Está en la página 1de 15

Revista de investigación de seguridad xxx (xxxx) xxx

Listas de contenidos disponibles en ScienceDirect

Revista de Investigación de Seguridad

página de inicio de la revista: www.el sevier. com / localizar / j sr

Gravedad de las lesiones de los pasajeros en los autobuses públicos: un análisis comparativo de las lesiones por
colisión y las lesiones sin colisión.

Hanchu Zhou una , si , Chen Yuan una , Ni Dong C , re , SC Wong mi , Pengpeng Xu mi , ⇑


una Escuela de Ingeniería de Tráfico y Transporte, Universidad Central del Sur, Changsha, Hunan, China
si Escuela de Ciencia de Datos, Universidad de la Ciudad de Hong Kong, Hong Kong, China

C Escuela de Transporte y Logística, Southwest Jiaotong University, Chengdu, Sichuan, China


re Departamento de Ingeniería Civil y Ambiental, Universidad de Washington, Seattle, WA, Estados Unidos
mi Departamento de Ingeniería Civil, Universidad de Hong Kong, Hong Kong, China.

información del artículo resumen

Historia del articulo: Introducción: Aunque los autobuses públicos han demostrado ser un medio de transporte relativamente seguro, el número de lesiones a los pasajeros de los
Recibido el 16 de noviembre de 2019 autobuses públicos está lejos de ser insignificante. Los estudios existentes sobre la seguridad del autobús público se han centrado principalmente en las lesiones
Recibido en forma revisada el 27 de marzo de 2020 Aceptado el
causadas por colisiones. Sorprendentemente, se ha dedicado un esfuerzo limitado a identificar los factores que aumentan la gravedad de las lesiones de los
16 de abril de 2020
pasajeros en incidentes sin colisión. Método:
Disponible en línea xxxx
Por lo tanto, nuestro estudio investigó el riesgo de lesiones de los pasajeros de autobuses públicos involucrados en incidentes de colisión e
incidentes de no colisión comparativamente, basado en un conjunto de datos reportado por la policía de 17.383 pasajeros heridos en
Palabras clave:
autobuses públicos franquiciados durante un período de 10 años en Hong Kong. Se estableció un modelo logístico de parámetros aleatorios
Autobús público
para estimar la probabilidad de lesiones fatales y graves a los pasajeros en función de varios factores. Resultados: Nuestros resultados
Gravedad de lesiones

Lesiones por colisión


indicaron inconsistencias sustanciales en los efectos de los factores de riesgo entre modelos de lesiones sin colisión y lesiones por colisión. La
Lesiones sin colisión gravedad de las lesiones de los pasajeros tendió a aumentar significativamente cuando ocurrieron incidentes sin colisión debido a la velocidad
Parámetros aleatorios excesiva de los conductores de autobuses, en autobuses de dos pisos, en áreas menos urbanizadas, en invierno, en fuertes lluvias, durante el
día y por la noche sin alumbrado público . También se descubrió que las pasajeros ancianas eran más propensas a sufrir lesiones fatales o
graves en incidentes sin colisión si perdían el equilibrio al abordar, bajarse o pararse en un autobús. En comparación, los siguientes factores se
asociaron con una mayor probabilidad de lesiones fatales o graves en incidentes de colisión: mujeres de edad avanzada, pasajeros de pie que
perdieron el equilibrio, autobuses fuera del control del conductor, autobuses de dos pisos, Aplicaciones prácticas: Con base en nuestro análisis
comparativo, se recomendaron contramedidas más específicas, a saber, '' 4E ”(ingeniería, aplicación, emergencia y educación) y '' 3A”
(conciencia, apreciación y asistencia), para mitigar las lesiones por colisión y las lesiones sin colisión. a pasajeros de autobuses públicos,
respectivamente.

2020 National Safety Council y Elsevier Ltd. Todos los derechos reservados.

1. Introducción 2014 ), mientras que hasta 2.000 pasajeros se lesionan inesperadamente cada año ( HKTD, 2018 ) La
seguridad de los pasajeros de los autobuses debe mejorarse sustancialmente si se promueve el
Aunque los autobuses públicos han demostrado ser un modo de transporte relativamente seguro autobús público como un modo de transporte seguro y atractivo. La mejora en la seguridad también
( Beck, Dellinger y O'Neil, 2007; Xu, Dong, Wong y Huang, 2019 ), el número de lesiones a los alentaría a más personas a tomar autobuses regularmente para viajar diariamente, fomentando así
pasajeros de autobuses públicos dista mucho de ser insignificante. Esto es particularmente cierto en una comunidad más habitable y sostenible.
áreas altamente urbanizadas con una población densa, donde los autobuses públicos desempeñan
un papel destacado para garantizar una movilidad asequible y adecuada para la mayoría de los
residentes locales. En Hong Kong, se estima que casi el 30% de los viajes diarios se realizan a En las últimas dos décadas, varios estudios han investigado la seguridad del autobús (ver: Tabla
través de autobuses públicos franquiciados ( HKTD, A1 en el Apéndice ) Los investigadores han intentado describir las características de las lesiones a los
pasajeros del autobús ( af Wåhlberg, 2002, 2004; Björnstig et al., 2005; Halpern, Siebzehner,
Aladgem, Sorkine y Bechar, 2005; McGeehan et al., 2006; Yang et al., 2009; Zunjic, Sremcevic,

⇑ Autor correspondiente en: Room LG-208, Composite Building, The University of Hong Kong, Pokfulam Road,
Sijacki y Sijacki, 2012; Barnes et al., 2016; Silvano y Ohlin, 2019; Siman-Tov,

Hong Kong, China.


Dirección de correo electrónico: pengpengxu@yeah.net (P. Xu).

https://doi.org/10.1016/j.jsr.2020.04.003
0022-4375 / 2020 National Safety Council y Elsevier Ltd. Todos los derechos reservados.

Cite este artículo como: H. Zhou, C. Yuan, N. Dong et al., Severidad de las lesiones de los pasajeros en los autobuses públicos: un análisis comparativo de las lesiones por colisión y las lesiones sin colisión, Journal of
Safety Research, https://doi.org/10.1016/j.jsr.2020.04.003
2 H. Zhou y col. / Revista de investigación de seguridad xxx (xxxx) xxx

Radomislensky, Marom, Kapra y Peleg, 2019 ); para cuantificar el riesgo de lesiones sin colisión
sufridas por los pasajeros del autobús ( Elvik, 2019 ); evaluar el desempeño de seguridad de las
compañías de autobuses ( Chang y Yeh, 2005; Strathman, Wachana y Callas, 2010 ) y conductores
de autobuses ( Salminen, Vartia y Giorgiani, 2009; Blower & Green, 2010; Tseng, 2012 ); para
investigar los factores que contribuyen a la frecuencia de los accidentes de autobús en las
intersecciones señalizadas ( Shahla, Shalaby, Persaud y Hadayeghi, 2009 ), en segmentos de
carretera ( Chimba, Sando y Kwigizile, 2010 ), y en los sistemas de tránsito rápido de autobuses ( Gómez
y Bocarejo, 2015 ); evaluar la efectividad de los programas de capacitación ( Rey, Hinebaugh y
Fernández, 2002 ) y estrategias de prioridad de autobús ( Goh, Currie, Sarvi y Logan, 2014 ) en la
reducción de accidentes de autobús; y para examinar los factores que afectan la gravedad de los
accidentes de autobús ( Barua y Tay, 2010; Chimba et al., 2010; Rahman, Kattan y Tay, 2011; Kaplan
y Prato, 2012; Prato y Kaplan, 2014; Feng

Li, Ci y Zhang, 2016; Yoon, Kho y Kim, 2017; Sam, Daniels, Brijs, Brijs y Wets, 2018 ) Los estudios Figura 1. El número de pasajeros de autobuses públicos heridos en incidentes sin colisión e incidentes de colisión en
existentes se han centrado principalmente en las lesiones causadas por colisiones ( af Wåhlberg, Hong Kong de 2008 a 2017 (fuente de datos: Fuerza de Policía de Hong Kong).

2002, 2004; Rey et al., 2002; McGeehan et al., 2006; Shahla et al., 2009; Yang et al., 2009; Barua y
Tay, 2010; Blower & Green, 2010; Chimba et al., 2010; Rahman et al., 2011; Kaplan y Prato, 2012;
Tseng, 2012; Goh et al., 2014; Prato y Kaplan, 2014; Gómez y Bocarejo, 2015; Feng et al., 2016;
y las diferencias de los efectos de numerosos factores sobre la gravedad de las lesiones de los
Yoon et al., 2017; Sam et al., 2018 ) Se ha dedicado un esfuerzo de investigación relativamente
pasajeros de autobuses entre incidentes sin colisión e incidentes de colisión. Tal análisis comparativo
limitado a investigar la causa de las lesiones sin colisión a los pasajeros del autobús ( Björnstig et al.,
es imperativo para mejorar el desempeño de seguridad de los autobuses públicos, ya que puede
2005; Halpern et al., 2005; Palacio, Tamburro, O'Neill y Simms, 2009; Strathman et al., 2010;
ayudar a las autoridades locales a desarrollar una gama de contramedidas basadas en evidencia
Kendrick, Drummond, Logan, Barnes y Worthington, 2015; Elvik, 2019; Silvano y Ohlin, 2019;
para reducir las bajas de los autobuses debido a diferentes mecanismos de lesiones.
SimanTov et al., 2019 ), a pesar de la considerable proporción de lesiones sufridas por los pasajeros
del autobús como resultado de incidentes sin colisión (por ejemplo, aproximadamente el 70% en
Hong Kong según los registros policiales; ver Figura 1 ) Como resultado, la influencia de los factores
de riesgo sobre la gravedad de las lesiones de los pasajeros de autobuses en estos incidentes sin
colisión sigue siendo en gran medida desconocida. 2. Métodos

Nuestros datos de lesiones se obtuvieron del Sistema de base de datos de accidentes de


tránsito, que es mantenido por la Policía de Hong Kong y el Departamento de Transporte de Hong
Kong. La información detallada sobre las víctimas, la participación del vehículo y el entorno del
accidente es recopilada rutinariamente por oficiales de policía en las escenas de accidentes ( Meng,
Estudios anteriores han sugerido que la gravedad de los accidentes de autobús tiende a Xu, Wong, Huang y Li, 2017; Zhai, Huang, Sze, Song y Hon, 2019 ) Solo se registran en la base de
aumentar significativamente durante el invierno ( Rahman et al., 2011; Feng et al., 2016 ), en los fines datos aquellos que resultan en lesiones en la vía pública ( Xu, Xie, Dong, Wong y Huang, 2019 ) Al
de semana ( Barua y Tay, 2010; Feng et al., 2016; Sam et al., 2018 ), Por la noche ( Prato y Kaplan, recuperar simultáneamente información sobre los tipos de víctimas y vehículos, se descubrió que
2014; Feng et al., 2016; Sam et al., 2018 ), en curva ( Yoon et al., 2017; Sam et al., 2018 ) o carreteras 18,859 pasajeros resultaron heridos en los autobuses públicos de 2008 a 2017. Después de eliminar
resbaladizas ( Prato y Kaplan, 2014 ), en calles de doble sentido ( Barua y Tay, 2010 ), en calles con muestras con registros incompletos, se retuvieron 17.383 (92,17%) observaciones válidas para un
límites de velocidad más altos ( Kaplan y Prato, 2012; Prato y Kaplan, 2014; Feng et al., 2016; Yoon análisis más detallado. De estos, el 73.74% involucraron lesiones de 11,852 incidentes sin colisión,
et al., 2017 ), en calles sin medianas ( Barua y Tay, 2010; Sam et al., 2018 ) y con la participación de mientras que los 4,565 pasajeros restantes resultaron heridos en 2,350 incidentes de colisión. Figura
usuarios vulnerables de la carretera ( Barua y Tay, 2010; Prato y Kaplan, 2014; Sam et al., 2018 ) o 2 ilustra todo el proceso de extracción de lesiones por colisión y lesiones sin colisión sufridas por
vehículos pesados ​( Prato y Kaplan, 2014 ) Sin embargo, estos hallazgos no deberían generalizarse pasajeros de autobuses públicos en Hong Kong.
directamente a incidentes sin colisión, porque estos dos tipos de eventos tienen mecanismos
esencialmente diferentes. A diferencia de las colisiones que generalmente resultan de impactos con
vehículos, peatones u objetos, las lesiones de los pasajeros de los autobuses en incidentes que no
son de colisión implican principalmente caídas dentro de un autobús debido a una aceleración,
frenado o giro abruptos, y caídas al subir o bajar de un autobús ( Elvik, 2019 ) Por lo tanto, para lograr La policía de Hong Kong define la gravedad de las lesiones como fatales, graves y leves. 1 . Dado
una comprensión profunda de los factores que contribuyen a las lesiones graves de los pasajeros en que las muertes representaron mucho menos del 1% de nuestras muestras, no se esperaba que la
los autobuses públicos, se deben desarrollar modelos separados para incidentes de colisión y sin combinación de muertes con lesiones graves para formar una sola categoría afectara
colisión. sustancialmente la inferencia ( Yau, 2004; Xu et al., 2016; Meng et al., 2017; Chang, Xu, Zhou, Chan
y Huang, 2019; Wang, Huang, Xu, Xie y Wong, 2019; Zhai et al., 2019 ) La variable dependiente en
nuestros modelos se convirtió así en un resultado de lesión dicotómica en la naturaleza, en la cual la
respuesta de interés se refería a fatalidad y lesiones graves, con lesiones leves como

Nuestro estudio, por lo tanto, tiene la intención de investigar comparativamente el riesgo de lesiones la referencia (es decir, Y yo ¼ 1 si el yo th re yo ¼ 1; 2; :::; norte Þ autobús público pasajero
de los pasajeros de autobuses públicos involucrados en incidentes sin colisión e incidentes de colisión, ger murió o resultó gravemente herido, de lo contrario Y yo ¼ 0 0 Þ. Al agregar los perfiles de
en base a un conjunto completo de datos de accidentes, vehículos y accidentes, el predictor
17.383 pasajeros heridos en autobuses públicos franquiciados durante un período de 10 años en variables que reflejan las características de los pasajeros lesionados (por ejemplo, edad, sexo,
Hong Kong. Se establece un modelo logístico de parámetros aleatorios para estimar la probabilidad asiento ocupado, uso del cinturón de seguridad y fallas), junto con las características de los
de lesiones fatales y graves a los pasajeros de autobuses públicos en función de varios factores, autobuses que transportan a los pasajeros (por ejemplo,
incluidas las características demográficas de los pasajeros, las fallas de los conductores de
autobuses, las características de los autobuses y los factores ambientales. Hasta donde sabemos, 1 Las víctimas que mueren inmediatamente en la escena o dentro de los 30 días posteriores al accidente se cuentan

nuestro estudio es uno de los primeros en explorar las similitudes. como víctimas fatales, mientras que los que ingresan en hospitales por más de 24 horas o menos se definen como

lesiones graves o leves, respectivamente.

Cite este artículo como: H. Zhou, C. Yuan, N. Dong et al., Severidad de las lesiones de los pasajeros en los autobuses públicos: un análisis comparativo de las lesiones por colisión y las lesiones sin colisión, Journal of
Safety Research, https://doi.org/10.1016/j.jsr.2020.04.003
H. Zhou y col. / Revista de investigación de seguridad xxx (xxxx) xxx 3

Víctima Participación del vehículo Ambiente de Accidentes

Numero de accidente Numero de accidente Numero de accidente


Número en el vehículo Número de vehículo Gravedad del accidente

Número de víctimas Edad del conductor Lugar del accidente


Papel de la víctima Sexo conductor Tiempo de accidente

Edad de la víctima Error del conductor Pega y corre


Sexo casual Año del vehículo Numero de vehiculos
Gravedad de la lesión Año de manufactura Movimiento de vehículos

Lugar de la lesión. Tipo de vehiculo Numero de bajas


Uso del cinturón de seguridad Colisión de vehículos con Dentro de 70m de una unión
Asiento ocupado Primer punto de impacto Dentro de 20m de una unión
Culpa de la víctima Maniobra del vehículo Control de unión
Tipo de unión
Tipo de camino

Límite de velocidad

La congestión del tráfico

Clima
Pasajero Autobús público
Lluvia
Superficie de la carretera

Luz natural
luz de la calle
si
si

Pasajeros heridos en autobuses públicos


(Número = 18,859)

Eliminar muestras con registros incompletos

Pasajeros heridos en autobuses públicos


(Número = 17,383)

Vehículo
colisión con

Ninguna Otros

Lesiones sin colisión Lesiones por colisión


(Número = 12,818) (Número = 4,565)

Figura 2. El diagrama de flujo solía extraer lesiones sin colisión y lesiones por colisión sufridas por pasajeros de autobuses públicos en Hong Kong.

la edad, el género y las fallas del conductor del autobús, el tipo de autobús, la compañía de afectar la gravedad de las lesiones casi no están disponibles para recoger. Si estos factores no
autobuses, el año de fabricación y la maniobra del vehículo) y los factores ambientales (por ejemplo, observados se correlacionan con los observados, se estimarán los parámetros sesgados y se
hora, ubicación, clima, tipo de carretera y condiciones de iluminación) se extrajeron colectivamente pueden extraer inferencias incorrectas ( Xu y Huang, 2015; Mannering, Shankar y Bhat, 2016; Xu,
y se resumen en tabla 1 . Aprovechando la función de enlace logit, tenemos: Huang, Dong y Wong, 2017; Huang, Chang, Zhou y Lee, 2019; Waseem, Ahmed y Saeed, 2019 )
Siguiendo Milton

Y yo Binomio re pags yo; norte Þ


Shankar y Mannering (2008), Anastasopoulos y Mannering (2011), Wu et al. (2014, 2016), Chang et
K
PAGS ð1Þ al. (2019), Guo, Wu, Lu y Zhou (2019), Zhai et al. (2019) y Zhang y Hassan (2019) , desarrollamos un
logit re pagsyoÞ ¼ Iniciar sesión re pags yo Þ 0¼0 si þ
1 pags yo
si k X ik
k¼1 modelo logístico de parámetros aleatorios para abordar esta heterogeneidad entre individuos debido
a factores no observados:
dónde pags yo denota la probabilidad de Y yo ¼ 1) X ik es el
k th kð ¼ 1; 2; :::; K Þ variable independiente. si 0 0 es la intercepción, y si k
se refiere a k El coeficiente de regresión a estimar.
Eq. (1) asume que el efecto de cada variable es constante en
si ik ¼ si k þ tu ik re 2 Þ
Las observaciones. Esta hipótesis fundamental es casi siempre
violado, porque en realidad cada víctima bajo investigación exhi-
bits características únicas que pueden influir en la severidad dónde tu ik es un término distribuido al azar, por ejemplo, normalmente

Viene diferencialmente. El impacto de las variables explicativas es término distribuido con media de cero y varianza de r 2 k. En la práctica,
por lo tanto, se espera que varíe individualmente ( Mannering & Bhat, 2014; un parámetro aleatorio se usa siempre r k es significativamente mayor que cero; de lo contrario, el
Cai, Abdel-Aty, Lee, Wang y Wang, 2018; Dong, Ma, Chen y parámetro debe fijarse a través de la observación
Chen, 2018; Chen, Song y Ma, 2019; Li et al., 2019 ) Adicionalmente, iones Basado en el método simulado de máxima verosimilitud, los parámetros se pueden estimar
algunos factores (por ejemplo, la posición de pie de los pasajeros heridos) que adecuadamente a través de 200 sorteos Halton

Cite este artículo como: H. Zhou, C. Yuan, N. Dong et al., Severidad de las lesiones de los pasajeros en los autobuses públicos: un análisis comparativo de las lesiones por colisión y las lesiones sin colisión, Journal of
Safety Research, https://doi.org/10.1016/j.jsr.2020.04.003
44 H. Zhou y col. / Revista de investigación de seguridad xxx (xxxx) xxx

tabla 1
Las características de los pasajeros heridos en autobuses públicos en Hong Kong, estrati fi cados por lesiones sin colisión y lesiones por colisión, 2008–2017.

Lesiones sin colisión Lesiones por colisión

Variables norte % norte %

Niveles de lesiones de los pasajeros.

Mortalidad o lesiones graves 1,348 10,52 263 5,76


Lesiones leves 11,470 89,48 4,302 94,24
Características del pasajero
Edad de la víctima (unidad: año)

6 6 17 771 6.01 302 6.62


18–34 1,125 8.78 1,065 23,33
35-49 2,159 16,84 1,167 25,56
50-64 4,388 34,23 1,261 27,62
PAGS sesenta y cinco 4,375 34,13 770 16,87
Género de la víctima

Masculino 3,649 28,47 1,526 33,43


Hembra 9,169 71,53 3,039 66,57
Lesión craneal
si 5,264 41,07 1,922 42,10
No 7,554 58,93 2,643 57,90
Lesión en el brazo

si 2,749 21,45 898 19,67


No 10,069 78,55 3,667 80,33
Lesión en una pierna

si 2,757 21,51 967 21,18


No 10,061 78,49 3,598 78,82
Uso del cinturón de seguridad

si 1,690 13,18 1,147 25,13


No 10,628 82,92 2,951 64,64
Desconocido 500 3.90 467 10,23
Asiento ocupado
Sentado 4,071 31,76 2,801 61,36
Parado en cubierta 8,338 65.05 1,369 29,99
Desconocido 409 3.19 395 8.65
Factor contribuyente de siniestros
Ninguna 4,202 32,78 3,480 76.23
Perder el equilibrio al subir o bajar Perder el equilibrio en la 1,125 8.78 40 0.88
escalera 1,664 12,98 70 1.53
Perder el equilibrio, no en la escalera Otros 4,171 32,54 206 4.51
1,656 12,92 769 16.85
Características del conductor del autobús

Edad del conductor del autobús (unidad: año)

18-24 70 0.55 30 0.66


25-34 1,235 9.63 496 10.87
35-44 2,892 22.56 944 20.68
45-54 5,250 40.96 1,850 40.53
55-64 3,299 25.74 1,191 26.09
PAGS sesenta y cinco 72 0.56 54 1.18
Género del conductor del autobús

Masculino 12,201 95.19 4,349 95.27


Hembra 617 4.81 216 4.73
Factor contribuyente del conductor del autobús

Ninguna 8,632 67.34 2,089 45.76


Conduciendo demasiado rápido 134 1.05 1 0.02
No mantener una distancia segura Arranque 5 0.41 649 14.22
o parada negligente Violación del control de 190 1.48 55 1.20
tráfico 3 0.02 49 1.07
Abriendo la puerta negligentemente 70 0.55 6 0.13
Cambiar de carril negligentemente 78 0.61 88 1.93
Girando negligentemente 23 0.18 71 1.56
Perder el control del vehículo 15 0.12 68 1.49
Conducir sin atención 539 4.21 905 19.82
Desviarse o detenerse repentinamente 645 5.03 61 1.34
No garantizar la seguridad de los pasajeros Otros 1,273 9.93 36 0.79
1,164 9.08 487 10.67
Características del bus
Año de manufactura
1989-1995 1,079 8.42 471 10.32
1996-2000 5,465 42.64 2,017 44.18
2001-2005 2,697 21.04 828 18.14
2006-2010 1,601 12.49 571 12.51
2011-2017 1,976 15.42 678 14.85
Antigüedad del vehículo (unidad: año)

65 2,977 23.38 977 21.40


6–10 2,501 19.51 955 20.92
11–15 4,638 36.18 1,829 40.07
P 16 2,682 20.93 804 17.61

Please cite this article as: H. Zhou, C. Yuan, N. Dong et al., Severity of passenger injuries on public buses: A comparative analysis of collision injuries and non-collision injuries, Journal of Safety Research, https://doi.org/10.1016
H. Zhou et al. / Journal of Safety Research xxx (xxxx) xxx 5

Table 1 ( continued)

Non-collision injuries Collision injuries

Variables N % N %

Vehicle type
Single-decker 304 2.37 173 3.79
Double-decker 9,396 73.30 3,919 85.85
Unknown 3,118 24.33 473 10.36
Vehicle company
New World First Bus Kowloon 1,990 15.53 604 13.23
Motor Bus 5,410 42.21 2,454 53.76
New Lantau Bus 74 0.58 54 1.18
City Bus 2,040 15.92 819 17.94
Long Win Bus 186 1.45 161 3.53
Unknown 3,118 24.33 473 10.36
Vehicle collision with
None 12,818 100.00 0 0.00
Vehicle 0 0.00 3,795 83.13
Pedestrian 0 0.00 117 2.56
Object 0 0.00 653 14.31
First point of impact
Front 8 0.06 2,604 57.04
Back 0 0.00 756 16.56
Offside 3 0.02 597 13.08
Nearside 7 0.05 480 10.51
No impact 12,800 99.86 128 2.80
Vehicle maneuver
Going straight ahead 8,330 64.99 3,395 74.37
Changing lanes or merging Overtaking 92 0.72 103 2.26
5 0.04 11 0.24
Making turn 674 5.26 263 5.76
Slowing or stopping 2,029 15.83 431 9.44
Stopped in traffic 502 3.92 174 3.81
Starting in traffic 881 6.87 61 1.34
Parked 87 0.68 10 0.22
Others 218 1.70 117 2.56
Environmental characteristics
Number of vehicles involved
1 11,528 89.94 685 15.01
2 1,255 9.79 2,990 65.50
P3 35 0.27 890 19.50
Vehicle movement
One moving vehicle 11,787 91.96 985 21.58
Two moving vehicles from same direction Two moving vehicles 884 6.90 2,291 50.19
from opposite direction Two moving vehicles on different roads 30 0.23 215 4.71
102 0.80 280 6.13
More than two moving vehicles from same direction More than two moving 14 0.11 762 16.69
vehicles from opposite direction More than two moving vehicles on different roads 0 0.00 17 0.37
Hit and run 1 0.01 15 0.33

Yes 95 0.74 35 0.77


No 12,723 99.26 4,530 99.23
Area
Hong Kong Island 3,687 28.76 1,158 25.37
Kowloon 5,203 40.59 1,377 30.16
New Territories 3,928 30.64 2,030 44.47
Year
2008 1,024 7.99 381 8.35
2009 1,196 9.33 378 8.28
2010 1,153 9.00 503 11.02
2011 1,213 9.46 487 10.67
2012 1,250 9.75 526 11.52
2013 1,374 10.72 601 13.17
2014 1,453 11.34 415 9.09
2015 1,430 11.16 409 8.96
2016 1,437 11.21 374 8.19
2017 1,288 10.05 491 10.76
Season
Winter 3,628 28.30 1,083 23.72
Spring 3,204 25.00 1,094 23.96
Summer 2,869 22.38 1,275 27.93
Autumn 3,117 24.32 1,113 24.38
Weekend
Yes 3,616 28.21 1,276 27.95
No 9,202 71.79 3,289 72.05
Hour of day
00:00–03:59 297 2.32 110 2.41

( continued on next page)

Please cite this article as: H. Zhou, C. Yuan, N. Dong et al., Severity of passenger injuries on public buses: A comparative analysis of collision injuries and non-collision injuries, Journal of Safety Research, https://doi.org/10.1
6 H. Zhou et al. / Journal of Safety Research xxx (xxxx) xxx

Table 1 ( continued)

Non-collision injuries Collision injuries

Variables N % N %

04:00–07:59 721 5.62 336 7.36


08:00–11:59 2,988 23.31 1,306 28.61
12:00–15:59 3,322 25.92 1,095 23.99
16:00–19:59 3,420 26.68 1,064 23.31
20:00–23:59 2,070 16.15 654 14.33
Within 20 m of a junction
Yes 4,617 36.02 1,268 27.78
No 8,201 63.98 3,297 72.22
Within 70 m of a junction
Yes 5,266 41.08 1,416 31.02
No 7,552 58.92 3,149 68.98
Junction control
No control 924 7.21 267 5.85
Stop sign 51 0.40 33 0.72
Give way 281 2.19 104 2.28
Traffic signal 3,358 26.20 861 18.86
Police 3 0.02 3 0.07
Not junction 7,552 58.92 3,149 68.98
Unknown 649 5.06 148 3.24
Junction type
Roundabout 166 1.30 66 1.45
T-junction 3,315 25.86 805 17.63
Y-junction 130 1.01 58 1.27
Crossroad 1,191 9.29 348 7.62
Other 464 3.62 139 3.04
Not junction 7,552 58.92 3,149 68.98
Road type
One-way street 5,616 43.81 1,869 40.94
Two-way street 3,388 26.43 1,135 24.86
Dual carriageway 2,819 21.99 1,128 24.71
More than two carriageways Speed limit 995 7.76 433 9.49

6 30 km/h 29 0.23 10 0.22


Between 40 and 60 km/h 12,282 95.82 3,510 76.89
P 70 km/h 507 3.96 1,045 22.89
Traffic congestion
None 6,865 53.56 2,352 51.52
Moderate 3,745 29.22 1,367 29.95
Severe 2,074 16.18 804 17.61
Unknown 134 1.05 42 0.92
Weather
Clear 12,284 95.83 4,250 93.10
Dull 420 3.28 238 5.21
Fog/mist 98 0.76 62 1.36
Strong wind 16 0.12 15 0.33
Rain
Not raining 11,849 92.44 3,979 87.16
Light rain 865 6.75 521 11.41
Heavy rain 104 0.81 65 1.42
Road surface
Wet 1,033 8.06 628 13.76
Dry 11,785 91.94 3,937 86.24
Natural light
Daylight 8,995 70.17 3,212 70.36
Dawn/Dusk 343 2.68 135 2.96
Dark 3,480 27.15 1,218 26.68
Street light
Good 5,133 40.05 1,819 39.85
Poor 22 0.17 9 0.20
Not lit or none 2,796 21.18 1,107 24.25
Daylight 4,867 37.97 1,630 35.71

( Train, 2009 ). The econometric software NLOGIT 5.0 was used to calibrate the models. facture and vehicle age, between vehicle type and vehicle company, between the number of vehicles
involved and vehicle movement, between rain and road surface, and between natural light and street
lighting condition, with the Spearman’s correlation parameters estimated at greater than 0.70.
Similarly, within 20 m of a junction, within 70 m of a junction, junction control, and junction type were
3. Results and discussion highly correlated, suggesting that these four variables should not be simultaneously added to the
models. Other variables showed weak collinearity, as their Spearman’s correlation
For model specification, a correlation test was conducted first to ensure the non-inclusion of
highly correlated variables. The correlation analysis indicated a high correlation between year of
manu-

Please cite this article as: H. Zhou, C. Yuan, N. Dong et al., Severity of passenger injuries on public buses: A comparative analysis of collision injuries and non-collision injuries, Journal of Safety Research, https://doi.org/10.1016
H. Zhou et al. / Journal of Safety Research xxx (xxxx) xxx 7

parameters were all less than 0.40. In the initial model, we included all uncorrelated variables ( Chang standing on the deck, winter, two-way streets, and streets without lighting at night were statistically
et al., 2016; Xie, Dong, Wong, Huang, & Xu, 2018 ). The likelihood ratio test was then used to significant in the models of noncollision injuries, but became insignificant in the models of collision
guarantee that each added variable significantly improved the overall model performance. After injuries. Similarly, the variable of within 70 m of a junction was significant solely in the collision injuries
comparing several random parameter density functions including normal, lognormal, triangular, and models. Second, relative to the non-collision injuries models, the effects of several risk factors, such
uniform distributions, the normal distribution was found to be statistically superior and was thus as double-decker buses and Hong Kong Island, changed substantially in the collision injuries models.
selected to fit our random parameters model. Specifically, the OR of double-decker buses in the random parameters noncollision injuries model
increased sharply from 1.52 to 2.59 when applied to collision injuries. These findings highlight that
translation of the findings directly from collision injuries models to the non-collision injuries would lead
to biased estimates and inadequate inferences.
For the purpose of comparison, in addition to the random parameters logistic model, we
developed fixed parameters logistic model. As a result, four models were eventually established with
two for non-collision injuries and the other two for collision injuries. The performance of these models
is compared below, followed by the presentation and interpretation of the parameter estimates.

More interestingly, a comparison between the fixed parameters and random parameters models
indicated that the explanatory power of the models improved considerably after adjustment for
unobserved heterogeneity, as the number of significant variables increased greatly in the random
3.1. Model performance comparison
parameters models. For example, the insignificant variables of male, double-decker buses, and
two-way streets in the fixed parameters non-collision injuries model became highly significant when
Table 2 shows the results of goodness-of-fit measures for the calibrated models. Noticeably,
random parameters were specified. The same conclusion held true for the variables of head injury,
although the performance of random parameters model was comparable to that of the fixed
double-decker buses, within 70 m of a junction, and light rain in the collision injuries models. Given
parameters model in terms of Akaike information criterion, the likelihood ratio test indicated that the
that the random parameters model resulted in a better performance, we chose it to interpret our
random parameters model was statistically superior at the 95% confidence level with a p- value of 0.00
results in the subsequent section.
and 0.02 for non-collision injuries and collision injuries, respectively, suggesting that explicitly
accounting for the heterogeneous effects of risk factors would be conducive to a substantial
improvement in goodness-of-fit. This result is expected, because unlike the fixed parameters model
with constant parameters for all explanatory variables, the random parameters model is more flexible
as its parameters are allowed to vary across observations.

3.2.1. Severity model results for non-collision injuries


As Tables 3 and 4 show, according to the results of random parameters logistic model, a total of
12 factors were significantly associated with the severity of injuries to bus passengers in noncollision
incidents. The signs of these parameters were generally consistent with empirical judgments.
3.2. Parameter estimates

Table 3 summarizes the parameter estimates in the fixed parameters and random parameters The likelihood of fatal and severe injuries to passengers was found to gradually increase with
models applied to the severity of non-collision injuries and collision injuries to public bus passengers. Fig. passenger age. Compared with those under 17 years of age, passengers aged 65 or above were
3 illustrate the distribution of the random parameters in the models of non-collision injuries and
collison injuries, respectively. A 5% level of significance was used as the threshold to determine 8.17 times more likely to be fatally or severely injured if involved in a non-collision incident. This
whether the parameters differed significantly from zero. Variables that were insignificant in all four finding is largely expected, given the increasing fragility of the body, slower gait, longer reaction time,
models were then excluded. To facilitate the interpretation of the effects of these independent and poorer ability to maintain balance due to aging ( Kendrick et al., 2015; Schubert, Liebherr,
variables, Table 4 presents the corresponding odds ratios (OR). Kersten, & Haas, 2017; Silvano & Ohlin, 2019 ).

The variable of male resulted in a normally distributed parameter with a mean of 0.31 and a
variance of 1.27. Given these distributional parameters, 59.64% of male passengers were more likely
Several general observations are worthy of mention. First, the significant variables were not to be fatally or severely injured in a non-collision incident. This heterogeneous effect likely reflects the
entirely identical between the models of non-collision injuries and collision injuries. For example, variations in physical

Table 2
The goodness-of-fit measures for fixed and random parameters logistic models for severity of injuries to public bus passengers in non-collision incidents and collision incidents.

Non-collision injuries Collision injuries

FP model y RP model FP model y RP model

Number of observations, N 12,828 12,828 4,565 4,565


Number of parameters, K 31 36 33 37
Log-likelihood at zero, LL ð 0 Þ – 4310.50 – 4310.50 – 1005.88 – 1005.88
Log-likelihood at convergence, LL ð b Þ – 4032.99 – 4024.75 – 928.01 – 922.10
AIC * 8127.98 8121.50 1922.02 1918.20
Likelihood-ratio test

X 2 ¼ 2 ½ LL ð b FP Þ LL ð b Þ 16.48 11.82
RP

Degree of freedom 5 4
p- value 0.00 0.02

y Fixed parameters logistic model. Random

parameters logistic model.


*
Akaike Information Criterion, AIC ¼ 2 LL ð b Þ þ 2 K:

Please cite this article as: H. Zhou, C. Yuan, N. Dong et al., Severity of passenger injuries on public buses: A comparative analysis of collision injuries and non-collision injuries, Journal of Safety Research, https://doi.org/10.1
8 H. Zhou et al. / Journal of Safety Research xxx (xxxx) xxx

Table 3
The model results for severity of passenger injuries on public buses, stratified by non-collision injuries and collision injuries.

Non-collision injuries Collision injuries

FP model y RP model FP model y RP model

Mean SD p- value Mean SD p- value Mean SD p- value Mean SD p- value

Intercept – 4.51 * 0.34 0.00 – 4.19 * 0.29 0.00 – 5.16 * 0.69 0.00 – 4.70 * 0.69 0.00
SD (Intercept) 0.77 * 0.04 0.00
Passenger casualty characteristics
Casualty age (unit: year; reference group: 6 17) 18–34
0.67 * 0.27 0.01 0.71 * 0.23 0.00 1.34 * 0.53 0.01 1.32 * 0.46 0.00
35–49 1.03 * 0.25 0.00 1.04 * 0.21 0.00 1.48 * 0.52 0.00 1.47 * 0.46 0.00
50–64 1.43 * 0.24 0.00 1.43 * 0.20 0.00 1.53 * 0.52 0.00 1.47 * 0.46 0.00
P 65 2.09 * 0.24 0.00 2.10 * 0.20 0.00 1.99 * 0.52 0.00 1.18 * 0.48 0.01
SD ( P 65) 2.27 * 0.23 0.00
Male (Yes = 1, No = 0) – 0.06 * 0.07 0.36 – 0.31 * 0.07 0.00 – 0.32 * 0.15 0.03 – 0.37 * 0.14 0.01
SD (Male) 1.27 * 0.08 0.00
Head injury (Yes = 1, No = 0) – 0.23 * 0.06 0.00 – 0.63 * 0.06 0.00 0.24 * 0.13 0.07 – 1.49 * 0.22 0.00
SD (Head injury) 1.61 * 0.08 0.00 3.25 * 0.23 0.00
Seat occupied (reference group: seated)
Standing in deck 0.17 * 0.07 0.02 0.16 * 0.06 0.01 0.16 * 0.15 0.31 0.20 * 0.13 0.12
Unknown – 0.33 * 0.22 0.13 – 0.41 * 0.20 0.04 0.81 * 0.20 0.00 0.82 * 0.20 0.00
Contributory factor (reference group: none)
Losing balance when boarding or alighting Losing balance on 0.69 * 0.11 0.00 0.68 * 0.10 0.00 – – – – – –
stairway 0.90 * 0.10 0.00 0.87 * 0.09 0.00 0.70 * 0.39 0.07 0.83 * 0.33 0.01
Losing balance, not on stairway Others 0.41 * 0.09 0.00 0.42 * 0.07 0.00 0.56 * 0.28 0.05 0.56 * 0.25 0.02
0. 0.05 0. 0.02
0.22 * 0.11 0.22 * 0.09 – 0.25 * 0.21 0.24 – 0.20 * 0.18 0.28
Bus driver characteristics
Contributory factor (reference group: none)
Driving too fast 2.98 * 0.31 0.00 3.12 * 0.27 0.00 – – – – – –
Losing control of vehicle – – – – – – 1.56 * 0.38 0.00 1.48 * 0.39 0.00
Failing to ensure passenger safety Others 0.47 * 0.11 0.00 0.48 * 0.09 0.00 0.46 * 0.57 0.42 0.13 * 0.60 0.83
0.08 * 0.08 0.36 0.10 * 0.07 0.17 0.02 * 0.14 0.87 0.01 * 0.13 0.92
Bus characteristics
Vehicle type (reference group: single-decker bus)
Double-decker bus 0.38 * 0.22 0.08 0.42 * 0.18 0.02 0.79 * 0.44 0.08 0.95 * 0.42 0.02
Unknown 0.20 * 0.23 0.38 0.25 * 0.19 0.18 0.39 * 0.49 0.43 0.66 * 0.46 0.15
Vehicle collision with (reference group: pedestrian)
Vehicle – – – – – – 1.12 * 0.30 0.00 1.19 * 0.28 0.00
Object – – – – – – 0.41 * 0.18 0.02 0.42 * 0.18 0.02
Environmental characteristics
Area (reference group: New Territories)
Hong Kong Island – 0.19 * 0.08 0.02 – 0.19 * 0.07 0.00 – 0.52 * 0.19 0.00 – 0.60 * 0.16 0.00
Kowloon – 0.31 * 0.08 0.00 – 1.07 * 0.09 0.00 – 0.68 * 0.17 0.00 – 0.68 * 0.16 0.00
SD (Kowloon) 2.15 * 0.09 0.00
Season (reference group: spring)
Summer 0.11 * 0.08 0.19 0.09 * 0.07 0.22 0.06 * 0.20 0.76 0.04 * 0.18 0.81
Autumn 0.14 * 0.09 0.11 0.12 * 0.07 0.11 0.25 * 0.19 0.19 0.27 * 0.17 0.11
Winter 0.19 * 0.08 0.02 0.17 * 0.07 0.01 0.24 * 0.19 0.23 0.25 * 0.18 0.16
Within 70 m of a junction (Yes = 1, No = 0) – 0.11 * 0.07 0.10 – 0.11 * 0.06 0.05 – 0.19 * 0.15 0.21 – 1.10 * 0.21 0.00
SD (within 70 m of junction) 2.24 * 0.21 0.00
Road type (reference group: one-way street)
Two-way street 0.11 * 0.07 0.12 – 0.50 * 0.08 0.00 0.06 * 0.17 0.72 0.14 * 0.15 0.33
SD (Two-way street) 2.02 * 0.09 0.00
Dual carriageway 0.04 * 0.08 0.60 0.05 * 0.07 0.68 0.28 * 0.16 0.08 0.29 * 0.15 0.05
More than two carriageways Rain (reference –* – – –* – – – 0.46 * 0.30 0.12 – 0.30 * 0.24 0.21
group: not raining)
Light rain – 0.34 * 0.13 0.01 – 0.35 * 0.11 0.00 – 0.13 * 0.22 0.53 – 0.48 * 0.24 0.04
SD (Light rain) 1.30 * 0.27 0.00
Heavy rain 0.10 * 0.31 0.74 0.22 * 0.25 0.39 0.58 * 0.41 0.15 0.60 * 0.37 0.11
Street light (reference group: good)
Poor 0.58 * 0.64 0.36 0.60 * 0.57 0.29 – – – – – –
Not lit or none 0.26 * 0.08 0.00 0.27 * 0.07 0.00 – 0.26 * 0.18 0.14 – 0.30 * 0.16 0.06
Daylight 0.15 * 0.07 0.03 0.18 * 0.06 0.00 – 0.08 * 0.15 0.57 – 0.04 * 0.14 0.79

y Fixed parameters logistic regression model. Random parameters

logistic regression model.


*
Significant at 95% confidence level.

and physiological conditions between male and female passengers. Similarly, the variable of head judgement of injury locations may be somewhat arbitrary. The heterogeneous effect of head injury on
injury produced a normally distributed random parameter with a mean of 0.31 and a variance of 1.61, the severity of passenger injuries in non-collision incidents may also result partially from some
indicating that 65.22% of passengers with head injuries experienced a higher likelihood of fatal or unobserved factors (e.g., injury type and outcome of injury) and merit detailed investigations in future
severe injuries in noncollision incidents. Note that the location of injury to passengers was visually studies.
recorded by the police at the scenes of incidents. This
Given that standing passengers on public buses have greater difficulties in maintaining balance,
it is not surprising that they

Please cite this article as: H. Zhou, C. Yuan, N. Dong et al., Severity of passenger injuries on public buses: A comparative analysis of collision injuries and non-collision injuries, Journal of Safety Research, https://doi.org/10.1016
H. Zhou et al. / Journal of Safety Research xxx (xxxx) xxx 9

Fig. 3. The distribution of random parameters in the severity model of injuries to public bus passengers.

were 1.17 times more likely than seated passengers to be fatally or severely injured in a non-collision and Kowloon. This result is expected to some extent, as Hong Kong Island and Kowloon are relatively
incident. This result is in line with a previous study ( Siman-Tov et al., 2019 ) that standing passengers developed areas with more favorable road facilities, lower speed limits, and better emergency
were found to sustain a significantly higher risk of injury in non-collision incidents. Furthermore, our services ( Xu, Dong, et al., 2019 ). Meanwhile, it is worth mentioning that the variable of Kowloon
estimated results indicated that a loss of balance had a positive effect on the outcomes of resulted in a normally distributed random parameter with a mean of 1.07 and a variance of 2.15.
non-collision injuries to passengers. Specifically, passengers who lost their balance while boarding or Given these distributional parameters, 69.06% of non-collision incidents in Kowloon were associated
alighting experienced a 97% greater likelihood of fatal and severe injuries. Likewise, a loss of balance with an increased likelihood of fatal and severe injuries to public bus passengers. This heterogeneous
on stairways or decks increased the odds of fatal and severe injuries by 139% and 52%, respectively. effect resulting from large variations in urban settings within the Kowloon area is supported by a
These findings are intuitively reasonable, as the loss of balance usually leads to a fall immediately previous study of Yau (2004) , in which the severity of single-vehicle crashes was found to decrease in
among standing passengers ( Silvano & Ohlin, 2019 ), thereby increasing the severity of injuries. Kowloon East whereas it became to increase in Kowloon West.

Also not surprisingly, the faults of bus drivers showed a positive association with the severity of With regards to season, the positive sign of the estimated coefficient for winter suggests that the
passenger injuries in non-collision incidents. Comparatively, bus drivers who drove too fast likelihood of fatal and severe injuries to public bus passengers tended to be higher in winter than in
dramatically increased the odds of their passengers being fatally or severely injured by as much as spring. During wintertime with the two most important festivals (i.e., Christmas and Chinese New
2165%. A well-designed biomechanical experiment by Karekla and Tyler (2018) proved that Year), greater pressure due to tight timetables inevitably motivates bus drivers to drive as fast as
passengers were almost impossible to control their balance if a bus decelerated at a rate of 2.0 m/s 2 or possible to end a run ( Kendrick et al., 2015 ). As a consequence, the frequent acceleration and
more. Our finding thus seems fairly plausible, because driving too fast often results in a sudden and deceleration maneuvers undoubtedly increase the severity of non-collision injuries to passengers.
abrupt braking to avoid a collision.

Vehicle type was closely related to the severity of injuries to passengers in non-collision As illustrated in Fig. 3 , the variable of two-way streets resulted
incidents. Notably, the likelihood of fatal and severe injuries was found to increase by 52% if in a normally distributed random parameter with a mean of 0.31 and a variance of 1.61. Given these
passengers were injured on a double-decker bus. One plausible explanation is that double-deckers distributional parameters, nearly half of the passengers were more likely to be fatally or severely
buses are intentionally designed to have fewer seats and more standing space, resulting in an injured if non-collision incidents occurred on a two-way street. This heterogeneous result arises very
over-representation of passengers who have to stand during some or all of the journey. As another, likely from unobserved factors, such as the number of lanes, the presence of bus lanes, the presence
the large size and heavy mass of double-decker buses also reduce their maneuverability and stability, of road medians, the prevalence of pedestrian activities, and the presence of bus stops.
especially during the process of making a turn. These facts may help to explain the increased risk of
non-collision injuries to passengers on doubledecker buses.

Light rain was also identified as a significant variable. Accordingly, the odds of fatal and severe
injuries to public bus passengers decreased by about 30% if the non-collision incidents occurred on
days with light rain. This result is expected because bus drivers usually drive more cautiously at a
The location of the incident also had a significant influence on the outcomes of non-collision slower speed on rainy days ( Rahman et al., 2011 ). However, this safety benefit from the behavior
injuries to public bus passengers. According to our results, the severity of injuries to passengers in adjustments of bus drivers was totally compensated by reduced
non-collision incidents seemed to decrease in Hong Kong Island

Please cite this article as: H. Zhou, C. Yuan, N. Dong et al., Severity of passenger injuries on public buses: A comparative analysis of collision injuries and non-collision injuries, Journal of Safety Research, https://doi.org/10.1
10 H. Zhou et al. / Journal of Safety Research xxx (xxxx) xxx

Table 4
The odds ratios for severity models of passenger injuries on public buses, stratified by non-collision injuries and collision injuries.

Non-collision injuries Collision injuries

FP model y RP model FP model y RP model

OR 95% CI OR 95% CI OR 95% CI OR 95% CI

Passenger casualty characteristics


Casualty age (unit: year; reference group: 6 17) 18–34
1.95 * (1.15, 3.35) 2.03 * (1.30, 3.19) 3.82 * (1.36, 10.70) 3.74 * (1.51, 9.21)
35–49 2.80 * (1.72, 4.57) 2.83 * (1.86, 4.26) 4.39 * (1.57, 12.30) 4.35 * (1.79, 10.59)
50–64 4.18 * (2.61, 6.69) 4.18 * (2.80, 6.23) 4.62 * (1.65, 12.81) 4.35 * (1.77, 10.70)
P 65 8.08 * (5.05, 12.81) 8.17 * (5.47, 12.18) 7.32 * (2.64, 20.49) 3.25 * (1.27, 8.41)
Male (Yes = 1, No = 0) 0.94 * (0.83, 1.07) 0.73 * (0.65, 0.84) 0.73 * (0.55, 0.97) 0.69 * (0.53, 0.90)
Head injury (Yes = 1, No = 0) 0.79 * (0.70, 0.90) 0.53 * (0.47, 0.60) 1.27 * (0.98, 1.65) 0.23 * (0.15, 0.35)
Seat occupied (reference group: seated)
Standing in deck 1.19 * (1.03, 1.35) 1.17 * (1.04, 1.31) 1.17 * (0.86, 1.58) 1.22 * (0.94, 1.58)
Unknown 0.72 * (0.47, 1.11) 0.66 * (0.45, 0.97) 2.25 * (1.52, 3.32) 2.27 * (1.55, 3.35)
Contributory factor (reference group: none)
Losing balance when boarding or alighting Losing balance on 1.99 * (1.60, 2.48) 1.97 * (1.63, 2.39) 2.59 * (1.06, 6.23) – –
stairway 2.46 * (2.01, 3.00) 2.39 * (2.01, 2.86) 2.01 * (0.93, 4.39) 2.29 * (1.20, 4.39)
Losing balance, not on stairway Others 1.51 * (1.27, 1.79) 1.52 * (1.31, 1.75) 1.75 * (1.01, 3.03) 1.75 * (1.07, 2.83)
1.25 * (1.00, 1.54) 1.25 * (1.03, 1.49) 0.78 * (0.52, 1.19) 0.82 * (0.57, 1.17)
Bus driver characteristics
Contributory factor (reference group: none)
Driving too fast 19.69 * (10.70, 35.87) 22.65 * (13.46, 38.47) – – – –
Losing control of vehicle – – – – 4.76 * (2.27, 9.87) 4.39 * (2.03, 9.58)
Failing to ensure passenger safety Others 1.60 * (1.30, 1.99) 1.62 * (1.35, 1.93) 1.58 * (0.52, 4.85) 1.14 * (0.35, 3.71)
1.08 * (0.91, 1.27) 1.11 * (0.96, 1.27) 1.02 * (0.78, 1.35) 1.01 * (0.79, 1.30)
Bus characteristics
Vehicle type (reference group: single-decker bus)
Double-decker bus 1.46 * (0.95, 2.25) 1.52 * (1.06, 2.18) 2.20 * (0.92, 5.26) 2.59 * (1.13, 5.93)
Unknown 1.22 * (0.79, 1.92) 1.28 * (0.89, 1.88) 1.48 * (0.57, 3.86) 1.93 * (0.79, 4.71)
Vehicle collision with (reference group: pedestrian)
Vehicle – – – – 3.06 * (1.68, 5.53) 3.29 * (1.90, 5.75)
Object – – – – 1.51 * (1.05, 2.16) 1.52 * (1.07, 2.16)
Environmental characteristics
Area (reference group: New Territories)
Hong Kong Island 0.83 * (0.70, 0.97) 0.83 * (0.73, 0.94) 0.59 * (0.41, 0.85) 0.55 * (0.40, 0.75)
Kowloon 0.73 * (0.63, 0.85) 0.35 * (0.29, 0.41) 0.51 * (0.37, 0.70) 0.51 * (0.38, 0.69)
Season (reference group: spring)
Summer 1.12 * (0.95, 1.31) 1.09 * (0.95, 1.25) 1.06 * (0.72, 1.57) 1.04 * (0.73, 1.49)
Autumn 1.15 * (0.97, 1.35) 1.13 * (0.97, 1.30) 1.28 * (0.89, 1.86) 1.31 * (0.94, 1.80)
Winter 1.21 * (1.03, 1.42) 1.19 * (1.03, 1.36) 1.27 * (0.87, 1.86) 1.28 * (0.90, 1.82)
Within 70 m of a junction (Yes = 1, No = 0) Road type (reference 0.90 * (0.79, 1.02) 0.90 * (0.80, 1.00) 0.83 * (0.62, 1.12) 0.33 * (0.22, 0.50)
group: one-way street)
Two-way street 1.12 * (0.97, 1.28) 0.61 * (0.52, 0.70) 1.06 * (0.76, 1.48) 1.15 * (0.86, 1.54)
Dual carriageway 1.04 * (0.90, 1.22) 1.05 * (0.91, 1.20) 1.32 * (0.97, 1.82) 1.34 * (0.99, 1.79)
More than two carriageways Rain (reference –* – –* – 0.63 * (0.35, 1.13) 0.74 * (0.46, 1.19)
group: not raining)
Light rain 0.71 * (0.55, 0.92) 0.70 * (0.57, 0.87) 0.88 * (0.57, 1.34) 0.62 * (0.39, 0.99)
Heavy rain 1.11 * (0.61, 2.03) 1.25 * (0.76, 2.05) 1.79 * (0.81, 3.97) 1.82 * (0.87, 3.78)
Street light (reference group: good)
Poor 1.79 * (0.51, 6.23) 1.82 * (0.60, 5.53) – – – –
Not lit or none 1.30 * (1.11, 1.51) 1.31 * (1.15, 1.51) 0.77 * (0.54, 1.09) 0.74 * (0.54, 1.02)
Daylight 1.16 * (1.01, 1.34) 1.20 * (1.06, 1.34) 0.92 * (0.68, 1.23) 0.96 * (0.74, 1.26)

y Fixed parameters logistic regression model. Random parameters

logistic regression model.


*
Significant at 95% confidence level.

visibility, slippery road surfaces, and reduced friction during heavy rain, as evidenced by the ules during daytime, are very likely to trigger road rage among bus drivers ( Montoro, Useche, Alonso,
insignificant effect of heavy rain on the severity of passenger injuries in non-collision incidents. & Cendales, 2018 ). This angry driving behavior, combined with the overcrowding of passengers on
board during rush hours, may contribute to this greater risk of non-collision injuries to bus passengers.
Finally, the variable of streets without lighting at night was found to be positively associated with
the severity of noncollision injuries to public bus passengers. As presented in Table 4 , the likelihood
of passengers being fatally or severely injured in a non-collision incident increased by 31% on streets
without lighting at night. This elevated injury risk is probably attributed to the restricted visibility and
fatigue driving during nighttime without lighting ( Anarkooli & Hosseinlou, 2016 ). In contrast, bus 3.2.2. Severity model results for collision injuries
passengers traveling in daytime were also found to sustain a higher likelihood of fatal and severe As shown in Tables 3 and 4 , a total number of 11 factors was
injuries in non-collision incidents. Indeed, the frequent interactions with other road users, together with found to be significantly associated with the severity of passenger injuries in collision incidents. Most
the high level of traffic congestion and the tight time sched- of the results were similar to those derived from the models of non-collision injuries, except for three
variables (i.e., loss of vehicle control, bus collided with vehicles or objects, and within 70 m of a
junction) that were significant solely in the models of collision injuries.

Please cite this article as: H. Zhou, C. Yuan, N. Dong et al., Severity of passenger injuries on public buses: A comparative analysis of collision injuries and non-collision injuries, Journal of Safety Research, https://doi.org/10.1016
H. Zhou et al. / Journal of Safety Research xxx (xxxx) xxx 11

Unlike the models of non-collision injuries in which the variable of driving too fast had a when boarding, alighting from, or standing on a bus. In addition, the variables of male, head injuries,
significant relationship, it is the variable of loss of vehicle control that played an indispensable role in Kowloon, and two-way streets resulted in heterogeneous effects on the severity of non-collision
the severity of collision injuries to public bus passengers. Specifically, passengers were 4.39 times injuries to passengers. By comparison, the following factors were found to be significantly associated
more likely to be fatally or severely injured if the collisions were a result of the bus drivers’ loss of with a higher likelihood of fatal or severe injuries to passengers as a result of collision incidents:
control of their vehicles, such as skidding, brake-failure, and loss of steering. Therefore, bus drivers elderly female passengers, standing passengers who lost balance, buses out of driver control,
should be retrained to better handle this emergency situation to minimize damage and ensure double-decker buses, collisions with vehicles or objects, and less urbanized areas. The variables of
passenger safety. elderly passengers, head injuries, within 70 m of a junction, and light rain had mixed effects on the
severity of collision injuries to passengers.

Compared with collisions with pedestrians, public buses colliding with vehicles and objects
resulted in a higher likelihood of fatal and severe injuries to passengers. As the most vulnerable road
users, pedestrians are typically victims of collisions with negligible threats to their collision partners ( Xu Based on these findings, traditional ‘‘4E” strategies (i.e., engineering, enforcement, emergency,
et al., 2016 ). As a result, buses colliding with vehicles and objects would generally generate more and education) are suggested to mitigate collision injuries to public bus passengers. Such
kinetic energy, leading to more serious outcomes for their passengers. countermeasures might include the provision of exclusive bus lanes to separate buses from mixed
traffic, improving the maneuverability and stability of double-decker buses, enhancing enforcement of
no-standing rules on the stairways or upper floors of doubledecker buses, improving emergency care,
and providing systematic training to bus drivers to improve their driving performance, particularly in
Since conflicts usually concentrate at intersections, bus crashes at intersections were found to be cases of losing control of vehicles and driving on heavily rainy days.
more serious than those at other locations ( Rahman et al., 2011; Kaplan & Prato, 2012 ). However,

Sam et al. (2018) reported an inconsistent result that intersections significantly reduced the severity of
bus crashes, because bus drivers tended to be more cautious when approaching intersections. This
contradictory finding from previous studies is well reflected in our study that the variable of within 70 Unlike injuries due to accidental impacts with vehicles, objects, or pedestrians, passenger injuries
m of a junction resulted in a normally distributed parameter with a mean of 1.10 and a variance of caused by non-collision incidents could be largely preventable. In addition to upgrading bus facilities
2.24. Given these distributional parameters, 68.83% of bus crashes near junctions were associated to better accommodate the needs of passengers, such as lowering bus steps to be more accessible
with a greater likelihood of fatal and severe injuries to passengers. to the elderly ( Björnstig et al., 2005 ), replacing horizontal metal seat handles with vertical ones ( Palacio
et al., 2009 ), and the use of textured floors to prevent slips ( Halpern et al., 2005 ), we introduce a ‘‘3A”
approach (i.e., awareness, appreciation, and assistance) here to aid in the formulation of safety
programs for both bus drivers and passengers to reduce non-collision injuries. First, it is important to
raise awareness among bus drivers and passengers of their inherent limitations under certain
situations. For example, the awareness of increasing fragility of the body, longer reaction time, slower
4. Conclusions gait, and weakened ability to maintain balance due to aging would encourage bus drivers to be more
considerate and responsible to their elderly passengers by avoiding abrupt starts, sudden
This study sought to comparatively examine the severity of passenger injuries on public buses acceleration, and unexpected deceleration. Likewise, keep in mind that the decks may be wet and
due to non-collision incidents and collision incidents, using a dataset of 17,383 passengers injured on slippery on rainy days, passengers would be more cautious while boarding and alighting from a bus.
public buses over a 10-year period in Hong Kong. Detailed characteristics of the passengers were Second, bus drivers and passengers should be educated to foster a greater appreciation of situations
integrated with demographic information and faults of bus drivers, characteristics of buses, and that significantly increase the severity of non-collision incidents. According to our results, passengers
environmental factors to construct our dataset. A random parameters logistic model was developed to were more likely to be fatally or severely injured if the non-collision incidents occurred in rural areas
explicitly account for the heterogeneous effects of risk factors. with less favorable road conditions, during winter holidays with greater pressure for bus drivers to
keep to timetables, on heavily rainy days with restricted visibility and slippery decks, at daytime with a
higher level of traffic congestion and overcrowding of passengers on board, and at night without street
lighting. It would therefore be beneficial if such a list of scenarios is generated to alert both bus drivers
Several key findings are worthy of note. First, the omission of unobserved heterogeneity in the and passengers when traveling in these particularly highrisky situations. Last, it would be useful to
fixed parameters model would result in reduced goodness-of-fit and biased estimations. Our empirical introduce appropriate
results also indicated substantial inconsistences in the effects of several risk factors between the
models of non-collision injuries and collision injuries. For example, the variables of standing on the
deck, winter, two-way street, and street without lighting at night were insignificant in the collision
injuries models, but became highly significant in the non-collision injuries models. Similarly, the OR of
double-decker buses increased by as much as 70% between the random parameters models of
non-collision injuries and collision injuries.

Although a large number of injuries sustained by public bus passengers resulted from
non-collision incidents, little is known about the role played by risk factors in the outcomes of driver assistance systems to reduce the burden of driving tasks without adding distractions to bus
passengers injured in these non-collision incidents. Our study provides additional evidence to existing drivers. For example, a passive collision avoidance system with audio warnings of close following or
research that the severity of passenger injuries tended to increase significantly when non-collision over-speeding may be effective in helping bus drivers to keep a safe distance from the preceding
incidents occurred due to bus drivers’ fault of driving too fast, on double-decker buses, in less vehicle. A voice prompt system could also be helpful to give a timely reminder to standing passengers
urbanized areas, in winter, on days with heavy rain, in daytime, and at night without street lighting. to grasp the handrails before the bus makes a turn.
Elderly female passengers were also found more likely to be fatally or severely injured in non-collision
incidents if they lost balance
Some limitations of this study should be well acknowledged. Our injury data were obtained from
police reports and may thus be subject to some degree of under-reporting. However, injuries

Please cite this article as: H. Zhou, C. Yuan, N. Dong et al., Severity of passenger injuries on public buses: A comparative analysis of collision injuries and non-collision injuries, Journal of Safety Research, https://doi.org/10.1
12 H. Zhou et al. / Journal of Safety Research xxx (xxxx) xxx

to public bus passengers are more likely to be recorded by the police than injuries to other road users et al., 2017; Karekla & Tyler, 2018, 2019 ), is highly advocated to provide deeper insights into the
( Loo & Tsui, 2007 ). As the only representative data source publicly available for such a long time causation of injuries sustained by public bus passengers.
span, these police records are also widely used by local authorities for safety policymaking ( Xu et al.,
2016; Xu, Dong, et al., 2019, Xu, Xie, et al., 2019; Meng et al., 2017; Xie et al., 2018; Sze, Su, & Bai,
2019; Wang et al., 2019; Zhai et al., 2019 ). Moreover, our developed random parameters logistic Acknowledgements
models assumed the parameters as normal distributions with fixed values of means and variances.
More advanced models, such as the random parameters logit models with heterogeneity in means We would like to thank the Hong Kong Police Force and the Hong Kong Transport Department for
and variances ( Seraneeprakarn et al., 2017; Waseem et al., 2019 ), could be used to track the source providing access to the database used in this study. The views expressed are the authors’ own and
of unobserved heterogeneity in future studies. Although we included a host of variables during model do not necessarily represent the views of the Hong Kong Police Force and the Hong Kong Transport
specification, several potential risk factors, such as the positions of standing passengers and the Department.
actions of passengers at the time of injury, are unavailable from police reports and are thus not
considered in our study. Further effort towards the integration of police reports with other data Funding: This work was supported by grants from the Joint Research Scheme of National Nature
sources, such as the records from hospital admissions and bus operators, questionnaire surveys, Science Foundation of China/Research Grants Council of Hong Kong (Project No. 71561167001 and
naturalistic driving studies, accident reconstruction simulations ( Palacio et al., 2009 ), and N_HKU707/15) and the Natural Science Foundation of China (Project No. 71671100). The funders
biomechanical experiments ( Schubert had no role in study design, data collection and analysis, decision to publish, or preparation of the
manuscript.

Competing interests

We declare that no competing interests exist.

Appendix

Table A1
Summary of studies investigating bus injuries within the past two decades.

Authors Study region Study period Observations Data source Research Research purpose
method

Bus injury characteristics analysis


af Wåhlberg (2002) Uppsala, Sweden 1986–2000 2,237 bus crashes Police records Descriptive Description of characteristics of lowspeed crashes
af Wåhlberg (2004) statistics involving buses Comparison of characteristics of
Björnstig et al. (2005) Umeå, Sweden 1994–2003 284 passengers injured Hospital injuries sustained by bus and coach passengers due
in buses and coaches admission to collisions and noncollisions
records

Halpern et al. (2005) Israel March to 120 non-collision Description of characteristics of noncollision injuries
October, 2000 injured passengers in sustained by public bus passengers
public buses
McGeehan et al. (2006) US 2001–2003 51,100 children and To investigate the characteristics of children and
teenagers injured in teenagers injured in school buses
school buses
Yang et al. (2009) Iowa, US 2002–2005 749 school bus crashes Police records To estimate crash, fatality, and injury rates of school
buses
To investigate the characteristics of crashes
involving school buses
Blower and Green US 1999–2005 2,252 fatal crashes Police records Logistic model To identify factors that contribute to bus drivers being
(2010) involving a bus at fault in fatal crashes, stratified by bus types

Zunjic et al. (2012) Belgrade, Serbia 2008–2010 287 non-collision Hospital Descriptive To investigate the characteristics of non-collision
injured passengers in admission statistics injuries sustained by bus passengers
city buses records
Barnes et al. (2016) England 2008–2012 10,010 passengers aged Police records Descriptive Description of characteristics of injuries sustained by
60 or above injured in buses statistics passengers aged 60 or above in public buses and
and coaches coaches
1999–2009 1,016 passengers aged Linked police
60 or above injured in buses and hospital
and coaches records
Silvano and Ohlin Sweden January 2015 316 standing bus Hospital Descriptive Description of characteristics of noncollision injuries
(2019) to August passengers injured due admission statistics sustained by standing passengers due to
2018 to acceleration and records
braking maneuvers acceleration and braking maneuvers in buses

Siman-Tov et al. (2019) Iran 2015–2017 704 hospitalized non- Hospital Descriptive Description of characteristics of noncollision injuries
collision bus casualties admission statistics sustained by public bus passengers
records

Bus crash frequencies analysis


Rey et al. (2002) Hillsborough, 1996–1999 524 transit bus crashes Police records Descriptive To analyze the effectiveness of training programs
Florida, US statistics and safety investment in reducing transit bus
Orlando, Florida, 751 transit bus crashes crash frequencies
US

Please cite this article as: H. Zhou, C. Yuan, N. Dong et al., Severity of passenger injuries on public buses: A comparative analysis of collision injuries and non-collision injuries, Journal of Safety Research, https://doi.org/10.1016
H. Zhou et al. / Journal of Safety Research xxx (xxxx) xxx 13

Table A1 ( continued)

Authors Study region Study period Observations Data source Research Research purpose
method

Chang and Yeh (2005) Taiwan, China 2001 42 bus companies Questionnaire Poisson model To identify factors affecting the safety performance of
survey bus companies
Salminen et al. (2009) Finland May 2005 to 176 Finnish and 130 immigrant Questionnaire Descriptive To examine whether immigrant bus drivers were
June 2006 bus drivers survey statistics more likely to be involved in injury crashes than local
ones

Shahla et al. (2009) Toronto, Canada 1995–2003 865 transit bus crashes at 1,665 Police records Negative To develop a crash prediction model for transit public
signalized binomial model buses at signalized intersections
intersections
Chimba et al. (2010) Florida, US 2003–2007 4,528 bus crashes at Police records Negative To identify factors affecting the frequencies of
1,285 road segments binomial model bus crashes at road segments

Strathman et al. (2010) Portland, US 2006–2009 4,628 incidents during Transit provider Poisson model To identify factors contributing to collision and
12 signup periods for Negative non-collision
1,502 transit operators binomial model involvement of transit operators
Tseng (2012) Taiwan, China 2010 2,023 tour bus drivers Questionnaire Logistic model To identify factors contributing to atfault crash
survey involvement of tour bus drivers

Goh et al. (2014) Melbourne, 2009–2011 1,099 transit bus Transit Mixed-effects To investigate the effect of bus priority strategy in
Australia crashes at 99 road providers negative reducing transit bus crash frequencies
segments binomial model
Neural network
model
Gómez and Bocarejo Bogotá, 2010 Bus crashes at 104 road Police records Poisson model To identify factors affecting the frequency of bus
(2015) Colombia segments in bus rapid transit Negative crashes in bus rapid transit system
corridors binomial model
Neural network
model

Bus injury severity analysis


Barua and Tay (2010) Dhaka, 1998–2005 2,662 transit bus Police records Ordered probit To identify factors affecting bus crash severities
Bangladesh crashes model
Chimba et al. (2010) Florida, US 2003–2007 4,528 bus crashes at Multinomial
1,285 road segments logit model
Rahman et al. (2011) Kaplan Alberta, Canada 2000–2007 9,485 bus crashes Logistic model
and Prato US 2005–2009 2,576 bus crashes Generalized
(2012) ordered logit
Prato and Kaplan Denmark 2002–2011 3,434 bus crashes model
(2014)
Feng et al. (2016) US 2006–2010 1,380 fatal crashes
involving a bus
Yoon et al. (2017) South Korea 2010–2014 27,731 local bus Hierarchical
crashes ordered probit
model
Sam et al. (2018) Ghana 2011–2015 33,694 crashes Generalized
involving public buses ordered logit
and minibuses model

References segment and intersection crash modeling. Analytic Methods in Accident Research,
19, 1–15 .
Chang, F., Li, M., Xu, P., Zhou, H., Haque, M. M., & Huang, H. (2016). Injury severity of
af Wåhlberg, A. E. (2002). Characteristics of low speed accidents with buses in
motorcycle riders involved in traffic crashes in Hunan, China: A mixed ordered logit approach. International
public transport. Accident Analysis and Prevention, 34, 637–647 .
Journal of Environmental Research and Public Health,
af Wåhlberg, A. E. (2004). Characteristics of low speed accidents with buses in
13, 714 .
public transport: Part II. Accident Analysis and Prevention, 36( 1), 63–71 .
Chang, F., Xu, P., Zhou, H., Chan, A. H. S., & Huang, H. (2019). Investigating injury
Anastasopoulos, P. C., & Mannering, F. L. (2011). An empirical assessment of fixed
severity of motorcycle riders: A two-stepmethod integrating latent class cluster analysis and random parameters
and random parameter logit models using crash- and non-crash-specific injury data. Accident Analysis and
logit model. Accident Analysis and Prevention,
Prevention, 43( 3), 1140–1147 .
131, 316–326 .
Anarkooli, A. J., & Hosseinlou, M. H. (2016). Analysis of the injury severity of crashes
Chang, H., & Yeh, C. (2005). Factors affecting the safety performance of bus
by considering different lighting conditions on two-lane rural roads. Journal of Safety Research, 56, 57–65 .
companies—the experience of Taiwan bus deregulation. Safety Science, 43,
323–344 .
Barnes, J., Morris, A., Welsh, R., Summerskill, S., Marshall, R., Kendrick, D., ... Bell, J.
Chen, F., Song, M., & Ma, X. (2019). Investigation on the injury severity of drivers in
(2016). Injuries to older users of buses in the UK. Journal of Public Transportation,
rear-end collisions between cars using a random parameters bivariate ordered probit model. International Journal
8, 25–38 .
of Environmental Research and Public Health, 16( 14), 2632 .
Barua, U., & Tay, R. (2010). Severity of urban transit bus crashes in Bangladesh.
Journal of Advanced Transportation, 44, 34–41 .
Chimba, D., Sando, T., & Kwigizile, V. (2010). Effect of bus size and operation to crash
Beck, L. F., Dellinger, A. M., & O’Neil, M. E. (2007). Motor vehicle crash injury rates by
occurrences. Accident Analysis and Prevention, 42, 2063–2067 .
mode of travel, United States: Using exposure-based methods to quantify differences. American Journal of
Dong, B., Ma, X., Chen, F., & Chen, S. (2018). Investigating the differences of single-
Epidemiology, 166( 2), 212–218 .
vehicle and multivehicle accident probability using mixed logit model. Journal of Advanced Transportation, 50, 2015–2028
Björnstig, U., Albertsson, P., Björnstig, J., Bylund, P., Falkmer, T., & Petzäll, J. (2005).
.
Injury evens among bus and coach occupants: Non-crash injuries as important as crash injuries. IATSS
Elvik, R. (2019). Risk of non-collision injuries to public transport passengers:
Research, 29( 1), 79–87 .
Synthesis of evidence from eleven studies. Journal of Transport & Health, 13,
Blower, D., & Green, P. E. (2010). Type of motor carrier and driver history in fatal bus
128–136 .
crashes. Transport Research Review, 2194, 37–43 .
Feng, S., Li, Z., Ci, Y., & Zhang, G. (2016). Risk factors affecting fatal bus accident
Cai, Q., Abdel-Aty, M., Lee, J., Wang, L., & Wang, X. (2018). Developing a grouped
severity: Their impact on different types of bus drivers. Accident Analysis and Prevention, 86, 29–39 .
random parameters multivariate spatial model to explore zonal effects for

Please cite this article as: H. Zhou, C. Yuan, N. Dong et al., Severity of passenger injuries on public buses: A comparative analysis of collision injuries and non-collision injuries, Journal of Safety Research, https://doi.org/10.1
14 H. Zhou et al. / Journal of Safety Research xxx (xxxx) xxx

Goh, K. C. K., Currie, G., Sarvi, M., & Logan, D. (2014). Bus accident analysis of routes Siman-Tov, M., Radomislensky, I., Marom, I., Kapra, O., & Peleg, K. (2019). A nation-
with/without bus priority. Accident Analysis and Prevention, 65, 18–27 . wide study on the prevalence of non-collision injuries occurring during use of public buses. Journal of Transport &
Gómez, F., & Bocarejo, J. P. (2015). Accident prediction models for bus rapid transit Health, 13, 164–169 .
systems: Generalized linear models compared with a neural network. Strathman, J. G., Wachana, P., & Callas, S. (2010). Analysis of bus collision and non-
Transportation Research Record, 2512, 38–45 . collision incidents using transit ITS and other archived operations data. Journal of Safety Research, 41, 137–144 .
Guo, Y., Wu, Y., Lu, J., & Zhou, J. (2019). Modeling the unobserved heterogeneity in E-
bike collision severity using full Bayesian random parameters multinomial logit regression. Sustainability, 11, 2071 Sze, N. N., Su, J., & Bai, L. (2019). Exposure to pedestrian crash based on household
. survey data: Effect of trip purpose. Accident Analysis and Prevention, 128, 17–24 .
Halpern, P., Siebzehner, M. I., Aladgem, D., Sorkine, P., & Bechar, R. (2005). Non- Train, K. E. (2009). Discrete choice: Methods with simulation. Cambridge, UK:
collision injuries in public buses: A national survey of a neglected problem. Cambridge University Press .
Emergency Medicine Journal, 22, 108–110 . Tseng, C. (2012). Social-demographics, driving experience and yearly driving
Hong Kong Transport Department. 2014. Travel characteristics survey 2011 final distance in relation to a tour bus driver’s at-fault accident risk. Tourism Management, 33( 4), 910–915 .
report. https://www.td.gov.hk/filemanager/en/content_4652/tcs2011_eng.pdf. Hong Kong Transport Department.
2018. Road traffic accident statistics. 2017. Wang, J., Huang, H., Xu, P., Xie, S., & Wong, S. C. (2019). Random parameter probit
Available from: https://www.police.gov.hk/info/doc/ models to analyze pedestrian red-light violations and injury severity in pedestrian-motor vehicle crashes at
statistics/traffic_report_2017_en.pdf. signalized crossings. Journal of
Huang, H., Chang, F., Zhou, H., & Lee, J. (2019). Modeling unobserved heterogeneity Transportation Safety & Security. https://doi.org/10.1080/
for zonal crash frequencies: A Bayesian multivariate random-parameters model with mixture components for 19439962.2018.1551257 .
spatially correlated data. Analytic Methods in Accident Research, 24 100105 . Waseem, M., Ahmed, A., & Saeed, T. U. (2019). Factors affecting motorcyclists’ injury
severities: An empirical assessment using random parameters logit model with heterogeneity in means and
Kaplan, S., & Prato, C. G. (2012). Risk factors associated with bus accident severity in variances. Accident Analysis and Prevention, 123,
the United States: A generalized ordered logit model. Journal of Safety Research, 12–19 .
43, 171–180 . Wu, Q., Chen, F., Zhang, G., Liu, X. C., Wang, H., & Bogus, S. M. (2014). Mixed logit
Karekla, X., & Tyler, N. (2018). Reducing non-collision injuries aboard buses: model-based driver injury severity investigations in single- and multi-vehicle crashes on rural two-lane highways. Accident
Passenger balance whilst walking on the lower deck. Safety Science, 105, Analysis and Prevention, 72,
128–133 . 105–115 .
Karekla, X., & Tyler, N. (2019). Reducing non-collision injuries aboard buses: Wu, Q., Zhang, G., Zhu, X., Liu, X. C., & Tarefder, R. (2016). Analysis of driver injury
Passenger balance whilst climbing the stairs. Safety Science, 112, 152–161 . severity in single-vehicle crashes on rural and urban roadways. Accident Analysis and Prevention, 94, 35–45 .
Kendrick, D., Drummond, A., Logan, P., Barnes, J., & Worthington, E. (2015).
Systematic review of the epidemiology of non-collision injuries occurring to older people during use of public Xie, S. Q., Dong, N., Wong, S. C., Huang, H., & Xu, P. (2018). Bayesian approach to
buses in high income countries. Journal of Transport & Health, 2, 394–405 . model pedestrian crashes at signalized intersections with measurement errors in exposure. Accident Analysis
and Prevention, 121, 285–294 .
Li, Z., Wu, Q., Ci, Y., Chen, C., Chen, X., & Zhang, G. (2019). Using latent class analysis and mixed logit model to Xu, P., & Huang, H. (2015). Modeling crash spatial heterogeneity: Random
explore risk factors on driver injury severity in singlevehicle crashes. Accident Analysis and Prevention, 129, 230–240 parameter versus geographically weighting. Accident Analysis and Prevention,
. 75, 16–25 .
Loo, B. P. Y., & Tsui, K. L. (2007). Factors affecting the likelihood of reporting road Xu, P., Huang, H., Dong, N., & Wong, S. C. (2017). Revisiting crash spatial
crashes resulting in medical treatment to the police. Injury Prevention, 13( 3), 186–189 . heterogeneity: A Bayesian spatially varying coefficients approach. Accident Analysis and Prevention, 98, 330–337
.
Mannering, F. L., & Bhat, C. R. (2014). Analytic methods in accident research: Xu, P., Dong, N., Wong, S. C., & Huang, H. (2019). Cyclists injured in traffic crashes in
Methodological frontier and future directions. Analytic Methods in Accident Research, 1, 1–22 . Hong Kong: A call for action. PLoS ONE, 14( 8) e0220785 .
Xu, P., Xie, S., Dong, N., Wong, S. C., & Huang, H. (2019). Rethinking safety in
Mannering, F. L., Shankar, V., & Bhat, C. R. (2016). Unobserved heterogeneity and the numbers: Are intersections with more crossing pedestrians really safer? Injury Prevention, 25, 20–25 .
statistical analysis of highway accident data. Analytic Methods in Accident Research, 11, 1–16 .
Xu, X., Xie, S., Wong, S. C., Xu, P., Huang, H., & Pei, X. (2016). Severity of pedestrian
McGeehan, J., Annest, J. L., Vajani, M., Bull, M. J., Agran, P. E., & Smith, G. A. (2006). injuries due to traffic crashes at signalized intersections in Hong Kong: A Bayesian spatial logit model. Journal of
School bus-related injuries among children and teenagers in the United States, 2001–2003. Pediatrics, 118, 1984–1987 Advanced Transportation, 50, 2015–2028 .
. Yang, J., Peek-Asa, C., Cheng, G., Heiden, E., Falb, S., & Ramirez, M. (2009). Incidence
Meng, F., Xu, P., Wong, S. C., Huang, H., & Li, Y. C. (2017). Occupant-level injury and characteristics of school bus crashes and injuries. Accident Analysis and Prevention, 41, 336–341 .
severity analyses for taxis in Hong Kong: A Bayesian space-time logistic model.
Accident Analysis and Prevention, 108, 297–307 . Yau, K. K. W. (2004). Risk factors affecting the severity of single vehicle traffic
Milton, J., Shankar, V., & Mannering, F. L. (2008). Highway accident severities and accidents in Hong Kong. Accident Analysis and Prevention, 36( 3), 333–340 .
the mixed logit model: An explanatory empirical analysis. Accident Analysis and Prevention, 40( 1), 260–266 . Yoon, S., Kho, S., & Kim, D. (2017). Effect of regional characteristics on injury
severity in local bus crashes: Use of hierarchical ordered model. Transportation Research Record, 2647, 1–8 .
Montoro, L., Useche, S., Alonso, F., & Cendales, B. (2018). Work environment, stress,
and driving anger: A structural equation model for predicting traffic sanctions of public transport drivers. International Zhai, X., Huang, H., Sze, N. N., Song, Z., & Hon, K. K. (2019). Diagnostic analysis of the
Journal of Environmental Research and Public Health, 15( 3), 497 . effects of weather condition on pedestrian crash severity. Accident Analysis and Prevention, 122, 318–324 .

Rahman, M. M., Kattan, L., & Tay, R. (2011). Injury risk in collisions involving buses Zhang, K., & Hassan, M. (2019). Crash severity analysis of nighttime and daytime
in Alberta, Canada. Transportation Research Record, 2265, 13–26 . highway work zone crashes. PLoS ONE, 14( 8) e0221128 .
Rey, J. R., Hinebaugh, D., & Fernandez, J. (2002). Analysis of Florida transit bus Zunjic, A., Sremcevic, V., Sijacki, V. Z., & Sijacki, A. (2012). Research of injuries of
crashes. Transportation Research Record, 1791, 26–34 . passengers in city buses as a consequence of non-collision. Work, 41,
Palacio, A., Tamburro, G., O’Neill, D., & Simms, C. K. (2009). Non-collision injuries in 4943–4950 .
urban buses-strategies for prevention. Accident Analysis and Prevention, 41, 1–9 .
Prato, C. G., & Kaplan, S. (2014). Bus accident severity and passenger injury:
Hanchu Zhou received the Bachelor’s degree in Traffic Engineering and Software Engineering from Dalian Jiaotong
Evidence from Denmark. European Transport Research Review, 6, 17–30 .
University, Liaoning, China, in 2014. And he received the master’s degree in Communication and Transportation
Salminen, S., Vartia, M., & Giorgiani, T. (2009). Occupational injuries of immigrant
Engineering from Central South University, Hunan, China, in 2017. He is currently working toward a joint PhD degree
and Finnish bus drivers. Journal of Safety Research, 40, 203–205 .
with Central South University, Hunan, China and City University of Hong Kong, Hong Kong, China. His research area
Sam, E. F., Daniels, S., Brijs, K., Brijs, T., & Wets, G. (2018). Modelling public bus/
includes road safety, intelligent transportation systems, and statistical analysis.
minibus transport accident severity in Ghana. Accident Analysis and Prevention,
119, 114–121 .
Schubert, P., Liebherr, M., Kersten, S., & Haas, C. T. (2017). Biomechanical demand
analysis of older passengers in a standing position during bus transport. Journal of Transport & Health, 4, 226–236
Chen Yuan is a PhD student in the School of Traffic and Transportation Engineering, Central South University. He
.
received his MSc degree in Transportation Engineering from the Department of Civil Engineering, The University of
Seraneeprakarn, P., Huang, S., Shankar, V., Mannering, F., Venkataraman, N., &
Hong Kong and a BSc degree in Traffic Engineering from the School of Highway, Chang’an University. His research
Milton, J. (2017). Occupant injury severities in hybrid-vehicle involved crashes: A random parameters approach
interests mainly include road safety, accident analysis, data analytics and machine learning.
with heterogeneity in means and variances.
Analytic Methods in Accident Research, 15, 41–55 .
Shahla, F., Shalaby, A. S., Persaud, B. N., & Hadayeghi, A. (2009). Analysis of transit
safety at signalized intersections in Toronto, Ontario, Canada. Transportation Research Record, 2102, 108–114 .
Ni Dong received the BSc degree in transportation engineering from Changsha University of Science and Technology,
China, in 2006, and the MSc and PhD degrees in transportation engineering from Central South University, China, in
Silvano, A. P., & Ohlin, M. (2019). Non-collision incidents on buses due to
2009 and in
acceleration and braking manoeuvers leading to falling events among standing passengers. Journal of Transport
2015, respectively. She is a visiting scholar at University of Washington, Seattle during 2018–2020. She is currently an
& Health, 14 100560 .
assistant professor at the Department of

Please cite this article as: H. Zhou, C. Yuan, N. Dong et al., Severity of passenger injuries on public buses: A comparative analysis of collision injuries and non-collision injuries, Journal of Safety Research, https://doi.org/10.1016
H. Zhou et al. / Journal of Safety Research xxx (xxxx) xxx 15

Transportation and Logistics, Southwest Jiaotong University. Her current research interest is the statistical modeling of Pengpeng Xu is now a PhD student in the Department of Civil Engineering, The University of Hong Kong. He received
human mobility, human factor, traffic safety and public health. his BSc degree from Wuhan University of Technology in 2012, and the MSc degree from Central South University in
2015. His research interests mainly include transportation and health, transportation safety planning, health
geography, and applied Bayesian statistical models. Since
S.C. Wong is Chair Professor of Transportation Engineering, and holds Francis S Y Bong Professorship in
Engineering, at the Department of Civil Engineering of the University of Hong Kong. He received his BSc and MPhil 2014, he has completed over 80 reviewer assignments for more than 15 international journals, and was rewarded as
degrees from the University of Hong Kong and a PhD in Transport Studies from University College London. His the Publons Top Peer Reviewer 2019 by Web of Science Group and the Outstanding Reviewer of Accident Analysis &
research interests cover a wide range of traffic and transportation subjects, including traffic signal control, Prevention by Elsevier.
continuummodeling of traffic problems, traffic flow theory, transportation network modeling, land-use and transportation
problem, urban taxi services, and road safety.

Please cite this article as: H. Zhou, C. Yuan, N. Dong et al., Severity of passenger injuries on public buses: A comparative analysis of collision injuries and non-collision injuries, Journal of Safety Research, https://doi.org/10.1

También podría gustarte