Data source
Using the Taiwan National Traffic Crash Dataset for the period 2011–2016, the current study examined fatalities and head injuries sustained by pedestrians in against- or with-traffic crashes. The Taiwan National Traffic Crash Dataset is owned and maintained by the National Police Agency, and the data are recorded after every road traffic crash is reported to the police. To record crash data, qualified and experienced police crash investigators complete three files, namely accident, vehicle, and victim files. Accident files contain general information regarding an accident, such as the time and date of the crash, weather condition, and road type. Vehicle files contain information regarding vehicle type, the first point of impact, and vehicle manoeuvres. Victim files contain data regarding casualties involved in crashes, such as age, sex, injury severity, injured body regions, licence status, walking direction, and alcohol use. Injury severity is classified into two categories: fatality and injury. In the victim file, pedestrians walking directions, namely crossing a street, walking against or with traffic, are recorded. Victims who die within 24 h as a result of an accident are classified as cases of fatality, whereas victims who sustain injuries, whether mild or severe, are classified as cases of injury. Crash investigators track injury data from hospitals for 30 days and thus obtain data on primary diagnosis on injured body regions and injury severity.
Figure 1 presents a flow chart of the sample selection from the Taiwan National Traffic Crash Dataset for the period 2011–2016. We extracted data of 98,269 pedestrian casualties from traffic crashes during this period. We excluded crashes in which pedestrian were crossing the streets rather than walking along the streets (n = 83,208). Furthermore, we focused on pedestrian crashes in which the crash partner was a private car, taxi, heavy-goods vehicle, bus or coach, or a motorcycle. As a result, crashes in which pedestrians were struck with bicycles were excluded (n = 321). A total of 15,061 pedestrians were involved in against- or with-traffic crashes. After removing crashes with missing data on pedestrian age, sex, and time or date of crash (n = 679), 14,382 pedestrian causalities remained. Of the 14,382 pedestrian casualties, 3633 were walking against traffic and 10,749 were walking with traffic, respectively.
Definitions of variables
We collected the following demographic data for casualties: sex, age (four groups: < 18, 18–40, 41–64, and ≥ 65 years), alcohol use (yes: breathalyser test results ≥0.15 mL/L or blood-alcohol consumption [BAC] level > 0.03%; no: breathalyser test results < 0.15 mL/L or BAC level ≤ 0.03%), and pedestrian walking direction (against traffic; with traffic). In Taiwan, people aged < 18 years are identified as teenagers, and they are not legally permitted to ride motorcycles or drive cars. People aged ≥65 years are identified as elderly individuals. In this study, we classified the remaining individuals into two age groups: 18–40 and 41–64 years. BAC data were obtained from police who conducted breathalyser tests or followed up for blood tests at hospitals. According to Taiwanese law, drivers with either breathalyser test results of ≥0.15 mL/L or BAC levels of > 0.03% are considered to be drunk driving. Data obtained from breathalyser tests or BAC levels were available only for motorists and not for pedestrians because, by law, only motorists involved in crashes are mandated to be tested for alcohol consumption. Data on injured body regions examined included injuries to the head or neck, upper or lower extremities, chest or abdomen, and spine.
The vehicle attribute considered was the crash partner (large vehicle: including buses, coaches, or heavy-goods vehicles; car: including private cars and taxis; and motorcycle). We examined three temporal factors, namely month of crash (spring/summer: March–August; autumn/winter: September–February), day of crash (weekday: Monday–Friday; weekend: Saturday–Sunday), and time of crash (rush hours: 0700–0859 and 1700–1859; nonrush hours: 0900–1659; evening hours: 1900–2359; and midnight/early morning: 0000–0659). The following roadway factors were considered: crash location (rural: roadways with speed limits of ≥51 km/h; urban: roadways with speed limits of ≤50 km/h), weather conditions (fine weather; adverse weather: rain or fog), street lighting condition (daylight, lit streets in darkness, and unlit streets in darkness), road surface condition (dry; slippery), and sight distance (adequate: sight distance was not obstructed; limited: sight distance was obstructed by obstacles such as road curvature, building, or tree).
Statistical analysis
The distribution of pedestrian injury severity according to a set of variables (e.g., human attributes, roadway or environmental factors, and vehicle characteristics) is first reported. We conducted chi-squared tests to examine the association between independent variables and pedestrian injury severity. We used chi-squared tests to discover variables that were significantly associated with the outcome variables (p < 0.2). These variables were then incorporated into the multivariate stepwise logistic regression models. To detect multi-collinearity among the variables (all categorical), we conducted a chi-squared independent test and estimated Cramer’s V [16].
Injuries to the head, which are generally devastating, were the focus of this study. Odds of head injuries were then estimated by using stepwise logistic regression models after controlling for a set of variables.