基于人工智能的重症监护室患者出院后健康风险预测模型构建与应用研究OA
Research on the construction and application of a health risk prediction model for ICU patients after discharge based on artificial intelligence
目的 构建基于人工智能的重症监护室(ICU)患者出院后健康风险预测模型,评估其性能及临床应用价值.方法 采用人工智能技术开发ICU患者健康管理信息系统,通过医院信息系统接口和智能移动终端采集 2024 年1～12 月于厦门大学附属第一医院ICU接受治疗且顺利出院的164 例患者的临床数据.采用派森语言进行开发并结合机器学习方法对特征进行筛选及模型训练.以2024 年1～6 月的数据构建训练集,2024 年 7～12 月的数据则用于模型验证.通过Logistic回归分析筛选预测因素,构建列线图预测模型,采用自助法进行内部验证,并通过校准曲线、受试者工作特征曲线和决策曲线分析评估模型性能.结果 ICU内是否发生感染、ICU内是否气管切开、镇静药物使用、血管活性药物使用、入ICU 24 h急性生理学和慢性健康状况评价Ⅱ评分和住ICU总时长均与ICU患者出院后的健康风险相关(P均<0.05).预测模型的训练集和验证集曲线下面积分别为0.935(95%CI为0.912～0.959)和0.875(95%CI为0.822～0.927),校准曲线显示模型具有良好的校准度,决策曲线分析证实模型具有较高的临床应用价值.结论 基于人工智能构建的ICU患者出院后健康风险预测模型具有良好的预测性能和临床实用性,可为患者出院后的健康管理提供决策支持.
Objective To construct a health risk prediction model for patients in intensive care units(ICU)after discharge based on artificial intelligence and evaluate its performance as well as clinical application value.Methods The ICU patient health management information system was developed by adopting artificial intelligence technology.Clinical data of 164 patients who received treatment and were successfully discharged from the ICU of the First Affiliated Hospital of Xiamen University from January to December 2024 were collected through hospital information system interfaces and intelligent mobile terminals.The Python programming language was used for development,and machine learning methods were used for feature screening and model training.Data from January to June 2024 were used for developing an training set,while data from July to December 2024 were used for model validation.Logistic regression analysis was applied to screen forecasting factors,and a nomogram prediction model was constructed.Internal validation was performed using a self-service method,and the model performance was evaluated through calibration curves,receiver operating characteristic curves,and decision curves.Results The occurrence of infection in the ICU,tracheostomy during ICU stay,use of sedatives,use of vasoactive drugs,the Acute Physiology and Chronic Health EvaluationⅡscore within 24 hours of admission to the ICU,and the total length of stay in the ICU were all related to the health risks of ICU patients after discharge(P<0.05).The forecasting model demonstrated the areas under the curve of 0.935(95%CI:0.912～0.959)for the training set and 0.875(95%CI:0.822～0.927)for the validation set.The calibration curve indicated that the model had good calibration,and decision curve analysis confirmed that the model had high clinical application value.Conclusion The health risk prediction model for ICU patients after discharge based on artificial intelligence exhibits robust forecasting performance and clinical practicality,which can provide decision-making support for the health management of patients after discharge.
黄洋洋;许文萍
厦门大学附属第一医院,福建 厦门,361000厦门大学附属第一医院,福建 厦门,361000
医药卫生
重症监护室人工智能预测模型健康风险预后
intensive care unitartificial intelligenceprediction modelhealth riskprognosis
《中西医结合护理（中英文）》 2025 (7)
5-9,5
评论