Issue of Data Imbalance on Low Birthweight Baby Outcomes Prediction and Associated Risk Factors Identification: Establishment of Benchmarking Key Machine Learning Models With Data Rebalancing Strategies.

BACKGROUND: Low birthweight (LBW) is a leading cause of neonatal mortality in the United States and a major causative factor of adverse health effects in newborns. Identifying high-risk patients early in prenatal care is crucial to preventing adverse outcomes. Previous studies have proposed various machine learning (ML) models for LBW prediction task, but they were limited by small and imbalanced data sets. Some authors attempted to address this through different data rebalancing methods. However, most of their reported performances did not reflect the models' actual performance in real-life scenarios. To date, few studies have successfully benchmarked the performance of ML models in maternal health; thus, it is critical to establish benchmarks to advance ML use to subsequently improve birth outcomes. OBJECTIVE: This study aimed to establish several key benchmarking ML models to predict LBW and systematically apply different rebalancing optimization methods to a large-scale and extremely imbalanced all-payer hospital record data set that connects mother and baby data at a state level in the United States. We also performed feature importance analysis to identify the most contributing features in the LBW classification task, which can aid in targeted intervention. METHODS: Our large data set consisted of 266,687 birth records across 6 years, and 8.63% (n=23,019) of records were labeled as LBW. To set up benchmarking ML models to predict LBW, we applied 7 classic ML models (ie, logistic regression, naive Bayes, random forest, extreme gradient boosting, adaptive boosting, multilayer perceptron, and sequential artificial neural network) while using 4 different data rebalancing methods: random undersampling, random oversampling, synthetic minority oversampling technique, and weight rebalancing. Owing to ethical considerations, in addition to ML evaluation metrics, we primarily used recall to evaluate model performance, indicating the number of correctly predicted LBW cases out of all actual LBW cases, as false negative health care outcomes could be fatal. We further analyzed feature importance to explore the degree to which each feature contributed to ML model prediction among our best-performing models. RESULTS: We found that extreme gradient boosting achieved the highest recall score-0.70-using the weight rebalancing method. Our results showed that various data rebalancing methods improved the prediction performance of the LBW group substantially. From the feature importance analysis, maternal race, age, payment source, sum of predelivery emergency department and inpatient hospitalizations, predelivery disease profile, and different social vulnerability index components were important risk factors associated with LBW. CONCLUSIONS: Our findings establish useful ML benchmarks to improve birth outcomes in the maternal health domain. They are informative to identify the minority class (ie, LBW) based on an extremely imbalanced data set, which may guide the development of personalized LBW early prevention, clinical interventions, and statewide maternal and infant health policy changes.

Group Prenatal Care Results in Medicaid Savings with Better Outcomes: A Propensity Score Analysis of CenteringPregnancy Participation in South Carolina.

Objectives This study was undertaken to determine the cost savings of prevention of adverse birth outcomes for Medicaid women participating in the CenteringPregnancy group prenatal care program at a pilot program in South Carolina. Methods A retrospective five-year cohort study of Medicaid women was assessed for differences in birth outcomes among women involved in CenteringPregnancy group prenatal care (n = 1262) and those receiving individual prenatal care (n = 5066). The study outcomes examined were premature birth and the related outcomes of low birthweight (LBW) and neonatal intensive care unit (NICU) visits. Because women were not assigned to the CenteringPregnancy group, a propensity score analysis ensured that the inference of the estimated difference in birth outcomes between the treatment groups was adjusted for nonrandom assignment based on age, race, Clinical Risk Group, and plan type. A series of generalized linear models were run to estimate the difference between the proportions of individuals with adverse birth outcomes, or the risk differences, for CenteringPregnancy group prenatal care participation. Estimated risk differences, the coefficient on the CenteringPregnancy group indicator variable from identity-link binomial variance generalized linear models, were then used to calculate potential cost savings due to participation in the CenteringPregnancy group. Results This study estimated that CenteringPregnancy participation reduced the risk of premature birth (36 %, P < 0.05). For every premature birth prevented, there was an average savings of $22,667 in health expenditures. Participation in CenteringPregnancy reduced the incidence of delivering an infant that was LBW (44 %, P < 0.05, $29,627). Additionally, infants of CenteringPregnancy participants had a reduced risk of a NICU stay (28 %, P < 0.05, $27,249). After considering the state investment of $1.7 million, there was an estimated return on investment of nearly $2.3 million. Conclusions Cost savings were achieved with better outcomes due to the participation in CenteringPregnancy among low-risk Medicaid beneficiaries.

Association of Hospital Incentive Care Management Partnerships for Uninsured Patients With Emergency Department Utilization.

Importance: The South Carolina (SC) Healthy Outcomes Plan (HOP) program aimed to expand access to health care to individuals without insurance; it remains unknown whether there is an association between the SC HOP program and emergency department (ED) use among patients with high health care costs and needs. Objectives: To determine whether participation in the SC HOP was associated with reduced ED utilization among uninsured participants. Design, Setting, and Participants: This retrospective cohort study included 11 684 HOP participants (ages 18-64 years) with at least 18 months of continuous enrollment. Generalized estimating equations and segmented regression of interrupted time-series analyses of ED visits and charges were conducted from October 1, 2012, to March 31, 2020. Exposures: Time intervals related to the HOP were 1 year before and 3 years after participation. Main Outcomes and Measures: ED visits per 100 participants per month and ED charges per participant per month overall and by subcategory. Results: The mean (SD) age of the 11 684 participants in the study was 45.2 (10.9) years; 6293 (54.5%) were women; 5028 (48.4%) were Black participants and 5189 (50.0%) were White participants. Over the study period, the mean (SE) number of ED visits decreased by 44.1%, from 48.1 (5.2) to 26.9 (2.8) per 100 participants per month. The mean (SE) ED charges were reduced to $858 ($46) per participant per month, a decrease from a mean (SE) of $1583 ($88) per participant per month 1 year before HOP implementation. There was an immediate level decrease of 40% (relative risk [RR], 0.61; 99.5% CI, 0.48-0.76; P < .001) from the preenrollment period, with a sustained reduction trend of 8% (RR 0.92; 99.5% CI, 0.89-0.95; P < .001) during the postenrollment period. A level change for ED charges was detected, at a decrease of 40% (RR 0.60; 99.5% CI, 0.47-0.77; P < .001) directly after HOP enrollment with a subsequent downward trend of 10% (RR 0.90; 99.5% CI, 0.86-0.93; P < .001) for the postenrollment period. Conclusions and Relevance: In this retrospective cohort study, proportions and charges of ED visits by uninsured patients saw immediate and sustained decreases after HOP enrollment. Reducing ED charges may have been driven by decreasing the ED as the primary point of patient care, especially for high-frequency users. These findings have implications for other nonexpansion states seeking to maximize uninsured compensation for low-income populations through improved outcomes.