Tracking Fatigue and Health State in Multiple Sclerosis Patients Using Connnected Wellness Devices
Tracking Fatigue and Health State in Multiple Sclerosis Patients Using Connnected Wellness Devices
复制标题
使用互联健康设备跟踪多发性硬化症患者的疲劳和健康状况
DOI：
10.1145/3351264
复制
发表时间：
2019
影响因子：
--
通讯作者：
Tong C
中科院分区：
文献类型：
--
作者：
Tong C
多发性硬化症需要长期的疾病管理，但通过使用临床调查来跟踪患者受到高成本和患者负担的阻碍。在这项工作中，我们调查了使用无处不在的感知数据来预测MS患者疲劳和健康状态的可行性，通过疲劳严重程度量表(FSS)和EQ-5D指数来衡量。我们收集了198名MS患者的数据，这些患者使用联网健康设备超过6个月。我们使用一组回归变量检验收集的数据如何准确地预测每个患者报告的FSS和EQ-5D得分。在对FSS和EQ-5D进行预测时，我们能够获得与仪器标准测量误差(SEM)一致的误差，以及预测值和地面真实值之间强大而显著的相关性。我们还展示了一种简单的自适应方法，通过使用仅使用用户提供的1个基本事实数据点，大大减少了预测误差。对于FSS(SEM0.7)，通用模型预测每周的MAE为1.00，而改进的模型预测MAE为0.58。对于EQ-5D(结构方程0.093)，通用模型预测MAE为0.097，而修正模型预测MAE为0.065。我们的研究代表了第一组结果，表明MS患者的疲劳和健康状态可以使用连接的健康设备和少量背景特征的数据来测量，在广泛使用的临床验证问卷中具有良好的预测性能，误差在可接受的误差范围内。我们结果的未来扩展和潜在应用将对多发性硬化症患者的疾病管理和支持临床研究产生积极影响。
Multiple Sclerosis requires long-term disease management, but tracking patients through the use of clinical surveys is hindered by high costs and patient burden. In this work, we investigate the feasibility of using data from ubiquitous sensing to predict MS patients' fatigue and health status, as measured by the Fatigue Severity Scale (FSS) and EQ-5D index. We collected data from 198 MS patients who are given connected wellness devices for over 6 months. We examine how accurately can the collected data predict reported FSS and EQ-5D scores per patient using an ensemble of regressors. In predicting for both FSS and EQ-5D, we are able to achieve errors aligning with the instrument' standard measurement error (SEM), as well as strong and significant correlations between predicted and ground truth values. We also show a simple adaptation method that greatly reduces prediction errors through the use of just 1 user-supplied ground truth datapoint. For FSS (SEM 0.7), the universal model predicts weekly scores with MAE 1.00, while an adapted model predicts with MAE 0.58. For EQ-5D (SEM 0.093), the universal model predicts weekly scores with MAE 0.097, while an adapted model predicts with MAE 0.065. Our study represents the first sets of results showing that fatigue and health state of MS patients can be measured using data from connected wellness devices and a small number of background features, with promising prediction performance with errors within the accepted range of error in the widely used clinically-validated questionnaires. Future extensions and potential applications of our results can positively impact MS patient disease management and support clinical research.
登录
查看更多内容
影响因子：
11
作者：
Kobelt, G.;Berg, J.;Jonsson, B.
通讯作者：
Jonsson, B.
影响因子：
9.9
作者：
KRUPP, LB;COYLE, PK;GRIMSON, R
通讯作者：
GRIMSON, R
影响因子：
4.7
作者：
Schreurs, KMG;de Ridder, DTD;Bensing, JM
通讯作者：
Bensing, JM
影响因子：
5.8
作者：
Fogarty, Emer;Walsh, Cathal;Tubridy, Niall
通讯作者：
Tubridy, Niall
影响因子：
2
作者：
Pereira T;Correia C;Cardoso J
通讯作者：
Cardoso J