Implementation of artificial intelligence-enabled, obstetric point-of-care ultrasounds into routine antenatal care in Lusaka Province of Zambia.
Authors
Affiliations (6)
Affiliations (6)
- Department of Obstetrics and Gynecology, University of Virginia, Charlottesville, Virginia, USA.
- University of North Carolina Global Projects Zambia, LLC, Lusaka, Zambia.
- Department of Obstetrics and Gynecology, University of North Carolina at Chapel Hill, Chapel Hill, North Carolina, USA.
- Department of Nutrition, Gillings School of Global Public Health, University of North Carolina at Chapel Hill, Chapel Hill, North Carolina, USA.
- Women and Newborn Hospital, University Teaching Hospitals, Lusaka, Zambia.
- Assistant Director, Reproductive Health, Zambian Ministry of Health, Lusaka, Zambia.
Abstract
Access to obstetric ultrasound remains limited in many countries owing to the limited availability of equipment and specialist operators. Artificial intelligence (AI)-enabled obstetric point-of-care ultrasound (OB POCUS) may expand access to reliable gestational age (GA) dating within routine antenatal care (ANC). To evaluate the feasibility and clinical utility of implementing AI-enabled OB POCUS for GA assessment across four public-sector facilities in Lusaka Province, Zambia. We integrated routine AI-enabled OB POCUS into ANC using a training-of-trainers model, followed by standardized competency-based training for facility healthcare providers and ongoing mentorship. We abstracted data from facility ANC registers for women scanned between September 2024 and March 2025. Among patients with both last menstrual period (LMP)- and AI-enabled OB POCUS-based dating, we calculated the discrepancy between the two estimates and applied American College of Obstetricians and Gynecologists (ACOG) GA-specific thresholds to classify each patient according to whether the discrepancy was large enough to warrant revising the LMP-based estimated due date (EDD) to the AI-enabled OB POCUS-based EDD. We estimated the proportion meeting these re-dating criteria and used logistic regression to assess factors associated with meeting them. Among 1068 ultrasound recipients, 859 (80.4%) had both a recorded LMP and a documented AI-enabled OB POCUS-based EDD. Of these, the median EDD discrepancy was 11 days (IQR = 5-29 days), and 51% met the ACOG criteria for EDD revision. The proportion meeting re-dating criteria did not differ by GA at presentation and was not affected by facility or demographic characteristics. Implementation of AI-enabled OB POCUS within public-sector ANC in Zambia was feasible. Half of women scanned met the criteria for revising an LMP-based due date, indicating that such technologies could substantially change pregnancy care where ultrasound is otherwise unavailable.