新生儿G6PD缺乏症的风险预测模型的构建及验证  

Development of a diagnostic prediction model for newborns with G6PD deficiency

作　　者：刘佳欣 刘荣秋 李婷[1] 罗朝莲 刘宣辰 魏丽 LIU Jiaxin;LIU Rongqiu;LI Ting;LUO Chaolian;LIU Xuanchen;WEI Li(Department of Pediatrics,Sichuan Provincial Maternity and Child Health Care Hospital,Chengdu,Sichuan 610045,China;Children’s Hospital of Chongqing Medical University,Chongqing 400015,China;NHC Key Laboratory of Birth Defects and Reproductive Health,Chongqing Population and Family Planning Science and Technology Research Institute,Chongqing 400020,China)

机构地区：[1]四川省妇幼保健院儿科,四川成都610045 [2]重庆医科大学附属儿童医院,重庆400015 [3]国家卫健委出生缺陷与生殖健康重点实验室/重庆市人口和计划生育科学技术研究院,重庆400020

出　　处：《中国优生与遗传杂志》2024年第2期237-246,共10页Chinese Journal of Birth Health & Heredity

基　　金：重庆市自然基金(CSTB2022NSCQ-MSX0227)。

摘　　要：目的识别合并G6PD缺乏症的新生儿以完善筛查策略。方法本研究为观察性类别、病例对照研究。自2022年1月1日—12月31日,连续性纳入四川省新生儿疾病筛查中心接受G6PD筛查的G6PD缺乏症新生儿为研究对象,通过PSM 1∶4匹配对照人群,并将其拆分为训练集、验证集。通过训练集样本,应用单因素和多因素回归以最终确定模型,通过区分度、校准度、适用度、合理度,全面评价预测模型。将模型信息代入验证集样本,汇总判定预测模型的严谨性和稳定性,再使用Nomogram图可视化呈现模型。结果共获得640例受试者,其中训练集448例、验证集192例。单因素-多因素Logistic回归显示:新生儿G6PD缺乏症的诊断预测模型包括母乳喂养、男性、Hb、既往贫血史(P<0.05)。经过4个维度评价训练集预测模型,具有良好的区分度、校准度、适用度、合理度。代入验证集样本予以内部验证,发现区分度、校准度均表现良好,而适用度、合理度则表现尚可。且验证集与训练集Nomogram图的赋值评分及趋势基本相同。结论本研究构建了新生儿G6PD缺乏症的诊断预测模型,识别高危风险新生儿的能力可靠。Objective To identify newborns with G6PD deficiency and improve screening strategies.Methods This study is an observational case-control study.From January 1,2022,to December 31,2022,newborns with G6PD deficiency undergoing G6PD screening at the Newborn Disease Screening Center in Sichuan Province were consecutively included as the study population.A 1:4 matching with control subjects was performed using propensity score matching(PSM),and the dataset was split into a training set and a validation set.Single-factor and multiple-factor regression analyses were conducted on the training set to determine the final model.The predictive model was comprehensively evaluated for discrimination,calibration,applicability,and reasonableness.The model information was then applied to the validation set to assess its rigor and stability.A Nomogram graph was used to visualize the model.Results A total of 640 subjects were included,with 448 in the training set and 192 in the validation set.Single-factor and multiple-factor Logistic regression showed that the diagnostic prediction model for newborn G6PD deficiency included factors such as breastfeeding,male gender,Hb levels,and a history of anemia(P<0.05).The training set predictive model demonstrated good discrimination,calibration,applicability,and reasonableness across all four dimensions.Internal validation using the validation set showed good discrimination and calibration,with acceptable applicability and reasonableness.The allocation scores and trends in the Nomogram graphs for the validation set were similar to those of the training set.Conclusion This study has developed a diagnostic prediction model for newborn G6PD deficiency,which reliably identifies high-risk newborns.

关 键 词：G6PD 新生儿 筛查 病例对照 诊断预测模型 

分 类 号：R722.1[医药卫生—儿科]