Adaptive multi-infusion decision support for the multivariable circulatory management of critically ill patients

摘要

We have developed a novel adaptive multi-infusion advisory system for circulatory management of critically ill patients which co-ordinates infusion adjustments to ensure safe trajectories. This system should reduce patient hospital stay and improve patient outcome by enhancing the quality of patient circulatory control; alleviating the clinical cognitive load, giving staff more time for direct patient care, while also reducing infusion adjustment errors. We have applied three derived circulatory variables which relate to the three main types of cardiovascular infusions (inotropic, vasoactive and fluid). A lumped parameter steady flow model of the human circulatory system and the effects of cardiovascular infusions was constructed for algorithm development, clinical experts providing feedback on a representative test set of simulated patients in circulatory shock. Independent self-learning fuzzy logic controllers (SLFLC) were found to give good adaptation to variable patient infusion sensitivities. A supervisory, rule-based module co-ordinates infusion adjustments to ensure safe circulatory trajectories. Monitoring of manual infusion adjustments allows timely advice and also a critiquing capability which can train junior staff and reduce infusion adjustment errors. A physical mock circulatory loop was used to construct and test our physical advisory system. Preliminary clinical results show good clinical utility of our adaptive multi-infusion advisory system.