This paper is aimed at identifying the risk factors that mainly contribute to reckless driving and other related causes of road accidents along the Douala-Dschang highway of Cameroon. The research work started with the collection of accident reports for 2018 and 2019 from security officials in charge of road safety and the police stations of the different localities included in the sample of the study. Three hundred and eighty-two (382) road accidents re ports were collected and analyzed using the 2020 version logit regression model of XLSTAT. From these analyses, it appears that, of the 382 accidents recorded during this period, six factors were identified and classified as follows: causes of accidents related to speed and carelessness, location of the accident, type of vehicle at fault, day the accident occurred, time of the accident and the age of drivers involved. These results could contribute to reduce the gravity of accidents along the Douala-Dschang highway and develop other policies in the program for road safety. In addition, this study can as much as possible equally contribute to reorienting road construction trends and development techniques in our environment.
Reducing road accident injuries, disabilities and casualties is a global social and public health challenge. The World Health Organization (WHO) [
Africa is not left out of this problem of traffic accidents. Every day, nearly a thousand people are killed and tens of thousands injured, not to mention the considerable economic losses they provoke [
These aggregate figures do indicate the magnitude of the problem as a whole, but they do not identify sufficiently enough the nature of the problem in order to point to possible remedies. Similarly, the inadequacy of the results which might have been expected, given the immense efforts made throughout the world to collect as much data as possible on accidents, can be explained in part by the gaps left in certain fields, such as that of exposure to risk [
However, the accident is only the outcome of a chain of probabilities that characterize a situation that is not less purely complex. A classical representation consists of making correspondence to this situation in a space with several dimensions, divided into three subspaces namely of the road, the vehicle and the driver. The determination of the factors of road accidents within the framework of this paper will be based on this density [
Accidents and their casualties on roads are the results of a number of factors. Road users in Cameroon highways in general and those of the Douala-Dschang highway, in particular, are heterogeneous in nature, ranging from pedestrians, moto bikes, tourism cars, buses, trucks, and multi-axle commercial vehicles, etc. Besides the traffic movement of trucks transporting goods within this stretch of road, steadily increases during weekends due to the high movements of passengers by transport and personal vehicles to attend funeral services and celebrations in the west region. Increase in vehicle population with limited road space and its poor nature used by a large variety of vehicles has heightened the need and urgency for a well-thought-out policy on the issue of road safety.
In Cameroon, most road accidents occur along the Yaounde-Douala-Bafoussam-Yaounde triangle. The Douala-Dschang highway which is our sample of the study is part of the National Road Number five (RN5) which is within the said triangle. The triangle is called the “death triangle” because it records over 46% [
The review of literature relevant to road traffic accidents explains some basic concepts: types of accidents, position of vehicle and a consequence which imposes both tangible and intangible costs [
The traffic and road transport system as a whole is composed of three interactive parts which are the vehicles, the road environment and the road users. In the same way, any traffic situation is the result of the interaction among these three systems and in certain circumstances this interaction generates a critical situation or a crash. In general, the number of crashes is affected by three factors: the road environment, the condition of vehicles using the road system, the skill-concentration-physical state of road users [
Although statistics do not allow us to determine with certainty the causes of road accidents, it is undeniable that, among the factors usually distinguished, the human factor remains paramount [
Although the literature on the subject is abundant, one should not expect to find rigorously established facts about driver behavior. There are reflections and observations, but no data that to justify. The subjects of studies are numerous but not enough in Africa, even less in Cameroon, although the phenomenon of accidents is prevalent in this country. It becomes imperative to conduct a causal study of road accidents in its entirety.
After carrying out detailed comparative studies of the different types of research methodology [
According to the Catalogue of the classification of roads in Cameroon [
We are going to limit our study between Bikoko and the end of the Dschang cliff, passing through the following towns: Mbanga, Loum, Nkongsamba, Melong, Santchou and the Dschang cliff as presented in
| Stretch | Description | Interceptive Towns and Villages | Distance |
|---|---|---|---|
| Stretch 1 | Bikoko-Mbanga | Bomono, Nkapa, Souza, Malek, Kompina, Kombe, Muyuka. | 45 km |
| Stretch 2 | Mbanga-Loum | Mombo, Njombe, Penja, | 33 km |
| Stretch 3 | Loum-Melong | Nlohe, Kolla, Mantem, Manjo, Manengole, Nlonako, Nkongsamba, Bare, Melong I, Melong II | 49.9 km |
| Stretch 4 | Melong-end of Dschang Cliff | Lelem, Santchou, Fombap, Dschang Cliff | 49.1 km |
| Total | 177 km | ||
The dilapidation of the road and road signs as well are some factors that could lead to reckless driving and other causes of road accidents [
The information on accident reports along the stretch of roads were obtained from the motorized road safety patrol of the “Gendarmerie” and the police services of the Menoua (Dschang, Santchou) and Moungo (Mbanga, Loum, Melong) divisions. A total of 177 and 171 accidents reports were collected for the year 2018 and 2019 respectively.
The independent variables (age of the driver at fault, day, time, kilometric point, type of vehicle at fault) and the dependent variable (nature of accident) had two categories namely 1 for fatal and 0 for non-fatal [
Six significant variables were statistically used. From
The logistic regression model XLSTAT 2020 version was the method used for the description of the variables and modeling. According to Yin [
| Age of driver | 19 to 70 years |
|---|---|
| Day | Weekly (from Monday to Friday) |
| Time | Hours in four categories (6 am - 12 noon, 12 noon - 6 pm, 6 pm - 12 midnight and 12 midnight - 6 am) |
| Location | Kilometric point/four group of categories: |
| S1: PK25 − PK70 | |
| S2: PK77 + 800 − PK103 | |
| S3: PK109 + 300 − PK152 + 900 | |
| S4: PK170 + 500 − PK201 | |
| Type of vehicle | Four groups of categories (bus, moto cycle, truck and VT) |
| Cause of accident | 1 for fatal and 0 for non-fatal |
The principle of the logic regression model is to link the occurrence or non-occurrence of an event to explanatory variables. XLSTAT can display the classification table used to calculate the percentage of well-classified observations for a given cut off point.
In this study, we code y = 1 (non-fatal or corporal) and y = 0 (fatal) [
ln [ p 1 − p ] = exp β 0 + ∑ i = 1 n β i ⋅ x i = exp β 0 ⋅ exp ∑ i = 1 n β i ⋅ x i (1)
The fundamental equation for the logistic regression shows that when the value of an independent variable increases by one unit and all other values are
held constant, the new probability ratio ( p 1 − p ) is given in Equation (2) as follows:
ln [ p 1 − p ] = exp β 0 + ∑ i = 1 n β 0 ( x i + 1 ) = exp β 0 ⋅ exp ∑ i = 1 n β i ⋅ x i ⋅ exp ∑ i = 1 n β i (2)
where:
β 0 = The constant model.
β i = The parameter estimates for the independent variables.
x i = The set of independent variables ( i = 1 , 2 , ⋯ , n ).
p = The probability ranges from 0 to 1.
ln ( p 1 − p ) = The natural logarithm ranges from negative infinity to positive infinity.
Since the model contains explanatory terms, variables are specified in the model. For the jth observation, let π be the estimated probability of the observed response. The three goodness-of-fit test criteria used for comparing different models for the same data in logistic procedures are: Roc curve and Wald Statistic test.
The odds ratio (OR) is a ratio of two odds. Odds ratio was used to give us an idea of how strongly a given variable may be associated with the outcome of interest compared to other variables. For a probability of successπ, the odds (likelihood) of success (in our case with risk factors) are defined as shown in Equation (3) below.
O R = o d d s 1 o d d s 2 = π 1 / ( 1 − π 1 ) π 2 / ( 1 − π 2 ) (3)
An odd ratio of 1 indicates that the success or desired event under study is equally likely in both groups. An odd ratio greater than >1 indicates that the successful event is more likely to occur in the first group than in the second group and an odd ratio less than <1 indicates that the successful event is less likely in the first group [
The analyses of the data were realized in two phases: the descriptive statistics of the data and the prediction of the nature of accidents.
The number of observations stands at a sum of 382 accidents for the years 2018 and 2019. The outcome of the descriptive statistics of fitness is presented in the following subheadings.
The different measurement in the box plots of
The frequency of accident per day is described as shown in
The time frequency of accident was categorized into four groups namely, the accidents from 6 am to 12 noon, 1 pm - 6 pm, 6 pm to 12 midnight and from 12 midnight to 6 am. As shown on the pie chart
The causes were grouped into eight categories such as bad roads, car defects, drowsiness, excessive speed, imprudence/inattention, natural disaster, non-respect of road signs and poor overtaking. The bar chart of
Causes related to car defects recorded a total of 36 accidents representing 9.499%. The records of the environmental factors were of 10 accidents for bad roads and 5 accidents for natural disaster representing 2.63% and 1.39% respectively should not be neglected. It should be noted that 322 out of 379 accidents were attributed to human factors (drowsiness, excessive speed, imprudence/inattention, non-respect of road signs and poor overtaking) making a total score of about 90%.
Vehicles involved in the accidents were categorized into four; Tourism Vehicles (VT), Trucks, Buses and Motorcycles. The pie chart of
38% (140 accidents) of vehicles at fault were trucks; the tourism vehicles 31% (115 accidents) and 24% (86 accidents) vehicles at fault were buses. Surprisingly, 7% was also recorded for 24 accidents of motorcycles.
The study of accident hot spot was categorized into four sections and in 26 defined spontaneous kilometric points. The description in
respectively, meanwhile S4 recorded 24 accidents at Dschang Cliff, 13 at Lelem representing 6.33% and 3.43% respectively.
Null hypothesis test
This test approves fitness of the model since the Pr > Khi2 value is significantly less than the Wald value 1. The validation test is presented on
The ROC curve Test
ROC curves in logistic regression are used for determining the best cut off value for predicting whether a new observation is a “failure” (0) or a “success” (1). Your observed outcome in logistic regression can only be 0 or 1. The ROC Curve is a plot of values of the False Positive Rate (FPR) versus the True Positive Rate (TPR) for a specified cutoff value. The curve of our model fit is presented in
It was observed that the higher the ROC curve (the closer to line y = 1) the better the fit. In fact, the area under the curve (AUC) can be used for this purpose. The closer AUC is to 1 (the maximum value) the better the fit. The ROC curve of model shows that the predictive power of the model is about 88.6% meaning about 88.6% of performance. Therefore, the model is of high
| Statistic | DDL | Khi2 | Pr > Khi2 |
|---|---|---|---|
| −Log (true resemblance) | 88 | 411.249 | <0.0001 |
| Wald | 88 | 42.177 | 1.000 |
performance hence the fit for the prediction of road accident along the Douala-Dschang highway.
From the analyses, the logic model with the significant variables is:
Pred ( NATURE_ACCIDENT ) = 1 / ( 1 + exp ( − ( − 6.819 × 10 − 2 − 3.367 × 10 − 2 × AGE_DRIVER_CAUSE + DayFact + TimeFact + CauseFact + TypeVehFact + KPFact + cte ) ) )
where:
DayFact is a factor that takes the value corresponding to the day of accident;
TimeFact is a factor that takes the value corresponding to the time of accident;
CauseFact is a factor that takes the value corresponding to the cause of accident;
TypeVeh Fact is a factor that takes the value corresponding to the type of vehicle;
KPFact is a factor that takes the value corresponding to the Kilometric Point.
Cause of Accident
CA-Excessive speed Vrs Natural disaster = exp ( − 0.068 + 0.619 ) / exp ( − 0.068 − 2.029 ) = 14.105
CA-Excessive speed Vrs car defects = exp ( − 0.068 + 0.619 ) / exp ( − 0.068 + 0.053 ) = 2.558
CA-Inattention / Imprudence Vrs Bad roads = exp ( − 0.068 + 0.539 ) / exp ( − 0.068 − 2.029 ) = 13.016
This shows that the impact of accidents due to reasons assigned for excessive speed is about 14 times higher than accidents due to natural disaster and about 2.6 times higher than accidents due to car defects. The impact of accidents due to Inattention/imprudence is about 13 times higher than that due to bad roads.
Location of Accident (LA)
Pk 152 + 9 00 VrsPk 201 = exp ( − 0.068 + 3.520 ) / exp ( − 0.068 + 2.987 ) = 1.7
Pk 152 + 9 00 VrsPk122 = exp ( − 0.068 + 3.52 ) / exp ( − 0.068 + 0.801 ) = 13.45
Pk2 0 1VrsPk122 = exp ( − 0.68 + 2.987 ) / exp ( − 0.068 + 0.801 ) = 8.9
In the case of accidents location, accidents at Pk152 are about 1.7 times higher than accidents at Pk201 and about 13.45 times more than accidents at Pk122 meanwhile accidents at Pk201 are about 8.9 times more than those at Pk122.
Type of Vehicle at Fault
Type Veh-Truck Vrs Type Veh-Bus = exp ( − 0.068 + 0.522 ) / exp ( − 0.068 + 0.597 ) = 0.927
Type Veh-Truck Vrs Type Veh-Motocycle = exp ( − 0.0 68 + 0. 522 ) / exp ( − 0.0 68 − 1 . 283 ) = 6 .00 7
Type Veh-VT Vrs Type Veh-Bus = exp ( − 0.06 − 0.562 ) / exp ( − 0.068 + 0.597 ) = 0.31
In this case, trucks are about 0.927 times at fault than buses and about 6 times than motorcycles meanwhile tourism vehicles (VT) are at fault 0.31 times than buses.
Day Cause
Day-Saturday Vrs Day-Wednesday = exp ( − 0.068 + 0.575 ) / exp ( − 0.068 + 1.039 ) = 0.629
Day-Saturday Vrs Monday = exp ( − 0.068 + 0.575 ) / exp ( − 0.068 + 0.679 ) = 0.901
Day-Wednesday Vrs Day-Monday = exp ( − 0.068 + 1.039 ) / exp ( − 0.068 + 0.679 ) = 1.433
On Saturdays and Thursdays, accidents occur about 0.629 times more than on Wednesdays and about 0.901 times more than on Mondays whereas it occurs 1.433 times more on Wednesdays than on Mondays.
Time Cause
Time 12 noon-6 pm Vrs 6 pm-12 midnight = exp ( − 0.068 + 0.868 ) / exp ( − 0.068 + 0.765 ) = 1.108
Time 12 noon-6 pm Vrs 6 am-12 noon = exp ( − 0.068 + 0.868 ) / exp ( − 0.068 + 0.424 ) = 1.559
Time 6 pm-12 midnight Vrs 6 am-12 noon = exp ( − 0.068 + 0.765 ) / exp ( − 0.068 + 0.424 ) = 1.406
Accidents between 12 noon - 6 pm are about 1.108 times more than those between 6 pm - 12 midnight and about 1.559 times more than those from 6 pm -12 noon meanwhile, from 6 pm - 12 midnight it is about 1.406 times more than accidents from 6 am - 12 noon.
Age of Driver Cause
The Odd ratio was used to interpret the values age of driver at fault.
The odd ratio age of driver below 40 years being a factor for accident is shown as e − 0.207 = 0.813 . This means that accidents due to drivers’ age above 40 years is 81% higher than drivers’ age below 40 years old.
These results reflect the national and global trends in road accident causality studies and even their categorisation. These results are similar to those obtained by the psychologist Robert Ngueutsa [
Moreover, on certain location or stretch, the level of risk will be higher than the general level of risk in other surrounding areas. Crashes will tend to be concentrated at these relatively high-risk locations. Locations that have an abnormally high number of crashes are described as crash concentrated, high hazard, hazardous, hot spot or black spot sites [
This paper examined the risk factors involved in reckless driving and related causes of traffic accidents along the Douala-Dschang highway. The research work applied a quantitative methodology approach by using a logistic regression model of XLSTAT, 2020 version to predict the nature of accidents and to identify the various factors involved. The null hypothesis test and ROC curve test proves that the model is of high performance. The results obtained identified that most causes of road accidents along the Douala-Dschang highway are attributed to human behavior (drowsiness, excessive speed, imprudence/inattention, non-respect of road signs and poor overtaking) that gives a percentage of about 85%. Excessive speed alone recorded 34.3% followed by imprudence/inattention with 30.3%. The model identified that the hot zones are at Melong, Manengole and Bare representing 8.97%, 3.95% and 3.43% respectively. The groups of vehicles at fault most involved in accidents are trucks and tourism vehicles with 38% and 31% respectively. Saturdays and Fridays were identified as being the days during which most of the accidents occur with 21% and 16% respectively meanwhile the most accidental interval was from 12 noon - 6 pm and 6 pm to midnight with 35% and 30% respectively.
The authors are thankful to Dr. ONOMO Cyrille (ESSEC, University of Douala), Simone OWONA (Ministry of communication, Yaounde, Cameroon) and Bide Ivo NGOME (College Catholique St Charles Borromee, Douala) for their critical review and valuable advice which improved the presentation of this paper. We also thank “Gendarmerie” and police officials and their units in the studied sample.
The authors declare no conflicts of interest regarding the publication of this paper.
Zogo Tsala, S.A., Ayissi, M.Z., Azeh, G., Noah, P.A., Betene Ebanda, F. and Ayina Ohandja, L.M. (2021) An In-Depth Analysis of the Causes of Road Accidents in Developing Countries: Case Study of Douala-Dschang Highway in Cameroon. Journal of Transportation Technologies, 11, 455-470. https://doi.org/10.4236/jtts.2021.113030