Pedestrian injury risk: unanswered questions and a developing research agenda

Literature on pedestrian injuries is extensive and predates current interest in walking as a transport mode. Broadly speaking, four different categories of risk factor are discussed. Firstly, characteristics associated with the victim, including behaviour (such as “distracted walking”) and demographic factors (such as gender and ethnicity). Secondly, socio-cultural and household factors, such as family characteristics and household or local area deprivation. Thirdly, characteristics of the physical environment, such as speed limits or presence of pedestrian crossings. Finally, characteristics of the driver or rider colliding with the pedestrian (demographic, behavioural, or vehicle type).

Where different types of risk factors are associated with elevated injury risk, policy implications may differ. Planners can directly modify the physical environment: crossings can be added or removed, speed limits increased or reduced, footways widened or narrowed. Thus, if we find lower speed limits reduce pedestrian injury risk, this gives us a direct policy lever. Other factors are not directly modifiable in this way: so, for instance, if lower speeds (driver behaviour) are associated with reduced injury risk, this does not in itself tell us what we need to do. We also need information about the impacts of causal pathways connecting interventions (education, engineering, enforcement, etc.) to changes in driving speeds.

Causal pathways are likely to be linked, with feedback potentially acting in unpredictable ways. An intriguing aspect of injury risk is that some individually protective factors may not feed through into similarly positive results at societal level, implying potentially negative unintended consequences of apparently benign interventions. The most obvious example is the cycle helmet debate. A bike helmet may sometimes reduce injury severity for an individual experiencing a crash or collision, yet among rich countries, those safest for cycling have the lowest levels of helmet use. In safer countries, policy-makers focus on making cycling feel easy and safe, rather than on persuading or coercing cyclists to wear personal protective equipment – equipment which may even encourage drivers to take less care (Walker, 2007).

Why do such unintended consequences occur? Policy, law, and infrastructure are cultural. They all continuously send messages about who matters on our roads, and who is to blame when things go wrong. If pedestrian mobile phone use is associated with injury, surely a campaign against distracted walking is a good thing, or at least does no harm? However, a distracted walking campaign may implicitly place responsibility and blame on injured pedestrians, and so subtly encourage other road users to take less care around pedestrians. This would tap into the historic and continuing privileging of automobility in planning and road safety, which helps to remove responsibility from drivers (Freund & Martin, 1997).

There are other reasons why we struggle to untangle causality in injury risk research. Traditionally, research has focused on factors predicting frequency or presence of pedestrian injuries (or in some cases predicting injury severity, given injury). However, this only tells us part of the story. For instance, are children from lower income and/or ethnic minority communities disproportionately represented among injured child pedestrians because they walk more, or because they are more likely when walking to be injured (Steinbach, Green, Edwards, & Grundy, 2010)? Such questions can in principle be resolved using travel survey data comparing walking levels among different demographic groups. Results suggest disadvantaged pedestrians do indeed experience greater risk while travelling. Using British travel survey data, I found that lower income pedestrians have over twice the risk of being injured by a motor vehicle than those living in higher income households, after adjusting for walking frequency (Aldred, 2018). For disabled pedestrians, the risk gap was even larger.

Reasons for such inequalities will continue to be debated and may involve different types of risk factor. For example, highway design may not be inclusive of those with different needs and abilities. Disabled or older pedestrians may then struggle to travel safety, for instance if standard crossing times suit the “average adult” but not those with mobility impairments. Lower income people and those from ethnic minority communities may be more likely to live in areas with more motor traffic and/or greater levels of community severance caused by main roads. And drivers may not treat all pedestrians equally. Research from Portland, Oregon found drivers less likely to stop for Black than for White pedestrians at crossings (Goddard, Kahn, & Adkins, 2015).

Separating exposure from risk is at least relatively straightforward in principle for demographic variables. Considering other types of posited risk factor, the problem is greater. If we are for instance interested in mobile phone use, we need some comparator telling us about levels of mobile phone use while walking among the broader population (i.e. among the uninjured as well as the injured). There are ways of dealing with this through controlled study designs (for instance, case-control and case-crossover methods – see below for examples) but these remain relatively unusual.

For physical environment characteristics – the most promising in terms of direct modification – the exposure problem is greatest. And yet, if we do not deal with exposure, we risk undermining attempts to support and increase walking, an active mode which improves health and poses little risk to other road users. The lack of exposure-based analysis reinforces the reactive focus of road safety policy. We tend to intervene only after there have been deaths and/or serious injuries, which means ignoring places so hostile for pedestrians that few people with other options choose to walk.

The problem here is that we generally have little if any data on pedestrian presence across our route network. In Britain, we do not have official data (as we do for cyclists and motorists) on what proportion of pedestrian distance is along A roads, and what proportion is along minor roads. This means that we cannot even be sure that walking along an A road is riskier than walking along a minor road. Intuitively it seems likely, but without controlling for walking levels on different road types we cannot know. Compounding the problem, we often lack data on route network characteristics, particularly those that may matter for pedestrians but are of little interest to car drivers, such as footway quality.

Data on pedestrian highway injuries is also usually incomplete. Although some countries have reasonably good spatially referenced data on motor vehicle injuries, they lack similarly good spatial data for analysing on-highway falls. In Britain police injury data excludes pedestrian falls, instead only captured in hospital episode data, which is much harder to access and is not geo-referenced. This means that even if we have data on characteristics of the highway environment that matter to pedestrians, we cannot (given current injury data) examine how such characteristics correlate with pedestrian falls injuries.

This feeds into the exclusion of falls from road safety research and policy. Pedestrian falls – a substantial health burden on the older population, particularly women – are seen as separate from motor vehicle injuries. Recognising this, many academics are calling for falls to be incorporated into road injury statistics, allowing much more detailed investigation, including spatial analysis (Methorst et al., 2017). This would allow us to start identifying locational factors associated with falls injury risk and to act to mitigate these.

Even with good quality spatial injury data, however, exposure problems remain. Different approaches have been tried to deal with this. One “case-control” study examined child pedestrian injuries that happened close to home in relation to speed hump presence, comparing children injured by motor vehicles to children seen in emergency departments for other reasons (Tester, Rutherford, Wald, & Rutherford, 2004). Children injured by motor vehicles were less likely to live in a street with speed humps than the “control” group of children.

An alternative “case-crossover” approach (e.g. Roberts, Marshall, & Lee-Joe, 1995) looks to compare places where people were injured with randomly selected places on their routes prior to injury. This requires finding out (or modelling, potentially) a person’s route, which can be difficult and expensive. However, the method permits studying a wide range of factors, as such factors can be looked up for both “control” and “injury” sites. The growing availability of street imagery (e.g. Google Street View) is promising here, offering remote lookups that can identify a range of variables complementing other data held by transport authorities or available from public sources such as OpenStreetMap. Street imagery can be analysed either manually or (increasingly likely and possible) using machine learning.

Other new “big data” sources may make aggregate injury risk analysis possible, such as that conducted recently for cyclists in London (Aldred, Goodman, Gulliver, & Woodcock, 2018). This involved using a model of cyclist flows across the network to generate “control” sites representing where people might be injured if risk were evenly distributed. Control sites were then compared with police recorded injury sites in terms of route environment characteristics. Few cities currently have equivalent pedestrian flow models, but the growing availability of mobile phone data alongside other sources is potentially promising. Transport for London already have a pedestrian density model although at small area rather than route segment level. As such data sources and modelling techniques improve, it will become increasingly possible to conduct risk-based analysis of pedestrian injuries, including detailed spatial representation and (if pedestrian falls are integrated into road injury statistics) covering falls and motor vehicle injuries.

But risk-based analysis should go both ways. Road danger reduction perspectives highlight the importance of addressing the source of danger, rather than only focusing on the victims. Relatively little research examines which drivers pose most risk to vulnerable road users, and the role of road conditions/route environment in contributing to driver risk is even less well researched. Scholes, Wardlaw, Anciaes, Heydecker, and Mindell (2018) have explored the former question, confirming for instance the high risk posed by young male drivers to others. Given the increasingly good data available on motor vehicle flows across the road network, spatial analysis of factors associated with driver risk is also possible and merits further research. Could future sat-navs (and autonomous vehicle routing algorithms) route vehicles so as to prioritise child pedestrian safety over small time savings for drivers? Such research would make that possible.

Improving pedestrian injury data could help cities transform transport planning and traffic engineering. Too often, practitioners must rely on outdated techniques and data, inconsistent with visionary headline policy goals. For instance, London has ambitious targets for increasing walking and cycling; yet while the city measures and regularly reports on delays to drivers across the network, it does not do the same for pedestrian delays. Hence despite the aim of prioritising walking above driving, planners lack tools to minimise pedestrian delays and fall back on minimising delays to motor traffic. They cannot quantify and include in evaluations how longer waits to cross the road may impact pedestrians, for instance by discouraging walking and encouraging risky crossing behaviour among those who persist.

Finally, pedestrian injury research could learn from recent developments in cycling research exploring relationships between safety and comfort, such as the role of everyday “near miss” experiences in shaping perceived risk. When considering all injuries (not just those reported to hospitals or to police), most are slight and in health economic terms, impose a low cost. However, the direct health impact is only a small part of the injury risk burden. Research on children’s independent mobility shows how children have lost freedom and the right to physical activity as motor traffic levels have grown, and adults become more fearful (Shaw et al., 2013). Pedestrian safety must mean more than not being hurt: it must also mean creating and maintaining an inclusive, comfortable, and welcoming environment.

Readers might be interested in the Special Issue of Transport Reviews on the theme of Walking – How, Where, Why, and for Whom? http://explore.tandfonline.com/cfp/pgas/ttrv-cfp-walking. Papers can be submitted online up to the end of 2018, and the issue will be published in 2020.

Disclosure statement

No potential conflict of interest was reported by the author.