Investigation of Factors Associated with Heavy Vehicle Crashes in Iran (Tehran–Qazvin Freeway)

Abstract

With the growing demand for transportation and cargo between cities, the proportion of heavy vehicles in freeway traffic has been increasing in Iran and worldwide during the past decade. The impact of heavy vehicles on crash severity has long been a concern in the crash analysis literature for the prevalence of crashes in freeway traffic. The purpose of this study is to investigate the contribution of heavy vehicles to freeway crashes and uncover other causal factors. Using the comprehensive crash and traffic data from the Qazvin–Tehran freeway in Iran, from 2013 to 2018, 1350 crashes involving heavy vehicles were extracted regarding the weather conditions, weekday, main cause of the crash, driver gender, and culprit side. Considering crash severity calculation, the applied coefficient weights in this study for a person were considered as 3 for an accident resulting in injury and 5 for a fatal crash. A binary logit model was estimated using the data to determine if there was a significant correlation between recognized factors and the likelihood of the crash. The logit modeling results clearly illustrate important relationships between various risk factors and occupant injury, in which heavy vehicles were recognized as one of the most important factors in this study. Other variables associated with crash severity were weather conditions and driver attention. Results indicate that the number of crashes is simultaneously dependent on the total vehicle volume and average speed of heavy vehicles.

1. Introduction

In transportation, road accidents are a serious threat to public health worldwide. The situation has become exacerbated in developing countries, for instance, in 2002, in Iran, the fatality rate was 7.3 people per 10,000 vehicles, which is more than twice the worldwide rate, and this has increased every year [1]. Although freight transportation systems play a significant role in the economic vitality of countries, the specific features of heavy vehicles impose significant safety issues on roads [2]. The increase in the proportion of heavy vehicles has raised traffic safety concerns due to the severity of injuries and economic losses. In other words, accidents involving types of large vehicles, such as trucks, lead to traffic-flow disruption and negative economic effects. Identifying factors affecting the safety of their transportation can have a positive effect by reducing fatalities and on the economy in the long term. According to the National Highway Traffic Safety Administration in the USA (NHTSA), of the truck-involved crashes in a decade, 72% of the fatalities were other vehicle occupants and 11% of the fatalities were not vehicle occupants (pedestrians, pedal cyclists, etc.).

This clearly shows that large trucks significantly affect the safety of other road users. This study considered freight and goods vehicles and long buses with a gross vehicle weight of 3.5 tons or more as heavy vehicles, which includes pickup trucks, trailers, and buses that travel with more than 16 seats. Acquiring official governmental data for transportation on the Qazvin–Tehran freeway allowed us to research three main goals: (1) investigate the factors affecting crash severity using the logit method, (2) determine the impact of heavy vehicles on highway crashes, and (3) create a prediction model for accidents. The next section is an overview of the various models used to predict influencing factors in the crash injury severity literature with a focus on truck crashes. The study area and data description are presented in the third section. The third section also provides a brief explanation of the logit model. The results and discussion are presented in the fourth section. The conclusions are then stated, and the empirical concept is the final part of this article. As road weather cloudy, rainy, foggy, snow, and blizzard conditions were investigated in this study on the Tehran–Qazvin highway. The crash possibility on different days of the week and the driving mistakes that lead to crashes were also studied. Therefore, this study aims to identify the factors affecting the severity of injuries for different occupants from the aspects of heavy vehicles on the highway and to determine whether the mentioned factors in the presence of heavy vehicles can lead to a risk of death or injury. We reviewed the literature considering the factors affecting the injury severities in crashes involving heavy vehicles. Below is a detailed summary of the literature review.

Recent studies have made progress in examining factors such as heavy vehicle presence, weather conditions, and day of the week in relation to accidents, which can inform the development of traffic policies to mitigate traffic congestion. However, little attention has been given to understanding the distribution of accidents at different hours of the day and night under varying conditions and its impact on traffic patterns. Additionally, the type of accident experienced is closely tied to specific time periods, necessitating further research to develop effective strategies for improving road safety and optimizing traffic flow. This research investigates the relationship between accidents and the hours they occur, as well as the association between vehicle speed and accidents, with a specific focus on heavy vehicles. Consequently, analyzing the temporal distribution of accidents and examining the impact of speed on heavy vehicle accidents significantly enhances our understanding of road safety dynamics. In other words, high-speed truck accidents during rush hours on this route result in significant injury and loss of life, highlighting the urgent need to address this issue. By investigating the contributing factors, such as speed and congestion during specific timeframes, effective measures can be implemented to improve road safety and reduce the severity and frequency of such incidents.

2. Previous Literature

Since the invention of the automobile in transportation history, the issue of road injuries has been considered an important issue in this industry. The growing number of road accidents in the world has led to sensible concerns. In addition to safety engineering and traffic volume control on the roads, which lead to a reduction in deaths and injuries from accidents on the roads, other factors such as the weather conditions, the number of vehicles, and the type of vehicles and their occupants are also influential in accidents and injury severities. It is clear that trucks are an important component of a nation’s highway traffic, and there is a relationship between truckload capacity and traffic accidents across the world. Due to their physical and operational characteristics, they can significantly affect the traffic system performance, safety, and travel experience of non-truck drivers [3]. The results from Kaiyang Freeway research in China suggested significant spatial correlations in the crash severity data [4]. In the United States, the reported occurrence of log trucks involved in a fatal crash increased by 41% between 2011 and 2015 [5], and road deaths, injuries, and property damage place a huge burden on the economy of most nations [6]. In particular, the gradually increasing truck proportion will inevitably affect freeway traffic volumes and lead to different levels of traffic safety [7]. The Wang research in Connecticut indicates that injury severity and vehicle damage are highly consistent, and there are correlations between crash counts among severity levels, as well as the crash counts in total and crash proportions by different severity levels [8]. Based on the findings of recent studies, several countermeasures were suggested such as focusing on safety programs for female truck drivers, lower speeds for trucks during rainy conditions, and truck transportation prohibition [9]. The operational problems of road traffic, risk factor identification, and road safety measures are used for statistical analyses on the severity of motor vehicle accidents and provide a beneficial assessment [10]. It was found that truck-involved crashes are affected by factors such as road geometry, weather, driver and vehicle characteristics, and traffic volume using statistical analyses and methodologies [11]. In the scientific literature, all of the independent factors are recognized as model variables that could have a direct or indirect effect on crash frequency. In the past decade, different researchers worked on developing methods for analyzing risk variables in crashes. Consequently, the models can estimate the presence of correlations among variables and crash counts at different severity levels.

According to Chang’s study, a comparison of results indicates that injury severity and vehicle damage are highly consistent, and the consistency among results, crash data, and injury severity was significant when comparing accident characteristics between truck-involved and non-truck accidents. Vehicles with higher occupancies had an increased likelihood of someone being seriously injured based on the logit model. The effects of large trucks have a significant impact on the most severely injured vehicle occupants, which are separately estimated using nested logit models for truck-involved accidents and non-truck-involved accidents [12]. In addition, the random parameter binary logit (RPBL) model for real-time road safety analysis and the support vector machine (SVM) model also can estimate multivehicle (MV) truck-involved crashes, in which large truck drivers are at fault, MV truck-involved crashes, in which large truck drivers are not at fault, and single-vehicle (SV) large truck crashes [13]. Classification and regression tree (CART) findings indicated that the average daily traffic volume and variables such as precipitation, degree of horizontal curve, vertical grade, and vehicle type are the key determinants for freeway accident frequencies [14]. In a truck at-fault crash analysis, the logistic regression model used truck violations as a proxy for truck crashes by considering crash characteristics, driver behavior, speed limit, and driver violation records [3]. A study using data on large truck crashes in Los Angeles over 7 years examined the effect of time-of-day and periods on resulting injury severities in large-truck crashes. In Australia, the finding of a multinomial logit model showed that factors such as occupant variables (e.g., driver age and gender), collision characteristics (e.g., collisions with fixed objects and truck overturns), temporal characteristics (e.g., early morning, midnight), spatial characteristics (e.g., urban or rural areas), environmental factors (e.g., lights condition) increase the probability of fatal/serious injuries in heavy vehicle crashes [15]. In another study, two regression models were developed to study both the maximum injury severity from a crash (over all involved individuals) and the maximum injury severity for occupants in all involved vehicles. The result showed that the fatality likelihood for a two-trailer is higher than that for a single trailer [16], and a similar investigation in New York City confirmed differences between single-vehicle and multi-vehicle truck crashes [17]. The crash severity estimation and crash frequency results in Jung’s study were determined by examining factor impacts on Wisconsin highway safety in rainy weather using a negative binomial regression [18].

For the prediction of a driver’s injury severity, boosted regression trees with the random forest method were used to determine the importance ranking of variables including crash characteristics and environmental, vehicle, road, and traffic characteristics. To consider the site correlation between single-vehicle and multiple-vehicle crashes, a bivariate negative binomial conditional autoregressive model (BNB-CAR) was developed by Wang [19]. Moomen used a random parameter negative binomial regression model to evaluate and predict the impact of geometric variables on crash frequency [6], and the parametric and nonparametric negative binomial model (NB) Poisson was used for crash prediction on freeways in some Wyoming mountains, USA [20]. Furthermore, since the dataset contained a large number of zero truck crashes, a zero-inflated negative binomial (ZINB) model was used to predict crash characteristics in the mountains [21]. By studying Korean freeway data between 2008 and 2010 with a binary logistic regression technique, the factors affecting truck crash severity were identified under normal and adverse weather conditions [22].

To analyze the high over-dispersion problem in accident data and develop the relationship between a truck accident and the geometric design of road segments, a negative binomial regression model was used in Bangladesh [23]. Schneider considered four factors in rural road accidents: the log length of each horizontal curve, the truck, the passenger vehicle, and the degree of curvature, which are statistically significant. These factors were analyzed using a Bayesian analysis and a negative binomial model [24]. For the Kaiyang Freeway, China, other variables were used, including monthly total crash count, monthly vehicle kilometer traveled, average monthly wind speed, monthly and daily average precipitation, and monthly visibility average, in a Bayesian spatio-temporal model [25]. The model was used with spatial correlation regarding crash count, segment length, seasonal traffic volume, and different vehicle classes proportion as independent variables [26]. An injury severity assessment, also using a hierarchical Bayesian random intercept approach for truck-involved crashes, was proposed for crash frequency analyses [27]. Using a binary logit model, remarkable effort merged crash data with weather stations to develop models related to wind conditions and the probability of overturning truck crashes in Wyoming, USA [28].

Some other methods, such as the multinomial logit (MNL) method, also examine the impacts of traffic, driver, vehicle, environment, and geometry factors on the crash severities resulting from truck crashes, for example, the analysis of crash data from Tennessee, USA [29]. To obtain better parameter estimation results, the maximum likelihood (ML) methssod and (MNL) are also used to analyze crash severity and frequency, respectively [30]. Moreover, the gradient boosting method, known as multiple additive trees (MATs), is a novel advancement in data mining, proposed by Friedman at Stanford University, that improves the decision tree (DT) model and benefits from stochastic gradient boosting to analyze crash injury severity [31]. Recently, data mining techniques such as support vector machines (SVMs) have been used in safety research, mostly due to their practical advantages [2]. In research from Alabama, USA, factors such as vehicle model, roadway, accident, and environmental characteristics were used in a fault-tree analysis to predict single- and multi-vehicle large truck crashes using probit and logit formulations [32]. On the other hand, the geographical information system (GIS) has been used to avoid further damage [33].

In Italy, other accident prediction models (APMs) were developed by considering road type, dual carriageway, site type, freeway segment, crash type (single-vehicle and multiple-vehicle crashes), and crash severity including fatal and injury crashes [34]. The serious and general conflict prediction model based on the adaptive network-based fuzzy inference system (ANFIS) was proposed to determine the impact of the truck proportion on freeway traffic safety in Shanxi, China. The network considered truck and other vehicle proportions and the average speed of vehicles [7]. Predicting unsafe driving risk among commercial trucks using a machine learning framework was also performed by Mehdizadeh to analyze crash frequency and injury consequences [35].

In other research, co-clustering is another method that discovers hidden latent patterns and utilizes the duality of clustering by generating a compact representation of the dataset [36]. Moreover, hotspot identification is possible when using spatial crash clustering analysis methods [37]. Modeling road traffic accidents usually includes accident frequency and injury severity forecasting using statistics, whereas when problems get complicated, artificial neural network models should be used as they can better handle nonlinearity in the data, such as Zeng and Huang’s proposed training algorithm and network structure optimization for injury severity prediction [38]. Based on the simulation and hybrid model, in an empirical analysis to predict crash frequency, four different structures covered a wide range of dependency structures, including radial symmetry and asymmetry, asymptotic tail independence, and dependence [39]. Over the years, important data and different methodological issues have been identified in highway crashes. Considering heavy vehicle importance, some studies from the recent decades look for crash parameter estimation. The differences in the types of freeways, weather conditions in different seasons, the volume of traffic during a specific period, speed, and characteristics of crashes that occurred on highways have created diverse conditions for the studies. By reviewing the studies conducted in the last decade, we identified the most effective parameters, as presented in

Table 1. Input and Output of Recent Research.

Authors	Year	INPUT										OUTPUT						Country	Case
		Heavy Vehicle Crashes	Roadway Characteristic	Environment Conditions	Vehicle Volume	Speed	Injury	Driver Characteristic	Temporal Parameters	Crash Characteristic	Damage to Vehicle	Parameters Estimation	Injury Severity	Crash Type	Crash Severity	Finding Major Factors	Hot Spot Identification
Abrari Vajari et al. [15]	2021	✓	✓	✓		✓	✓	✓	✓				✓					Australia	Victoria
Bhowmik et al. [39]	2021		✓														✓	USA	Central Florida
Hosseinzadeh et al. [2]	2021	✓	✓	✓		✓	✓	✓	✓	✓		✓	✓	✓				Iran	Qazvin, Tehran, Ghom
Orsini et al. [13]	2021					✓			✓			✓						-	-
Mehdizadeh et al. [35]	2021			✓		✓		✓	✓							✓		USA	USA
Wang et al. [8]	2020						✓				✓	✓	✓		✓			USA	-
Shaik et al. [23]	2020	✓			✓			✓	✓	✓				✓				Bangladesh	Khulna City
Zandi et al. [37]	2020				✓	✓						✓						Iran	Qazvin, Tehran
La Torre et al. [34]	2019		✓	✓	✓		✓	✓	✓				✓	✓	✓			Italy	Italian Freeways
Zeng et al. [4]	2019		✓	✓				✓	✓	✓								China	Kaiyang Freeway
Cole et al. [5]	2019	✓	✓	✓		✓	✓					✓	✓					USA	Highways
Wen et al. [25]	2019		✓	✓								✓				✓		China	Kaiyang Freeway
Uddin et al. [9]	2020		✓	✓	✓	✓	✓		✓	✓		✓	✓					USA	Ohio
Moomen et al. [6]	2019	✓	✓	✓	✓		✓					✓						USA	Wyoming
Zeng et al. [4]	2019		✓		✓	✓			✓			✓						China	Kaiyang freeway
Rahimi et al. [36]	2019	✓		✓		✓	✓	✓		✓	✓	✓	✓	✓				USA	Florida
Zhang et al. [7]	2019	✓			✓	✓						✓						China	Shanxi
Rezapour et al. [6]	2018	✓	✓	✓	✓	✓	✓	✓	✓	✓		✓				✓		USA	Wyoming
Choi et al. [22]	2014	✓	✓	✓	✓	✓	✓	✓	✓			✓	✓		✓	✓		Korean	Korean freeway
Schneider et al. [24]	2006		✓							✓		✓	✓					USA	Ohio

.

To deal with the methodological issues and effective data, a variety of methodological techniques were applied to analyze crash data over the years. The statistical methods have primarily relied on the nature of the dependent variables. Regarding the ordinal nature of data, i.e., injury as an instance, dependent variables with multiple response outcomes have been treated as both ordinal and nominal.

Table 2. Applied Methods in Recent Research.

Authors	Year	METHOD																						Country
		Logit	Negative Binomial Regression	Bayesian Model	Multinomial Logit	Poisson Lognormal	Random Forest	Support Vector Machine	Data Mining	Heteroskedastic Ordered Probit	Fault Tree Analysis	Binary Outcome	Ordered Discrete Outcome	Union-Dered Multinomial Discrete Outcome	Simulation-Based Approach	Analytical Closed -Form Approach	Machine Learning	Neural Networks	Adaptive Network-Based Fuzzy	Gradient Boosting	Boosted Regression Trees	Multivariate NB	Liner Regression
Abrari Vajari et al. [15]	2021				✓							✓	✓	✓										Australia
Bhowmik et al. [39]	2021														✓	✓						✓		USA
Hosseinzadeh et al. [2]	2021	✓						✓																Iran
Orsini et al. [13]	2021						✓	✓																-
Mehdizadeh et al. [35]	2021																✓							USA
Wang et al. [8]	2020		✓			✓																		USA
Shaik et al. [23]	2020		✓			✓																		Bangladesh
Zandi et al. [37]	2020																						✓	Iran
La Torre et al. [34]	2019																							Italy
Zeng et al. [4]	2019	✓		✓																				China
Cole et al. [5]	2019	✓																						USA
Wen et al. [25]	2019			✓																				China
Uddin et al. [9]	2020	✓																						USA
Moomen et al. [6]	2019		✓																					USA
Rahimi et al. [36]	2019																							USA
Zhang et al. [7]	2019																		✓					China
Rezapour et al. [6]	2018																							USA
Choi et al. [22]	2014																							Korea
Schneider et al. [24]	2006		✓	✓																				USA

provides a listing of studies that used specific methods to analyze crash data along with the researchers and case locations.

We reviewed a number of research studies conducted within the past year, and the result as shown in Figure 1 showed depicted that researchers have utilized the following inputs to investigate accidents with heavy vehicles more comprehensively. Roadway characteristics and environmental conditions appeared to be the most frequently recorded factors in heavy vehicle crashes, both at 13%, indicating their importance in the occurrence of heavy vehicle crashes. Vehicle volume and speed followed those factors at 12%, respectively. Although injuries and fatalities seem to be important in heavy vehicle crashes, only 11% of the studies addressed them. The next priorities involved consideration of driver characteristics and temporal parameters, as well as crash characteristics and vehicle damage. As it shown in Figure 2 the results included parameter estimation, which was the most significant goal, at a percentage of 45.21%, indicating the potential impact on understanding the occurrence. Injury severity, at a percentage of 23.29%, was also a critical factor to consider for reducing the impact of crashes and ensuring the safety and well-being of individuals involved in the accident. Crash types and crash severity were observed in the results of recent research at 10.96% and 9.59%, respectively.

3. Methodology

In this study, a heavy vehicle crash is defined as a single heavy one- and any two-vehicle crash that included at least one light, medium, or heavy one. The probability of highway accidents can be considered as zero or non-zero, and thus the binary logit model could be utilized [40]. It provides a measure of the strength and direction of the relationship between the predictor variables and the binary outcome variable, which is used to identify risk factors. Due to the discrete nature of the independent binary variables of crash and fatality, the binary logit model seems more appropriate for identifying causal relationships between crash, severity, and their features [2]. Logistic regression is a maximum-likelihood method that has been used in hundreds of studies on crash outcomes [41]. It is a statistical regression model for binary dependent variables such as death or life. Being two-sided means that a random event occurred in two possible situations. In fact, in a logistic relationship, the function changes between about 0 and 1, as mentioned in Equation (1). This method is known as the natural logarithm, which can represent (g(x) = 1) in a fatal versus nonfatal (g(x) = 0) crash, as shown in Equation (2).

$$p(x_i)=\frac{e^{{\beta }_0+{\beta }_1{x}_1+{\beta }_2{x}_2+...+{\beta }_i{x}_i}}{1+e^{{\beta }_0+{\beta }_1{x}_1+{\beta }_2{x}_2+...+{\beta }_i{x}_i}}$$

(1)

$$g(x_i)=Ln(\frac{p_i}{1-p_i})=Ln(\frac{\frac{e^{{\beta }_0+{\beta }_1{x}_1+{\beta }_2{x}_2+...+{\beta }_i{x}_i}}{1+e^{{\beta }_0+{\beta }_1{x}_1+{\beta }_2{x}_2+...+{\beta }_i{x}_i}}}{1-\frac{e^{{\beta }_0+{\beta }_1{x}_1+{\beta }_2{x}_2+...+{\beta }_i{x}_i}}{1+e^{{\beta }_0+{\beta }_1{x}_1+{\beta }_2{x}_2+...+{\beta }_i{x}_i}}})=Ln(e^{{\beta }_0+{\beta }_1{x}_1+{\beta }_2{x}_2+...+{\beta }_i{x}_i})={\beta }_0+{\beta }_1{x}_1+{\beta }_2{x}_2+...+{\beta }_i{x}_i$$

(2)

where P represents the probability of a fatal crash, x_i represents the independent or predictor variable i, and β_i represents the model coefficient used to directly determine the odds ratio. In other words, it is commonly estimated with maximum likelihood estimation to optimize for the best fit of the log odds.

Serious injury and fatal crashes are classified as critical and severe crashes. Based on the available information, the binary logit model was applied to examine the effects of the driver, environmental factors, and roadway on severe and non-severe injuries of the occupants involved in truck-related crashes. The probability of the risk for every predictor was achieved using Equation (3)

$$p_i=\frac{1}{\exp [-({\beta }_0+{\beta }_1{x}_1+{\beta }_2{x}_2+...+{\beta }_i{x}_i)]}$$

(3)

Data Description

This study utilized the Qazvin province traffic police accident database and focused on the fatal and injury-causing accidents on the Qazvin–Karaj freeway in Qazvin province, Iran. Considering the Provincial Meteorological Department records of weather conditions and crash places, raw data were gathered in tables exceeding 70,000 records. Due to the large number of vehicles in transportation, about 31 million annually, and the vital connection to four provinces directly and to five other provinces indirectly, the freeway is the most important transportation route in Qazvin province [42]. Consequently, the number of casualties and injuries on this freeway includes a significant number of the total casualties of Iran. In Qazvin province, there were 224 fatalities and 6837 injuries just for one year (2021). In the last 4 years, the average number of fatalities and injuries on this freeway was 500, which is the highest in Qazvin province and is one of the largest numbers among all freeways in Iran (Qazvin province traffic police, accident data). After cleaning the gathered data spreadsheets, the prepared data were moved to SPSS for analysis. Due to the remarkable transport of heavy vehicles between Tehran and Qazvin, this research was designed to study the influence of heavy vehicle impact on crashes and severe injuries and fatalities between 2017 and 2021 and examine the critical conditions. The results can be used by transport managers for risk reduction in different seasons, weather conditions, and days of the week, and relative countermeasures could be determined to prohibit crashes.

4. Results

The results obtained are based on an analysis conducted over five years, using both inferential statistics and descriptive statistics. Using this rigorous examination, this research focused on exploring various aspects including the frequency of incident occurrence and the interrelationship between the collected data. By utilizing inferential statistics, this study delved into concluding the population based on the analyzed sample, while the descriptive statistics provided a detailed summary of the data, facilitating a deeper understanding of the investigated phenomena.

4.1. Descriptive Statistics

Crash occurrence is a result of casual variables, which could depend on weather conditions, weekday, vehicle class, etc. as it shown in

Table 3. Descriptive Statistic of the Model Variables on Crash Occurrence.

Variable	Variable Level	%	Variable	Variable Level	%
Weather	Sunny	79%	Culprit Vehicle	Heavy Vehicles	74%
	Cloudy	6%		Lack of Attention to the Front	66%
	Rainy	13%	Cause of Crash	Not Complying Distance	2%
	Frosty	0.5%		Inability to Control the Vehicle	10%
	Foggy	0.5%		Speed Limit Offense	1%
	Snowy and Blizzard	1%		Lane Deviation	14%
Weekday	Not Holiday	53%		Precipitate Direction Changing	1%
	Holiday	17%		Inattention to the Right of Way	3%
	Before and After the Holiday	30%		Illegal Overtaking	2%
Gender	Male	98%		Technical Failure of the Vehicle	1%

. The descriptive statistics for this study conducted on the Qazvin–Karaj highway show that the highest percentage of crashes occurred in sunny weather conditions (79%) due to prolonged sunny weather in this area, and the lowest frequency of crashes was related to frosty and foggy weather conditions (0.5%). Based on the recorded data, 17% of vehicle crashes occurred on holidays, 30% were related to days before and after holidays, and 53% of accidents were related to all non-holiday days. In other words, the contribution of each non-holiday day of the week was equal to 10.5%. The obtained data shows that 2% of the drivers included in the crash were women and 98% were men. For vehicle type, this study determined that heavy vehicles were involved in 74% of crashes, with a significant contribution, and the remaining accidents happened only among passenger cars.

Considering the main cause of a crash, the driver’s lack of attention to the front accounted for the highest percentage of all crashes 66%, which was followed by switching to the left lane and inability to control the vehicle at 14% and 10%, respectively.

4.2. Inferential Statistics

Regarding the discrete nature of the collected data in terms of statistical methods, this study used logit modeling to analyze the data. First, the impact of the independent variables on crash occurrence was investigated. Then, using the logit model, the effect of each of the independent variables on the severity of the accident, which is a combination of injury and death, was examined. In the binary model, the Wald test was used to determine the significant predictor variables and express whether there was any difference between the odds of a crash for vehicles with and without the risk factor under consideration [43]. The odds ratio shows the positive impact of 18 predictor variables among 20 on the crash that occurred on this highway, which is presented in

Table 4. The Effect of Independent Variables of The Model for Crash Occurrence.

Variable	Variable Level	β	(Wald)	(Sig)	OR = Exp(β)
Weather	Sunny	0.223	1.706	0.108	1.249
	Cloudy	0.603	5.706	0.001	1.827
	Rainy	0.703	6.805	0.001	2.019
	Frosty	0.849	8.336	0.001	2.337
	Foggy	0.369	2.041	0.033	1.446
	Snowy and Blizzard	0.951	9.025	0.001	2.588
Weekday	Not holiday	0.206	1.698	0.120	1.228
	Holiday	0.639	6.011	0.001	1.894
	Before and After the Holiday	0.753	7.056	0.001	2.123
Gender	Male	−0.233	1.711	0.105	0.792
Culprit Vehicle	Heavy Vehicles	0.876	8.609	0.001	2.401
Cause of Crash	Lack of Attention to the Front	0.882	8.805	0.001	2.415
	Not Complying Distance	0.487	4.546	0.001	1.627
	Inability to Control the Vehicle	0.554	5.506	0.001	1.74
	Speed Limit Offense	0.663	6.209	0.001	1.940
	Lane Deviation	0.641	6.006	0.001	1.898
	Precipitate Direction Changing	0.389	2.806	0.022	1.475
	Inattention to the Right of Way	0.501	5.203	0.001	1.650
	Illegal Overtaking	0.659	6.338	0.001	1.932
	Technical Failure of the Vehicle	0.236	1.729	0.100	1.266
Significant Statistics of the Model		−2 Log Likelihood = 9023.361			R2 = 0.705

.

According to this table, five out of six weather conditions have significant coefficients. The snowy and blizzard condition, with an odds ratio of 2.588, was identified as the most dangerous situation, which was followed by three states of frost, rain, and cloudy weather. Since the number of travels between Tehran, as a metropolis, and Qazvin surges considerably on the days before and after the holidays, there was a significant coefficient for these days and an odds ratio of 2.123, which indicate the highest probability compared to other days. Regarding the driver’s gender, the odds ratio for men was lower than 1, and their effect on crash severity was lower for women. As mentioned in the descriptive statistics, heavy vehicles had a significant presence in crashes, and their odds ratio of 2.401 from the logit model suggests that this issue is remarkable. Among the recognized causes, the highest risk was associated with a lack of attention to the front, followed by the speed limit offense, illegal overtaking, lane deviation, inability to control the vehicle, etc.

Considering the descriptive and inferential statistics, the presence of heavy vehicles is one of the main factors in the occurrence of crashes, and other identified parameters contribute to an accident. If the general flow of traffic between cities is considered, then the probability of crashes in both heavy and low traffic conditions is significantly different. Increasing the speed of heavy vehicles on average and the presence of light traffic increases the number of accidents per hour. According to Figure 3, when total vehicle volume is between 150 and 250 per hour and the average speed of the heavy vehicle is between 95 and 105 km/h, crash frequency surges dramatically.

Given the substantial amount of traffic on the Tehran–Qazvin highway and the occurrence of accidents at various times, it is feasible to investigate the correlation between the time of day or night and the type, primary cause, and speed of accidents. Figure 4 provides a visual representation of the average crash type, including front-to-backside crash overturned and rollovers and others considering different hours of the day in which the rear-end crash rate was significantly high.

Figure 5 indicates that the leading cause of crashes among all other recognized factors was the lack of attention to the front. This finding highlights the importance of driver attentiveness in preventing accidents. Figure 6 clearly illustrates the disparity between the average number of accidents during day and night hours, as well as the correlation between vehicle speed and accident frequency. Notably, the majority of accidents occurred when vehicles were traveling at speeds between 90 and 100 km per hour.

5. Discussion

Examining collisions involving trucks is crucial for identifying the factors that significantly impact the severity of such crashes. The present study focuses specifically on large truck-involved collisions for predicting the influential factors contributing to crash injury severity. To accomplish this, a method known as the binary logit is used for analyzing and forecasting these influencing factors. The statistical results obtained indicate that drivers’ inattention to the front vehicles, speed violations, and sudden lane changes in the presence of heavy vehicles on the Tehran–Qazvin highway significantly increase the risk of accidents. These results are also supported by previous studies [2]. Rear-end collisions with vehicles cause over 23% of the observed accidents on this route, which could be rooted in driver fatigue. Furthermore, more than 20% of accidents occur at speeds between 90 and 100 km/h, resulting in a substantial loss of life and financial damages due to the weight of heavy vehicles. This highlights the potential for heavy vehicles to exacerbate injuries or fatalities in accidents, emphasizing the need to identify the underlying factors [3].

Based on the observations shown in Figure 3, increasing the speed of heavy vehicles leads to an increase in risks and accidents associated with them. Additionally, transitioning from light to heavy traffic conditions has a negative effect on accident rates. Furthermore, higher speeds reduce the driver’s control over the vehicle. When traffic conditions change from light to heavy, the density of vehicles increases, resulting in shorter distances between vehicles. In this situation, even at reduced speeds, the risk of accidents remains high. However, when the traffic load is reduced and the density of cars decreases, with increased distances between vehicles, the number of accidents decreases compared to periods of heavy traffic.

To reduce accidents, policymakers should prioritize addressing the identified main factors, which include promoting driver attentiveness to the road ahead, curbing illegal speeding, and discouraging dangerous lane deviations. These factors were also mentioned by Chen, although they used deep learning methods [44]. Raising driver awareness using media education can effectively educate people about the consequences. Additionally, many researchers have expressed that installing enhanced pavement markers such as rumble strips and reflectors, along with warning signs, can increase driver vigilance. Moreover, the probability of accidents varies significantly during different hours of the day and night, reflecting distinct driver behaviors during these periods. Although Hosseinzadeh [2] cited significant cases in his research, he did not mentioned the accident-prone hours in this area on this highway.

The sideslip angle plays a vital role in vehicle stability and control, especially at higher speeds, carrying a significant safety risk. However, its impact on accidents may be less pronounced at lower speeds influenced by factors such as straight roads, heavy traffic, or driver behavior. Furthermore, the presence of turns and road deviations greatly affects the sideslip angle. Nevertheless, considering the limited number of turns or curves on this specific highway, the relative importance of the sideslip angle is diminished compared to other contributing factors and thus was not discussed.

Additionally, given the climatic characteristics and traffic volume fluctuations between the two cities, prediction is achievable [45]. The importance and impact of weather conditions and traffic patterns are shown in Table 4 and Figure 3, respectively. As policymakers and researchers continue to improve road safety on the Tehran–Qazvin highway, future studies should also consider the types of heavy vehicles and their electronic stability control systems along the route to assess their impact. Optimal decisions should be made regarding the movement of certain vehicles on this route based on the type of stability system they possess. Lastly, including accident-prone areas in the aforementioned measures can further reduce the likelihood of accidents.

6. Conclusions

Crash severity and relative factors are and will remain essential topics of traffic safety research, and numerous efforts should be made to evaluate the scope. This paper examined factors that could interfere with crashes on one of the most important freeways in Iran (Qazvin–Tehran). Although this freeway plays an important role in Iran’s economy as a transport route in the west of the country, there is a significant number of crashes due to heavy vehicle transportation. Based on the logit model results and statistical findings, which were gained using a broad range of variables including, environmental conditions, vehicle type, culprit vehicle, weekday, and main cause, heavy vehicle transport led to a high probability of crashes. The incident of crashes and injury severity is dependent on the traffic volume level on this freeway. Consequently, it can be observed, generally, that the main cause of accidents on this highway is the driver’s lack of attention to the front. Speed violations and lane changes are ranked next, respectively, regardless of the direction of movement of other vehicles, and this type of collision often happens in the form of a rear car hitting the back of the front car on average every 10 h. Other types of crashes include vehicle overturning, side-by-side collisions, road departure collisions, and collisions with motorcycles. The majority of accidents occur at speeds exceeding 90 km per hour, and the rate of accidents decreases with reduced traffic load.