Search papers, labs, and topics across Lattice.
This paper introduces a novel clinical decision support system for optimizing the switch from intravenous (IV) to oral antibiotics by forecasting patient vital sign trajectories using neural processes. The system predicts switch-readiness by comparing these forecasts against clinical guidelines, rather than learning from historical decisions, thus avoiding the reproduction of existing delays and inconsistencies. Validated on MIMIC-IV and UCLH datasets, the system significantly outperforms random selection, identifying 2.2-3.2 times more relevant patients for potential antibiotic switching.
Ditch the backward-looking decision support systems: this new approach forecasts patient physiology to proactively identify candidates for IV-to-oral antibiotic switching, dramatically improving relevance compared to random selection.
Timely transition from intravenous (IV) to oral antibiotic therapy shortens hospital stays, reduces catheter-related infections, and lowers healthcare costs, yet one in five patients in England remain on IV antibiotics despite meeting switching criteria. Clinical decision support systems can improve switching rates, but approaches that learn from historical decisions reproduce the delays and inconsistencies of routine practice. We propose using neural processes to model vital sign trajectories probabilistically, predicting switch-readiness by comparing forecasts against clinical guidelines rather than learning from past actions, and ranking patients to prioritise clinical review. The design yields interpretable outputs, adapts to updated guidelines without retraining, and preserves clinical judgement. Validated on MIMIC-IV (US intensive care, 6,333 encounters) and UCLH (a large urban academic UK hospital group, 10,584 encounters), the system selects 2.2-3.2$\times$ more relevant patients than random. Our results demonstrate that forecasting patient physiology offers a principled foundation for decision support in antibiotic stewardship.
