Seminar Presented by Leanne Smith and Robin Hutchings of the London Ambulance Service
Article by Chaoyu Hou and Jack Humble.
London Ambulance Service Summary
Recently, the London Ambulance Service (LAS), one of the oldest ambulance services in the world, has been increasing its utilization of new and higher quality data to develop an improved strategic programme for dispatching its ambulances and staff. On the 3rd October 2018, Head of Forecasting and Planning, Leanne Smith, and Head of Business Intelligence, Robin Hutchings, visited Kings College London to give a talk on the developments and changes to the LAS in recent years. Being the busiest ambulance service in the UK with an increasing service demand year on year, the LAS needs to be at the forefront of the digital change to health services. Changes to the way in which data is collected and analyzed, as well as new techniques used for inferring information from this data are just some of the method the LAS is aiming to provide a more holistic approach to the problem of ambulance dispatch. In this summary, we will look at the main departments of the LAS and how they are adapting to a world in which understanding data is now more significant than ever.
Both internal and external sources of data are available for the LAS to use. Internally, they have information on the 999 calls, whether it was called by the patient or not, as well as information about the incident and what decisions were made once there. The times of the calls, dispatches, and arrivals are also recorded. The vehicle that was chosen to respond and the experience of the staff on board are also collected. Externally, the LAS is looking to incorporate data outside of their control to improve the service. Transport Data offered by London’s Intelligent Transport System could help reduce the journey time for ambulances. Tourism data as well as data on air quality, provided by the London Air Quality Network, will help predict demand. Finally, weather conditions and forecast data will be considered for implementation into the dispatch system.
The Ambulance Response Programme (ARP), established in 2015, conducted one of the largest studies of an ambulance service in the world. Based on 14 million calls collected over 18 months, they proposed a new set of performance targets. Its main objectives were to prioritize the sickest patients, ensuring they could receive the fastest possible response as well as distributing resources effectively so no patients had to wait unacceptably large amounts of time. However, this new system changed the way in which start and stop clock data was collected, meaning 18 years of previous data was now redundant and the new data collection method conforming to the ARP specifications replaced the old system.
As well as changing the way in which data is collected, the ARP also announced a series of updated ambulance management measures. The biggest change is the format of the call, allowing more time to triage the patients and provide the most appropriate response. Four categories of call were proposed: life-threatening illnesses or injuries calls (response within 7 minutes); emergency calls (within 18 min); urgent calls (within 120 min) and less urgent calls (within 180 min). The aim of this is to save time and staff labour for emergency services. Another change is to provide more time to understand the condition of the patients and identify their demands, allowing for quicker identification in an emergency. Moreover, they try to balance the workforce with appropriate levels of experience to better match demand and capacity of that time.
The LAS spreads its many types of data analytics across two teams. Business Intelligence (BI) serves as the core, working with diagnostic and descriptive analysis. Forecasting and planning, on the contrary, focuses on the predictive and prescriptive side of the analysis. Supplementary to this is the geo-services team who aim to allow this analysis to be done with a spatial element. Additional teams include Data Quality Assurance, Clinical Audit and Workforce Analytics.
Business Intelligence provides a more diagnostic and descriptive form of analysis. Prior to the new system introduced by the ARP, the business intelligence team had been capturing and analysing multiple datasets for over 18 years. They developed a Business Intelligence Portal to host as many as 256 reports which any member of staff could access. These reports included information about 111 services, A&E, fleet and logistics, patient transport service and emergency bed services to help NHS healthcare staff find hospital beds for seriously ill patients. These reports and data can then be shared with internal and external stakeholders, NHS England, NHS Improvement, Clinical Commissioning Groups, Local Governments as well as the other emergency services (London Fire Brigade, Metropolitan Police). With the advancements in technology over the past 18 years, so too has the quality of the data. The ARP proposed a new set of response time targets, now considering the mean, median and 90th percentile of response times as opposed to the traditional method of trying to meet 75 percent of calls within a certain time period as seen in Figure 1. This new start posed a high risk to the patients and organisation. However, once new reports are developed, agreed upon and peer-reviewed, the updated system will provide a much more holistic overview of all frontline operations. These reports are shared again through the Trusts Intranet site allowing all staff access and represents a gold standard for the organisation. On top of this, Dashboards are created for the Clinical Commissioning Groups (CCGs) using Tableau to visualise much of the data.
The forecasting and planning team has only been in place for two and a half years, currently consisting of four members. They aim to operate on a strategic, tactical and operational level. Both planning for short-term events such as Christmas and New Year or extreme weather conditions, as well as long-term planning to account for changing populations. However, due to the ARP, only 11 months of data are available to work with so they do not yet have the data to represent a full seasonal cycle. When implementing a predictive model, a vital step is to choose relevant features in which to train your algorithm. Features found to improve accuracy include demand, number of staff and A&E queue length among others. A linear regression model was chosen as a starting algorithm. Initial iterations of the model showed promise, accurately predicting the demand for a couple of weeks. But unaccounted for snowfall in March caused a large error in the model. This posed the question of how the developer’s resources should be allocated. Focusing attention on these rare events to greatly improve prediction accuracy for such occasions or spending time improving the model as a whole so small improvements can be seen across the whole year. These models are built in R and Python. They are then converted and stored in excel so they are accessible to frontline colleagues and senior managers.
Future aspirations for the team include sector-specific models. If a large event was happening in one part of the city, LAS could increase supply for specific locations rather than London as a whole. Accumulating the information from the BI team to create a full demonstration of analytics is also an ambition by displaying past, present and future predictions within one space. Finally, having access to electronic care records could improve the time taken to provide a more appropriate response. The work that the forecasting and planning team is doing will provide a much more sophisticated way of dealing with high demand as well as possibly providing a financially optimal solution, further increasing the level of performance the London Ambulance Service provides.
- Data and Predictive Analytics at LAS – October 2018, Leanne Smith Segment of the presentation, the head of forecasting and planning at the London Ambulance Service NHS Trust(Figure 1)
- Business Intelligence ARP at LAS – October 2018, Robin Hutchings Segment of the presentation, the head of Business Intelligence
- Our strategic intent 2018/19 – 2022/23, London Ambulance Service NHS Trust