Factors Influencing Injury Severity in Road Traffic Collisions: A Comprehensive Analysis from Libya

The escalating rates of fatalities and injuries associated with Road Traffic Collisions (RTCs) have escalated road safety into a pressing global issue. Developing countries, particularly those rich in Eastern Mediterranean resources, are witnessing a surge in RTC-related death rates. As depicted in Figure 1, Libya has reported the second-highest road traffic mortality rate among these countries. With an estimated death rate of 26.1 per 100,000 population in 2018 [1], RTCs in Libya account for 11% of all hospital deaths, becoming the third leading cause of hospital morbidity [2]. Therefore, the identification and understanding of primary factors related to RTC injury severity are fundamental in tackling this issue, given its significant contribution to injury, death, and economic loss.

A multitude of research efforts, particularly in developed countries, have been dedicated to discerning the factors associated with RTC severity. Comprehensive literature reviews, such as the one conducted by Christoforou et al. [3], have demonstrated that certain factors are intrinsically linked to increased RTC injury severity. Concurrently, other factors have been associated with a decrease in injury severity, while the effects of some remain controversial. Interestingly, Garrido et al. [4] observed that the influence of certain factors on RTC injury severity varies across countries, likely due to differences in enforced laws, driver behavior, and vehicle fleets.

Despite RTC death rates in Libya being more than three times that of the European Union [2], there is a paucity of studies focusing on RTCs within the country. The initial efforts to study traffic accidents in Libya were taken by Makki in the early 1980s, who based his studies [5, 6] on summary statistical reports issued by the Traffic and Licensing Department (containing information such as population, number of vehicles, number of accidents, number of deaths and number of infections). Subsequent research, including studies by Hamza [7] and Belker and Bensaleh [8], also relied on these statistical reports. More recent studies, such as the one by Yahya and Ismail [9], analyzed traffic accidents in Libya using accident statistics reports issued by the Traffic and Licensing Office in Tripoli. The only deviation in data source was observed in the second part of Hamza's study [7], where injury patterns were examined using a sample of car occupants gathered through questionnaires and medical records.

Previous analyses by Makki [6], Hamza [7], Belker and Bensaleh [8] and Yahya and Ismail [9] employed a simple linear regression model, considering only two independent factors (population and number of vehicles). It should be noted that Makki [5] and the second part by Hamza [7] were merely descriptive studies, as tabulation and the chi-square test were used to achieve the rest of the study's objectives.

All regression analysis results indicated a correlation between the increasing vehicle number and increasing crash rates and severity; also, there is a linear link between population and deaths. Only analysis result in Yahia and Ismail [9] population factor was insignificant. Compared to the investigated developing countries, Libya has been found to have the highest rate of road fatalities among the countries investigated [6].

Descriptive studies further revealed that RTCs are one of the leading causes of death, especially among males aged 15-25 years, and account for nearly half of all deaths on main coastal roads, which are inadequately equipped with ambulance services [5]. Additionally, only 8.2% of drivers and 7% of front-seat passengers were found to be wearing seatbelts; 53% of unbelted drivers sustained head and neck injuries (neck and head trauma), and 93% of those injuries were transported to the hospital by private cars without receiving first aid at the scene of the accidents [7].

A review of the literature highlights the limited scope and data inadequacy of injury studies in Libya. The illogical is that the previous studies only relied on two independent factors (i.e., population and the number of vehicles), which rendered the linear regression model incapable of explaining the actual state of the problem. In conclusion, no previous study using detailed and reliable crash data has been implemented.

To address this gap, the present study employs a logistic regression model (i.e., BLM) to investigate the factors influencing injury severity in RTCs occurring on the rural freeways of Ajdabiya, Libya. The model utilizes detailed and reliable data extracted from relevant RTC investigation reports.

Reliable and accurate data regarding the circumstances of RTCs, road geometry, traffic characteristics, and the weather should be used to guarantee the success of any traffic safety study. The scarcity of such studies in Libya could be attributed to the lack of a comprehensive database for such information. Due to this limitation, the RTCs database was constructed based on a study area with an RTCs data archive that could reflect the state of traffic safety in the country. As it was not feasible to develop a detailed state-level database, The rural freeway roads located in the municipality of Ajdabiya (i.e., El Brega Road, Tobruk Road, El Kufra Road, Benghazi Road, and Maradah Road) were selected as a representative of the road traffic safety situation in Libya given their geographical location importance, functioning to link the east to the west and the north to the south, and the availability of a full investigation reports archive at Ajdabiya Traffic Police Department.

Traffic crash data were extracted from 2,328 handwritten RTCs investigation reports for both urban and rural freeway crashes from the year 2001 to 2010. A manual transcribing methodology was adopted to establish the RTCs database from these reports since data mining (i.e., software) was not feasible as they were written in the Arabic language. Initially, data was collected separately for each year (2001 to 2010). The ten years were merged into one Excel file to build the database, then translated into English, classified, and coded. Data merging and translation processes resulted in some duplicate cases and the emergence of structural errors discovered during data cleaning using IBM SPSS Statistics 22 software. Duplicate cases were removed, and coding errors were corrected. Accident severity (dependent variable) was defined as a binary variable representing two levels of injury severity (0 or 1). No outliers of accident severity were observed, and no missing data was found since the data was collected on a case-by-case basis using the police report number.

The database was organized based on the chronological and spatial data of RTCs. Various factors related to the general information on collision, such as vehicle, driver, and passenger(s) information and the injury severity, as stated in the accompanying medical reports, were used to describe each case. Meanwhile, the traffic data, road geometry, and pavement information were excluded due to the absence of such information and a database at the Road Authority in Libya.

According to the raw data obtained, despite only 53% of RTCs occurring on the rural freeways, they recorded more than 80% of overall deaths. Therefore, only 1,255 crash cases that occurred on the rural freeways were considered in this study. The rural freeway data sample revealed that 67.65% (i.e., two-thirds) of RTCs are solely single-vehicle crashes.

Crash severity is the level of injury a victim is exposed to in a traffic crash and can be classified in several ways. Typically, the KABCO scale is used to classify the injury levels: fatal injury (K), incapacitating injury (A), non-incapacitating injury (B), minor injury (C), and property damage-only injury (O). Nevertheless, in states with a large area, long roads, and low traffic volume, crash data usually have very low frequencies for some of the categories of the KABCO scale. Therefore, combining fatal and severe injury could prove useful in identifying prevalent risk factors [10].

In this study, the KABCO scale for injury severity in road traffic crashes was defined as a binary variable of two levels of injury: 1) non-severe injury=0 and 2) severe injury or fatal =1. In other words, the dichotomous classification was conducted by classifying the response variable as a binary target variable to ensure decreased bias via selection [11].

The variable descriptions are shown in Table 1, and all binary predictors were reasonable with a 0/1 code, no data were missing, and the continuous variable revealed a reasonable range. We also looked at how a continuous variable

(Driver age) related to accident severity and found no discernible effect (Table 2). However, the following analysis employed driver age as a control variable.

This study examined 1,255 rural freeway crash records, whereby the odds ratio (OR) was used for model interpretation purposes. The results of the final BLM model are shown in Table 4, with details on the estimated parameter and p-values of the predictors.

This study aims to conduct an injury severity analysis employing reliable and detailed RTCs data. To achieve this objective, RTCs data for ten years (2001-2010) was obtained from investigative reports housed in the archive of Ajdabiya Traffic Police Department. More than 2,300 handwritten investigation reports were reviewed and then transformed into an electronic format to create an RTCs database. Even while just 53% of RTCs occurred on highways, they accounted for more than 80% of total fatalities, according to the database. Therefore, only 1,255 rural freeway-related collisions were considered for this investigation. BLM model was developed based on 1,255 RTCs records. It was found that risk factors such as weekends, head-on crashes, high-speed, and travel by private car were more likely to increase the likelihood of injury severity in rural freeway crashes. On the other hand, rear-end collisions, crashes involving camels, and being a tractor-trailer occupant were most likely to result in less severe injuries. Furthermore, the review of police investigation reports revealed that reports are handwritten, lack a uniform format, and are not computerized. Also, police officers are low performance, lack traffic safety concepts, and neglect the implementation of laws.

The present study concluded that driving on weekends, Head-on collisions, high speed, and travelling in a private car contribute to an increase in injury severity. Additionally, the lack of efficiency of police officers and the failure of law enforcement. All These combined factors played a role in high road traffic death rate in Libya.

The findings of this study could help transportation agencies in Libya to better understand the effects of various factors influencing crash severity and thus identify effective countermeasures to reduce severe RTCs consequences.

Therefore, the following measures should be considered:

(1) Improvement of the data recording system of RTCs, starting from the police investigation reporting processes and working towards establishing a multisource database.

(2) Enforcement of seat belt use regulations could have a meaningful impact on saving lives, especially in head-on collisions.

(3) Speed control measures on freeways, particularly on weekends, may substantially influence the reduction of injury severity.

(4) Activation of ambulance services, especially on long rural freeways such as Kufra Road and Tobruk Road, where the distance between residential communities may reach 300 km or more.

(5) Legalizing and regulating the import of second-hand cars and ensuring their compliance with safety standards.

Nevertheless, this study was not without limitations, which should be considered when applying the findings. Lack of data such as traffic data, traffic volume, roadway geometry, and pavement information resulted in some factors being excluded from the analysis, which limited predictability and hampered the future strategies' effectiveness in reducing injury severity. A multisource database would be needed to determine these fundamental factors' effects on the injury severity of rural freeway RTCs. Such a multisource database should be implemented for future work to understand RTCs' injury severity better.

In addition, suggestions were made for future research methodologies in the analysis of accident severity and frequency to enhance road safety, and it is as follows:

(1) To account for individual-level heterogeneity in data, a random parameter model such as Mixed logit model is suggested for future injury severity research to reveal unobserved heterogeneity. The analysis should include all categories of factors affecting accident severity.

(2) Count data regression models allow the researcher to estimate the expected number of events (traffic accidents) for an observation unit (number). Count regression models, such as negative binomial models, are suggested for future research to analyse accident frequency in Libya.