重症监护病房住院时长关联规则预测模型的临床应用价值  

Application and evaluation of association rule prediction model in predicting the hospitalization duration of patients at intensive care unit

作　　者：陈璇 温鸿毅 田龙[1] 王晨宇[1] CHEN Xuan;WEN Hongyi;TIAN Long;WANG Chenyu(Department of Intensive Care Unit,the First Affiliated Hospital of Hebei North University,Zhangjiakou 075000,China)

机构地区：[1]河北北方学院附属第一医院重症监护病房,张家口075000

出　　处：《浙江医学》2025年第3期263-267,273,共6页Zhejiang Medical Journal

基　　金：河北省医学科学研究课题计划项目(20220589)。

摘　　要：目的探讨重症监护病房(ICU)患者住院时长关联规则预测模型的临床应用价值。方法回顾性收集2012年3月至2024年3月河北北方学院附属第一医院医院信息系统中1000例ICU患者的基线资料。按照7∶3的比例将患者分为模型组700例和验证组300例。采用FP-Growth算法构建住院时长关联规则预测模型。采用一致性指数和校准曲线对模型进行基于模型组基线资料的内部验证;采用ROC曲线对模型进行基于模型组和验证组基线资料的外部验证。结果关联规则预测模型显示,相应的前项基线资料组合存在时,患者住院时长为1、2、3、4周的概率分别为69.00%、43.00%、22.00%、8.00%。关联规则预测模型的一致性指数为0.727~0.872且具有良好的一致性。关联规则模型预测模型组和验证组患者住院时长的ROC曲线拟合较为理想(均P>0.05)且AUC的差异均无统计学意义(均P>0.05)。结论本研究所构建的关联规则预测模型在ICU患者住院时长的预测中具有一定的应用价值,其预测结果能够为ICU医疗资源配置的优化提供一定的参考。Objective To explore the clinical application value of association rule prediction model for hospitalization duration of patients at intensive care unit(ICU).Methods A total of 1000 ICU patients included in the First Affiliated Hospital of Hebei North University information system from March 2012 to March 2024 were selected.The patients were randomly divided into a model group(n=700)and a validation group(n=300)randomly according to a ratio of 7∶3.FP-Growth algorithm was used to construct the association rule prediction model for the hospitalization duration.The internal validation of the model was performed based on the baseline data of the model group by using the C-index and calibration curve.The external validation of the model was performed based on the baseline data of the model and validation groups by using ROC curve.Results The association rule prediction model showed that the probabilities of the patients with the combination of first item baseline data being hospitalized for 1,2,3 and 4 weeks were 69.00%,43.00%,22.00%,and 8.00%,respectively.The C-index of the association rule prediction model was 0.727-0.872,which showed good consistency.The ROC for the hospitalization duration predicted by the association rule model for the patients in the model and validation groups fitted well(all P>0.05),and their differences in AUC were not statistically significant(all P>0.05).Conclusion The association rule prediction model constructed by the research has certain application value in predicting the hospitalization duration of ICU patients,and its prediction results can provide references for optimizing ICU medical resource allocation.

关 键 词：重症监护病房 住院时长 关联规则 预测 模型 

分 类 号：R47[医药卫生—护理学]