.60 (±3.77)	34	1.48
CPT
ADHD Clinical Index	60.85 (±33.68)	38.20 (±26.02)	31	4.75**
Hit reaction time	52.28 (±8.27)	48.47 (±9.27)	31	2.87**
Omissions	4.67 (±5.96)	0.80 (±1.14)	31	3.65**
Detectability	0.56 (±0.50)	0.57 (±0.44)	30	0.18

*p < .05.

**p < .01.

To assess the relationship between cognition and the joint text messaging and sleep-deprived condition of pedestrian behavior, linear regressions were performed with cognitive outcomes predicting pedestrian behavior for each of the three pedestrian variables in which the sleep deprivation and texting condition was significantly different than another condition (hits and close calls, start delay, and time to contact). UFOV selective attention, and PVT reaction time and PVT lapse count accounted for 33% of the variance in predicting the text messaging and sleep-deprivation condition of hits and close calls, F(2,24) = 3.69, p = .03. CPT detectability and PVT lapse count accounted for 20% of the variance in the text messaging and sleep-deprivation condition of time to contact, F(2,30) = 3.63, p = .04. CPT reaction time and total UFOV score accounted for 40% of the variance in start delay in the text messaging and sleep deprivation condition (Table 5).

Table 5.

Summary of linear regression analysis predicting pedestrian behavior in the text messaging and sleep-deprivation condition

Measure	B	SE B	β	t
Hits and close calls
PVT reaction time	0.00	0.00	0.87	2.13*
PVT lapse count	−0.09	0.05	−0.66	−1.60
UFOV selective attention	0.01	0.00	0.37	1.97
Adjusted R2		0.24
F for change in R2		3.69*
Time to contact
CPT detectability	0.35	0.30	0.19	1.16
PVT lapse count	−0.03	0.01	−0.38	−2.29*
Adjusted R2		0.20
F for change in R2		3.63*
Start delay
Total UFOV	0.00	0.00	−0.49	−2.86**
CPT Reaction Time	−0.001	0.00	0.58	3.42**
Adjusted R2		0.40
F for change in R2		7.65**

*p < .05.

**p < .01.

Discussion

Sleep deprivation and text messaging both significantly influenced university students’ pedestrian behavior and negatively impacted their cognitive functioning. After one night of complete sleep deprivation, participants who were sleepy crossed significantly closer to oncoming cars as compared with when normally rested. Similarly, participants who were text-messaging crossed closer to oncoming cars than they did when they crossed undistracted. Further, pedestrian safety was impacted to a greater extent when participants were both sleep deprived and distracted through text messaging. When both sleep deprived and text messaging, participants crossed significantly closer to cars and experienced almost twice the number of virtual hits and close calls than when they were normally rested and not text messaging. It is noteworthy that sleep significantly affected both the number of hits and close calls and time to contact, the two measures of safety during crossings.

Our results concerning delays entering safe traffic gaps after a safe gap appeared were intriguing. Texting seemed to delay participants’ entry into safe traffic gaps whether they were sleepy or not, and sleep deprivation had minimal influence on start delays in our sample. Finally, the poorest attention to traffic, as measured by looks left and right toward traffic, occurred when rested and texting; in that condition, participants looked toward traffic significantly less often than when they were rested and undistracted and when they were sleepy and undistracted. Participants who were sleepy and distracted may have compensated for their sleepiness by attending to traffic more often, whereas participants who were rested and texting became distracted by their texting and failed to attend to traffic as carefully.

The results of the cognitive tests generally support the findings from the virtual reality pedestrian crossings. When sleep deprived, students demonstrated significant impairments in divided attention, or the ability to attend to two things at once, and selective attention, or the ability to attend to only relevant stimuli and ignore irrelevant information on the UFOV. Additionally, students demonstrated poorer overall attention and processing speed on the UFOV when sleep deprived. On the PVT, students displayed a slower reaction time and had a greater number of lapses when sleep deprived, indicating diminished ability to react in response to digital signals when tired that may translate to diminished ability to detect and react to stimuli in traffic environments. This is consistent with previous research demonstrating a positive relationship between cognitive lapses and sleep deprivation [28]. On the CPT, participants displayed a substantial increase in the ADHD Clinical Index when sleep deprived, suggesting their performance resembled the symptoms of ADHD after a night of sleep deprivation. Specifically, they failed to respond to targets (omissions) and had slower hit reaction times. This suggests participants experienced slower reaction time and attention when tired.

When assessing relations between cognitive scores and pedestrian safety, cognitive variables significantly predicted hits and close calls, start delay and time to contact when participants were both distracted and tired. Specifically, slower reaction time, poorer selective attention and decreased attention predicted greater hits and close calls. Hit reaction time on the CPT emerged as a significant independent predictor for hits and close calls, suggesting that participants who were slower to react were more likely to be hit or almost hit. This is consistent with the results for start delay, in which slower reaction time and poorer overall attention and processing speed resulted in greater delay start delay. Additionally, poorer discriminability and poorer attention were associated with decreased time to contact. Lapse count was a significant independent predictor for time to contact, indicating that participants with poorer attention crossed closer to cars.

There are several implications for these results. First, primary care providers must assess and promote sleep health in patients. University students are known to have poor sleep health [29], and our findings suggest implications of this poor sleep hygiene may extend to personal safety while crossing the street, especially when crossing the street while sleepy and distracted, along with the many other negative health outcomes that have been reported (e.g. increased risk for road traffic accidents while driving; impairments in attention, working memory, and decision making; and higher rates of obesity, diabetes, stroke, and depression) [11, 30–33].

The finding that combined sleep deprivation and distraction through text messaging created elevated risk of pedestrian safety for young adults increases the need for intervention among university students. Pedestrian injuries are among the few public health outcomes that are currently increasing in frequency in the United States, and experts attribute the increased risk to distracted pedestrian and driver behavior [34, 35], perhaps exacerbated by pervasive sleepiness among university students. The scope of distracted pedestrian behavior, especially among college students, is now well-documented [36–38]. Efforts to reduce distracted pedestrian behavior, however, are few. Policy interventions have been attempted in a few jurisdictions, with unclear results [39]; behavioral interventions also have been attempted, with mixed results [40]. Future research to unravel the dual risks of sleep deprivation and distracted pedestrian behavior, with attention toward rapid dissemination of effective programs, is urgently needed.

These data may also have implications for scheduling decisions at universities. Combined with other research demonstrating sleep timing is associated with poorer academic performance, our findings support the notion that universities might consider scheduling classes later for students. Initial evidence with a group of freshmen enrolled at the United States Air Force Academy supports this notion [32, 41, 42]; that experiment did not include pedestrian safety outcomes, but did find that a 50-min delay in school start time was associated with an increase in academic performance similar to the effect of raising teacher quality [41].

This experiment had strengths and limitations. Strengths include a rigorous repeated-measures experimental design to study the combined effects of sleep deprivation and distraction through text-messaging within a validated virtual reality pedestrian environment. We chose to distract participants with a rather routine and mundane text-messaging conversation. This is a strength, as it allows us to test distraction with what we perceive as a more typical texting exchange among university students, but it also presents a fairly low cognitive load. A more emotional and engaging text messaging conversation might add to the cognitive load and further impact pedestrian safety.

Several other study limitations should be mentioned. First, the study included only university students, who comprise a high-risk group for both sleep deprivation and pedestrian injury but may not permit generalization to other age groups or to young adults not enrolled in a university. Additionally, our sample of university students was 74% female and older than a typical college student population (mean age = 22.96 years, partly because we included graduate students in the sample); this sample is not fully representative of the student population at UAB or other universities. A second limitation was the relatively small sample size included, although the within-subjects experimental design offers increased statistical power over a between-subjects design. Third, because our virtual environment switched to a third-person perspective upon stepping off the curb, participants were unable to accelerate their pace to compensate for an anticipated close call with an oncoming vehicle. Alternative virtual environment apparatuses might be considered in future or replicative research. Fourth, many participants did not obtain adequate sleep (7–9 h based on National Sleep Foundation recommendations) [43] during the normal sleep condition. Sleep restriction in these participants may have limited the effect comparing performance among the sleep deprived and normally rested groups. Last, although most appointments for participants after the normal night’s sleep were scheduled in the morning, scheduling complications led to occasional afternoon sessions after the normal night’s sleep, resulting in possible circadian effects on the outcomes. Relatedly, we did not collect data on participants’ typical sleep patterns and while “all-nighters” commonly occur in university students, students more often have erratic sleep patterns.

Conclusion

Safe pedestrian behavior requires complex cognitive functioning, and sleep deprivation significantly impairs cognition. Additionally, risky pedestrian behaviors are exacerbated by distraction when university student pedestrians are both text messaging and sleep deprived.

Funding

This project was sponsored by the Kaul Pediatric Research Institute and the Alabama Children’s Hospital Foundation. A.D.R. reports a grant from NIH/NIDDK number (1K23DK106570). The content is solely the responsibility of the authors and does not necessarily represent the official views of the Kaul Pediatric Research Institute or the Alabama Children’s Hospital Foundation.

Conflict of interest statement. None declared.