Development of a model to predict reported and observed safety behaviors among Chinese city taxi drivers based on the PRECEDE-PROCEED framework

Abstract

Background

Driver behavior plays a crucial role in road safety and is considered a vital variable in accident prevention. High-risk driving is a complex behavior influenced by various factors, including individual knowledge, resources, and skills, attitude, and psychological factors.

Objective

This study aimed to develop a model for the simultaneous prediction of factors affecting the reported and observed safety behaviors of Chinese urban taxi drivers based on the PRECEDE-PROCEED model using structural equation modeling (SEM).

Methods

This cross-sectional study was conducted among 1160 urban taxi drivers in Yong'an city, China. A questionnaire was developed to evaluate the educational diagnosis variables of the PRECEDE-PROCEED model, including knowledge, reinforcing, and enabling factors. The Driving Attitude Questionnaire (DAQ) was used to assess driving attitude. The Wiener Fahrprobe (WF) technique and Driving Behavior Questionnaire (DBQ) were employed to assess the observed and self-reported safety behaviors. All statistical analyzes of the data were performed using IBM SPSS-23 Statistics and AMOS-23 software.

Results

A significant relationship was found between educational diagnosis and driving attitude. Although this variable showed both direct and indirect relationships with drivers’ performance, its direct relationship with driving behavior was not confirmed. However, driving attitude was directly related to driving behavior. Drivers’ attitude had an indirect relationship, mediating the role of behavior with performance.

Conclusions

In summary, considering the simultaneous role of several educational diagnostic variables based on the PRECEDE-PROCEED model can be an appropriate predictor of drivers’ behavior and particularly performance. Therefore, the implementation of training intervention programs based on this model can improve drivers’ safety performance.

Keywords

driver behavior dangerous behavior driving attitude driving performance road safety traffic accidents

Introduction

Driving a public transit vehicle is one of the most hazardous and fatal occupations.¹ Statistics reveal that taxi drivers, with a fatalities rate of 14.9 deaths per 100,000 drivers compared to 3.3 fatalities per 100,000 workers in other occupations, are at a significantly higher risk of death from crashes.² To mitigate this risk, a multifaceted approach is necessary, encompassing changes in driver behavior, ensuring safe road infrastructure, maintaining safe vehicles, and implementing stringent rules and regulations for safe driving, as well as effective crime preventive measures. Among these factors, drivers’ behavior warrants particular attention.³

Driver behavior plays a crucial role in road safety and is considered a vital variable in accident prevention.^3,4 Driving behavior includes behaviors that the driver chooses as a model for his driving, such as speed, concentration, and keeping a standard following-distance. These behaviors are divided into two categories: positive behaviors and negative or unsafe behaviors.⁵ Positive behaviors are friendly behaviors towards other drivers and road users, whose the main intention is to facilitate smooth driving.⁶ Unsafe driving behaviors include repetitive tasks, inactivity, carelessness of drivers, not wearing a seat belt, running a red light, sudden turns without signaling, left turn, improper speed, not paying attention to the front lane, insufficient following-distance to the car in front, using mobile phones while driving, and inappropriate sitting position.^7,8 Drivers who drive patiently and carefully tend to drive safely and are less likely to crash.⁶ Traffic accidents, particularly among adolescents, are attributed to a greater number of negative driving behaviors, including decision-making errors (e.g., improper speed for conditions), high-risk driving behaviors, inadequate hazard perception, distracted driving (e.g., cell phone use), and the presence of peers.⁹

Self-report methodology by the driver is one of the most widely used and important methods of evaluating driving behaviors, which despite the many advantages of this method, such as practicality and ease of implementation, it has certain limitations and sometimes the expression of behaviors is affected by social desirability.¹⁰ Individuals are also careless in their judgments and are criticized in this regard. Self-report surveys are prone to cognitive biases (e.g., recent priorities and effects), affective biases (e.g., mood and emotional states), and self-presentational biases (e.g., self-deception).¹¹ In this regard, measuring the observed behavior (performance), as another method of assessing driving behavior, can complement the shortcomings and sometimes confirm the self-report surveys of behavior.^8,11 Observer assessment techniques (OST) are much more useful for evaluating the effects of attitude structures than self-report surveys; behavioral observations can also provide richer data on drivers’ driving behavior (positive and negative).¹¹ While Observer OST are valuable for evaluating attitude structures, they are not without limitations. OSTs are susceptible to observer bias, particularly the Hawthorne effect, where individuals may alter their behavior simply because they are being observed.¹² In this regard, comparing self-reported and observed behaviors can provide a more realistic picture of drivers’ attitudes and behaviors.^8,11 Therefore, on the one hand, the simultaneous assessment of both behavior survey methods and the factors affecting them can be more practical and less error-prone, and on the other hand, utilizing theories and models as determinants of the path framework will be helpful in recognizing these factors.

Theories and models can serve as a roadmap, provide guidance for educational diagnosis and review, educational planning methods and intervention design, and facilitate evaluation.^13,14 One of the most comprehensive planning models is the eight-step PRECEDE-PROCEED model.^8,15 Educational diagnosis step of this model includes determining predisposing factors (e.g., beliefs, attitudes, awareness), reinforcing factors (e.g., rewards) and enabling factors (e.g., facilities), influencing behavior that provide a platform for educational interventions to modify behavior. This model is a framework that can help health planners and policy makers to design health programs efficiently based on the evaluation and analysis of situations.^8,15 Although several studies have been conducted to identify and determine the factors influencing safe driving behavior^2,16 and sometimes intrapersonal and interpersonal theories have been used,⁸ few studies have been conducted to assess these factors in the context of a comprehensive model, e.g., the PRECEDE–PROCEED model and to measure safe driving behaviors in the form of expressed and observed behaviors.

Human behavior reflects various factors, including attitudes toward behavior and subjective norms, both of which lead to the formation of behavioral intentions and ultimately lead to behavior.¹⁷ Attitude is defined as “a person's favorable or unfavorable evaluation of behavior” and is formed based on a person's behavioral beliefs.¹⁸ Behavioral beliefs of individuals are based on knowledge and knowledge is a strong predictor of people's attitudes toward behavior.¹³ Knowledge is a prerequisite for any behavioral modification, although the knowledge acquisition does not necessarily lead to behavior change, it is undoubtedly a necessary component of behavior change.^14,19 Therefore, attitude in the relationship between knowledge and behavior can play a mediating role.

Previous literature on safe driving behavior has mostly relied on self-reported behavior,^5,6,16 but there may be a difference between the self-reported behavior and the performance (observed behavior). On the other hand, although behavior and sometimes performance have been studied in some similar studies, both behavior and performance are less considered together in the safe driving behavior and in the context of a comprehensive model such as the PRECEDE–PROCEED model in these studies. Furthermore, self-reported behavior is considered as outcome variable in these studies while the performance is not considered.⁸ Any study did not use a comprehensive model such as the PRECEDE–PROCEED model in the field of safe driving behavior. This model is a comprehensive framework often used to guide health promotion and behavior change interventions. In this study, the model's educational and ecological evaluation phases were employed specifically to design the research questionnaire.^8,20 This approach allowed for a systematic assessment of the factors influencing taxi drivers’ safety behaviors. The educational diagnosis phase of the model was particularly useful in identifying and categorizing predisposing, reinforcing, and enabling factors, which were then used to construct a robust questionnaire. By focusing on these key aspects, the model provided a structured framework that ensured all relevant variables were captured in the study. As there is no study so far analyzed all the features of the reported behavior and performance, especially in the platform of the PRECEDE–PROCEED model, the results of this study can provide an overview of all effective factors and the weight of each factor on the reported and observed behavior of urban taxi drivers. The hypothetical model of the study is presented in Figure 1.

Figure 1.

The initial hypothetical model.

The research hypotheses were as follows:

H1: There is a direct and positive relationship between educational diagnosis and driving attitude.

H2: Educational recognition is related to driving behavior directly and also indirectly through the mediating role of driving attitude.

H3: Educational recognition is related to driving performance directly and also indirectly through the mediating role of driving attitude

H4: Driving attitude has a direct and positive relationship with driving behavior.

H5: Driving attitude is related to driving performance directly and also indirectly through the mediating role of driving behavior.

H6: Driving behavior has a direct and positive relationship with driving performance.

Materials and methods

Study design and participants

This cross-sectional study was conducted during 2023 in Yong'an, China. The total population of taxi drivers was 1600 people in Yong'an where 1510 people were eligible to enter the study. Inclusion criteria included being employed in the urban taxi driving job with a code registered in the Yong'an Taxi Organization, having at least one year of experience as a taxi driver, and not suffering from severe mental illness. The questionnaire was designed on the Internet platform to comply with health protocols and maintain the health of taxi drivers as much as possible. Then, the questionnaires reached the target group by forming a group in social media. The objectives of the research and how to complete the questionnaire were fully explained to the taxi drivers before distributing the questionnaire. Then, 1220 drivers announced their readiness to participate in the study. The data of 60 people were deleted due to incompleteness and distorted data, and finally the final analysis was performed on 1160 people. It should be noted that the taxi drivers were not obliged to participate in the study and all of them were given informed written consent to participate in the study. All questionnaires were anonymous. This study was approved by the ethics committee of Fujian Institute of Water Conservancy and Electricity Vocational Technology.

Research tools

Basic demographic characteristics

In this study, a researcher-made questionnaire was used to collect demographic data including age, gender, place of residence, marital status, number of family members, level of education, number of years of driving (driving history), number of years of driving license, driving per week, crash history in the last month, smoking, drug use, income, specific disease, eye weakness, and need for glasses.

Educational diagnosis questionnaire

This questionnaire was a researcher-made questionnaire and was designed in accordance with the educational and ecological evaluation phase of the PRECEDE–PROCEED model. In the process of developing this questionnaire, risk factors related to safe driving behavior were identified based on the educational diagnosis phase of the PRECEDE–PROCEED model. Enabling factors include available resources and new skills, while reinforcing factors include the influence of others and their feedback. A comprehensive literature review was performed in various databases to achieve the conceptual framework of tool development, and the relevant factors were extracted from related articles. Finally, a panel of experts consisting of 10 safety faculty member and driving experts came together and discussed the influential factors in three sessions. In the end, the conceptual framework and classification of causal factors for each behavioral target, prioritization among classifications, and the creation of intra-class priorities were discussed, and a general consensus was reached for each dimension. Accordingly, knowledge among the predisposing factors, required resources and skills among the enabling factors, and the reinforcing role of factors such as family, friends, other drivers, and the taxi organization among the reinforcing factors were agreed. Then the primary proportional items were developed. The initial version of the questionnaire consisted of 28 questions in 3 dimensions including knowledge (10 items) from predisposing factors, enabling factors (8 items) and reinforcing factors (10 items). In order to increase the face validity, the items were tried to have the following characteristics: (i) brief content, (ii) not ambiguous, (iii) no negative verbs, and if necessary, their meaning should be negative, (iv) being one-part, and (v) not being inducible.²¹

The knowledge dimension included items of taxi drivers’ knowledge of traffic rules, such as safe speed on city streets, following-distance to the front car, and information on driving at night, as well as questions to measure people's knowledge of anger management while driving. These items had only one correct answer and the correct answer received one point and the wrong answer zero points. Thus, the maximum score was 10 and the minimum score in this section was zero. Enabling factors included resource accessibility such as safe driving instruction and knowledge of traffic rules and regulations, as well as information on resource accessibility for anger management. The response option for items in this dimension was in the form of a Likert scale with options always (score 5) / often (score 4)/ sometimes (score 3)/ rarely (score 2)/ never (score 1). Thus, the highest score obtained from the total of this section was 50, while the lowest score was 10. Reinforcing factors included items of encouragement or punishment received from family, friends, other taxi drivers, the traffic police, and the taxi drivers’ organization. Responding to items in this dimension was in the form of Likert scale with options always (5)/ often (4)/ sometimes (3)/ rarely (2)/ never (1). Thus, the highest score of this section was 40 and the lowest score was 10. A five-point Likert scale was utilized for its simplicity and effectiveness in capturing a broad spectrum of attitudes and behaviors. This scale is widely recognized for its reliability and ease of interpretation, making it suitable for the diverse demographic of taxi drivers surveyed in this study.

Driving attitude questionnaire (DAQ)

Measuring attitudes, particularly in the context of safety, involves addressing various intervening factors. The tools used in this study were carefully selected to ensure validity and reliability in capturing these complex variables. As demonstrated in previous research,²² understanding the mediating role of factors such as safety climate is crucial for accurate assessment. Our approach incorporated similar considerations to ensure comprehensive evaluation. So, driving attitude questionnaire developed by developed and applied by Batool et al.¹¹ was selected. The questionnaire consists of 58 items, the responses of which are scored on a 5-point Likert scale (from strongly agree (5) to strongly disagree (1)). Thus, the highest and lowest scores in this questionnaire were 290 and 58, respectively. A higher score indicates a better attitude in this questionnaire. This questionnaire measured drivers’ attitudes about speed, seat belt, lane change, running the red line, one-way driving, driving at close distance, drinking alcohol and drug use, overtaking, decentralized driving, proper vehicle driving, driving rules and regulations, aggressive driving, driving self-assessment, other constitutions, driving environment, and social norms.

Driving behavior questionnaire (DBQ)

The driving behavior questionnaire of Newnam et al. was used to evaluate self-report driving behavior.²³ This questionnaire focuses on aggressive violations and Highway Code violations, which was revised Batool.¹¹ The questionnaire consists of 29 items that in the Likert scale are never (0), rarely (1), occasionally (2), often (3), frequently (4), and almost most of the time (5). This questionnaire was translated into Chinese in the present study and its psychometric properties were confirmed.

Driving performance

The Wiener Fahrprobe (WF) technique was used to assess driving performance by the observer. The WF technique is one of the most appropriate and practical techniques for assessing the actual performance of drivers.¹¹ This technique is an in-vehicle observation to study the actual performance of drivers in different traffic conditions. This questionnaire measures four categories of variables including standardized variables, errors, interaction and communication, and traffic collisions. Standardized variables are the types of behaviors that are likely to occur while driving, such as adapting to speeds at intersections/obstacles, changing lanes, interacting with other road users, and driving excessively. Errors are events that indicate a serious violation of the law and/or a hazard. The process of interaction/communication means intentional neglect of the rules. Traffic collisions are situations in which at least two road users are on the collision route and can only be prevented by an unavoidable action by at least one of them.¹¹

Translation and analysis of psychometric properties of study questionnaires

The psychometrics and translation of driving attitude, driving behavior and WF questionnaires included the following process. After permission for intercultural compatibility and observance of the rights to use the questionnaires of driving attitude, driving behavior and WF technique from the original authors, the process of intercultural adjustment of the questionnaire was performed. In the first step (Forward translation), two bilingual experts familiar with English and Chinese, independently and double-blind, translated the questionnaires from English to Chinese. The translators and the research team then compared the translated versions, discussed on the vague and unfamiliar terms, applied and agreed upon the intended modifications, and obtained a temporary Chinese version of each questionnaire. At this step, an attempt was made to prepare a simple, appropriate version while maintaining the semantic value of the main questionnaires. The questionnaires were then given to two bilingual translators in English and Chinese who did not know the contents of the original English questionnaires and were asked to translate the Chinese version of the questionnaires into English independently and double-blindly. The English questionnaires translated by the research team and translators were reviewed and analyzed in one session, and finally a single temporary English version of each questionnaire agreed upon by the research team and translators was obtained. These versions of questionnaires with the ambiguities and disagreements were sent to the developers of the original version for further clarification and explanation. The final version of the questionnaires was approved after applying the required amendments. For cognitive debriefing process, a pre-test was provided to 20 taxi drivers to identify and resolve cognitive problems in the text of the questionnaires, e.g., possible ambiguities and the complexity of items, incorrect sentences, unnecessary questions, embarrassment or exhaustion caused to respondents, etc. Moreover, ten people from the population were interviewed about their perception of the questionnaire items. Structured interviews were conducted with them, each lasting approximately 30 min. Participants were selected based on their willingness to participate and their representation of diverse driving experiences within the urban taxi community, including years of experience and accident history. The results of these two steps were discussed in an expert committee consisting of the research team, 11 health promotion specialists, occupational health professionals, driving instructors, and English translators, and the necessary revises were made to the items. Finally, the final versions were prepared for further consideration of psychometric properties.

To assess the face validity of the questionnaires, a group of 20 urban taxi drivers were randomly selected. First, the purpose of the study was explained to the taxi drivers informed consent was then obtained from them to participate in the study. The questionnaires were given to the participants anonymously and voluntarily. Minor modifications were made after reviewing the questionnaires in terms of comprehensibility and difficulty, wording type, interpretations, cultural issues and item clarity by the participants.

Both qualitative and quantitative methods were used to determine the content validity. To this end, 10 professors of health promotion, occupational health and safety were invited to cooperate. For assessing the validity of qualitative content, experts were asked to review the standards for grammar, wording and item allocations for each item, and to indicate their suggestions for improving the items if these principles are not followed. On the other hand, two indicators including content validity index (CVI) and content validity rate (CVR) were evaluated to assess the quantitative content validity. Experts assessed the fitness rate for each item to evaluate CVI. According to the instructions, a CVI greater than 0.79 is appropriate, between 0.7 and 0.79 requires revision, and less than 0.7 is unacceptable and the item should be removed.²⁴ Experts assessed the necessity of each item to review the CVR. According to the table provided by Lawshe, which is based on the number of experts panel, items with C VR > 0.62 (for 10 experts) are significant at P < 0.05 and items with a lower CVR are removed.²⁵

Internal consistency of researcher-made questionnaires, driving attitude and driving behavior were assessed using Cronbach's alpha index. Cronbach's alpha of 0.9 or higher is excellent, 0.8–0.9 is good, 0.7–0.8 is acceptable, 0.6–0.7 is debatable, 0.5–0.6 is weak and less than 0.5 is unacceptable.²⁶ Inter-rater reliability index based on the opinions of two raters was used to assess WF reliability. Both raters analyzed a situation based on the questionnaire items. Internal correlation is affected by the number of items, which in the level between 0.7 and 0.9 is appropriate.²⁷ On the other hand, the stability of the WF checklist w assessed by intra-rater approach. Thus, 10 taxi drivers in the same situation were evaluated by two raters. The Kappa rate of 0–0.2 is disagreement, 0.21–0.39 is the minimum agreement, 0.4–0.59 is the weak agreement, 0.6–0.79 is the moderate agreement, 0.8–0.9 is the strong agreement, and above 0.9 is the excellent agreement.²⁸

Statistical analysis and software

Kolmogorov-Smirnov test was used to assess the normality of research data. Preliminary statistical analysis including demographic characteristics (e.g., age, gender, education, etc.) and descriptive values of variables were expressed based on mean and standard deviation indicators. More advanced analyzes were applied using the SEM method to test the hypotheses in the form of equations between variables, to take into account measurement error, to explain the relationships between variables, and to eliminate competing models or to present the final model in general. Significance level in all tests was considered 0.05. All statistical analyzes of the data were performed using IBM SPSS-23 Statistics and AMOS- 23 software. In the CFA, goodness-of-fit was investigated based on the root mean square error of approximation (RMSEA), the root mean square residuals (RMR), the goodness-of-fit Index (GFI), the adjusted goodness-of- fit index (AGFI), the comparative fit index (CFI), and the chi-square/degrees of freedom ratio (χ²/df); if CFI ≥ 0.90, RMSEA < 0.08, GFI ≥ 0.8, AGFI ≥ 0.8, and χ²/df < 2, the model goodness-of-fit is considered appropriate.²⁹

Results

All participants in the study were male with the mean age of 35.06 ± 7.63, the mean number of years of driving of 9.35 ± 5.26, and the mean number of years of smoking of 2.22 ± 3.90. Among the studied population, 910 people (77.8%) were urban residents and 810 people (69.2%) were married. About 400 (34.5%) Taxi drivers reported the disease, the most common of which was knee pain 220 (55%). Other demographic data related to the research population are presented in Table 1.

Table 1.

Demographic information of taxi drivers participating in this study.

Variables	Subcategory	Frequency (%)
Marital status	Single	350 (30.2)
Marital status	Married	810 (69.8)
Residence place	Urban	910 (77.8)
Residence place	Rural	250 (22.2)
Education	Under diploma	460 (39.7)
	Diploma	370 (31.9)
	Collage education	330 (28.4)
Smoking	No	730 (62.9)
	Yes	190 (16.4)
	Missing data	240 (20.7)
Drug abuse	No	960 (82.1)
Drug abuse	Yes	200 (17.9)
Income (in USD)	Under 20	120 (10.3)
	550 (47)
	Between 80–120	160 (13.7)
	Up than 120	200 (17.1)
	Missing data	130 (11.1)
Expiry of driver's license	Yes	100 (8.4)
	No	950 (81.2)
	I don’t know	110 (10.4)
Accident last month	Yes	920 (78.6)
Accident last month	No	260 (21.4)
Specific disease	Yes	400 (34.5)
Specific disease	No	760 (65.5)
Type of specific disease	Back ache	150 (37.5)
	knee pain	220 (55)
	High blood pressure	30 (7.5)
Weak eyes	Yes	250 (22.2)
Weak eyes	No	910 (77.8)
Mandatory use of glasses	Yes	110 (44)
Mandatory use of glasses	No	140 (66)
Use of glasses if mandatory	Yes	40 (36.3)
Use of glasses if mandatory	No	70 (63.7)

The results of psychometric properties of the tools used in the study showed that all study tools have the desired validity and reliability. The CVI and CVR scores for the educational diagnosis questionnaire were 0.87 and 0.88, respectively. The total mean scores of CVI and CVR values of the DAQ were 0.81 and 0.76, respectively, indicating the excellent content of the scale from the experts’ point of view; and for DBQ were 0.80 and 0.79, respectively. CVI and CVR were also performed for the WF checklist, which was 0.85 for CVI and 0.82 for CVR.

The results showed that the educational diagnosis, driving attitude, and driving behavior questionnaires had a very good internal consistency and their Cronbach's alpha coefficient was 0.79, 0.75 and 0.70, respectively. The reliability of the WF checklist was conducted in the intra-rater method. The coefficient of Intra-class correlation coefficient (ICC) between raters was 0.76, indicating a satisfactory consistency.

Figure 2 shows the results of the final coefficients of the model in the relationship and correlation between the study variables and Table 2 shows the direct, indirect, and total effects of endogenous variables with exogenous variables. SEM results showed that educational diagnosis has a strong and positive relationship with driving attitude (β = 0.88). Although this variable had an indirect relationship with driving behavior through the mediating role of driving attitude (β = 0.68), its direct relationship with driving behavior was not verified. Furthermore, it was found that there was a relationship between educational diagnosis and drivers’ performance both directly (β = 0.62) and indirectly (β = 0.15). Therefore, hypotheses H1 and H2 were fully verified and hypothesis H3 was partially approved.

Figure 2.

Study model path coefficients.

Table 2.

Direct, indirect and total effects of exogenous variables on endogenous variables.

Endogenous variables	Exogenous variables	Direct effect	Indirect effect	Total effect
Attitude	Educational diagnosis	0.88	–	0.88
Behavior	Educational diagnosis	–	0.68	0.68
Behavior	Attitude	0.77	–	0.77
Performance	Educational diagnosis	0.62	0.15	0.77
	Attitude	–	0.17	0.17
	Behavior	0.22	–	0.22

SEM results also showed a direct relationship between driving attitude and driving behavior (β = 0.77). Although the direct relationship between drivers’ attitude and their performance was not approved, their indirect relationship with the mediating role of behavior was verified (β = 0.17). Moreover, the relationship between drivers’ behavior and their performance was confirmed (β = 0.22). Therefore, hypotheses H4 and H6 were fully verified and hypothesis H5 was partially approved.

According to the results of the CFA, the goodness-of-fit obtained from responses to the Chinese version of the SQOL-M was relatively acceptable. The goodness-of-fit indicators were found to be as follows: RMSEA = 0.071; GFI = 0.90; AGFI = 0.88; CFI = 0.92; and, χ²/df = 1.93. With the exception of χ²/df, almost all these indicators indicated an acceptable goodness-of-fit.

Discussion

This study aimed to develop a model to simultaneously predict the factors affecting the reported behavior and performance of drivers using educational diagnostic variables in the context of the PRECEDE–PROCEED model. According to the results of the study, educational diagnosis directly predicts the drivers’ performance and indirectly predicts reported behavior of drivers through the mediating role of driving attitude. Attitude was also a predictor of drivers’ reported behavior and performance. In addition, the drivers’ reported behavior was predictive of their performance.

According to the results of the study, educational diagnosis is a strong predictor of driving attitude. A direct and strong relationship between educational diagnosis, of which driving knowledge was an important component, with drivers ‘attitudes was to be expected because many previous studies as well as classical behavioral models have shown that increasing individuals’ knowledge through educational programs can lead to Improve their attitude.⁸ However, some studies have found that changing drivers’ attitudes through training programs alone is low and difficult and requires long-term planning. In other words, although increasing staff knowledge through training programs plays a pivotal role, simply increasing knowledge cannot necessarily lead to a change in attitude or change in behavior directly or indirectly.^30,31

According to the findings of the study, educational diagnosis was only indirectly related to driving behavior through the mediating role of driving attitude. However, as can be seen in the findings of the study, educational diagnosis was only indirectly related to driving behavior through the mediating role of driving attitude. But, the relationship between educational diagnosis and drivers’ performance was confirmed in both direct and indirect path. The relationship between educational diagnosis and drivers’ behavior and performance showed that drivers’ performance can be improved through PRECEDE–PROCEED model-based training. In this regard, the direct, strong effect of educational diagnosis on drivers’ performance will be more important than its indirect effect due to the mediating role of driving attitude because some researchers believe that the process of behavior change through attitude change will be a time-consuming and long process.³⁰

Results of this study were consistent with the findings of several studies on the role of training and knowledge on drivers’ behaviors with other educational methods. For example, in a study on the impact of simulation-based training on knowledge, attitudes, and risky behaviors in German drivers, Prohn and Herbig offered that training increases drivers’ knowledge and reduces some of their risky behaviors.³⁰ In a study, Agrawal et al. on a training program based on risk awareness on young drivers found the effectiveness of this program in predicting drivers for hidden hazards and threats.³² However, some studies reported the ineffectiveness of training programs to modify or improve driver behaviors.^21,33

As abovementioned, it seems that the training programs to change drivers’ behavior are more complex than behavior change in other dimensions, mainly because of the multiplicity and variety of risk factors affecting the behavior and knowledge of drivers. Obviously, the effectiveness of educational programs depends on the type of educational program and how it is implemented.^8,15 According to the results of the present study, considering the simultaneous role of several factors, educational diagnosis based on the PRECEDE–PROCEED model can be a predictor of drivers’ behavior and performance. Therefore, it seems that the training programs based on this model, considering the various risk factors affecting driver behavior, if properly implemented can improve driver behavior.

SEM results also showed a direct relationship between driving attitude and driving behavior. Although the direct relationship between drivers’ attitudes and their performance was not verified, their indirect relationship with the mediating role of behavior was approved. The relationship between negative attitudes towards driving and risky behaviors has been extensively studied, revealing three key aspects: attitudes toward rule violations and speeding, carelessness of other drivers, and drinking while driving. Each of these attitudes has been shown to directly correlate with risky driving behaviors.^34–36

The findings of the current study also revealed that there is a correlation between the reported behavior of employees and their performance. However, this correlation was unexpectedly weak. Previous studies reported weak or moderate correlations or inconsistencies in the results related to the relationship between observed driving behavior and drivers’ performance, whose findings are consistent with the recent results.⁵ In sum, the application of self-reported (subjective) methods or objective methods has been one of the main challenges in studies in the field of traffic safety. It is impossible to determine exactly which of the above methods represents the actual performance of the individual. Some studies, particularly primary traffic safety studies, generally used retrospective objective methods, such as crash or near crashes rates recorded by legal authorities, to measure drivers’ performance. Although it seems easy to focus on these indicators to measure driver performance, there are several problems in this regard. First, the incidence rate of crashes is generally low, and some may not be reported for reasons such as low severity. Purposefully, their database is commonly incomplete and has a skewed distribution, for which researchers inevitably have to use weaker statistical methods (e.g., non-parametric statistical tests), which usually have less power to discover a significant relationship. In addition, the ultimate consequences of behaviors such as crash rates have a small contribution to drivers’ performance; i.e., low crash rates do not necessarily indicate good driver performance.³⁷ For all the above-mentioned reasons, reported or observed behaviors were used as outcome variables in traffic safety studies in subsequent studies. Self-report behavior is the most common method used in traffic safety studies. Despite the advantages such as lower cost, higher generality and the possibility of obtaining more detailed information, its application has been criticized by some researchers due to the possibility of participants not responding properly, which itself stems from various causes. In contrast, although objective methods do not have the mentioned limitations, they are less popular for traffic safety studies due to their higher cost, time-consuming and also the possibility of different driver behavior during the presence of an observer or control systems.^30,37 Anyway, a weak correlation between reported behavior and observed behavior was not expected in this study. Social desirability biases can be considered as one of the possible causes of this issue. In this phenomenon, a person would like to express himself in other views as a desirable, reasonable and prosocial person. Of course, various psychological reasons can be suggested for this contradiction, and as mentioned, drivers may even behave contrary to their normal driving behavior due to the presence of a person.³⁷

Research limitations and suggestions for future studies

Since this study conducted in a particular city and also due to the lack of similar study in the group of drivers, it is recommended to conduct similar studies to increase the generalizability of the results in other cities. Despite efforts to consider the risk factors affecting drivers’ behavior in the educational diagnosis phase, it is impossible to identify and consider all the factors in the form of a PRECEDE–PROCEED model in one study. Another limitation of the present study was the large number of items of tools used in the study, which may be effective on the respondents and may cause bias in the final result of the study due to exhaustion. Despite the wide advantages of the SEM technique, this method cannot demonstrate causality, it is required to conduct longitudinal studies for several years and use methods such as neural network or neuro-fuzzy to achieve this objective.

In this study, self-reporting by taxi drivers could have introduced bias, as participants may have altered their responses to appear more favorable. This phenomenon, known as social desirability bias, can affect the accuracy of self-reported data. To mitigate this, future studies should consider incorporating more objective measures, such as direct observations or data from driving monitoring systems, to complement self-reported data. Additionally, the use of mixed methods could provide a more comprehensive understanding of driver behavior by combining quantitative and qualitative data.

Conclusion

A valid tool was designed and developed to assess the educational diagnosis variables affecting drivers’ behavior based on the PRECEDE–PROCEED model in this study. The findings revealed that these variables could well predict the reported behavior and performance of drivers, although their predictive power was higher for drivers’ performance. Therefore, considering the simultaneous role of several educational diagnosis variables including knowledge, reinforcing factors and enabling factors can be a proper predictor of drivers’ behavior and performance. Furthermore, the intervention programs based on the PRECEDE–PROCEED model and its structures can improve the safety performance of drivers and its implementation is recommended in future studies.

Although this study was conducted among urban taxi drivers in Yong'an, China, the findings may be applicable to other regions or similar populations. However, cultural, social, and demographic differences should be considered when evaluating the generalizability of the results. For instance, the specific traffic regulations, driving culture, and socioeconomic status in Yong'an might differ from those in other cities or countries, which could affect the applicability of the findings. Future research should test the model in different settings to explore its broader applicability. This would help in understanding how contextual factors influence driver behavior and in refining the model for use in various regions.

Footnotes

Acknowledgements

We extend our sincere gratitude to the taxi drivers who participated in this study, generously sharing their time and experiences to advance our understanding of safety behaviors in urban transportation

Ethical considerations

Ethical approval for this study was granted by the Ethics Committee of the Fujian Institute of Water Conservancy and Electricity Vocational Technology. While no specific approval code was issued, the study adhered strictly to the ethical guidelines and policies of the institute, including compliance with the Declaration of Helsinki and relevant national regulations governing human subjects research.

Informed consent

Informed consent was obtained from all participants involved in this study. Prior to participation, taxi drivers were provided with a detailed explanation of the research objectives, procedures, potential risks, and benefits. Participants were assured of their right to withdraw at any stage without penalty and were informed that their responses and observational data would remain strictly confidential, anonymized, and used solely for academic purposes. Written consent was obtained from drivers participating in surveys and interviews. Observational data collection in public spaces followed ethical standards for non-intrusive research practices.

Funding

The author(s) received no financial support for the research, authorship, and/or publication of this article.

Conflicting interests

The author(s) declared no potential conflicts of interest with respect to the research, authorship, and/or publication of this article.

References

Llamazares

Useche

Montoro

, et al. Commuting accidents of Spanish professional drivers: when occupational risk exceeds the workplace. Int J Occup Saf Ergon 2021; 27: 754–762.

Burgel

Gillen

White

. Health and safety strategies of urban taxi drivers. J Urban Health 2012; 89: 717–722.

Mokarami

Alizadeh

Pordanjani

, et al. The relationship between organizational safety culture and unsafe behaviors, and accidents among public transport bus drivers using structural equation modeling. Transp Res Part F: Traffic Psychol Behav 2019; 65: 46–55.

Song

Yin

Cao

, et al. The mediating effect of driver characteristics on risky driving behaviors moderated by gender, and the classification model of driver’s driving risk. Accid Anal Prev 2021; 153: 106038.

Singh

Kathuria

. Analyzing driver behavior under naturalistic driving conditions: a review. Accid Anal Prev 2021; 150: 105908.

Shen

, et al. The relationship between personalities and self-report positive driving behavior in a Chinese sample. PLoS One 2018; 13: e0190746.

Bowen

Budden

Smith

. Factors underpinning unsafe driving: a systematic literature review of car drivers. Transp Res Part F: Traffic Psychol Behav 2020; 72: 184–210.

Eisapareh

Nazari

Kaveh

, et al. Effects of an educational intervention program based on the PRECEDE–PROCEED model for anger management and driving performance of urban taxi drivers: a comparison of traditional and online methods. Saf Sci 2023; 157: 105933.

Walshe

Ward McIntosh

Romer

, et al. Executive function capacities, negative driving behavior and crashes in young drivers. Int J Environ Res Public Health 2017; 14: 1314.

10.

Gupta

Goswami

Kumar

. Analysis of driver behaviours towards road safety measures using DBQ in the Indian context. Trans Transp Sci 2021; 12: 12–18.

11.

Batool

. Attitudes towards road safety and aberrant behaviour of drivers in Pakistan. Leeds, UK: The University of Leeds, Institute for Transport Studies, 2012.

12.

Sporrong

Kalleberg

Mathiesen

, et al. Understanding and addressing the observer effect in observation studies. In: Desselle

Cardenas

Aslani

, et al. (eds) Contemporary research methods in pharmacy and health services. Cambridge, MA: Elsevier, 2022, pp.261–270.

13.

Noroozi

Eisapareh

Bahadori

, et al. Development and validation of dust exposure prevention questionnaire for cardiovascular patients based on the health belief model. BMC Public Health 2020; 20: 1–11.

14.

Glanz

Rimer

Viswanath

. Health behavior: theory, research, and practice. Hoboken, NJ: John Wiley & Sons, 2015.

15.

Khorsandi

Dehghan

Pirposhteh

, et al. Development and validation of assessment tool of knowledge, attitude, and practice of outdoor workers regarding heat stress. Work 2022; 71: 671–680.

16.

Nguyen-Phuoc

De Gruyter

Nguyen

, et al. Risky behaviours associated with traffic crashes among app-based motorcycle taxi drivers in Vietnam. Transp Res Part F: Traffic Psychol Behav 2020; 70: 249–259.

17.

Ajzen

. The theory of planned behavior: frequently asked questions. Hum Behav Emerg Technol 2020; 2: 314–324.

18.

Nilsson

Fielding

Dean

. Achieving conservation impact by shifting focus from human attitudes to behaviors. Conserv Biol 2020; 34: 93–102.

19.

Kalteh

Salesi

Mokarami

. The mediator role of safety motivation and knowledge between safety culture and safety performance: the results of a sociotechnical and macroergonomics approach. Work 2022; 72: 707–717.

20.

Crosby

Noar

. What is a planning model? An introduction to PRECEDE-PROCEED. J Public Health Dent 2011; 71: S7–S15.

21.

Gopaul

Singh-Gopaul

Sutherland

, et al. Knowledge, attitude and practice among drivers in trinidad, West Indies. J Transp Technol 2016; 6: 405–419.

22.

Karimpour

Jahani

Shirali

GAJW

. The relationship between safety attitude and safety performance in the petrochemical industry: the mediating role of safety climate. Work 2021;68:507–517.

23.

Abbasi

Monazzam

Karanika-Murray

, et al. Development and validation of an individual job performance questionnaire (IJPQ). Work 2022; 73: 309–320.

24.

Yuan

Zhao

You

, et al. Reliability and validity analyses of the postoperative comfort scale for patients with lung cancer undergoing endoscopic surgery and an evaluation of patient comfort. Technol Health Care 2024; 32: 2331–2343.

25.

Lawshe

. A quantitative approach to content validity. Oakland, CA: Personnel Psychology/Berrett-Koehler Publishers, 1975.

26.

Jain

Angural

. Use of Cronbach's alpha in dental research. Med Res Chron 2017; 4: 285–291.

27.

Mokarami

Eskandari

Cousins

, et al. Development and validation of a nurse station ergonomics assessment (NSEA) tool. BMC Nurs 2021; 20: 83.

28.

McHugh

. Interrater reliability: the kappa statistic. Biochem Med 2012; 22: 276–282.

29.

Maasoumi

Mokarami

Nazifi

, et al. Psychometric properties of the Persian translation of the sexual quality of life–male questionnaire. Am J Men's Health 2017; 11: 564–572.

30.

Prohn

Herbig

. Evaluating the effects of a simulator-based training on knowledge, attitudes and driving profiles of German ambulance drivers. Accid Anal Prev 2020; 138: 105466.

31.

Isler

Starkey

Sheppard

. Effects of higher-order driving skill training on young, inexperienced drivers’ on-road driving performance. Accid Anal Prev 2011; 43: 1818–1827.

32.

Agrawal

Knodler

Fisher

, et al. (eds). Advanced virtual reality based training to improve young drivers’ latent hazard anticipation ability. In: Proceedings of the human factors and ergonomics society annual meeting. Los Angeles, CA: Sage Publications Sage CA, 2017.

33.

Boele-Vos

de Craen

. A randomized controlled evaluation study of the effects of a one-day advanced rider training course. Accid Anal Prev 2015; 79: 152–159.

34.

Jing

Shan

Zhang

. Risk preference, risk perception as predictors of risky driving behaviors: the moderating effects of gender, age, and driving experience. J Transp Saf Secur 2023; 15: 467–492.

35.

Love

Larue

Rowland

. Development and validation of the driver attention regulation scale: a measure of the perceived ability to regulation attention on the road. Transp Res Part F: Traffic Psychol Behav 2024; 102: 199–212.

36.

Tanglai

Chen

C-F

Rattanapan

, et al. The effects of personality and attitude on risky driving behavior among public van drivers: hierarchical modeling. Saf Health Work 2022; 13: 187–191.

37.

Liu

Wang

Guo

. The moderating effects of emotions on the relationship between self-reported individual traits and actual risky driving behaviors. Psychol Res Behav Manag 2021; 14: 423–447.