IN BRIEF

VISUALISATION London’s burning: Assessing the impact of the 2014 fire station closures What effect have fire station closures had on response times in central London?Benjamin M. Taylor investigates

n the face of ongoing austerity To its credit, the LFB collects boroughs can be divided further into measures imposed by the British information on all call-outs – including wards; however, my analysis sought government, the consequences of the type of fire, date, time of day, spatial to understand the effect of the station Icuts to public services are beginning location and response time – which closures at the sub-ward level. to unfold. One of the most contentious allows researchers to analyse the impact Deciding which spatial areas may decisions taken in 2014 related to the that station closures have had. In fact, be at risk of increased response times

closure of 10 fire stations and the loss in preparation for the closures, the Open Benjamin Taylor following the closures is not just a of 552 fire-fighters around the central Data Institute used this information to completed his question of producing maps of the London area. These measures were a work out that average response times PhD, “Sequential raw data. The time an engine takes result of the London Fire Brigade (LFB) across London were likely to increase Methodology with to respond to a call depends at what Applications in Sports being tasked with making savings of only marginally, from 5 min 34 s to 5 min Rating” at Lancaster time of day and on which day of the almost £29 million. 37 s. However, in parts of the city directly University in 2010. He week the fire occurs, as road traffic and

The affected stations were Belsize, affected by the station closures, the effect joined the CHICAS staffing levels will vary accordingly. Main image: Gary Blakeley/Hemera/Thinkstock research group, Bow, Clerkenwell, Downham, was expected to be more pronounced headed by Professor Response times also depend on the Kingsland, Knightsbridge, Silvertown, (london-fire.labs.theodi.org/). Peter Diggle, as a level of service provision in the vicinity Southwark, Westminster and Woolwich. Now, two years on, I have sought to research associate in – that is, the number of fire stations These stations were selected for closure bring some statistical rigour to these 2010 and subsequently able to respond in a reasonable time as a lecturer in (presumably) so as to minimise any efforts and to map the response times biostatistics in 2013. to a call – and the demands on those disruption to coverage and service at a more granular level, using data on stations being placed by other calls in affected areas. Indeed, the LFB dwelling fires only – as this type of fire This article is at the time of the fire, the effect of said at the time that services would accounts for over 80% of fire-related based on Benjamin which might be that an engine has not be impacted. The Fire Brigades fatalities. Across the UK, there were Taylor’s paper, to be deployed from a station further “Spatial Modelling of Union (FBU), however, had a different 39 600 dwelling fires in 2013–2014 Emergency Service out. There will, of course, be other perspective. Paul Embery, regional (bit.ly/1rf6yNu). Response Times”, influencing factors not measured by secretary of the FBU, said: “It will be which is published the available covariates. only a matter of time before someone Areas at risk in the Journal of the The main aims of my analysis were Royal Statistical dies because a fire engine did not get to London is divided into 33 districts – 32 Society, Series A at to identify spatial regions that had them in time” (bbc.in/1reZEb7). boroughs plus the City of London. The bit.ly/1TBkmLA potentially been affected by the fire

10 SIGNIFICANCE June 2016 © 2016 The Royal Statistical Society IN BRIEF

● ● ● FIGURE 1 Maps of dwelling fire locations

● and deploying stations ● ● ● ● ● ● ● ● connected by coloured lines in 2013, 2014 ● and 2015. Different ● ● ● ● ● ● ● colours are used to ● ● ● ● help distinguish the

● responses from each ● ● station. Solid markers

● are operational fire ● ●

● stations, hollow ● ● markers are the locations of the 10 fire stations closed ● ● ● ● ● ● in 2014. These plots 2013 2014 2015 ● show how the London ● ● ● Fire Brigade has ● ● adjusted its response station closures and to determine if strategy in light of the is necessary to link response time affected by station closures. Two closures; note that the response times had changed in those strategy employed data to data from the ambulance years on, the impact of the closures areas. Response times are partially in 2015 is different service and from hospitals to follow would seem to be more substantial governed by the redeployment strategy from that employed up health outcomes as a result of than anticipated. n in 2014. For example, adopted by the LFB which is constantly consider Old Kent Road fires, comparing the health outcomes evolving, as can be seen in Figure 1. station, highlighted of victims in 2012–2013 to those in Acknowledgements Specifically, I looked for areas in by an arrow. In 2014 2014–2015. Changes in the cost of Map tiles by Stamen Design, under CC which the response times (accounting this station was damage to property resulting from fires BY 3.0. Data by OpenStreetMap, under responding mainly to for time of day, a weekend effect and fires to the east, but could also be compared if the right ODbL. The London Fire Brigade data the level of local service provision) in 2015 responded in data were available. are subject to the Open Government seemed higher in 2014 and 2015 all directions, as it did Two years ago, the LFB declared Licence: http://www.nationalarchives. in 2013 compared with 2013 and 2012. publicly that services would not be gov.uk/doc/open-government-licence/. Figure 2 gives an idea of the size and location of regions potentially affected by the closures and how these have changed between 2014 and 2015. For residents living in the affected areas, the consequences of the closures could be serious. Following the closures, the LFB has only been able to respond to around 50% of calls in some of these regions (marked as red, blue or green) within its 6-minute target. Before the closures, there was no problem with response times at all. FIGURE 2 Areas where response times were higher in 2014, Fire alarm 2015, and both 2014 In my analysis, I assume a proportional and 2015 than they hazards spatial survival model for the were in 2012–2013, adjusting for time of response times. While there may be day, a weekend effect other sensible modelling alternatives and level of service and methods for identifying potentially provision. The bold line marks a 4 km buffer affected regions, my analyses indicate around the convex that the 2014 closures have had a hull of the closed negative impact on response times in fire stations; we only consider differences the central London area. in response times To find out the substantive within this region in implications of these closures – that the present article. The is, whether increased response times results are presented on a 500 m by 500 m have led to more injuries and deaths, square grid of small as the FBU suggested it would – it areas

June 2016 significancemagazine.com 11