Texting while driving: A study of 1211 U.S. adults with the Distracted Driving Survey

Abstract

Texting and other cell-phone related distracted driving is estimated to account for thousands of motor vehicle collisions each year but studies examining the specific cell phone reading and writing activities of drivers are limited. The objective of this study was to describe the frequency of cell-phone related distracted driving behaviors. A national, representative, anonymous panel of 1211 United States drivers was recruited in 2015 to complete the Distracted Driving Survey (DDS), an 11-item validated questionnaire examining cell phone reading and writing activities and at what speeds they occur. Higher DDS scores reflect more distraction. DDS scores were analyzed by demographic data and self-reported crash rate. Nearly 60% of respondents reported a cell phone reading or writing activity within the prior 30 days, with reading texts (48%), writing texts (33%) and viewing maps (43%) most frequently reported. Only 4.9% of respondents had enrolled in a program aimed at reducing cell phone related distracted driving. DDS scores were significantly correlated to crash rate (p < 0.0001), with every one point increase associated with an additional 7% risk of a crash (p < 0.0001). DDS scores were inversely correlated to age (p < 0.0001). The DDS demonstrated high internal consistency (Cronbach's alpha = 0.94). High rates of cell phone-related distraction are reported here in a national sample. Distraction is associated with crash rates and occurs across all age groups, but is highest in younger drivers. The DDS can be used to evaluate the impact of public health programs aimed at reducing cell-phone related distracted driving.

Highlights

• •Nearly 60% of respondents reported a cell phone reading or writing activity.
• •Reading texts (48%), writing texts (33%) and viewing maps (43%) were most frequent.
• •Only 4.9% of respondents had a program to reduce cell-related distracted driving.
• •Distracted Driving Survey Scores significantly correlated with crash rate (p < 0.0001).

1. Introduction

Texting and other cell phone use while driving is a major risk factor for motor vehicle collisions and associated injury and death (Wilson & Stimpson, 2010). In 2012, distracted driving was associated with 3300 deaths and 421,000 injuries in collisions in the US; there is evidence that smartphone use is increasingly contributing to these numbers (US Department of Transportation National Highway Traffic Safety Administration, 2014).

Simulation and instrumented vehicle studies have shown that drivers who are viewing information on or writing with cell phones have significantly increased risk of collision or near-collision events, (Yannis et al., 2014, Owens et al., 2011, Caird et al., 2014) and the problem is exacerbated in younger drivers (Caird et al., 2014, Hosking et al., 2009). Rigorous instrumented vehicle naturalistic studies have confirmed these results (Klauer et al., 2014, Olson et al., 2009).

In spite of the risk, texting and driving is widespread; among US adults 18 to 64 years old, 31% reported reading or sending text messages or emails while driving in prior last 30 days (Centers for Disease Control and Prevention (CDC), 2013). The issue is even more pronounced in younger drivers with nearly half of young drivers reporting texting in just the past 30 days (Olsen et al., 2013). Observational studies on college campuses have also confirmed high rates of texting and driving (Cook & Jones, 2011). In our prior work, 59.2% and 71.5% of young adults wrote and read text messages, respectively, while driving in the last 30 days (Bergmark et al., 2016).

The purposes of this study were to describe the frequency of cell phone related distracted driving behaviors and self-reported accident rate by relevant demographic subgroups and confirm reliability in a national sample of drivers of all ages.

2. Methods

2.1. Metrics

The cell phone focused Distracted Driving Survey (DDS, ©Massachusetts Eye and Ear, 2013, Table 1, with responses) is an 11-item validated driver-reported questionnaire assessing common cell phone reading and writing tasks, such as writing and reading text messages and email, social media site use, and GPS use (Bergmark et al., 2016). The DDS includes questions about the speeds at which drivers have completed each task in the past 30 days as well as a question about perceived risk. It has been validated among drivers 18–24 years old (Bergmark et al., 2016). A scoring algorithm is used to produce a score 0–44, with 44 being the highest risk survey result. The details of the questionnaire and scoring algorithm have been previously published (Bergmark et al., 2016).

Table 1.

Distracted Driving Survey and responses (N = 1211 drivers, 2015, responses as percentage).

Do you think that you can safely text and drive?	Always 4	Most of the time 4	Some of the time 10	Rarely 17	Never 65
	Every time I drive	Most of the times I drive	Some of the times I drive	Rarely	Never
In the last 30 days have you written text messages while driving?a	2	3	9	19	67
In the last 30 days have you read text messages while driving?a	2	5	18	24	52
In the last 30 days have you written email messages while driving?a	1	1	2	8	88
In the last 30 days have you read email messages while driving?a	1	3	7	11	79
In the last 30 days have you viewed maps or directions on your phone while driving?a	3	7	20	13	57
In the last 30 days have you read messages or viewed information on social media apps or sites while driving? (e.g. Facebook, Twitter, Snapchat, etc.)a	1	4	7	10	78
	Driving at any speed	Driving at low speeds	In stop-and-go traffic	Stopped at a red light	Never
In the last 30 days, when have you written texts while driving?b	5	5	8	22	60
In the last 30 days, when have you read TEXTS while driving?b	9	7	9	26	49
In the last 30 days, When have you written email messages while driving?b	2	3	3	8	84
In the last 30 days, when have you read Email messages while driving?b	3	3	5	13	76

^aQuestion phrased, “For each of the following questions, please choose the answer that best applies”.

^bQuestion phrased, “For each question below, please indicate the highest speed that you have performed the action (meaning, the column furthest to the left that is applicable) in the last 30 days”.

© 2013 Clinical Outcomes Research Unit, Massachusetts Eye and Ear Foundation. Reprinted with permission.

Additional questions covering topics such as crash rates, driving while intoxicated, and demographic information were also described in the initial validation study. Crash rate reporting has been previously described (Bergmark et al., 2016) and was self-reported according to a single question, “In the last 12 months, have many car accidents have you been in with you as the driver? (Answers 0, 1, 2, 3, 4, 5 or more).” The colloquial term “accident” rather than the more modern term crash” was used based on our pilot testing.

2.2. Study design and oversight

The DDS was used to capture major reading and writing activities associated with smartphone use while driving. (Bergmark et al., 2016) Items to evaluate driving while intoxicated, use of smartphone applications aimed at reduction of texting while driving, self-reported crashes in the previous 12 months, and demographic information were included.

The questionnaire was set up as a web-based survey using standard, Health Information Portability and Accountability Act compliant software, SurveyGizmo (Boulder, CO). After submitting the survey, the system was set up to provide a ‘thank you’ page that included the derived DDS score for that participant.

Sample size calculations were based on the ability to compare 4 major US Census divisions with 95% confidence and estimated 267 respondents per group or 1068 in total. The study was approved by the Massachusetts Eye and Ear Institutional Review Board.

2.3. Study population

Subjects were recruited using a third party survey panel (SurveyGizmo, Boulder, CO) and enrolled online through a generic link. Subjects received nominal incentives to participate (i.e. participation in sweepstakes) and were informed that through participation they would receive their DDS scores. Subjects who chose not to continue after reviewing the consent or who reported not having driven a motor vehicle in the prior 30 days were disqualified. Preset limits on subjects based on age cohorts, U.S. Census division and gender were also in place to ensure representativeness. These limits were established with demographic questions. For example, only the first 400 respondents per geographic area were allowed to complete the full survey. Other respondents were excluded, leading to a large number of excluded participants.

In all, 6370 people responded to the survey; 5117 respondents were disqualified primarily to obtain appropriate geographical diversity and 42 were eliminated for partial responses (survey was never finished or submitted). The remaining 1211 respondents constituted the analytical sample.

2.4. Survey reliability

Internal consistency was measured with the method of Cronbach (reported as Cronbach's alpha coefficient). Each item was further evaluated for its contribution to Cronbach's alpha (based on the overall DDS Cronbach's alpha coefficient with each variable deleted).

2.5. Statistical analysis

All data from the testing were transferred to SAS v. 9.0 (Cary, NC) for analysis. The Distracted Driving Survey score was generated as described previously (Bergmark et al., 2016). Logistic regression was performed to evaluate the relationship between the DDS score and other variables as independent variables with a dependent variable of self-reported accidents. All items demonstrating correlations to DDS scores were evaluated using the Wilcoxon test. As there were many respondents with scores of zero or low scores as expected per our validation study, a nonparametric test was selected for analysis. Demographic questions were used to compare the sample to the 2010 U.S. Census for assessment of representativeness and to complete correlation analysis. Due to the infrequency of multiple crashes, analysis was performed comparing respondents with any crashes to respondents with no crashes, and therefore logistic regression was used for analysis.

2.6. Survey reliability and timing

In this study, Cronbach's alpha for the survey was excellent at 0.94 and demonstrates high levels of internal consistency at the individual and population levels. No individual item significantly changed Cronbach's alpha with deletion indicating the relatively equal contribution of each item. This result was similar to the initial validation study of this instrument which had a Cronbach's alpha of 0.93 (Bergmark et al., 2016). In automated testing, the 11-item DDS took approximately 2 minutes to complete, and the full research survey (DDS, demographic questions, and several additional driving-related questions) required approximately four and a half minutes to complete.

3. Results

3.1. Study population

1211 participants completed the survey from 50 states. There were 608 male (50.2%) and 603 female (49.8%) respondents. Mean age was 43.1 years (SD = 15.3 years; range 18 to 78 years; median = 42 years). Respondents were reasonably well distributed between major U.S. census divisions with 23% from West, 23% from Midwest, 19% from Northeast, and 35% from South. In terms of primary driving setting, 45% reported primarily urban driving, 40% suburban driving and 15% rural driving. Age, gender, geography and driving setting distributions were very similar to those reported for the general population in the 2010 U.S. Census (US census). Respondent education levels were somewhat higher than for the U.S. population in categories such as Bachelor's degree (25% versus 18%) and some college, no degree (27% versus 19%).

3.2. Reading and writing behaviors

Mean DDS score was 6.3 (SD = 8.39) with a range from 0 to 44. A non-zero DDS score indicating at least one distracted driving behavior was entered on the survey by nearly 60% of respondents.

As shown in Table 1, reading texts was the most commonly reported distracted driving behavior (48%), followed by viewing maps (43%) and writing texts (33%). Reading and writing email and viewing social media sites were less common. By comparison, 11% of respondents reported having driven while impaired by any substance over the prior 30 days, half of which said they had done so rarely. Only 4.9% of respondents reported having enrolled in a program aimed at reducing cell phone related distracted driving.

3.3. DDS scores are strongly and inversely correlated with age

Age was significantly and inversely correlated with DDS score, indicating that younger drivers reported higher levels of cell phone-related distraction (r = − 0.46, p < 0.0001) (Fig. 1). Total DDS scores were also significantly associated with whether a respondent believes that they can safely text and drive (r = 0.76, p < 0.0001). Further, in comparing respondents less than or equal to 24 years versus those older than 24 years, differences were significant (p < 0.0001). As expected, the mean scores and standard deviations were also lower in the 55–64 and 65 + age groups. No one age 55 or over had a score over 27 (all other age groups had maximum score 41–44).

Fig. 1.

Mean DDS score versus age cohort.

Age was significantly and inversely correlated with DDS score (r = − 0.46, p < 0.0001) among 1211 U.S. drivers in 2015, indicating that younger drivers reported higher levels of cell phone-related distraction.

3.4. DDS scores do not correlate with other demographic variables

DDS scores were not significantly associated with other demographic variables including gender, geography, driving setting or educational levels. As anticipated, DDS scores were also not significantly associated with driving while impaired by any substance in the prior 30 days (p = 0.09), further confirming that this survey is specific to risk from cell phone-related distraction.

3.5. DDS scores correlate strongly with self-reported crash rate

DDS scores were significantly correlated to self-reported crash rates in the prior 12 months (r = 0.18, p < 0.0001). To control for driving while intoxicated, logistic regression was performed with reported car accidents as the dependent variable and DDS and driving under the influence as independent variables. A higher DDS was significantly associated with the number of reported car crashes (p < 0.0001) even when driving under the influence was controlled for, and driving under the influence was still not significantly associated (p = 0.08).

For every increase of 1 point of the DDS, the odds of a reported car accident increased 7% (OR 1.07, 95% CI 1.05–1.10). In order to better characterize the relative risk of higher DDS, a two-way table of car accidents and DDS total was evaluated using the median DDS (3 points) compared to scores above the third quartile (10 and above). The odds of an accident being reported by subjects with a DDS > 10 is 4.3 times (95% CI 2.65–7.05) higher than of subjects with DDS ≤ 3. Differences in crash rates between respondents less than or equal to 24 years and those older than 24 years approached but did not reach significance (p = 0.06).

4. Discussion

In a large and representative sample of U.S. drivers, nearly 60% of drivers admitted to at least one cell-phone related distraction while driving in the past 30 days, higher than in other national studies (Centers for Disease Control and Prevention (CDC), 2013; Olsen et al., 2013). Scores were significantly correlated to self-reported 12-month accident rate, and inversely correlated with age, with 18–24 year old drivers having the highest rates of cell-phone related distraction. The finding of higher rates in younger drivers is consistent with other studies (Tison et al., 2011–12).

Respondents in the highest tercile of scores were > 4 times as likely to have had a crash than subjects with scores in the lowest tercile of risk. The odds of a reported car accident increased 7% for every increase of one point of the DDS score, demonstrating a ‘dose response’ relationship. The effect persisted when controlling for driving under the influence. This correlation would indicate that the DDS could be used as a measure of risk.

The DDS had excellent internal consistency in this national sample (Cronbach's alpha of 0.94) (Kline, 1999). The DDS is brief and easy to administer, with the 11-item scale taking just over 2 minutes to complete in automated testing. Through inclusion of multiple specific reading and writing activities, this survey may allow for a more accurate and specific analysis of cell phone related distraction.

Because survey recruitment was undertaken through the internet it is possible that the sample was not truly representative of all U.S. drivers. The respondent demographic profile is closely aligned with the U.S. census data, which may differ slightly from the US adult driving population. Respondents were more likely than the general U.S. population to have achieved a higher educational status although the lack of differences in DDS scores by educational status makes it unlikely that this significantly impacted the results. Distracted drivers tend to exhibit higher risk behaviors in multiple ways. Further research is required to elucidate the effect size of these reading and writing related distractions versus other distracted behaviors. This study associates DDS score with crash rates and other methodology would be required to determine causality.

Although multiple smartphone applications and other interventions aimed at reducing texting and driving have been created, few of these interventions have been closely studied to assess impact on behavior (AT&T, 2014, Verizon Wireless, 2014, Lee, 2007, Moreno, 2013). In this study, fewer than 5% of respondents reported participation in these programs and those who did participate were more likely to have been in an accident in the prior 12-months. As a brief validated instrument with excellent reliability, the DDS could be used in large populations of drivers to begin to evaluate such efforts.

5. Conclusion

Cell phone reading and writing activities are common in the general U.S. population and vary by activity, with reading and writing text messages and use of GPS being the most common. < 5% of respondents participate in any type of program, such as a cell phone application or pledge, to reduce or limit texting and driving. Higher DDS scores, indicating higher rates of cell-phone related distraction, are significantly correlated to higher self-reported crash rates and are inversely related to age. The DDS may be used to evaluate individual risk and the impact of public health programs aimed at reducing texting and other cell-phone related distracted driving.

Conflicts of interest

The authors declare that there are no conflicts of interest.

Transparency document

Transparency document.