Injury severity analysis of electric bike crashes in Changsha, Hunan Province: taking different lighting conditions into consideration

Abstract

With the increasing use of electric bikes, electric bike crashes occur frequently. Analysing the influencing factors of electric bike crashes is an effective way to reduce mortality and improve road safety. In this paper, spatial analysis is performed by geographic information system (GIS) to present the hot spots of electric bike crashes during daylight and darkness in Changsha, Hunan Province, China. Based on the Ordered Probit (OP) model, we studied the risk factors that led to different severities of electric bike crashes. The results show that the main influencing variables include age, illegal behaviour, collision type and road factors. During daylight and darkness, elderly electric bike riders over the age of 65 years have a higher probability of fatal crashes. Not following traffic signals and reverse driving are significantly related to the severity of riders' injuries. In darkness, frontal collisions are significant factors causing rider injury. In daylight, more serious crashes will be caused in bend and slope road segments. In darkness, roads with no physically separated bicycle lanes increases the risk of riders being injured. Electric bike crashes are mainly concentrated in the commercial, public service and residential areas in the main urban area. In suburbs at darkness, electric bike riders are more likely to be seriously injured. Adding protection measures, such as improved lighting, non-motorized lane facilities and interventions targeting illegal behaviour in the hot spot areas can effectively reduce the number of electric bike crashes in complex traffic.

1. Introduction

Globally, cyclists and pedestrians contribute to 26% of all deaths related to traffic accidents and motorized two- and three-wheeled vehicles contribute to 28%. In Southeast Asia and the Western Pacific, most deaths occurred among two- and three-wheeled motorized bicycle passengers, contributing to 43% and 36% of all deaths related to traffic accidents, respectively [1]. With the public's attention to a low-carbon lifestyle, the electric bike, which takes economy and convenience into consideration, has become an important means of transportation for many people. In China, from 2007 to 2016, more than 30 million electric bikes were added to the road each year, and car ownership increased by an average of 15 million vehicles during the same period. Electric bikes have become one of the main modes of travel for urban residents in China; especially in small and medium-sized cities, the proportion of electric bikes travel is as high as 10%–30% [2]. The traffic safety risk of electric bike is accumulating rapidly. From 2012 to 2016, there were 193,000 road traffic crashes involving electric bikes in China, resulting in 37,700 deaths. Among them, the proportion of active crashes and fatalities of electric bikes was as high as 29.1% and 22.3%, respectively [3]. Electric bike crash is the fastest rising group in the death toll of road traffic crashes.

At the same time, night traffic crashes have become a major problem in road traffic safety rectification. Riding in daylight and at darkness has obvious risks for powered two-wheelers (PTW) users [4]. In China, traffic crashes at night contribute to 41.76% of the number of urban road traffic crashes, and the death toll contributes to 42.63% of the total number of traffic deaths [5]. Some studies have explored the relationship between ambient light and crash risk from the perspective of illumination. Due to the low light after dark, drivers often face the challenge of looking at the external road environment from the inside of the vehicle. The light condition is related to the increase of injury severity, and the death toll at night is much higher than that during daylight [6,7]. The night-time road death rate per distance travelled is about three to four times that of the daytime [8]. Compared with daytime crashes, drivers caused serious injuries at night, especially in younger age groups [9,10]. Moreover, Ariannezhad and Wu [11] considered the influence of light and weather characteristics on crash severity and used four independent polynomial logit models to estimate the specific weather (rainstorm or sunny) and lighting conditions (day or night).

Vulnerable road users, such as pedestrians, electric bike riders and motorcyclists, are vulnerable to injury and death when they collide with vehicles at night due to poor visibility [12,13]. For example, Hoque [14] has found that adult cyclists over the age of 17 are easily involved in fatal traffic crashes. The proportion of cyclists who die at night increases with age, and almost 80% of night crashes occur on arterial roads, most of which are in high-speed restricted areas. Robbins and Fotios [15] found that the risk of motorcycle crashes is much higher after dark compared to that during daylight. In the case of low-speed restrictions and T-shaped intersections and intersections controlled by road signs, the risk to motorcycles after darkness is significantly higher than that of cars. Pedestrians and electric bike riders died because the driver did not see them and take evasive action in time [16]. Visibility of vulnerable road users is an important road safety issue, especially in streets without lighting at night.

Non-visual factors, such as drinking, fatigue and speeding, usually occur at the same time as driving at night, increasing the risk of fatal crashes [17,18]. People who are active during the day are usually at greater risk of driving crashes at night than during the day. The causes of crashes include not only worker-centred determinants, such as circadian time related to cognitive and physical performance rhythm, or recent sleep history related to sleep tendency and fatigue [19]. Abegaz et al. [20] found that fatigue leads to a higher severity of crashes and a higher risk of injury for drivers who drink alcohol. In the absence of street lights, rainy weather and night driving will increase the severity of the injury. Oksanen et al. [21] found that most of the injuries related to electric bikes are caused by driving under the strong influence of alcohol, and most of them occur on weekend nights when the riders do not wear a helmet. Further, Liu et al. [22] found that the type of crash and the central isolation facilities are key factors affecting crash severity in darkness. Chen and Fanny [23] found that riders involved in rear-end, frontal, angular and left-turn collisions may be seriously injured when studying the influencing factors of different injury outcomes of night collisions.

Location information is usually included in electric bike crashes, which gives the crashes spatial characteristics. However, the existing research seldom details spatial analysis of electric bike crashes. Night traffic crashes are usually more serious than daytime traffic crashes, but risk factors related to such crashes and the mode of night traffic fatalities and injuries have not been recorded. Therefore, this study discusses the spatial relationship of electric bike crashes based on the geographic information system (GIS) technology. By using an Ordered Probit model to identify and quantify the important variables that affect the severity of riders’ injuries between daylight and darkness, this paper provides a basis for improving the traffic safety of electric bikes and put forward countermeasures to curb the increasingly serious electric bike crashes.

2. Analysis

2.1 Data preparation

The crash data in this paper come from the traffic crash database of the traffic management department of Hunan Province. This study extracted the traffic crash cases in Changsha, Hunan Province from 2014 to 2016 from the database. The crash data that meet the following condition is selected: the crash involved an electric bike and a car. Finally, we extracted 846 electric bike crashes. The crash data include information, such as crash time, location, characteristics of electric bike riders, drivers, road and environmental characteristics.

2.1.1 The basis for dividing the time between daylight and darkness

Some studies tend to compare the frequency of road traffic crashes during daylight and darkness to reveal the impact of ambient light levels on the occurrence of crashes. However, the definitions of darkness and daylight in previous studies are not always precise and there is no clear definition. Most classifications are carried out by the police at the scene of the crash [24–26]. In this study, we searched the sunrise and sunset times in the historical meteorological records of Changsha from 2014 to 2016, and compared them with the time and ambient light conditions recorded by the police. In this study, the crash period from sunrise to sunset is divided into daylight and the crash period from sunset to sunrise the next day is divided into darkness. Finally, 633 crashes occurred in daylight and 213 in darkness.

2.1.2 Characteristics of electric bike crash

We considered various indicators, such as the following crash information which shows the number of cases of electric bike crashes in darkness and daylight. Table 1 includes electric bike riders and collision characteristics, such as rider's age, gender, household type, personnel type, collision type, crash responsibility, illegal behaviour and injury severity. At the same time, driver characteristics include driver's age, gender, driving age, whether alcohol had been used and whether seat belts had been used. It also includes road and environmental characteristics, such as weather condition, visibility, type of land use, surface condition, season, type of day, road type, road alignment, types of road section, road physical isolation, median strip, traffic control and position of the road cross-section.

Table 1.

Open in new tab

Descriptive statistics of crash characteristics.

Category	Characteristic	Description	Daylight		Darkness
Category	Characteristic	Description	Count	Percentage/%	Count	Percentage/%
Injury severity	N/A	1: No injury	23	3.6	16	7.5
		2: Slight	559	88.3	167	78.4
		3: Serious	8	1.3	3	1.4
		4: Fatal	43	6.8	27	12.7
Rider and collision	Rider gender	1: Male	376	59.4	145	68.1
characteristics		2: Female	257	40.6	68	31.9
	Rider age	1: < = 25	47	7.4	28	13.1
		2: 26–35	108	17.1	32	15.0
		3: 36–45	120	19.0	57	26.8
		4: 46–55	164	25.9	46	21.6
		5: 56–65	142	22.4	42	19.7
		6: 65+	52	8.2	8	3.8
	Rider household type	1: Agricultural	585	92.4	196	92.0
		2: Non-agricultural	48	7.6	17	8.0
	Rider personnel type	1: Express delivery employees	101	16.0	26	12.2
		2: Worker	33	5.2	10	4.7
		3: Farmer	93	14.7	19	8.9
		4: Self-employed	225	35.5	86	40.4
		5: Other	181	28.6	72	33.8
	Illegal behaviour of riders	1: Not following traffic signals	64	10.1	22	10.3
		2: Reverse driving	52	8.2	18	8.5
		3: Cross the motorway without getting off	46	7.3	14	6.6
		4: Failure to give way as required	48	7.6	9	4.2
		5: Driving off the right side of the roadway¹	31	4.9	21	9.9
		6: Not driving in the non-motorized Lane	35	5.5	14	6.6
		7: Other illegal behaviour	37	5.8	14	6.6
		8: Absent	320	50.6	101	47.3
	Collision type	1: Side impact (angle uncertain)	110	17.4	29	13.6
		2: Side impact (opposite)	60	9.5	20	9.4
		3: Side impact (same direction)	115	18.2	41	19.2
		4: Side impact (right angle)	164	25.9	42	19.7
		5: Frontal impact	68	10.7	23	10.8
		6: Rear impact	36	5.7	24	11.3
		7: Scratch	10	1.6	5	2.3
		8: Other angles crash	70	11.0	29	13.7
	Crash responsibility	1: No responsibility	263	41.5	76	35.7
		2: Main responsibility	117	18.5	42	19.7
		3: Equal	132	20.9	42	19.7
		4: Secondary	115	18.2	48	22.6
		5: Full responsibility	6	0.9	5	2.3
Driver characteristics	Driver gender	1: Male	542	85.6	184	86.4
		2: Female	91	14.4	29	13.6
	Driver age	1: < = 25	81	12.8	36	16.9
		2: 26–35	231	36.5	96	45.1
		3: 36–45	196	31.0	48	22.5
		4: 46–55	109	17.2	29	13.6
		5: 56–66	16	2.5	4	1.9
	Driver driving age	1: 0–4	269	42.5	99	46.5
		2: 5–9	195	30.8	62	29.1
		3: 10–14	98	15.5	28	13.1
		4: 15–19	42	6.6	18	8.5
		5: >= 20	29	4.6	6	2.8
	Drunk driving	1: Yes	128	20.2	53	24.9
		2: No	505	79.8	160	75.1
	Use seat belts	1: No	69	10.9	19	8.9
		2: Yes	564	89.1	194	91.1
Road and environment	Road alignment	1: Bend and slope	98	15.5	31	14.6
characteristics		2: Straight	535	84.5	182	85.4
	Road type	1: General urban roads	417	65.8	149	70.0
		2: Urban expressway	43	6.8	24	11.3
		3: Highway²	110	17.4	26	12.2
		4: Other roads	63	10.0	14	6.5
	Types of road section	1: Intersections	142	22.5	53	24.9
		2: Ordinary road segment	442	69.8	145	68.1
		3: Special segments³	49	7.7	15	7.0
	Type of land use	1: Commercial area	143	22.6	54	25.4
		2: Industrial area	38	6.0	17	8.0
		3: Public service area⁴	159	25.1	50	23.4
		4: Suburb	124	19.6	31	14.6
		5: Residential area	169	26.7	61	28.6
	Road physical isolation	1: Absent	389	61.4	102	47.9
		2: PSBR⁵	21	3.3	9	4.2
		3: Median divider	165	26.1	69	32.4
		4: MD&PSBR⁶	58	9.2	33	15.5
	Median strip	1: Absent	377	59.6	103	48.4
		2: Greenbelt	125	19.7	35	16.4
		3: Guardrail	131	20.7	75	35.2
	POTRCS⁷	1: Motorway	433	68.4	163	76.5
		2: Mixed road	134	21.2	35	16.5
		3: Bicycle lane	35	5.5	9	4.2
		4: Other	31	4.9	6	2.8
	Traffic control	1: Absent	131	20.7	29	13.6
		2: Present	502	79.3	184	86.4
	Surface condition	1: Stagnant water	9	1.4	2	.9
		2: Wet	96	15.2	54	25.4
		3: Dry	528	83.4	157	73.7
	Weather condition	1: Rainy/foggy/snowy	90	14.2	43	20.2
		2: Overcast	127	20.1	64	30.0
		3: Sunny	416	65.7	106	49.8
	Visibility	1: <50 m	38	6.0	55	25.8
		2: 50–100 m	178	28.1	91	42.7
		3: 100–200 m	183	28.9	33	15.5
		4: >= 200 m	234	37.0	34	16.0
	Season	1: Spring	164	25.9	46	21.6
		2: Summer	177	28.0	60	28.2
		3: Autumn	166	26.2	64	30.0
		4: Winter	126	19.9	43	20.2
	Type of day	1: Weekday	486	76.8	163	76.5
		2: Weekend	147	23.2	50	23.5

Category	Characteristic	Description	Daylight		Darkness
Category	Characteristic	Description	Count	Percentage/%	Count	Percentage/%
Injury severity	N/A	1: No injury	23	3.6	16	7.5
		2: Slight	559	88.3	167	78.4
		3: Serious	8	1.3	3	1.4
		4: Fatal	43	6.8	27	12.7
Rider and collision	Rider gender	1: Male	376	59.4	145	68.1
characteristics		2: Female	257	40.6	68	31.9
	Rider age	1: < = 25	47	7.4	28	13.1
		2: 26–35	108	17.1	32	15.0
		3: 36–45	120	19.0	57	26.8
		4: 46–55	164	25.9	46	21.6
		5: 56–65	142	22.4	42	19.7
		6: 65+	52	8.2	8	3.8
	Rider household type	1: Agricultural	585	92.4	196	92.0
		2: Non-agricultural	48	7.6	17	8.0
	Rider personnel type	1: Express delivery employees	101	16.0	26	12.2
		2: Worker	33	5.2	10	4.7
		3: Farmer	93	14.7	19	8.9
		4: Self-employed	225	35.5	86	40.4
		5: Other	181	28.6	72	33.8
	Illegal behaviour of riders	1: Not following traffic signals	64	10.1	22	10.3
		2: Reverse driving	52	8.2	18	8.5
		3: Cross the motorway without getting off	46	7.3	14	6.6
		4: Failure to give way as required	48	7.6	9	4.2
		5: Driving off the right side of the roadway¹	31	4.9	21	9.9
		6: Not driving in the non-motorized Lane	35	5.5	14	6.6
		7: Other illegal behaviour	37	5.8	14	6.6
		8: Absent	320	50.6	101	47.3
	Collision type	1: Side impact (angle uncertain)	110	17.4	29	13.6
		2: Side impact (opposite)	60	9.5	20	9.4
		3: Side impact (same direction)	115	18.2	41	19.2
		4: Side impact (right angle)	164	25.9	42	19.7
		5: Frontal impact	68	10.7	23	10.8
		6: Rear impact	36	5.7	24	11.3
		7: Scratch	10	1.6	5	2.3
		8: Other angles crash	70	11.0	29	13.7
	Crash responsibility	1: No responsibility	263	41.5	76	35.7
		2: Main responsibility	117	18.5	42	19.7
		3: Equal	132	20.9	42	19.7
		4: Secondary	115	18.2	48	22.6
		5: Full responsibility	6	0.9	5	2.3
Driver characteristics	Driver gender	1: Male	542	85.6	184	86.4
		2: Female	91	14.4	29	13.6
	Driver age	1: < = 25	81	12.8	36	16.9
		2: 26–35	231	36.5	96	45.1
		3: 36–45	196	31.0	48	22.5
		4: 46–55	109	17.2	29	13.6
		5: 56–66	16	2.5	4	1.9
	Driver driving age	1: 0–4	269	42.5	99	46.5
		2: 5–9	195	30.8	62	29.1
		3: 10–14	98	15.5	28	13.1
		4: 15–19	42	6.6	18	8.5
		5: >= 20	29	4.6	6	2.8
	Drunk driving	1: Yes	128	20.2	53	24.9
		2: No	505	79.8	160	75.1
	Use seat belts	1: No	69	10.9	19	8.9
		2: Yes	564	89.1	194	91.1
Road and environment	Road alignment	1: Bend and slope	98	15.5	31	14.6
characteristics		2: Straight	535	84.5	182	85.4
	Road type	1: General urban roads	417	65.8	149	70.0
		2: Urban expressway	43	6.8	24	11.3
		3: Highway²	110	17.4	26	12.2
		4: Other roads	63	10.0	14	6.5
	Types of road section	1: Intersections	142	22.5	53	24.9
		2: Ordinary road segment	442	69.8	145	68.1
		3: Special segments³	49	7.7	15	7.0
	Type of land use	1: Commercial area	143	22.6	54	25.4
		2: Industrial area	38	6.0	17	8.0
		3: Public service area⁴	159	25.1	50	23.4
		4: Suburb	124	19.6	31	14.6
		5: Residential area	169	26.7	61	28.6
	Road physical isolation	1: Absent	389	61.4	102	47.9
		2: PSBR⁵	21	3.3	9	4.2
		3: Median divider	165	26.1	69	32.4
		4: MD&PSBR⁶	58	9.2	33	15.5
	Median strip	1: Absent	377	59.6	103	48.4
		2: Greenbelt	125	19.7	35	16.4
		3: Guardrail	131	20.7	75	35.2
	POTRCS⁷	1: Motorway	433	68.4	163	76.5
		2: Mixed road	134	21.2	35	16.5
		3: Bicycle lane	35	5.5	9	4.2
		4: Other	31	4.9	6	2.8
	Traffic control	1: Absent	131	20.7	29	13.6
		2: Present	502	79.3	184	86.4
	Surface condition	1: Stagnant water	9	1.4	2	.9
		2: Wet	96	15.2	54	25.4
		3: Dry	528	83.4	157	73.7
	Weather condition	1: Rainy/foggy/snowy	90	14.2	43	20.2
		2: Overcast	127	20.1	64	30.0
		3: Sunny	416	65.7	106	49.8
	Visibility	1: <50 m	38	6.0	55	25.8
		2: 50–100 m	178	28.1	91	42.7
		3: 100–200 m	183	28.9	33	15.5
		4: >= 200 m	234	37.0	34	16.0
	Season	1: Spring	164	25.9	46	21.6
		2: Summer	177	28.0	60	28.2
		3: Autumn	166	26.2	64	30.0
		4: Winter	126	19.9	43	20.2
	Type of day	1: Weekday	486	76.8	163	76.5
		2: Weekend	147	23.2	50	23.5

Note: ¹Driving off the right side of the roadway means that electric bike riders do not rider on the right side of the roadway when there is no non-motorized lane.

²Highway include Grade one, Grade two, Grade three, Grade four and Substandard way.

³Special segments include tunnels, bridges, elevated segments and narrowed segments.

⁴Public service area includes educational and cultural facilities, hospitals, subway entrance, bus stop and parks.

⁵PSBR denotes physically separated bicycle roadways.

⁶MD&PSBR denotes the median divider and physically separated bicycle roadways.

⁷POTRCS denotes the variable ‘Position of the road cross-section’.

Table 1.

Open in new tab

Descriptive statistics of crash characteristics.

Category	Characteristic	Description	Daylight		Darkness
Category	Characteristic	Description	Count	Percentage/%	Count	Percentage/%
Injury severity	N/A	1: No injury	23	3.6	16	7.5
		2: Slight	559	88.3	167	78.4
		3: Serious	8	1.3	3	1.4
		4: Fatal	43	6.8	27	12.7
Rider and collision	Rider gender	1: Male	376	59.4	145	68.1
characteristics		2: Female	257	40.6	68	31.9
	Rider age	1: < = 25	47	7.4	28	13.1
		2: 26–35	108	17.1	32	15.0
		3: 36–45	120	19.0	57	26.8
		4: 46–55	164	25.9	46	21.6
		5: 56–65	142	22.4	42	19.7
		6: 65+	52	8.2	8	3.8
	Rider household type	1: Agricultural	585	92.4	196	92.0
		2: Non-agricultural	48	7.6	17	8.0
	Rider personnel type	1: Express delivery employees	101	16.0	26	12.2
		2: Worker	33	5.2	10	4.7
		3: Farmer	93	14.7	19	8.9
		4: Self-employed	225	35.5	86	40.4
		5: Other	181	28.6	72	33.8
	Illegal behaviour of riders	1: Not following traffic signals	64	10.1	22	10.3
		2: Reverse driving	52	8.2	18	8.5
		3: Cross the motorway without getting off	46	7.3	14	6.6
		4: Failure to give way as required	48	7.6	9	4.2
		5: Driving off the right side of the roadway¹	31	4.9	21	9.9
		6: Not driving in the non-motorized Lane	35	5.5	14	6.6
		7: Other illegal behaviour	37	5.8	14	6.6
		8: Absent	320	50.6	101	47.3
	Collision type	1: Side impact (angle uncertain)	110	17.4	29	13.6
		2: Side impact (opposite)	60	9.5	20	9.4
		3: Side impact (same direction)	115	18.2	41	19.2
		4: Side impact (right angle)	164	25.9	42	19.7
		5: Frontal impact	68	10.7	23	10.8
		6: Rear impact	36	5.7	24	11.3
		7: Scratch	10	1.6	5	2.3
		8: Other angles crash	70	11.0	29	13.7
	Crash responsibility	1: No responsibility	263	41.5	76	35.7
		2: Main responsibility	117	18.5	42	19.7
		3: Equal	132	20.9	42	19.7
		4: Secondary	115	18.2	48	22.6
		5: Full responsibility	6	0.9	5	2.3
Driver characteristics	Driver gender	1: Male	542	85.6	184	86.4
		2: Female	91	14.4	29	13.6
	Driver age	1: < = 25	81	12.8	36	16.9
		2: 26–35	231	36.5	96	45.1
		3: 36–45	196	31.0	48	22.5
		4: 46–55	109	17.2	29	13.6
		5: 56–66	16	2.5	4	1.9
	Driver driving age	1: 0–4	269	42.5	99	46.5
		2: 5–9	195	30.8	62	29.1
		3: 10–14	98	15.5	28	13.1
		4: 15–19	42	6.6	18	8.5
		5: >= 20	29	4.6	6	2.8
	Drunk driving	1: Yes	128	20.2	53	24.9
		2: No	505	79.8	160	75.1
	Use seat belts	1: No	69	10.9	19	8.9
		2: Yes	564	89.1	194	91.1
Road and environment	Road alignment	1: Bend and slope	98	15.5	31	14.6
characteristics		2: Straight	535	84.5	182	85.4
	Road type	1: General urban roads	417	65.8	149	70.0
		2: Urban expressway	43	6.8	24	11.3
		3: Highway²	110	17.4	26	12.2
		4: Other roads	63	10.0	14	6.5
	Types of road section	1: Intersections	142	22.5	53	24.9
		2: Ordinary road segment	442	69.8	145	68.1
		3: Special segments³	49	7.7	15	7.0
	Type of land use	1: Commercial area	143	22.6	54	25.4
		2: Industrial area	38	6.0	17	8.0
		3: Public service area⁴	159	25.1	50	23.4
		4: Suburb	124	19.6	31	14.6
		5: Residential area	169	26.7	61	28.6
	Road physical isolation	1: Absent	389	61.4	102	47.9
		2: PSBR⁵	21	3.3	9	4.2
		3: Median divider	165	26.1	69	32.4
		4: MD&PSBR⁶	58	9.2	33	15.5
	Median strip	1: Absent	377	59.6	103	48.4
		2: Greenbelt	125	19.7	35	16.4
		3: Guardrail	131	20.7	75	35.2
	POTRCS⁷	1: Motorway	433	68.4	163	76.5
		2: Mixed road	134	21.2	35	16.5
		3: Bicycle lane	35	5.5	9	4.2
		4: Other	31	4.9	6	2.8
	Traffic control	1: Absent	131	20.7	29	13.6
		2: Present	502	79.3	184	86.4
	Surface condition	1: Stagnant water	9	1.4	2	.9
		2: Wet	96	15.2	54	25.4
		3: Dry	528	83.4	157	73.7
	Weather condition	1: Rainy/foggy/snowy	90	14.2	43	20.2
		2: Overcast	127	20.1	64	30.0
		3: Sunny	416	65.7	106	49.8
	Visibility	1: <50 m	38	6.0	55	25.8
		2: 50–100 m	178	28.1	91	42.7
		3: 100–200 m	183	28.9	33	15.5
		4: >= 200 m	234	37.0	34	16.0
	Season	1: Spring	164	25.9	46	21.6
		2: Summer	177	28.0	60	28.2
		3: Autumn	166	26.2	64	30.0
		4: Winter	126	19.9	43	20.2
	Type of day	1: Weekday	486	76.8	163	76.5
		2: Weekend	147	23.2	50	23.5

Category	Characteristic	Description	Daylight		Darkness
Category	Characteristic	Description	Count	Percentage/%	Count	Percentage/%
Injury severity	N/A	1: No injury	23	3.6	16	7.5
		2: Slight	559	88.3	167	78.4
		3: Serious	8	1.3	3	1.4
		4: Fatal	43	6.8	27	12.7
Rider and collision	Rider gender	1: Male	376	59.4	145	68.1
characteristics		2: Female	257	40.6	68	31.9
	Rider age	1: < = 25	47	7.4	28	13.1
		2: 26–35	108	17.1	32	15.0
		3: 36–45	120	19.0	57	26.8
		4: 46–55	164	25.9	46	21.6
		5: 56–65	142	22.4	42	19.7
		6: 65+	52	8.2	8	3.8
	Rider household type	1: Agricultural	585	92.4	196	92.0
		2: Non-agricultural	48	7.6	17	8.0
	Rider personnel type	1: Express delivery employees	101	16.0	26	12.2
		2: Worker	33	5.2	10	4.7
		3: Farmer	93	14.7	19	8.9
		4: Self-employed	225	35.5	86	40.4
		5: Other	181	28.6	72	33.8
	Illegal behaviour of riders	1: Not following traffic signals	64	10.1	22	10.3
		2: Reverse driving	52	8.2	18	8.5
		3: Cross the motorway without getting off	46	7.3	14	6.6
		4: Failure to give way as required	48	7.6	9	4.2
		5: Driving off the right side of the roadway¹	31	4.9	21	9.9
		6: Not driving in the non-motorized Lane	35	5.5	14	6.6
		7: Other illegal behaviour	37	5.8	14	6.6
		8: Absent	320	50.6	101	47.3
	Collision type	1: Side impact (angle uncertain)	110	17.4	29	13.6
		2: Side impact (opposite)	60	9.5	20	9.4
		3: Side impact (same direction)	115	18.2	41	19.2
		4: Side impact (right angle)	164	25.9	42	19.7
		5: Frontal impact	68	10.7	23	10.8
		6: Rear impact	36	5.7	24	11.3
		7: Scratch	10	1.6	5	2.3
		8: Other angles crash	70	11.0	29	13.7
	Crash responsibility	1: No responsibility	263	41.5	76	35.7
		2: Main responsibility	117	18.5	42	19.7
		3: Equal	132	20.9	42	19.7
		4: Secondary	115	18.2	48	22.6
		5: Full responsibility	6	0.9	5	2.3
Driver characteristics	Driver gender	1: Male	542	85.6	184	86.4
		2: Female	91	14.4	29	13.6
	Driver age	1: < = 25	81	12.8	36	16.9
		2: 26–35	231	36.5	96	45.1
		3: 36–45	196	31.0	48	22.5
		4: 46–55	109	17.2	29	13.6
		5: 56–66	16	2.5	4	1.9
	Driver driving age	1: 0–4	269	42.5	99	46.5
		2: 5–9	195	30.8	62	29.1
		3: 10–14	98	15.5	28	13.1
		4: 15–19	42	6.6	18	8.5
		5: >= 20	29	4.6	6	2.8
	Drunk driving	1: Yes	128	20.2	53	24.9
		2: No	505	79.8	160	75.1
	Use seat belts	1: No	69	10.9	19	8.9
		2: Yes	564	89.1	194	91.1
Road and environment	Road alignment	1: Bend and slope	98	15.5	31	14.6
characteristics		2: Straight	535	84.5	182	85.4
	Road type	1: General urban roads	417	65.8	149	70.0
		2: Urban expressway	43	6.8	24	11.3
		3: Highway²	110	17.4	26	12.2
		4: Other roads	63	10.0	14	6.5
	Types of road section	1: Intersections	142	22.5	53	24.9
		2: Ordinary road segment	442	69.8	145	68.1
		3: Special segments³	49	7.7	15	7.0
	Type of land use	1: Commercial area	143	22.6	54	25.4
		2: Industrial area	38	6.0	17	8.0
		3: Public service area⁴	159	25.1	50	23.4
		4: Suburb	124	19.6	31	14.6
		5: Residential area	169	26.7	61	28.6
	Road physical isolation	1: Absent	389	61.4	102	47.9
		2: PSBR⁵	21	3.3	9	4.2
		3: Median divider	165	26.1	69	32.4
		4: MD&PSBR⁶	58	9.2	33	15.5
	Median strip	1: Absent	377	59.6	103	48.4
		2: Greenbelt	125	19.7	35	16.4
		3: Guardrail	131	20.7	75	35.2
	POTRCS⁷	1: Motorway	433	68.4	163	76.5
		2: Mixed road	134	21.2	35	16.5
		3: Bicycle lane	35	5.5	9	4.2
		4: Other	31	4.9	6	2.8
	Traffic control	1: Absent	131	20.7	29	13.6
		2: Present	502	79.3	184	86.4
	Surface condition	1: Stagnant water	9	1.4	2	.9
		2: Wet	96	15.2	54	25.4
		3: Dry	528	83.4	157	73.7
	Weather condition	1: Rainy/foggy/snowy	90	14.2	43	20.2
		2: Overcast	127	20.1	64	30.0
		3: Sunny	416	65.7	106	49.8
	Visibility	1: <50 m	38	6.0	55	25.8
		2: 50–100 m	178	28.1	91	42.7
		3: 100–200 m	183	28.9	33	15.5
		4: >= 200 m	234	37.0	34	16.0
	Season	1: Spring	164	25.9	46	21.6
		2: Summer	177	28.0	60	28.2
		3: Autumn	166	26.2	64	30.0
		4: Winter	126	19.9	43	20.2
	Type of day	1: Weekday	486	76.8	163	76.5
		2: Weekend	147	23.2	50	23.5

Note: ¹Driving off the right side of the roadway means that electric bike riders do not rider on the right side of the roadway when there is no non-motorized lane.

²Highway include Grade one, Grade two, Grade three, Grade four and Substandard way.

³Special segments include tunnels, bridges, elevated segments and narrowed segments.

⁴Public service area includes educational and cultural facilities, hospitals, subway entrance, bus stop and parks.

⁵PSBR denotes physically separated bicycle roadways.

⁶MD&PSBR denotes the median divider and physically separated bicycle roadways.

⁷POTRCS denotes the variable ‘Position of the road cross-section’.

In this study, we focused on the impact on the injury severity of electric bike riders. Generally, more crashes occurred during daylight (n = 633; 74.8%) than in darkness (n = 213; 25.2%). The number of casualties during daylight (n = 610; 72.1%) was higher than the number of casualties in darkness (n = 197; 23.3%). There are four types of injury results in Fig. 1: no injury, slight, serious and fatal. However, the proportion of serious injuries is greater in darkness than during daylight (1.4% vs. 1.3%) and the proportion of fatal injuries is greater in darkness than during daylight (12.7% vs. 6.8%). According to the ratio, serious traffic crashes for riders in darkness are 1.1 times the serious traffic crashes for riders during daylight and the fatal ones are 1.9 times, which indicates that the consequences of electric bike crashes in darkness are more serious than those during daylight. Fig. 2 shows the age group distribution of the severity of injuries among electric bike riders. Older riders are more vulnerable to death. People in the 46–55 age group have the highest probability of serious crashes, while in the 56–65 age group the probability of fatal crashes is the highest.

Fig. 1.

Crash rate of electric bikes in daylight and darkness.

Open in new tab Download slide

Fig. 2.

Age distribution of electric bike riders.

Open in new tab Download slide

2.2 Method

2.2.1 Geographic Information System

ArcGIS is a GIS basic software developed by the American Institute of Environmental Systems. ArcMap is a major application in ArcGIS for desktop, which undertakes all mapping and editing tasks, as well as map-based query and analysis functions. ArcMap provides two types of map views: geographic data view and map layout view. In the geographic data view, users can symbolically display, analyse and edit GIS data sets. In the map layout view, users can design and process map pages, including geographic data view and map elements, such as scale bar, legend, compass and geographic reference.

GIS technology is a useful platform for storing, analysing and managing data to provide visual graphic output. According to the crash location provided in the data, the study uses AutoNavi map to geocode all crash locations and records the corresponding longitude and latitude. In ArcGIS 10.2 software, the longitude and latitude of the crash point are matched with the corresponding position on the map, and then the location of each rider's crash is marked on the map. According to the input feature data set, the kernel density analysis tool is used to calculate the aggregation state of regional point data, and the region with high crash rate is obtained.

2.2.2 Ordered Probit model

The Ordered Probit (OP) model is an ideal estimation method for analysing discrete multiple classification problems. The order of injury data is taken as an important consideration in road crash severity modelling [27,28]. The probit model is a popular specification of ordinal or binary response model. According to the classification number of dependent variables, it can be divided into multivariate and binary probit models. In addition, it can also be divided into OP model and disordered probit model according to whether the dependent variables have order characteristics.

In this paper, the severity of traffic crash between electric bike and car is gradually increasing, so the focus is on the OP model. The OP model judges the influence injury severity of specific variables on decision variables by estimating the marginal effect, which is suitable for the analysis of variables of electric bike crashes. Therefore, considering that the influencing factors of different ambient light are quite different, the OP models of daylight and darkness are established, respectively.

Assuming that the severity of a traffic crash between an electric bike and a car is expressed by |$y$| and various factors affecting the severity of crashes are expressed by |$x$|⁠, the general expression of the model is:

$$\begin{eqnarray*} y_i^* = \beta {x_i} + {\varepsilon _i} \end{eqnarray*}$$

(1)

where |$y_i^*$| is a latent variable for measuring injury severity of riders i in a crash, |$\beta $| is the vector composed of covariate coefficients, which is the parameter corresponding to the influencing factor. |${x_i}$| is the vector of independent variables describing several factors. |${\varepsilon _i}$| represents the random disturbance term, which is independent of the covariate |${x_i}$|⁠. It is a combination of other factors that cannot be ignored that have an impact on the severity of the crash.

The ordered probability model treats the potential and difficult to observe continuous variables, |$y_i^*$| is mapped into an observable ordered variable |$y$|⁠, which represents the severity of the crash, |$y$| is continuous with |${y_i}$| and |$y_i^*$| in equation 2, and |$y_i^*$| is an invisible variable.

$$\begin{eqnarray*} y_i &=& 1\ if\ {\mu _0} < y_i^* \le \mu _1\nonumber\\ &=& 2\ if\ \mu _1 < y_i^* \le \mu _2\nonumber\\ &=& \cdots \nonumber\\ &=& J\ if\ \mu _{j - 1} < y_i^* \le \mu _j \end{eqnarray*}$$

(2)

where |$I = ( {{\mu _0},{\rm{\ }}{\mu _1} \cdots {\mu _j}} )$| is the set of severity classification points of electric bike crashes, |$({\mu _0} < {\mu _1} < {\mu _2} \cdots < {\mu _j},{\rm{\ }}{\mu _0} = - \infty ,{\rm{\ }}{\mu _j} = + \infty )$|⁠.

When the conditional distribution of crash influencing factors of |${\varepsilon _i}$| is assumed to be standard normal distribution, its probability density function |$f( {{\varepsilon _i}} ) = \phi ( {{\varepsilon _i}} )$|⁠, cumulative distribution function |$F( {{\varepsilon _i}} ) = {\rm{\Phi }}( {{\varepsilon _i}} )$|⁠, i.e. |${\varepsilon _i}|{X_i}\sim{\rm{N}}( {0,1} )$|⁠, then the conditional probability that the severity of electric bike crash is j is:

$$\begin{eqnarray*} P(y_i = j|x_i) &=& P(\mu _j - 1 < y_i^* \le \mu _j)\nonumber\\ &=& P(\mu _j - 1 - \beta x_i < \varepsilon _i \le \mu _j - \beta x_i)\nonumber\\ &=& \rm{\Phi} \left( \mu _j - \beta x_i \right) - \rm{\Phi} \left( \mu _j - 1 - \beta x_i \right) \end{eqnarray*}$$

(3)

Due to the limitation of OP model, the parameter β cannot be used to explain the relationship between crash variables and the severity of electric bike crashes. Therefore, for the established OP model, after estimating the parameters, it is necessary to calculate the marginal effect value of each variable. Marginal effect value means: if other crash variables take the mean value, when one variable increases or decreases by one unit, the probability change of the severity is a category, and its calculation formula is:

$$\begin{eqnarray*} \frac{{\partial p({y_i} = j|{x_i})}}{{\partial {x_k}}} = \beta \cdot \left[ {{\rm{\Phi }}\left( {{\mu _j} - \beta {x_i}} \right) - {\rm{\Phi }}\left( {{\mu _{j - 1}} - \beta {x_i}} \right)} \right] \end{eqnarray*}$$

(4)

3. Results

3.1 Temporal and spatial distribution characteristics of crashes

3.1.1 Time distribution characteristics

Aiming at 846 traffic crashes of electric bike riders in Changsha, the time distribution characteristics were analysed from three scales of month, week and hour. The time distribution of electric bike crashes is shown in Fig. 3.

Fig. 3.

Electric bike rider injury severity by month (a), by week (b) and by hour (c).

Open in new tab Download slide

It can be seen from Fig. 3(a) that the fatality rate is the highest in January and November, while July has the highest number of crashes. January and November are the cold season in Changsha. The crashes are seriously affected by the weather and road condition, so the fatality rate is high. July belongs to the summer with clear weather and easy travel by electric bike during summer vacation. The high frequency of use increases the probability of crashes. As can be seen from Fig. 3(b), the total number of crashes of electric bike riders on weekdays is much higher than that on weekends, of which the number of crashes on wednesday is the largest and the number of crashes on saturday is the least. The highest fatality rate was 10.77% on monday, followed by crashes on sunday, with a fatality rate of 9.32%. It can be seen from Fig. 3(c) that there are multiple peaks in the number of crashes within one day (24 h). Among them, the number of crashes from 17:00 to 18:00 is the highest, followed by 16:00–17:00, 18:00–19:00 and 07:00–08:00 and there is an obvious peak of fatal crashes from 23:00 at darkness to 03:00 the next day. Although the total number of crashes in daylight is high, the fatality rate of traffic crashes at darkness is significantly higher than that in daylight. 07:00–19:00 has the high possibility of crashes, which is due to the large number of residents going out during this period, the large traffic flow and the high severity of freedom of driving.

3.1.2 Spatial distribution characteristics

By combining the injury crashes of electric bike riders with the GIS data, the distribution of traffic crashes can be roughly distinguished. Fig. 4(a) shows the map of Hunan Province with region names. In the study, we focus on the spatial distribution characteristics of the crash in Changsha. Fig. 4(b) shows the map of Changsha with region names.

Fig. 4.

A map of Hunan Province with region names (a) and a map of Changsha with region names (b). (In the picture, tag 1 stands for Furong District and tag 2 stands for Tianxin District.)

Open in new tab Download slide

Fig. 5 shows the overall spatial distribution of electric bike crashes in Changsha. According to the analysis, the crashes spread from the central area along the urban trunk road to the surrounding areas. The crashes mainly occurred in the main urban area, and the main type of land uses with high crash rates were the commercial, public service and residential areas. During daylight, the administrative divisions of Changsha with a higher crash rate are the Wangcheng District, Yuelu District, Tianxin District and Yuhua District. The density of crashes in darkness is slightly lower than that during daylight, but the Yuhua District in darkness is still the area with a high incidence of crashes during the corresponding period; the density of crashes in the Furong District increases in darkness, and the density in other areas is basically stable during daylight and darkness.

Fig. 5.

Kernel density analysis of electric bike crash in Changsha. Total distribution of all day crashes (a), crash distribution in daylight (b), crash distribution in darkness (c).

Open in new tab Download slide

Furthermore, from the distribution of crash density of electric bike riders depicted in Fig. 5(a), major hot spot areas can be identified in Figs. 5(b) and 5(c), with the tag numbers from 1 to 8. Tag 1 is located near the bridge and is mostly in residential areas with elementary schools. The high crash rate may be due to the large traffic volume and residents are prone to crashes when using electric bikes during daylight. Tags 2, 3, 6 and 7 represent the most prosperous business district in Changsha. There is little difference in the crash rate between daylight and darkness, which is in line with the situation that the Wuyi Square and Furong Square have many people and vehicles all day. It is worth noting that compared with daylight, the crash density in the area where tag 5 is located is significantly higher in darkness. It is an old residential area with a long history. Incomplete transportation facilities, narrow roads and blocked sight of riders will lead to more people crashing at darkness. Tags 4 and 8 are in the congested section of the Shaoshan South Road in Changsha. It can be found that in darkness, the crash gathering area of tag 8 moves up relative to tag 4. There are residential areas and overpasses near the school in this area, which hinder riders from crossing the road.

3.2 Analysis results from the OP model

3.2.1 Independent variable screening

In this study, the injury severity of riders is taken as the dependent variable, which is divided into four grades, and there is a certain order among the grades, which are no injury, slight, serious and fatal. The independent variables are classified in Table 1. If all the independent variables are utilized in the analysis, the fitting accuracy of the model will not be high. As shown in Fig. 6, we first use Pearson correlation analysis to screen out the variables with strong correlation with injury severity. Therefore, some non-significant variables, such as rider gender, driver gender, driver driving age, driver age, POTRCS, surface condition, visibility and type of day can be eliminated by correlation analysis. Variables, such as rider age, rider household type, crash responsibility, personnel type, collision type, illegal behaviour, type of land use, road type, road alignment, road physical isolation and median strip, etc. will continue to be studied. The reference levels of variables in OP model are shown in Table 2.

Fig. 6.

Correlation matrix. (In the picture, POTRCS denotes the variable ‘Position of the road cross-section’.)

Open in new tab Download slide

Table 2.

Open in new tab

Reference levels of OP model variables.

Variable	Category
Rider age	≤ 25, 26–35,* 36–45, 46–55, 56–65, 65+
Rider household type	Non-agricultural,* Agricultural
Rider personnel type	Express delivery employees, Worker, Farmer, Self-employed, Other*
Illegal behaviour of riders	Not following traffic signals, Reverse driving, Cross the motorway without getting off, Failure to give way as required, Driving off the right side of the roadway, Not driving in the non-motorized lane, Other illegal behaviour, Absent*
Crash responsibility	Full responsibility,* Main responsibility, Equal, Secondary, No responsibility
Collision type	Side impact (angle uncertain), Side impact (opposite), Side impact (same direction), Side impact (right angle), Frontal impact, Rear impact, Scratch, Other angles crash*
Use seat belts	Yes,* No
Drunk driving	Yes, No*
Type of land use	Commercial area, Industrial area, Public service area, Residential area,* Suburb
Road physical isolation	PSBR, Median divider, MD&PSBR,* Absent
Road type	General urban roads, Urban expressway, Highway, Other roads*
Road alignment	Straight,* Bend and slope
Types of road section	Intersections, Ordinary road segment, Special segments*
Traffic control	Present,* Absent
Median strip	Greenbelt, Guardrail,* Absent
Weather condition	Sunny,* Overcast, Rainy/foggy/snowy
Season	Spring, Summer, Autumn, Winter*

Variable	Category
Rider age	≤ 25, 26–35,* 36–45, 46–55, 56–65, 65+
Rider household type	Non-agricultural,* Agricultural
Rider personnel type	Express delivery employees, Worker, Farmer, Self-employed, Other*
Illegal behaviour of riders	Not following traffic signals, Reverse driving, Cross the motorway without getting off, Failure to give way as required, Driving off the right side of the roadway, Not driving in the non-motorized lane, Other illegal behaviour, Absent*
Crash responsibility	Full responsibility,* Main responsibility, Equal, Secondary, No responsibility
Collision type	Side impact (angle uncertain), Side impact (opposite), Side impact (same direction), Side impact (right angle), Frontal impact, Rear impact, Scratch, Other angles crash*
Use seat belts	Yes,* No
Drunk driving	Yes, No*
Type of land use	Commercial area, Industrial area, Public service area, Residential area,* Suburb
Road physical isolation	PSBR, Median divider, MD&PSBR,* Absent
Road type	General urban roads, Urban expressway, Highway, Other roads*
Road alignment	Straight,* Bend and slope
Types of road section	Intersections, Ordinary road segment, Special segments*
Traffic control	Present,* Absent
Median strip	Greenbelt, Guardrail,* Absent
Weather condition	Sunny,* Overcast, Rainy/foggy/snowy
Season	Spring, Summer, Autumn, Winter*

Note: *represents the reference level in the OP model.

Table 2.

Open in new tab

Reference levels of OP model variables.

Variable	Category
Rider age	≤ 25, 26–35,* 36–45, 46–55, 56–65, 65+
Rider household type	Non-agricultural,* Agricultural
Rider personnel type	Express delivery employees, Worker, Farmer, Self-employed, Other*
Illegal behaviour of riders	Not following traffic signals, Reverse driving, Cross the motorway without getting off, Failure to give way as required, Driving off the right side of the roadway, Not driving in the non-motorized lane, Other illegal behaviour, Absent*
Crash responsibility	Full responsibility,* Main responsibility, Equal, Secondary, No responsibility
Collision type	Side impact (angle uncertain), Side impact (opposite), Side impact (same direction), Side impact (right angle), Frontal impact, Rear impact, Scratch, Other angles crash*
Use seat belts	Yes,* No
Drunk driving	Yes, No*
Type of land use	Commercial area, Industrial area, Public service area, Residential area,* Suburb
Road physical isolation	PSBR, Median divider, MD&PSBR,* Absent
Road type	General urban roads, Urban expressway, Highway, Other roads*
Road alignment	Straight,* Bend and slope
Types of road section	Intersections, Ordinary road segment, Special segments*
Traffic control	Present,* Absent
Median strip	Greenbelt, Guardrail,* Absent
Weather condition	Sunny,* Overcast, Rainy/foggy/snowy
Season	Spring, Summer, Autumn, Winter*

Variable	Category
Rider age	≤ 25, 26–35,* 36–45, 46–55, 56–65, 65+
Rider household type	Non-agricultural,* Agricultural
Rider personnel type	Express delivery employees, Worker, Farmer, Self-employed, Other*
Illegal behaviour of riders	Not following traffic signals, Reverse driving, Cross the motorway without getting off, Failure to give way as required, Driving off the right side of the roadway, Not driving in the non-motorized lane, Other illegal behaviour, Absent*
Crash responsibility	Full responsibility,* Main responsibility, Equal, Secondary, No responsibility
Collision type	Side impact (angle uncertain), Side impact (opposite), Side impact (same direction), Side impact (right angle), Frontal impact, Rear impact, Scratch, Other angles crash*
Use seat belts	Yes,* No
Drunk driving	Yes, No*
Type of land use	Commercial area, Industrial area, Public service area, Residential area,* Suburb
Road physical isolation	PSBR, Median divider, MD&PSBR,* Absent
Road type	General urban roads, Urban expressway, Highway, Other roads*
Road alignment	Straight,* Bend and slope
Types of road section	Intersections, Ordinary road segment, Special segments*
Traffic control	Present,* Absent
Median strip	Greenbelt, Guardrail,* Absent
Weather condition	Sunny,* Overcast, Rainy/foggy/snowy
Season	Spring, Summer, Autumn, Winter*

Note: *represents the reference level in the OP model.

Furthermore, we test the multicollinearity between the independent variables in daylight and darkness condition model to improve the accuracy of variable selection. The variance inflation factor (VIF) is used to quantify the severity of collinearity and ensure the independence of each variable [29]. According to the judgement standard of multiple collinearities, the result is shown in Table 3. It can be seen from the table that the variance expansion factor is less than 10 and the tolerance of each variable is greater than 0.2, so there is no multiple collinearity relationship between the variables.

Table 3.

Open in new tab

Collinearity analysis between influence factors.

Variable	Daylight		Darkness
Variable	VIF	Tolerance	VIF	Tolerance
Rider age	1.03	0.97	1.08	0.92
Rider household type	1.11	0.90	1.17	0.85
Rider personnel type	1.15	0.87	1.08	0.93
Illegal behaviour of riders	1.33	0.75	1.27	0.79
Crash responsibility	1.25	0.80	1.27	0.79
Collision type	1.04	0.96	1.11	0.90
Use seat belts	1.07	0.93	1.09	0.91
Drunk driving	1.12	0.89	1.13	0.89
Type of land use	1.07	0.94	1.06	0.95
Road physical isolation	1.56	0.64	1.44	0.70
Road type	1.54	0.65	1.46	0.69
Road alignment	1.21	0.83	1.16	0.87
Types of road section	1.05	0.95	1.12	0.90
Traffic control	1.22	0.82	1.19	0.84
Median strip	1.56	0.64	1.49	0.67
Weather condition	1.03	0.97	1.05	0.95
Season	1.04	0.96	1.06	0.94
Mean VIF	1.20		1.19

Variable	Daylight		Darkness
Variable	VIF	Tolerance	VIF	Tolerance
Rider age	1.03	0.97	1.08	0.92
Rider household type	1.11	0.90	1.17	0.85
Rider personnel type	1.15	0.87	1.08	0.93
Illegal behaviour of riders	1.33	0.75	1.27	0.79
Crash responsibility	1.25	0.80	1.27	0.79
Collision type	1.04	0.96	1.11	0.90
Use seat belts	1.07	0.93	1.09	0.91
Drunk driving	1.12	0.89	1.13	0.89
Type of land use	1.07	0.94	1.06	0.95
Road physical isolation	1.56	0.64	1.44	0.70
Road type	1.54	0.65	1.46	0.69
Road alignment	1.21	0.83	1.16	0.87
Types of road section	1.05	0.95	1.12	0.90
Traffic control	1.22	0.82	1.19	0.84
Median strip	1.56	0.64	1.49	0.67
Weather condition	1.03	0.97	1.05	0.95
Season	1.04	0.96	1.06	0.94
Mean VIF	1.20		1.19

Table 3.

Open in new tab

Collinearity analysis between influence factors.

Variable	Daylight		Darkness
Variable	VIF	Tolerance	VIF	Tolerance
Rider age	1.03	0.97	1.08	0.92
Rider household type	1.11	0.90	1.17	0.85
Rider personnel type	1.15	0.87	1.08	0.93
Illegal behaviour of riders	1.33	0.75	1.27	0.79
Crash responsibility	1.25	0.80	1.27	0.79
Collision type	1.04	0.96	1.11	0.90
Use seat belts	1.07	0.93	1.09	0.91
Drunk driving	1.12	0.89	1.13	0.89
Type of land use	1.07	0.94	1.06	0.95
Road physical isolation	1.56	0.64	1.44	0.70
Road type	1.54	0.65	1.46	0.69
Road alignment	1.21	0.83	1.16	0.87
Types of road section	1.05	0.95	1.12	0.90
Traffic control	1.22	0.82	1.19	0.84
Median strip	1.56	0.64	1.49	0.67
Weather condition	1.03	0.97	1.05	0.95
Season	1.04	0.96	1.06	0.94
Mean VIF	1.20		1.19

Variable	Daylight		Darkness
Variable	VIF	Tolerance	VIF	Tolerance
Rider age	1.03	0.97	1.08	0.92
Rider household type	1.11	0.90	1.17	0.85
Rider personnel type	1.15	0.87	1.08	0.93
Illegal behaviour of riders	1.33	0.75	1.27	0.79
Crash responsibility	1.25	0.80	1.27	0.79
Collision type	1.04	0.96	1.11	0.90
Use seat belts	1.07	0.93	1.09	0.91
Drunk driving	1.12	0.89	1.13	0.89
Type of land use	1.07	0.94	1.06	0.95
Road physical isolation	1.56	0.64	1.44	0.70
Road type	1.54	0.65	1.46	0.69
Road alignment	1.21	0.83	1.16	0.87
Types of road section	1.05	0.95	1.12	0.90
Traffic control	1.22	0.82	1.19	0.84
Median strip	1.56	0.64	1.49	0.67
Weather condition	1.03	0.97	1.05	0.95
Season	1.04	0.96	1.06	0.94
Mean VIF	1.20		1.19

3.2.2 Daylight condition model

Table 4 shows the results from the OP model for the daylight condition. Age has a great influence on the injury severity of electric bike riders in daylight. There are two age groups with significance (P < 0.05), namely, the age group 56–65 years old (P = 0.048) and the elderly over 65 (P = 0.047), which shows that middle-aged and elderly electric bike riders are vulnerable. Compared with those aged 26–35, elderly riders aged over 65 have a lower probability of slight injuries during daylight, but a higher probability of serious and fatal injuries. Farmers (P = 0.008) are more vulnerable in electric bike crashes, and the number of crashes of farmers contributes 14.7% of the total crashes in daylight. The occurrence of illegal behaviour of riders also increases the risk of injury. When the riders do not follow traffic signals to ride (P = 0.062), reverse driving (P = 0.051) and cross the motorway without getting off (P = 0.071), these have a significant impact on the severity of injuries to riders. Whether the car driver is drunk driving also has a significant impact on the severity of the crash. The car driver's drunk driving (P = 0.000) led to the probability of fatal of electric bike riders increases by 6.46%.

Table 4.

Open in new tab

OP injury severity model for daylight condition.

Variable	Daylight				Marginal effects
Variable	Coefficient (β)	SE	Z-value	P-value	No injury	Slight	Serious	Fatal
Rider aged 56–65	0.466	0.236	1.98	0.048**	−0.0313	−0.0194	0.0065	0.0441
Rider aged 65+	0.587	0.296	1.98	0.047**	−0.0393	−0.0244	0.0082	0.0555
Farmer	0.614	0.232	2.65	0.008***	−0.0412	−0.0255	0.0086	0.0581
Not following traffic signals	−0.671	0.360	−1.86	0.062*	0.0449	0.0278	−0.0094	−0.0634
Reverse driving	−0.723	0.371	−1.95	0.051*	0.0484	0.0301	−0.0101	−0.0684
Cross the motorway without getting off	−0.653	0.362	−1.81	0.071*	0.0437	0.0271	−0.0091	−0.0617
Drunk driving (Yes)	0.683	0.194	3.52	0.000***	−0.0457	−0.0284	0.0095	0.0646
Bend and slope	0.521	0.215	2.43	0.015**	−0.0349	−0.0216	0.0073	0.0493
Traffic control (Absent)	−0.422	0.200	−2.11	0.034**	0.0283	0.0175	−0.0059	−0.0399
Median divider	0.680	0.290	2.35	0.019**	−0.0455	−0.0282	0.0095	0.0642
Side impact (angle uncertain)	−0.579	0.283	−2.05	0.041**	0.0388	0.0241	−0.0081	−0.0548
Side impact (right angle)	−0.533	0.271	−1.97	0.049**	0.0357	0.0221	−0.0074	−0.0504
Cut1	−1.771	0.915	–	–	–	–	–	–
Cut2	2.342	0.922	–	–	–	–	–	–
Cut3	2.470	0.923	–	–	–	–	–	–

Variable	Daylight				Marginal effects
Variable	Coefficient (β)	SE	Z-value	P-value	No injury	Slight	Serious	Fatal
Rider aged 56–65	0.466	0.236	1.98	0.048**	−0.0313	−0.0194	0.0065	0.0441
Rider aged 65+	0.587	0.296	1.98	0.047**	−0.0393	−0.0244	0.0082	0.0555
Farmer	0.614	0.232	2.65	0.008***	−0.0412	−0.0255	0.0086	0.0581
Not following traffic signals	−0.671	0.360	−1.86	0.062*	0.0449	0.0278	−0.0094	−0.0634
Reverse driving	−0.723	0.371	−1.95	0.051*	0.0484	0.0301	−0.0101	−0.0684
Cross the motorway without getting off	−0.653	0.362	−1.81	0.071*	0.0437	0.0271	−0.0091	−0.0617
Drunk driving (Yes)	0.683	0.194	3.52	0.000***	−0.0457	−0.0284	0.0095	0.0646
Bend and slope	0.521	0.215	2.43	0.015**	−0.0349	−0.0216	0.0073	0.0493
Traffic control (Absent)	−0.422	0.200	−2.11	0.034**	0.0283	0.0175	−0.0059	−0.0399
Median divider	0.680	0.290	2.35	0.019**	−0.0455	−0.0282	0.0095	0.0642
Side impact (angle uncertain)	−0.579	0.283	−2.05	0.041**	0.0388	0.0241	−0.0081	−0.0548
Side impact (right angle)	−0.533	0.271	−1.97	0.049**	0.0357	0.0221	−0.0074	−0.0504
Cut1	−1.771	0.915	–	–	–	–	–	–
Cut2	2.342	0.922	–	–	–	–	–	–
Cut3	2.470	0.923	–	–	–	–	–	–

Note: Number of obs = 633; Log likelihood = −227.09 763; LR chi²(51) = 138.50; Prob > chi² = 0.0000; Pseudo R² = 0.2337. *P-value significant at 90% confidence interval; **P-value significant at 95% confidence interval; ***P-value significant at 99% confidence interval; – indicates a dummy variable or the value does not exist.

Table 4.

Open in new tab

OP injury severity model for daylight condition.

Variable	Daylight				Marginal effects
Variable	Coefficient (β)	SE	Z-value	P-value	No injury	Slight	Serious	Fatal
Rider aged 56–65	0.466	0.236	1.98	0.048**	−0.0313	−0.0194	0.0065	0.0441
Rider aged 65+	0.587	0.296	1.98	0.047**	−0.0393	−0.0244	0.0082	0.0555
Farmer	0.614	0.232	2.65	0.008***	−0.0412	−0.0255	0.0086	0.0581
Not following traffic signals	−0.671	0.360	−1.86	0.062*	0.0449	0.0278	−0.0094	−0.0634
Reverse driving	−0.723	0.371	−1.95	0.051*	0.0484	0.0301	−0.0101	−0.0684
Cross the motorway without getting off	−0.653	0.362	−1.81	0.071*	0.0437	0.0271	−0.0091	−0.0617
Drunk driving (Yes)	0.683	0.194	3.52	0.000***	−0.0457	−0.0284	0.0095	0.0646
Bend and slope	0.521	0.215	2.43	0.015**	−0.0349	−0.0216	0.0073	0.0493
Traffic control (Absent)	−0.422	0.200	−2.11	0.034**	0.0283	0.0175	−0.0059	−0.0399
Median divider	0.680	0.290	2.35	0.019**	−0.0455	−0.0282	0.0095	0.0642
Side impact (angle uncertain)	−0.579	0.283	−2.05	0.041**	0.0388	0.0241	−0.0081	−0.0548
Side impact (right angle)	−0.533	0.271	−1.97	0.049**	0.0357	0.0221	−0.0074	−0.0504
Cut1	−1.771	0.915	–	–	–	–	–	–
Cut2	2.342	0.922	–	–	–	–	–	–
Cut3	2.470	0.923	–	–	–	–	–	–

Variable	Daylight				Marginal effects
Variable	Coefficient (β)	SE	Z-value	P-value	No injury	Slight	Serious	Fatal
Rider aged 56–65	0.466	0.236	1.98	0.048**	−0.0313	−0.0194	0.0065	0.0441
Rider aged 65+	0.587	0.296	1.98	0.047**	−0.0393	−0.0244	0.0082	0.0555
Farmer	0.614	0.232	2.65	0.008***	−0.0412	−0.0255	0.0086	0.0581
Not following traffic signals	−0.671	0.360	−1.86	0.062*	0.0449	0.0278	−0.0094	−0.0634
Reverse driving	−0.723	0.371	−1.95	0.051*	0.0484	0.0301	−0.0101	−0.0684
Cross the motorway without getting off	−0.653	0.362	−1.81	0.071*	0.0437	0.0271	−0.0091	−0.0617
Drunk driving (Yes)	0.683	0.194	3.52	0.000***	−0.0457	−0.0284	0.0095	0.0646
Bend and slope	0.521	0.215	2.43	0.015**	−0.0349	−0.0216	0.0073	0.0493
Traffic control (Absent)	−0.422	0.200	−2.11	0.034**	0.0283	0.0175	−0.0059	−0.0399
Median divider	0.680	0.290	2.35	0.019**	−0.0455	−0.0282	0.0095	0.0642
Side impact (angle uncertain)	−0.579	0.283	−2.05	0.041**	0.0388	0.0241	−0.0081	−0.0548
Side impact (right angle)	−0.533	0.271	−1.97	0.049**	0.0357	0.0221	−0.0074	−0.0504
Cut1	−1.771	0.915	–	–	–	–	–	–
Cut2	2.342	0.922	–	–	–	–	–	–
Cut3	2.470	0.923	–	–	–	–	–	–

Compared with straight road segments, poor road segments, such as bend and slope lines will cause more serious crashes, with the probability of slight injuries reduced by 2.16% and the probability of fatalities increased by 4.93%. Roads without traffic control, such as roads without traffic lights, signs, markings and other safety facilities have a significant impact on the severity of the crash (P = 0.034). When the road physical isolation is only a median divider (P = 0.019), compared with the roads with both median divider and physically separated bicycle roadways, the probability of slight injury is reduced by 2.82% and the probability of fatalities is increased by 6.42%. The installation of non-motor vehicle isolation facilities can effectively avoid traffic violations, such as non-motor vehicles crossing the road at will and motor vehicles rushing to the road, and further prevent and reduce road traffic crashes. When an electric bike has a side impact (angle uncertain) or side impact (right angle) with a car, it is closely related to the severity of injury of riders (P < 0.05). Compared with other angle collision types, in daylight, the probability of slight injury caused by side impact increases and the probability of serious and fatal injury crash decreases.

3.2.3 Darkness condition model

Table 5 shows the results from the OP model for the darkness condition. Under the condition of darkness, the ages of 46–55, 56–65 and 65+ were significant (P < 0.05). Among them, the fatal probability of elderly aged 65+ (P = 0.009) increased by 26.14% compared with riders aged 26–35, indicating that people over 65 are most vulnerable to fatal injury in darkness. This may be because the elderly often wear dark coats when they go out, which is not easy to be seen in darkness. In addition, the elderly have poor operating skills when riding and slow responses when encountering situations, so it is easy to cause traffic crashes. However, different from the daylight, the probability of slight injury of the elderly aged 46–55 is reduced by 4.32%, the probability of serious injury is increased by 1.01% and the probability of fatal injury is increased by 12.69%. Illegal behaviours are more likely to occur in darkness, among which driving off the right side of the roadway (P = 0.003) and failure to give way as required (P = 0.009) cause the most serious injuries. On roads without non-motorized lanes, failure to drive on the right side can easily lead to electric bike crashes. At the same time, violations of road priority, such as electric bikes that do not allow cars to go straight when turning, also increase the severity of rider injuries. Compared with riders who did not break the law, the fatal crash probability of not following traffic signals, not driving in the non-motorized lane and reverse driving are also increased by 17.69%, 17.44% and 13.54%, respectively. The crash responsibility of electric bike riders is no responsibility (P = 0.052), that is, when the car driver is fully responsible, it has a significant impact on the injury severity of riders, which indicates that the driver's own fault is more likely to cause injury at darkness. When the responsibility is entirely the rider's, the probability of slight injury is reduced by 7.68% and the probability of fatal is increased by 22.6%. Consistent with daylight, car driver drunk driving (P = 0.001) will still greatly increase the probability of serious and fatal injury of riders.

Table 5.

Open in new tab

OP injury severity model for darkness condition.

Variable	Daylight				Marginal effects
Variable	Coefficient (β)	SE	Z-value	P-value	No injury	Slight	Serious	Fatal
Rider aged 46–55	0.927	0.429	2.16	0.031**	−0.0939	−0.0432	0.0101	0.1269
Rider aged 56–65	0.902	0.422	2.14	0.033**	−0.0913	−0.0420	0.0098	0.1235
Rider aged 65+	1.910	0.733	2.61	0.009***	−0.1933	−0.0890	0.0208	0.2614
Not following traffic signals	1.292	0.572	2.26	0.024**	−0.1308	−0.0602	0.0141	0.1769
Reverse driving	0.989	0.569	1.74	0.082*	−0.1001	−0.0460	0.0108	0.1354
Failure to give way as required	1.934	0.747	2.59	0.009***	−0.1961	−0.0902	0.0211	0.2652
Driving off the right side of the roadway	1.725	0.576	3.00	0.003***	−0.1746	−0.0803	0.0188	0.2361
Not driving in the non-motorized Lane	1.274	0.651	1.96	0.050**	−0.1290	−0.0593	0.0139	0.1744
Other illegal behaviour	1.308	0.627	2.09	0.037**	−0.1323	−0.0609	0.0142	0.1790
No responsibility	1.651	0.848	1.95	0.052*	−0.1671	−0.0768	0.0180	0.2260
Drunk driving (Yes)	1.061	0.310	3.42	0.001***	−0.1074	−0.0494	0.0116	0.1452
Suburb	1.151	0.548	2.10	0.036**	−0.1165	−0.0536	0.0125	0.1575
General urban roads	1.091	0.550	1.98	0.047**	−0.1104	−0.0508	0.0119	0.1493
Urban expressway	1.918	0.647	2.96	0.003***	−0.1942	−0.0893	0.0209	0.2626
Median divider	0.653	0.383	1.70	0.089*	−0.0661	−0.0304	0.0071	0.0894
Median strip (Absent)	0.612	0.330	1.85	0.064*	−0.0619	−0.0285	0.0067	0.0838
Frontal impact	1.766	0.519	3.40	0.001***	−0.1788	−0.0822	0.0192	0.2418
Scratch	1.575	0.857	1.84	0.066*	−0.1595	−0.0733	0.0172	0.2156
Rainy/foggy/snowy	−0.668	0.334	−2.00	0.046**	0.0676	0.0311	−0.0073	−0.0914
Overcast	−0.572	0.280	−2.05	0.041**	0.0579	0.0266	−0.0062	−0.0783
Cut1	2.239	1.393	–	–	–	–	–	–
Cut2	5.943	1.501	–	–	–	–	–	–
Cut3	6.045	1.502	–	–	–	–	–	–

Variable	Daylight				Marginal effects
Variable	Coefficient (β)	SE	Z-value	P-value	No injury	Slight	Serious	Fatal
Rider aged 46–55	0.927	0.429	2.16	0.031**	−0.0939	−0.0432	0.0101	0.1269
Rider aged 56–65	0.902	0.422	2.14	0.033**	−0.0913	−0.0420	0.0098	0.1235
Rider aged 65+	1.910	0.733	2.61	0.009***	−0.1933	−0.0890	0.0208	0.2614
Not following traffic signals	1.292	0.572	2.26	0.024**	−0.1308	−0.0602	0.0141	0.1769
Reverse driving	0.989	0.569	1.74	0.082*	−0.1001	−0.0460	0.0108	0.1354
Failure to give way as required	1.934	0.747	2.59	0.009***	−0.1961	−0.0902	0.0211	0.2652
Driving off the right side of the roadway	1.725	0.576	3.00	0.003***	−0.1746	−0.0803	0.0188	0.2361
Not driving in the non-motorized Lane	1.274	0.651	1.96	0.050**	−0.1290	−0.0593	0.0139	0.1744
Other illegal behaviour	1.308	0.627	2.09	0.037**	−0.1323	−0.0609	0.0142	0.1790
No responsibility	1.651	0.848	1.95	0.052*	−0.1671	−0.0768	0.0180	0.2260
Drunk driving (Yes)	1.061	0.310	3.42	0.001***	−0.1074	−0.0494	0.0116	0.1452
Suburb	1.151	0.548	2.10	0.036**	−0.1165	−0.0536	0.0125	0.1575
General urban roads	1.091	0.550	1.98	0.047**	−0.1104	−0.0508	0.0119	0.1493
Urban expressway	1.918	0.647	2.96	0.003***	−0.1942	−0.0893	0.0209	0.2626
Median divider	0.653	0.383	1.70	0.089*	−0.0661	−0.0304	0.0071	0.0894
Median strip (Absent)	0.612	0.330	1.85	0.064*	−0.0619	−0.0285	0.0067	0.0838
Frontal impact	1.766	0.519	3.40	0.001***	−0.1788	−0.0822	0.0192	0.2418
Scratch	1.575	0.857	1.84	0.066*	−0.1595	−0.0733	0.0172	0.2156
Rainy/foggy/snowy	−0.668	0.334	−2.00	0.046**	0.0676	0.0311	−0.0073	−0.0914
Overcast	−0.572	0.280	−2.05	0.041**	0.0579	0.0266	−0.0062	−0.0783
Cut1	2.239	1.393	–	–	–	–	–	–
Cut2	5.943	1.501	–	–	–	–	–	–
Cut3	6.045	1.502	–	–	–	–	–	–

Note: Number of obs = 213; Log likelihood = −103.93 075; LR chi²(51) = 93.35; Prob > chi² = 0.0003; Pseudo R² = 0.3099. *P-value significant at 90% confidence interval; **P-value significant at 95% confidence interval; ***P-value significant at 99% confidence interval; – indicates a dummy variable or the value does not exist.

Table 5.

Open in new tab

OP injury severity model for darkness condition.

Variable	Daylight				Marginal effects
Variable	Coefficient (β)	SE	Z-value	P-value	No injury	Slight	Serious	Fatal
Rider aged 46–55	0.927	0.429	2.16	0.031**	−0.0939	−0.0432	0.0101	0.1269
Rider aged 56–65	0.902	0.422	2.14	0.033**	−0.0913	−0.0420	0.0098	0.1235
Rider aged 65+	1.910	0.733	2.61	0.009***	−0.1933	−0.0890	0.0208	0.2614
Not following traffic signals	1.292	0.572	2.26	0.024**	−0.1308	−0.0602	0.0141	0.1769
Reverse driving	0.989	0.569	1.74	0.082*	−0.1001	−0.0460	0.0108	0.1354
Failure to give way as required	1.934	0.747	2.59	0.009***	−0.1961	−0.0902	0.0211	0.2652
Driving off the right side of the roadway	1.725	0.576	3.00	0.003***	−0.1746	−0.0803	0.0188	0.2361
Not driving in the non-motorized Lane	1.274	0.651	1.96	0.050**	−0.1290	−0.0593	0.0139	0.1744
Other illegal behaviour	1.308	0.627	2.09	0.037**	−0.1323	−0.0609	0.0142	0.1790
No responsibility	1.651	0.848	1.95	0.052*	−0.1671	−0.0768	0.0180	0.2260
Drunk driving (Yes)	1.061	0.310	3.42	0.001***	−0.1074	−0.0494	0.0116	0.1452
Suburb	1.151	0.548	2.10	0.036**	−0.1165	−0.0536	0.0125	0.1575
General urban roads	1.091	0.550	1.98	0.047**	−0.1104	−0.0508	0.0119	0.1493
Urban expressway	1.918	0.647	2.96	0.003***	−0.1942	−0.0893	0.0209	0.2626
Median divider	0.653	0.383	1.70	0.089*	−0.0661	−0.0304	0.0071	0.0894
Median strip (Absent)	0.612	0.330	1.85	0.064*	−0.0619	−0.0285	0.0067	0.0838
Frontal impact	1.766	0.519	3.40	0.001***	−0.1788	−0.0822	0.0192	0.2418
Scratch	1.575	0.857	1.84	0.066*	−0.1595	−0.0733	0.0172	0.2156
Rainy/foggy/snowy	−0.668	0.334	−2.00	0.046**	0.0676	0.0311	−0.0073	−0.0914
Overcast	−0.572	0.280	−2.05	0.041**	0.0579	0.0266	−0.0062	−0.0783
Cut1	2.239	1.393	–	–	–	–	–	–
Cut2	5.943	1.501	–	–	–	–	–	–
Cut3	6.045	1.502	–	–	–	–	–	–

Variable	Daylight				Marginal effects
Variable	Coefficient (β)	SE	Z-value	P-value	No injury	Slight	Serious	Fatal
Rider aged 46–55	0.927	0.429	2.16	0.031**	−0.0939	−0.0432	0.0101	0.1269
Rider aged 56–65	0.902	0.422	2.14	0.033**	−0.0913	−0.0420	0.0098	0.1235
Rider aged 65+	1.910	0.733	2.61	0.009***	−0.1933	−0.0890	0.0208	0.2614
Not following traffic signals	1.292	0.572	2.26	0.024**	−0.1308	−0.0602	0.0141	0.1769
Reverse driving	0.989	0.569	1.74	0.082*	−0.1001	−0.0460	0.0108	0.1354
Failure to give way as required	1.934	0.747	2.59	0.009***	−0.1961	−0.0902	0.0211	0.2652
Driving off the right side of the roadway	1.725	0.576	3.00	0.003***	−0.1746	−0.0803	0.0188	0.2361
Not driving in the non-motorized Lane	1.274	0.651	1.96	0.050**	−0.1290	−0.0593	0.0139	0.1744
Other illegal behaviour	1.308	0.627	2.09	0.037**	−0.1323	−0.0609	0.0142	0.1790
No responsibility	1.651	0.848	1.95	0.052*	−0.1671	−0.0768	0.0180	0.2260
Drunk driving (Yes)	1.061	0.310	3.42	0.001***	−0.1074	−0.0494	0.0116	0.1452
Suburb	1.151	0.548	2.10	0.036**	−0.1165	−0.0536	0.0125	0.1575
General urban roads	1.091	0.550	1.98	0.047**	−0.1104	−0.0508	0.0119	0.1493
Urban expressway	1.918	0.647	2.96	0.003***	−0.1942	−0.0893	0.0209	0.2626
Median divider	0.653	0.383	1.70	0.089*	−0.0661	−0.0304	0.0071	0.0894
Median strip (Absent)	0.612	0.330	1.85	0.064*	−0.0619	−0.0285	0.0067	0.0838
Frontal impact	1.766	0.519	3.40	0.001***	−0.1788	−0.0822	0.0192	0.2418
Scratch	1.575	0.857	1.84	0.066*	−0.1595	−0.0733	0.0172	0.2156
Rainy/foggy/snowy	−0.668	0.334	−2.00	0.046**	0.0676	0.0311	−0.0073	−0.0914
Overcast	−0.572	0.280	−2.05	0.041**	0.0579	0.0266	−0.0062	−0.0783
Cut1	2.239	1.393	–	–	–	–	–	–
Cut2	5.943	1.501	–	–	–	–	–	–
Cut3	6.045	1.502	–	–	–	–	–	–

When the type of land use is suburb, the fatality probability increases by 15.75%, compared with the electric bike crashes in residential areas. The road type of general urban roads (P = 0.047) and urban expressway (P = 0.003) will lead to more serious crashes. Compared with other roads, the fatal probability will increase by 14.93% and 26.26%, respectively. Like daylight, when the road physical isolation is only a median divider (P = 0.089), compared with the roads with both median divider and physically separated bicycle roadways, the probability of slight injury is reduced by 3.04% and the probability of fatalities is increased by 8.94%. When there is no median strip on the road (P = 0.064), compared with guardrails belt on the road, the probability of no injury is reduced by 6.19%, and the probability of serious and fatal injury are increased by 0.67% and 8.38%, respectively. When an electric bike has a front impact (P = 0.001) or scratch (P = 0.066) with a car, it is closely related to the severity of injury of riders. Compared with other angle collision types, in darkness, the probability of slight injury caused by front impact decreased by 8.22%, and the probability of fatal increased by 24.18%. And the probability of slight injury caused by scratch decreased by 7.33%, and the probability of fatal increased by 21.56%. Crashes that occur in darkness and when the weather conditions are rainy, foggy, snowy or overcast also deserve attention. Compared with the clear night, the probability of slight injury increases by 3.11% and 2.66%, respectively, and the probability of serious and fatal injury decreases.

4. Discussion

Through the analysis of the GIS and the OP model, this study obtained temporal and spatial distribution characteristics of crashes and significant factors in the injury severity of electric bike riders in traffic crashes under different light conditions. It is necessary to take targeted preventive measures from the perspective of electric bike riders, car driver, roads and law enforcement management.

4.1 Relevant factors of both parties to crash

We observed that although the total number of crashes during daylight is large, the possibility of injury to riders is lower than that of darkness. In darkness, the death toll of electric bike crashes is significantly higher than that in daylight. The crash rate is very high in summer. More electric bike trips may lead to more traffic crashes. However, due to the cold weather and slippery roads in winter, the fatality rate is high. Compared with weekends, there are more electric bike crashes on weekdays, which may be because more people ride electric bikes to work on weekdays. The fatality rate is high on Sundays and Mondays. Therefore, educational activities on the safety of electric bikes are more effective in weekdays and summer when the rate of collision is the highest. However, safety education in weekends and winter also deserves attention due to the high fatality rate [30,31]. The discovery of crash hot spots is conducive to more accurate management of electric bikes. Electric bike and car traffic crashes in Changsha, Hunan Province are mainly concentrated in the commercial, public service and residential areas in the main urban area. In suburbs in darkness, it is more likely to cause serious injury to electric bike riders. In reality, our previous study [32] examined the characteristics of pedestrian collision severity in the main urban area and pedestrian collision crash clusters in urban areas. The results of these two studies strongly suggest that urban traffic safety management should comprehensively consider the influence factors of daylight and darkness and the differences of spatial distribution characteristics.

From the electric bike riders’ point of view, people aged 16–65 are prone to traffic crashes during daylight, among which the elderly aged 56–65 and over 65 are more prone to fatal injury, and the probability of fatal injury is similar. At darkness, the elderly aged 46–55, 56–65 and over 65 are more prone to fatal injury, and the elderly over 65 have a higher probability of fatalities. In daylight, the farmer has a significant impact on the injury severity, contributing 14.7% of the total number of crashes. At the same time, the proportion of self-employed and express delivery employee crashes in daylight and darkness is also very high, contributing 51.5% and 52.6%, respectively. This shows that we should strengthen the safety publicity and education for the middle-aged and elderly, self-employed, express delivery employees and farmers who are riding electric bikes. Some studies also advocate wearing helmets and using electric bikes to protect riders in accordance with the new national standards [33,34].

Among riders' illegal behaviour, not following traffic signals and reverse driving are significantly related to the severity of riders' injuries. Traffic crashes caused by not following traffic signals accounted for the highest proportion, accounting for 10.1% and 10.3%, respectively, during daylight and darkness. The illegal behaviour of riders crossing the motorway without getting off is only significant in daylight, while in darkness, there are more types of rider illegal behaviour, which can lead to more serious injuries. For example, failure to give way as required, it means that the riders violate the road priority: pedestrian > straight ahead > turn left > turn right. On roads without non-motorized lanes, electric bike riders drive off the right side of the roadway, and on roads with non-motorized lanes, electric bike riders do not drive in the non-motorized lane, both of which will hinder motor vehicle driving and increase the crash risk of electric bike riders. Among them, driving off the right side of the roadway accounts for the second in the total number of illegal behaviour of riders at darkness, indicating that the accurate planning and design of non-motorized lane facilities should also be strengthened in road construction.

In order to rectify the illegal behaviours, we should strictly investigate and deal with the road violations and focus on strengthening the on-site law enforcement. For example, we can combine the temporal and spatial characteristics of electric bike crashes, focus on the road segments and intersections where electric bikes pass in a centralized way and strictly control and investigate the illegal behaviours, such as non-compliance with traffic lights, reverse driving, not giving way as required and taking motor lanes. At the same time, increasing investment in scientific and technological facilities, such as actively exploring the application of new scientific and technological systems to capture traffic violations, such as electric bikes running red lights and developing electronic licence plates to track riders' driving tracks and perform real-time monitoring. We should also strengthen publicity, education and guidance. For example, increase traffic safety publicity stations, and strengthen the management of non-motor vehicle lanes, improper operation, not giving way and other illegal behaviours through a combination of education and law enforcement. Encouraging them to wear helmets and other protective equipment to improve riders' traffic safety awareness.

In darkness, when the rider's crash responsibility is no responsible, e.g., when the car driver's crash responsibility is full responsibility, it has a significant impact on injury severity of rider, which indicates that the driver's own reasons are more likely to cause injury to the rider in darkness. Drunk driving increased the fatal probability of riders by 6.46% and 14.52% in daylight and darkness, respectively, and drunk driving was more likely to cause fatalities in darkness. Drinking will distract drivers and reduce their driving ability. We should strengthen the investigation and punishment of drunk driving and improve the visibility of electric bike riders at darkness. Law enforcement and education on car drivers' dangerous behaviuor should also be improved. In electric bike crashes, the four types of side impact in the same direction, in the opposite direction, right angle impact and uncertain side angle can be classified as side impact. The number of side impact crashes during daylight and darkness is the largest, contributing to 71.0% and 61.9% of the total injury crashes of electric bike riders, respectively. In the traffic crash of electric bike riders during daylight, the side impact of right angle and uncertain angle cause significant injury to riders. Side impact is mainly caused by the limited vision of electric bikes at intersections or various serious traffic safety violations, such as violation of traffic lights, driving on motor vehicle roads [35]. In darkness, frontal collision and scratch crashes account for a low proportion, but they are significant factors causing serious riders' crash injuries, which also need attention.

4.2 Road and environmental factors

In terms of roads and environment, in daylight, electric bike and car traffic crashes are easy to occur in bend and slope segment, which increases the possibility of serious and fatal injury to riders. Compared with the straight road segment, due to the limited line of sight of the bend and slope, it is unable to respond to the oncoming car in time, which increases the possibility of crash. It can be considered to add traffic signs, lay anti-skid pavement in bend and slope segment, and improve driving sight distance by clearing obstacles at curves. On roads without traffic control, the impact of crash severity is significant, so traffic control can be increased, such as signal lights, signs, markings, and other safety facilities.

During daylight and darkness, when road physical isolation has only median divider, compared with the road with both median divider and physically separated bicycle roadways, the probability of serious and fatal injury of riders increases. The mixed traffic of motor vehicles and non-motor vehicles is an important cause of traffic safety crashes. Proper isolation between motor vehicles and non-motor vehicles should be implemented to reduce the probability of conflict and casualties between them. On roads without physical separation conditions, coloured non-motor vehicle lane pavements can be adopted, and eye-catching mixed motor vehicle and non-motor vehicle road signs can be set to clarify the driving space of non-motor vehicles. Eye-catching traffic warning signs should be set at the beginning of the separation belt of non-motor vehicles to improve the visibility of electric bike riders. In darkness, the absence of a median divider facility also increases the risk of riders being injured. The median strip of the road should be improved to isolate the traffic moving in different directions or at different speeds, such as the anti-collision barrier can separates the traffic lane. Crash barriers and other physical measures should be ‘soft’ and help in the event of a collision and make the edges safer.

On general urban roads and urban expressways, the traffic environment is good and the driving speed of car drivers is high, so the severity of traffic crashes is more serious than that of other roads. Under darkness conditions, the possibility of serious crash between rainy, foggy, snowy and cloudy days is higher than that on the clear night. Under the influence of adverse weather, electric bike riders should slow down on slippery roads because the speed is very high and it is easy to sideslip when turning. Crashes in the suburbs at darkness lead to a high fatal rate for riders, which may be due to poor lighting conditions and poor road conditions. Most substandard roads are in suburban and rural areas. The damage of some roads and incomplete renewal of road signs lead to high crash mortality. In the future, the technical requirements of the road shall meet the actual needs, and the damaged pavement, signboard and protective wall shall be repaired in time. Finally, night road lighting facilities, such as streetlamps are essential, and the construction of streetlamps should be strengthened and improved.

5. Conclusions

We analysed the electric bike crash in Changsha, using the police crash report provided by the Ministry of Transportation of Hunan Province. We noted the difference of the time and space distribution, collision characteristics and impact factors of electric bike collisions. In addition, we have made recommendations on implementation and education to prevent and reduce future crashes. Compared with previous studies, this study investigates the factors influencing the injury severity of riders by dividing the darkness and daylight periods and studies the spatial distribution characteristics of the corresponding regions. On this basis, the protection measures of the main roads in the hot spot areas of crashes can effectively reduce electric bike crashes in complicated traffic. Improving lighting conditions at darkness and the safety awareness of riders and drivers is also an effective strategy to improve the travel environment. Reasonable use of non-motorized lanes is conducive to ensure the safety of driving. In addition, illegal behaviour is an important factor leading to collision and casualties. The relevant laws and regulations should be combined with the safety education activities of electric bike riders and car drivers to improve their understanding of traffic regulations.

However, the study has several limitations. First, because the speed of electric bike was not recorded in the police report, the influence of the speed factor could not be tested. Second, we only analyse crashes between electric bikes and cars; the crash may also be caused by other reasons, such as falling and collision with a stationary object. In addition, some uninjured and small property damage may not be reported to the police, combined with the data of the hospital emergency department can provide better diagnostic performance.

ACKNOWLEDGEMENTS

This work was supported by the National Natural Science Foundation of China (Grant No. 52172399/51875049); the Key Research and Development Program of Hunan Province (Grant No. 2020SK2099); the Scientific Research Fund of Hunan Provincial Education Department (Grant No. 21A0193).

Conflict of interest

Authors declare no conflict of interest.

References

World Health Organization

Global Status Report on Road Safety 2018

World Health Organization

2018

https://www.who.int/violence_injury_prevention/road_safety_status/2018/en/

Month:	Total Views:
August 2022	54
September 2022	102
October 2022	120
November 2022	61
December 2022	35
January 2023	39
February 2023	55
March 2023	1,753
April 2023	61
May 2023	17
June 2023	25
July 2023	30
August 2023	24
September 2023	37
October 2023	22
November 2023	29
December 2023	39
January 2024	31
February 2024	26
March 2024	45
April 2024	49
May 2024	31
June 2024	23
July 2024	29
August 2024	26
September 2024	29
October 2024	25
November 2024	27
December 2024	4
January 2025	41
February 2025	21
March 2025	31
April 2025	9
May 2025	2

Article Contents

Injury severity analysis of electric bike crashes in Changsha, Hunan Province: taking different lighting conditions into consideration

Abstract

1. Introduction

2. Analysis

2.1 Data preparation

2.1.1 The basis for dividing the time between daylight and darkness

2.1.2 Characteristics of electric bike crash

2.2 Method

2.2.1 Geographic Information System

2.2.2 Ordered Probit model

3. Results

3.1 Temporal and spatial distribution characteristics of crashes

3.1.1 Time distribution characteristics

3.1.2 Spatial distribution characteristics

3.2 Analysis results from the OP model

3.2.1 Independent variable screening

3.2.2 Daylight condition model

3.2.3 Darkness condition model

4. Discussion

4.1 Relevant factors of both parties to crash

4.2 Road and environmental factors

5. Conclusions

ACKNOWLEDGEMENTS

Conflict of interest

References

Citations

Views

Altmetric

Email alerts

Citing articles via

Most Read

Latest

This Feature Is Available To Subscribers Only