Abstract: Background: Cardiovascular metabolic diseases are closely associated with depression. Although the management of cardiovascular metabolic diseases at the community level has been established, psychological issues such as depression in patients have not received sufficient attention. Moreover, there is a lack of simple, accurate, and efficient screening and assessment tools for depression. Objective: To apply single-lead wearable electrocardiographic devices to predict the risk of depression in elderly patients with cardiovascular metabolic diseases at the community level. Methods: A total of 3,121 elderly patients (aged over 65) with hypertension, diabetes, coronary heart disease, and other cardiovascular metabolic diseases were selected from 20 primary medical institutions in Ningxia between January 2022 and June 2023. Electrocardiographic data collected via single-lead wearable electrocardiographic devices were uploaded to a cloud platform. Additionally, sociodemographic, lifestyle, and mental health data were collected from the same platform. The data were divided into a training set (2,341 cases) and a validation set (780 cases) using a simple random sampling method at a 3:1 ratio. LASSO regression analysis and cross-validation were performed using RStudio software to identify the best predictors. A multivariable logistic regression model was then established using the predictors selected by LASSO regression. A nomogram model for predicting the risk of depression in elderly patients with cardiovascular metabolic diseases was constructed. The model's efficacy was evaluated using the Receiver Operating Characteristic (ROC) curve, calibration, and decision curve analysis. Results: In the training set, LASSO regression combined with logistic regression analysis identified several significant factors associated with depression in elderly patients with cardiovascular metabolic diseases: gender (OR=1.747, 95%CI 1.258-2.434), BMI (OR=1.073, 95%CI 1.024-1.125), urban-rural residence (OR=1.684, 95%CI 1.172-2.456), exercise (OR=0.61, 95%CI 0.460-0.799), anxiety (OR=3.041, 95%CI 1.597-5.484), coronary heart disease (OR=2.743, 95%CI 1.971-3.815), premature beats (OR=4.745, 95%CI 1.681-19.977), SDANN (OR=4.745, 95%CI 1.681-19.977), rMSSD (OR=0.986, 95%CI 0.972-0.999), and sleep efficiency (OR=0.988, 95%CI 0.982-0.995). The differences were statistically significant (P<0.05). The optimal model lambda.1se was selected, with an optimal lambda value of 0.00828878. The AUC, sensitivity, and specificity for predicting the risk of depression in the training and validation sets were 0.748 (0.707-0.786, P<0.001), 75.2%, 63.4% and 0.751 (0.692-0.809), 76.7%, 60.6%, respectively. The clinical decision curve analysis showed that when the probability threshold for depression risk was between 8% and 35% in the training set and between 8% and 37% in the validation set, the net benefit of predicting the risk of depression in elderly patients with cardiovascular metabolic diseases was higher. Conclusion: This study successfully constructed a nomogram model for predicting the risk of depression in elderly patients with cardiovascular metabolic diseases at the community level, based on single-lead wearable electrocardiographic devices. The model demonstrated good predictive efficacy and clinical application value. It can assist primary medical institutions in conducting depression screening and formulating individualized intervention measures for patients, thereby aiding in the prevention and control of cardiovascular diseases at the community level.

Key words: Community level Elderly Cardiovascular Metabolic Diseases, Depression, Single-Lead Wearable Electrocardiographic Device, Nomogram, Prediction Model

摘要： 背景 心血管代谢疾病与抑郁密切相关。基层心血管代谢疾病管理工作已初具规模，但对于患者抑郁等心理问题一方面缺乏关注，另一方面缺乏简单、准确、高效的筛查评估工具。目的 应用单导联可穿戴心电设备预测基层老年心血管代谢疾病患者抑郁发生风险。方法 选取 2022年1月—2023年6月宁夏20家基层医疗机构应用单导联可穿戴心电设备采集并上传至云平台的3121例65岁以上高血压、糖尿病、冠心病等心血管代谢疾病患者的心电图资料及云平台收集的患者社会人口学、生活行为方式、心理健康等资料。采用简单随机抽样方法按3:1比例分为训练集(2341例)和验证集(780例)。通过运行RStudio软件,使用LASSO回归分析以及交叉验证,来筛选最佳预测因子,采用多因素 logistic 回归建立预测模型,引入从LASSO 回归筛选出的预测因子,进而构建预测老年心血管代谢疾病患者抑郁风险列线图模型,采用受试者工作特征（ROC）曲线、校正及决策曲线评估模型的效能。结果 在训练集中,LASSO结合logistic 回归分析结果显示,性别(OR=1.747,95%CI 1.258-2.434)、BMI(OR=1.073,95%CI 1.024-1.125)、城乡(OR=1.684,95%CI 1.172-2.456)、运动(OR=0.61,95%CI 0.460-0.799)、焦虑(OR=3.041，95%CI 1.597-5.484)、冠心病(OR=2.743.95%CI 1.971-3.815)、早搏(OR=4.745.95%CI 1.681-19.977)、SDANN(OR=4.745.95%CI 1.681-19.977)、rMSSD(OR=0.986.95%CI 0.972-0.999)、睡眠效率(OR=0.988.95%CI0.982-0.995)是老年心血管代谢疾病合并抑郁的影响因素（P<0.05）。选择最佳模型lambda.1se，最佳lambda值为0.00828878。训练集和验证集预测老年慢病患者抑郁发生风险的AUC、灵敏度、特异度分别为 0.748(0.707-0.786,P<0.001))、75.2%、63.4%；0.751(0.692-0.809) 、76.7%、60.6%。临床决策曲线显示,在训练集和验证集中当抑郁风险闽值概率分别在8%到35%、8%到37%之间时,预测老年心血管代谢疾病患者抑郁发病风险的净收益更高。结论　本研究基于单导联可穿戴心电设备构建的基层老年心血管代谢疾病发生抑郁风险的列线图模型，有较好预测效能及临床应用价值，不但有助于在基层医疗机构对患者进行抑郁筛查与个体化干预措施的制定、还可助力基层心血管疾病防控工作。

关键词: 基层老年心血管代谢疾病, 抑郁, 单导联可穿戴心电设备, 列线图, 预测模型