Development of a clinical decision support tool using machine learning to assess a patient’s individual risk of extubation failure in mechanically-ventilated surgical ICU patients

重症监护病房(ICU)的患者由于各种原因需要插管和机械通气。长时间的机械通气可能会导致严重的并发症，如呼吸机相关性肺炎，可能需要进行气管切开术。然而，过早拔管的患者可能需要在不理想的条件下重新插管，这反过来可能会带来更高的严重并发症风险。这两种情况都可能导致ICU住院时间延长和死亡率增加。确定ICU患者何时准备拔管是一个关键决定，这取决于患者护理团队所有成员之间讨论的许多不同因素。拔管失败的危险因素在不同的ICU人群和个体患者之间可能有很大差异。该项目的目的是开发一种使用机器学习的临床决策支持工具，以预测机械通气外科ICU患者拔管失败的风险。该工具可以引导围绕患者拔管准备情况的跨学科讨论。通过准确识别拔管失败的高风险患者，并提供导致患者个体高风险的参数信息，该工具将支持重症监护团队根据患者的个体风险状况定制护理计划，从而帮助减少拔管失败，同时避免长时间的机械通气。关注外科ICU患者允许整合这一人群特有的危险因素，由于非手术患者中有很大比例的缺失值，因此在混合ICU人群中可能不会考虑这些危险因素。

Patients in the intensive care unit (ICU) require intubation and mechanical ventilation for various reasons. Prolonged mechanical ventilation can result in serious complications such as ventilator-associated pneumonia, and tracheostomy may become necessary. However, patients who are extubated prematurely may need to be reintubated under suboptimal conditions, which in turn may carry a higher risk of serious complications. Both scenarios may lead to extended ICU length of stay and increased mortality. Determining when an ICU patient is ready to be extubated is a critical decision that depends on many different factors discussed among all members of the patient's care team. Risk factors for extubation failure may vary greatly among different ICU populations and individual patients. The aim of the proposed project is to develop a clinical decision support tool using machine learning to predict the risk of extubation failure in mechanically-ventilated surgical ICU patients. This tool may guide the interdisciplinary discussion around a patient’s extubation readiness. By accurately identifying patients at high risk of extubation failure and providing information on the parameters that contribute to the patient’s individual high risk, this tool would support the critical care team in tailoring the care plan to the patient’s individual risk profile, thus helping to reduce extubation failure while avoiding prolonged mechanical ventilation. Focusing on surgical ICU patients allows for the integration of risk factors specific to this population that may not be considered in a mixed ICU population due to a large proportion of missing values in non-surgical patients.