构建Nomogram预测模型探讨慢性阻塞性肺疾病患者并发呼吸衰竭的危险因素

To explore the risk factors for concomitant respiratory failure in the patients with chronic obstructive pulmonary disease based on nomogram prediction model

目的 分析慢性阻塞性肺疾病(COPD)患者并发呼吸衰竭的相关危险因素并构建Nomogram预测模型。方法 回顾性分析2019年4月至2021年8月于永康市中医院进行治疗的200例COPD患者临床资料,根据是否并发呼吸衰竭分为呼吸衰竭组(n=68)和非呼吸衰竭组(n=132)。对比两组患者病历资料,经独立样本t检验及χ^(2)检验分析影响呼吸衰竭发生的相关指标;MedCalc软件对差异有统计学意义的指标进行ROC曲线分析,探讨其对呼吸衰竭发生的预测价值;经Logistic回归分析影响呼吸衰竭发生的独立危险因素;R语言软件4.0“rms”包构建呼吸衰竭发生的Nomogram预测模型,校正及决策曲线对Nomogram预测模型进行内部验证及临床预测效能评估。结果 呼吸衰竭组体重指数(BMI)、尿酸、白蛋白均低于非呼吸衰竭组,血白细胞计数、呼吸困难比例、1年内急性发作次数均高于非呼吸衰竭组(P<0.05)。BMI、尿酸、白蛋白的AUC分别为0.796、0.739、0.835,最佳截断值分别为20.22 kg/m^(2)、275.42μmol/L、30.10 g/L。Nomogram模型预测呼吸衰竭的C-index为0.645(95%CI0.578~0.714),阈值>0.17,Nomogram模型提供临床净收益并高于独立预测因子。结论 基于COPD并发呼吸衰竭的独立危险因素即BMI、尿酸、白蛋白、血白细胞计数、呼吸困难、1年内急性发作次数构建的Nomogram模型,对临床COPD患者并发呼吸衰竭有较好的预测价值。

Objective To analyze the risk factors of respiratory failure in the patients with chronic obstructive pulmonary disease(COPD) and to establish a nomogram prediction model.Methods The clinical data of 200 patients with COPD treated in our hospital from April 2019 to August 2021 were retrospectively analyzed,and the patients were divided into respiratory failure group(n=68) and non-respiratory failure group(n=132) according to whether they were complicated with respiratory failure.The medical records of the two groups were compared,and the related indexes affecting the occurrence of respiratory failure were analyzed by independent sample t test and chi-square test.MedCalc software was used to analyze the ROC curve of the measurement indicators with statistical significance,and their predictive values for the occurrence of respiratory failure were discussed.The independent risk factors of respiratory failure were analyzed by Logistic regression.R language software 4.0 "rms" package was used to construct a nomogram prediction model for respiratory failure,and calibration curves and decision curves were used for internal verification and the evaluation of clinical prediction effectiveness of the nomogram prediction model.Results BMI,uric acid and albumin in the respiratory failure group were lower than those in the non-respiratory failure group.The white blood cell count,the proportion of dyspnea and the number of acute attacks within 1 year in the respiratory failure group were higher than those in the non-respiratory failure group(P<0.05).The AUC of BMI,uric acid and albumin were 0.796,0.739 and 0.835,respectively.The optimal truncation values were 20.22 kg/m^(2),275.42 μmol/L and 30.10 g/L,respectively.The C-index of the nomogram model for predicting respiratory failure was 0.645(95%CI 0.578-0.714),and the threshold was0.17.The nomogram model provided a higher clinical net benefit than independent predictors.Conclusions This study establishes a nomogram model based on the independent risk factors of respiratory failure,BMI,uric acid,albumin,white blood cell count,dyspnea,and the number of acute attacks within 1 year,which has good predictive value for respiratory failure in the patients with COPD.