Our three-part blog series explores the greater role transport data can play in supporting ambulance dispatch, in the context of health devolution in London.


Archie Drake

The purpose of the London Ambulance Service (LAS) is to make sure people get the help they need. LAS saves lives. Our ‘Data Awareness for Sending Help’ (DASH) project looks at how technology is changing the data which LAS can draw on to do this crucial job.

We’ve found that our collaborators at LAS are enthusiastic about what the emergent world of ‘big data’ means for their work. They’ve been working hard to explore and unlock its potential, for example through their Pre-Hospital Emergency Department Data Sharing project, which aims to ‘better understand what happens to patients after ambulance staff leave them at the hospital’s emergency department’, and through their partnership with the GoodSAM app, which enables properly trained volunteers to sign up as responders for life-threatening emergency calls. It’s not easy; there are all sorts of legitimate concerns like ensuring patient confidentiality and making sure responders have the right skills. But these sorts of initiatives are helping LAS respond more effectively. The picture below is a snapshot from GoodSAM’s website showing ‘Community Life Saver Density’ (the location of available responders) across London.


We’ve also found that LAS are constantly evolving their approach to service provision in order to adapt to change. Most recently, the NHS England Ambulance Response Programme has introduced some very sensible reforms to decision making around ambulance dispatch, including more time for information-gathering at the triage stage (helping improve the quality of decisions about which incidents are life-threatening) and adjustments to the way response times are measured (helping to undercut incentives which serve the target more than the patient).

What’s missing at present is effective city-level coordination support for LAS. Things get more difficult for LAS when the data they need comes from outside the healthcare domain. As things stand, one of the clearest challenges appears to be access to rapidly evolving sources of real-time transport data. Given that the Mayor directs Transport for London (TfL), health devolution in London creates an opportunity for the city to address this and demonstrate the potential benefits.

The example of transport data: why and how this matters for London’s ambulances

The biggest transport problem for ambulances is traffic. Even with the priority accorded by their blue lights and sirens, ambulances find it difficult to get through London’s congested streets to emergency incidents. Like other ambulance services working in congested areas, LAS’ reaction has been thoughtful and profound: they have developed more agile resources like fast response car, motorbike and bicycle units to get through traffic more quickly.

Photo: BBC

Transport data also offers ambulance services an opportunity to understand Londoners’ mobility patterns, which is important because understanding where people are is a key starting point for working out when ambulances are likely to be needed. That means LAS is interested in the same ‘heartbeat’ that London’s transport planners monitor to keep the city moving. The animation below is produced by the mapping and transport data firm HERE, using TfL data to show the London tube’s daily heartbeat.

Tube Heartbeat

There are several points in LAS’ operational dispatch decision-making process at which these data types matter. One is the ‘routing engine’ element of the computer assisted dispatch (CAD) system, which generates estimates of how long it will take a selection of nearest available ambulances to reach an incident and recommends the one that will take the least time to reach it. Another is the routing information passed through to ambulance drivers and other LAS responders on their on-vehicle Mobile Data Terminals (MDTs) to support them in finding the fastest route.

The key here is the systematic accuracy of predicted route cost estimates. A system which uses less accurate route cost estimates will tend to make sub-optimal decisions about which resource to allocate to an incident, i.e., ambulances will be allocated that will take longer than expected to arrive, potentially endangering the person needing emergency medical attention. Companies like Uber are relentlessly focused on developing more efficient solutions to this problem so that people get taxis as quickly as possible. Why should we not apply the same line of enquiry to save lives?

The image below is a chart from an Uber Engineering blog post from 2015. It summarises the benefit of using improved route cost estimates (Estimated Times of Arrival – ETAs) in a dispatch system. Where more accurate estimates can be used, with less difference from Actual Time of Arrival (ATA) less frequently than before, the overall efficiency of the dispatch system can be improved to the benefit of users as well as the provider.

Uber Engineering

Transport data, including traffic prediction, is also significant at the tactical and strategic levels of LAS’ dispatch decision-making. Knowing which areas of the city are experiencing unusually high levels of traffic, for example, or which areas are likely to be particularly affected by congestion over the coming weeks and months, can support resource deployment planning. And knowing whether London is successfully meeting its congestion-reduction targets would help LAS test the assumptions it uses for fleet planning.

What types of transportation data have become available, and from where

London is considered to lead the world in intelligent transportation and mobility. In more concrete terms, the emergence of two notable types of data denote the transformation of transport data in London since the turn of the millennium. The first is ‘live’ traffic data, crowdsourced in real time from GPS-determined locations of devices in smartphones and vehicles. The second is e-ticketing data derived from Londoners’ use of Oyster cards.

The vast majority of us are now familiar with the image below (or something like it) in our day to day lives, which is an example of the first kind of data. It is a snapshot of Google Maps’ live traffic layer and shows the current traffic speed on main London roads (red for slow, green for fast) at a given moment in time. For an example of the second kind of data, see the HERE animation already shown above.

Google Maps (displaying traffic, 30 Oct 2017 5.39pm GMT)

Although the traffic data landscape is still largely dominated by private actors, TfL is at the cutting edge with both types of data. Surprisingly little information is available to the public about TfL’s Surface Transport and Traffic Operations Centre (STTOC), but it apparently ‘uses a variety of technologies to monitor and control traffic’. TfL operates in close partnership with the Google-owned Waze on traffic data. Its work on Oyster data, meanwhile, appears to focus on a partnership with MIT targeted mainly at customer analytics, operational management and strategic planning.

The chart below is from the 2016 edition of the TfL’s ‘Travel in London’ statistical series and gives a quick impression (though only static and at an aggregated level) of the kind of knowledge about congestion that TfL is able to generate thanks to its data on transport patterns. Its shows how London (with its road network carrying close to maximum capacity) ‘shows large differences in the degree of variability of traffic congestion by both area and time period’.


Like many cities now, TfL makes a large amount of data available through its open data programme, including a unified API designed to simplify access which drives many of the applications Londoners use every day to plan and guide their mobility. It also engages vigorously with the ‘developer community’ in business to build a user base through which to realise the economic value of its data. But the API only provides current data (not past or predictive), and provides only a fraction of TfL’s wider awareness of transport dynamics in the city.

A sense of the LAS opportunity on transport data

Despite the heroic efforts of LAS technical staff and control room management to go above and beyond their core challenge of maintaining and operating CAD through daily pressures, LAS is not currently able to take advantage of live traffic data in its routing engine. Routing information on MDTs is not dynamically adjusted according to traffic conditions to support ambulance drivers in getting to incidents efficiently. LAS tactical and strategic planners do not currently have access to historical transport data to support their decision-making.

What research has been done on the opportunity here for ambulances has used ‘dynamic’ models to suggest quantified gains. Routing estimates that are sensitive to time variations in traffic help direct individual responses better, and can also help improve the way that ambulances are distributed around the city given the overall demand from the whole. A 2013 paper from Poulton & Roussos (Univ of London), based on LAS data from 2012, suggested that this approach could enable the majority of Category A (most serious, time sensitive) incidents to be reached in 254 seconds rather than the actual 360 seconds (i.e., 29% faster). A 2010 paper which used data from Vienna quantified the whole-system ‘solution improvement’ at more than 10% on average.

Category A arrival times compared: historic (2012) vs. Birkbeck’s ‘CARD’ variable-traffic model (Poulton & Roussos, 2013)

Subsequent work by Poulton and others at LAS to make a system known as Geotracker (now Quest) available to LAS in addition to CAD has begun to make progress towards improved route cost estimates based on historical ambulance travel times. This is likely to represent a significant improvement and it should be noted as a real LAS accomplishment given the challenges.

But there is further to go; further improvements can be realised through the integration of live traffic data. This is not simply a question of LAS tapping into TfL’s API. Effective integration of live traffic data into LAS dispatch decision-making systems would require a programmatic partnership effort of the type TfL uses for its own exploitation of traffic and ticketing data, probably including cooperation across public-private and practice-research lines.

Why? Firstly because, from a technical standpoint, effective practical extension of the state-of-the-art in fleet management for this purpose is non-trivial:

‘… many fleet management problems correspond to combinatorial optimisation problems (e.g., vehicle routing, scheduling, network design) that are notoriously difficult to solve, even in a static context. This, coupled with real-time requirements, explains to a large extent the reliance up to now on human dispatchers.’ (Bielli, Bielli & Rossi, 2011)

Secondly because the effort requires both specialist transport modelling expertise not currently available within LAS itself. And thirdly because the opportunity extends beyond the control of LAS and into the potential for TfL to manage London’s transport system in a way that actively accounts for LAS requirements as a priority user of the system. The clearest example of how this might be useful is dynamic traffic light adjustment to help ambulances reach priority emergencies faster even in heavy traffic. This idea has been current in the US since at least 2006, and forms the subject of a current Innovate UK-funded trial in Liverpool.

In the absence of more a focused effort, the potential here remains largely theoretical and TfL is rightly concentrated on the vital job of keeping London moving every day. However with relatively little TfL analytical attention there is the potential for significant system efficiency gains here. They would materialise as improvements to London ambulance response times and therefore, at least in some cases, lives saved. In more qualitative policy terms, initiatives which could achieve results like changing traffic lights to help ambulances respond quicker could make a strong impression as to the quality of proactive leadership of a devolved city health system. TfL may also find quantifiable gains within its own remit, for example maybe through using the closer partnership to reduce transport delays from passenger illness.

In the meantime, we are faced with a bizarre situation in which most Londoners have better traffic information available on the smartphone in their pockets than is used in dispatching London’s ambulances, and a city transport system that knows much more about the movements and behaviours of its customers than its counterpart emergency ambulance service.