连续性肾脏替代治疗患者低体温发生风险列线图预测模型构建与验证

Construction and validation of nomogram predictive model for risk of hypothermia in patients undergoing continuous renal replacement therapy

目的构建连续性肾脏替代治疗患者低体温发生风险列线图预测模型并验证。方法采取回顾性研究,采用便利抽样的方法选取2019年1月—2022年12月浙江省某三级甲等综合性医院符合标准的450例连续性肾脏替代治疗患者,根据是否发生低体温将其分为低体温组(n=140)与非低体温组(n=310),采用单因素分析和logistic回归分析风险因素,构建风险预测模型,绘制列线图。采用前瞻性队列研究,收集2023年1—9月某三级甲等医院重症监护室的120例连续性肾脏替代治疗患者作为外部验证集对模型进行外部验证。结果通过多因素分析,最终纳入基线体温(OR=0.092)、APACHE-Ⅱ评分(OR=2.499)、是否机械通气(OR=2.578)、连续性肾脏替代治疗持续时间(OR=3.483)构建风险预测模型。训练集和验证集的ROC曲线下面积为0.856、0.819;Bootstrap内部验证其校准曲线实际值与预测值间的平均绝对误差为0.027;Hosmer-Lemeshow检验P=0.063(P>0.05),表明模型拟合优度良好;Calibration校准曲线与决策曲线分析显示模型具有较好的校准度与临床实用性。结论本研究构建的列线图风险预测模型具有较好的区分度、校准度和临床实用性,可有效预测连续性肾脏替代治疗患者的低体温发生风险。

Objective To construct a nomogram prediction model of the risk of developing hypothermia in patients undergoing continuous renal replacement therapy and validate the model.Methods A retrospective study was conducted,and 450 patients undergoing consecutive renal replacement therapy in a tertiary grade-A hospital in Zhejiang Province from January 2019 to December 2022 were selected by convenience sampling and they were divided into hypothermia group(n=140)and non-hypothermia group(n=310)according to the occurrence of hypothermia or not.One-way analysis of variance and logistic regression analysis were used to analyze the risk factors,construct risk prediction model,and plot a nomogram.A prospective cohort study was used to collect 120 patients undergoing consecutive renal replacement therapy in the intensive care unit of a tertiary grade-A hospital from January to September 2023 as an external validation set to externally validate the model.Results The risk prediction model was constructed by multifactorial analysis incorporating baseline body temperature(OR=0.092),APACHE-II score(OR=2.499),whether or not mechanical ventilation(OR=2.578),and duration of continuous renal replacement therapy(OR=3.483).The areas under the ROC curves of the training and validation set were 0.856 and 0.819,respectively;Bootstrap internal validation of the mean absolute error between the actual and predicted value of its calibration curves was 0.027;Hoer-Lemeshow test of P=0.063(P>0.05)indicated high goodness of fit of the model;The calibration curve and the decision curve showed that the model had the calibration degree and clinical usefulness.Conclusion The nomogram risk prediction model constructed in this study has the differentiation,calibration and clinical practicability,and can effectively predict the risk of hypothermia in patients undergoing continuous renal replacement therapy.