Maternal health decision support
County-level access to risk-appropriate maternal care across the United States: drive time to each ACOG level, maternity care deserts and the effect of distance on postpartum complications.
Key terms
- Levels of maternal care
ACOG and SMFM define four levels of maternal care, from basic (Level I) to regional perinatal health care centers (Level IV), so that each pregnant patient delivers where her risks can be managed.
- Maternity care desert
A maternity care desert is a county with no hospital providing obstetric care, no birth center and no obstetric clinicians, a definition popularized by the March of Dimes.
- Calibration of a prediction model
A calibrated model's predicted risks match observed rates: of patients given a 20% risk, about 20% have the event. Calibration matters as much as discrimination for clinical use.
- SHAP (Shapley additive explanations)
SHAP explains an individual prediction by splitting it into contributions from each input feature, based on Shapley values from game theory.
- Graph neural network (GraphSAGE)
A graph neural network learns from data arranged as a network. GraphSAGE builds each node's representation from its own features and a sample of its neighbors' features.
- Double machine learning (DoubleML)
Double machine learning estimates a causal effect from observational data by using machine learning to remove the influence of confounders, with cross-fitting to avoid overfitting bias.
Research and educational tools. Not for clinical decision-making.