Continuous Fall Risk Monitoring System: Walking vs Activities of Daily Living

描述(申请人提供)：跌倒是美国急诊科与伤害相关就诊的主要原因，也是65岁以上人群意外死亡的主要原因1。物理治疗、调整药物或改变行为等干预措施可以降低老年人的跌倒几率。需要确定跌倒风险，以确定谁可能从干预措施中受益。跌倒风险的变化可能突然或逐渐发生，更有可能在家庭环境中变得明显，因为个人在进行正常的日常生活活动(ADL)时，而不是在有限和定期的护理提供者评估期间。因此，需要一个持续的、一体化的系统，监测家中的老年人是否有迹象表明他们更容易摔倒，以减少老年人的跌倒。为了有效，监测设备必须是非侵入性的，并为社会所接受。该项目的长期目标是[扩展现有的非侵入性、商业可用的监测系统，能够进行位置跟踪、生理监测和警报，以实时评估跌倒的风险；该系统将考虑频繁使用卫生间、间歇性睡眠和步态特征变化等因素]。这项拟议研究的显示器具有手表外形因素，可佩戴在手腕上，并已证明老年用户的接受率很高。研究表明，异常步态是跌倒风险的标志，导致在确定跌倒风险时使用了各种步态测量方法。为了从ADL期间收集的数据中分析步态，有必要区分步行的阶段，然后可以分析异常特征。这项第一阶段建议的目的是测试使用从商业手腕监测器收集的三轴加速度数据来识别行走周期的可行性。然后，行走数据可以与其他系统数据结合使用，以]对跌倒风险的变化做出推断。将招募30名老年人(65岁及以上)，居住在独立生活设施中的流动志愿者。志愿者将被要求进行正常的ADL，同时接受4小时的监测。在研究期间，志愿者将被录像，使用手腕设备进行监测，并使用弗吉尼亚大学开发的身体区域传感器网络技术进行监测。该项目的具体目标将是：1)确定是否可以使用手腕收集的加速度数据中通常与步行相关的频率作为区分标准来区分步行周期和其他ADL；2)确定个人是否典型地表现出比针对整个人群建议的步行频率范围更窄的步行频率范围；以及3)确定使用机器学习和时间序列技术来使用从步行和非步行数据中学习的特征来区分步行周期和其他ADL是否可行。 与公共卫生相关：在美国，跌倒是与伤害相关的急诊科就诊的主要原因，也是651岁以上人群意外死亡的主要原因。为了减少跌倒的次数，需要开发一个连续的、非侵入性的、方便的、外形可接受的监测系统，监测家中环境中的老年人是否有迹象表明他们更容易摔倒。这个项目的长期目标是开发一个这样的系统；拟议的项目将采取必要的步骤，识别和区分步行和其他正常活动[以便在步行过程中收集的数据可以用来识别个人稳定性的变化，然后可以与已经实时自动收集的其他系统数据一起使用，以识别在步行或其他活动过程中摔倒的可能性增加。]

DESCRIPTION (provided by applicant): Falls are the leading cause of injury-related visits to U.S. emergency departments and the primary etiology of accidental deaths in persons over the age of 65 years1. Interventions such as physical therapy, adjusting medications, or behavior changes can reduce the elderly fall rate2. Fall risk determination is needed to identify who may benefit from interventions. Changes in fall risk may occur suddenly or gradually, and are more likely to become apparent in the home environment as an individual goes about their normal activities of daily living (ADLs) rather than during a limited and periodic care provider assessment. Consequently, a continuous, all-in-one system that would monitor elderly individuals in the home for signs they are becoming more susceptible to falls is needed to reduce falls in the elderly. To be effective, the monitoring device must be non-intrusive and socially acceptable. The long-term goal of this project is to [extend an existing non-intrusive, commercially available monitoring system capable of location tracking, physiologic monitoring, and alerting to also assess fall risk in real time; the system would consider factors such as frequent bathroom use, fitful sleep and changes in gait characteristics]. The monitor for the proposed research has a watch form factor, is worn at the wrist, and has demonstrated high acceptance rates by elderly users. Research has shown that abnormal gait is indicative of fall risk, leading to the use of a variety of measurements of gait in fall risk determinations. To analyze gait from data gathered during ADLs, it is necessary to differentiate periods of walking, which can then be analyzed for abnormal characteristics. The purpose of this Phase I proposal is to test the [feasibility of using tri-axial acceleration data gathered from a commercially available wrist monitor to recognize periods of walking. Walking data can then be used in conjunction with other system data to] make inferences about changes in fall risk. Thirty elderly (aged 65 and over), ambulatory volunteers residing in an independent living facility will be recruited. The volunteers will be asked to engage in normal ADLs while being monitored over a 4-hour time period. During the study, volunteers will be videotaped, monitored using the wrist device, and monitored using a body-area sensor network technology developed at the University of Virginia. The specific aims of this project will be to: 1) determine if it is feasible to distinguish between periods of walking and other ADLs using the presence of frequencies generally associated with walking in wrist- gathered acceleration data as the differentiator, 2) determine whether individuals typically demonstrate a narrower range of walking frequencies than that suggested for the entire population, and 3) determine if it is feasible to use machine learning and time-series techniques to distinguish between periods of walking and other ADLs using characteristics learned from walking and non-walking data. PUBLIC HEALTH RELEVANCE: Falls are the leading cause of injury-related visits to emergency departments in the United States and the primary cause of accidental deaths in persons over age 651. In order to reduce the number of falls, a continuous, non-intrusive, convenient monitoring system in an acceptable form factor which will monitor elderly individuals in their home environment for signs they are becoming more susceptible to falls needs to be developed. The long-term goal of this project is to develop such a system; the proposed project would take the necessary step of identifying and differentiating between walking and other normal activities [so that data gathered during walking can be used to recognize changes in stability for an individual, and then can be used in conjunction with other system data already being automatically collected in real-time to recognize an increased probability of falling both during walking or other activities.]