Maternal mortality in the U.S. is rising, and over 50% of these deaths occur after delivery—in the postpartum period, when structured care is minimal and risk signals are easily missed. Clinicians are often the last to know when a patient begins to deteriorate.

At Phia, we built the Maternal Outcomes Monitor (MOM) to change that. This AI-powered clinical decision support tool continuously monitors postpartum patients for high-acuity complications, enabling early, trajectory-aware intervention before symptoms spiral.

This system isn't theoretical. It runs in production today, supporting licensed care teams with risk scoring, alert generation, and timeline-modified triage logic. Here's how it works.

Clinical Background

Maternal mortality is most commonly associated with delivery, but the data show otherwise:

• 53% of pregnancy-related deaths occur between 7 and 365 days postpartum (CDC, 2020)

• Only 60% of commercially insured moms attend a postpartum visit within 8 weeks (PubMed, 2022)

• SMM-related readmissions cost over $19,000 per case and are 13x more likely in patients with known risk trajectories (Blue Cross Blue Shield Association)

Despite this, structured postpartum surveillance remains fragmented or absent. Our system fills that gap.

How the Maternal Outcomes Monitor (MOM) System Works

1. Persistent Memory-Based Monitoring

Every patient interaction—onboarding, chat, vitals, clinical scales—feeds into a longitudinal memory context engine. This allows our system to:

• Track symptom progression across days or weeks

• Detect re-emerging risk patterns (e.g., "I feel dizzy again")

• Correlate social and clinical signals to build full risk profiles

2. Mortality Trajectory Detection

Our AI continuously screens for 9 postpartum mortality trajectories:

• Hemorrhage

• Preeclampsia/eclampsia

• Cardiomyopathy

• Pulmonary embolism

• Sepsis

• Stroke

• Mental health crisis (suicide risk)

• Intimate partner violence (IPV)

• Substance use disorder (SUD) crisis

Each is defined by evidence-based symptom clusters and timing-weighted urgency curves aligned with CDC, ACOG, and NIH guidance.

3. Timeline-Aware Risk Scoring

Phia's detection model uses weighted scores adjusted for postpartum timing, trajectory type, baseline symptom changes, and contextual amplifiers (e.g., housing instability, IPV, substance use).

Example:
A patient reporting dizziness and weakness 3 days postpartum with prior hemorrhage risk receives a 3.0x urgency multiplier and critical alert within 4 hours.

Scoring is continuous and updates with each new data point. Unmanaged trajectories drive score acceleration, while resolution or care acknowledgment pauses progression.

4. Clinical Alerting and Care Plan Integration

When a risk score crosses threshold:

• A structured clinical summary is generated

• Trajectory type and score are surfaced

• Recommended actions are embedded

These alerts integrate into Phia's care dashboard, allowing RNs, care coordinators, and MDs to intervene immediately.

Built for Clinical Realities

This system accounts for:

• Low patient follow-through: We detect risk even if a patient doesn’t complete every form or respond perfectly.

• Indirect symptom language: Our LLM prompts detect temporal, non-clinical phrasing ("My heart's racing," "I'm so tired I can't move")

• Multiple conditions: Alerts are cross-scored for concurrent risk (e.g., IPV + SUD amplifies score by +40 pts)

All logic is explainable, traceable, and built for audit.

Evidence-Aligned Detection Models

We mapped our trajectories to national mortality data and risk factor weightings from:

• CDC Pregnancy Mortality Surveillance System

• ACOG Practice Bulletins on postpartum hypertension, hemorrhage, perinatal depression

• Peer-reviewed SMM cost and timing studies (e.g., HCCI, PubMed)

Every trajectory has a time-weighted decay curve grounded in real-world outcomes. For example:

• Preeclampsia: 3.0x urgency in first 48h postpartum, 2.0x through 6 weeks

• Mental health crisis: Persistent 1.5x urgency through 12 months postpartum

• IPV/SUD dual risk: +40-point amplifier based on co-morbidity mortality studies

Why It Works for Providers

Phia’s AI doesn’t replace clinicians—it makes them faster and safer.

• Prioritized alerts: No more inbox scanning for red flags. Our alerts rank by mortality probability.

• Context summaries: See the last 7 days of relevant data, not just the last message.

• Asynchronous care: Nurses, therapists, and NPs act on alerts without waiting for scheduled visits.

Final Note from the Team

The Maternal Mortality Risk Detection System is Phia’s core commitment to safer postpartum care. Built from clinical data, refined with real-world evidence, and operated with licensed oversight, it gives every care team the ability to act early, escalate fast, and reduce risk without increasing burden.

This is what clinical AI looks like when it’s built for outcomes—not automation.