Distracted mobile device use among street-crossing college student pedestrians: an observational approach

Abstract

Objective:

To estimate the incidence of mobile device use among street-crossing pedestrians and explore differences by sex and intersection type at a large public South-eastern university in the United States.

Participants:

All instances of campus pedestrians crossing the street during the observation period (N = 4,878).

Methods:

Video recordings of crosswalk activity at four locations were analyzed for pedestrian use of a mobile device while crossing.

Results:

Device use while crossing was observed 1,201 (24.6%) times. Of male crossing instances, 277 (16.8%) were coded as using a device. Of female instances, 924 (28.6%) were coded as using a device. Differences in device use while crossing were found between sexes and some intersection types.

Conclusions:

This study estimates mobile device use while crossing the street and suggests differences by sex and intersection type. Future research should focus on improving understanding of the problem and evaluation of interventions to address the issue.

Introduction

In the U.S., unintentional injury is the third leading cause of death overall and is the leading cause for persons aged 1–44 years.¹ Regarding death from unintentional injury, motor vehicle traffic crashes are the leading cause of death for persons aged 5 through 24 years.² Recent data show that pedestrian fatalities have been consistently increasing since 2009.³ According to the National Highway Traffic Safety Administration (NHTSA), 5,977 pedestrians were killed in motor vehicle traffic crashes in 2017.³ For the same year, an estimated 137,189 pedestrians were treated in the emergency department for non-fatal traffic crash-related injuries.⁴ The economic cost of pedestrian fatalities in 2010 is estimated to be over $4.85 billion.⁵ In 2010, nonfatal pedestrian injuries of those hospitalized and those treated and released from the emergency department are estimated to have cost $6.75 billion and $1.06 billion respectively.⁵ Another trend can be illustrated by the percentage of pedestrian deaths expressed as a percentage of total traffic fatalities. In 2008, pedestrian fatalities in traffic crashes made up 12% of the total 37,423 traffic fatalities. Since 2008, the number of pedestrian fatalities in traffic crashes has steadily increased and, in 2017, made up 16% of the 37,133 total motor vehicle traffic fatalities.³

While traffic-related incidents have been a primary focus of distracted pedestrian research, it is important to consider the potential impact of other injuries that can occur because of distracted walking. Through secondary analysis of U.S. emergency department data obtained from the National Electronic Injury Surveillance System (NEISS), Nasar and Troyer⁶ estimated that 1,506 nonfatal injuries occurred in 2010 due to mobile phone use among pedestrians in public places; an estimate which includes traffic-related and non-traffic-related events (e.g., a pedestrian collides with a tree). Nasar and Troyer⁶ discovered that the frequency of such injuries increased with statistical significance between 2004 and 2010. Injuries included concussions, seizures, fractures, contusions, lacerations, dislocations, abrasions, sprains, strains, and pain in various parts of the body. Overall, more injuries related to talking than texting with talking accounting for 69.5% of injuries and texting accounting for 9.1%. Pedestrian injuries more frequently involved people under the age of 31 (54.7%) and more men (52.9%) than women. A similar study with expanded criteria that included injuries in settings beyond public spaces found an additional 280 estimated telephone-related cases in the year 2010 while reporting a statistically significant increase during a study period of 2000 to 2011.⁷ The authors of both studies argue that the true number of non-traffic-related pedestrian injuries are likely higher due to underreporting.^6,7

Research suggests that conditions of distraction, including mobile device use, have a negative impact on gait^8–18 and are linked to an increase in unsafe crossing behaviors.^{8,14–16,18–21} Examples of such unsafe crossing behaviors include failure to look both ways before crossing, failure to wait for traffic to stop before crossing, and not looking at traffic while crossing. A lack of situational awareness has also been linked to pedestrian distraction.^9,20 In one demonstration of this effect, Hyman et al⁹ found that distracted pedestrians crossing a central university plaza failed to notice a brightly-colored clown riding a unicycle more often than their non-distracted peers.

College students frequenting campus may be at risk of distracted walking injury due to regular street-crossing on campus,²² high regard for text-based messaging (texting) as socially acceptable means of communication,²³ and heavy use of mobile communication technology among the traditional college student age group.^24,25 The heavy vehicular and pedestrian traffic inherent of many campus environments leads to numerous instances of vehicle-pedestrian interaction and thus provides abundant opportunities for conflict.

Although there is growing evidence to support increased risk associated with walking while distracted, there currently exists only a small body of recent literature investigating the incidence of the behavior using real-world observation.^26–28 Improved data collection is needed to move toward an understanding of the problem of distracted walking.²⁹ Regarding distracted pedestrians, formal data collection on the national level does not occur. At the state level, there is a lack of uniformity among reporting systems which makes in-depth analysis difficult.³⁰ The most recent studies^26,27 have used video recordings to capture pedestrian behavior; thereby facilitating analysis. Recent technological advancements have resulted in recording devices with improved portability, higher resolution, increased frame rates, and lower cost. These advancements have increased the feasibility of incorporating video-recording devices into the data collection process.

Russo et al²⁶ performed video observations of 3,038 pedestrians at four signalized intersections (one in New York and three in Flagstaff, Arizona) to examine pedestrian behavior. From the video, trained data collectors extracted information that included pedestrian age, gender, group size, crossing time, and pedestrian distraction. Pedestrian distraction was recorded as: no distraction, talking on mobile phone, texting on mobile phone, listening to headphones, or other (non-phone distractions such as looking in a purse or reading a newspaper). In this study, 13.5% of pedestrians were deemed distracted while crossing the street. Pedestrians at the New York location were more likely to cross while talking and texting, while pedestrians at the Arizona location were more likely to be wearing headphones.

Schultz et al²⁷ explored pedestrian behavior by capturing video data of 2,061 crossing events from 15 signalized intersections in Utah. Each crossing event was categorized by pedestrian age group, gender, mobility status (regular, wheelchair, cane, pet, etc …), alertness, and distraction. Distraction included behaviors such as wearing headphones, talking on a cellphone, pedestrians texting or looking down at their cellphone, and other (non-phone distractions such as looking in a purse or reading a newspaper). Results of this study indicate 8.8% of crossing events involved distracted pedestrians.

It is important to note that the aforementioned studies only investigated signalized intersections and did not explore behavioral differences that might exist between different types of street crossings. The studies also differed in their definition of a crossing event. Russo et al²⁶ analyzed individual pedestrians while Schultz et al²⁷ defined a crossing event as one or more persons moving at the same rate across a pedestrian crosswalk; meaning that a group of crossing pedestrians was counted the same as a single pedestrian for the purpose of data analysis.

The purpose of the present study is to use video observations to provide an estimate of the incidence of distracted mobile device use among street-crossing pedestrians while also examining differences in pedestrian behavior by sex and across different types of intersections at a large, South-eastern university in the United States.

Materials and methods

Prior to performing the procedures described herein, Institutional Review Board (IRB) approval was obtained. Data were collected during the fall 2016 semester. The researcher utilized video recording equipment to facilitate data collection and analysis. The process of recording video involved the use of a magnetic mount to temporarily secure a recording device to existing traffic infrastructure. Two GoPro^® brand recording devices were used to capture video data in the present study; A GoPro^® HERO3+ Black Edition set to record at a resolution of 4096 by 2160 with a frame rate of 12 frames per second (fps) and a GoPro^® HERO5 Session set to record at a resolution of 3840 by 2160 with a frame rate of 30 fps. Each recording device’s field of view was positioned toward the observation location and left stationary to record for one hour.

Four locations that differ mainly by method of traffic control were chosen for observation. The first selected location was an intersection largely controlled by lighted signals (a “signalized” location). The second location was a mid-block crossing that was not part of an intersection. Vehicle and pedestrian traffic at this location were controlled by lighted signals and was, therefore, considered a “signalized mid-block” location. The third location was a group of crosswalks at a 3-way intersection controlled by stop signs (an “un-signalized” location). The fourth location was a mid-block crossing that was not part of an intersection. Vehicle and pedestrian traffic at this un-signalized mid-block location were controlled only by “yield to pedestrian” signage. All locations had a speed limit of 25 miles per hour and vehicular traffic consisted of passenger vehicles, mass transit busses, golf carts, and campus facilities vehicles. Although locations were not explicitly selected based on their vehicular traffic volume, the total number of vehicles crossing the intersection during the one-hour observation period was counted and is reported in Table 1. Traffic composition was majority passenger vehicles. University mass transit busses and university maintenance trucks were also present in smaller numbers.

Table 1.

Frequencies and percentages of male and female pedestrian crossing instances at observed locations during observation period (N = 4,878).

	Crossing instances		% total of each location
Location (pedestrian crossing instances; vehicular traffic volume)	Male	Female	Male	Female
Signalized intersection (2,299; 1,032)	618	1,681	26.9	73.1
Signalized mid-block (543; 792)	86	457	15.8	84.2
Un-signalized intersection (1,411; 564)	688	723	48.8	51.2
Un-signalized mid-block (625; 354)	260	365	41.6	58.4

Note. Signalized intersection = A four-way intersection with vehicle and pedestrian traffic controlled by lighted signals. Signalized mid-block = A single mid-block crosswalk with vehicle and pedestrian traffic controlled by lighted signals. Un-signalized intersection = A three-way intersection with only vehicle traffic controlled by stop signs. Un-signalized mid-block = A single mid-block crosswalk with only vehicle traffic controlled by passive “yield to pedestrian” signage.

At each location, the hour of observation began 30 minutes prior to a scheduled class release time and continued 30 minutes past the scheduled class release time. For example, a one-hour observation that took place from 11:45 AM to 12:45 PM covered the time before, during, and after the class release time of 12:15 PM. This timing scheme was chosen because it was expected that pedestrian traffic would be heaviest during times between consecutive classes.

Coding

Coding of observation data followed a modified version of an electronic device use observation protocol used as part of the National Occupant Protection Use Survey.³¹ This protocol has been used to observe electronic device use among motor vehicle drivers and was modified for use in the present study. Consistent with the existing observation protocol, participants were considered to be “using a mobile device” if they exhibited any of the following while crossing: (1) holding a device to their ears, (2) speaking with visible headsets on, or (3) visibly manipulating a handheld device. While the intended targets of observation throughout this study were the crosswalks at each observation location, pedestrians captured within the frame of the video camera but outside the physical boundaries of the crosswalk markings were still coded and analyzed. Thus, for simplicity, reference to participants in the “crosswalk” or “crossing” includes a small number of participants crossing outside of crosswalk markings.

Rather than coding and reporting the incidence of distracted mobile device use in terms of number of pedestrians observed engaging in the behavior, all coding and reporting was performed in terms of crossing instances where the behavior of distracted mobile device use was observed. For example, the same pedestrian at a four-way intersection crossing a road running East and West, then crossing the perpendicular road running North and South was coded as two separate crossing “cases” for the purpose of data analyses and reporting. Though a rare occurrence, pedestrians crossing an intersection diagonally and thus leaving the first crosswalk before reaching the other side of the road were also coded as two crossing cases. To differentiate case 1 from case 2 for these pedestrians, an imaginary bisecting line was created by the coder that extended from the corner of two adjoining crosswalks at a 45-degree angle. The bisecting line served as the point where one case ended and the other began. Each case was coded independently for the behavior of distracted mobile device use.

Coding in terms of crossing instances instead of number of pedestrians was necessary to address a limitation introduced by attempting to keep track of unique pedestrians. It is possible that a pedestrian crossed the same street more than once during an observation period. For example, a pedestrian could cross the street on her way to a 10-minute meeting, then cross the same intersection as she leaves the meeting. In such scenarios, it would be difficult for a coder to identify every repeat crosser and exclude them from further analyses.

All pedestrians were coded as either male or female based on a combination of such contextual factors as clothing, hairstyle, and body figure. Pedestrians were classified as holding a device to their ear if they were seen holding a device to their ear. Activities in this category included speaking, listening to messages, or conducting voice-activated dialing with a device held to the ear.

Pedestrians were coded as speaking with visible headsets on if they appeared to be speaking and wearing headsets with microphones. Activities in this category include engaging in conversation or conducting voice-activated dialing via a wireless earpiece on the ear or via an ear bud connected by wire to a mobile device. Talking via a visible Bluetooth headset is also an included activity.

Pedestrians were coded as visibly manipulating a handheld device if they appeared to be manipulating some type of electronic device such as a cell phone, a smart phone, personal digital assistant (PDA), MP3 player, video game, or some other device. Activities in this category included text messaging, using a Web-capable smart phone, or a PDA to view travel directions, checking e-mails or calendar appointments, or surfing the Internet. Also included were manual dialing, playing handheld games, and holding phones in front of their face to converse or check messages via speakerphone or use voice-activated dialing. Given the passive nature of the headphone use/listening to music, evidence from studies that the activity tends to have a different or no effect on gait and unsafe crossing behavior,^9,18,21,32 and the fact that the behavior is not included as a criterion of distraction in the National Occupant Protection Use Survey,³¹ headphone use was not considered distracted mobile device use for the purpose of the present study.

Both rater-interrater and intra-rater reliability analyses were performed during analysis of observational data to provide an indicator of protocol reliability. Rater-interrater reliability is a measure of agreement between two or more observers making independent coding judgements at the same time.³³ The procedure for determining inter-rater reliability involved a second researcher coding a randomly selected video segment across which 100 pedestrians traversed the trafficway. Inter-rater reliability was assessed by calculating a Cohen’s kappa coefficient.³⁴

Intra-rater reliability analyses provide information on the extent of agreement of a single researcher’s observations over time.^33,35 The procedures for intra-rater reliability were identical to those for rater-interrater reliability except that the researcher’s initial coding was compared to his own coding 21 days later. The time period of 21 days between coding sessions was consistent with literature describing video reliability analyses.³⁶

Results from the rater-interrater reliability analyses indicated perfect agreement between the two researchers, κ = 1.00, p < .001. Results from the intra-rater reliability analyses indicated perfect agreement between the researcher’s coding at time 1 and the researcher’s coding 21 days later at time 2, κ = 1.00, p < .001.

Data analyses

Frequencies and percentages were calculated for each observation location to provide an estimated incidence of distracted mobile device use among street-crossing pedestrians. Binary logistic regression analyses were utilized to test for differences in observed street-crossing while using a mobile device between observation locations. Chi-square analyses were utilized to test for differences in observed street-crossing while using a mobile device between males and females. Statistical procedures were conducted using International Business Machines (IBM^®) Statistical Package for Social Sciences (SPSS), version 23.0.

Results

Observation sessions across all locations yielded a total of 4,878 crossing instances. In total, 1,652 (33.9%) of the crossing instances were male and 3,226 (66.1%) were female. Pedestrian crossing instances were observed at: (1) the signalized intersection (n = 2,299); (2) the signalized mid-block crosswalk (n = 543); (3) the un-signalized intersection (n = 1,411); and (4) the un-signalized mid-block crossing (n = 625). Frequencies and percentages of crossing instances as well as traffic volume are summarized in Table 1. Across all observation locations, device use while crossing was observed 1,201 (24.6%) times. Of all male crossing instances, 277 (16.8%) were coded as using a device. Of all female crossing instances, 924 (28.6%) were coded as using a device.

Results of the chi-square analyses indicate that of the 1,652 male crossing instances, 277 (16.8%) instances of distracted mobile device use were observed. Of the 3,226 female crossing instances, 924 (28.6%) instances of distracted mobile device use were observed. This association between gender and device use while crossing was statistically significant as assessed by Fisher’s exact test, p < .001, odds ratio = .50, 95% CI [.43, .58]. With males set as the reference group, females were almost twice as likely to be observed using a device while crossing the street, odds ratio = 1.99, 95% CI [1.72, 2.32]. Table 2 provides frequencies and percentages of observed distracted mobile device use while crossing.

Table 2.

Frequencies and percentages of distracted mobile device use crossing instances among males (n = 1,652) and females (n = 3,226) overall and by observation location.

Device use	Male	Female	% of male*	% of female*
Yes (n = 1,201)	277	924	16.8	28.6
Signalized intersection (n = 585)	109	476	17.6	29.4
Signalized mid-block (n = 177)	16	161	18.6	35.2
Un-signalized intersection (n = 308)	119	189	17.3	26.1
Un-signalized mid-block (n = 131)	33	98	12.7	26.8

Note.

^*Percentage distracted out of total male and female crossing instances at each observation location; Signalized intersection = A four-way intersection with vehicle and pedestrian traffic controlled by lighted signals. Signalized mid-block = A single mid-block crosswalk with vehicle and pedestrian traffic controlled by lighted signals. Un-signalized intersection = A three-way intersection with only vehicle traffic controlled by stop signs. Un-signalized mid-block = A single mid-block crosswalk with only vehicle traffic controlled by passive “yield to pedestrian” signage.

Bivariate logistic regression analyses were employed to test for a relationship between pedestrians engaged in distracted mobile device use while crossing the street on campus and observation location. Results indicated a statistically significant relationship, χ2(3) = 29.079, p < .001.

Since some observation locations had a high percentage of female crossing instances compared to male and significantly more females were observed using a mobile device, loglinear analyses were utilized to assess the possible confounding influence of gender on frequency of device use. Results from loglinear analyses indicated that there was not a significant 3-way interaction between observation location, distracted mobile device use, and gender, p = .061. Further, after controlling for gender, a significant difference between observation location and device use remained χ2(3) = 8.866, p = .031.

To explore the influence of gender on the results of logistic regression analyses, gender was included in an additional logistic regression model. Results from this model indicated a statistically significant relationship, χ2(4) = 99.975, p < .001. Across all comparisons, the signalized mid-block location emerged as significantly different (p < .01). Odds ratios and 95% confidence intervals for the locations are shown in Table 3.

Table 3.

Logistic Regression Predicting Likelihood of Distracted Mobile Device Use Based on Observation Location and Gender.

							95% CI for odds ratio
Location	B	SE	Wald	df	p	Odds ratio	Upper	Lower
Signalized intersection *	-	-	-	-	-	-	-	-
Signalized mid-block	.287	.104	7.616	1	.006	1.333	1.087	1.634
Un-signalized intersection	−.069	.082	.700	1	.403	.933	.794	1.097
Un-signalized mid-block	−.167	.111	2.269	1	.132	.847	.682	1.051
Gender	.643	.079	66.989	1	< .001	1.902	1.630	2.218
Constant	−1.565	.079	394.021	1	< .001	.209	–	–
Signalized intersection	−.287	.104	7.616	1	.006	.750	.612	.920
Signalized mid-block *	-	-	-	-	-	-	-	-
Un-signalized intersection	−.356	.115	9.640	1	.002	.700	.559	.877
Un-signalized mid-block	−.454	.136	11.114	1	.001	.635	.487	.829
Gender	.643	.079	66.989	1	< .001	1.902	1.630	2.218
Constant	−.726	.092	62.966	1	< .001	.484	–	–
Signalized intersection	.069	.082	.700	1	.403	1.071	.912	1.259
Signalized mid-block	.356	.115	9.640	1	.002	1.428	1.140	1.788
Un-signalized intersection *	-	-	-	-	-	-	-	-
Un-signalized mid-block	−.098	.119	.678	1	.410	.907	.719	1.144
Gender	.643	.079	66.989	1	< .001	1.902	1.630	2.218
Constant	−1.634	.080	412.941	1	< .001	.195	–	–
Signalized intersection	.167	.111	2.269	1	.132	1.181	.951	1.467
Signalized mid-block	.454	.136	11.114	1	.001	1.574	1.206	2.055
Un-signalized intersection	.098	.119	.678	1	.410	1.103	.874	1.391
Un-signalized mid-block *	-	-	-	-	-	-	-	-
Gender	.643	.079	66.989	1	< .001	1.902	1.630	2.218
Constant	−1.732	.112	237.836	1	< .001	.177	–	–

Note.

^*For each set of comparisons, the reference group has been bolded. Signalized intersection = A four-way intersection with vehicle and pedestrian traffic controlled by lighted signals. Signalized mid-block = A single mid-block crosswalk with vehicle and pedestrian traffic controlled by lighted signals. Un-signalized intersection = A three-way intersection with only vehicle traffic controlled by stop signs. Un-signalized mid-block = A single mid-block crosswalk with only vehicle traffic controlled by passive “yield to pedestrian” signage.

Discussion

In the United States, pedestrian injuries are an increasing public health concern and use of mobile devices while walking has been implicated as a contributor to the problem of pedestrian injuries. The purpose of this study was to estimate the incidence of distracted mobile device use on campus among street-crossing college student pedestrians and explore differences by sex and intersection type. Also emerging from the present study is an observation protocol informed by methods used in the NHTSA National Occupant Protection Use Survey (NOPUS). The NOPUS is NHTSA’s primary method of estimating cell phone use by automobile drivers.³¹ Despite the use of the word “survey” in its title, its administration relies solely upon unobtrusive observation.

In the present study, observation sessions across all locations yielded a total of 4,878 crossing instances. In total, 33.9% of the crossing instances were male and 66.1% were female. Across all observation locations, device use while crossing was observed 24.6% of the time. Of all male crossing instances, 16.8% were coded as using a device. Of all female crossing instances, 28.6% were coded as using a device. Consequently, it is estimated that female pedestrians on campus use mobile devices while crossing the street almost twice as much as their male counterparts. Further, it is estimated that nearly one quarter of all pedestrians on campus use a mobile device while crossing the street.

Comparison of the present results to recent observation studies is limited given differences in operationalization of the term “distracted,” differences in observation methods, and differences in time periods from which data were collected. All studies discussed in this paragraph considered headphone use/listening to music a distraction. The authors of the present study decided to omit cases of headphone use/listening to music in their analyses given the passive nature of the activity and evidence from studies that the activity tends to have a different or no effect on gait and unsafe crossing behavior.^9,18,21,32

Despite these limitations, an observation of 1,102 pedestrians conducted in 2012 in Seattle, Washington reported that 29.8% of participants were distracted while crossing the street. Live coding was employed in that study.¹⁸ Study data collected in 2015 from Norfolk, Virginia and Birmingham, Alabama revealed distraction in 41.2% of cases. Live coding was also used in that study.²⁸ The two previously mentioned studies to analyze recorded video reported distraction in 13.5% of cases (data collected in 2017)²⁶ and 8.8% of cases (data collected in 2018).²⁷ In both of these studies, headphone use while crossing was considered to be distracted crossing behavior.

In the present study, proportions of distracted mobile device use while crossing differed between male and female crossing instances. A significantly higher proportion of female crossing instances involved mobile device use compared to their male counterparts (28.6% versus 16.8%). Statistical analyses revealed that female crossing instances were almost twice as likely to involve mobile device use. One reason for a difference between males and females might be the tendency for females to initiate contact with friends and family more often than males.³⁷ Despite difficulty in providing an explanation of the gender differences in the present study, an overall observation made by the researcher during coding is worth noting. It appeared that many of the female pedestrians did not have pockets as a feature of their clothing, while many more male pedestrians did. One area of future research investigating gender differences could include exploration of clothing differences and whether the absence of pockets makes a person more likely to hold their device in their hand; and whether holding a device in a person’s hand due to lack of convenient storage (whether pockets, handbags, or similar types of storage) makes them more likely to use the device.

After controlling for the influence of gender on distraction at each observation location, differences were found among the locations. Comparison of all locations revealed one location that differed significantly from the rest. One characteristic that may be important in attempting to explain the observed difference between observation locations is the type of traffic control at the location. Results were such that the signalized mid-block location was significantly more likely to have device use when compared to all other locations. The greatest magnitude difference was found between the signalized mid-block location and the un-signalized mid-block location. The signalized mid-block location consisted of only one crosswalk that crossed two lanes of bi-directional traffic and both pedestrian and vehicular traffic were controlled by lighted signals. Thus, drivers at this location were commanded to stop by a traffic signal before pedestrians were advised to cross via a crosswalk signal. Similar to the signalized mid-block location, the un-signalized mid-block location featured only one crosswalk that crossed two lanes of bi-directional traffic, though vehicular traffic was controlled by a passive “yield to pedestrians” sign and no crossing instruction was provided to pedestrians. Further, despite characteristic similarities between the two aforementioned locations, statistical analyses revealed that the un-signalized mid-block location was not significantly different from the others. Though further exploration is required, this suggests that the presence of lighted traffic control signals and simplified traffic flow may increase the tendency to cross while distracted by a mobile device. It may be that pedestrians at such locations perceive a lower risk due to the increased traffic control afforded by lighted signals and simpler bi-directional traffic patterns.

Future research

The university setting, with its high concentration of pedestrians as well as a younger population that is more likely to take risks, has been identified as an important setting in which pedestrian safety interventions can be delivered.³⁸ Pedestrian safety campaigns have been attempted on some college campuses,^39,40 though formal and rigorous evaluation is often lacking.³⁸ Evaluated interventions that are known to exist are often behavioral in nature.^40,41 One week-long intervention on a college campus involved a virtual reality experience as a means of raising awareness of the dangers of distracted street-crossing among students.⁴⁰ While the authors considered the reach of the intervention to be broad, they found that the intervention failed to produce meaningful behavior changes. Another intervention study not on a college campus used stenciled text at crosswalks that read, “Heads Up, Phones Down.”⁴¹ The authors of that study found a reduction of some pedestrian behaviors and an increase in others. Other, non-behavioral interventions have recently been proposed, such as improving accuracy of wearable devices to detect when a pedestrian might be crossing the street while distracted so intervention strategies can be delivered while the behavior is occurring.⁴² Another proposal suggests that Bluetooth beacons placed at intersections frequented by pedestrians may be effective in reducing distracted pedestrian behavior.⁴³ As pedestrians approach a crossing, these beacons are designed to send messages to their mobile devices that remind them to put their phones down and focus on the crossing environment. Future research should continue to explore effective intervention strategies.

We also advocate for standardized evaluation methods that are replicable. The present study offers a data collection method that relies on high-quality video data and is based on a nationally implemented driver observation protocol that has been used in its current form since 2005 to collect data on driver electronic device use.³¹ An effort to create standardized data collection procedures will: (1) allow for a better understanding of the risk associated with device use while crossing the street and (2) allow for assessment of intervention effectiveness in achieving long-term outcomes.²⁹ Though a significant proportion of pedestrian crossings result in injury or death, most pedestrian crossings are uneventful. This makes the chances of observing even one vehicle-pedestrian crash at sampled crossing locations unlikely; making statistical analyses of risk difficult. To alleviate this, future research can focus on traffic conflicts (events where a pedestrian or driver had to perform a mild-to-severe maneuver to avoid a crash) as a proxy measure of actual crashes.^44,45

Future research should also focus on understanding which distracted walking countermeasures will be most effective given the target population demographic and crossing environment.²⁹ Research on interventions to reduce college student alcohol use have successfully used focus groups with members of the target population to inform intervention strategies.⁴⁶ Further, qualitative follow-up with members of the target population after observation could help explain any differences between gender, observation location, and previous injury exposure that are identified during future investigation. Such qualitative follow-up with intent to explain previous quantitative findings is a form of mixed methods research that is often used to help explain phenomena about which little is known.⁴⁷

Limitations

Several important limitations are important to consider when interpreting results of the present study. Though several documented pedestrian incidents take place outside of crosswalks, this investigation did not explore those crossing the street outside of crosswalks. Further, observations were delimited to daylight hours. While a number victims of documented fatal and nonfatal pedestrian injuries are under the influence of alcohol at the time of injury, it was assumed that those observed were not under the influence of alcohol. Exploration of the aforementioned factors and their association with distracted pedestrian injuries was beyond the scope of the present investigation. The cross-sectional nature of the observations made it difficult to determine whether a behavior exhibited by an observed participant was a behavior in which they engaged frequently or rarely. This introduced the possibility of a phenomenon known as regression toward the mean. For example, a person observed engaging in use of a mobile phone while crossing could have engaged in this behavior infrequently while another person observed crossing while not using a mobile phone could have engaged in this behavior very frequently. Despite this potential limitation, it was expected that any such extreme cases exerted relatively little influence given their dispersion amongst the large number of observed cases. Given that analyses of observational data in the present study relied on the researcher’s judgment of whether participants appeared to be engaging in distracted mobile device use, there was a possibility of misclassification of pedestrian behavior in the present study. To address this limitation, both rater-inter-rater and intra-rater reliability analyses were performed to provide an indicator of protocol reliability. The researcher’s determination of participant gender based on factors such as clothing, hairstyle, and body figure introduced the possibility for misclassification of gender. Although the observation locations differed by traffic control devices and traffic flow, the small sample of observation locations made it difficult to determine whether any differences between locations were due to these characteristics or whether other factors were at play. Widespread use of a standardized observation protocol would help to reduce the impact of these limitations.

Regarding the observation analyses, an important note should be made about the assumption that chi-square and logistic regression analyses be performed across independent samples. Since analyses were performed on crossing instances and not members of the population, it was possible that some of the same pedestrians could have been present at multiple observation locations; thus violating assumptions of independent samples to an undeterminable degree. Despite this limitation, observation studies with similar methods have deemed such analyses appropriate.⁴⁸

Conclusions

The present study contributes to the literature by increasing understanding of the incidence of distracted mobile device use while crossing the street. The results of this study bolster the need for further exploration in this emerging area of research. Through observations, the present study found that a sizeable proportion of observed pedestrians use a mobile device while crossing the street; a behavior that has been implicated as a risk factor for traffic-related injury. Further, differences in distracted mobile device while crossing the street were found between gender as well as by observation location. College campuses possess characteristics that make them a fitting venue for intervention and should be considered as we move to address the problem of distracted street-crossing.