Study population A total of 415 women hospitalized at the Fujian Provincial Maternity and Children's Hospital between May 2012 to September 2020 were included in this retrospective study. All the women had delivered two sequential, live, singleton infants and were diagnosed with GDM during pregnancy. The inclusion criteria were (1) a diagnosis of GDM during the two pregnancies and glucose intolerance that was relieved post-partum. The exclusion criteria were (1) pre-existing diabetes diagnosed before pregnancy. (2) Multiparity. (3) Incomplete clinical information in our hospital. (4) During the two pregnancies the women had different husbands. Figure 1 shows that a total of 12,849 primiparous women were diagnosed with GDM between May 2012 and September 2020 at Fujian Maternal and Child Health Hospital. Of these, we excluded 765 women for late miscarriages, stillbirths, and multiple pregnancies. A total of 1811 women had a second pregnancy and delivered in our hospital during the observation period, of which 896 women were diagnosed with recurrent GDM. During the observation period, 43 women were excluded due to pre-pregnancy diabetes. Additionally, we excluded 22 cases involving multiple pregnancy, 37 cases of late abortion or still death, and 19 cases due to three or more times of delivery. Furthermore, 379 cases lack of basic information or 75g OGTT outcome in our hospital. Finally, 415 women were eligible and included in further analysis.
Clinical measurements and definitions
The data recorded at baseline included gestational age (weeks), age, height and weight (pre-pregnancy and antepartum), abdominal circumference of the mother during early pregnancy, and medical history. Laboratory tests were also carried out which included FBG, HDL, TG levels in early pregnancy. Maternal and neonatal complications were also investigated. The maternal complications included cesarean section, postpartum hemorrhage, premature rupture of the membranes, hypothyroidism, hypertension, preeclampsia, fetal distress, and premature birth, while the neonatal complications included hypoglycemia, hyperglycemia, macrosomia, neonatal infection, hyperbilirubinemia, respiratory failure, congenital heart disease, and neonatal sepsis. Pre-pregnancy BMI was calculated by the weight and height before pregnancy. Gestational weight gain was defined as the antepartum weight minus pre-pregnancy body weight, while the pregnancy interval referred to the interval between delivery and the next pregnancy.
Statistical analysis
A machine learning technique called Random Forest tree classification was used for the statistical analyses. In Python 3.7 environment, after simple preprocessing of formatted table data, the Random Forest model in SK-Learn was used for training and testing. At the same time, the Random Forest model under SK-Learn provided the use of information entropy to judge the impact of various feature attributes (risk factors) on the category (outcome).