The Effects of Acute Sleep Restriction on Adolescents' Pedestrian Safety in a Virtual Environment

Abstract

Purpose

Over 8,000 American adolescents ages 14-15 require medical attention due to pedestrian injury annually. Cognitive factors contributing to pedestrian safety include reaction time, impulsivity, risk-taking, attention, and decision-making. These characteristics are also influenced by sleep restriction. Experts recommend adolescents obtain 8.5 hours of uninterrupted sleep each night, but most American adolescents do not. Inadequate sleep may place adolescents at risk for pedestrian injury.

Method

Using a within-subjects design, fifty-five 14- and 15-year-olds engaged in a virtual reality pedestrian environment in two conditions, scheduled a week apart: sleep-restricted (4 hours sleep previous night) and adequate sleep (8.5 hours). Sleep was assessed using actigraphy and pedestrian behavior via four outcome measures: time to initiate crossing, time before contact with vehicle while crossing, virtual hits/close calls and attention to traffic (looks left and right).

Results

While acutely sleep restricted, adolescents took more time to initiate pedestrian crossings, crossed with less time before contact with vehicles, experienced more virtual hits/close calls and looked left and right more often compared to when adequately rested. Results were maintained after controlling for age, gender, ethnicity and average total sleep duration prior to each condition.

Discussion

Adolescent pedestrian behavior in the simulated virtual environment was markedly different, and generally more risky, when acutely sleep restricted compared to adequately rested. Inadequate sleep may influence cognitive functioning to the extent that pedestrian safety is jeopardized among adolescents capable of crossing streets safely when rested. Policy decisions might be educated by these results.

Introduction

Unintentional pedestrian injury is the sixth leading cause of death in US adolescents ages 14-15, injuring 8,133 and killing 105 adolescents in 2008.¹ Safe pedestrian behavior depends on advanced cognitive and self-regulatory skills, including attention, reaction time, decision-making, risk-taking and impulsivity.^2-5 These skills diminish with sleep restriction, suggesting tired adolescents may behave in dangerous ways while crossing streets.

To function well, experts currently suggest the typical teenager requires at least 8.5 hours of uninterrupted sleep nightly,⁶ although the recommended time is somewhat controversial and has changed over the years,^7,8 and there is some individual variation in the amount of sleep each person needs.⁹ Adolescents in the US have been found to require more sleep than younger children,^10,11 and they reportedly obtain an average of 7.5 hours on school nights, leading to chronic sleep restriction patterns.¹² American adolescents obtain inadequate sleep for multiple biological, cultural and psychosocial reasons. Inadequate sleep on school nights is likely due to the change in circadian rhythm during adolescence, causing them to fall asleep later, having a biological need to sleep later to obtain adequate sleep,^13-15 but being required to awaken early due to early school start times.¹⁴ Other factors affecting acute adolescent sleep deprivation include decreased supervision by parents, adolescents' desire for independence, increased academic demands, involvement in social and extracurricular activities and response to peer pressure.^13-15 Chronically sleep deprived individuals demonstrate increased impulsivity and risk-taking, slower reaction times, decreased attention, and impaired decision-making.^16-19 Citing these cognitive influences on functioning, scientists have reported increased unintentional injury risk among children with chronically inadequate sleep or sleep disturbance.^20-22

While many adolescents suffer from chronic sleep deprivation over days or weeks,²³ others suffer from highly inadequate sleep on single occasions, causing acute sleep restriction. Although there is little documented research on this area, since four percent of adolescents report sleeping for 5 hours or less each night on a regular basis,²⁴ it is plausible that the number of adolescents who obtain this little sleep on single occasions is higher. Adolescents experience acute sleep restriction for various psychosocial and cultural reasons, including evening social events (e.g., a late-night football game or high school dance) and academic deadlines (e.g., staying up late to prepare for an exam). Cognitive consequences of acute sleep restriction are not understood as well by researchers as those of chronic sleep deprivation, though some evidence links acute sleep restriction with decreased attention²⁵ and with increased unintentional injury risk.^26,27

This study used a within-subjects experimental design to study associations between acute sleep restriction and adolescent pedestrian behavior. Fifty-five 14- and 15-year-olds engaged in a pedestrian environment the mornings after an adequate night's sleep (8.5 hours sleep) and an acutely sleep restricted night (4 hours sleep). We hypothesized adolescents would exhibit riskier pedestrian behavior when sleep restricted.

Method

Participants

Fifty-five 14-15 year olds participated. They were recruited through community sources, including a laboratory database of names of local families interested in participating in research and through distribution of information and permission slips at middle schools in the Birmingham, Alabama area. The sample was 58% female, 53% African American and 47% Caucasian, and had an average age of 14.89 years (SD=0.62). 36% of the sample came from families with a household income at or greater than $80,000, 45% between $40,000 and $79,999, and 19% below $40,000. Signed informed consent was obtained from a parent and signed informed assent from adolescents. The protocol was approved by the university's Institutional Review Board.

Study Protocol

Participants visited the laboratory 3 times, each about one week apart (M days apart = 8.12, SD = 2.02). The first visit served as a baseline. During this visit, parents and adolescents completed several questionnaires (details below), and adolescents received and actigraph and sleep diaries to track sleep for the remainder of the study (about 2 weeks; details below). During the baseline visit, participants also were assigned randomly to an experimental order for the next two scheduled visits. Half (n=27) were scheduled to return for a morning appointment in about a week after obtaining a full (8.5 hour) night's sleep the evening before. That visit would be followed by a morning appointment about two weeks after the initial visit following a restricted (4 hour) night's sleep the evening before. The other half of the sample (n=28) had the order reversed, with sleep restriction scheduled first and a full night's sleep second. All participants were scheduled for standardized morning appointments. Average time between wake time and appointment time was 75.38 minutes (SD = 31.40) for the sleep restricted condition and 70.81 minutes (SD = 34.92) for the adequate sleep condition, and the variation between conditions was not statistically significant (t(44) = 0.92, p = 0.36). Caffeine consumption was prohibited in the mornings before visits. To ensure safety of adolescents following participation in the study, we strongly recommended that they take a nap on the day of the sleep restriction prior to engaging in any activities for which sleepiness may place them at risk of injury; all were non-drivers and left the laboratory with a parent whose sleep was not restricted for research purposes.

During the two morning visits, actigraphy data were reviewed to confirm the adolescent had slept the required amount of time (details below). Participants then engaged in 25 street-crossing trials within the virtual pedestrian environment (details below). At the end of each appointment, participants received a modest incentive ($20 cash for parents; $10 gift card for adolescents).

Equipment and Measures

Virtual reality environment

The virtual reality pedestrian environment replicates a 2-lane bidirectional mid-block street crossing near a local elementary school. It was validated as an accurate measure of actual pedestrian behavior in a real world environment.²⁸

To use the virtual environment, participants first walked along a 25-foot distance in a hallway 5 times. This provided average walking speed that was used for the virtual avatar. Participants then watched the experimenter complete two crossing trials, one successful and one in which the experimenter was purposely struck by a vehicle to reduce curiosity. Next, participants performed 8 practice trials in the virtual reality environment. These trials reduced learning effects. For each trial, participants stood on a wooden “curb,” immersed inside three screens simulating the street environment, and watched vehicles pass from both directions. After deciding it was safe to cross, participants stepped down off the curb. The step triggered an avatar to cross the street at the pre-determined walking speed. If the crossing was safe, a character provided positive feedback and the next trial was initiated. If the crossing was a “close call” (defined as being within 1 second of a collision), cautionary feedback was provided prior to initiation of the next trial. If the virtual avatar was struck by a vehicle, the screen froze, cautionary feedback was provided, and then the next trial was initiated.

Participants completed 25 experimental crossing trials during each visit (that is, 25 trials while sleep restricted and 25 trials while adequately rested). Vehicle traffic moved at 30 miles per hour and appeared at randomized intervals with an average density of 12 vehicles per minute.

Four indicators of safe pedestrian street crossing, adapted from previous research,²⁹ assessed safety of pedestrian crossings in the virtual environment: (1) average start delay (time in seconds after a car passed and before participants initiated crossing); (2) average time to contact (the smallest temporal gap, in seconds, between the avatar and any oncoming vehicle during the crossing); (3) hits/close calls (when participants would have been struck by a vehicle in the real environment, or when the gap between the participants and the oncoming vehicle was less than 1 second); and (4) attention to traffic (times participants looked left plus looked right before crossing, divided by wait time).

For analytic purposes, the start delay, time to contact, and attention to traffic measures were averaged across the 25 street crossings for each condition (sleep restricted and adequate sleep). The hits/close calls pedestrian behavior was summed over the 25 trials. Thus, each participant yielded 4 pedestrian behavior scores (start delay, time to contact, hits/close calls, attention to traffic) in each of 2 conditions (sleep restricted, adequate sleep).

Actigraphy and sleep diaries

Participants wore an Actiwatch 2 actigraph by Respironics on the non-dominant wrist for the full 2-week experiment. Actigraphs are small, wristwatch-sized devices containing a piezoelectric accelerometer that yields estimates of sleep and wake patterns over time. Previous work demonstrates validity of actigraphy data. For example, correlations of sleep duration data measurements using actigraphy and sleep duration data using traditional polysomnography are >.80.^30,31

In this study, data downloaded from the actigraph were analyzed to assess total sleep duration for each night. This was calculated by the actual minutes of scored sleep, excluding any awakenings after sleep onset. Most relevant, of course, was sleep duration the night prior to the laboratory visits. We accepted as valid a range of one hour more or less than the desired sleep durations of 4 and 8.5 hours for the sleep restricted and adequate sleep conditions, respectively. Across the full sample, participants slept a mean of 4.13 hours (SD = 0.67, range = 3.16-4.97) in the 4-hour condition and a mean of 8.28 hours (SD = 0.52, range = 7.51-9.43) in the 8.5 hour condition. In instances when participants arrived for a session having slept over one hour more or less than the required sleep amount, their appointment was rescheduled. About 17% of the sample was rescheduled for this reason one time due to inadequate sleep, and it was not necessary to reschedule any participants more than once.

Actigraphy data also assessed nightly total sleep duration for the three nights prior to the target night for each condition. Because chronic sleep deprivation is known to influence cognition, average total sleep duration served as a covariate of the primary analyses. This was measured by averaging the total sleep duration over the three nights prior to the target night for each condition. The three-night window was chosen to permit full “recovery” from the sleep restriction/adequate sleep night that had occurred about 7 nights prior. Across the full sample, sleep duration prior to the sleep restricted condition averaged 7.61 hours (SD = 1.25) and 7.25 hours (SD = 1.09) prior to the adequate rest condition. The two rates were not statistically different.

As a back-up in case the actigraphy equipment failed, adolescents completed sleep diaries every morning throughout the study. The sleep diary was brief, simply asking adolescents to record when they fell asleep and woke up daily. Research suggests 28% of actigraph data are lost on average, primarily for technical problems;³² in this study, 18.9% of data were lost due to battery failure. In those cases, total sleep duration recorded on the sleep diary was used instead. In instances when actigraph data were valid, actigraph data correlated well with diary reports (r (88) = .96, p < 0.001), providing evidence of diary data validity.

Questionnaires

Parents completed a demographics questionnaire and the Pediatric Sleep Questionnaire during the baseline visit. Adolescents completed the Pediatric Daytime Sleepiness Scale. The demographics questionnaire assessed adolescents' age, gender, and race/ethnicity, and household income. The Pediatric Sleep Questionnaire (PSQ) is a 22-item instrument that assesses sleep-related breathing disorders ranging from a score of 0 to 22 (higher scores indicate greater disorder). All items are dichotomous (yes/no) and internal reliability data are adequate (Cronbach's alphas = .89).³³ Adolescents completed the Pediatric Daytime Sleepiness Scale (PDSS) to obtain a level of daytime sleepiness.³⁴ The PDSS is an 8-item instrument. Items are answered on a 5-point scale ranging from Always (4) to Never (0), yielding scores ranging from 0-32, with higher scores indicating greater daytime sleepiness. Internal reliability is adequate (Cronbach's alpha = .80).³⁴

Results

Descriptive statistics were examined first. The Shapiro-Wilk test suggested the start delay, hits/close calls and attention to traffic variables were non-normal in both conditions, so they were transformed using square root transformations. Outliers were removed ±2 SD from the mean for all variables. See Table 1 for descriptive statistics.

Table 1.

Descriptive Statistics.

Variable	M (SD) (sleep restricted condition)	M (SD) (adequate rest condition)	df	t	Cohen's d
Start Delay (mean, in sec)	1.22 (±.36)	1.07 (±.38)	47	2.68*	.36
Time to Contact (mean, in sec)	4.58 (±.92)	4.91 (±.96)	49	−2.38*	.35
Hits/Close Calls (number out of 25 crossings)	2.22 (±2.15)	1.42 (±1.47)	51	2.17*	.32
Attention to Traffic (Looks/Wait Time)	.51 (±.15)	.46 (±.15)	50	2.08*	.31

^*p<.05

To test our primary hypothesis that adolescents would engage in riskier pedestrian behaviors when sleep-restricted, repeated measures t tests were computed for each of the four pedestrian variables. No covariates were included. As hypothesized, adolescents took more time to initiate crossings (M = 1.22 (SD = 2.15) when sleep restricted; M = 1.42 (SD = 1.47) when adequately rested; t(47) = 2.68, p < 0.05), crossed with less time to contact between vehicles (M = 4.58 (SD = 0.92) when sleep restricted; M = 4.91 (SD = 0.96) when adequately rested; t(49) = −2.38, p < 0.05), experienced more hits/close calls (M = 2.22 (SD = 0.36) when sleep restricted; M = 1.42 (SD = 0.38) when adequately rested; t(51) = 2.17, p < 0.05), and looked left and right more often before crossing (M = 0.51 (SD = 0.15) when sleep restricted; M = 0.46 (SD = 0.15) when adequately rested; t(50) = 2.08, p < .05) in the sleep restricted condition versus the adequate sleep condition (Table 1).

Given results suggesting sleep restriction impacted pedestrian behavior, we conducted further analyses to test whether various factors may influence the effect of sleep restriction on pedestrian behavior. The outcome measures for these analyses were difference scores, computed by subtracting pedestrian behavior following adequate sleep from pedestrian behavior following restricted sleep. Thus, significant findings would indicate differential effects of acute sleep restriction on subsamples of interest.

We first considered randomized experimental order (sleep restriction first vs. second). Randomized order served as an independent variable in a series of independent-sample t tests, with the four difference scores (start delay, time to contact, hits/close calls, attention to traffic) as dependent variables. No statistically significant effects emerged.

Second, we considered whether age (M = 14.89 years, SD=0.62), the average of the two sleep durations over the three nights prior to each condition (M = 7.43 hours, SD = 1.17), PSQ scores (M = 2.87, SD = 2.77) or PDSS scores (M = 15.07, SD = 4.57) influenced the effect of sleep restriction on pedestrian behavior. Bivariate Pearson correlations were performed between these variables and pedestrian behavior difference scores, and no statistically significant correlations emerged. Similarly, to examine race/ethnicity and gender, two sets of independent samples t tests were performed with race (Caucasian, n = 26, vs. African American, n = 29) and with gender (boys, n = 23 vs. girls, n = 32) as the independent variables and pedestrian behavior difference scores as the dependent variable. No statistically significant effects emerged. This set of bivariate analyses was then extended using multivariate linear regression. Each of the four pedestrian behavior difference scores was used as a dependent variable in a model with gender, race, average sleep duration over the three nights prior to each condition, PDSS score, and PSQ score as independent variables. No models were significant (See Table 2).

Table 2. Summary of Linear Regression Analysis for Pedestrian Difference Scores (N = 55).

Start Delay	B	SE B	B	t	p
Gender	−.09	.07	−.24	−1.37	.18
Race	−.05	.06	−.13	−.78	.44
Prior 3 Nights' Sleep	.00	.00	−.03	−.21	.84
PDSS	.00	.01	−.03	−.18	.86
PSQ	−.01	.01	−.03	−.21	.28
Time to Contact
Gender	.13	.31	.07	.43	.67
Race	−.24	.30	−.13	−.78	.44
Prior 3 Nights' Sleep	.00	.00	.14	.87	.39
PDSS	.02	.03	.12	.74	.46
PSQ	.03	.06	.08	.48	.63
Hits/Close Calls
Gender	−.31	.28	−.18	−1.08	.29
Race	.36	.27	.22	1.32	.19
Prior 3 Nights' Sleep	.00	.00	−.08	−.49	.63
PDSS	.00	.03	−.01	−.07	.94
PSQ	−.03	.05	−.10	−.61	.54
Attention to Traffic
Gender	.19	.11	.32	1.82	.93
Race	.01	.10	.01	.06	.32
Prior 3 Nights' Sleep	.00	.00	.12	.71	.59
PDSS	−.02	.01	−.25	−1.50	.40
PSQ	.03	.02	.26	1.55	.26

Discussion

Sleep restriction influenced adolescents' pedestrian behavior. When fatigued, adolescents had greater temporal delays prior to entering a crossing gap to initiate street-crossing, left less time between themselves and oncoming vehicles, experienced more hits/close calls with virtual vehicles and attended less carefully to traffic before crossing. Stated numerically, adolescents who were sleep restricted – who had slept less than 5 hours the night before – had over a 50% increase in hits/close calls with virtual vehicles while sleep restricted compared than when they were adequately rested. While sleep restricted, adolescents experienced a hit or close call on 2.22 of the 25 simulated crossings, or 8.9% of the time, whereas they experienced a hit or close call on 1.42 of the 25 crossings (5.6%) after an adequate night of sleep. Results did not vary across any of the covariates tested, including randomized experimental condition order, average sleep duration prior to each condition, baseline measures of self-reported sleep quality, age, gender or race/ethnicity.

The generally riskier pedestrian behaviors among sleep-restricted adolescents were likely caused by the cognitive effects of sleep restriction.^16-19 Safe pedestrian behavior requires multiple aspects of cognitive function. First, safe pedestrian behavior requires impulse control; we found sleep-restricted adolescents made risky decisions by crossing in potentially dangerous traffic gaps. Second, safe pedestrian behavior requires efficient, rapid, and precise decision-making; a pedestrian must perceive and judge the safety of a traffic gap very quickly and then initiate movement into it. As measured by the start delay variable, adolescents who were tired also performed poorly at quickly initiating the crossing when a safe opportunity was present.

Finally, safe pedestrian behavior requires attention to oncoming traffic. ³⁵ It appears that the tired adolescents looked more frequently at traffic but made risky and poor decisions, yielding more risky behavior overall when acutely sleep restricted. Adolescents may have increased their looking at traffic while acutely sleep restricted as a means of compensation. They may have recognized their deficient processing and tried to compensate by looking more frequently at the pedestrian environment. Simply looking at the environment does not translate into safer pedestrian behavior however, as a safe pedestrian must not only perceive the environment but also process the perceived information and decide how to act. Given our findings on the cognitive and decision-making aspects of pedestrian behavior that we measured, it appears the acutely sleep-restricted adolescents demonstrated poor cognitive processing and decision-making about the pedestrian environment despite the fact that they looked back and forth more frequently.

Our results may have implications for adolescents engaging in everyday activities such as crossing streets in at least three domains. First, our results confirm the benefit of pediatricians promoting sleep health among adolescent patients. Well-patient visits typically involve counseling on preventative health topics,³⁶ and such counseling can instigate healthy behavior change.^37,38 Sleep health is often not included in such counseling, but given the present results, the scope of unintentional injury as the leading cause of adolescent death in the US, and the other consequences of inadequate sleep (e.g., on school performance), ^{39- 41} it may be worthwhile for pediatricians to consider including sleep health among topics addressed during preventative counseling discussions with adolescent patients and their parents.

Second, this information could educate policy decisions about school start times.⁴² Changes in the circadian rhythms of adolescents cause many teens to fall asleep later, and early school start times prevent them from achieving adequate amounts of sleep.^13-15 Previous work suggests inadequate sleep leads to decreased academic performance,^39-41 and later school start times are associated with longer sleep duration in adolescents.⁴³ Our findings offer initial data of another benefit that might arise from later school start times: reduced pedestrian injury risk among adolescents walking to and from school.

Finally, the results might have implications for parents concerning adolescent bedtime enforcement. Parent-adolescent conflict at bedtime is common. Adolescents are more likely to feel controlled by parents if they make decisions such as bedtimes, for them,⁴⁴ and in a study of family conflict, parents and adolescents identified bedtimes as an issue causing conflict.⁴⁵ Research suggests most parents do not understand developmental sleep patterns such as the shift in adolescents' circadian rhythms, ⁴⁶ and adolescents also have low knowledge of sleep and sleep hygiene themselves.⁴⁷ However, appropriate parental-set bedtimes are associated with longer sleep duration in adolescents⁴³, behavior that may help adolescents in multiple domains of functioning, including pedestrian safety.

This experiment had strengths and limitations. Methodological strengths include the use of actigraphy to measure sleep and a virtual reality environment to assess pedestrian behavior. Another strength is the repeated measures research design, which minimized measurement error across samples. One limitation is that it studied only one developmental stage (14-15 year olds), and the results may not generalize to other age cohorts. Another limitation was the omission of other sleep-restriction categories. We studied only 4 versus 8.5 hours of sleep and it is unknown how adolescents might function with 6 hours of sleep, 7.5 hours of sleep, 2 hours of sleep, or with other sleep accumulations or debts. Finally, we relied on a somewhat artificial laboratory design. Sleep was restricted for an experiment, not for “natural” reasons; similarly, pedestrian behavior was measured in the ethically-sound but not perfectly realistic virtual environment. Future research might seek research designs with greater ecological validity.

In conclusion, safe pedestrian behavior requires precise cognitive functioning, and human cognition is compromised by sleep restriction. This study demonstrated the risk of acute sleep restriction on multiple aspects of adolescents' pedestrian safety.

Implications and Contribution.

The results have implications in at least three areas. They may 1) confirm the benefit for pediatricians to address sleep health during preventative counseling discussions with adolescents and parents; 2) inform policymakers to delay school start times, providing safer travel to school; and 3) influence parents to enforce appropriate bedtimes.