SPHERE-PPL NHS Severe Patient Harm Forecasting Contest

The SPatial, Health & Environmental REsearch using Probabilistic Programming Languages (SPHERE-PPL) team are excited to announce the launch of the SPHERE-PPL NHS Severe Patient Harm Forecasting Contest.  

SPHERE-PPL is a community of researchers and data scientists focused on advancing the use of AI forecasting in health and environmental science.   

Every four hours of delay in Emergency Department (ED) admission is associated with an 8% increase in 30-day mortality risk — roughly 25 potentially avoidable deaths per month. Accurate forecasting can help hospital managers take pre-emptive actions to reduce this risk.  

The goal of the contest is to develop an algorithm that accurately forecasts the number of estimated avoidable deaths over 1–10 day horizons.  

Data: Estimated avoidable deaths as the outcome variable, alongside 220 explanatory healthcare variables, covering March 2023 – September 2025 (development set) and October 2025 – March 2026 (assessment set).

Timeline:

• Final algorithms due: 5 June
• Assessment dataset released: 6 June
• Final submissions (including performance on assessment dataset) due: 20 June

Tools:

• Models should be implemented in R or Python.

Evaluation:

• Accuracy assessed via Mean Squared Error (MSE) over short-term (1–5 day) and medium-term (6–10 day) horizons.  

For full contest details and participation instructions, please visit the GitHub repository: https://github.com/SPHERE-PPL/NHS-EAD-forecast

This is a unique opportunity to make a real-world impact. The winning model will be used daily by Bristol NHS managers to provide advance warnings and support proactive decision-making.  

Further information:

• Email: [email protected]
• https://sphere-ppl.org