临床资料联合CT建模预测COVID-19普通型肺炎向重型转化

Development of a model for predicting the progress from common to severe COVID-19 pneumonia using clinical characteristics and CT imaging features

目的:探讨新型冠状病毒肺炎(COVID-19)由普通型向重症型转化相关危险因素并建立有效预测模型,提高对新型冠状病毒肺炎认识和诊治水平。方法:分析153例COVID-19患者入院临床及CT资料,依据第7版新型NCP诊疗方案将其分为普通型组及重症型组,普通型肺炎组101例;普通型肺炎转重型组52例。通过CT表现、临床症状、基础病、血常规、肝功及凝血功能行综合分析,运用logistic回归建模并绘制ROC曲线。结果:临床资料中男性、高龄、呼吸困难、纳差、高血压、糖尿病、心血管疾病、淋巴细胞、C反应蛋白、白细胞及中性粒细胞计数、白蛋白及纤维蛋白原计数两组间比较有统计学意义(P<0.05);CT表现肺叶受累数量、病灶形态、邻近胸膜情况两组间比较有统计学意义(P<0.05)。临床资料建立预测模型灵敏度82.56%,特异度75.00%,曲线下面积0.881;CT表现建立预测模型灵敏度68.18%,特异度47.62%,曲线下面积0.666;结合临床资料及CT表现综合建立预测模型灵敏度85.98%,特异度80.43%,曲线下面积0.922。结论:CT结合临床特征、实验室检查在早期预测COVID-19临床分型转变中具有重要意义,可对疾病严重程度行早期评估。

Objective:To explore the related risk factors in predicting the progress of coronavirus disease 2019(COVID-19)pneumonia from common type to severe type and construct an effective predictive model,further improve the understanding,diagnosis and treatment of COVID-19 pneumonia.Methods:The clinical characteristics and CT imaging features on admission of 153 patients with COVID-19 pneumonia were retrospectively analyzed.Patients were divided into common type and severe type based on guideline of COVID-19(Trial Version 7),including 101 common type and 52 severe type.CT imaging features,clinical symptoms,underlying diseases,laboratory tests of routine blood,liver function and coagulation function were comprehensively analyzed.Logistic regression was used to construct the model and receiver operating characteristic(ROC)curve was also performed.Results:Regarding the clinical characteristics,there were statistically significant differences in male,old age,dyspnea,anorexia,hypertension,diabetes,cardiovascular disease,lymphocyte,C-reactive protein,white blood cell and neutrophil count,albumin and fibrinogen count between the two groups(all P<0.05).The numbers of lung lobe involvement,lesion morphology and presentations of the adjacent pleura were statistically significant between the two groups(P<0.05).Models based on clinical characteristics and CT imaging features achieved a sensitivity,specificity and area under the curve of 82.56%and 68.18%,75.00%and 47.62%,0.881 and 0.666 respectively.When combined the clinical characteristics and CT imaging features,the predictive model yield a sensitivity of 85.98%,specificity of 80.43%and area under curve of 0.922.Conclusion:Using the combination of CT imaging features and clinical characteristics,laboratory examinations is of great significance in early prediction of the progress of COVID-19 pneumonia,which could be used for early assessment of disease severity.