Development and validation of prediction models for gestational diabetes treatment modality using supervised machine learning: a population-based cohort study

Table 4 Final models developed by simplified logistic regression

Level 1	− 0.856 to 0.005 * history of GDM + 0.741 * BMI obese + 0.800 * prediabetes before pregnancy
Levels 1–2	− 1.001 + 0.572 * history of GDM + 0.579 * pre-pregnancy obesity + 0.774 * prediabetes before pregnancy + 0.733 * screening value^a − 0.323 * history of GDM * pre-pregnancy obesity − 0.577 * history of GDM * screening value^a + 0.480 * pre-pregnancy obesity * screening value^a
Levels 1–3	− 4.468 + 0.074 * oral glucose tolerance test^b − 0.063 * week of gestational age^c − 1.435 diagnosis by C–C criteria^d
Levels 1–4	− 2.645 to 0.810 * meeting glycemic control goal^e + 0.167 * number of SMBG tests taken − 0.076 * week of gestational age^c + 0.044 * oral glucose tolerance test^b − 0.234 * meeting glycemic control goal^e * number of SMBG tests taken

BMI, body mass index; GDM, gestational diabetes mellitus; SMBG, self-monitored blood glucose
The outcome is in log odds form, and coefficients have been rounded to the third decimal point
^aGlucose levels at 50-g, 1-h glucose challenge test for GDM screening (≥ 200 mg/dL)
^bFasting glucose value at 100-g, 3-h oral glucose tolerance test
^cGestational week at GDM diagnosis (continuous)
^dGDM diagnosis by Carpenter-Coustan criteria (versus by fasting hyperglycemia)
^eSMBG control status for the fasting test measured during first week after GDM diagnosis

ISSN: 1741-7015