Strengthening Maternal Mortality Review Committees with Better Data

Reliable data and structured reviews improve recommendations—and save lives when follow‑through is real. Begin by naming outcomes and definitions in plain language; the checklist in choosing outcomes that matter aligns committees and facilities. Because reviews rely on routine clinical and vital records, set expectations with real‑world evidence in healthcare decision‑making and tighten inputs using EHR data quality for real‑world evidence.

Case ascertainment and data preparation

Cast a wide net: hospital deaths, postpartum deaths in the community, emergency transfers, and deaths flagged by vital records. Reconcile across sources; publish match rates and “data notes.” Normalize key fields (gestational age, blood pressure units) and ensure timelines are credible.

Consistent abstraction and preventability criteria

Use a shared abstraction tool with clear definitions and required fields. Train abstractors and run inter‑rater reliability checks. Apply standardized preventability criteria and document rationale. Where free text is heavy, use light LLM assistance with human verification to extract structured fields—mirroring practices in AI for maternal health surveillance.

Review process that respects time and voice

Prepare concise case packets with timelines, key vitals, interventions, and social context (transport, interpreter need). Include summaries of respectful care concerns. In meetings, use disciplined facilitation that centers facts and avoids blame. Where helpful, include de‑identified community or family perspectives.

From findings to action

Translate recommendations into specific changes with owners, timelines, and resources: order sets, discharge checklists, interpreter‑first outreach, transport support, privacy improvements. Track adoption and outcomes monthly. For operational outreach, reuse the capacity‑matched design in AI for population health management.

Equity focus

Disaggregate cases and recommendations by language, neighborhood, facility type, and payer. If certain groups face recurring barriers, specify remedies and monitor results. For autonomy‑preserving counseling fixes, align with AI‑supported contraceptive counseling.

Case vignette: reducing severe postpartum hypertension deaths

Findings: delayed recognition and lack of interpreter support in two facilities.

Actions: interpreter‑first discharge scripts, same‑day postpartum BP checks, and home‑cuff programs. Results: day‑10 checks rise to 67%; severe events drop by 24%; families report better understanding and respect.

Common pitfalls (and fixes)

• Incomplete case capture → reconcile sources; report match rates and gaps.
• Vague recommendations → name owners, resources, and deadlines.
• No feedback loop → track adoption and outcomes; report monthly.
• Equity as an aside → disaggregate findings and remedies by subgroup.

Implementation checklist

• Publish plain‑language definitions and a shared abstraction tool.
• Reconcile cases across sources; normalize units and timelines.
• Train abstractors; check inter‑rater reliability; standardize preventability criteria.
• Translate recommendations into owned changes; monitor adoption and outcomes.
• Disaggregate and address inequities explicitly.