Evaluating the Usability and Practicality of AI-Enabled Smartphone-based Obstetric Ultrasound in Sierra Leone: A Mixed-Methods Study

Abstract Background: In Sierra Leone, access to WHO-recommended obstetric ultrasound before 24 weeks of gestation is limited. The introduction of Artificial Intelligence (AI)-enabled smartphone-based obstetric ultrasound offers potential for improving accessibility and quality of maternal healthcare. This study, conducted in 2023, evaluates the usability, practicality, and sustainable adoption of this technology in the Tonkolili district, where it has been implemented as part of a pilot program since 2020. Methods: A mixed-methods approach was used, analyzing quantitative data from 2,315 obstetric scans performed by 81 healthcare providers across seven health facilities in Tonkolili district from October 2020 to March 2023. Data were analyzed using Stata version 17, applying a mixed-effects logistic regression model to identify factors influencing scan quality. Primary qualitative data were collected through interviews, focus group discussions, and observations with healthcare providers and clients. The Non-Adoption, Abandonment, Scale-up, Spread, and Sustainability (NASSS) framework informed the topic guide and guided qualitative analysis, which was conducted using NVIVO software. Results: The study revealed that 83.8% of obstetric scans were of sufficient quality for remote or AI analysis, with 59% of healthcare providers achieving the 80% threshold of sufficient quality scans, based on data distribution. Tutorial videos (OR: 14.1, 95% CI: 6.40-31.12, p < 0.001) and healthcare provider's skill sets were key factors in improving scan quality. The device was easily adopted by healthcare providers and well-received by clients. However, infrastructure limitations, supply chain issues, lack of comprehensive guidelines, and funding constraints were identified as barriers to its broader impact. Conclusion: AI-enabled obstetric ultrasound is a practical tool for healthcare providers in Sierra Leone. With the help of tutorial videos, healthcare providers were able to perform quality scans despite varying skill levels. The device’s simplicity and ability to produce clear, analysable scans make it a valuable tool for improving maternal healthcare at the PHC level. However, challenges remain, including the need for more comprehensive training, better integration of ultrasound findings into broader care, and infrastructure improvements. Addressing these barriers and ensuring proper guidelines will enhance its effectiveness in improving maternal and child health outcomes.