Ross, Magnus; Swanepoel, Nel; Luintel, Akish; McGuire, Emma; Cox, Ingemar; Harris, Steve; Lampos, Vasileios; (2025) Expediting Switches from Intravenous to Oral Antimicrobial Therapy with Neural Processes. In: Proceedings of the Machine Learning for Health (ML4H). (In press).

Preview

Text camera-ready.pdf - Published Version Download (2MB) | Preview

Abstract

Intravenous (IV) antimicrobial therapy often continues after an oral alternative would be safe, harming patients and increasing costs. The key decision—when to switch from IV to oral (IVOS)—is complex. Past machine-learning approaches are trained on labels constructed from prescription records that indicate when switches actually happened, not when they were first clinically appropriate; they therefore reproduce delays and suboptimal practice. We take a different approach, forecasting each patient’s physiological state using a probabilistic neural process and applying established IVOS criteria to those forecasts to estimate switch-readiness. The system ranks patients currently on IV therapy by the probability they will meet all criteria within the next 12 hours, highlighting candidates for clinician review while keeping the final decision with the prescriber. By anchoring recommendations to clinical criteria rather than past actions, the tool targets a reduction in unnecessary IV days instead of reinforcing the status quo.

Download activity - last month

Export as CSVXMLJSON

Download activity - last 12 months

Export as CSVXMLJSON

Downloads by country - last 12 months

Export as XMLJSONCSV

Archive Staff Only

View Item