先天性心脏病术后患儿营养不良风险预测模型的构建及验证研究

Construction and verification of a risk prediction model for postoperative malnutrition in infants with congenital heart disease

目的构建先天性心脏病术后患儿营养不良风险预测模型,进行内部及外部验证。方法采用便利抽样法,选取2018年1月—2021年12月山东省某三级甲等医院收治的300例6月龄以下的先天性心脏病术后患儿作为建模组,选取2022年1月—2023年6月收治的129例患儿作为验证组。回顾性收集患儿的一般资料、疾病相关资料、治疗和护理相关资料,通过单因素、Logistic回归分析构建先天性心脏病术后患儿营养不良风险预测模型,绘制列线图并评价其预测效果。结果先天性心脏病术后患儿营养不良发生率为33.10%,Logistic回归分析结果显示,出生体重<2.5 kg、术前营养不良、术后1周液体负平衡、体外循环时间较长是患儿发生营养不良的危险因素(P<0.05)。该模型建模组受试者操作特征曲线下面积为0.933,灵敏度为83.30%,特异度为90.90%,Hosmer-Lemeshow检验显示,χ^(2)=7.765(P=0.457)。验证组受试者操作特征曲线下面积为0.918,灵敏度为87.20%,特异度为90.00%,Hosmer-Lemeshow检验显示,χ^(2)=4.947(P=0.763)。两组校准曲线均显示,该模型具有较好的校准度;临床决策曲线显示,该模型具有较好的临床实用性。结论该研究构建的风险预测模型具有较好的预测能力,可为医护人员早期识别先天性心脏病术后营养不良高危患儿,制订针对性干预措施提供参考。

Objective To develop a risk prediction model for postoperative malnutrition in children with congenital heart disease(CHD)and to verify it both internally and externally.Methods By a convenience sampling method,300 CHD children treated at a tertiary hospital in Shandong Province from January 2018 to December 2021 were selected as a modeling group,and 129 children from January 2022 to June 2023 were selected as a validation group.Data on patient demographics,disease-specific variables,therapeutic interventions,and nursing care parameters were collected.Single factor and logistic regression were employed to construct a risk prediction model for postoperative malnutrition in CHD children,and the nomogram was drawn and its prediction effect was evaluated.Results The incidence of postoperative malnutrition among CHD children was 33.10%.Logistic regression analysis revealed that risk factors for malnutrition in children included birth weight<2.5 kg,preoperative malnutrition,negative fluid balance 1 week after surgery,and long duration of cardiopulmonary bypass(P<0.05).The area under the receiver operating characteristic curve of the modeling group was 0.933;the sensitivity was 83.30%;the specificity was 90.90%.The Hosmer-Lemeshow test showed thatχ^(2)=7.765(P=0.457).The AUC of the validation group was 0.918;the sensitivity was 87.20%;the specificity was 90.00%.The Hosmer-Lemeshow test showed thatχ^(2)=4.947(P=0.763).Calibration curves for both groups indicated good calibration of the model,and the clinical decision curves demonstrated its practical clinical utility.Conclusion The risk prediction model developed in this study exhibits good predictive ability,which can provide a reference for medical staff to early identify high-risk infants for postoperative malnutrition following CHD surgery and to formulate targeted intervention measures.