分类树模型在实验室指标预测重症手足口病中的应用效果

Application Effect of Classification Tree Model in Predicting Severe Hand-Foot-Mouth Disease by Laboratory Indicators

目的用分类树模型对实验室指标预测重症手足口病(HFMD)的效果进行评价。方法共选取2013—2018年4—6月于郑州大学附属儿童医院感染性疾病科住院治疗的100例HFMD患儿(其中重症50例为病例组,轻症50例为对照组),收集患儿的相关实验室检查指标,采用CHAID分类树建立模型以预测重症HFMD的发生风险,之后在对该模型的价值进行评价时,评价指标包括索引图、错分概率Risk值及ROC曲线。结果模型共4层,共13个节点,共筛选出CD19^(+)、α-羟丁酸脱氢酶(α-HBDH)、CD4^(+)/CD8^(+)、S100、谷草转氨酶共5个解释变量;其中在该模型中最重要的预测因素是CD19^(+)。模型错分概率Risk值为0.220。结论分类树模型不仅可以对重症HFMD的发生风险进行有效的拟合,亦可以对变量间的交互作用进行筛选。

Objective To evaluate the effect of laboratory indicators in predicting severe hand-foot-mouth disease(HFMD)by using classification tree model.Methods A total of 100 children with HFMD who were hospitalized in the Department of Infectious Diseases,Childre’s Hospital Affiliated of Zhengzhou University from April to June 2013 to 2018 were selected(the case group included 50 severe HFMD patients,while the control group included 50 mild HFMD patients).Relevant laboratory examination indicators of the children were collected,and the CHAID classification tree was used to build the model to predict the risk of severe HFMD.Later,when evaluating the value of the model,the evaluation indexes include the index graph,the misscore probability Risk value and the ROC curve.Results There were 4 stratum and 13 nodes in the model,in which 5 explanatory variables were selected,CD19^(+)level,α-hydroxybutyrate dehydrogenase(α-HBDH),CD4^(+)/CD8^(+),S100 protein and aspartate aminotransferase involved,among which CD19^(+)level was the most important.The Risk value of misclassification probability was 0.220.Conclusion The classification tree model can not only effectively fit the risk of severe HFMD,but also screen the interaction between variables.