Abstract
Introduction: In the past 20 years, the concept of the digital twin has been utilized in other fields to track the operations of wind turbines, monitor the status of spacecraft, and even create a model of the Earth for climate research While artificial intelligence holds much promise for the neurocritical care unit, these models should be clearly based on an understanding of underlying physiological variables rather than be opaque “black box” models. Our group has previously created such a digital twin model to predict acute response to treatment of sepsis. This project expands on our previous work by developing a similar model for the critical care of acute ischemic stroke. Methods: A panel of 18 experts in the field of critical care reviewed the clinical practice and pathophysiological statements related to neurology critical care. The core group of investigators developed statements based on a Directed Acyclic Graph (DAG) describing pathophysiology surrounding acute neurological issues encountered in the practice of Neurocritical Care (NCC). A modified Delphi method (3 rounds) was used to gauge agreement on 20 statements (120 sub statements) using a 7-point Likert scale. Agreement was defined a-priori by >80% selection of a 6 (“agree”) or 7 (“strongly agree”). Results: Consensus was achieved on 58 (48.3%) expert statements after completion of 2 rounds. After completion of 3rd round, the level of consensus increased to 93 (77.5%) expert statements. Some statements garnered 100% consensus in the first round of DELPHI such as “Reperfusion of ischemic stroke can lead to improvement of stroke.” Other statements required revision to capture the nuances of clinical practice before gaining consensus. For example, when the severity of infection was incorporated into the statement “Infection can lead to low blood pressure.” consensus increased from 56% to 100%. Conclusions: Our study demonstrated the feasibility of application of the DELPHI process to generate consensus among experts for use in development of a “digital twin” artificial intelligence model for use in the Neuro ICU. Compared to other models which rely on “black-box” associative artificial intelligence, our proposed causal AI model is based on a solid foundation of expert rules and casual mechanisms.
Talk to us
Join us for a 30 min session where you can share your feedback and ask us any queries you have