Assisted living communities- transforming predictive data into proactive care for COVID-19

摘要 最近的数据表明，接触COVID-19的美国老年人住院的风险最大， 可怜的结果。此外，由于高龄和他们患有多种慢性疾病的可能性很高， 老年生活设施中的成年人患COVID-19的风险最高，这是其最严重的并发症， 死了自西雅图发现美国首例新型冠状病毒2019疾病（COVID-19）病例以来， 华盛顿，在长期护理和辅助生活设施中发现了几起疫情， 迅速蔓延。老年居民和长期护理辅助生活设施的工作人员以及公共卫生 官员们面临着众多挑战，导致早期发现COVID-19感染变得困难， 这些设施对控制感染蔓延的努力构成了重大障碍。增加 在这些挑战中，超过一半的COVID-19检测结果呈阳性的居民当时没有症状 测试，进一步促进传播。迫切需要制定监测战略， 长期护理和辅助生活设施的居民，以便及早发现感染 使用需要最少人与人接触的手段。 虽然COVID-19感染传播的动态正在通过几个接触者追踪应用程序来解决， 评估这些社区居民出现严重症状和住院的风险 需要积极的生理监测和生态瞬时评估的背景下，预先存在的 临床条件，并提出了一个迫切的未满足的需求。在这个项目中，我们建议提供一个用户- 友好的COVID-19早期检测警报平台（COVID-Alert），集成了：1）生物传感器集成， 非侵入性和连续地监测和记录关键的生命体征（体温、心率、呼吸率， 氧饱和度和活动水平）; 2）使用5个问题集的生态瞬时评估（EMA） 由CDC发布，并在美国被医疗保健提供者和健康保险行业采用; 3）人工 智能框架，其基于实时生理生物感测数据馈送的合成来触发警报， EMA对症状进行监测，并根据电子数据对既存疾病进行个性化风险评估 该设施保存的健康记录。 将COVID-19临床决策支持纳入长期护理机构的工作流程将确保 居民得到适当和及时的护理（居民一级）和持续的监测，以防止爆发 （设施一级），同时避免不必要的工作人员接触。这项研究汇集了强大的跨学科 工程、信息学、数据科学、机器学习和CDS领域的专家团队。先进的数据驱动 预测模型将使用高维EHR数据和临床医生反馈进行训练和验证。的 将密切监测和评估算法开发和临床实施的过程 通过形成性和总结性评估。

ABSTRACT Recent data suggests that that older Americans who contact COVID-19 are at greatest risk for hospitalization and poor outcomes. Additionally, due to advanced age and their high likelihood of having multiple chronic conditions, adults in senior living facilities are at highest risk for developing COVID-19, its most serious complications, and dying. Since the identification of first US case of novel coronavirus 2019 disease (COVID-19) in the Seattle, Washington, several outbreaks have been identified in long-term care and assisted living facilities with evidence of rapid spread. Older residents and the staff of long-term care assisted living facilities as well as public health officials are facing a multitude of challenges which render early detection of COVID-19 infections difficult in these facilities and which have posed a major barrier to the efforts to control the spread of infection. Adding to these challenges, more than half of residents with positive COVID-19 test results are asymptomatic at the time of testing, further contributing to transmission. There is an urgent unmet need for strategies for monitoring of residents in long-term care and assisted living facilities to facilitate early detection of the infection using means that require minimal person-to-person contact. While the dynamics of COVID-19 infection spread is being addressed by several contact tracing apps, assessing the risk for development of severe symptoms and hospitalization in these community residents requires active physiological monitoring and ecological momentary assessment in the context of preexisting clinical conditions and presents an immediate unmet need. With this project, we propose to deliver a user- friendly COVID-19 early detection alert platform (COVID-Alert) that integrates: 1) biosensor ensemble that noninvasively and continuously monitor and record critical vital signs (temperature, heart rate, respiratory rate, oxygen saturation, and activity level); 2) ecological momentary assessment (EMA) using the 5-question set released by CDC and adopted across US by healthcare providers and health insurance industry; 3) artificial intelligence framework that triggers an alert based on synthesis of real-time physiological biosensing data feed, EMA monitoring of symptoms, with personalized risk profiles of preexisting conditions derived from electronic health record maintained by the facility. COVID-19 clinical decision support integrated into the workflow of long-term care facilities will ensure that residents receive appropriate and timely care (resident level) and ongoing surveillance to prevent an outbreak (facility level) while avoiding unnecessary staff exposure. This study brings together a strong interdisciplinary team of experts in engineering, informatics, data science, machine learning, and CDS. The advanced data-driven predictive model will be trained and validated using both high-dimensional EHR data and clinician feedback. The process of the algorithm development and clinical implementation will be closely monitored and evaluated through formative and summative evaluation.