Abstract:

Objective　To investigate the relationships between maternal-infant factors and neonatal hyperbilirubinemia by Lasso regression model, and explore its preventive measures. Methods　A retrospective analysis was conducted on the medical records of 200 pregnant women and their newborns in Yangling Demonstration Zone Hospital from January to December 2023. Among the pregnant women, there were 113 primiparas and 87 multiparas, the age ranged from 23 to 40 (31.42±5.64) years old. Among the newborns, there were 105 boys and 95 girls; the birth weight ranged from 1 511 to 4 035 (2 506.50±1 019) g. All newborns were divided into an occurrence group (108 cases) and a non-occurrence group (92 cases) by the results of hyperbilirubinemia. The baseline data of pregnant women and newborns (such as number of deliveries, delivery methods, gestational age, birth weight, etc.) of two groups were collected. Lasso regression was used to select variables, and then logistic regression analysis was used to identify the risk factors influencing the occurrence of hyperbilirubinemia in newborns. R language was used to establish a nomogram prediction model and the efficiency of the prediction model was analyzed. Statistical analysis was conducted using the χ² test and independent sample t test. Results　After conducting a single-factor screening of the 22 most representative risk factors that affected the occurrence of hyperbilirubinemia in newborns, and analyzing 22 variables using Lasso regression and completing ten-fold cross-validation, it was found that a total of 11 variables were selected when λ = 0.035 090. The obtained variables were included in the multivariate logistic regression analysis. The results showed that cesarean section (OR = 2.217), blood type incompatibility of mother and infant (OR = 2.863), gestational age < 34 weeks (OR = 1.842), and onset time of neonatal jaundice < 24 h (OR = 1.799) were risk factors for neonatal hyperbilirubinemia (all P<0.05). In addition, Bootstrap method (B=1000) was used to verify the model. The area under the curve (AUC) of the predicted model was 0.833, 95%CI 0.779 - 0.887. The mean absolute error and mean square error of the calibration curve were 0.033 and 0.001 42. There was no statistically significant difference between the actual and predicted values (χ²= 5.407, P=0.713). The net benefit rate of the decision curve analysis (DCA) curve was good. Conclusions　Maternal and infant factors such as delivery modes, incompatibility of blood group, gestational age, and onset time of neonatal jaundice may increase the risk of neonatal hyperbilirubinemia. Clinical intervention can be strengthened to reduce the risk of neonatal hyperbilirubinemia.

Key words:

Newborns, Hyperbilirubinemia, Lasso regression model, Maternal and infant outcomes, Preventive countermeasures

摘要：

目的　基于Lasso回归模型研究母婴因素与新生儿高胆红素血症的关系，并探讨预防对策。方法　回顾性分析杨凌示范区医院2023年1至12月200例孕妇及其新生儿病历资料。孕妇资料：初产妇113例，经产妇87例；年龄23～40（31.42±5.64）岁。新生儿资料：男105例，女95例；出生体重1 511～4 035（2 506.50±1 019）g。按是否发生新生儿高胆红素血症分组，分为发生组（108例）与未发生组（92例），收集两组孕妇、新生儿基线资料（生产次数、分娩方式、胎龄、出生体重等），先采用Lasso回归筛选变量，再以logistic回归分析影响新生儿发生高胆红素血症的危险因素，并用R语言建立列线图预测模型，分析预测模型效能。采用χ²检验、独立样本t检验进行统计学分析。结果　经单因素筛选影响新生儿发生高胆红素血症的22项最具代表性的危险因素，在对22个变量以Lasso回归进行分析，完成十折交叉验证后，发现当λ=0.035 090时，共筛选出11个变量，将所得变量纳入多因素logistic回归分析，结果表明，剖宫产（OR=2.217）、母婴血型不合（OR=2.863）、胎龄<34周（OR=1.842）、新生儿黄疸发生时间<24 h（OR=1.799）是新生儿发生高胆红素血症的危险因子（均P<0.05）。采用Bootstrap法（B=1 000）对模型进行内验证，预测模型曲线下面积（AUC）为0.833，95%CI 0.779～0.887；校准曲线平均绝对误差0.033，均方误差0.001 42，模型实际观测值与预测值差异无统计学意义（χ²=5.407，P=0.713）；且决策曲线分析（DCA）曲线的净获益率较好。结论　分娩方式、母婴血型不合、胎龄、新生儿黄疸发生时间均可能增加新生儿高胆红素血症发生风险，临床可从上述因素加强干预，以降低其发生风险。

关键词:

新生儿, 高胆红素血症, Lasso回归模型, 母婴结局, 预防对策