At Mid Yorkshire NHS Teaching Trust, pharmacy teams traditionally spent hours manually identifying high-risk patients across 1,100 beds, a process that was time-intensive and ripe for digital transformation.
This session shares the journey of developing the Clinical Pharmacy Prioritisation Software Tool (CPPST), a proof-of-concept that leverages Machine Learning and Generative AI to autonomously flag high-priority patients with 93-97% accuracy. Attendees will learn how the project team moved from a digitised dataset to a predictive model that boosts staff morale and improves efficiency. We go behind-the-scenes of a responsible AI deployment, detailing how the team overcame clinician nervousness towards machine learning through innovative, play-based education and human-centred design. From the Discovery phase to iterative data modelling, we explore how to ensure AI outputs remain human-readable and explainable, supporting rather than replacing expert clinical judgement.
- A blueprint for Responsible AI: How to use ethical frameworks and human-centred design to move from a data hypothesis to a viable clinical decision-support tool.
- Building clinician trust: Practical techniques (including gamified workshops playing "Guess Who") to demystify machine learning concepts for frontline staff and overcome cultural barriers.
- Technical iteration and explainability: Insights into the transition from statistical modelling to using Generative AI to provide human-readable reasoning for why a patient is flagged as high-risk.
- Data governance and security: Navigating the complexities of NHS data profiling and pseudonymisation to ensure patient privacy while fostering innovation.
- Scaling impact: How identifying complex cases with high accuracy (<2.2% false positives) allows frontline staff to devote more time to direct patient treatment.
- Anna Lay, Principal Consultant, Hippo
