Utilizing high resolution physiological data and artificial intelligence to develop a pediatric cardiac arrest prediction tool for integration into bedside clinical practice

重症儿童有心肺骤停（CPA）的风险。在加拿大儿科重症监护病房（PICU）中，这些毁灭性事件每年发生300-700次，并与高死亡率（45 - 85%）、残疾和更高的护理成本相关。CPA的有效管理需要临床医生立即做出反应。然而，在CPA之前，患者状态的变化是微妙的。如果这些都得到赞赏，干预措施将是可能的，有可能防止CPA和相关的不良结果和成本。该项目的目标是使用人工智能开发和实施一种工具来预测重症儿童的CPA。自2013年4月以来，我们一直在PICU收集和存储连续的生理信号。我们已经开展了初步工作，使用机器学习来分析CPA患者的这些数据。我们的研究结果表明，CPA的早期预测是可能的。我们将使用这笔资金：1）进一步验证，培训和测试预测CPA的模型，利用包括临床医生，工程师，计算机科学家在内的多学科专业知识，2）使用可视化平台和与人为因素工程师和伦理学家开发的实施计划将模型翻译到床边。该项目是预测，个性化医学的新领域的一部分，并与新的伦理困境有关的沟通预测的重要性，病人和他们的家人，我们将探讨和定义。一个准确的CPA模型将挽救加拿大重症儿童的生命，并减少与这些事件相关的负担。它的发展将为加拿大研究人员扩展预测建模和重新定义医疗团队创造机会。这一举措将降低医疗成本，并使加拿大成为使用人工智能解决关键医疗问题的领导者。

Critically ill children are at risk of cardio-pulmonary arrest (CPA). These devastating events occur 300-700 times per year in Canadian pediatric intensive care units (PICUs) and are associated with high mortality (45 to 85%), disability and higher costs of care. Effective management of CPA requires immediate reaction from clinicians. However, changes in apatients state that precede the CPA are subtle. If these were appreciated, interventions would be possible, potentially preventing CPA and the associated poor outcomes and costs. The objective of this project is to use artificial intelligence to develop and implement a tool to predict CPA in critically ill children. We have been collecting and storing continuous physiologic signals in our PICU since April 2013. We have undertaken preliminary work using machine learning to analyze these data in patients who suffered a CPA. Our results indicate that earlier prediction of CPA is possible.We will use funding from this grant to: 1) Further validate, train and test the model predicting CPA, utilizing multidisciplinary expertise in a team that includes clinicians, engineers, computer scientists, and 2) Translate the model to the bedside using a visualization platform and implementation plan developed with human factors engineers and ethicists. This project is part of the new field of predictive, personalized medicine, and is associated with novel ethical dilemmas related to communicating the significance of predictions to patients and their families, which we will explore and define. An accurate model for CPA will save the lives of critically ill Canadian children and reduce the burden associated with these events. Its development will create opportunities for Canadian researchers to extend predictive modelling and redefine the healthcare team. This initiativewill reduce healthcare costs, and positions Canada as a leader in using artificial intelligence to solve crucial healthcare problems.