新型冠状病毒感染肺炎患者辅助诊断预测模型的建立

目的:建立一种快速、合理,且识别率高的新型冠状病毒感染肺炎的辅助诊断模型。方法:来自8个医疗机构的30例确诊病例的血清样本检测血常规指标,选取被排除COVID-19的其他患者和健康体检者的血清样本作为对照组,采用随机森林(random forest)方法建立识别模型,最终选取了8个重要指标,模型总准确率86.57%,对阳性样本的预测正确率(即敏感性)可达91.67%,使用内部、外部交互检验方法分别对模型进行了验证,结果证明了所选模型的稳定性和可靠性。结论:本工作提出了一种快速、经济、低人工成本且方便的COVID-19预诊断工具,有助于临床医生提供有价值的诊断信息。

Establishment of a prediction model and auxiliary diagnosis for patients with COVID-19

Objective:The aim of this work is to establish a rapid,reasonable and high recognition model for COVID-19,which could be as an auxiliary diagnostic tool for clinicians.Results:Thirty confirmed cases from eight different medical institutions are included in this work,and 808 cases that are excluded patients with COVID-19 and healthy physical examination are considered as control group.Random forest(RF)algorithm was used to establish the recognition model based on blood routine indexes.ultimately the top-eight important indexes are selected.The accuracy of final model is 86.57%,and the positive samples(i.e.sensitivity)accuracy can achieve 91.67%.The internal and external validation methods are applied to verify the built model,respectively,and the results proved that the stability and reliability of the selected model.Conclusion:This work proposes a fast,economical,low labor cost and convenient COVID-19 pre-diagnosis tool,which is helpful for clinicians to provide valuable diagnostic information.