Abstract
.webp)
Authors: Kathan Vyas, Timothy Wen, Mia Charifson, Adesh Kadambi, Shreyas Kadambi, Isabel Fulcher, Sara Sauer
Conference: Society for Maternal and Fetal Medicine Annual Pregnancy Meeting 2026
Objective: Excessive or inadequate gestational weight gain (GWG) are associated with increased risk of adverse maternal and neonatal outcomes. Our goal was to develop and validate a machine learning model for predicting GWG at term using clinical and self-reported weight data through 24 weeks gestational age (wga).
Study Design: We analyzed pregnant individuals from four community clinics using a digital health platform (2021–2025), excluding preterm births, multiple gestations, and insufficient weight data. We combined electronic health records and self-reported data to predict GWG category (inadequate, within range, excessive per IOM guidelines) using a gradient boosting classifier, trained with 5-fold cross-validation. Predictors included pre-pregnancy body-mass index (BMI), weight trajectory calculated as the slope between pre-pregnancy and 16-24 weeks, maternal age, parity, chronic hypertension, and pregestational diabetes. We assessed model performance with area under the curve (AUC) and compared it across BMI and race categories using DeLong’s test with Bonferroni correction.
Results: Among the 1,088 included pregnancies, 287 (26%), 240 (22%), and 561 (52%) had adequate, inadequate, and excessive GWG, respectively. The final model achieved an overall AUC of 0.85 (95% CI: 0.80–0.87), with class-specific AUCs of 0.76 (CI: 0.70–0.81) for "Within Range", 0.90 (CI: 0.86–0.93) for "Inadequate," and 0.90 (CI: 0.87–0.93) for "Excessive" (Figure 1). Performance was stable across BMI and race groups, with no statistically significant differences in AUC (all Bonferroni-adjusted p > 0.05).
Conclusion: Machine learning can be leveraged to construct pragmatic and parsimonious models to accurately predict GWG using clinical and self-reported weight data prior to 24wga with limited bias across BMI or race groups. Early identification of pregnant individuals at risk of either excessive or inadequate GWG enables timely initiation of preventive interventions.
Abstract
Authors: Kathan Vyas, Timothy Wen, Mia Charifson, Adesh Kadambi, Shreyas Kadambi, Isabel Fulcher, Sara Sauer
Conference: Society for Maternal and Fetal Medicine Annual Pregnancy Meeting 2026
Objective: Excessive or inadequate gestational weight gain (GWG) are associated with increased risk of adverse maternal and neonatal outcomes. Our goal was to develop and validate a machine learning model for predicting GWG at term using clinical and self-reported weight data through 24 weeks gestational age (wga).
Study Design: We analyzed pregnant individuals from four community clinics using a digital health platform (2021–2025), excluding preterm births, multiple gestations, and insufficient weight data. We combined electronic health records and self-reported data to predict GWG category (inadequate, within range, excessive per IOM guidelines) using a gradient boosting classifier, trained with 5-fold cross-validation. Predictors included pre-pregnancy body-mass index (BMI), weight trajectory calculated as the slope between pre-pregnancy and 16-24 weeks, maternal age, parity, chronic hypertension, and pregestational diabetes. We assessed model performance with area under the curve (AUC) and compared it across BMI and race categories using DeLong’s test with Bonferroni correction.
Results: Among the 1,088 included pregnancies, 287 (26%), 240 (22%), and 561 (52%) had adequate, inadequate, and excessive GWG, respectively. The final model achieved an overall AUC of 0.85 (95% CI: 0.80–0.87), with class-specific AUCs of 0.76 (CI: 0.70–0.81) for "Within Range", 0.90 (CI: 0.86–0.93) for "Inadequate," and 0.90 (CI: 0.87–0.93) for "Excessive" (Figure 1). Performance was stable across BMI and race groups, with no statistically significant differences in AUC (all Bonferroni-adjusted p > 0.05).
Conclusion: Machine learning can be leveraged to construct pragmatic and parsimonious models to accurately predict GWG using clinical and self-reported weight data prior to 24wga with limited bias across BMI or race groups. Early identification of pregnant individuals at risk of either excessive or inadequate GWG enables timely initiation of preventive interventions.
Abstract
Authors: Kathan Vyas, Timothy Wen, Mia Charifson, Adesh Kadambi, Shreyas Kadambi, Isabel Fulcher, Sara Sauer
Conference: Society for Maternal and Fetal Medicine Annual Pregnancy Meeting 2026
Objective: Excessive or inadequate gestational weight gain (GWG) are associated with increased risk of adverse maternal and neonatal outcomes. Our goal was to develop and validate a machine learning model for predicting GWG at term using clinical and self-reported weight data through 24 weeks gestational age (wga).
Study Design: We analyzed pregnant individuals from four community clinics using a digital health platform (2021–2025), excluding preterm births, multiple gestations, and insufficient weight data. We combined electronic health records and self-reported data to predict GWG category (inadequate, within range, excessive per IOM guidelines) using a gradient boosting classifier, trained with 5-fold cross-validation. Predictors included pre-pregnancy body-mass index (BMI), weight trajectory calculated as the slope between pre-pregnancy and 16-24 weeks, maternal age, parity, chronic hypertension, and pregestational diabetes. We assessed model performance with area under the curve (AUC) and compared it across BMI and race categories using DeLong’s test with Bonferroni correction.
Results: Among the 1,088 included pregnancies, 287 (26%), 240 (22%), and 561 (52%) had adequate, inadequate, and excessive GWG, respectively. The final model achieved an overall AUC of 0.85 (95% CI: 0.80–0.87), with class-specific AUCs of 0.76 (CI: 0.70–0.81) for "Within Range", 0.90 (CI: 0.86–0.93) for "Inadequate," and 0.90 (CI: 0.87–0.93) for "Excessive" (Figure 1). Performance was stable across BMI and race groups, with no statistically significant differences in AUC (all Bonferroni-adjusted p > 0.05).
Conclusion: Machine learning can be leveraged to construct pragmatic and parsimonious models to accurately predict GWG using clinical and self-reported weight data prior to 24wga with limited bias across BMI or race groups. Early identification of pregnant individuals at risk of either excessive or inadequate GWG enables timely initiation of preventive interventions.
Abstract
Authors: Kathan Vyas, Timothy Wen, Mia Charifson, Adesh Kadambi, Shreyas Kadambi, Isabel Fulcher, Sara Sauer
Conference: Society for Maternal and Fetal Medicine Annual Pregnancy Meeting 2026
Objective: Excessive or inadequate gestational weight gain (GWG) are associated with increased risk of adverse maternal and neonatal outcomes. Our goal was to develop and validate a machine learning model for predicting GWG at term using clinical and self-reported weight data through 24 weeks gestational age (wga).
Study Design: We analyzed pregnant individuals from four community clinics using a digital health platform (2021–2025), excluding preterm births, multiple gestations, and insufficient weight data. We combined electronic health records and self-reported data to predict GWG category (inadequate, within range, excessive per IOM guidelines) using a gradient boosting classifier, trained with 5-fold cross-validation. Predictors included pre-pregnancy body-mass index (BMI), weight trajectory calculated as the slope between pre-pregnancy and 16-24 weeks, maternal age, parity, chronic hypertension, and pregestational diabetes. We assessed model performance with area under the curve (AUC) and compared it across BMI and race categories using DeLong’s test with Bonferroni correction.
Results: Among the 1,088 included pregnancies, 287 (26%), 240 (22%), and 561 (52%) had adequate, inadequate, and excessive GWG, respectively. The final model achieved an overall AUC of 0.85 (95% CI: 0.80–0.87), with class-specific AUCs of 0.76 (CI: 0.70–0.81) for "Within Range", 0.90 (CI: 0.86–0.93) for "Inadequate," and 0.90 (CI: 0.87–0.93) for "Excessive" (Figure 1). Performance was stable across BMI and race groups, with no statistically significant differences in AUC (all Bonferroni-adjusted p > 0.05).
Conclusion: Machine learning can be leveraged to construct pragmatic and parsimonious models to accurately predict GWG using clinical and self-reported weight data prior to 24wga with limited bias across BMI or race groups. Early identification of pregnant individuals at risk of either excessive or inadequate GWG enables timely initiation of preventive interventions.
Abstract
Authors: Kathan Vyas, Timothy Wen, Mia Charifson, Adesh Kadambi, Shreyas Kadambi, Isabel Fulcher, Sara Sauer
Conference: Society for Maternal and Fetal Medicine Annual Pregnancy Meeting 2026
Objective: Excessive or inadequate gestational weight gain (GWG) are associated with increased risk of adverse maternal and neonatal outcomes. Our goal was to develop and validate a machine learning model for predicting GWG at term using clinical and self-reported weight data through 24 weeks gestational age (wga).
Study Design: We analyzed pregnant individuals from four community clinics using a digital health platform (2021–2025), excluding preterm births, multiple gestations, and insufficient weight data. We combined electronic health records and self-reported data to predict GWG category (inadequate, within range, excessive per IOM guidelines) using a gradient boosting classifier, trained with 5-fold cross-validation. Predictors included pre-pregnancy body-mass index (BMI), weight trajectory calculated as the slope between pre-pregnancy and 16-24 weeks, maternal age, parity, chronic hypertension, and pregestational diabetes. We assessed model performance with area under the curve (AUC) and compared it across BMI and race categories using DeLong’s test with Bonferroni correction.
Results: Among the 1,088 included pregnancies, 287 (26%), 240 (22%), and 561 (52%) had adequate, inadequate, and excessive GWG, respectively. The final model achieved an overall AUC of 0.85 (95% CI: 0.80–0.87), with class-specific AUCs of 0.76 (CI: 0.70–0.81) for "Within Range", 0.90 (CI: 0.86–0.93) for "Inadequate," and 0.90 (CI: 0.87–0.93) for "Excessive" (Figure 1). Performance was stable across BMI and race groups, with no statistically significant differences in AUC (all Bonferroni-adjusted p > 0.05).
Conclusion: Machine learning can be leveraged to construct pragmatic and parsimonious models to accurately predict GWG using clinical and self-reported weight data prior to 24wga with limited bias across BMI or race groups. Early identification of pregnant individuals at risk of either excessive or inadequate GWG enables timely initiation of preventive interventions.
Abstract
Authors: Kathan Vyas, Timothy Wen, Mia Charifson, Adesh Kadambi, Shreyas Kadambi, Isabel Fulcher, Sara Sauer
Conference: Society for Maternal and Fetal Medicine Annual Pregnancy Meeting 2026
Objective: Excessive or inadequate gestational weight gain (GWG) are associated with increased risk of adverse maternal and neonatal outcomes. Our goal was to develop and validate a machine learning model for predicting GWG at term using clinical and self-reported weight data through 24 weeks gestational age (wga).
Study Design: We analyzed pregnant individuals from four community clinics using a digital health platform (2021–2025), excluding preterm births, multiple gestations, and insufficient weight data. We combined electronic health records and self-reported data to predict GWG category (inadequate, within range, excessive per IOM guidelines) using a gradient boosting classifier, trained with 5-fold cross-validation. Predictors included pre-pregnancy body-mass index (BMI), weight trajectory calculated as the slope between pre-pregnancy and 16-24 weeks, maternal age, parity, chronic hypertension, and pregestational diabetes. We assessed model performance with area under the curve (AUC) and compared it across BMI and race categories using DeLong’s test with Bonferroni correction.
Results: Among the 1,088 included pregnancies, 287 (26%), 240 (22%), and 561 (52%) had adequate, inadequate, and excessive GWG, respectively. The final model achieved an overall AUC of 0.85 (95% CI: 0.80–0.87), with class-specific AUCs of 0.76 (CI: 0.70–0.81) for "Within Range", 0.90 (CI: 0.86–0.93) for "Inadequate," and 0.90 (CI: 0.87–0.93) for "Excessive" (Figure 1). Performance was stable across BMI and race groups, with no statistically significant differences in AUC (all Bonferroni-adjusted p > 0.05).
Conclusion: Machine learning can be leveraged to construct pragmatic and parsimonious models to accurately predict GWG using clinical and self-reported weight data prior to 24wga with limited bias across BMI or race groups. Early identification of pregnant individuals at risk of either excessive or inadequate GWG enables timely initiation of preventive interventions.
Abstract
