Leveraging Machine Learning to Identify and Address Maternal Health Disparities in Underserved U.S. Communities

Abstract

This study has examined how machine learning has supported the identification of maternal health disparities and the delivery of timely responses within underserved U.S. communities. The problem has concerned persistent inequities in maternal outcomes that are intensified by fragmented healthcare data, insufficient integration of social determinants of health, limited algorithmic transparency, and uneven organizational capacity to translate predictive signals into effective care actions. The study has aimed to determine the effects of machine-learning capability, healthcare-data quality, social determinants of health integration, algorithmic fairness and transparency, and organizational readiness on maternal health disparity identification and response effectiveness. A quantitative, cross-sectional, multiple-case-study design has been used across four maternal-care organizational cases: an urban safety-net hospital, a rural maternal-health network, community health centers, and public or nonprofit maternal-care programs. The final sample has included 312 valid professional responses from 380 distributed questionnaires, producing an effective response rate of 82.1%. Participants have included obstetric, nursing, midwifery, public-health, informatics, data, information-technology, and administrative professionals. Data have been collected through a 30-item five-point Likert-scale questionnaire and analyzed using descriptive statistics, Cronbach’s alpha, exploratory factor analysis, Pearson correlation, multiple regression, analysis of variance, and diagnostic tests. The instrument has demonstrated strong reliability, with an overall Cronbach’s alpha of .93, construct coefficients between .85 and .91, KMO = .89, and Bartlett’s test χ²(435) = 4,286.37, p < .001. Machine-learning capability has recorded the highest predictor mean, M = 3.89, SD = 0.62, while maternal health disparity identification and response effectiveness has recorded M = 3.83, SD = 0.61. All five predictors have shown positive relationships with the outcome. The regression model has explained 61.0% of outcome variance, R² = .610, adjusted R² = .604, F(5, 306) = 95.72, p < .001. Machine-learning capability has been the strongest predictor, β = .29, followed by organizational readiness, β = .24, healthcare-data quality, β = .20, algorithmic fairness and transparency, β = .16, and social determinants integration, β = .12. The findings have implied that equitable maternal-care improvement requires technically capable systems combined with data quality, fairness governance, social-context integration, and organizational readiness.