Using machine learning to predict the risk of developing hypertensive disorders of pregnancy using a contemporary nulliparous cohort.

Background: Hypertensive disorders of pregnancy (HDP) are significant drivers of maternal and neonatal morbidity and mortality. Current management strategies include early identification and initiation of risk mitigating interventions facilitated by a rules-based checklist. Advanced analytic techniques, such as machine learning, can potentially offer improved and refined predictive capabilities. Objective: To develop and internally validate a machine learning prediction model for hypertensive disorders of pregnancy (HDP) when initiating prenatal care. Study Design: We developed a prediction model using data from the prospective multisite cohort Nulliparous Pregnancy Outcomes Study: Monitoring Mothers-to-Be (nuMoM2b) among low-risk individuals without a prior history of aspirin utilization for preeclampsia prevention. The primary outcome was the development of HDP. Random forest modeling was utilized to develop predictive models. Recursive feature elimination (RFE) was employed to create a reduced model for each outcome. Area under the curve (AUC), 95% confidence intervals (CI), and calibration curves were utilized to assess discrimination and accuracy. Sensitivity analyses were conducted to compare the sensitivity and specificity of the reduced model compared to existing risk factor-based algorithms. Results: Of 9,124 assessed low risk nulliparous individuals, 21% (n=1,927) developed HDP. The prediction model for HDP had satisfactory discrimination with an AUC of 0.73 (95% CI: 0.70, 0.75). After RFE, a parsimonious reduced model with 30 features was created with an AUC of 0.71 (95% CI: 0.68, 0.74). Variables included in the model after RFE included body mass index at the first study visit, pre-pregnancy weight, first trimester complete blood count results, and maximum systolic blood pressure at the first visit. Calibration curves for all models revealed relatively stable agreement between predicted and observed probabilities. Sensitivity analysis noted superior sensitivity (AUC 0.80 vs 0.65) and specificity (0.65 vs 0.53) of the model compared to traditional risk factor-based algorithms. Conclusion: In cohort of low-risk nulliparous pregnant individuals, a prediction model may accurately predict HDP diagnosis at the time of initiating prenatal care and aid employment of close interval monitoring and prophylactic measures earlier in pregnancy.

Betamethasone Exposure and Neonatal Respiratory Morbidity Among Late Preterm Births by Planned Mode of Delivery and Gestational Age.

OBJECTIVE: To estimate the effect of late preterm antenatal steroids on the risk of respiratory morbidity among subgroups of patients on the basis of the planned mode of delivery and gestational age at presentation. METHODS: This was a secondary analysis of the ALPS (Antenatal Late Preterm Steroid) Trial, a multicenter trial conducted within the Eunice Kennedy Shriver National Institute of Child Health and Human Development Maternal-Fetal Medicine Units Network of individuals with singleton gestations and without preexisting diabetes who were at high risk for late preterm delivery (34-36 weeks of gestation). We fit binomial regression models to estimate the risk of respiratory morbidity, with and without steroid administration, by gestational age and planned mode of delivery at the time of presentation. We assumed a homogeneous effect of steroids on the log-odds scale, as was reported in the ALPS trial. The primary outcome was neonatal respiratory morbidity, as defined in the ALPS Trial. RESULTS: The analysis included 2,825 patients at risk for late preterm birth. The risk of respiratory morbidity varied significantly by planned mode of delivery (adjusted risk ratio [RR] 1.90, 95% CI, 1.55-2.33 for cesarean delivery vs vaginal delivery) and week of gestation at presentation (adjusted RR 0.56, 95% CI, 0.50-0.63). For those planning cesarean delivery and presenting in the 34th week of gestation, the risk of neonatal respiratory morbidity was 39.4% (95% CI, 30.8-47.9%) without steroids and 32.0% (95% CI, 24.6-39.4%) with steroids. In contrast, for patients presenting in the 36th week and planning vaginal delivery, the risk of neonatal respiratory morbidity was 6.9% (95% CI, 5.2-8.6%) without steroids and 5.6% (95% CI, 4.2-7.0%) with steroids. CONCLUSION: The absolute risk difference of neonatal respiratory morbidity between those exposed and those unexposed to late preterm antenatal steroids varies considerably by gestational age at presentation and planned mode of delivery. Because only communicating the relative risk reduction of antenatal steroids for respiratory morbidity may lead to an inaccurate perception of benefit, more patient-specific estimates of risk expected with and without treatment may inform shared decision making.

Development of a Practical Prediction Model for Adverse Neonatal Outcomes at the Start of the Second Stage of Labor.

OBJECTIVE: To develop a prediction model for adverse neonatal outcomes using electronic fetal monitoring (EFM) interpretation data and other relevant clinical information known at the start of the second stage of labor. METHODS: This was a retrospective cohort study of individuals who labored and delivered at two academic medical centers between July 2016 and June 2020. Individuals were included if they had a singleton gestation at term (more than 37 weeks of gestation), a vertex-presenting, nonanomalous fetus, and planned vaginal delivery and reached the start of the second stage of labor. The primary outcome was a composite of severe adverse neonatal outcomes. We developed and compared three modeling approaches to predict the primary outcome using factors related to EFM data (as interpreted and entered in structured data fields in the electronic health record by the bedside nurse), maternal comorbidities, and labor characteristics: traditional logistic regression, LASSO (least absolute shrinkage and selection operator), and extreme gradient boosting. Model discrimination and calibration were compared. Predicted probabilities were stratified into risk groups to facilitate clinical interpretation, and positive predictive values for adverse neonatal outcomes were calculated for each. RESULTS: A total of 22,454 patients were included: 14,820 in the training set and 7,634 in the test set. The composite adverse neonatal outcome occurred in 3.2% of deliveries. Of the three modeling methods compared, the logistic regression model had the highest discrimination (0.690, 95% CI, 0.656-0.724) and was well calibrated. When stratified into risk groups (no increased risk, higher risk, and highest risk), the rates of the composite adverse neonatal outcome were 2.6% (95% CI, 2.3-3.1%), 6.7% (95% CI, 4.6-9.6%), and 10.3% (95% CI, 7.6-13.8%), respectively. Factors with the strongest associations with the composite adverse neonatal outcome included the presence of meconium (adjusted odds ratio [aOR] 2.10, 95% CI, 1.68-2.62), fetal tachycardia within the 2 hours preceding the start of the second stage (aOR 1.94, 95% CI, 1.03-3.65), and number of prior deliveries (aOR 0.77, 95% CI, 0.60-0.99).

Postpartum Primary Care Engagement Using Default Scheduling and Tailored Messaging: A Randomized Clinical Trial.

Importance: More than 30% of pregnant people have at least 1 chronic medical condition, and nearly 20% develop gestational diabetes or pregnancy-related hypertension, increasing the risk of future chronic disease. While these individuals are often monitored closely during pregnancy, they face major barriers when transitioning to primary care following delivery, due in part to a lack of health care support for this transition. Objective: To evaluate the impact of an intervention designed to improve postpartum primary care engagement by reducing patient administrative burden and information gaps. Design, Setting, and Participants: An individual-level randomized clinical trial was conducted from November 3, 2022, to October 11, 2023, at 1 hospital-based and 5 community-based outpatient obstetric clinics affiliated with a large academic medical center. Participants included English- and Spanish-speaking pregnant or recently postpartum adults with obesity, anxiety, depression, diabetes, chronic hypertension, gestational diabetes, or pregnancy-related hypertension and a primary care practitioner (PCP) listed in their electronic health record. Intervention: A behavioral economics-informed intervention bundle, including default scheduling of postpartum PCP appointments and tailored messages. Main Outcome and Measures: Completion of a PCP visit for routine or chronic condition care within 4 months of delivery was the primary outcome, ascertained directly by reviewing the patient's electronic health record approximately 5 months after their estimated due date. Intention-to-treat analysis was conducted. Results: A total of 360 patients were randomized (control, 176; intervention, 184). Individuals had a mean (SD) age of 34.1 (4.9) years and median gestational age of 36.3 (IQR, 34.0-38.6) weeks at enrollment. The distribution of self-reported race and ethnicity was 6.8% Asian, 7.4% Black, 68.6% White, and 15.0% multiple races or other. Most participants (75.4%) had anxiety or depression, 16.1% had a chronic or pregnancy-related hypertensive disorder, 19.5% had preexisting or gestational diabetes, and 40.8% had a prepregnancy body mass index of 30 or greater. Medicaid was the primary payer for 21.2% of patients. Primary care practitioner visit completion within 4 months occurred in 22.0% (95% CI, 6.4%-28.8%) of individuals in the control group and 40.0% (95% CI, 33.1%-47.4%) in the intervention group. In regression models accounting for randomization strata, the intervention increased PCP visit completion by 18.7 percentage points (95% CI, 9.1-28.2 percentage points). Intervention participants also had fewer postpartum readmissions (1.7% vs 5.8%) and increased receipt of the following services by a PCP: blood pressure screening (42.8% vs 28.3%), weight assessment (42.8% vs 27.7%), and depression screening (32.8% vs 16.8%). Conclusions and Relevance: The findings of this randomized clinical trial suggest that the current lack of support for postpartum transitions to primary care is a missed opportunity to improve recently pregnant individual's short- and long-term health. Reducing patient administrative burdens may represent relatively low-resource, high-impact approaches to improving postpartum health and well-being. Trial Registration: ClinicalTrials.gov Identifier: NCT05543265.

Stratifying Risk for Postpartum Depression at Time of Hospital Discharge.

Objective: Postpartum depression (PPD) represents a major contributor to postpartum morbidity and mortality. Beyond efforts at routine screening, risk stratification models could enable more targeted interventions in settings with limited resources. Thus, we aimed to develop and estimate the performance of a generalizable risk stratification model for PPD in patients without a history of depression using information collected as part of routine clinical care. Methods: We performed a retrospective cohort study of all individuals who delivered between 2017 and 2022 in one of two large academic medical centers and six community hospitals. An elastic net model was constructed and externally validated to predict PPD using sociodemographic factors, medical history, and prenatal depression screening information, all of which was known before discharge from the delivery hospitalization. Results: The cohort included 29,168 individuals; 2,703 (9.3%) met at least one criterion for postpartum depression in the 6 months following delivery. In the external validation data, the model had good discrimination and remained well-calibrated: area under the receiver operating characteristic curve 0.721 (95% CI: 0.707-0.734), Brier calibration score 0.088 (95% CI: 0.084 - 0.092). At a specificity of 90%, the positive predictive value was 28.0% (95% CI: 26.0-30.1%), and the negative predictive value was 92.2% (95% CI: 91.8-92.7%). Conclusions: These findings demonstrate that a simple machine-learning model can be used to stratify the risk for PPD before delivery hospitalization discharge. This tool could help identify patients within a practice at the highest risk and facilitate individualized postpartum care planning regarding the prevention of, screening for, and management of PPD at the start of the postpartum period and potentially the onset of symptoms.

Increasing Postpartum Primary Care Engagement through Default Scheduling and Tailored Messaging: A Randomized Clinical Trial.

Importance: Over 30% of pregnant people have at least one chronic medical condition, and nearly 20% develop gestational diabetes or pregnancy-related hypertension, increasing the risk of future chronic disease. While these individuals are often monitored closely during pregnancy, they face significant barriers when transitioning to primary care following delivery, due in part to a lack of health care support for this transition. Objective: To evaluate the impact of an intervention designed to improve postpartum primary care engagement by reducing patient administrative burden and information gaps. Design: Individual-level randomized controlled trial conducted from November 3, 2022 to October 11, 2023. Setting: One hospital-based and five community-based outpatient obstetric clinics affiliated with a large academic medical center. Participants: Participants included English- and Spanish-speaking pregnant or recently postpartum adults with obesity, anxiety, depression, diabetes mellitus, chronic hypertension, gestational diabetes, or pregnancy-related hypertension, and a primary care practitioner (PCP) listed in their electronic health record (EHR). Intervention: A behavioral economics-informed intervention bundle, including default scheduling of postpartum PCP appointments and tailored messages. Main Outcome: Completion of a PCP visit for routine or chronic condition care within 4 months of delivery. Results: 360 patients were randomized (Control: N=176, Intervention: N=184). Individuals had mean (SD) age 34.1 (4.9) years and median gestational age of 36.3 weeks (interquartile range (IQR) 34.0-38.6 weeks) at enrollment. The distribution of self-reported races was 7.4% Asian, 6.8% Black, 15.0% multiple races or "Other," and 68.6% White. Most (75.8%) participants had anxiety or depression, 15.9% had a chronic or pregnancy-related hypertensive disorder, 19.8% had pre-existing or gestational diabetes, and 40.4% had a pre-pregnancy BMI ≥30 kg/m2. Medicaid was the primary payer for 21.9% of patients. PCP visit completion within 4 months occurred in 22.0% in the control group and 40.0% in the intervention group. In regression models accounting for randomization strata, the intervention increased PCP visit completion by 18.7 percentage points (95%CI 10.7-29.1). Intervention participants also had fewer postpartum readmissions (1.7 vs. 5.8%) and increased receipt of the following services by a PCP: blood pressure screening (42.8 vs. 28.3%), weight assessment (42.8 vs. 27.7%), and depression screening (32.8 vs. 16.8%). Conclusions and Relevance: In this randomized trial of pregnant individuals with or at risk for chronic health conditions, default PCP visit scheduling, tailored messages, and reminders substantially improved postpartum primary care engagement. The current lack of support for postpartum transitions to primary care is a missed opportunity to improve recently pregnant individual's short- and long-term health. Reducing patient administrative burdens may represent relatively low-resource, high-impact approaches to improving postpartum health and wellbeing. Trial Registration: NCT05543265.

Increased Rates of Hypertensive Disorders of Pregnancy during the COVID-19 Pandemic.

OBJECTIVE: Hypertensive disorders of pregnancy (HDP) are common complications associated with severe maternal and neonatal morbidity. One goal of prenatal care, especially at term, is to screen for HDP. As treatment of HDP centers on delivery when appropriate, timely diagnosis is crucial. We postulated that reduced in-person visits during the coronavirus disease 2019 (COVID-19) pandemic may have resulted in delayed diagnosis of HDP with concomitant higher rates of maternal morbidity. We sought to investigate the prevalence of HDP during the COVID-19 pandemic, as well as median gestational age at time of delivery as compared with the prepandemic median. STUDY DESIGN: This was a retrospective cohort analysis comparing singleton deliveries at four large-volume hospitals during the COVID-19 pandemic (April-July 2020 during a statewide "stay-at-home" order) to those in a pre-COVID era (April-July 2019). Deliveries complicated by HDP were identified by International Classification of Disease, Tenth Revision codes. Rates of HDP and markers of severe disease were the primary outcomes compared between the groups; multivariate regression was used to calculate the odds ratio of severe disease among women with any diagnosis of HDP. RESULTS: The cohort included 9,974 deliveries: 5,011 in 2020 and 4,963 in 2019. Patient characteristics (age, body mass index, race, ethnicity, and insurance type) did not differ significantly between the groups. There was an increase in HDP during the COVID era (9.0 vs. 6.9%; p < 0.01), which was significant even when controlling for patient parity (odds ratio = 1.41, 95% confidence interval: 1.20-1.66). Among women with HDP, gestational age at delivery did not differ between the cohorts, nor did the proportion of patients with severe disease. CONCLUSION: We found a statistically significant increase in the rate of HDP during the COVID-19 pandemic. However, there was no change in the proportion of severe disease, suggesting that this increase did not significantly impact clinical morbidity. KEY POINTS: · Rates of HDP increased during the COVID-19 pandemic.. · There was no change in the proportion of severe HDP.. · HDP-related maternal/neonatal morbidity was unchanged..

Performance of a Maternal Risk Stratification System for Predicting Low Apgar Scores.

OBJECTIVE: Maternal risk stratification systems are increasingly employed in predicting and preventing obstetric complications. These systems focus primarily on maternal morbidity, and few tools exist to stratify neonatal risk. We sought to determine if a maternal risk stratification score was associated with neonatal morbidity. STUDY DESIGN: Retrospective cohort study of patients with liveborn infants born at ≥24 weeks at four hospitals in one health system between January 1, 2020, and December 31, 2020. The Expanded Obstetric Comorbidity Score (EOCS) is used as the maternal risk score. The primary neonatal outcome was 5-minute Apgar <7. Logistic regression models determined associations between EOCS and neonatal morbidity. Secondary analyses were performed, including stratifying outcomes by gestational age and limiting analysis to "low-risk" term singletons. Model discrimination assessed using the area under the receiver operating characteristic curves (AUC) and calibration via calibration plots. RESULTS: A total of 14,497 maternal-neonatal pairs were included; 236 (1.6%) had 5-minute Apgar <7; EOCS was higher in 5-minute Apgar <7 group (median 41 vs. 11, p < 0.001). AUC for EOCS in predicting Apgar <7 was 0.72 (95% Confidence Interval (CI) 0.68, 0.75), demonstrating relatively good discrimination. Calibration plot revealed that those in the highest EOCS decile had higher risk of neonatal morbidity (7.6 vs. 1.7%, p < 0.001). When stratified by gestational age, discrimination weakened with advancing gestational age: AUC 0.70 for <28 weeks, 0.63 for 28 to 31 weeks, 0.64 for 32 to 36 weeks, and 0.61 for ≥37 weeks. When limited to term low-risk singletons, EOCS had lower discrimination for predicting neonatal morbidity and was not well calibrated. CONCLUSION: A maternal morbidity risk stratification system does not perform well in most patients giving birth, at low risk for neonatal complications. The findings suggest that the association between EOCS and 5-minute Apgar <7 likely reflects a relationship with prematurity. This study cautions against intentional or unintentional extrapolation of maternal morbidity risk for neonatal risk, especially for term deliveries. KEY POINTS: · EOCS had moderate discrimination for Apgar <7.. · Predictive performance declined when limited to low-risk term singletons.. · Relationship between EOCS and Apgar <7 was likely driven by prematurity..

Regional Variation in Antenatal Late Preterm Steroid Use Following the ALPS Trial.

Importance: The publication of the Antenatal Late Preterm Steroids (ALPS) trial in February 2016 demonstrated that antenatal administration of betamethasone in the late preterm period (between 34 to 36 weeks of gestation) for individuals with a high risk of delivery decreased neonatal respiratory morbidity. National estimates have suggested the trial did change obstetric practice, but little is known if the evidence was adopted uniformly or equitably. Objective: To assess regional variation in the use of late preterm steroids after the publication of the Antenatal Late Preterm Steroids (ALPS) Trial and to understand factors associated with a region's pace of adoption. Design, Setting, and Participants: This cross-sectional study used US natality data from February 2015 to October 2017 from hospital referral regions (HRRs) within the US. Inclusion criteria included live-born, nonanomalous, singleton, late preterm (34 to 36 completed weeks of gestation) neonates born to individuals without pregestational diabetes. This study was conducted from November 15, 2022, to January 13, 2023. Main Outcome and Measures: HRRs were categorized as either a slower adopter or faster adopter of antenatal late preterm steroids based on the observed vs expected pace of antenatal steroid adoption in a 1-year period after the trial's dissemination. Patient and regional factors hypothesized a priori to be associated with the uptake of late preterm steroids were compared between faster and slower adopters. Comparisons were made using Student t test or Wilcoxon rank-sum test, as appropriate. A multivariable logistic regression was constructed to identify factors associated with faster adopter status in the postperiod. Results: There were 666 097 late preterm births in 282 HRRs. The mean (SD) maternal age in HRRs was 27.9 (1.2) years. The median (IQR) percentage of births by race categories in HRRs for patients identifying as American Indian or Alaskan Native was 0.5% (0.2%-1.3%); Asian or Pacific Islander, 3.0% (1.7%-5.3%); Black, 12.9% (5.1%-29.1%); and White, 78.6% (66.6%-87.0%). The median percentage of births in HRRs to patients of Hispanic ethnicity was 11.2% (6.3%-27.4%). In this study, 136 HRRs (48.2%) were classified as faster adopters and 146 (51.8%) were classified as slower adopters. Faster adopters increased their steroid use by 12.1 percentage points (from 5.9% to 18.0%) compared with a 5.5 percentage point increase (from 3.7% to 9.2%) among slower adopters (P < .001). Most examined patient and regional factors were not associated with a region's pace of adoption, with the exception of the regional prevalence of prior preterm birth (adjusted odds ratio [aOR], 2.04 [95% CI, 1.48-2.82]) and the percentage of deliveries at 34 to 35 weeks of gestation (aOR, 0.68 [95% CI, 0.47-0.99]) compared with 36 weeks. Conclusions and Relevance: In this cross-sectional study, there was widespread geographic variation in the adoption of antenatal steroid administration for late preterm births that largely remained unexplained by population factors. These findings should prompt further investigations to barriers to timely or equitable access to new evidence-based practices and guide future dissemination strategies with the goal of more uniform adoption.

External Validation of a Prediction Model for External Cephalic Version Success.

External cephalic version (ECV) success correlates with numerous maternal and pregnancy factors. A prior study developed an ECV success prediction model based on body mass index, parity, placental location, and fetal presentation. We performed external validation of this model using a retrospective cohort of ECV procedures from a separate institution between July 2016 and December 2021. Four hundred thirty-four ECV procedures were performed, with a 44.4% success rate (95% CI 39.8-49.2%), which was similar to the derivation cohort (40.6%, 95% CI 37.7-43.5%, P =.16). There were significant differences in patients and practices between cohorts, including the rate of neuraxial anesthesia (83.5% derivation cohort vs 10.4% our cohort, P <.001). The area under the receiver operating characteristic curve (AUROC) was 0.70 (95% CI 0.65-0.75), which was similar to that in the derivation cohort (AUROC 0.67, 95% CI 0.63-0.70). These results suggest the published ECV prediction model's performance is generalizable outside the original study institution.

The Application of a Standard Risk Threshold for the Stratification of Maternal Morbidity among Population Subgroups.

OBJECTIVE: The aim of this study was to determine if a universally applied risk score threshold for severe maternal morbidity (SMM) resulted in different performance characteristics among subgroups of the population. STUDY DESIGN: This is a retrospective cohort study of deliveries that occurred between July 1, 2016, and June 30, 2020, in a single health system. We examined the performance of a validated comorbidity score to stratify SMM risk in our cohort. We considered the risk score that was associated with the highest decile of predicted risk as a "screen positive" for morbidity. We then used this same threshold to calculate the sensitivity and positive predictive value (PPV) of this "highest risk" designation among subgroups of the overall cohort based on the following characteristics: age, race/ethnicity, parity, gestational age, and planned mode of delivery. RESULTS: In the overall cohort of 53,982 women, the C-statistic was 0.755 (95% confidence interval [CI], 0.741-0.769) and calibration plot demonstrated that the risk score was well calibrated. The model performed less well in the following groups: non-White or Hispanic (C-statistic, 0.734; 95% CI, 0.712-0.755), nulliparas (C-statistic, 0.735; 95% CI, 0.716-0.754), term deliveries (C-statistic, 0.712; 95% CI, 0.694-0.729), and planned vaginal delivery (C-statistic, 0.728; 95% CI, 0.709-0.747). There were differences in the PPVs by gestational age (7.8% term and 29.7% preterm) and by planned mode of delivery (8.7% vaginal and 17.7% cesarean delivery). Sensitivities were lower in women who were <35 years (36.6%), non-White or Hispanic (40.7%), nulliparous (38.9%), and those having a planned vaginal delivery (40.9%) than their counterparts. CONCLUSION: The performance of a risk score for SMM can vary by population subgroups when using standard thresholds derived from the overall cohort. If applied without such considerations, such thresholds may be less likely to identify certain subgroups of the population that may be at increased risk of SMM. KEY POINTS: · Predictive risk models are helpful at condensing complex information into an interpretable output.. · Model performance may vary among different population subgroups.. · Prediction models should be examined for their potential to exacerbate underlying disparities..