More than 80% of pregnancy-related deaths in the United States are preventable. Yet maternal mortality continues to rise, particularly among Medicaid beneficiaries who account for nearly half of all U.S. births. The challenge isn't a lack of effective interventions—community-based care coordination, home visiting programs, and enhanced prenatal services have all proven successful. The problem is timing.

Current risk stratification tools wait for clinical red flags to appear in claims data – but by the time these indicators show up, opportunities for early prevention have often passed. For pregnant patients with limited or late prenatal care access, this reactive approach systematically misses those at highest risk.

A New Approach to Early Detection

Today, we're announcing Signal for Maternity Risks, the newest addition to the Waymark Signal™ platform. Published in Nature's npj Digital Public Health, our peer-reviewed research demonstrates that Signal for Maternity Risks predicts adverse pregnancy outcomes with 89.4% accuracy – a median of 55 days earlier than traditional clinical indicators emerge.

Our study analyzed data from 190,698 pregnant women receiving Medicaid benefits across 26 states and Washington, D.C. We found that individual clinical factors alone poorly predicted outcomes. Instead, adverse pregnancy outcomes arose from complex interactions between biological, social, and structural factors, all combinations that traditional screening approaches miss.

Signal for Maternity Risks integrates comprehensive clinical data (including pharmacy claims, emergency visits, and historical pregnancy outcomes) with social determinants of health information such as healthcare workforce availability, distance to obstetric care, and community infrastructure. This integration achieved 81.3% sensitivity in identifying women who would experience adverse outcomes including preterm birth, severe maternal morbidity, or neonatal intensive care unit admission.

Perhaps most significantly, Signal for Maternity Risks eliminates algorithmic bias. Our baseline model using only clinical data showed lower sensitivity for Black patients compared to white patients (71.5% vs. 73.0%). Incorporating social determinants of health achieved equivalent performance for both groups (81.3% vs. 81.5%).

This drives an extremely high impact; Black women face three times higher maternal mortality rates than white women, disparities that persist regardless of education or income level. Our simulation showed that addressing modifiable social determinants could reduce adverse pregnancy outcomes by 31.8%, with the greatest absolute benefit for Black women.

Signal for Maternity Risks joins our Signal Suite of predictive models designed specifically for Medicaid populations:

• Signal for Rising Risk identifies which Medicaid patients need proactive support before acute events occur. Published in Nature Scientific Reports, our research demonstrates >90% accuracy in predicting future avoidable hospital and emergency department visits.
• Signal for Quality Improvement demonstrated 85% accuracy in predicting which patients will benefit from proactive outreach. In our analysis of 14 million Medicaid enrollees, this precision targeting approach increased preventive care gap closure by 35% compared to traditional methods.
• Signal for Dual-Eligible Populations tackles the complexity of Medicare-Medicaid beneficiaries by integrating patient, provider, facility, and area-level social determinants. Signal demonstrates 80% accuracy in predicting future avoidable acute care, with 75% improvement in cost prediction over traditional risk models. 

Together, these tools enable health plans and providers to shift from reactive crisis management to proactive, early intervention. For the 40 million women of reproductive age covered by Medicaid, this represents a fundamental change in how we approach maternal health – a shift from responding to complications after they occur to preventing them.

Read the full paper in npj Digital Public Health.