NHS Foundation Trust Unlocks AI Healthcare Potential with Dash Enterprise

Introduction

Somerset NHS Foundation Trust (SFT) provides Acute, Mental Health, Learning Disabilities, Community, and General Practice healthcare services to the county of Somerset, England. It runs services out of two main Acute hospitals, 13 Community Hospitals, and a quarter of GP Practices in the local area. The Trust employs around 13,500 staff including Doctors, Clinical Specialists, Nurses, Operational Leaders, and Administrative staff.

The Data Science & Operational Research Team at Somerset NHS Foundation Trust is a dedicated function for providing prescriptive and predictive insights to the organization. The team prioritizes the use of open-source software to promote inter-organizational sharing of models and techniques in the wider National Healthcare Service (NHS). The Trust chose Dash Enterprise as the solution for building enterprise level data science applications that could service the models and techniques created by the team to the entire organization.

Challenge

The team at Somerset NHS Foundation Trust has vast experience in building predictive and prescriptive models to assist clinicians and operational leaders in making decisions. However, the scale of the organization, 13,500 staff across 20+ sites, makes deploying models and reaching all the staff who need assistance problematic. The structure of the NHS means that any individual organization can be trialing new and innovative ways of delivering care, meaning, in most cases national models don’t fit the nuances of individual organizations, therefore these models need to be built bespoke. To overcome this challenge, the team began looking for a tool that could deploy their models, securely, without losing functionality and ideally enhancing them. This is where Plotly came in. 

Successful data science operations at this scale require a data application solution that is:

• Built for Data Science, not Business Intelligence: Data scientists run complex artificial intelligence/machine learning models while leveraging the vast open-source ecosystem of Python libraries and packages.
• Secure: With sensitive patient care and operational data at stake, features like active directory user authentication, authentication middleware, and user activity IAM logging are ideally included rather than implemented as a separate layer.
• Real-time rendering: Critical information updates and data views render without delay so that decisions can be made promptly.

Solution

Our trust chose to work with Dash Enterprise, because the Data Science & Operational Team quickly builds and deploys data applications for a wide range of care provider and organizational needs.

Causal Impact Analysis & Statistical Process Charts

The team was spending a large proportion of their time performing causal impact analysis and statistical process charts on change programs at the hospital. By utilizing Plotly’s ability to ingest data and build front ends, the team was able to create a web-based tool where users can self-service.

Musculoskeletal Waiting List

Planning services to meet their demand is a recurrent and widespread workstream. In this application, the team focused on the musculoskeletal department, helping clinicians and operational leaders to effectively manage their capacity and test various assumptions to reduce waiting times in the same application.

Call Centre Modelling (Erlang-C)

Somerset NHS Foundation Trust manages several call lines for remote monitoring and community services. Planning to ensure that these phone lines are manned and able to meet call handling and answering metrics was previously a manual job. The team built a call center model based on Erlang-C to empower operational planners to make these adjustments on the fly.

Inpatient Anomaly Detection

Detecting anomalies in real time allows operational leaders to intervene before issues escalate. Automated detection scans daily and hourly information and highlights these issues in an app that our colleagues found simple and intuitive to use.

Inpatient Forecasting

The team produces forecasts on inpatient occupancy to allow our patient flow teams to plan bed capacity throughout the year. By making these forecasts available through Dash Enterprise, staff can see the latest daily forecast and adjust where needed.

Results

Plotly has enabled our trust to deploy prescriptive and predictive methods to users across departments and at multiple sites.

For our trust the benefits have been:

• Better-informed decisions: Department metrics are visualized in real-time. Admissions teams have more time and information to admit patients and assign beds. Managers can schedule on-call personnel with greater accuracy.
• Wider understanding of decision impacts: A single application grants cross-departmental users the ability to run advanced analytics techniques, putting the power of data science into hands across the organization.
• Performant apps without delays: Dash Enterprise’s built-in caching ensures fast UX/UI rendering for even complex data apps, so NHS staff can make decisions quickly and efficiently.

Built-in app security: Dash Enterprise includes a level of security above other solutions, including active directory user authentication, authentication middleware and user activity IAM logging.