论文部分内容阅读
目的探讨Logistic回归与ROC曲线在儿童巨细胞病毒性肺炎临床筛查中应用的可行性。方法采用Logistic回归筛选出对儿童巨细胞病毒性肺炎发生影响有统计学意义的指标,利用ROC曲线对这些指标进行分析,评价其诊断价值并确定截断值,最后对样本进行预测,以评估模型预测的准确性。结果筛选出3个对儿童巨细胞病毒性肺炎发生影响有统计学意义的指标:血红蛋白、淋巴细胞计数、血小板计数,ROC曲线下的面积分别为0.779±0.047、0.788±0.042、0.816±0.041,截断值分别为118.5 g/L、4.61×109/L、257×109/L,对样本预测的灵敏度、特异度和诊断符合率分别为84.62%、69.57%、78.38%;83.08%、58.69%、72.97%;86.15%、65.22%、77.48%;综合预测模型的灵敏度、特异度和诊断符合率分别为83.08%、76.09%、80.18%。结论联合运用Logistic回归与ROC曲线筛查儿童巨细胞病毒性肺炎是可行的,血红蛋白、淋巴细胞计数、血小板计数3个独立指标对临床筛查都具有诊断价值,但综合预测模型更具诊断价值且预测效果更为理想。
Objective To explore the feasibility of Logistic regression and ROC curve in the clinical screening of children with cytomegalovirus pneumonia. Methods Logistic regression was used to screen out the indicators that had a significant effect on the occurrence of cytomegalovirus pneumonia in children. The ROC curves were used to analyze the indicators to evaluate their diagnostic value and determine the cut-off value. Finally, the samples were predicted to evaluate the model predictive value Accuracy Results Three indicators of cytomegalovirus pneumonia were found to be statistically significant: hemoglobin, lymphocyte count, platelet count, area under the ROC curve were 0.779 ± 0.047,0.788 ± 0.042,0.816 ± 0.041, respectively The sensitivity, specificity and diagnostic accuracy for the prediction of the samples were 84.62%, 69.57%, 78.38%, 83.08%, 58.69% and 72.97, respectively, with the values of 118.5 g / L, 4.61 × 109 / L and 257 × 109 / L, respectively %, 86.15%, 65.22%, 77.48%. The sensitivity, specificity and diagnostic coincidence rate of integrated prediction model were 83.08%, 76.09% and 80.18% respectively. Conclusion Combining Logistic regression and ROC curve to screen cytomegalovirus pneumonia in children is feasible. Three independent indexes such as hemoglobin, lymphocyte count and platelet count have diagnostic value for clinical screening. However, the comprehensive prediction model is more diagnostic value Predict the effect is more ideal.