Center for Weight, Eating, and Lifestyle Sciences (WELL Center), Drexel University, Stratton Hall, 3141 Chestnut Street Philadelphia, PA, 19104, USA; Department of Psychology, Drexel University, Stratton Hall, 3141 Chestnut Street, Philadelphia, PA, 19104, USA.
Center for Weight, Eating, and Lifestyle Sciences (WELL Center), Drexel University, Stratton Hall, 3141 Chestnut Street Philadelphia, PA, 19104, USA; Department of Psychology, Drexel University, Stratton Hall, 3141 Chestnut Street, Philadelphia, PA, 19104, USA.
Appetite. 2020 Sep 1;152:104698. doi: 10.1016/j.appet.2020.104698. Epub 2020 Apr 9.
Emotion dysregulation is a known risk factor for a variety of maladaptive eating behaviors, including emotional eating (Crockett, Myhre, & Rokke, 2015; Lavender et al., 2015). New passive sensing technologies offer the prospect of detecting emotion dysregulation in real-time through measurement of heart rate variability (HRV), a transdiagnostic bio-signal of emotion regulation, which may in turn signal risk of engaging in a maladaptive eating behavior. In the current study, our primary aim was to test the hypothesis that momentary changes in HRV can be used to detect risk of experiencing an emotional eating episode in an ecologically valid setting using a wrist worn sensor with acceptable classification accuracy. Participants were 21 adults with clinically significant emotional eating behaviors. Participants wore the Empatica E4 wrist-sensor and tracked all emotional eating episodes using ecological momentary assessment for four weeks. Time and frequency domain features of HRV were extracted in the 30-min period preceding emotional eating episodes and control cases (defined as the 30 min prior to an EMA survey that did not contain an emotional eating episode). Support vector machine (SVM) learning models were implemented using time domain and frequency domain features. SVM models using frequency domain features achieved the highest classification accuracy (77.99%), sensitivity (78.75%), and specificity (75.00%), consistent with standards deemed acceptable for the prediction of event-level health behavior. SVM models using time domain features still performed above chance, though were less accurate at classifying episodes (accuracy 63.48%, sensitivity 62.68%, and specificity 70.00%) and did not meet acceptable classification accuracy. Wearable sensors that assess HRV show promise as a tool for capturing risk of engaging in emotional eating episodes.
情绪失调是各种适应不良的进食行为的已知风险因素，包括情绪性进食（Crockett、Myhre 和 Rokke，2015；Lavender 等人，2015）。新的被动感应技术提供了通过测量心率变异性（HRV）实时检测情绪失调的前景，HRV 是情绪调节的一种跨诊断生物信号，这可能反过来表明发生适应不良进食行为的风险。在当前的研究中，我们的主要目的是测试假设，即使用腕戴式传感器，通过测量心率变异性（HRV），以可接受的分类准确性，在生态有效的环境中检测到经历情绪性进食发作的风险。参与者为 21 名具有明显情绪性进食行为的成年人。参与者佩戴 Empatica E4 腕式传感器，并使用生态瞬时评估（EMA）在四周内跟踪所有情绪性进食发作。在情绪性进食发作和对照案例（定义为在不包含情绪性进食发作的 EMA 调查前 30 分钟）前 30 分钟提取 HRV 的时域和频域特征。支持向量机（SVM）学习模型使用时域和频域特征来实现。使用频域特征的 SVM 模型实现了最高的分类准确性（77.99%）、敏感性（78.75%）和特异性（75.00%），与预测事件级健康行为的可接受标准一致。使用时域特征的 SVM 模型仍然表现出高于机会的水平，但在分类发作方面的准确性较低（准确性 63.48%，敏感性 62.68%，特异性 70.00%），且未达到可接受的分类准确性。评估 HRV 的可穿戴传感器有望成为捕捉情绪性进食发作风险的工具。
