Black-White Differences in Chronic Stress Exposures to Predict Preterm Birth: Interpretable, Race/Ethnicity-Specific...

Area of Science:

• Reproductive Health
• Machine Learning in Healthcare
• Health Disparities

Background:

• Preterm birth (PTB) remains a significant public health concern with complex contributing factors.
• Existing predictive models may not fully capture the nuances of PTB risk across different racial/ethnic groups.
• Identifying specific chronic stressors is crucial for targeted PTB prevention strategies.

Purpose of the Study:

• To develop accurate machine learning models for predicting preterm birth (PTB).
• To identify key predictors of PTB, distinguishing between non-Hispanic (N-H) Black and N-H White women.
• To pinpoint the influence of chronic stressors on PTB risk within these populations.

Main Methods:

• Multivariate Adaptive Regression Splines (MARS) machine learning models were developed.
• Models utilized data from the Pregnancy Risk Assessment Monitoring System (2012-2017).
• Model performance was assessed using 5-fold cross-validation and Area Under the Curve (AUC).

Main Results:

• MARS models demonstrated high predictive accuracy for PTB (AUC: 0.754-0.765).
• Key predictors across populations included prenatal care visits, premature rupture of membranes, and medical conditions.
• Chronic stressors (e.g., low maternal education, violence) significantly impacted PTB prediction specifically for N-H Black women.

Conclusions:

• Interpretable, race/ethnicity-specific MARS models accurately predict PTB.
• The models elucidate the magnitude of effect of life stressors on PTB risk.
• Findings support the development of tailored interventions addressing social determinants of health for PTB prevention.