慢性乙型病毒性肝炎患者合并抑郁的COX风险预测模型

背景　慢性乙型病毒性肝炎是我国流行最广、危害最重的传染病，抑郁是其常见并发症，轻者情绪低落，重者不仅可加重固有疾病、甚至可出现轻生行为，目前常用治疗措施疗效不理想，积极预防护理成为医护人员关注的重点。目的　构建慢性乙型病毒性肝炎患者合并抑郁的COX风险预测模型。方法　选择2017年6月-2018年5月川北医学院附属医院收治的慢性乙型病毒性肝炎患者245例为研究对象，随访慢性乙型病毒性肝炎患者合并抑郁的情况。完成数据预处理后，所有因素均进入单因素及多元COX风险因素分析，并构建风险预测模型，采用列线图展示预测模型，受试者工作特征（ROC）曲线评价模型区分度，采用calibration plot曲线评价模型准确度，采用临床决策曲线（DCA）评价模型的有效性。结果　单因素分析结果显示不同年龄、职业、学历、乙肝分度、感染时间、确诊时间、复发次数、家庭地位、婚姻满意度、担心疾病难以根治、担心住院环境、疾病分期、有无并发症、对治疗是否有信心慢性乙型病毒性肝炎患者的抑郁发生率比较，差异有统计学意义（P<0.05）。多元COX风险因素分析结果显示：RR=-1.446 1×（职业为知识分子）-0.688 7×（学历高中或中专）-2.043 0×（经常饮酒）-0.783 5×（偶尔吸烟）-1.068 2×（经常吸烟）-0.894 0×（确诊时间0.5~5年）-1.092 4×（确诊时间<0.5年）+1.335 2×（家庭地位不满意）+1.345 1×（婚姻不满意）-0.574 3×（不担心住院环境不适宜）。本研究构建的COX风险预测模型的ROC曲线下面积（AUC）为0.979 8，模型预测特异度0.972 5，灵敏度0.940 7，准确度为0.954 9，阳性似然比为34.180 2，阴性似然比为0.060 9，诊断价值比为560.916 7，阳性预测值为0.976 9，阴性预测值为0.929 8，模型的区分度较高。在模型准确度评价上：当事件发生率在16%以下时，模型高估风险；当事件发生率在16%~40%时，模型低估风险；当事件发生率在40%~80%时，模型高估风险；当事件发生率在80%~100%时，模型低估风险；而在16%、40%、80%时候，模型预测和观察值完全一致，整体上看本模型构建的准确度较好。临床决策曲线显示模型的净获益（NB）值较高，提示基于本模型预测结果开展临床决策产生的效果能给患者病情带来较好的获益值。结论　本次构建的慢性乙型病毒性肝炎合并抑郁的风险预测模型可用于预测新诊断慢性乙型病毒性肝炎合并抑郁的风险，从而指导临床医护人员进行针对性的干预措施，最终避免或降低患者合并抑郁的可能性，值得临床推广应用。

The COX Risk Prediction Model for Depression in Patients with Chronic Hepatitis B

Background　Chronic hepatitis B is the most widespread and most harmful infectious disease in China，often accompanied by depression. Patients with mild depression have a depressed mood，and those with severe symptoms have not only aggravated inherent diseases，but may even commit suicide. At present，the commonly used treatment measures have unsatisfactory effects，and active preventive nursing has become the focus among medical staff. Objective　To construct a Cox risk prediction model for depression in patients with chronic hepatitis B. Methods　A total of 245 patients with chronic hepatitis B admitted to the Affiliated Hospital of North Sichuan Medical College were selected as research subjects from June 2017 to May 2018，and their depression situation were followed up. After the data pre-processing was completed，all factors were conducted univariate analysis and multivariate COX risk factor analysis，and a risk prediction model was constructed. The nomogram was used to show the prediction model，and the receiver operating characteristic（ROC） curve was used to evaluate the model discrimination. The calibration curve was used to evaluate the accuracy of the model，and the clinical decision curve analysis was used to evaluate the validity of the model. Results　The results of univariate analysis showed that there were significant differences in the incidence of depression among patients with different ages，occupations，education background，hepatitis B grades，infection time，diagnosis time，recurrence times，family status，marriage satisfaction，fear of difficult to cure the disease，fear of hospitalization environment，disease stage，complications，and confidence in treatment（P<0.05）. The results of multivariate COX risk factor analysis showed that RR=-1.446 1×（the occupation is intellectual）-0.688 7×（education background was high school or technical secondary school）-2.043 0×（frequent drinking）-0.783 5×（occasional smoking）-1.068 2×（frequent smoking）-0.894 0×（diagnosis time was 0.5-5 years）-1.092 4×（diagnosis time was less than 0.5 year）+1.335 2×（dissatisfaction with family status）+1.345 1×（dissatisfaction with marriage）-0.574 3×（not worried about unsuitable hospitalization environment）. The area under the ROC curve（AUC） of the COX risk prediction model constructed in this study was 0.979 8. The specificity of the model was 0.972 5，the sensitivity was 0.940 7，and the accuracy was 0.954 9. The model had a positive likelihood ratio of 34.180 2，and the negative likelihood ratio was 0.060 9. The diagnostic value ratio was 560.916 7. The positive predictive value was 0.976 9 and the negative predictive value was 0.929 8. The model had great discrimination.In terms of model accuracy evaluation，when the event rate was below 16%，the model overestimated the risk；when the event rate was between 16% and 40%，the model underestimated the risk；when the event rate was between 40% and 80%，the model overestimated the risk；when the event rate was between 80% and 100%，the model underestimated the risk；when the event rate was 16%，40% and 80%，the predicted value and observed value were exactly the same. On the whole，this model had good accuracy. The clinical decision curve showed that the net benefit value was high，which showed the effect of clinical decision-making based on the prediction results of this model could bring great benefit to patient's condition. Conclusion　The risk prediction model for depression in patients with chronic hepatitis B constructed in this paper can be used to predict the risk of depression in patients with chronic hepatitis B so as to guide the clinical medical staff to carry out targeted intervention measures，and ultimately avoid or reduce the possibility of depression，which is worthy of clinical application.