American journal of translational research · 2026
Construction and validation of a multicenter predictive model for neonatal hypoglycemia risk in infants of mothers with gestational diabetes mellitus.
СтатьяНазвание и аннотация: Перевод в очереди
Обработка статьи · этапы, ошибки и управление
Загрузка состояния…
Ошибки и ограничения
История · было → стало
Состояние обновляется каждые 5 секунд, пока блок открыт. Процент показывается только там, где известен объём работы. Пауза останавливает продолжение этапа; текущий запрос к источнику может завершиться.
Geng Xiuchen, Hou Mingyan, Zhang Shuai, Yang Wei, Geng Zhijie
Аннотация
OBJECTIVE: This study aimed to develop and validate a predictive model for neonatal hypoglycemia (NH) risk in infants born to mothers with gestational diabetes mellitus (GDM).
METHODS: A multicenter retrospective cohort study included 3215 GDM mother-infant pairs from five hospitals in Tianjin. Data were split into training, internal, and external validation sets. Risk factors were selected by expert consultation and univariate analysis. Logistic regression, random forest (RF), and radial basis function neural network models were built and evaluated using AUC, sensitivity, and specificity.
RESULTS: Fourteen risk factors were identified, with hypothermia (OR = 2.31), insulin treatment during pregnancy (OR = 2.15), and large for gestational age (OR = 2.08) being the strongest. The RF model performed best, with AUC values of 0.896, 0.872, and 0.865 across validation groups. In external validation (threshold = 0.38), sensitivity was 82.19% and specificity 79.38%. Subgroup analysis by maternal age, gestational week, and neonatal sex showed stable performance (AUC 0.848-0.895). A simplified RF model using five key predictors retained 97.34% of performance (AUC = 0.842) and reduced assessment time to 2-3 minutes.
CONCLUSION: The RF model effectively predicts NH risk in GDM newborns with strong generalizability, supporting early clinical identification and intervention. Hypothermia, insulin use, and large for gestational age are core risk factors.
METHODS: A multicenter retrospective cohort study included 3215 GDM mother-infant pairs from five hospitals in Tianjin. Data were split into training, internal, and external validation sets. Risk factors were selected by expert consultation and univariate analysis. Logistic regression, random forest (RF), and radial basis function neural network models were built and evaluated using AUC, sensitivity, and specificity.
RESULTS: Fourteen risk factors were identified, with hypothermia (OR = 2.31), insulin treatment during pregnancy (OR = 2.15), and large for gestational age (OR = 2.08) being the strongest. The RF model performed best, with AUC values of 0.896, 0.872, and 0.865 across validation groups. In external validation (threshold = 0.38), sensitivity was 82.19% and specificity 79.38%. Subgroup analysis by maternal age, gestational week, and neonatal sex showed stable performance (AUC 0.848-0.895). A simplified RF model using five key predictors retained 97.34% of performance (AUC = 0.842) and reduced assessment time to 2-3 minutes.
CONCLUSION: The RF model effectively predicts NH risk in GDM newborns with strong generalizability, supporting early clinical identification and intervention. Hypothermia, insulin use, and large for gestational age are core risk factors.