基于支持向量机算法的重症患者压力性损伤分级预测模型的构建及应用价值

Construction and application value of the grading prediction model of pressure injury in critically ill patients based on support vector machine algorithm

目的基于支持向量机算法构建并验证重症监护病房(ICU)患者压力性损伤(PI)分级预测模型。方法收集2020年12月至2022年12月157例在重庆市某三级甲等医院重症医学科住院患者的临床资料,采用χ~2检验和kruskal-Wallis H检验筛选PI分级的影响因素。再将数据以7∶3的比例随机分为训练组和验证组,基于训练组数据利用支持向量机算法建立ICU患者PI分级预测模型,并采用五折交叉验证法进行参数优化。训练好的模型在验证组数据集中进行内部验证,对前后结果进行混淆矩阵分析,采用准确率、精确率、召回率、F1值和受试者工作特征曲线的曲线下面积(AUC)评估模型的性能。结果初步确定10个影响PI分级的因素,在gamma=0.1、cost=2.2时模型的性能最佳,PI分级预测模型的准确率为81.25%,精确率为79.70%,召回率为80.30%,F1值为79.90%,受试者工作特征曲线的AUC为0.939。结论构建的PI分级预测模型预测性能良好,可为临床医护人员制订ICU患者预防PI的分级护理干预方案提供参考依据。

Objective To construct and verify the grading prediction model of pressure injury(PI)in intensive care unit(ICU)patients based on support vector machine(SVM)algorithm.Methods The clinical data of 157 inpatients in the Department of Critical Care Medicine of a tertiary hospital in Chongqing from December 2020 to December 2022 were collected.The influencing factors of PI grading were screened by the Chi-square test and kruskal-Wallis H test.Then,the data were randomly divided into the training group and the validation group at a ratio of 7∶3.Based on the training group data,the support vector machine algorithm was used to establish the PI grading prediction model of ICU patients,and the five-fold cross-validation method was used to optimize the parameters.The trained model was internally validated in the validation group data set,and the confusion matrix analysis was performed on the results before and after.The performance of the model was evaluated by the accuracy rate,precision rate,recall rate,F1 value and the area under the curve(AUC)of the receiver operating characteristic value.Results The 10 factors affecting PI grading were preliminarily determined.The performance of the model was the best when gamma=0.1 and cost=2.2.The accuracy rate of the PI grading prediction model was 81.25%,the accuracy rate was 79.70%,the recall rate was 80.30%,and the F1 value was 79.90%.The AUC of the receiver operating characteristic was 0.939.Conclusion The constructed PI grading prediction model has good predictive performance,which can provide reference for clinical medical staff to formulate grading nursing intervention programs for preventing PI in ICU patients.