AECOPD患者发生低钠血症风险列线图预测模型的构建

Construction of a nomogram predictive model for hyponatremia in AECOPD patients

目的 调查分析慢性阻塞性肺疾病急性加重期(AECOPD)患者低钠血症的发生率和危险因素,并构建风险列线图预测模型以指导临床实践。方法 回顾性总结2020年1月1日—2023年7月31日张家港市第一人民医院诊断为AECOPD的患者401例为研究对象,住院期间将其分为低钠血症组95例和正常血钠组306例。比较两组的临床资料和血生化,采用LASSO回归模型和多因素Logistic回归分析危险因素,并构建列线图。受试者工作特征(ROC)曲线评估预测能力,Bootstrap法计算模型一致性指数(C-index)评估区分度;校准曲线评估校准度。结果 单因素比较发现,两组吸烟率、胸腔积液和社区获得性肺炎(CAP)、发热、白细胞计数、中性粒细胞和淋巴细胞百分比、中性粒细胞与淋巴细胞比值(NLR)、降钙素原(PCT)、红细胞沉降率(ESR)、pH值、阴离子间隙(AG)、血钙和血镁、白蛋白存在显著差异(P<0.05)。LASSO回归和Logistic回归显示,CAP(OR=2.790,95%CI=2.101~3.236,P<0.001)和ESR(OR=1.266,95%CI=1.001~1.524,P<0.001)是AECOPD患者发生低钠血症的危险因素,而AG(OR=0.444,95%CI=0.121~0.768,P<0.001)和血镁(OR=0.373,95%CI=0.096~0.865,P<0.001)是保护因素。构建的列线图模型总分为200分。ROC显示,列线图预测低钠血症的曲线下面积(AUC)为0.886(95%CI=0.823~0.936,P<0.001)。内部验证C-index为0.924,校准曲线显示一致性良好。结论 AECOPD患者有较高的低钠血症发生率,CAP、ESR、AG和血镁与低钠血症的发生密切相关。构建的列线图模型对指导临床早期、准确筛选低钠血症的高危患者有较好的应用潜能。

Objective To investigate and analyze the incidence and risk factors of hyponatremia in AECOPD patients,so as to construct a risk nomogram predictive model to guide clinical practice.Methods A retrospective summary was conducted on 401 AECOPD patients diagnosed in our hospital from January 1,2020 to July 31,2023.During hospitalization,they were divided into a hyponatremia group of 95 cases and a normal blood sodium group of 306 cases.The clinical data and blood biochemistry of two groups were compared,the LASSO regression model and ten fold cross validation method were used to obtain the optimal subset of risk predictors for hyponatremia.The risk predictors were determined using a multivariate logistic regression model.Based on the risk prediction factors,apply R software to construct a column chart prediction model.Evaluate the predictive ability of the column chart prediction model using ROC curve.Using the Bootstrap method for intra model validation,calculate the consistency index(C-index) to evaluate model differentiation;Draw calibration curves to evaluate the calibration of the model.Results Univariate comparison revealed significant differences in smoking rates,pleural effusion and community acquired pneumonia(CAP),fever,white blood cell count,percentage of neutrophils and lymphocytes,neutrophil to lymphocyte ratio(NLR),procalcitonin(PCT),erythrocyte sedimentation rate(ESR),pH value,anion gap(AG),blood calcium and magnesium,and albumin between the two groups(P<0.05).LASSO regression and logistic regression screening showed that CAP(OR=2.790,95% CI=2.101-3.236,P<0.001) and ESR(OR=1.266,95% CI=1.001-1.524,P<0.001) were risk factors for developing hyponatremia in AECOPD patients,while AG(OR=0.444,95% CI=0.121-0.768,P<0.001) and blood magnesium(OR=0.373,95% CI=0.096-0.865,P<0.001) were protective factors.The total score of the constructed column chart model was 200 points.ROC showed AUC predicted by the nomogram for hyponatremia was 0.886(95% CI=0.823-0.936,P<0.001).The internal validation C-index was 0.924,and the calibration curve showed good consistency between the predicted results of the model and the actual incidence rate.Conclusion AECOPD patients have a higher incidence of hyponatremia,and CAP,ESR,AG,and blood magnesium are closely related to the occurrence of hyponatremia.The constructed column chart model has good application potential for guiding early and accurate screening of high-risk patients with hyponatremia in clinical practice.