Exploring background risk factors for fatigue crashes involving truck drivers on regional roadway networks: a case control study in Jiangxi and Shaanxi, China

Background Fatigue driving is a leading cause of traffic fatalities and injuries in China, especially among heavy truck drivers. The present study tried to examine which and how factors within the human-vehicle-roadway-environment system contribute to the occurrence of crashes involving fatigued truck drivers. Findings To reduce such risk on the road, a total of 9168 crashes which occurred in Jiangxi and Shaanxi between 2003 and 2014 were selected to measure the effects of potential factors on fatigue related truck crashes using a case control study. Pearson Chi-square test was used to determine the relationship between crash risk and independent factors, and a stepwise logistic regression model was developed to determine the significant risk factors. According to the data analysis results, driver’s gender, age, driving experience, and overspeeding behavior, vehicle’s commercial status, overloading conditions and brake performance, road’s type, slippery pavement and existence of sharp curve and long steep grade, and time of day, season, weather and visibility conditions, etc. were identified to be significantly associated with fatigue related truck crashes on Jiangxi and Shaanxi highways. Moreover, it is found that (a) in Jiangxi, an employed truck driver has a higher risk of crash involving multi-vehicles or a passenger car at bridge locations, and (b) in Shaanxi, the adult, tunnel location, summer and winter days prohibit statistically significant association with the occurrence of multi-vehicle and single-vehicle run-off-road/rollover crashes. Conclusions Young employed male truck drivers with less experience are at high risk, especially while driving across sharp curves, down long steep grades, over bridge or through tunnels, during the midnight period, on rainy, snowy or foggy days in rural areas. All these help recommend potential policy initiatives as well as effective safety promotion strategies at the public health scale for professional truck drivers.

to danger, and falling asleep at the wheel. All of these symptoms induce important performance degradation, and thus increase the likelihood of crashes (Chu 2012;Islam et al. 2014). It is reported that trucks account for only 7.8 % of all motor vehicles, but have caused 17.5 % of total road crashes as well as 22 % of total deaths in China (Chen and Xie 2014). Addressing the high risk of road accidents for this huge group, therefore, represents an important public health issue for the population.
Given the human, social, and economic impacts of these crashes worldwide, it is essential to ascertain the crash characteristics and determine the risk factors in order to better understand how and why they happen (Pei and Fu 2014). In previous studies, drivers' demographic characteristics are a significant contributory factor affecting fatigue-related truck crashes (Di Milia et al. 2011). Among these factors, young male drivers with less driving experience exhibited the highest risk for being involved in fatigue-related crashes (Davenne et al. 2012). This is perhaps because male drivers are often overconfident in their driving skills and therefore, they exhibit dangerous behaviors frequently (Shinara and Comptonb 2004), such as reckless driving, fatigued driving, and drunk driving, etc. In addition, young, inexperienced drivers are also considered high risk, because they often underestimate the potential risks associated with driving situations, and therefore, they cannot tackle driving challenges properly (Friswell and Williamson 2013;Wang et al. 2014). A driver's occupation is another risk factor that affects the occurrence of fatigue-related crashes (Rodriguez et al. 2003). Truck drivers usually have to face stressful working conditions, with long work hours and tight schedules (Arnold et al. 1997;Coeugnet et al. 2013), such as night time driving, irregular rest times, etc., and even do the non-driving physical work (i.e. loading, unloading, heavy lifting, etc.), too. As a result, they have an insufficient amount of time to rest and recover between shifts, which can easily lead to fatigue (Di Milia 2006;Choi et al. 2014).
Likewise, roadway geometrics and environmental factors significantly affect the likelihood of fatigue crashes (Friswell and Williamson 2013;Crum et al. 2001;Jamroz and Smolarek 2013;Du et al. 2015). Monotonous driving situations, with less demanding tasks, would increase fatigue; thus, drivers can easily get fatigued when driving for a long time on straight, smooth and low-traffic roads (Thiffault and Bergeron 2003). Moreover, each individual has his/her own internal "body clock, " also known as "biological rhythm", continuous driving when one should be sleeping leads to degraded performance, especially during from 22:00 to 6:00, which indicates that the time of day is an important risk factor (Di Milia and Kecklund 2013). Certainly, drivers' fatigue patterns should be studied further, due to their significant impact on driving performance. In addition, poor driving conditions, such as hot, rainy, or foggy weather and noise, etc., can also affect the driver's concentration, and thus are more likely to increase fatigue (Chipman and Jin 2009).
Despite various efforts, however, many problems related to fatigue-related truck crashes remain unsolved today. Few previous studies focus on an impact analysis of the various factors that contribute to fatigue-induced crashes among heavy trucks in China. Using reported and reliable crash data from Jiangxi and Shaanxi over a recent 10-year time frame, therefore, this study aims to determine the specific background risk factors associated with (a) the occurrence of fatigue-related truck crashes and (b) the severity of injuries caused by such crashes, and then to propose the corresponding traffic legislation and technical measures targeting different truck groups to help reduce injuries caused by road crashes. Due to these two provinces' specific geographical positions (as shown in Fig. 1), crash features, corresponding risk factors, traffic flow characteristics, and driving environments are more representative when compared with data from 29 other provinces, and thus current studies focusing on Jiangxi and Shaanxi's road safety and corresponding policy recommendations will help develop nationwide traffic policies and legislative measures.

Methods
This paper utilizes 12 years of crash data between 2003 and 2014 from the "traffic accident database" (TADS), released by the Ministry of Public Security of the People's Republic of China and maintained by Jiangxi and Shaanxi Transport Policy Bureau. This system provides precise crash messages that include the time, location, driver and vehicle features, major causes, and environment factors of crashes. Figure 1 presents the location of two studied provinces (Jiangxi and Shaanxi) in China.
The original crash data in TADS consists of 71,695 officially reported crashes in Jiangxi and 63,484 crashes in Shaanxi, among which 12,583 records involved at least one truck. Since the primary purpose of this study is to examine fatigue-related crashes with fatal injuries and fatalities among large trucks, then 9168 records (i.e., 5447 cases from Jiangxi and 3721 cases from Shaanxi) were selected for further analysis based on the original judgment on driving fatigue from traffic police officers, accounting for 6.78 % of total observations. According to the crash data, the injury severity of each individual involved is coded on a four-point ordinal scale: (1) fatal injury, (2) serious injury, (3) moderate injury, and (4) slight injury. Fatal injury refers to crashes that result in immediate death or subsequent death within 30 days of the accident; serious injury refers to crashes that require hospitalization for 2 weeks or more; moderate injury refers to crashes that require hospitalization from 1 day to 2 weeks; and slight injury refers to crashes that require hospitalization for 24-h or less, or no hospitalization at all.
A case control study was conducted using the selected crash sample to examine the potential factors that may contribute to fatigue-related truck crashes, in which 'case' refers to crashes caused by fatigued driving, and 'control' refers to crashes caused by other reasons. Contingency table test methods are used to statistically examine the distribution of each explanatory variable and model the relationship between crash characteristics and contributory factors through Pearson Chi-square (χ 2 ) test. In addition, we conducted a stepwise logistic regression to measure the effects of different factors that contribute to the likelihood of crash occurrences, to identify the significant factors and examine the effect magnitude of adjusted odds ratios for each risk factor, while controlling for other factors. The adjusted odds ratios and the corresponding 95 % confidence intervals are determined by the stepwise logistic regression model, in which all risk factors are initially entered and the insignificant factors were subsequently removed by the stepwise procedure with probability of 5 %. Table 1 summarizes the distribution of crash sample characteristics examined in this study, including driver factors, vehicle factors, road factors, environmental factors, and crash characteristics.

Demographics characteristics of fatigue related truck crashes
As listed in Table 1, the characteristics of fatigue related truck crashes in Jiangxi and Shaanxi are generally similar.
Both truck drivers' age and gender are important factors that are associated with their crashes. The sample has an extremely high proportion of males, accounting for 92.05 and 94.01 % of the total number of crashes for Jiangxi and Shaanxi, respectively, and about two-thirds of truck drivers are 30-55 years old, making up the largest population group of drivers. In view of driving experience, it can be inferred that the more driving years that truck drivers have, the less likely they are to be involved in fatigue crashes (crash proportion: 'below 3 years' = 29.26 and 24.81 %; 'between 3 and 5 years' = 44.65 and 51.81 %; 'between 6 and 15 years' = 19.33 and 17.55 %; and 'above 15 years' = 6.76 and 5.83 %, for Jiangxi and Shaanxi, respectively). Additionally, employed workers are found to have an increased risk of truck crashes while driving when fatigued, and we estimate this group to make up 91.41 and 88.44 % of the total truck crashes in Jiangxi and Shaanxi, respectively.
For vehicle factors, vehicle type, commercial transport status, insurance, overloading conditions, safety status, and brake performance have a positive correlation with fatigued-related truck crashes. Specifically, , 1902 km of a first-class highway, 9941 km of a second-class highway, 10,619 km of a third-class highway, and an additional 101,315 km of a fourth-class highway. As one of the cradles of Chinese civilization, Shaanxi is located along the middle reaches of the Yellow River and a gateway to northwest China. It has an area of 205,600 sq km and its total population reaches 37.75 million by end of 2014. Nowadays, the registered motor vehicles in Shaanxi amounted to more than 6.96 million, and the total highway network exceeds 167,000 km, which includes 4512 km of expressway (c), 1235 km of first class highway, 9268 km of second class highway, 18,164 km of third class highway, and 120,445 km of fourth class highway Among road factors, both the type of road and design geometrics show a strong association with fatigue related truck crashes. As shown in Table 1, a large proportion of such crashes occur on expressways (41.53 and 37.19 %), the first class highways (28.77 and 26.44 %) and the second class highways (19.33 and 21.85 %). In particular, they are more likely to occur on curves, grades, bridges, in tunnels or at urban intersections (highway The results indicate that the time of day is one of the most significant contributing factors in fatigue-related truck crashes. Such crashes are more likely to take place during the evening or at dawn (crash proportion: '0:00-6:00' = 35.01 and 32.55 %; and '20:00-24:00' = 16.14 and 18.44 %), presumably due to the higher level of sleepiness and fatigue associated with human circadian rhythm during these periods. Additionally, a majority of the surveyed crashes took place on weekdays (79.66 and 82.67 %), but not on public holidays (96.53 and 97.88 %). In particular, truck crashes caused by driving fatigue occurred more frequently under adverse weather conditions (31.61 and 23.17 %), such as wet pavements (33.34 and 31.52 %) and decreased visibility (27.48 and 16.47 %), due to the increased brake distance on slippery pavements. Furthermore, winter and summer demonstrate the highest prevalence of fatigue related truck crashes, and each of these two seasons account for more than 30 % of the total crashes for each province.
In view of crash forms, the total proportion of sideswipes, rear-end collisions, and head-on collisions account for approximately 70 % or more (crash proportion: 'sideswipes' = 26.20 and 30.37 %; 'rear end' = 24.34 and 21.58 %; and 'head-on' = 21.68 and 25.05 %) of the outcomes reported in the sample. Notably, fatal crashes in Shaanxi account for only 24.64 % of all of fatigue-related truck crash records, where serious crashes account for 29.11 %, moderate crashes account for 31.93 %, and slight crashes account for 14.32 %, over the surveyed 12 years. However, the conditions become more serious when considering crash injuries and severities in Jiangxi, where the proportion of fatal and serious crashes increased to 27.46 and 32.59 %, respectively; but the proportion of moderate injuries decreased to 24.36 %, while slight crashes remained at 15.59 %, during the same period. Table 1 also shows that more than a half of all fatigue related truck crashes is associated with multiple vehicles (50.21 and 56.89 %), and over one-third (25.30 %) of truck crash records in Jiangxi involves a passenger vehicle, but that in Shaanxi decreases to 12.12 %. Additionally, those causing injuries or fatalities to vulnerable road user account for 3.32 and 4.70 % in Jiangxi and Shaanxi, respectively. Table 2 presents the Chi-square test of independence and the results of the association examination between crash occurrences and risk factors, which indicate that drivers' gender, age, and driving experience, disregarding speed limits and risky following behaviors, vehicle's type, commercial operations, insurance status, overloading conditions, and brake performance, road's type, specific locations and pavement condition, time of day, seasons, weather condition, and visibility condition are potential factors that significantly affect the fatigue-related truck crashes in Jiangxi, China. Table 3 presents the results of estimating the impact of risk factors on fatigue-induced truck crashes by a logistic regression model, in which human, vehicle, road and environment factors all significantly contribute to the occurrence of fatigue related truck crashes in Jiangxi, China. In terms of crash severity and type, such crashes are also identified to have obvious characteristics, as shown in Table 3

Risk factors for fatigue related truck crashes in Shaanxi
Upon analyzing the effects of significant risk factors in fatigue-related truck crashes in Jiangxi, a total of 3721 crash samples in Shaanxi were subsequently examined. From Table 2, 15 factors demonstrate a significant association with fatigue-related truck crashes in Shaanxi, including drivers' gender, age, driving experience and overspeeding behaviour, vehicle's type, commercial operation, overloading conditions and brake performance, road's type, specific locations, and pavement condition, time of day, season, weather condition, and visibility condition.
In terms of crash characteristics, fatal (OR 1.767, 95 % CI 1.422-2.897) and serious (OR 0.752, injuries are negatively associated with fatigueinduced truck crashes. Among these crashes, rear-end collisions (OR 2.061, 95 % CI 1.414-3.665), sideswipes (OR 1.665, 95 % CI 1.439-3.078), and head-on collisions (OR 1.128, 95 % CI 0.866-2.743) rank as the top three most common forms of collision leading to injuries and fatalities. Specially, run off the road crashes (OR 0.764,) are also found to happen more frequently due to truck driver's driving fatigue in Shaanxi expressways, especially within the Qinba areas. Moreover, such crashes are most likely to occur involving multiple vehicles (OR 2.057, 95 % CI 0.896-2.963), which keeps consistent with the previous findings in Jiangxi, but those involving single trucks also often happen in runoff-road and rollover crashes in Shaanxi (OR 2.057, 95 % CI .896-2.963).

Conclusions and discussions
Consistent with present findings in the literature, our results also reveal that the commercial male truck drivers with less experience exhibit a propensity for fatigue crashes, because they are doing a high-risk job and have to drive more often and longer on the road, under time constraints, as well as are more frequent to engage in risky driving behaviours, such as over-speeding and over-loading. All these in turn increase the likelihood of being in a fatigue crash. Specially, employed truck drivers exhibit higher probabilities of involvement in fatigue crashes in Jiangxi, China, because a large percentage of traffic are pass through vehicles, especially the bidirectional traffic between Guangdong Province in South China and North China as well as those between the developed east coast provinces and the remote western areas in China. In Shaanxi, however, such effects are not observed.
Overall, Jiangxi and the majority of Shaanxi are mountainous regions, and truck drivers in these areas are more prone to fatigue-related truck crashes, because of poor roadway geometric design below existing guidelines, such as sharp horizontal curves, steep grades, and absence of tunnels and bridges, etc. Driving over these road segments, truck drivers have to tackle much more frequent, complex reactions and bear high workloads, which exhaust them more quickly. In the long run, such drivers will get fatigued and insensitive to emergency conditions. If they stay alert by keeping themselves in a proper driving posture (i.e. compliance with load and speed restriction and rules regarding continuous driving, breaks and total daily driving, et al.), however, it is generally possible for commercial truck drivers to have good safety records. On the other hand, we did find that fatigue-related crashes among truck drivers sometimes take place on unchallenging, straight road sections, where drivers can acclimatize to the monotonous environment, gradually fall asleep or get distracted and then lose the necessary arousal and vigilance to potential driving risk, which has important implications for roadway design and alignment. Accordingly, it is worthwhile to set multicolour lighting zones and other types of man-made landscape within long tunnel to change the monotonous environment and relieve fatigue (Wang et al. 2016), as performed in the 18.04 km long Zhongnanshan Tunnel in Shaanxi province, China, a section of G65 Baotou-Maoming Expressway (see Fig. 2).
Truck drivers in Shaanxi are found to have a higher probability of being involved in fatigue-related crashes on tunnel sections, especially on the freeways through the major mountains of Qinling, such as G5 (Beijing-Kunming Expressway). In Jiangxi, however, bridge locations seem to be more dangerous. Among environmental factors, time of day was a negative factor that could increase the risk for fatigue crashes among truck drivers. In particular, overnight and during early morning hours (12 midnight to 6 am) are the periods of highest risk, mainly because this is the time that drivers should be in the bed, rather than in the car. In addition, night-time driving may be associated with potential alcohol involvement or music listening, which can impair driver's performance or distract his/ her attention easily (Arnedt et al. 2001;Vanlaar et al. 2008;Brodsky and Slor 2013). On the other hand, bad weather and poor visibility are significantly associated with fatigueinduced crashes involving trucks, largely because trucks need longer distances to brake safely on slippery roads, thus truck drivers experience the highest risk of crashes on slippery roads; however, both drivers and passengers have a lower injury propensity, presumably because vehicles proceed at slower speeds and drivers are more cautious (Russo et al. 2014).
It was found that head-on, rear-end, and angle collisions are associated with more severe injuries among all occupants, while side-swipe collisions demonstrate a lower severity level for injuries. Multiple-vehicle crashes comprise more than half of all crashes involving at least Under the influence of fatigue, a heavy truck is much more likely to crash with a passenger car in Jiangxi, but in Shaanxi it has the higher risk of being involved in a single-vehicle run-off-road or rollover crashes. Furthermore, those involving vulnerable road users are less likely to occur, which indicates that truck crash associated with fatigue is not a major problem in urban areas.
As expected, a driver should be allowed to drive only as long as his/her ability to drive is not impaired (Zaranka et al. 2014;Wang et al. 2015). Therefore, the government and responsible departments should launch corresponding law enforcement programs to regulate driver's max weekly and fortnightly driving hours, maximum continuous driving time in 24 h, and standard minimum rest hours after maximum continuous driving, to reduce fatigue-related crashes. Thus, state-of-the-art technology should be developed to evaluate the drivers' driving performance, provide the warning or alerting messages and assess crash risk, if necessary (Jung and Shin 2014). Specially, drivers are strongly recommended to own vehicles with automatic in-vehicle safety equipment, to alarm drivers of their fatigue or physical impairment, etc.
On the other hand, an education program for safe driving can be developed for not only truck drivers, but also employers. Employers should be forced to extend delivery times and alleviate the burden of truck drivers, and drivers should be provided with intensive awareness training by local governments or employers about how to alert the danger of driving fatigue as well as how to deal with it. Governing agencies could provide local media advertising, cell phone information, brochures or schooling, particularly for long-distance truck drivers, to enhance drivers' knowledge, skills and ability to drive safely. Specifically, it is strongly suggested that improved services at rest stations and larger parking facilities should be offered to make breaks and rest hours more attractive for drivers.
Overall, our findings provide a thorough understanding of the various contributing factors that have a negative association with fatigue-related truck crashes in Jiangxi and Shaanxi, China, and thus can help improve overall truck safety on road. However, the study has some obvious methodological limitations in nature. First, the crash sample was only selected from Jiangxi and Shaanxi, and may be not representative of the overall situation of traffic safety in China. Second, the original data may potentially contain inaccuracies, due to unreported crashes and injuries, indecipherable hand writing, missing or incomplete messages, etc. Third, there is neither "breathalyzer" nor other direct and reliable tool that an investigating police officer can use to determine whether some level of fatigue was a major or contributory factor in a crash, and thus police findings during the original accident report may underestimate the level of fatigue through the "I-know-it-when-I-see-it" approach in identifying fatigue as a cause 1 ; thus, establishing a national monitoring system will help improve uniformity in data collection and analysis between provinces, especially in rural areas. Although traffic flow characteristics have been found to have significant effects on the occurrence of crashes in many previous researches (Kerner 2013(Kerner , 2015(Kerner , 2016, they have not been considered here due to the incomplete or missing messages in TADS, so further works should pay close attention to the association between traffic flow conditions and probability of truck crash occurrence. All these suggestions and future studies might be helpful for promulgating potential policy initiatives and promoting effective strategies to increase overall safety performance on roads in China.