The impact of artificial intelligence-driven point-of-care ultrasound (AI-POCUS) on antenatal care and maternal-newborn outcomes in Ethiopia

Objective This study investigates the application of Artificial Intelligence-Driven Point-of-Care Ultrasound (AI-POCUS) in antenatal care (ANC) services within low-resource settings, examining its effects on ANC adherence, referral decisions, and maternal and neonatal outcomes. The findings will provide empirical evidence to support the integration of digital interventions into primary healthcare systems. Methods A total of 706 pregnant women from the North Shewa Zone in the Amhara Region of Ethiopia were enrolled in the study between November 2024 and February 2025 using a consecutive sampling method and subsequently categorized into two groups based on their receipt of AI-POCUS examinations (AI-POCUS group, n=108; control group, n=598). Data were collected through face-to-face structured questionnaires, medical record extraction, and maternal follow-up, and differences in sociodemographic characteristics, referral patterns, and maternal and neonatal outcomes were compared between groups. Multivariate analyses using Firth logistic regression were performed to explore associations between AI-POCUS exposure and adequate ANC visits, maternal complications, and neonatal health outcomes. Results Compared with the control group, women in the AI-POCUS group were more likely to reside in urban areas, have employed occupations, and have partners with lower educational attainment ( P <0.001), while the proportion of elderly mothers (Age ≥ 35 years) was significantly lower ( P =0.042). No statistically significant differences were observed between the two groups in maternal age, marital status, or obstetric history ( P >0.05). Among those who underwent ultrasound examination, the referral rate during the second trimester (19.67%) was markedly higher than in the first (0%) and third (4.55%) trimesters. Women who received AI-POCUS were 4.92 times more likely to achieve the WHO -recommended number of ANC visits than those in the control group ( aOR = 4.923, 95% CI : 2.863-8.465, P <0.001). However, no statistically significant associations were observed between AI-POCUS use and maternal complications, neonatal complications, or neonatal mortality ( P >0.05). Conclusions AI-POCUS can serve as an important tool for ANC in low-resource settings, with potential value in improving ANC visits, strengthening risk management, and facilitating timely referrals. When scaling up digital health interventions for ANC, socioeconomic and cultural differences should be carefully considered to foster the development of sustainable and replicable models of digital perinatal care.