Cycling Injury Risk Dr. Rachel Aldred, Reader in Transport, University of Westminster Georgios Kapousizis, Research Associate, University of Westminster
Pedal cycle fatalities, 1949-2012 https://assets.publishing.service.gov.uk/government/uploads/system/uploads/attachment_data/file/447674/pedal-cyclists-2013-data.pdf
Pedal cycle fatalities, 1949-2012 A positive story? https://assets.publishing.service.gov.uk/government/uploads/system/uploads/attachment_data/file/447674/pedal-cyclists-2013-data.pdf
Pedal cycle traffic, 1949-2013 https://assets.publishing.service.gov.uk/government/uploads/system/uploads/attachment_data/file/447674/pedal-cyclists-2013-data.pdf
Fatality risk for people cycling, per bn veh miles Parliament 2012 - www.parliament.uk/briefing-papers/SN06224.pdf
Fatality risk for people cycling, per bn veh miles Parliament 2012 - www.parliament.uk/briefing-papers/SN06224.pdf
Without risk, we can’t identify dangerous route environments… Cycling injuries (Stats19)
Without risk, we can’t identify dangerous route environments… Cycling flows (Cynemon, modelled) Cycling injuries (Stats19)
Controlling for cycling volumes/flows: two methods for whole-network analysis Case-control Uses aggregate data Two data sources: data on where injuries happen (=cases) & data on where people cycle (→ controls) Usually don’t have data on where people cycle, but in London TfL have created the ‘Cynemon’ model, which can be used to identify a set of control sites Join to route environment data to identify and compare characteristics of control and injury sites
Example results: pilot case-control study
Example results: pilot case-control study
In the regression model… See Aldred et al (2018) in Accident Analysis & Prevention: https://www.sciencedirect.com/science/article/pii/S0001457518301076
Uses individual level data Controlling for cycling volumes/flows: two methods for whole-network analysis Case-control Case-crossover Uses aggregate data Two data sources: data on where injuries happen (=cases) & data on where people cycle (→ controls) Usually don’t have data on where people cycle, but in London TfL have created the ‘Cynemon’ model, which can be used to identify a set of control sites Join to route environment data to identify and compare characteristics of control and injury sites Uses individual level data Cases are their own controls: for each injured individual we have an injury site and a control site The control site is selected randomly from individuals’ routes prior to injury Hence, like the case-control method, you have a set of control sites representing where people might have been injured if all route environments equally safe/dangerous Like case-control method, then join to route environment data to identify high-risk factors
Reducing Cycling Injury Risk While Cycling Grows Case-crossover study for UK (no Cynemon model outside London!) Uses 2017 Stats19 data for weekday AM commuters alongside home postcode data from DfT/PSNI NTS confirms >92% of these commuters travelling from home We will then algorithmically model the commuters’ routes, using this to select our control sites Currently reviewing factors to include within our route environment dataset & available data (your advice welcome!) We also have the option to use Google Streetview to look up route environment characteristics where data does not exist
Route Environment Characteristics Potentially Affecting Injury Risk Route environment factors-safety in numbers* Example Findings Number of studies Intersection 50% - 60% of cycle injuries (Miranda-Moreno et al., 2011) Signalized intersections are 25% more risky than non-signalized intersections (Strauss et al. 2015) 15 Road type Bike infrastructure (Footway, bike lane, bike path) Injury risk is significantly higher in major streets with parked cars (Teschke et al., 2012, Winters et al. 2012) Minor roads increase injury probability by almost 60% compared to collector roads (Williams, 2018) Bike paths reduce injury risk by 50%-60% (Rodgers, 1997, Teschke, et al., 2012) 5 Lighting Poor lighting may be associated with 25% of cyclist fatalities in EU countries (Evgenikos et al., 2016) Increase in injury risk probably >30% (Wanvik, 2009) 4 Land use/population density, parked vehicles Commercial and mixed-use land leads to a 20%-30% increase in injury risk (Tin Tin, et al., 2013, Williams, 2018, P. Chen, et al., 2019) Bus stops Bus stops are associated with a 40% increase in injury (Strauss, et al., 2013) Conversely, bus stops may reduce risk (Adams, 2018) 7 *Some of the factors might be on the same study **Not exhaustive data
Modelling Routes Taken ArcGIS, Network Analyst: Dijkstra's algorithm used by the ArcGIS Network Analyst. It solves the shortest-path problem on a weighted, directed graph (G = V, E) Where: V: is a set whose elements (Vertices) E: is a set of ordered pairs of vertices (Edges) Alternative options: bespoke cyclist routing (Cyclestreets, etc.)
More information/ideas? Please contact Georgios Kapousizis: g.kapousizis@westminster.ac.uk or Rachel Aldred: r.aldred@westminster.ac.uk Control and injury sites in relation to bus lanes, pilot study (Aldred et al, 2018)