Author:
Weissman Gary E.,Bishop Nicholas S.,Schmid Benjamin E.,Madden Vanessa,Pirracchio Romain,Jentzer Jacob C.,Riman Kathryn A.,Krutsinger Dustin C.,DiGiovine Bruno,Ungar Lyle H.,Halpern Scott D.,Auriemma Catherine L.,Valley Thomas S.,Kerlin Meeta Prasad
Abstract
AbstractTreatment decisions for patients receiving invasive mechanical ventilation (IMV) are complex and depend simultaneously on the current ventilator settings, the function of multiple interrelated organ systems, and other treatments. An artificial intelligence (AI)-based clinical decision support system (CDSS) offers a promising approach to alleviate uncertainty in this complex management and provide personalized treatment recommendations. However, little is known about clinician preferences for which treatment decisions clinicians think would most benefit from a CDSS. Therefore, we conducted a cross-sectional electronic survey of practicing physicians, nurses, advanced practice providers, and respiratory therapists to identify key treatment decisions in IMV care. We sent the survey instrument to 132 clinicians across six geographically diverse health systems. Among 51 respondents (39% response rate), there were 24 (47%) physicians, 7 (14%) registered nurses, 8 (16%) advanced practice providers, and 12 (24%) respiratory therapists. Participants were from five US states including Pennsylvania, California, Michigan, Minnesota, and Nebraska. At least 50% of participants identified 18 distinct treatment decisions for IMV care asvery importantorabsolutely essential, including many that were outside of the ventilator settings themselves. The highest agreement about importance was for the decision to extubate (N=51, 100%), and the decisions to conduct a spontaneous awakening trial (N=48, 94%) and spontaneous breathing trial (N=48, 94%). The highest agreement about a decision beingnot importantor onlyslightly importantwas for the shape of the inspiratory flow pattern (N=15, 29%). These findings underscore the scope and complexity of clinical decision making across multidisciplinary teams in caring for people receiving IMV. Furthermore, they underscore the gap between important clinician-identified decisions and existing IMV CDSSs that provide suggestions for at most 5 ventilator settings without support for other related decisions. Future work is needed to identify which of these decisions might be most appropriate for inclusion in a CDSS to support IMV management.
Publisher
Cold Spring Harbor Laboratory