Background: Current postpartum hemorrhage (PPH) risk stratification is based on traditional statistical models or expert opinion. Machine learning could optimize PPH prediction by allowing for more complex modeling. Objective: We sought to improve PPH prediction and compare machine learning and traditional statistical methods. Methods: We developed models using the Consortium for Safe Labor data set (2002-2008) from 12 US hospitals. The primary outcome was a transfusion of blood products or PPH (estimated blood loss of ≥1000 mL). The secondary outcome was a transfusion of any blood product. Fifty antepartum and intrapartum characteristics and hospital characteristics were included. Logistic regression, support vector machines, multilayer perceptron, random forest, and gradient boosting (GB) were used to generate prediction models. The area under the receiver operating characteristic curve (ROC-AUC) and area under the precision/recall curve (PR-AUC) were used to compare performance. Results: Among 228,438 births, 5760 (3.1%) women had a postpartum hemorrhage, 5170 (2.8%) had a transfusion, and 10,344 (5.6%) met the criteria for the transfusion-PPH composite. Models predicting the transfusion-PPH composite using antepartum and intrapartum features had the best positive predictive values, with the GB machine learning model performing best overall (ROC-AUC=0.833, 95% CI 0.828-0.838; PR-AUC=0.210, 95% CI 0.201-0.220). The most predictive features in the GB model predicting the transfusion-PPH composite were the mode of delivery, oxytocin incremental dose for labor (mU/minute), intrapartum tocolytic use, presence of anesthesia nurse, and hospital type. Conclusions: Machine learning offers higher discriminability than logistic regression in predicting PPH. The Consortium for Safe Labor data set may not be optimal for analyzing risk due to strong subgroup effects, which decreases accuracy and limits generalizability.

Obesity has become increasingly prevalent, impacting up to 41 percent of women in the United States between 2021 and 2023, leading to a rise in short- and long-term adverse health events. With regard to reproductive health, obesity is associated with menstrual irregularities, poorer reproductive and obstetric outcomes, and an increased risk of endometrial cancer. Obesity can lead to hyperandrogenism and anovulation, which is consistent with polycystic ovarian syndrome (PCOS). The prevalence of obesity is higher in women with PCOS compared to the general population. Although PCOS increases the risk of obesity, not all women with PCOS are obese, and not all women with obesity develop PCOS. However, individuals with both PCOS and obesity often present with a more extreme phenotype, with increased risk of chronic anovulation, glucose intolerance, dyslipidemia, metabolic syndrome, vitamin D deficiency, and decreased fertility. Therefore, weight loss is the backbone of patient management in women with obesity and PCOS, and is associated with improvement in cardiovascular risk, as well as improvement in menstrual cycles, ovulation, and pregnancy rate. Lifestyle modifications are often the first-line intervention, with data supporting low glycemic index diets, including ketogenic and DASH diets, along with vitamin D supplementation to improve hormonal imbalances, insulin sensitivity, and menstrual cycles in those who do not have normal vitamin D levels. Furthermore, with the recent widespread adoption of newer FDA-approved medications for weight loss, including GLP-1 (glucagon-like peptide) receptor agonists, new data are emerging regarding the impact of PCOS and longer-term cardiovascular risk. The treatment of PCOS requires a personalized approach, with consideration of a patient’s reproductive goals, tolerance of risk, and acceptance of behavioral and financial commitments, as well as consideration of other medical comorbidities. This narrative review explores different weight loss treatment options, comparing lifestyle modifications (including diet, physical activity, mindfulness, stress management, and cognitive behavioral training), weight loss medications, and bariatric surgery and their respective impact on PCOS to assist clinicians in guiding their patients towards an effective, individualized intervention.

BACKGROUND Current postpartum hemorrhage (PPH) risk stratification is based on traditional statistical models or expert opinion. Machine learning could optimize PPH prediction by allowing for more complex modeling. OBJECTIVE We sought to improve PPH prediction and compare machine learning and traditional statistical methods. METHODS We developed models using the Consortium for Safe Labor data set (2002-2008) from 12 US hospitals. The primary outcome was a transfusion of blood products or PPH (estimated blood loss of ≥1000 mL). The secondary outcome was a transfusion of any blood product. Fifty antepartum and intrapartum characteristics and hospital characteristics were included. Logistic regression, support vector machines, multilayer perceptron, random forest, and gradient boosting (GB) were used to generate prediction models. The area under the receiver operating characteristic curve (ROC-AUC) and area under the precision/recall curve (PR-AUC) were used to compare performance. RESULTS Among 228,438 births, 5760 (3.1%) women had a postpartum hemorrhage, 5170 (2.8%) had a transfusion, and 10,344 (5.6%) met the criteria for the transfusion-PPH composite. Models predicting the transfusion-PPH composite using antepartum and intrapartum features had the best positive predictive values, with the GB machine learning model performing best overall (ROC-AUC=0.833, 95% CI 0.828-0.838; PR-AUC=0.210, 95% CI 0.201-0.220). The most predictive features in the GB model predicting the transfusion-PPH composite were the mode of delivery, oxytocin incremental dose for labor (mU/minute), intrapartum tocolytic use, presence of anesthesia nurse, and hospital type. CONCLUSIONS Machine learning offers higher discriminability than logistic regression in predicting PPH. The Consortium for Safe Labor data set may not be optimal for analyzing risk due to strong subgroup effects, which decreases accuracy and limits generalizability.