Incorporating the Theory of Planned Behaviour into Distracted Driving: Influencing Factors and Intervention Effects

. Tis study focuses on the psychological characteristics and empirically tests of the factors infuencing distracted driving behaviours. Tis information is used as a reference for an intervention on dangerous driving behaviours. First, a distracted driving scale is constructed based on the theory of planned behaviour (TPB). Te questionnaires are distributed in Chongqing, China, and 321 completed questionnaires are obtained. Data are analyzed using mean-variance analysis, one-way ANOVA, T -test, and multivariate test by SPSS 26.0 to determine the signifcance of distracted behaviours and demographic variables. We use a structural equation model to determine the path coefcients of each latent variable. Finally, we select the drivers with high tendency of distraction from the results of the questionnaires, conduct a four-stage rational emotional behaviour therapy (REBT) experiment, and use a repeated measures ANOVA analysis to test the validity and persistence of the intervention method. Results show that subjective norm is the most infuential psychological factor. Tere are signifcant diferences between the experimental group (2.38, SD � 0.41) and the control group (2.89, SD � 0.40) in the scores of distractions. Tis indicates that the distracted behaviour intervention achieves adequate validity and consistency. Educational research on distracted driving behaviour can help identify and correct drivers with high distraction tendency.


Introduction
Technological progress in the automobile industry and the application of advanced safety equipment has greatly improved vehicle safety [1].Standardization of road design and construction technology has further contributed to this improvement [2].An indispensable link in the drivervehicle-road coupling system, statistics have shown that drivers are the most active and unstable factor, and their bad driving behaviours are the primary cause of frequent trafc accidents [3,4].In recent years, the increase in road trafc fow density and information on various road signs have increased drivers' cognitive load.Additionally, the extensive use of smart phones and multimedia information systems have further consumed drivers' limited attention resources [5], which may lead to a decline in driving performance levels or an increase in accident risk [6,7].Terefore, in order to reduce the accident risk caused by distracted driving behaviour, it is necessary to conduct a questionnaire study on distracted behaviour among drivers.By looking for the signifcant internal causes, we can efectively intervene the occurrence of distracted driving behaviour.
Distracted driving is a hot research topic in the feld of road trafc safety in recent years.In this study, the core collection of Web of Science is used as the database, and a total of 3884 papers from 2002 to 2022 are obtained.Te tendency of published papers is shown in Figure 1.Previous studies explored the infuence of distracted driving behaviour on road safety from the aspects of driver gaze behaviour, saccade behaviour, vehicle speed, and trajectory [8][9][10][11].Other studies had made comprehensive reviews of distracted driving behaviour such as cause, frequency, and detection methods [12][13][14].
Causes of distracted driving behaviour can be divided into visual distractions such as reading mobile phone messages and roadside objects [15,16], cognitive distractions such as "daydreaming" and Bluetooth conversations [17,18], and physical distractions such as sending text messages, changing the car settings, and so on [9,19].In contrast to visual and physical distractions, cognitive distractions may occur when a driver's gaze is always on the road, which makes it covert and hard to detect [12].Terefore, exploring the characteristics and psychological causes of driving distraction may help improve the current research in the feld [20,21].
Schroeder et al. [22] analyzed the characteristics of American drivers' distracted behaviours to understand the underlying psychological mechanism.Talking among passengers and drivers, adjusting the radio, eating, and making a phone call are researched and ranked through a questionnaire survey to describe the occurrence regularity of distracted driving behaviour.However, the risk perception behind such distracted behaviour is not measured.To overcome this shortcoming, Prat et al. [23] used a semistructured scale and found that drivers could fully perceive only one dangerous behaviour and ignored other distracted behaviours.Rather than analyzing a single psychological factor infuencing distracted driving behaviour, Chen et al. [24] verifed the validity of a research framework of planned behaviour theory to analyze multiple factors that infuence distracted driving behaviour, specifcally age, gender, attitude, and prescriptive norms.However, owing to the differences in culture and trafc rules in diferent countries, the applicability of the above theoretical results needs to be studied in an environment with Chinese trafc characteristics.
In order to further elucidate the mechanism of distracted driving behaviour, a theoretical framework for this behaviour needs to be constructed.Te TPB can well predict and explain the internal causes of behaviour through latent variables such as attitude, subjective norms, and perceived behavioural control [25].It has been widely used in felds such as agricultural production [26], transit trip [27], consumption habit [28], and ftness exercise [29] et al.Terefore, based on the research framework of planned behaviour theory, this study constructs a driving distraction questionnaire and tries to fnd out the important factors that afect distraction behaviour.
After fnding the factors that infuence distracted driving behaviour, appropriate methods are needed to intervene the above factors to reduce the tendency of distracted behaviour.Behavioural psychology has recommended several theoretical methods for conducting intervention experiments which explore the mechanism of how diferent demographic and psychological factors infuence distracted driving behaviour [30].Brewster et al. [31] used the conscious awareness of intentions which can weaken the efect of habit, as an intervention in drivers' speeding behaviour.By summarizing the main results of the above studies, it is found that most of the current studies focus on the frequency and infuencing factors of distracted driving behaviour, but few studies carry out targeted educational intervention based on the results of the infuencing factors.Terefore, in order to improve the defciencies of existing research in the feld of distracted driving behaviour, we conduct the intervention experiment based on cognitive behaviour therapy (CBT).CBT has proposed that behavioural intention is not directly determined by events but by people's cognition of the events themselves.Terefore, this method helps people gradually realize their false beliefs and cognitions involved in past bad behaviour through debate.Ten, false beliefs are replaced with proper beliefs to minimize the bad behaviour caused by false cognition.Among the common cognitive behavioural therapies, REBT is the most widely used and accepted behavioural intervention technique [32].Terefore, by measuring the availability and reliability of various methods of training, REBT is fnally selected as the intervention method.
Te objectives of this study are to (i) develop a distracted driving scale based on the TPB as a research framework, (ii) use structural equation modeling to analyze the path coefcients of latent variables, and (iii) conduct a distraction intervention experiment based on REBT to evaluate the intervention's efectiveness and sustainability.

Methods
In order to carry out investigation and intervention on distracted driving behaviour, this study was divided into two stages: building a distracted driving scale based on TPB, analyzing behavioural and psychological characteristics, and fnding out the factors that afect distracted driving behaviour were presented in the frst stage.Drivers with high distraction tendency were divided into the experimental group and the control group.Te experimental group was conducted a distraction intervention experiment based on REBT, and the intervention efect was evaluated in the second stage.

Scale Items' Development.
Tis study collected original distracted behaviours by interviewing eight drivers with the use of self-reports and determined the classifcation of distraction types by consulting fve experts.We designed the driving distraction scale based on the following fve steps.
Step 1: based on the research framework of classic TPB [25], the driving distraction scale was divided into 5 subscales, as shown in Table 1.Step 2: the distracted behaviours were extracted from the interviews and literature studies [23,24], and the contents were modifed and supplemented.Step 3: the distracted behaviours were refned by operationalization.
Step 4: the driving distraction scale was given a title, response instructions, and acknowledgments.Step 5: ambiguous items were deleted or modifed through a small-scale trial investigation.Each item was presented on a 5-point Likert scale [33].After the preliminary deletion of several items, a validity test was performed by SPSS 26.0.Te Kaiser-Meyer-Olkin (KMO) fgure was 0.809, and the results of Mauchly's test of sphericity were signifcant (p < 0.001), indicating that the driving distraction scale has a good validity [34].

Intervention Procedure.
For the process of the intervention experiment, the experimental site was the Trafc Safety Education School of the Motor Vehicle Drivers Association in Nan'an District, Chongqing, China.Te 40 drivers participating in the intervention were divided equally into the experimental group and the control group, and the demographic variables of the two groups were similar.First, before the implementation of the education intervention course, both groups of drivers completed the frst questionnaire.Second, according to the intervention course's characteristics of psychological diagnosis stage, comprehension stage, working through stage, and reeducation stage, the drivers in the experimental group received four educational interventions within a week (they were set on Monday, Wednesday, Friday, and Sunday), with each course of intervention time lasting an hour (10 : 00 AM-11 : 00 AM).Subsequently, the experimental group completed the second questionnaire, while the control group flled the second questionnaire without going through the courses.Finally, in order to verify the sustainability of the intervention efect, all participants were contacted again half a month after the educational intervention and flled in the questionnaire for the third time.Te intervention experimental process is shown in Figure 2.
For the content of intervention experiment, according to the preset requirements of REBT experience, the intervention process was divided into four courses.In the frst lesson, the questionnaire data of each driver were analyzed, and the participants were pointed out through PowerPoint (PPT) pages about the common distractions in the driving process, the impact of distractions on normal driving, and the accidents caused by distractions so that the drivers could realize the high incidence and harmfulness of distractions.In the second lesson, we would analyze the risk perception and psychological motivation of drivers when they were distracted so that drivers could understand the mechanism of distraction.In the third lesson, gradually teaching correct driving knowledge to the participants would enable them to recognize and challenge their incorrect cognition, changed their interpretation of distracted driving behaviours, and transformed the bad behaviours into proper ones.In the fourth lesson, the proportion of distracted driving behaviours and their corresponding risk levels were listed so that drivers could further understand the dangers of distracted driving and generate self-correction ability (see Table 2 for details).

Participants.
To ensure the sufciently representative samples, 361 questionnaires were selected in the frst study stage.As a megacity in China, Chongqing contains diverse trafc scenes and drivers.We conducted our study in Nan'an District of Chongqing and randomly selected drivers to complete the questionnaire survey.After excluding invalid questionnaires such as those with missing answers, a total of 321 valid questionnaires were collected, a response rate of 88.9%.Te samples included 208 men and 113 women.Additionally, the number of professional drivers (those who drive for business) in this survey was relatively small (29), with most being nonprofessional drivers (292).Drivers of all ages were evenly distributed, ranging from 18 to 57 years old, with an average age of 34.8 years and a standard deviation of 10.1 years.Te number of driving experience ranged from 1 year to 31 years, with an average of 6.84 years and a standard deviation of 6.5 years.
In the second stage, with the assistance of the trafc police department, 40 drivers were selected from the frst stage of 321 ordinary drivers, as intervention experiment participants.Compared with the ordinary drivers, the 40 drivers with a total of 12 points were deducted from their driver's license.Such drivers are usually assumed to have a higher tendency to be distracted and need more trafc safety education.

Demographics and Descriptive Variables.
A total of 321 valid questionnaires were collected in this study.Please refer to Table 3 for more information about the participants.Regarding driving time per day, 82% of drivers drive for less than two hours a day.With the increase of age, the frequency of drivers driving every week has increased from sometimes driving to frequent driving.According to the statistics of driver's license point deduction, 89% of drivers were deducted less than 6 points, and 76% less than 3 points.In addition, driving experience increase with age, with an average of 6.84 years for all drivers.

Tendency of Distracted Driving Behaviours.
As seen in Table 4, the average value represents the tendency of each distraction behaviour, from 1 (never occur) to 5 (always occur).Cognitive distractions have the highest frequency (2.79, SD � 1.274) of the three types of distractions during driving.In this distraction type, "chatting with other people in the car" has the highest frequency (3.43, SD � 1.301).Te highest frequency may be because drivers generally believe that this behaviour has little infuence on driving in their daily driving scenarios.As a common type of distraction, physical distractions are ranked second in frequency, with an average of 2.21 (SD � 1.131).At the same time, "adjusting the air conditioner temperature and window" is the second most frequent distracted behaviour (3.31, SD � 1.206).Te least In statistical methods, one-way analysis of variance (ANOVA) is to test whether the mean value of a dependent variable in multiple groups of samples afected by a single factor has signifcant diference [35].Terefore, the ANOVA method is used to analyze the relationship between drivers' age and drivers' distracted behaviours.Similarly, the average value represents the tendency of each distraction behaviour.Tat is, whether there are signifcant diferences in the tendency of drivers of diferent age groups to engage in various distractions.Te results are presented in Table 5.With an age interval of seven years, the drivers are divided into four groups, and the diferences in cognitive distractions, visual distractions, physical distractions, and total distracted behaviours of drivers in diferent age groups were investigated.Cognitive distractions, visual distractions, and total distractions gradually increased from the 20-27 to 28-35 age group.However, for physical distractions, the largest mean value appears in the youngest age group and then decreases in the other age groups.

Path Coefcient from Latent Variables to Distractions.
To further quantitatively demonstrate the infuence of the relationships among the psychological factors in the questionnaire, a structural equation model was introduced to quantify the various latent variables.In addition, personal details were added to jointly explain the distracted behaviour and the assumptions H1-H4 were made in Figure 3. Te structural equation model of data included the fve latent variables of the driver's personal details, distractions attitude, subjective norms, perceived behavioural control, and distracted behaviours.Each latent variable corresponded to several items as observed variables [25].To facilitate the corresponding analysis in Amos   [36].Te coded item data were entered into the structural equation model (Figure 4) for path ftting, and the path was modifed twice according to the ft index.It was added a path from A12 to P12 and set the paths from P to P7 and P to P8 as the same value.Te ft index after the two modifcations is shown in Table 6.
Finally, a path graph that meets the requirements of the ft index was formed.
After the structure equation model and path modifcations were determined, model path parameter estimation was conducted and the results with standardized estimation are shown in Table 7.    8.As can be seen from the table, the diferences were not signifcant when p values were all greater than 0.05, which indicated that the belief cognition of the drivers regarding distracted driving behaviour had not changed during half a month.
Second, using the method of repeated measures ANOVA analysis, the scores of the three types of distraction behaviours in the experimental group and the control group were compared in the three-time questionnaire survey.Mauchly's test of sphericity was applied, as shown in Table 9.
Since the p values of the spherical test for the three types of distracted behaviours are all less than 0.05, multivariate analysis of variance is required.Te analysis results are shown in Tables 10-12 and in Figure 6.In fgure, time point 1 is the frst survey, time point 2 is the second survey, and time point 3 is the third survey.Te results of the multivariate analysis of variance showed that the scores of the three types of distractions in the intervention group were signifcantly diferent across the mentioned time points.Te results of interaction efect analysis between the time point and the intervention mode showed that the efects of intervention tests at diferent time points were maintained.

Questionnaire Validity and Reliability.
To further explore the psychological factors, it is necessary to carry out a targeted questionnaire survey on distracted driving behaviour [21,24,37,38].As the research basis of the questionnaire survey, this study constructed a driving distraction scale with good reliability and validity in line with a Chinese cultural background, trafc scenes, and trafc regulations.Compared with the questionnaire by Prat et al. [23], the driving distraction questionnaire established in this study incorporates more scene information to cover all three types of distracted behaviours.Drivers could understand the questionnaire contents more conveniently through using plentiful scene descriptions, which could further improve the reliability and validity of the distracted driving questionnaire.Additionally, through a small-scale preliminary survey, some vague items were deleted, which made the semantic expression of the driving distraction scale clearer and laid a good foundation for the later statistical analysis.

Infuence of Demographic Variables.
After the statistical analysis of the scale data, the analysis results of drivers' demographic information and driving information show that the scores of male drivers in the same age group are signifcantly higher than those of female drivers in terms of driving age, weekly driving frequency, and everyday driving time, which is consistent with other relevant studies [4,24].In terms of driver's license point deduction, male drivers in the same age group are deducted more points than female drivers.Te reason for this result may be related to the fact  that male drivers are more aggressive in driving style and more prone to speeding, illegal lane changing, and other violations [39].Furthermore, it is worth noting that female drivers in the same age group have a larger ratio of variance in mean driving experience than male drivers, indicating that the polarization of driving experience among female drivers is more obvious than the male ones.Among all types of distracted driving behaviours, the frequency of cognitive distraction (2.79) is the highest, comparing with physical distraction (2.21) and visual distraction (2.11).Cognitive distraction can be experienced when a driver is keeping sight on the road ahead, and often there is no action with one's hands or feet.Terefore, from the perspective of risk perception, the harmfulness of cognitive distraction is generally considered to be the lowest.In particular, the frequency of "talking to other people in the car" is the highest, so its potential risk cannot be ignored.Terefore, relevant trafc regulations have been established in many regions of China to prohibit public transport drivers from talking to passengers.
By analyzing the infuence of various demographic information pieces on driving distractions, it is found that the driver's age has a signifcant negative correlation with the distraction frequency.A possible reason for this result is that younger drivers tend to operate multimedia devices during driving compared with older groups.Additionally, weekly driving frequency and time have a signifcant positive correlation with the frequency of physical distraction.Tis indicates that drivers with higher driving profciency are more likely to participate in physical distraction, and such drivers usually think that the physical distraction has little adverse impact on themselves.
Relevant studies in social psychology have shown that human behaviours are infuenced by internal beliefs.Whether an individual will carry out a certain behaviour is ultimately determined by the mediating variable of behavioural intention [40,41].A similar mechanism exists for driver distraction.In recent years, with the in-depth research and practice of relevant theories, an increasing number of scholars believe that the TPB is an efective means to analyze the mechanisms of various behaviours [42,43].However, only the three core elements of TPB (attitude, subjective norms, and perceived behavioural control) were involved in the path ftting in the preliminary test.Many ftting resists were not fully explained by these paths.To further improve the overall explanatory of the model, this study explores multiple types of potential infuencing factors of the driving distraction scale.Relevant literature studies on attention characteristic factors show that drivers with low attention levels in daily life tend to be more distracted while driving.

Journal of Advanced Transportation
Terefore, this study adds the variables of drivers' attention characteristics to increase the explanatory of the model.However, according to the ftting results of the structural equation model, the path coefcient is −0.09, a very weak negative correlation which is not statistically signifcant.However, Tosi et al. [44] preliminarily confrmed that attention characteristics have a certain potential infuence on driving behaviour.
Journal of Advanced Transportation innovatively explores the internal mechanism of Chinese driver distraction behaviour from a psychological perspective by exploring the psychological motivation behind the distraction phenomenon.Furthermore, in order to make the intervention experiment more efective, educational intervention based on REBT was carried out for relevant psychological factors with a high path coefcient [33,45,46].Tis method has a better operability and allows us to conduct a centralized training intervention for 20 drivers in the form of courses, avoiding the psychological resistance of drivers involved in biomedical therapy.
According to the evaluation of the intervention course's efects, distraction attitude, subjective norms, and perceived behavioural control are signifcantly decreased, indicating that the educational intervention achieves a good validity.However, the decline in the three factors is not consistent.Only the decline in attitudes exceeds 20%, reaching 21.1%, while the decline in the other two factors was less than 20%.Tis result may be related to the shorter duration of the intervention course.Terefore, in the future research, it is necessary to explore the balance point of intervention design to ensure that the course validity and drivers' acceptance of intervention are of a high level.Additionally, from the perspective of the sustainability of the intervention efect, the three types of factors decrease by no more than 20%, but they all exceed 15%, indicating that the persistence efect of the intervention is slightly weaker than the efectiveness.Because the evaluation of the persistence efect is only carried out at two weeks after the end of the intervention, there is a lack of continuous tracking and comparison during longer periods.
Improving the driver's cognitive level of distraction harm fundamentally reverses the driver's distraction intention.Te extraction of typical infuencing factors is helpful for the distraction detection system to consider psychological factors such as driver personality attributes and distraction traits as early warning correction parameters.Matching the diferences of individual attention traits to the personalized early warning model can correct the early warning threshold, improve the accuracy of distraction detection, and further expand the detection system from the explicit visual distraction level to the more implicit cognitive distraction feld.However, only three types of signifcant psychological infuencing factors were identifed in this study.Relevant studies have added factors such as legal norms and risk perception [17,23], which constitute an extended TPB to explain driver behaviour.Future research will need to examine diferent psychological factors to identify their impact on driver distraction.
In the past, driver training was mostly guided by improvements in driving skills.Driver cognition teaching mainly focused on familiarity with regulations and road signs and did not target the correction of wrong driving habits [47].In this study, the driver distraction behaviour frequency data obtained through the distraction scale show that drivers participated in more distraction activities than expected, and among all psychological factors, subjective norm has the most signifcant impact on distracted behaviour.However,Chen et al. [24] regarded attitudes as the most signifcant impact and descriptive norm as the least important factor.In any case, in the driving training process, not only general education links such as driving skills but also the test links of relevant psychological scales need to be added, to comprehensively evaluating whether drivers have a potentially high risk of distraction.As the most important infuencing factor, driver's subjective norm needs to be paid close attention to.Since subjective norm represents the psychological pressure outlined by the surrounding people or society when drivers participate in distraction, the intervention course for subjective norm needs to start from the context level in the future to ensure the further improvement of the intervention efect.
Additionally, with the promotion of relevant national legislation, the majority of drivers have a certain understanding of the harm caused by distraction.However, many focus on the use of hand-held phone, ignoring that the distraction behaviours found in this study are diverse.Underestimating the harm caused by these behaviours will cause cognitive bias.Terefore, relevant cognitive intervention courses must be included in driver training.

Strengths and Limitations of the Research.
Te use of a questionnaire base on the TPB further improved our understanding of the psychological mechanism of driving distraction.Te educational intervention method adopted can well reduce the tendency of distracted driving behaviour.However, there are still some limitations in the current research.For example, in the analysis of the infuencing factors of driving distraction, the existing latent variables do not fully explain driving distraction.In addition, in the existing evaluation of the intervention efect, the sample size of divers is relatively small, the follow-up time is relatively short, and the long-term impact of educational intervention should be investigated in future research.

Conclusion
Overall, this study innovatively explored the internal mechanism of Chinese drivers' distracted driving behaviour and found that subjective norms are the most important latent variables that afect the distraction.It showed the important infuence of social orientation on driving distractions.Other factors, such as attitude and perceived behavioural control, also signifcantly afect distraction behaviour.Tere are signifcant diferences in the tendency of distracted behaviour among diferent age groups.Te efectiveness and persistence of intervention training are in good performance, suggesting that educational intervention has promotional value.It efectively increased the drivers' awareness of the risks of distracted driving.Te research on the infuencing factors and intervention efects of distracted driving behaviour can be applied to select drivers with a potential distraction risk and conduct educational intervention for such groups to improve trafc safety.

Figure 1 :
Figure 1: Annual number of articles published in web of science from 2002 to 2022.

Table 1 :
Framework and content of the distracted driving scale.Te frequency of various types of distracted behaviours over a period of time S Subjective norm Individually perceived social or other pushes or hindrances to a distracted behaviour P Perceived behavioural control Individual perceived ease or difculty of manipulation when a distracted behaviour occurs A Distracted behaviour attitude An individual's perceived liking for the outcome of a distracted behaviour 4 Journal of Advanced Transportation frequent distraction type is visual distraction (2.11, SD � 1.131); "answering an acquaintance's phone on the highway (hand-hold)" is the least frequent of all kinds of distracted behaviours, with a score of only 1.51 (SD � 0.979).

Figure 4 :
Figure 4: Structure of the psychological factor model of distracted driving behaviours.

Table 4 :
Driving distraction degree of drivers.

Table 5 :
One-way ANOVA table of distraction driving behaviour in diferent age groups.
3.4.1.Te Efectiveness of Intervention Training.Te second questionnaire survey was conducted immediately after the course training.Te score distribution of distracted behaviours of the drivers in the experimental group and the control group is plotted in Figure5.It can be seen that the average score of distracted behaviours was lower in the experimental group, and the diference values after training were −0.50 (cognitive distraction), −0.45 (visual distraction), and −0.51 (physical distraction).In addition, the maximum and minimum scores of each behaviour also decreased in varying degrees, and the distribution range of the experimental group was signifcantly smaller than that of the control group, indicating that the score distribution of the experimental group was more concentrated.

Table 6 :
Fit index distribution after twice modifcation.

Table 7 :
Structural equation model standardized path parameter table.

Table 8 :
TPB three factors' T-test in the second questionnaire survey and third questionnaire survey.