呼吸道传染性疾病中甲乙流新冠病毒感染患者出现并发症中西医结合预后模型探讨

Integrated Chinese and Western Medicine Prognosis Model of Complications in Patients with Influenza A/B or COVID-19

目的:基于中医真实世界临床数据探索呼吸道传染性疾病中甲乙流新冠病毒感染患者出现并发症的危险因素,构建Cox比例风险回归模型,并运用累积发生函数图形呈现预测输出。方法:收集2022年11月至2023年10月来源于黑龙江中医药大学附属第一医院包括新冠病毒感染、甲(或乙)型流感病毒感染在内的呼吸道传染性疾病患者病例,将其电子病例系统数据整合形成数据仓库并回顾性收集自2022年11月至2023年10月期间,涉及甲型、乙型流感及新冠病毒感染导致的呼吸道疾病患者的案例,涵盖了患者的年龄、性别、疾病进程、既往病史、实验室检测结果、舌象特征、脉搏状况、中医辨证分型及主要治疗用药等全面信息。通过电话回访和查阅再入院记录来获取患者后续是否发生并发症的相关结局指标。数据按照70%和30%的比例被划分为训练数据集和验证数据集。在训练集中,运用了Cox比例风险模型来识别影响患者并发症的关键因素。随后,通过逐步排除法优化变量组合,构造出一个高效的并发症风险评估模型,并以直方图的形式直观展示。模型的性能指标通过C-index、受试者工作特征(ROC)曲线、校准误差图及临床决策效能曲线逐一验证,全面衡量其预测能力。结果:慢性肺部疾病史[比例风险(HR)4.46,95%置信区间(95%CI)1.79~11.12]、病理因素气虚(HR 5.74,95%CI 2.14~15.39)、细弱脉(HR 4.45,95%CI 1.88~10.50)、激素用药史(HR 4.57,95%CI 2.04~10.23)、降钙素原(PCT)>10μg·L^(-1)(HR1.23,95%CI 0.06~0.86)、血清淀粉样蛋白(SAA)>100 mg·L^(-1)(HR 9.80,95%CI 7.24~59.75)、血小板(PLT)>303×109个/L(HR 5.66,95%CI 2.01~16.00)是患者出现并发症的危险因素;中药参与(HR 0.20,95%CI 0.06~0.70)是患者出现并发症保护因素;基于筛选出以上危险因素构建预测模型,在训练集中的C-index估计值为0.765,CI为0.667~0.859,而在验证集中这一指数为0.804,区间为0.773~0.855。进一步描绘了C-index的时间演变图形。在训练数据中,5、10、15个月的ROC曲线下面积(AUC)分别是0.61、0.72和0.79。同样,在验证数据中,相应的AUC为0.60、0.67和0.62。此外,对训练集和验证集都绘制了5、10、15个月的校准图及临床决策曲线,显示出模型具有良好的校准性能,并且在临床实践中具有实用性。结论:慢性肺部疾病史、病理因素气虚、细弱脉、激素用药史、PCT(>10μg·L^(-1))、SAA(>100 mg·L^(-1))、PLT(>303×10^(9)个/L)是呼吸道传染性疾病中甲乙流新冠病毒感染患者发生并发症的危险因素,治疗过程中药参与是患者出现并发症的保护因素,并以此建立临床预测模型。该模型展现出卓越的鉴别力、校准性能及显著的临床实用性,为预防和控制呼吸道病毒性感染引发的并发病症提供了坚实的科学支撑。

Objective:Based on real-world clinical data of traditional Chinese medicine(TCM),a Cox proportional hazards model was built to predict the risk factors of complications in patients with Corona Virus Disease 2019(COVID-19)or influenza A/B,and the cumulative occurrence function graph was used to present the prediction output.Method:The medical records of the patients with respiratory infectious diseases,including COVID-19 and influenza A/B,treated in the First Affiliated Hospital of Heilongjiang University of Chinese Medicine from November 2022 to October 2023 were collected.The data from the electronic medical record system were integrated into a data warehouse.The information of the patients with respiratory diseases caused by influenza A and B viruses and SARS-CoV-2 from November 2022 to October 2023 was retrospectively collected.The information involved age,gender,disease course,past medical history,laboratory test results,tongue manifestation,pulse manifestation,TCM syndrome,and main therapeutic drugs.The outcome indicators of whether complications occurred were obtained by telephone follow-up and review of readmission records.The data was divided into a training set and a validation set in a ratio of 70%and 30%,respectively.In the training set,the Cox proportional hazards model was used to identify the key factors affecting patient complications.Then,the combination of variables was optimized by stepwise elimination method,and an efficient complication risk assessment model was constructed,which was visualized in the form of histogram.The C-index,receiver operating characteristic(ROC)curve,calibration error graph,and decision curve analysis were employed to comprehensively measure the prediction performance of the model.Result:The history of chronic lung diseases[hazard ratio(HR)4.46,95% confidence interval(95%CI)1.79-11.12],Qi deficiency(HR 5.74,95%CI 2.14-15.39),thready and weak pulse(HR 4.45,95%CI 1.88-10.50),hormone use history(HR 4.57,95%CI 2.04-10.23),procalcitonin(PCT>10μg·L^(-1))(HR 1.23,95%CI 0.06-0.86),serum amyloid A(SAA)>100 mg·L^(-1)(HR 9.80,95%CI 7.24-59.75),and platelet(PLT)>303×109/L(HR 5.66,95%CI 2.01-16.00)were the risk factors for complications.Chinese medicine intervention(HR 0.20,95%CI 0.06-0.70)was the protective factor for complications.Based on the above risk factors,the prediction model was constructed.In the training set,the C-index was estimated to be 0.765,and the CI was within the range of 0.667 to 0.859.In the validation set,the C-index was 0.804,and the CI varied within the range of 0.773 to 0.855.The temporal variation graph of C-index was then described.The area under the ROC curve(AUC)at 5,10,15 months was 0.61,0.72,and 0.79 in the training set and 0.60,0.67,and 0.62 in the validation set,respectively.In addition,calibration and decision curves were drawn for 5,10,15 months for both training and validation sets,which showed that the model had good calibration performance and was effective in clinical practice.Conclusion:The history of chronic lung diseases,Qi deficiency,thready and weak pulse,hormone use history,PCT>10μg·L^(-1),SAA>100 mg·L^(-1),and PLT>303×10^(9)/L were risk factors for complications in patients with COVID-19 or influenza A/B,while Chinese medicine intervention was a protective factor.The prediction model was established based on the indicators above.The model showcased excellent distinguishing performance,calibration performance,and clinical practicability,providing scientific support for the prevention and control of complications caused by respiratory viral infections.