GEOGRAPIC INFORMATION SYSTEMS IN DETERMINING ROAD TRAFFIC CRASH ANALYSIS IN IBADAN, NIGERIA

Abstract

Background

Road traffic accidents are frequent in this environment, hence the need to determine the place of geographic information systems in the documentation of road traffic accidents.

Aim & Objectives

To investigate and document the variations in crash frequencies by types and across different road types in Ibadan, Nigeria.

Materials & Methods:

Road traffic accident data between January and June 2011 were obtained from the University College Hospital Emergency Department’s trauma registry. All the traffic accidents were categorized into motor vehicular, motorbike and pedestrian crashes. Georeferencing of accident locations mentioned by patients was done using a combination of Google Earth and ArcGIS software. Nearest neighbor statistic, Moran’s-I, Getis-Ord statistics, Student T-test, and ANOVA were used in investigating the spatial dynamics in crashes.

Results

Out of 600 locations recorded, 492 (82.0%) locations were correctly georeferenced. Crashes were clustered in space with motorbike crashes showing greatest clustering. There was significant difference in crashes between dual and non-dual carriage roads (P = 0.0001), but none between the inner city and the periphery (p = 0.115). However, significant variations also exist among the three categories analyzed (p = 0.004) and across the eleven Local Government Areas (P = 0.017).

Conclusion

This study showed that the use of Geographic Information System can help in understanding variations in road traffic accident occurrence, while at the same time identifying locations and neighborhoods with unusually higher accidents frequency.

Introduction

Road traffic crashes (RTCs) and the associated injuries are a major cause of death in developing countries1,2. Indeed, more than 90% of RTCs occurred in low and middle-income countries3,4. The World Health Organization (WHO) noted that, death resulting from RTCs would increase by 65% between 2002 and 2020, surpassing death from malaria and tuberculosis unless something was done4. Nigeria has the second highest RTC fatalities among the 193 countries in the world5. The increasing road crashes has the potential of adversely affecting the health system and hampering the attainment of Millennium Development Goals (MDGs) 4 and 5 since youths are mostly affected3,4. RTC prevention and mitigation should therefore be accorded greater attention to reduce the increasing human loss and injury.

There is paucity of data on locational characteristics of RTC sites in most developing countries. This could be attributed to the life threatening nature of crashes which typically makes interrogating affected patients directly impossible since most of them arrive hospitals either unconscious or in pain. Data on locational characteristics of crash sites, type of vehicle involved, time of the day and weather condition are required in order to comprehensively address RTCs most especially, where the cause of the RTC is locational or environmentally dependent. Geographic Information Systems (GISs) provide tools and techniques for identifying and analyzing the influence of location on phenomenon6. GIS applications in road crash analysis offers data management system as well as cartographic and analytical functions in support of RTCs management7,8. GIS has been widely used in crash analysis to identify high risk neighborhoods, areas of vehicle collision and pedestrian black spots9^-13. The growing use of GIS is based on the increasing availability of digital data coupled with the need to increase precision in the identification of crash locations while reducing time and money expended on such analysis. The use of GIS for crash analysis is limited in Nigeria perhaps because of the inadequate awareness of its potential. Furthermore, variations in crashes across different administrative units (Local Government Areas), types of roads (dual and non-dual carriage), mode of transportation (motorcycle, motor vehicles and pedestrian) and different part of the city (inner and outer city) have not been investigated within a singular framework. This study focused on variations in road traffic crashes (RTC) frequencies and patterns across different administrative spatial units in Ibadan metropolis.

Materials & Methods

The study was conducted in Ibadan metropolis (figure 1) the capital of Oyo State, which is the third largest metropolitan area by population in Nigeria after Lagos and Kano. It is made up of eleven Local Government Areas (LGAs). Ibadan is located in southwestern Nigeria, 128 km inland northeast of Lagos and 530 km southwest of Abuja.

Figure 1 . Administrative Map of Oyo State Showing Local Government Areas in the Ibadan Metropolis .

After obtaining institutional ethical approval, data on road crashes between January and June 2011 within the Ibadan metropolis was obtained from the University College Hospital Emergency Department’s Trauma Registry. The hospital receives over 10,000 unrestricted emergencies annually with an admission rate of 47% and it also serves as a referral center for other hospitals in Southern Nigeria. All crashes were categorized based on the mode of transportation: motor vehicle crash (MVC), motorbike crash (MBC) and pedestrian crash (PC). The Emergency Department (ED) Trauma registry was designed locally using Microsoft Access database software. Crash victims and social health workers who accompanied them to the hospital for treatment were administered a questionnaire which covered demographics, detailed injury causation/site of occurrence, pre-hospital state/care, referring hospital, clinical assessment/admission, procedures and primary payment source.

The analog administrative map showing boundaries of LGAs in Ibadan metropolis was obtained from the Oyo State Ministry of Land and Survey. The map was subsequently converted into digital format through the process of heads-up digitizing. Road network data extracted from Google Earth was used as backdrop for the map. The location of each crash was geo-referenced with the aid of Google Earth to its corresponding location using the address and landmark information provided by patients. Place marks were used in representing locations of crashes. Each place mark represents either a single or multiple crash events since a number of road traffic crashes (RTCs) may be reported at the same location. The place marks were subsequently imported into ArcGIS software as KMZ file extension and were represented as point locations. Town planners familiar with the neighborhood, road network and Google Earth navigated the areas described by the crash victims to locate the precise crash locations. The data was aggregated to produce counts for the three categories of crashes at each location and added to each of the point features as attribute information.

The distributional pattern road traffic crashes (RTCs) were analyzed by category using Nearest Neighbor Statistic (Rn), with this equation:

Where

R_n = the nearest neighbor statistic

= the mean observed nearest neighbour distance

n = the total number of points

a = the total area

Moran-I Index was used to measure the level of spatial autocorrelation in the reported crashes in the metropolis. The locational proximity of data events were measured as direct distance between two points, while, inverse distance weighting method was used as a measure of locational proximity among neighboring points. Moran I-Index was calculated using the following equation:

Where

W_ij = the proximity weight of location I and location j with w_ii=0

x_i = the severity index at location j

x = the global mean value

n = the total number of road traffic accident

In addition, the statistical significance of Moran’s I was calculated using Z-Score formulae

Where (E [I]) - Expected Value for a random pattern

VAR ([I]) = Variances

Apart from the global Moran’s-I index, the Local Indicator of Spatial Autocorrelation (LISA) was used to identify neighborhoods with high and low incidence of RTCs. Getis-Ord statistics implemented through the Hot Spot analytical function in the ArcGIS software was used in this regard. The z-scores below –2 standard deviations are rendered dark blue, light blue for z-scores between –2 and –1 standard deviations, neutral for z-scores between –1 and +1 standard deviations, pink for z-scores between 1 and 2 standard deviations and z-scores above 2 standard deviations are bright red.

However, in order to be a statistically significant hot spot, a crash location will have a high value and be surrounded by other crash locations with similarly high values. The local sum for a feature and its neighbors was compared proportionally to the sum of all features; when the local sum was very different from the expected local sum, and that difference was too large to be the result of random chance, statistically significant z-score results. For statistically significant positive z-scores, the larger the z-score was, the more intense the clustering of high values (hot spot), while for statistically significant negative z-scores, the smaller the z-score was, the more intense the clustering of low values (cold spot). The maximum distance threshold obtained through the Moran–I Index was used in the Getis-Ord statistical analysis. Getis-Ord statistics and its associated z-score statistics were mathematically expressed by the following equation:

Where W_ij = the weight for the target neighbor pair

d = distance threshold

X_j = the severity index at location j

Paired sample T-Test was used to explore variations in RTCs between dual and non-dual carriage roads as well as between the inner city and the periphery of the metropolis. Analysis of variance (ANOVA) was however used to explore variations in RTCs across the eleven LGAs in the metropolis.

Results

Distribution of Road Traffic Crashes in Ibadan Metropolis

Out of 600 road traffic crashes recorded in this study, 492 (82.0%) crash locations were correctly georeferenced. The Local Government Areas (LGAs) with the highest mean crashes were Ibadan North West (41.0 ± 27.2), Ibadan South East (24.3 ± 12.7), and Akinyele (22.3 ± 15.4), while Lagelu (0.0 ± 0.0), Ona Ara (2.7 ± 4.6) and Oluyole (3.7 ± 3.2) LGAs had the least. There was however a significant difference in the number of crashes that occurred in the various LGAs (F = 2.943, P < 0.05). The Duncan test of groups in homogenous subsets showed that Lagelu, Ona Ara, Oluyole, Ibadan North East, Ibadan North, and Ibadan South West had an almost homogenous mean, while Ido had a distinct mean and Egbeda, Akinyele, Ibadan South East and Ibadan North West also had an almost similar mean. Thus, LGAs with similar or near similar crash figures exhibited similar pattern of crashes and as such can be grouped together. It is also clear that no distinct pattern could be discerned in crashes between the inner city and peri-urban LGAs in the metropolis.

Crashes on dual carriage roads accounted for 54.3%, while 45.7% occurred on non-dual carriage roads. Motorcycles account for 30.3% of RTCs on dual-carriage roads because of their increasing use for commercial purposes. There was no tricycle crash recorded during the study period. The average number of crashes on dual carriage roads within the metropolis was 8.8 ± 6.9, while non-dual carriage was 149.2 ± 75.2, which was significant (T = -9.710, P = 0.001).

The mean crashes in the inner city LGAs was 18.5 ± 18.7, while the peri-urban LGAs had 9.9 ± 11.5. No significant difference was observed despite an almost double mean difference between the two (F = 2.630, P = 0.115).

Pattern of Road Traffic Crashes in Ibadan Metropolis

Generally, the pattern of crashes were clustered in space and localized in the metropolis. More specifically,Table 1 showed that the motorbike crashes (MBC), motor vehicular crashes (MVC) and pedestrian crashes (PC) are also clustered in the metropolis (P = 0.0001). However, motorbike crashes showed greater clustering compared to the motor vehicular and pedestrian crashes. The motorbike (MBC) hotspots are mainly concentrated within the metropolitan areas, while motor vehicular crashes (MVC) are common along the major highways (dual carriage road). The Ojoo-Moniya axis of the Lagos-Ibadan express road, Iwo road interchange, Akinyo street, Bishop Akinyele road, Parliament road, Queen Elizabeth road, Oba Salawu Aminu roads, Oba Adebimpe road, Fajuyi road are some of the identified hotspots for the motor vehicular crashes (figure 2,figure 3,figure 4 and figure 5).

Table 1. Descriptive pattern of road traffic crashes in Ibadan metropolis.

Nearest Neighbour Analysis	Observed Mean Distance (Meters)	Expected Mean Distance (Meters)	Nearest Neighbour Ratio	Z-Scores	P-Values	Description of Pattern
MVC	444.325889	2454.662908	0.181013	-24.221824	0.0000	Clustered
MBC	1180.610855	3880.527703	0.304240	-18.586935	0.0000	Clustered
PC	1482.823991	4015.884712	0.369240	-9.189867	0.0000	Clustered
All Crashes	408.925289	2518.904056	0.162343	-35.545113	0.0000	Clustered

Figure 2 . Hotspot result for all crashes .

Figure 3 . Hotspots for pedestrian crashes.

Figure 4 . Hotspots for motor vehicle crashes .

Figure 5 . Hotspots for motorbike crashes .

Discussion

Compared to the 70% georeferencing achieved by Razzak et al11 using paper based maps, the use of Google Earth and the involvement of town planning officers familiar with Ibadan metropolis correctly georeferenced road traffic crashes/accidents locations in 82% of cases. Our hospital data is important because the development of an efficient database of road crashes can provide a means of identifying contributing factors in the metropolis. So far, the inability to effectively address the challenge of road traffic crashes in developing countries like Nigeria could be linked to the absence of reliable data3,20. However, freely available image from Google Earth help in overcoming the problem of unavailability and inadequacy of data often encountered in developing countries.

Road crashes were clustered in space with motorbike crashes showing greater clustering compared to motor vehicle and pedestrian crashes, contrary to the data from four South African cities which ranked pedestrian crashes above the others14. This clustered pattern indicated that there were locations that experienced more crashes than others and that different type of crashes predominate in different places. The Local Indicators of Spatial Autocorrelation (LISA) clearly showed localities with higher occurrence of certain type of crashes compared to others. Furthermore, road crashes significantly differed between dual and non-dual carriage roads (P=0.001). This finding has shown that higher frequency of vehicular movement leads to higher number of crashes, given that dual carriage roads host a higher number of vehicles compared to single lane roads; with the vehicles on higher speed than they do on the single lanes.

Higher number of motorbike crashes recorded is congruous with the findings of Arosanyin et al15. This has resulted from the increasing use of motorbikes for commercial purposes in the city. Commercial motorbikes, which started in Cross Rivers State, have assumed a greater dimension with unemployed youths earning a living thereby15^-17. Commercial motorcyclists and their passengers often do not use protective helmets thereby increasing their risks. It has been observed that the use of crash helmet has the potential of reducing head injuries by 50%18,19. Our data did not include tricycle crashes because they were newly introduced for transportation during the 6-month study period.

We hereby state that we have not conducted an overall crash mapping of Ibadan metropolis because our data source being hospital-based imposes a limitation. An exhaustive database should have inputs from the Federal Road Safety Corp and the Motor Traffic Division of Nigeria Police. Identifying crash clusters within the metropolis using GIS-aided technique can highlight contributory factors which can assist the government and its agencies to implement remedial measures to make driving safe at such spots and improve these roads. The installation of traffic lights at road intersections, better traffic monitoring and enforcing traditional measures such as speed limits, mandatory wearing of seatbelts and helmets have been shown to reduce crashes and injuries4.

Conclusions

In conclusion, this study showed that GIS is useful in identifying crash-prone areas within Ibadan metropolis and assist our understanding of contributory factors in crash occurrences. The information obtained is necessary in the development of strategies to reduce road traffic crashes and the associated injuries.