Connecting risk prediction to intervention: evidence from a randomized trial testing patient navigation and education...

Racial inequities in maternal morbidity and mortality persist in the postpartum period. Postpartum hospital use (PHU), including readmissions and emergency department visits, occurs in 5% to 9% of births, with rates among Black individuals at the higher end of this range. Risk-targeted interventions may improve efficiency and equity by directing resources to those at greatest risk postpartum and addressing structural barriers to care. Building on prior prediction model development and community-engaged intervention design, we conducted a pilot randomized controlled trial using a validated PHU risk calculator to identify and enroll high-risk participants into a postpartum education and patient navigation intervention.

To evaluate feasibility and engagement with a postpartum educational and patient navigation intervention and generate preliminary estimates of effectiveness in reducing PHU within 30 days of delivery discharge.

We conducted a randomized controlled trial among Black individuals who gave birth at a Philadelphia academic medical center between February 2024 and April 2025. Recruitment was guided by a risk calculator derived from a previously validated prediction model for PHU. Study coordinators abstracted predictor data from medical charts; self-identified Black individuals with ≥5.5% predicted PHU were eligible (cut-point selected from validation). Participants were enrolled before discharge from delivery hospitalization and randomized 1:1 to receive a postpartum educational and patient navigation intervention or standard care. The intervention included bedside printed and electronic education materials, and patient navigation including an in-hospital educational session and support via text message and phone calls through 30 days postpartum. The primary outcome was hospital use within 30 days of discharge, ascertained via electronic medical record and self-report. Secondary outcomes included depressive symptoms, self-efficacy, and perceived access to care self-reported on surveys at 3 weeks postpartum. Feasibility measures, including recruitment yield, retention, and implementation challenges identified from study logs and research team debriefs.

Among 1051 individuals screened, 616 (58.6%) were eligible and approached for enrollment, and 220 were enrolled and randomized (110 intervention, 110 control). Use of the risk calculator resulted in a high social vulnerability cohort: 50.0% with annual household income <$45,000 and 14.2% unhoused. Clinical risk scores were similar in intervention and control groups. Coordinators reported challenges to risk calculator use, including time-intensive abstraction, discrepancies between variable definitions in administrative data (model development) and medical records (recruitment), and missing data. The overall 30-day cumulative incidence of PHU was 18.2%. Incidence was 20.0% among intervention participants and 16.4% among control participants (risk ratio [95% confidence interval]=1.2 [0.7-2.0]). Secondary outcomes were similar between groups. Among intervention participants, 79% reported that the educational materials were helpful and 61% reviewed them more than once.

Among Black participants at elevated risk, engagement with a postpartum patient navigation and education intervention was high, although preliminary estimates did not suggest reduced PHU. Risk-based recruitment identified a cohort with elevated PHU but posed challenges for real-time risk ascertainment. These findings can inform larger multisite trials of risk-targeted, multilevel postpartum interventions.