Geospatial analysis of mortality risk from road traffic crashes in Federal Capital Territory, Nigeria

Data collection

Injury data were analyzed from a HIPAA-compliant trauma registry of patients cared for at National Hospital Abuja (NHA) in FCT. Trauma registry data were de-identified and did not contain any protected health information. Trauma registry data collection began in 2010 following its inception as a joint project between Ahmadu Bello University Teaching Hospital in Zaria, NHA, and the Medical College of Wisconsin (MCW) (Cassidy et al., 2016). Currently only data from NHA in FCT are reported. The following data were included in the trauma registry analysis: presentation date, patient demographics, mechanism of injury, location of incident, and outcome (Holmes, 2019). The registry documents patient outcomes as alive (satisfactory, leaving against medical advice, admitted to an intensive care unit, awaiting surgery, pending, follow-up appointment, and discharged) and deceased (on arrival or while hospitalized).

Patient population for analysis

Patients sustaining traumatic injury between 1 January 2016 and 30 August 2018 were included in this analysis because these years had the most complete data on crash locations. Trauma patients who were not injured in a road traffic crash were excluded. Of 3 595 total trauma patients, 2 136 (59.4%) were injured in a road traffic crash. Crash patients whose crashes were geocoded (1 460 of 2 136 patients, 68.4%) were then assigned to one of the 60 wards within FCT based on the location of their crash, for spatial pattern mapping and cluster analysis.

Statistical and spatial analyses

Descriptive statistics were used to summarize the following patient characteristics: age, sex, mode of transportation, road type at crash location, and outcome (deceased or living). Age was compared between deceased persons and survivors using Wilcoxon rank-sum test. Wilcoxon rank-sum testing is preferable when comparing groups of continuous measurements with non-normal distributions. Sex, mode of transportation, and road type were compared using a chi-square test, which is widely used for comparing categorical data between groups. Multiple logistic regression analysis was used to evaluate risks of mortality in relation to age, sex, mode of transportation, and road type. Logistic regression analysis aims to isolate predictive factors of a binary variable, in this case traffic fatality. Goodness-of-fit of the logistic regression model was established using the Hosmer-Lemeshow test. Modes of transportation were divided into vehicle-users and pedestrians. Road types were divided into local roads and expressways or transition zones (areas where expressways transition into local roads, such as at junctions, bridges, and roundabouts).

Geographic Information Systems (GIS)-based exploratory spatial data analysis was conducted by comparing choropleth maps of road traffic mortality rate (defined as the number of patients involved in a crash who died after reaching NHA per 100 000 population), pedestrian injury rate (number of patients presenting to NHA who were injured in a crash as a pedestrian per 100 000 population), and the injury rate on expressway/transition zones (number of patients presenting to NHA who were injured in a crash on an expressway/transition zone per 100 000 population), as well as raw fatalities and injuries. This was followed by spatial autocorrelation analysis, in order to determine locations, distribution, patterns, and spatial associations of mortality, pedestrian injury, and expressways/transition zone injury rates. Spatial autocorrelation analysis measures the correlation of data values and locations simultaneously and determines whether spatial patterns, seen for instance in exploratory analysis, are significantly clustered or dispersed, or random.

The administrative level 3 geographic region (ward) is the smallest geographic unit in FCT available for mapping and spatial analysis. There are 60 wards throughout FCT, and ward boundaries were created from Nigeria Independent National Electoral Commission polling unit maps (INEC, 2023) (Figure 1). As no census data are available at ward levels, population counts were derived for each ward using the WorldPop Africa dataset, where people per hectare were estimated for 2015 by using the random forest regression tree-based mapping approach (Stevens et al., 2015), with national totals adjusted to align with United Nations population division estimates (DESA, 2017; WorldPop, 2018). The estimated ward-level population is used for calculating the crash, injury, and mortality rates per 100 000 population. When referring to population, the remainder of this paper removes the word ‘estimated’. Results were considered statistically significant at p < 0.05. SAS OnDemand for Academics software (SAS Institute, Cary, NC) was used to conduct statistical analysis, and ArcGIS (Esri, Redlands, CA) and GeoDa (GitHub, San Francisco, CA) software were used to conduct geospatial analysis.

Statistical analysis

A total of 2 136 patients treated at NHA for road traffic injuries were included in analysis. Table 1 summarizes descriptive statistics of patients and crashes. The median patient age was 30 years, ranging from 0 to 78 years. Most patients were male (n = 1 608, 75.3%). Approximately three quarters of the patients were injured in a vehicle (vehicles with ≥ four wheels, tricycles, motorcycles, and bicycles) (n = 1 615, 75.6%), while 521 (24.4%) of crash patients were injured as pedestrians. A total of 574 (26.9%) injuries occurred on expressways/transition zones (497 occurred on an expressway, and 77 occurred at junctions, bridges, and roundabouts), as compared to 1 562 (73.1%) on local roads. Nearly one quarter (n = 121) of pedestrians were injured on an expressway/transition zone. 118 crash patients (5.5%) died after reaching NHA.

Table 2 summarizes results of bivariate analysis of patient and crash characteristics by patient outcome (i.e., whether the patient lived or died in hospital). Mode of transportation and road type were both significantly associated with patient outcome. A significantly higher proportion of pedestrians died at the hospital (n = 51, 9.8%) than vehicle-users (n = 67, 4.1%) (p < .0001). A significantly higher proportion of patients involved in a crash on an expressway/transition zone died (n = 45, 7.8%) than those who crashed on a local road (n = 73, 4.7%) (p = 0.0045).

Results of multiple logistic regression are summarized in Table 3. Crash patient fatality at NHA was positively associated with both involvement in a crash as a pedestrian (p < .0001) and involvement on an expressway/transition zone.

Spatial analysis

Spatial patterns and clustering of the road traffic mortality rate, pedestrian injury rate, and expressway/transition zone injury rate were analyzed. Figure 2 shows the road traffic mortality rate of crash patients after presenting to NHA, by FCT administrative ward, per 100 000 population, demonstrating the highest rates in the Garki, Nyanya, and Kashi wards descendingly.

Figure 3 shows the road network in Abuja city center, comprising Garki, Nyanya, Wuse, City Centre, Karu, and Gwarimpa wards, and parts of Orozo and Dutse Alhaji, overlaying the mortality rate map, with the location of NHA.

Figure 4 depicts the raw number of crash patient deaths per administrative ward, in side-by-side comparison with the number of crash patients who presented to NHA per ward, demonstrating similar spatial patterning between death and crash occurrences.

Figure 5; Figure 4 compare the road traffic mortality rate with pedestrian and expressway/transition zone injury rates, respectively.

These choropleths suggest possible clustering and spatial autocorrelation of the road traffic mortality rate, pedestrian injury rate, and expressways/transition zone injury rate.

Spatial autocorrelation

Spatial autocorrelation analysis resulted in statistically significant clustering of the population-adjusted road traffic mortality rate throughout FCT at the level of 0.01 (Moran’s I = 0.31). This indicates that the crash location clustering pattern for patients who die in the hospital is not likely due to chance. Similarly, pedestrian and expressway/transition zone injury rates were significantly clustered at the level 0.01 (Moran’s I = 0.34 and 0.18, respectively).

Figure 9; Figure 8; Figure 7 show the maps of local spatial autocorrelation analyses, with high-high clustering in red (Dutse Alhaji, Wuse, Garki, City Centre, and Nyanya wards on each of the maps) and low-low in blue (Yenche ward on the pedestrian injury map). High-high clustering indicates areas of high mortality or injury rates surrounded by areas with similarly high rates, and low-low clustering indicates areas of low mortality or injury rates surrounded by areas with similarly low rates. Cluster maps are presented side-by-side with significance maps, which indicate that these clusters are statistically significant.

Mode of transportation

Although some researchers opine that the proportion of total crashes in Nigeria involving a pedestrian is comparatively low (Labinjo et al., 2009), others argue that pedestrian injuries are under-reported, as crashes involving pedestrians are frequently recorded as vehicular crashes (Aluko, 2018; Solagberu et al., 2015). Notably, in a recent survey exploring road-users’ attitudes in 32 countries, Nigeria had the second lowest safety perception score for pedestrians (Yannis et al., 2020). Pedestrian injuries may also more commonly result in death than injuries sustained by vehicle-users. Pedestrians do not wear helmets or protective clothing, and when contact with a vehicle occurs, impact on vital anatomical structures is unimpeded. Also, the high speeds at which vehicles are traveling on expressways when impact occurs with a pedestrian may lead to a higher likelihood of death. Crashes involving pedestrians have been shown to achieve greater than double the proportion of crash-related deaths in LMICs, compared to higher income countries (Sundet et al., 2018). Whereas just under 10% (n = 521) of total pedestrians in this study died at the hospital, 43% (n = 22) of pedestrians injured on expressways/transition zones died at the hospital. Further, severer injuries like these respond most favorably to immediate medical intervention, which may not be available in regions like FCT, of emergency medical services deficit (Solagberu et al., 2009). Pedestrian injuries and fatalities in Nigeria will likely increase over time as rising rural-urban migratory trends continue, population congestion increases, and higher numbers of pedestrians are exposed to a higher risk of vehicular contact (Buor, 2018).

Road type

The increased mortality from crashes on expressways/transition zones is not surprising given the traffic congestion on these types of roads and high-speed driving. Motorcycle and tricycle taxis are prohibited on expressways, so throngs of pedestrians are commonly found in transition zones, especially in highly populated areas in and around Abuja city center, as passenger interchange occurs here. Pedestrians are also commonly found crossing busy expressways on foot. Walking bridges have been built in some areas to provide safe transit for pedestrians crossing expressways, but it is conceivable that pedestrians may find them inconvenient to use and opt for a quicker method. A high number of pedestrians tend to cross expressways at night when markets along the expressways are open to customers who are just finishing their workday. This may place them at increased risk. Crashes on expressways during these times may be more severe, due to the higher numbers of pedestrians and limited visibility for both pedestrians and vehicle-users.

Next steps

There are several implications from these findings. Significant high-high spatial clustering of road traffic mortality rate, pedestrian injury rate, and injury rate on expressways/transition zones in the same five wards (Dutse Alhaji, Wuse, Garki, City Centre, and Nyanya) demonstrates the hotspots of injuries which lead to fatalities for crash patients presenting to NHA. A high density of transition zones is apparent in these five wards, which may place a large percentage of the pedestrian population in these wards at greater risk of injury and death and could explain why these specific wards comprise the cluster. These maps can serve as a blueprint of geographic prioritization for public and community health specialists and policymakers attempting to effect the highest impact at the lowest expense. Such efforts may include: studying pedestrian behaviorisms in greater depth using the Pedestrian Behavior Questionnaire (Bayomi et al., 2022), providing greater incentive for pedestrians in these five wards to use walking bridges over expressways; restructuring pedestrian/public transit interchange locations to offset them from junctions, bridges, and roundabouts in these five wards; and improving roadway emergency medical services, beginning in these five wards.

Limitations

This study has limitations. First, the data used were specific to NHA; they are not representative of the entire FCT. However, NHA houses FCT’s largest trauma center with a broad catchment area. These data can be triangulated with FRSC data to enhance the external validity of findings (Brennan & Shepherd, 2012). Second, this was a retrospective analysis and while the data were relatively complete upon admission, some entries were not geocoded, and there were missing data related to mortality, including post-discharge mortality. Descriptive and regression analyses were performed on both the total crash patient dataset (n = 2 136) and georeferenced data (n = 1 460) for comparison, with nearly identical results. Finally, although a spatial disparity in mortality rate was evident and there are known risk factors such as density of expressways, testing the effects of those factors was not feasible because of unavailable spatially matched data. It is important to continue the efforts to create reliable spatial datasets for further analyses, which would improve understanding of the interactions of environment and human behaviors and health outcomes.