耐碳氢酶烯肺炎克雷伯杆菌肺部感染后进展为急性呼吸窘迫综合症的临床特征及Nomogram预测模型构建

Clinical Features and Nomogram Prediction Model Construction for Progression to Acute Respiratory Distress Syndrome after Carbapenem-Resistant Klebsiella pneumoniae (CRKP) Lung Infection

摘要: 目的：分析耐碳青霉烯类肺炎克雷伯杆菌(Carbapenem-resistant Klebsiella pneumoniae, CRKP)肺部感染后进展为急性呼吸窘迫综合症(Acute respiratory distress syndrome, ARDS)的临床特征以及构建Nomogram模型。方法：回顾性收集合肥市第二人民医院2019年01月~2022年12月所有肺部感染后肺泡灌洗液分离出CRKP的患者共162例，其中有71例患者CRKP肺部感染后进展为ARDS，此为ARDS组；91例患者CRKP肺部感染后未进展为ARDS，此为非ARDS组。使用SPASS 26.0软件对于收集的患者的临床资料进行单因素以及Logistic多因素分析，受试者工作特征(receiver operator char-acteristic, ROC)曲线分析各指标诊断耐CRKP肺部感染后进展为ARDS的危险因素最佳截断值以及曲线下面积(area under the ROC curve, AUC)。并以此为基础应用R软件“rms”包构建其Nomogram模型，校正曲线对Nomogram模型进行内部验证，应用R软件“rmda”包构建决策曲线，并评估Nomogram模型的预测效能。P < 0.05为差异有统计学意义。结果：单因素分析提示，与非ARDS组相比，ARDS组患者的年龄、高血压病史、吸烟史、慢性阻塞性肺疾病(COPD)病史、查尔森共病指数评分(Charlson comorbidity index, CCI)、序贯器官衰竭评分(Sequential Organ Failure Assessment, SOFA)、肺炎严重程度评分(Pneumonia Severity Index, PSI)、入院第3天CRP计数、CRP/白蛋白比值(第3天)等10项指标差异均有统计学意义(P < 0.05)；多因素Logistic回归分析显示，与非ARDS组相比，ARDS组患者的年龄[OR = 1.307, 95% CI (1.005~1.069)]、SOFA评分[OR = 1.376, 95% CI (1.176~1.610)]、CCI评分[OR = 1.268, 95% CI (1.067~1.507)]具有统计学差异(P < 0.05)，是CRKP肺部感染后进展为ARDS的独立危险因素；将两组中有统计学意义的连续变量进行ROC曲线分析可知，患者年龄、SOFA评分、CCI评分的AUC分别为0.641、0.710、0.669；最佳截断值分别为67.5岁、1.5分、3.5分。Nomogram模型校正曲线及临床净收益分析：内部验证结果显示预测肺部感染CRKP后进展为ARDS的风险C-index为0.728 (95% CI: 0.656~0.801)，校正C-index为0.717。校准曲线显示观测值与预测值之间一致性较好。决策曲线结果显示，当风险阈值波动在0.367~0.567时，Nomogram模型提供临床净收益；此外，Nomogram模型临床净收益均高于年龄、SOFA评分、CCI评分。结论：年龄(>67.5岁)、SOFA (>1.5分)、CCI (>3.5分)是CRKP肺部感染后发生ARDS的独立危险因素(P < 0.05)。本研究基于此构建的Nomogram模型对于肺部感染肺炎克雷伯杆菌后进展为ARDS的早期诊断、早期干预提供了重要的指导意义。

Abstract: Objective: To analyze the clinical features of progression to acute respiratory distress syndrome (ARDS) after Carbapenem-resistant Klebsiella pneumoniae (CRKP) lung infection and Nomogram model was constructed. Methods: A total of 162 patients with CRKP isolated from alveolar lavage fluid after all lung infections were retrospectively collected in the Second People’s Hospital of Hefei City from January 2019 to December 2022, of which 71 patients progressed to ARDS after CRKP lung infection, which is the ARDS group, and 91 patients did not progress to ARDS after CRKP lung infection, which is the non-ARDS group. Using SPASS 26.0 software, the clinical data collected from the patients were analyzed by single-factor and logistic multifactorial analysis, and the receiver op-erator characteristic (ROC) curves were analyzed to determine the optimal cut-off value of the risk factors for progression to ARDS after diagnosis of CRKP-resistant lung infections as well as the area under the ROC curve (AUC). The Nomogram model was constructed using the R software “rms” package, the calibration curves were used for internal validation of the Nomogram model, and the decision curve was constructed using the R software “rmda” package, and the predictive efficiency of the Nomogram model was evaluated. And if P < 0.05, the difference between the Nomogram models was considered to be statistically significant. Results: Univariate analysis suggested that compared with the non-ARDS group, patients in the ARDS group differed in 10 indices, including age, history of hypertension, smoking history, history of chronic obstructive pulmonary disease (COPD), Charlson comorbidity index (CCI), Sequential Organ Failure Assessment (SOFA), Pneumonia Severity Index (PSI), CRP count on day 3 of admission, CRP/albumin ratio (day 3) and other indica-tors, were statistically significant (P < 0.05); multifactorial logistic regression analysis showed that, compared with the non-ARDS group, the age [OR = 1.307, 95% CI (1.005~1.069)], SOFA score [OR = 1.376, 95% CI (1.176~1.610)], and CCI score [OR = 1.268, 95% CI (1.067~1.507)] of the patients in the ARDS group showed statistically significant differences (P < 0.05), and were independent risk factor for progression to ARDS after CRKP lung infection; ROC curve analysis of continuous variables with statistical significance in the two groups showed that the AUCs of patient age, SOFA score, and CCI score were 0.641, 0.710, and 0.669, respectively; and the optimal cutoff values were 67.5 years old, 1.5 points, and 3.5 points, respectively. Nomogram modeling Calibration curve and net clinical benefit analysis: internal validation results showed a predicted risk C-index of 0.728 (95% CI: 0.656~0.801) and a calibrated C-index of 0.717 for progression to ARDS after CRKP of pulmonary infection. The calibration curves showed good agreement between observed and predicted values. Decision curve results showed that the Nomogram model provided a net clinical benefit when the risk threshold fluctuated from 0.367~0.567; in addition, the Nomogram model net clinical benefit was higher than age, SOFA score, and CCI score. Conclusion: Age (>67.5 years), SOFA (>1.5 points), and CCI (>3.5 points) were independent risk factors (P < 0.05) for the development of ARDS after CRKP lung infection. The Nomogram model constructed in this study based on this provides an im-portant guideline for early diagnosis and early intervention for progression to ARDS after pulmo-nary infection with Klebsiella pneumoniae.