Using artificial intelligence to identify and predict delirium among hospitalized medical patients

谵妄是一种急性神经认知障碍，影响多达一半的老年住院患者。它可能导致混乱，定向障碍，幻觉和激动，导致患者及其护理人员的显着痛苦。它可能导致痴呆症，住院时间延长，医疗费用增加和死亡。虽然谵妄可以预防和治疗，但很难识别和预测。目前的方法需要进行临床评估或使用复杂的筛选工具。这对大多数医院来说是昂贵的，而且往往不切实际。鉴于这些挑战，谵妄报告依赖于行政数据，这大大低估了谵妄的真实率。从护理质量的角度来看，这种不准确性使得难以比较各医院的谵妄发生率，并评估干预措施对改善谵妄护理的影响。在这个项目中，我们将利用人工智能开发两种自动化工具，这些工具将使用从电子病历中提取的常规数据来识别谵妄病例并预测患者谵妄风险。由于人们担心人工智能应用可能会恶化健康不平等，因为它们没有考虑社会和经济变量，我们将测试我们的工具在检测和预测谵妄方面是否存在偏见。我们的方法在谵妄护理和更普遍的医疗质量改善方面是新颖的，并且可以改变质量指标的测量方式。此外，它有可能改善患者的预后并降低医疗成本。与已经组装了一个多学科的研究团队，从计算机，健康和社会科学，并与健康质量安大略（HQO），在医疗保健质量的省级顾问合作。HQO积极参与规划该项目的方法和成果，以确保它们可用于改善系统和个体患者层面的医疗保健。

Delirium is an acute neurocognitive disorder which affects up to half of older hospitalized medical patients. It can cause confusion, disorientation, hallucinations, and agitation, leading to significant distress for patients and their caregivers. It can lead to dementia, longer hospital stays, increased health costs, and death. While delirium can be prevented and treated, it isdifficult to identify and predict. Current methods require either ongoing clinical assessment or use of complicated screening tools. This is expensive and often impractical for most hospitals. Given these challenges, delirium reporting relies on administrative data, which significantly underestimate true rates of delirium. From a quality of care perspective, this inaccuracymakes it difficult to compare rates of delirium across hospitals and evaluate the impact of interventions to improve delirium care. In this project, we will utilize artificial intelligence to develop two automated tools, which will use routine data extracted from electronic medical records to identify delirium cases and predict patient delirium risk. As there is concern that AIapplications may worsen health inequity because they do not take into consideration social and economic variables, we will test whether there is bias in the detection and prediction of delirium with our tools. Our approach is novel in delirium care and healthcare quality improvement more generally, and could change how quality indicators are measured. Moreover, it has the potential to improve patient outcomes and reduce healthcare costs. With has assembled a multidisciplinary research team from computer, health, and social sciences, and are partnering with Health Quality Ontario (HQO), a provincial advisor on quality in health care. HQO are an active participant in the planning the methods and outcomes for the project to ensure that they can be used to improve health care at both a systems and individual patient level.