A Qualitative Inquiry Exploring Perceptions of Artificial Intelligence to Improve Outcomes in Maternal, Sexual and Reproductive Health in Sub-Saharan Africa

A Qualitative Inquiry Exploring Perceptions of Artificial Intelligence to Improve Outcomes in Maternal, Sexual and Reproductive Health in Sub-Saharan Africa | Research Square window.SnipcartSettings = { analytics: { enabled: false } }; (function() { var accessVector = localStorage.getItem('access_vector') || ''; window.dataLayer = window.dataLayer || []; if (accessVector) { window.dataLayer.push({ user: { profile: { profileInfo: { snid: accessVector } } } }); } })(); (function(w,d,s,l,i){w[l]=w[l]||[];w[l].push({'gtm.start':new Date().getTime(),event:'gtm.js'});var f=d.getElementsByTagName(s)[0],j=d.createElement(s),dl=l!='dataLayer'?'&l='+l:'';j.async=true;j.src='https://www.googletagmanager.com/gtm.js?id='+i+dl;f.parentNode.insertBefore(j,f);})(window,document,'script','dataLayer','GTM-K279D39R'); Browse Preprints In Review Journals COVID-19 Preprints AJE Video Bytes Research Tools Research Promotion AJE Professional Editing AJE Rubriq About Preprint Platform In Review Editorial Policies Our Team Advisory Board Help Center Sign In Submit a Preprint Cite Share Download PDF Research Article A Qualitative Inquiry Exploring Perceptions of Artificial Intelligence to Improve Outcomes in Maternal, Sexual and Reproductive Health in Sub-Saharan Africa Rachel King, Elizabeth Oseku, Cecilia Akatukwasa, Moreen Nanyonjo, and 4 more This is a preprint; it has not been peer reviewed by a journal. https://doi.org/ 10.21203/rs.3.rs-7463408/v1 This work is licensed under a CC BY 4.0 License Status: Posted Version 1 posted You are reading this latest preprint version Abstract Artificial intelligence has the potential to transform healthcare in low- and middle-income countries, where access to quality care remains limited. Maternal, sexual, and reproductive health (MSRH) outcomes are especially poor due to resource shortages, financial barriers, and geographic inequities. With thoughtful implementation, AI could help address these gaps through innovations in diagnostics, health education chatbots, and telemedicine. However, responsible use is essential to ensure AI reduces—rather than exacerbates—health disparities between high- and low-income regions. Our study examines the perceptions, uses, benefits, and challenges of AI in MSRH among medical professionals, community members, and AI experts, guided by the Diffusion of Innovations Theory. Our findings will inform the development of a continent-wide AI hub for MSRH, highlighting barriers and opportunities for improving health care access. We aim to support policymakers, researchers, and implementers in using AI to promote equitable maternal and sexual and reproductive healthcare delivery across Africa. Artificial Intelligence and Machine Learning Figures Figure 1 INTRODUCTION The World Health Organization (WHO) estimates that poor reproductive health accounts for up to 18% of the global burden of disease, and 32% of the total burden of disease for women of reproductive age. 1 In Sub-Saharan Africa, death and disability resulting from reproductive health causes remain unacceptably high. The adult lifetime risk of maternal death has been estimated to be highest in Africa (1 in 26), while developed countries have been estimated to have the smallest lifetime risk (1 in 7,300). The prevention and control of reproductive tract infections is another area of concern; for example, Congenital Syphilis is the second leading cause of preventable stillbirth globally, preceded only by malaria. 2 The global HIV response continues to be undermined by a multitude of complex, interrelated challenges that are aggravated by limited domestic resources, declining donor assistance, including the recent reorganization of US global health commitment, which threatens the sustainability of all critical health programmes in the region. Vulnerable populations such as adolescents in Africa continue to be more susceptible to STIs, HIV, unwanted pregnancies and unsafe abortions. 3 , 4 In the last five to ten years, there has been an explosion in available health data in Africa due to improved infrastructure for electricity and internet as well as the widespread adoption of digital health technologies. 5 – 7 This provides an opportunity for data driven strategies and innovations that previously did not exist. To increase technology-driven solutions, we must consider the human research and implementation capacity to enable this valuable resource to improve health of the extremely diverse population. Due to the rapidly emerging technology around AI, many health care workers and policymakers are not aware of the opportunities and limitations of these technologies. The use of AI in health in LMICs is now emerging across sub-Saharan Africa. There is consensus in the literature that there is potential in using AI in expanding and extending healthcare access, by contributing to early disease detection and prevention, increasing diagnostic capability and drug development, disease surveillance, stock and healthcare management as well as clinical decision-making. 8 Thus, our aim was to map knowledge and awareness of AI among individuals working in health across the continent. To help frame and understand how individuals, and groups understand and adopt new innovations, the Diffusion of Innovations Theory (DOI) first documented by Everett Rogers, has been used as a structure specifically for technology innovations. 9 – 11 DOI speculates that the spread of new ideas, technologies, or practices within a social (or health) system occurs in a predictable and systematic way. The theory states that individuals within a population adopt innovations at varying degrees and rates, categorized into: innovators, early adopters, early majority, late majority, and stragglers. These groups are characterized by differing levels of risk tolerance, social influence, and access to information. The theory further highlights the role of communication channels, social networks, and perceived attributes of the innovation (e.g., relative advantages, compatibility, complexity, trialability, observability) in facilitating or hindering its adoption. While the theory does not explicitly underscore issues of cost and access to the innovation, it implicitly considers access through trialability which refers to the extent to which an innovation can be tested on a limited basis before a full commitment is made. An important feature of the theory is observability referring to the degree to which the results of an innovation are visible ‘in the wild’ meaning in everyday life and therefore easily communicated to others. If individuals cannot observe others using the innovation and experiencing its benefits, they are less likely to be interested in adopting it themselves. The theory also includes economic and political elements. 9 – 11 The overall goal of this study is to explore the early experiences, perceptions of health workers, policymakers and general population, as well as AI researchers and implementers around the opportunities, risks, limitations, and best practices for responsible artificial intelligence applications in maternal, sexual and reproductive health (MSRH) in sub-Saharan Africa. Diffusion of Innovations theory provides a framework for understanding how innovations are adopted, highlighting knowledge, attitudes, barriers and recommendations for adoption of AI in MRSH in the future within the numerous environments in sSA. We inspect our data through the lens of the Diffusion of Innovations theory to explore factors influencing stakeholder engagement in AI applications. METHODS We conducted a cross-sectional qualitative study to explore perceptions of health workers, policymakers, AI researchers and implementers, as well as seven community Advisory Board (CAB) members. We delved into the opportunities, risks, limitations, and best practices for responsible AI in maternal, sexual and reproductive health (MSRH) in sub-Saharan Africa (SSA). The study was conducted within the activities of HASH - The Hub for Artificial Intelligence in Maternal, Sexual and Reproductive Health in Sub-Saharan Africa. HASH was formed in 2021 by a multidisciplinary consortium of the Infectious Diseases Institute, the Makerere University College of Computing and Information Science, Sunbird AI through funding from the International Development Research Centre (IDRC) and the Swedish International Development Cooperation Agency (SIDA), as part of the Artificial Intelligence for Development in Africa Program (AI4D Africa) and the Global South AI for Global Health (AI4GH) INIT. HASH’s objective is to advance Maternal, Sexual, and Reproductive Health (MSRH) and rights while strengthening health systems in Sub-Saharan Africa (SSA) through the responsible development and deployment of Artificial Intelligence (AI) innovations. The Hub is based in Uganda but targets both anglophone and francophone stakeholders at various levels in the field of AI and/or MSRH across SSA. This includes innovators, researchers, health workers, policy makers, organizations and enthusiasts working in the AI for MSRH space in sub-Saharan Africa (SSA). The Hub provides support for and gains insight into the use of new and existing data and how AI and data technology can be leveraged to solve African health challenges. Grounded in the ethical and responsible application of AI, the Hub provides capacity building through technical and methodological assistance to its members and shares opportunities for mentorship and collaboration. This study used the HASH platform to undertake purposeful sampling of thought leaders in AI in MSRH for targeted focus group discussion and in-depth interviews. Additionally, we collected data from routine stakeholder engagement consultations (scoping workshops) conducted to inform the strategy and implementation of HASH. Data collection Key Informant interviews: As part of HASH activities in 2021 we conducted an online survey of 107 respondents from 25 countries of MSRH experts and AI researchers across Africa. We included individuals with evidence of expert knowledge, including but not limited to: actively working on AI projects in healthcare, working maternal, sexual and reproductive health, familiar with the concepts and methods of AI, based in anglophone or francophone sub-Saharan Africa. The survey identified key areas of MSRH that respondents considered to be key research priorities for AI (Table 1). Table 1. Ranking of top research and development in MSRH according to potential and viability of AI as a solution, 2021, sub-Saharan Africa THEME PRIORITY AREA TO IMPROVE HEALTH OUTCOMES MATERNAL HEALTH Identification of high-risk conditions in mothers and newborns Attracting and retaining skilled birth attendants in remote areas Maternal mortality Improving uptake of proven existing interventions Algorithms and point of care diagnostic tests for sepsis in the face of emergencies and epidemics Implementation research of interventions that improve quality of care during labour The impact of training on skills of health workers handling obstetric emergencies Improving access to emergency transport in hard-to-reach communities Improved maternal and new-born emergency care ADOLESCENT SEXUAL REPRODUCTIVE HEALTH Unmet need for contraception Inform how health services can be designed to effectively meet adolescents' health needs. Development of prevention technologies besides condoms Gender gaps in sexual health Examining health related behaviour (physical activity, diet) Examining health seeking behaviour Prevalence/incidence of different health outcomes (STIs, pregnancy) Access to comprehensive SRH services Evaluation of effectiveness of health service delivery models Understanding needs of vulnerable adolescents (street children) SEXUALLY TRANSMITTED INFECTIONS 1. Screening and case finding of STIs 2. Early diagnosis and treatment of STI cases 3. STI-prevention strategies 4. Issues around antimicrobial resistance HIV/AIDS Tracing and reengagement in care Diagnostics that meet the specifications for use in LMICs, especially point-of-care diagnostics targeting lower levels facilities Assessing ART adherence Assessing frequency of care visits/refills Task sharing of specimen collection and point-of-care testing Integration of HIV and sexual reproductive health services Assessing ART initiation outside the health facility Generation of data on drug safety and efficacy for PLHIV Assessing integration of HIV and diabetes and hypertension care Developing and introducing new paediatric drug formulations for treatment of HIV/AIDS Developing psychosocial interventions for adolescents & youth Balanced integration of diagnostic services Procedures: Participant selection KII : During the online survey participants were asked if they could be contacted for further one-on-one interviews. Ten participants were recruited by analyzing the results of questions of the HASH online survey to gauge the level of expertise, experience and willingness to participate. 12 The respondents who met the eligibility criteria were selected through purposive sampling using the above criteria to participate in the key informant interviews (KIIs). Respondents were contacted via email or telephone and appointments made. Before the interview, the participant received a consent form via email. During the scheduled interview time, informed consent was administered by the interviewer and the participant asked to voluntarily sign the consent form with an electronic signature and email it to the interviewer before the interview began. Participant selection Focus Group Discussion (FGD) : Participants were invited to the FGD if they were a member of the Academy for Health Innovations Community Advisory Board (CAB) at the Infectious Diseases Institute in Uganda. The CAB is composed of ministry of health officials, a youth representative, health worker from a lower health facility, two religious leaders (Muslem; Catholic), local council leader, a microfinance administrator, and a civil society organizational representative. The CAB’s main role is to advise and facilitate dialogue between the community and research team. An appointment was made for the FGD with all the CAB members either face to face or on-line through an email or phone. The FGD was conducted in March 2023, with seven participants (four men and three women). All respondents were consented in the same manner as for the interviews. Data collection Interviews and FGD: A standardized semi-structured guide was developed for FGD and KII. The guide was developed to explore the perceptions of the stakeholders on the future of AI in health care in Africa including perceived benefits and limitations. Exploration of the role in MSRH was based upon the priority areas identified in the survey (Table 1). Each interview was held with two social scientists; one to guide the discussion and one note-taker. The process was interactive and took between 30 and 60 minutes depending on experience with AI. All KIIs/FGD were conducted in English as we did not have anyone preferring French or any other language when they were asked. The interviews were conducted either face-to-face or by Zoom with video on for the participant, moderator and note-taker throughout the interview to allow for recognition of non-verbal cues. The FGD was face-to-face and was audio recorded. Interviews were audio recorded and both interview and FGD recordings were kept in a secured file in an online repository with access restricted only to the study team members. Scoping workshops (n=3): Three multi-stakeholder workshops with participants across sub-Saharan Africa convened diverse voices including healthcare providers, AI innovators, venture capitalists, health educators, policy makers, young people (15-24 years), health entrepreneurs, and religious leaders to design a path forward regarding AI innovations and MSRH in sub-Saharan Africa. Participants were identified through stakeholder mapping using purposive selection aiming to identify representatives across the above categories. The workshops included presentations on responsible AI, on the HASH project and on AI in medicine to facilitate a level playing field for fruitful discussion. The sessions were designed to capture diverse stakeholder perspectives on the role of AI in MSRH. One full-day physical workshop and two half-day virtual workshops were held in March and April 2025. Each workshop convened participants from across MSRH ecosystem to foster rich, multidisciplinary dialogue in the sub-Saharan contexts. Four social scientists were present as participant observers to critically observe and document the full discussions, reactions, and content of each session. For the scoping workshops, the insights from stakeholders were structured around five guiding questions 1) Why do we need AI in Maternal, Sexual and Reproductive Health? 2) What strategies enable AI to meet the needs of all stakeholders in MSRH? 3) What are the concerns and anticipated challenges of AI in MSRH? 4) What are the overall implications and way forward? 5) How do we make AI a mainstream tool for health in your setting? All discussions were held in English and lasted for two hours and 30 minutes for virtual workshops and four hours for physical workshops (divided into two breakout sessions). All workshops were audio recorded. Analysis: FGD and KII audio recordings were transcribed word for word. Data from the FGDs and interviews was analyzed using the Framework Analysis approach. 13-15 The method produces highly structured outputs of summarized data and is particularly useful for large quantities of data. Framework analysis is commonly used for the thematic analysis of semi-structured transcripts as reflexivity, rigor and quality are integral and critical. It can be adapted for use with deductive, inductive, or combined types of qualitative analysis. The key stages of Framework Analysis include familiarization, identification of a thematic framework or a codebook, coding, charting and interpretation. We combined scoping workshop, KII and FGD data. The preliminary findings were presented for feedback on three occasions: inaugural AfricaAI conference 2023 in Rwanda; AI4GH meeting in Nairobi in November 2023; and to the Technology and Innovation unit at the UK Foreign, Commonwealth Development Office in January 2025. Ethics Statement Ethical approval was obtained from the Infectious Diseases Institute Research Ethics Committee (IDIREC REF 011/2022) and the Uganda National Council for Science and Technology (HS2356ES). The study was conducted in accordance with the protocol, GCP guidelines, the Declaration of Helsinki and all applicable local regulatory requirements and laws. All research team members had verified GCP certification. FINDINGS We include 44 total participants as described above from Uganda, Nigeria, Kenya, Tanzania, Ghana, Zambia, United Kingdom, Senegal, Spain and almost equal numbers of men and women with slightly more men. We included 11 participants in the first in-person scoping workshop in Uganda and 16 participants in the two virtual scoping workshops where we had participants from across the continent and one person from the UK. The FGD included seven participants. We describe our findings based on the four main elements in the diffusion of innovations theory: the innovation, communication channels, time, and the social system (Figure 1). For innovation in this case, we consider a cluster of potential or existing AI innovations, for the time element, we look specifically at the implementation or actualization of AI/MSRH projects. THE INNOVATION: Current and suggested uses of AI When asked about participants’ current and suggested uses of AI in MSRH, helping to target or add specificity to high priority health activities or patients, was an overarching theme. Some participants mentioned how useful AI has been in tracking logistics such as tracking supply chain products and triggering action specifically for security purposes and when/if the terrain or environment was challenging. Some participants also mentioned predictive modeling, how powerful chatbots are in personalizing care by answering health-related questions. Additionally, AI could be used to rapidly triage numerous queries from health service users to prioritize questions that require urgent responses. Participants highlighted that these questions and answers could then be used to develop datasets such as ultrasound imaging. We try to develop predictive models to deliver care more efficiently and to deliver the care really to the women who need it. So, there are some services we deliver as a blanket care to everybody but if we identify certain risk factors or certain special needs, then we think, predictive modeling through machine learning or through AI, can help us identify women or children who need specialized care, then we could deliver that care to only these clients which makes the service delivery more targeted because we do not need to provide these generalized messages. We would only provide messages that are relevant to that client. (KII, TZ) In this quote, participants suggest predicting particular health concerns mothers are likely to suffer from so that they deliver services and information that is tailored to patient needs, rather than providing the same general service and information to everyone. This helps in improving patient-centered care and in resource allocation. “with the Q and A functionality ; what we used to do is answer questions in the order in which they arrived. If… we got…. in a day [more than] 90 questions that were not urgent, and number 91 was something that needed immediate attention, we would have to answer the questions 1 to 90 before we even realize that 91 was something that needed immediate attention.” (KII, Kenya) Many participants in the scoping workshops as well as in the interviews and FGD emphasized that they felt that use of AI was not to replace current roles or individuals but to improve efficiency and enhance numerous medical services. Those who are working in AI and those that felt and expressed a window of opportunity for uses of AI are likely be the early adopters of AI. Benefits of AI General benefits that participants stressed included using AI to fill gaps particularly when skilled and experienced health personnel were not continuously available, especially in rural and hard to reach areas. It was mentioned that AI can compete with the best human skills and could, in principle, eliminate human error, bias and corruption, and therefore should support equity as well as reducing time and resources spent on care. Participants felt that with these advantages, we could save money and reduce workload, while increasing coverage and quality of health services. Many felt that AI can increase accurate health information and improve collaboration between providers. AI is competing with the best human skills and so it is able to get you more accurate support in terms of diagnosis and is able to do it more quickly.. . . . , in terms of being able to support with diagnosis in a timely manner which high accuracy, AI is quite instrumental, and it is low cost especially if it can come as open source. Bias too is eliminated. . . not totally eliminated. Biases are always there but when we give a lot of our data to AI algorithms, let’s say Nigeria’s or South Africa’s data, they tend to, over time, be able to predict more accurately our needs regarding to patients’ diagnosis . (KII, Nigeria) When asked about which populations may benefit the most from an increase in use of AI, it was highlighted that high literacy populations because AI innovations often use technology through text. Additionally, rural populations who may have limited access to conventional medical services, adolescents who adopt technology quickly and fishermen who are highly mobile so may not easily access health facilities. Though one participant from Nigeria stated that all populations will benefit from AI. fishermen, as they are so migratory because today they are here, tomorrow they are there (CAB FGD) Content areas that KII participants reported benefitting most from AI included maternal and newborn health as well as sexual and reproductive health. Within MSRH, specific areas revolved around monitoring of health conditions including, pregnancy, blood sugar for gestational diabetes, post-natal monitoring and pregnancy complications. There are times when the baby is having a temperature. It may not necessarily need to go to hospital. . . .It (AI) can give you some form of first aid that can do (in the moment). (KII, Nigeria). I think one of the things that AI can help (with) at community level is to develop software that can tell this mother about the dangers of the baby (CAB FGD) Some growth areas that participants noted for further development would be in using images to build databases to use natural language processing for later transfer into voice. What I discovered about that ultrasound is… I saw that they take pictures and if that picture is well scanned, you can create a whole dataset. . . . If you know certain computations of how they show up in those pictures, they can be able to train a machine learning model that uses deep learning … Not by looking at only images and deep learning but let us look at natural language processing . (KII, Uganda) COMMUNICATION CHANNELS are a key element in relation to adoption of new innovations. An overarching, though not a surprising result, was the lack of understanding and awareness of AI among health workers and the general community. From health workers worried about job security to patients new to digital tools, the importance of user education was a key finding. Ensuring AI meets people’s needs also requires capacity building and education. A scoping workshop speaker, who works closely with frontline health workers, pointed out that many providers and patients currently have limited understanding of AI. To prevent AI tools from sitting unused or being misused, investment in training is vital. “ Even the best AI tool is useless if health workers don’t know how to use it or don’t trust it .” (scoping workshop) Addressing fear around trust in AI would necessitate building confidence through education and transparent processes. Encouraging partnerships and enforcing collaboration through donor requirements may improve outcomes. One strategy discussed in a scoping workshop was to incorporate AI literacy into medical and nursing education, as well as offering continuous training for existing staff where AI systems would be introduced. Once health workers decide to adopt AI then clear user guidance is necessary to mitigate harm from inaccurate queries or incomplete data. Health workers should learn not just how to operate AI-driven devices or apps, but also how to interpret AI outputs critically and integrate them with clinical judgment. One participant noted the significance of training on “ how to prompt AI…so if you get incorrect information, you are going to go with that wrong information .” (scoping workshop) Likewise, community health educators could help familiarize the public with new AI tools and services (for example, teaching expectant mothers how to interact with an AI-driven messaging service that sends them prenatal care advice). The goal of these capacity-building efforts is to ensure that AI becomes a help rather than a hindrance in the workflow, and that communities feel empowered rather than intimidated by new technology. Other suggestions in the FGD included educating the general public about what AI is and what AI is not . The CAB members also suggested using AI in an integration process for both traditional and biomedical health care systems. Traditional birth attendants (TBAs) are highly valued in some communities in sub-Saharan Africa (SSA) and may be an important communication channel for AI in MSRH. For example, including traditional birth attendants into the awareness building so that they understand better which high risk health issues they should refer mothers to health facilities for. I think the best way is to involve the TBAs, first, to educate them and show them what they can do and what they cannot do. That can give them contacts which they can refer to . . . it’s [their strength] is about the etiquette; it’s about the customer care that they show (CAB FGD) Ensuring a cultural fit especially with respect to linguistic barriers was a key point highlighted in the scoping workshops. Recognizing that MSRH is deeply intertwined with cultural norms and sensitivities is paramount. AI systems, particularly in diverse contexts like in Africa where open discussions about sex are often taboo, all potential solutions must undergo thorough cultural reviews. The communication style, information delivery, and even the platform’s interface should be sensitive to local customs and communication patterns. Creating a sense of trust and comfort is essential, as demonstrated by the strategy of designing chatbots with a friendly and approachable persona. “Young people often lack trusted sources for sexual health information. An AI chatbot available on their phone 24/7 can answer questions accurately and anonymously, which is a huge step forward.” Many participants cited rural populations with limited English proficiency as an example for ensuring localized AI solutions. One scoping workshop speaker asked, “ Is it possible for me to deposit my Ateso [local language in Uganda] somewhere so that it can be utilized… ?” (scoping workshop) This underscores the urgent need for AI-driven tools that incorporate local languages and dialects, ensuring that critical MSRH information is accessible to all. Co-creation and inclusive design were mentioned as key strategies to tackling the language and culture concern. Several contributors advocated for a co-creation process involving health workers, community members, and technology developers. “ It’s important to have a co-creation process in the way that AI is developed, ” (scoping workshop) as one attendee mentioned. As noted in a scoping workshop, this cultural tailoring is vital when implementing AI in Africa, where stigma around sexual health influences care-seeking behaviors. “ …ensuring it’s relatable to the daily experiences of these (specific) communities [is a key to successful uptake].” (scoping workshop) Such collaboration ensures user-friendly designs that accurately reflect community knowledge, practices and ethical priorities. Additionally, when communities feel a sense of ownership over a tool, they are more likely to trust it and use it. In addition to the language and cultural concerns, several scoping workshop participants also touched on the importance of interdisciplinary collaboration to breaking down silos between technology developers and healthcare practitioners, social scientists, and the intended beneficiaries of the technology. One scoping workshop participant emphasized: “ We learned that when engineers sit with midwives and doctors, they come up with much more practical solutions. Neither can do it alone — it has to be a joint effort. ” Through collaboration, AI solutions are more likely to address real-world problems in a feasible way. One illustrative example from the discussion in a scoping workshop was a pilot project mentioned by one scoping workshop participant, where an AI tool for predicting postpartum hemorrhage risk was developed with direct input from obstetricians: the clinicians specified what risk factors they saw as red flags, and the engineers used those insights to train a model that aligned with clinical intuition. Such collaborative models help ensure the resulting AI is not a “black box” but something clinicians feel connected to. Participants in scoping workshops, FGDs and KIIs were all asked about priorities for research and development. Some suggestions highlighted the importance of communication channels. CAB members signaled the area of how to incorporate AI in building awareness for high-risk health issues at the community level to prevent emergency situations. Another CAB member suggested not to focus only on the mother, but to include other family members such as the male partner in pregnancy-related awareness raising. TIME, IMPLEMENTATION AND ACTUALIZATION in AI solutions: Challenges that impact time from innovation to diffusion Participants’ concerns revolved around six main areas including: ethics and regulation, cost, specifically internet accessibility costs and sustainability, trust in new innovations, access including population literacy, data and bias as well as cultural and religious barriers. One participant mentioned that like current use of Google, there could be a risk of increase in self-medication, if people use AI alone to inform their medical choices. in situations where you are supposed to go and see your doctor, you are relying on what your chatbot or your application is telling you. You misuse it. . . . It may lead to abuse of information, abuse of access to information you are supposed to use positively but people tend to abuse it . (KI Nigeria) Regulation and Ethical Concerns : Marginalized populations, including women in patriarchal societies, could face barriers to benefiting from AI. Confidentiality breaches were mentioned as a concern with use of AI. Regulatory frameworks have not caught up to the pace of innovations in many SSA countries. AI expert participants described some of the complexities ranging from data literacy and technology access bottlenecks. Data and Bias Issues: In addition, both community and expert participants highlighted that AI models can face limited data sources in low-research languages and regions, compounded by biases in existing datasets. This can lead to inequitable outcomes, especially in marginalized communities. Access to health data is not straightforward because of the sensitivity of the information; privacy is very important. Stakeholders do not always understand the importance of AI tools and hence are skeptical about having health data shared with developers. Lack of a central repository for data resulting in segmented data. There is need for data, otherwise AI cannot exist. Data acquisition is a challenge. Data entrants may not fully understand the concept of unique identifiers and may choose to use traceable data e.g. phone numbers as a unique ID. . Medical workers and data entrants may not understand how to encrypt data; lack of knowledge on data encryption may limit data sharing where it would otherwise be possible. We must be careful what data we train AI on because what it is fed is what it gives. Not all data is good for AI. In many cases the data being used was not initially collected for AI use, so there may be a lot of missing data elements or assumptions. Data acquisition takes a long time and causes a lot of frustration to developers, especially because it involves a lot of stakeholder engagement. (KII Uganda) Cost, including connectivity and Infrastructure: All groups of participants noted that limited internet access and inconsistent electricity in low-income regions could lead to either low uptake or low capacity to sustain these projects. can we sustain it? Do we have the people we are going to work with? Are they well versed with the technology (CAB FGD) Trust and Literacy: It is common across the literature that initiation of new innovations often instill hesitancy. This was found in our data as well that adopting AI instill some fear of change, concerns about job loss, and low education levels, specifically literacy levels and more specifically in areas and populations with low data literacy among some population groups including community health workers. Some participants mentioned concern that their community would not know how to use AI correctly, thus could endanger their health instead of benefitting. A lack of collaboration and data sharing between stakeholders hinders progress. Cultural and Religious Barriers: Some communities resist technological innovations due to religious beliefs or cultural norms. Additionally, the belief in the importance of human-to-human connection and the risk of losing that with increased use of AI. “ someone can misinterpret the guidelines or the information that AI is using to maybe give treatment or diagnosis and come up with a decision that will affect the patient in due course” . . . certain people believe that if I go to the health center and I don’t find Stella, I might not be treated well, I would rather go to someone I know is not even a health person but just talking to her or him might cure me .” (CAB FGD) SOCIAL SYSTEM Generally, within the Diffusion of Innovations Theory, one considers the social system as including social norms, cultural values, social networks, political factors like government policies, ethics, regulations, and infrastructure. Our data suggested that policy makers need to think ahead for sustainability to ensure capacity building. There were multiple examples highlighted of knowledge, experience, and attitude gaps that will hinder sustainability. (for sustainability) we need to find ways to engage the opinion leaders at a community level early on, . . . , we need to explore the use of experts (to educate opinion leaders) (CAB FGD) we need to have a systems thinking approach however hard it is (CAB FGD) All groups of participants stressed that tailored engagement and inclusive strategies would be necessary to overcome barriers. Addressing challenges requires a holistic approach, combining technical solutions that work with societal, educational, and policy interventions to strengthen the diffusion of any of the AI maternal health innovations within the social system. A key point highlighted in a scoping workshop was the strategy of continuous evaluation and feedback. One speaker noted that implementing AI in MSRH should be an iterative process: “ We should deploy AI carefully and study its impact. Are mothers actually healthier? Are providers happier? We need to gather feedback and be ready to improve the tools. ” In practice, this would imply using traditional clinical evidence methods such as randomized controlled trials for efficiacy compared to humans alone, as well as monitoring outcomes once an AI system is rolled out (for example, tracking if an AI triage system actually reduced waiting times at clinics or improved maternal mortality rates in an area). Getting feedback from users would allow developers and health authorities to refine the system or even decide to scale it up or down. This approach would, if working well, ensure that AI remains aligned with the people’s needs over time, adapting to new information or changing circumstances in close to real time. With regards to the ethical concerns, participants mentioned strengthening ethical protocols by ensuring data protection and combating institutional biases. It was emphasized that it is key to ensure that there is always a human in the loop, this is important to the general population in addition to the specifics for MSRH. To alleviate infrastructural issues such as the cost of or lack of internet, participants suggested that future models would necessitate AI models that can function offline and remain efficient under resource constraints. In the FGD, some reflections on sustainability revolved around how use of cultural opinion leaders, scientific experts, and other stakeholders enable a longer term vision and can contextualize AI use for MSRH in SSA. Participants in the scoping workshop cautioned that using AI cannot instantly remove all the challenges in our MSRH services, it will only provide one piece of the complex puzzle. “We need AI precisely because we need to do better for mothers and young people. But AI is not a silver bullet for all our health system problems,” the participant cautioned. (scoping workshop) DISCUSSION Across the participant groups and data collection methods, we found a high level of diversity in the experiences and perceptions of AI in MSRH in sub-Saharan Africa. In high income countries AI in MSRH is predicted to touch all areas of health care from disease prediction to health promotion in the very near future. 16 Our data clearly highlighted that stakeholders agree that AI is needed in MSRH because it offers benefits though powerful tools to improve outcomes and equity. Our findings showed the importance of AI innovations as a support tool to strengthen MSRH healthcare delivery including: improving stock management and triage, increasing MSRH awareness, enhancing diagnosis, guiding treatments, and enabling personalized care. However, our findings suggest that this same level of confidence in AI for MSRH among early adopters has not been reached to allow for a wide diffusion and for the potential of AI in MSRH among the diversity of African stakeholders. A global systematic review of health care professional experience by Ayorinde et al identified health care professionals’ understanding of AI applications, level of trust and confidence in AI tools and judging the value added by AI as main areas raised in their study. 17 In Africa, there have been several quantitative interview studies of health worker perceptions including trainee radiologists . 18,19 A comprehensive mixed methods study by Asiedu with health workers and general public across Africa revealed more positive attitudes towards AI from general public compared to more cautious health workers who identified trust, ethics and systemic barriers to integration. For MSRH specific work, a review of midwife perceptions revealed eight studies highlighting potential for improved quality of care, particularly in perinatal and neonatal settings, with barriers to integration of ethical concerns and hesitation among midwives, due to low levels of digital health literacy. We have not found any previous comprehensive qualitative studies exploring AI in MSRH. Participants in our study, were steadfast in emphasizing the core principle that AI should augment, not replace human healthcare. As other studies have highlighted lack of awareness of health workers about the opportunities of health AI; health workers are thirsty for knowledge about how they can use AI, and how to make sure it is used safely and appropriately. A novel finding from our study is the emphasis on co-creation of AI tools with health workers, traditional health workers and communities as a way to ensure that the tools benefit those most in need and are safe. Our data reinforces previous experience with health technology adoption, that it must go hand-in-hand with strengthening healthcare infrastructure and workforce to effectively use the innovations. The consensus across our data, as well as others, cautions that AI must be implemented thoughtfully and carefully alongside improvements in capacity, regulation and infrastructure. 8,20 Rigorous evaluation of new AI tools pre and during implementation was highlighted by participants as an important process that would support trust in the tools. Ensuring adequate and non-biased data was a concern that was less prevalent in health worker facing studies; this may be due to the inclusion of AI experts in this study 18,21,22 This is particularly important in MSRH, as women are often missing in research data, formal medical records and other socio-economic datasets. Notably in our study and in others were the challenges related to ethics and regulation. 23 AI innovations have been highlighted as a potential to revolutionize maternal and other health care in Africa, yet large scale studies are limited across the continent. 8,24 Our study found that content area specialists believe there are diverse AI innovations that could improve, widen the reach and increase equity in maternal, sexual and reproductive health yet, their diffusion depends on a number of factors related to all four key pillars of the theory, the innovation itself that is built on complex infrastructure, the communication channels that depend on accessibility to technology, the speed with which access is available for new innovative projects and the social system comprised of cultural and language, challenges. We referred to the diffusion of innovations theory in our analysis of findings to help frame our results. We found that diffusion of AI innovations has been hampered by awareness in the general public of specific AI innovations, complexity of many of the innovations, trialability, observability and in the realm of ethics and regulations. 25 The strengths of this study are the triangulation of three different settings, key informant interviews with international African AI and MSRH experts, FGD with a community advisory board, and observations at a scoping workshop. This allows for a wide cross-sectional view of different perceptions of AI. The potential weakness of this study is that there was a large gap in time between KII-FGD and the scoping workshop, and we were initially expecting results to be very different and require separate reporting; however during analysis the themes were more similar than expected so the data were combined. Overall, the consensus of our participants suggested that AI has great potential to tackle the urgent MSRH challenges that disproportionality affect women and adolescent girls. AI for MSRH has potential to improve quality and efficiency of MSRH health care delivery in Africa and to widen the reach of health services. For the AI innovations to be disseminated equitably, increase in awareness of the characteristics of AI innovations, their layered complexities, capacity to understand and use them fairly and improving infrastructure will all be critical. Co-creation of AI tools with communities, heath workers and technologists and rigorous evaluation may help to create safe, trustworthy and easily adopted AI tools for MSRH. References WHO. World Health Organization Global adult estimates , (2020). WHO. Sexual and Reproductive Health , (2022). Adejumo, O. A., Malee, K. M., Ryscavage, P., Hunter, S. J. & Taiwo, B. O. Contemporary issues on the epidemiology and antiretroviral adherence of HIV-infected adolescents in sub-Saharan Africa: a narrative review. J Int AIDS Soc 18 , 20049 (2015). https://doi.org/10.7448/IAS.18.1.20049 Idele, P. et al. Epidemiology of HIV and AIDS among adolescents: current status, inequities, and data gaps. J Acquir Immune Defic Syndr 66 Suppl 2 , S144-153 (2014). https://doi.org/10.1097/QAI.0000000000000176 Mbadugha, N. Big Data and AI in Africa: Why the Future of the Continent Is Artificially Intelligent and Digitally Enabled , (2021). Bolarinwa, O., Adebisi, Y. A., Ajayi, K. V. & Boutahar, R. Sexual and reproductive health rights in the era of artificial intelligence. Lancet 404 , 120-121 (2024). https://doi.org/10.1016/S0140-6736(24)00696-2 Bolarinwa, O. A., Mohammed, A. & Igharo, V. The role of artificial intelligence in transforming maternity services in Africa: prospects and challenges. Ther Adv Reprod Health 18 , 26334941241288587 (2024). https://doi.org/10.1177/26334941241288587 Townsend, B. A. et al. Mapping the regulatory landscape of AI in healthcare in Africa. Front Pharmacol 14 , 1214422 (2023). https://doi.org/10.3389/fphar.2023.1214422 Valente, T. W. Diffusion of innovations. Genet Med 5 , 69 (2003). https://doi.org/10.1097/01.GIM.0000061743.67794.C4 Valente, T. W., Dyal, S. R., Chu, K. H., Wipfli, H. & Fujimoto, K. Diffusion of innovations theory applied to global tobacco control treaty ratification. Soc Sci Med 145 , 89-97 (2015). https://doi.org/10.1016/j.socscimed.2015.10.001 Valente, T. W. & Fosados, R. Diffusion of innovations and network segmentation: the part played by people in promoting health. Sex Transm Dis 33 , S23-31 (2006). https://doi.org/10.1097/01.olq.0000221018.32533.6d HASH. Baseline Quantitative Online Survey (2024). Collaco, N. et al. Using the Framework Method for the Analysis of Qualitative Dyadic Data in Health Research. Qual Health Res 31 , 1555-1564 (2021). https://doi.org/10.1177/10497323211011599 Gale, N. K., Heath, G., Cameron, E., Rashid, S. & Redwood, S. Using the framework method for the analysis of qualitative data in multi-disciplinary health research. BMC Med Res Methodol 13 , 117 (2013). https://doi.org/10.1186/1471-2288-13-117 Klingberg, S., Stalmeijer, R. E. & Varpio, L. Using framework analysis methods for qualitative research: AMEE Guide No. 164. Med Teach 46 , 603-610 (2024). https://doi.org/10.1080/0142159X.2023.2259073 Wilkins, A. Concerns raised over AI trained on 57 million NHS medical records , (2025). Ayorinde, A. et al. Health Care Professionals' Experience of Using AI: Systematic Review With Narrative Synthesis. J Med Internet Res 26 , e55766 (2024). https://doi.org/10.2196/55766 Nciki, A. I. & Hlabangana, L. T. Perceptions and attitudes towards AI among trainee and qualified radiologists at selected South African training hospitals. SA J Radiol 29 , 3026 (2025). https://doi.org/10.4102/sajr.v29i1.3026 Asiedu M, H. I., Dieng A, Kauer K, Ahmed T, Ofori F, Chan C, Pfohl S, Rostamzadeh N, Heller K. in ACM Conference on Fairness, Accountability, and Transparency (ACM, New York, NY, USA, Athens, Greece, 2025). Reid, M. J. A. et al. Announcing the Lancet Global Health Commission on artificial intelligence (AI) and HIV: leveraging AI for equitable and sustainable impact. Lancet Glob Health 13 , e611-e612 (2025). https://doi.org/10.1016/S2214-109X(25)00049-X Giaxi, P., Vivilaki, V., Sarella, A. & Gourounti, K. Artificial Intelligence in Midwifery: A Scoping Review of Current Applications, Future Prospects, and Midwives' Perspectives. Healthcare (Basel) 13 (2025). https://doi.org/10.3390/healthcare13080942 Giaxi, P., Vivilaki, V., Sarella, A., Harizopoulou, V. & Gourounti, K. Artificial Intelligence and Machine Learning: An Updated Systematic Review of Their Role in Obstetrics and Midwifery. Cureus 17 , e80394 (2025). https://doi.org/10.7759/cureus.80394 Fjeld, J., Achten, N., Hilligoss, H., Nagy, A., and Srikumar, M. Principled Artificial Intelligence: Mapping Consensus in Ethical and Rights-Based Approaches to Principles for AI. (2020). Chemisto, M., Gutu, TJL, Kalinaki, K, Bosco, DM, Egau, P, Fred, K, Oloya, IT, Rashid, K. (Nairobi, Kenya, 2023). Pham, T. Ethical and legal considerations in healthcare AI: innovation and policy for safe and fair use. R Soc Open Sci 12 , 241873 (2025). https://doi.org/10.1098/rsos.241873 Additional Declarations The authors declare no competing interests. Cite Share Download PDF Status: Posted Version 1 posted You are reading this latest preprint version Research Square lets you share your work early, gain feedback from the community, and start making changes to your manuscript prior to peer review in a journal. As a division of Research Square Company, we’re committed to making research communication faster, fairer, and more useful. We do this by developing innovative software and high quality services for the global research community. Our growing team is made up of researchers and industry professionals working together to solve the most critical problems facing scientific publishing. Also discoverable on Platform About Our Team In Review Editorial Policies Advisory Board Help Center Resources Author Services Accessibility API Access RSS feed Manage Cookie Preferences © Research Square 2026 | ISSN 2693-5015 (online) Privacy Policy Terms of Service Do Not Sell My Personal Information {"props":{"pageProps":{"initialData":{"identity":"rs-7463408","acceptedTermsAndConditions":true,"allowDirectSubmit":true,"archivedVersions":[],"articleType":"Research Article","associatedPublications":[],"authors":[{"id":508023608,"identity":"59138d12-5527-4f8a-bb6e-95583c4b09a7","order_by":0,"name":"Rachel King","email":"data:image/png;base64,iVBORw0KGgoAAAANSUhEUgAAAZAAAAAyAQMAAABI0h/eAAAABlBMVEX///8AAABVwtN+AAAACXBIWXMAAA7EAAAOxAGVKw4bAAAA50lEQVRIiWNgGAWjYHACNgaGAgYDIMn4AMjjB2IDIrQYgLUwg5RKNhCtBcSQIEqLefvhZw8+GDAY80m3X6vmqamTYGBv3iaBT4vMmTRzwxkGDGZsMmfKbvMcOyzBwHOsDK8WCYYcNmkeAwYbNomctNu8DQfqGCRyzPBr4X/DJv0HqqWYtwHoMPk3BLRIAG1hADlMIv0YM28DswSDBA8hLc/MJHsMJIyBtjBLzgH6hY0nrdgCv8OSn0n8qLAxnD8j/eGHN8AQ42c/vPEGPi0wnUDMA4kONiKUwwD7AxIUj4JRMApGwUgCAEbBOAXwDOCDAAAAAElFTkSuQmCC","orcid":"https://orcid.org/0000-0002-0085-3498","institution":"University of California, San Francisco","correspondingAuthor":true,"prefix":"","firstName":"Rachel","middleName":"","lastName":"King","suffix":""},{"id":508023956,"identity":"a7d94ecf-7748-48c1-80d8-338657da195a","order_by":1,"name":"Elizabeth Oseku","email":"","orcid":"","institution":"Infectious Diseases Institute, Kampala, Uganda","correspondingAuthor":false,"prefix":"","firstName":"Elizabeth","middleName":"","lastName":"Oseku","suffix":""},{"id":508024085,"identity":"8508a0ae-24a0-4c93-91ce-ea9f1ff0ab69","order_by":2,"name":"Cecilia Akatukwasa","email":"","orcid":"","institution":"Infectious Diseases Institute, Kampala, Uganda","correspondingAuthor":false,"prefix":"","firstName":"Cecilia","middleName":"","lastName":"Akatukwasa","suffix":""},{"id":508024086,"identity":"4efb7ee5-9e60-498d-94ec-f6a45576c6d1","order_by":3,"name":"Moreen Nanyonjo","email":"","orcid":"","institution":"Infectious Diseases Institute, Kampala, Uganda","correspondingAuthor":false,"prefix":"","firstName":"Moreen","middleName":"","lastName":"Nanyonjo","suffix":""},{"id":508024087,"identity":"c1668c47-6798-4eda-8c09-46895ba2e78a","order_by":4,"name":"Joshua Beinomugisha","email":"","orcid":"","institution":"Infectious Diseases Institute, Kampala, Uganda","correspondingAuthor":false,"prefix":"","firstName":"Joshua","middleName":"","lastName":"Beinomugisha","suffix":""},{"id":508024088,"identity":"60d31e51-da80-4c34-b613-014f724674d6","order_by":5,"name":"Joan Akullo","email":"","orcid":"","institution":"Infectious Diseases Institute, Kampala, Uganda","correspondingAuthor":false,"prefix":"","firstName":"Joan","middleName":"","lastName":"Akullo","suffix":""},{"id":508024089,"identity":"23444f83-c2a0-4109-a120-99f12e4a5790","order_by":6,"name":"Jackie Ssemata","email":"","orcid":"","institution":"Infectious Diseases Institute, Kampala, Uganda","correspondingAuthor":false,"prefix":"","firstName":"Jackie","middleName":"","lastName":"Ssemata","suffix":""},{"id":508024090,"identity":"323f4f73-51a1-4847-8762-1e9225493766","order_by":7,"name":"Rosalind Parkes-Ratanshi","email":"","orcid":"https://orcid.org/0000-0001-9297-1311","institution":"Queens University, Belfast, Northern Ireland","correspondingAuthor":false,"prefix":"","firstName":"Rosalind","middleName":"","lastName":"Parkes-Ratanshi","suffix":""}],"badges":[],"createdAt":"2025-08-26 13:23:36","currentVersionCode":1,"declarations":{"humanSubjects":true,"vertebrateSubjects":false,"conflictsOfInterestStatement":false,"humanSubjectEthicalGuidelines":true,"humanSubjectConsent":true,"humanSubjectClinicalTrial":false,"humanSubjectCaseReport":false,"vertebrateSubjectEthicalGuidelines":false},"doi":"10.21203/rs.3.rs-7463408/v1","doiUrl":"https://doi.org/10.21203/rs.3.rs-7463408/v1","draftVersion":[],"editorialEvents":[],"editorialNote":"","failedWorkflow":false,"files":[{"id":90469151,"identity":"2acc3996-0153-4702-bc1e-59acda0a8530","added_by":"auto","created_at":"2025-09-03 06:09:42","extension":"png","order_by":1,"title":"Figure 1","display":"","copyAsset":false,"role":"figure","size":158518,"visible":true,"origin":"","legend":"\u003cp\u003eDiffusion of Innovations Theory\u003c/p\u003e","description":"","filename":"1.png","url":"https://assets-eu.researchsquare.com/files/rs-7463408/v1/c204496fbd8acd217fef8e47.png"},{"id":90470935,"identity":"2edd654d-7d82-4a62-bf65-1369f96addd3","added_by":"auto","created_at":"2025-09-03 06:17:46","extension":"pdf","order_by":0,"title":"","display":"","copyAsset":false,"role":"manuscript-pdf","size":784895,"visible":true,"origin":"","legend":"","description":"","filename":"manuscript.pdf","url":"https://assets-eu.researchsquare.com/files/rs-7463408/v1/bd539b2a-0e7a-434f-9a09-ee3720398a2c.pdf"}],"financialInterests":"The authors declare no competing interests.","formattedTitle":"\u003cp\u003e\u003cstrong\u003eA Qualitative Inquiry Exploring Perceptions of Artificial Intelligence to Improve Outcomes in Maternal, Sexual and Reproductive Health in Sub-Saharan Africa\u003c/strong\u003e\u003c/p\u003e","fulltext":[{"header":"INTRODUCTION","content":"\u003cp\u003eThe World Health Organization (WHO) estimates that poor reproductive health accounts for up to 18% of the global burden of disease, and 32% of the total burden of disease for women of reproductive age.\u003csup\u003e\u003cspan citationid=\"CR1\" class=\"CitationRef\"\u003e1\u003c/span\u003e\u003c/sup\u003e In Sub-Saharan Africa, death and disability resulting from reproductive health causes remain unacceptably high. The adult lifetime risk of maternal death has been estimated to be highest in Africa (1 in 26), while developed countries have been estimated to have the smallest lifetime risk (1 in 7,300). The prevention and control of reproductive tract infections is another area of concern; for example, Congenital Syphilis is the second leading cause of preventable stillbirth globally, preceded only by malaria.\u003csup\u003e\u003cspan citationid=\"CR2\" class=\"CitationRef\"\u003e2\u003c/span\u003e\u003c/sup\u003e The global HIV response continues to be undermined by a multitude of complex, interrelated challenges that are aggravated by limited domestic resources, declining donor assistance, including the recent reorganization of US global health commitment, which threatens the sustainability of all critical health programmes in the region. Vulnerable populations such as adolescents in Africa continue to be more susceptible to STIs, HIV, unwanted pregnancies and unsafe abortions.\u003csup\u003e\u003cspan citationid=\"CR3\" class=\"CitationRef\"\u003e3\u003c/span\u003e,\u003cspan citationid=\"CR5\" class=\"CitationRef\"\u003e4\u003c/span\u003e\u003c/sup\u003e\u003c/p\u003e\u003cp\u003eIn the last five to ten years, there has been an explosion in available health data in Africa due to improved infrastructure for electricity and internet as well as the widespread adoption of digital health technologies.\u003csup\u003e\u003cspan additionalcitationids=\"CR6\" citationid=\"CR6\" class=\"CitationRef\"\u003e5\u003c/span\u003e\u0026ndash;\u003cspan citationid=\"CR7\" class=\"CitationRef\"\u003e7\u003c/span\u003e\u003c/sup\u003e This provides an opportunity for data driven strategies and innovations that previously did not exist. To increase technology-driven solutions, we must consider the human research and implementation capacity to enable this valuable resource to improve health of the extremely diverse population.\u003c/p\u003e\u003cp\u003eDue to the rapidly emerging technology around AI, many health care workers and policymakers are not aware of the opportunities and limitations of these technologies. The use of AI in health in LMICs is now emerging across sub-Saharan Africa. There is consensus in the literature that there is potential in using AI in expanding and extending healthcare access, by contributing to early disease detection and prevention, increasing diagnostic capability and drug development, disease surveillance, stock and healthcare management as well as clinical decision-making.\u003csup\u003e\u003cspan citationid=\"CR9\" class=\"CitationRef\"\u003e8\u003c/span\u003e\u003c/sup\u003e Thus, our aim was to map knowledge and awareness of AI among individuals working in health across the continent.\u003c/p\u003e\u003cp\u003eTo help frame and understand how individuals, and groups understand and adopt new innovations, the \u003cem\u003eDiffusion of Innovations Theory (DOI)\u003c/em\u003e first documented by Everett Rogers, has been used as a structure specifically for technology innovations.\u003csup\u003e\u003cspan additionalcitationids=\"CR10\" citationid=\"CR10\" class=\"CitationRef\"\u003e9\u003c/span\u003e\u0026ndash;\u003cspan citationid=\"CR11\" class=\"CitationRef\"\u003e11\u003c/span\u003e\u003c/sup\u003e DOI speculates that the spread of new ideas, technologies, or practices within a social (or health) system occurs in a predictable and systematic way. The theory states that individuals within a population adopt innovations at varying degrees and rates, categorized into: innovators, early adopters, early majority, late majority, and stragglers. These groups are characterized by differing levels of risk tolerance, social influence, and access to information. The theory further highlights the role of communication channels, social networks, and perceived attributes of the innovation (e.g., relative advantages, compatibility, complexity, trialability, observability) in facilitating or hindering its adoption. While the theory does not explicitly underscore issues of cost and access to the innovation, it implicitly considers access through \u003cem\u003etrialability\u003c/em\u003e which refers to the extent to which an innovation can be tested on a limited basis before a full commitment is made. An important feature of the theory is \u003cem\u003eobservability\u003c/em\u003e referring to the degree to which the results of an innovation are visible \u0026lsquo;in the wild\u0026rsquo; meaning in everyday life and therefore easily communicated to others. If individuals cannot observe others using the innovation and experiencing its benefits, they are less likely to be interested in adopting it themselves. The theory also includes economic and political elements.\u003csup\u003e\u003cspan additionalcitationids=\"CR10\" citationid=\"CR10\" class=\"CitationRef\"\u003e9\u003c/span\u003e\u0026ndash;\u003cspan citationid=\"CR11\" class=\"CitationRef\"\u003e11\u003c/span\u003e\u003c/sup\u003e\u003c/p\u003e\u003cp\u003eThe overall goal of this study is to explore the early experiences, perceptions of health workers, policymakers and general population, as well as AI researchers and implementers around the opportunities, risks, limitations, and best practices for responsible artificial intelligence applications in maternal, sexual and reproductive health (MSRH) in sub-Saharan Africa.\u003c/p\u003e\u003cp\u003eDiffusion of Innovations theory provides a framework for understanding how innovations are adopted, highlighting knowledge, attitudes, barriers and recommendations for adoption of AI in MRSH in the future within the numerous environments in sSA. We inspect our data through the lens of the Diffusion of Innovations theory to explore factors influencing stakeholder engagement in AI applications.\u003c/p\u003e"},{"header":"METHODS","content":"\u003cp\u003eWe conducted a cross-sectional qualitative study to explore perceptions of health workers, policymakers, AI researchers and implementers, as well as seven community Advisory Board (CAB) members. We delved into the opportunities, risks, limitations, and best practices for responsible AI in maternal, sexual and reproductive health (MSRH) in sub-Saharan Africa (SSA).\u003c/p\u003e\n\u003cp\u003eThe study was conducted within the activities of HASH - The Hub for Artificial Intelligence in Maternal, Sexual and Reproductive Health in Sub-Saharan Africa. HASH was formed in 2021 by a multidisciplinary consortium of the Infectious Diseases Institute, the Makerere University College of Computing and Information Science, Sunbird AI through funding from the International Development Research Centre (IDRC) and the Swedish International Development Cooperation Agency (SIDA), as part of the Artificial Intelligence for Development in Africa Program (AI4D Africa) and the Global South AI for Global Health (AI4GH) INIT. HASH\u0026rsquo;s objective is to advance Maternal, Sexual, and Reproductive Health (MSRH) and rights while strengthening health systems in Sub-Saharan Africa (SSA) through the responsible development and deployment of Artificial Intelligence (AI) innovations.\u003c/p\u003e\n\u003cp\u003eThe Hub is based in Uganda but targets both anglophone and francophone stakeholders at various levels in the field of AI and/or MSRH across SSA. This includes innovators, researchers, health workers, policy makers, organizations and enthusiasts working in the AI for MSRH space in sub-Saharan Africa (SSA). The Hub provides support for and gains insight into the use of new and existing data and how AI and data technology can be leveraged to solve African health challenges. Grounded in the ethical and responsible application of AI, the Hub provides capacity building through technical and methodological assistance to its members and shares opportunities for mentorship and collaboration.\u003c/p\u003e\n\u003cp\u003eThis study used the HASH platform to undertake purposeful sampling of thought leaders in AI in MSRH for targeted focus group discussion and in-depth interviews. Additionally, we collected data from routine stakeholder engagement consultations (scoping workshops) conducted to inform the strategy and implementation of HASH.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eData collection\u0026nbsp;\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003e\u003cem\u003e\u003cu\u003eKey Informant interviews:\u003c/u\u003e\u003c/em\u003e As part of HASH activities in 2021 we conducted an online survey of 107 respondents from 25 countries of MSRH experts and AI researchers across Africa. We included individuals with evidence of expert knowledge, including but not limited to: actively working on AI projects in healthcare, working maternal, sexual and reproductive health, familiar with the concepts and methods of AI, based in anglophone or francophone sub-Saharan Africa. The survey identified key areas of MSRH that respondents considered to be key research priorities for AI (Table 1).\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003e\u003cem\u003eTable 1. Ranking of top research and development in MSRH according to potential and viability of AI as a solution, 2021, sub-Saharan Africa\u0026nbsp;\u003c/em\u003e\u003c/strong\u003e\u003c/p\u003e\n\u003ctable border=\"0\" cellspacing=\"0\" cellpadding=\"0\" width=\"611\"\u003e\n \u003ctbody\u003e\n \u003ctr\u003e\n \u003ctd valign=\"top\"\u003e\n \u003cp\u003e\u003cstrong\u003eTHEME\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\"\u003e\n \u003cp\u003e\u003cstrong\u003ePRIORITY AREA\u003c/strong\u003e \u003cstrong\u003eTO IMPROVE HEALTH OUTCOMES\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd valign=\"top\"\u003e\n \u003cp\u003e\u003cstrong\u003eMATERNAL HEALTH\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\"\u003e\n \u003col\u003e\n \u003cli\u003eIdentification of high-risk conditions in mothers and newborns\u003c/li\u003e\n \u003cli\u003eAttracting and retaining skilled birth attendants in remote areas\u003c/li\u003e\n \u003cli\u003eMaternal mortality\u003c/li\u003e\n \u003cli\u003eImproving uptake of proven existing interventions\u003c/li\u003e\n \u003cli\u003eAlgorithms and point of care diagnostic tests for sepsis in the face of emergencies and epidemics\u003c/li\u003e\n \u003cli\u003eImplementation research of interventions that improve quality of care during labour\u003c/li\u003e\n \u003cli\u003eThe impact of training on skills of health workers handling obstetric emergencies\u003c/li\u003e\n \u003cli\u003eImproving access to emergency transport in hard-to-reach communities\u003c/li\u003e\n \u003cli\u003eImproved maternal and new-born emergency care\u003c/li\u003e\n \u003c/ol\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd valign=\"top\"\u003e\n \u003cp\u003e\u003cstrong\u003eADOLESCENT SEXUAL REPRODUCTIVE HEALTH\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\"\u003e\n \u003col\u003e\n \u003cli\u003eUnmet need for contraception\u003c/li\u003e\n \u003cli\u003eInform how health services can be designed to effectively meet adolescents\u0026apos; health needs.\u003c/li\u003e\n \u003cli\u003eDevelopment of prevention technologies besides condoms\u003c/li\u003e\n \u003cli\u003eGender gaps in sexual health\u003c/li\u003e\n \u003cli\u003eExamining health related behaviour (physical activity, diet)\u003c/li\u003e\n \u003cli\u003eExamining health seeking behaviour\u003c/li\u003e\n \u003cli\u003ePrevalence/incidence of different health outcomes (STIs, pregnancy)\u003c/li\u003e\n \u003cli\u003eAccess to comprehensive SRH services\u003c/li\u003e\n \u003cli\u003eEvaluation of effectiveness of health service delivery models\u003c/li\u003e\n \u003cli\u003eUnderstanding needs of vulnerable adolescents (street children)\u003c/li\u003e\n \u003c/ol\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd valign=\"top\"\u003e\n \u003cp\u003e\u003cstrong\u003eSEXUALLY TRANSMITTED INFECTIONS\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\"\u003e\n \u003cp\u003e1. Screening and case finding of STIs\u003c/p\u003e\n \u003cp\u003e2. Early diagnosis and treatment of STI cases\u003c/p\u003e\n \u003cp\u003e3. STI-prevention strategies\u003c/p\u003e\n \u003cp\u003e4. Issues around antimicrobial resistance\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd valign=\"top\"\u003e\n \u003cp\u003e\u003cstrong\u003eHIV/AIDS\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\"\u003e\n \u003col\u003e\n \u003cli\u003eTracing and reengagement in care\u003c/li\u003e\n \u003cli\u003eDiagnostics that meet the specifications for use in LMICs, especially point-of-care diagnostics targeting lower levels facilities\u003c/li\u003e\n \u003cli\u003eAssessing ART adherence\u003c/li\u003e\n \u003cli\u003eAssessing frequency of care visits/refills\u003c/li\u003e\n \u003cli\u003eTask sharing of specimen collection and point-of-care testing\u003c/li\u003e\n \u003cli\u003eIntegration of HIV and sexual reproductive health services\u003c/li\u003e\n \u003cli\u003eAssessing ART initiation outside the health facility\u003c/li\u003e\n \u003cli\u003eGeneration of data on drug safety and efficacy for PLHIV\u003c/li\u003e\n \u003cli\u003eAssessing integration of HIV and diabetes and hypertension care\u003c/li\u003e\n \u003cli\u003eDeveloping and introducing new paediatric drug formulations for treatment of HIV/AIDS\u003c/li\u003e\n \u003cli\u003eDeveloping psychosocial interventions for adolescents \u0026amp; youth\u003c/li\u003e\n \u003cli\u003eBalanced integration of diagnostic services\u003c/li\u003e\n \u003c/ol\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003c/tbody\u003e\n\u003c/table\u003e\n\u003cp\u003e\u003cstrong\u003e\u003cem\u003e\u003cu\u003eProcedures:\u003c/u\u003e\u003c/em\u003e\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003e\u003cem\u003e\u003cu\u003eParticipant selection KII\u003c/u\u003e\u003c/em\u003e: During the online survey participants were asked if they could be contacted for further one-on-one interviews. Ten participants were recruited by analyzing the results of questions of the HASH online survey to gauge the level of expertise, experience and willingness to participate.\u003csup\u003e12\u003c/sup\u003e The respondents who met the eligibility criteria were selected through purposive sampling using the above criteria to participate in the key informant interviews (KIIs). Respondents were contacted via email or telephone and appointments made.\u003c/p\u003e\n\u003cp\u003eBefore the interview, the participant received a consent form via email. During the scheduled interview time, informed consent was administered by the interviewer and the participant asked to voluntarily sign the consent form with an electronic signature and email it to the interviewer before the interview began.\u003c/p\u003e\n\u003cp\u003e\u003cem\u003e\u003cu\u003eParticipant selection Focus Group Discussion (FGD)\u003c/u\u003e\u003c/em\u003e: Participants were invited to the FGD if they were a member of the Academy for Health Innovations Community Advisory Board (CAB) at the Infectious Diseases Institute in Uganda. The CAB is composed of ministry of health officials, a youth representative, health worker from a lower health facility, two religious leaders (Muslem; Catholic), local council leader, a microfinance administrator, and a civil society organizational representative. The CAB\u0026rsquo;s main role is to advise and facilitate dialogue between the community and research team. An appointment was made for the FGD with all the CAB members either face to face or on-line through an email or phone. The FGD was conducted in March 2023, with seven participants (four men and three women). All respondents were consented in the same manner as for the interviews.\u003c/p\u003e\n\u003cp\u003e\u003cem\u003e\u003cu\u003eData collection Interviews and FGD:\u003c/u\u003e\u003c/em\u003e A standardized semi-structured guide was developed for FGD and KII. The guide was developed to explore the perceptions of the stakeholders on the future of AI in health care in Africa including perceived benefits and limitations. Exploration of the role in MSRH was based upon the priority areas identified in the survey (Table 1). Each interview was held with two social scientists; one to guide the discussion and one note-taker. The process was interactive and took between 30 and 60 minutes depending on experience with AI. All KIIs/FGD were conducted in English as we did not have anyone preferring French or any other language when they were asked.\u003c/p\u003e\n\u003cp\u003eThe interviews were conducted either face-to-face or by Zoom with video on for the participant, moderator and note-taker throughout the interview to allow for recognition of non-verbal cues. The FGD was face-to-face and was audio recorded. Interviews were audio recorded and both interview and FGD recordings were kept in a secured file in an online repository with access restricted only to the study team members.\u003c/p\u003e\n\u003cp\u003e\u003cem\u003e\u003cu\u003eScoping workshops (n=3):\u0026nbsp;\u003c/u\u003e\u003c/em\u003eThree multi-stakeholder workshops with participants across sub-Saharan Africa convened diverse voices including healthcare providers, AI innovators, venture capitalists, health educators, policy makers, young people (15-24 years), health entrepreneurs, and religious leaders to design a path forward regarding AI innovations and MSRH in sub-Saharan Africa. Participants were identified through stakeholder mapping using purposive selection aiming to identify representatives across the above categories. The workshops included presentations on responsible AI, on the HASH project and on AI in medicine to facilitate a level playing field for fruitful discussion. The sessions were designed to capture diverse stakeholder perspectives on the role of AI in MSRH. One full-day physical workshop and two half-day virtual workshops were held in March and April 2025. Each workshop convened participants from across MSRH ecosystem to foster rich, multidisciplinary dialogue in the sub-Saharan contexts. Four social scientists were present as participant observers to critically observe and document the full discussions, reactions, and content of each session.\u003c/p\u003e\n\u003cp\u003eFor the scoping workshops, the insights from stakeholders were structured around five guiding questions 1) Why do we need AI in Maternal, Sexual and Reproductive Health? 2) What strategies enable AI to meet the needs of all stakeholders in MSRH? 3) What are the concerns and anticipated challenges of AI in MSRH? 4) What are the overall implications and way forward? 5) How do we make AI a mainstream tool for health in your setting? All discussions were held in English and lasted for two hours and 30 minutes for virtual workshops and four hours for physical workshops (divided into two breakout sessions). All workshops were audio recorded.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003e\u003cem\u003e\u003cu\u003eAnalysis:\u003c/u\u003e\u003c/em\u003e\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eFGD and KII audio recordings were transcribed word for word. Data from the FGDs and interviews was analyzed using the Framework Analysis approach.\u003csup\u003e13-15\u003c/sup\u003e The method produces highly structured outputs of summarized data and is particularly useful for large quantities of data. Framework analysis is commonly used for the thematic analysis of semi-structured transcripts as reflexivity, rigor and quality are integral and critical. It can be adapted for use with deductive, inductive, or combined types of qualitative analysis. The key stages of Framework Analysis include familiarization, identification of a thematic framework or a codebook, coding, charting and interpretation.\u003c/p\u003e\n\u003cp\u003eWe combined scoping workshop, KII and FGD data. The preliminary findings were presented for feedback on three occasions: inaugural AfricaAI conference 2023 in Rwanda; AI4GH meeting in Nairobi in November 2023; and to the Technology and Innovation unit at the UK Foreign, Commonwealth Development Office in January 2025.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eEthics Statement\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eEthical approval was obtained from the Infectious Diseases Institute Research Ethics Committee (IDIREC REF 011/2022) and the Uganda National Council for Science and Technology (HS2356ES). The study was conducted in accordance with the protocol, GCP guidelines, the Declaration of Helsinki and all applicable local regulatory requirements and laws. All research team members had verified GCP certification.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eFINDINGS\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eWe include 44 total participants as described above from Uganda, Nigeria, Kenya, Tanzania, Ghana, Zambia, United Kingdom, Senegal, Spain and almost equal numbers of men and women with slightly more men. We included 11 participants in the first in-person scoping workshop in Uganda and 16 participants in the two virtual scoping workshops where we had participants from across the continent and one person from the UK. The FGD included seven participants.\u003c/p\u003e\n\u003cp\u003eWe describe our findings based on the four main elements in the diffusion of innovations theory: the innovation, communication channels, time, and the social system (Figure 1). For innovation in this case, we consider a cluster of potential or existing AI innovations, for the time element, we look specifically at the implementation or actualization of AI/MSRH projects.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003e\u003cu\u003eTHE INNOVATION: Current and suggested uses of AI\u003c/u\u003e\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eWhen asked about participants\u0026rsquo; current and suggested uses of AI in MSRH, helping to target or add specificity to high priority health activities or patients, was an overarching theme.\u003c/p\u003e\n\u003cp\u003eSome participants mentioned how useful AI has been in tracking logistics such as tracking supply chain products and triggering action specifically for security purposes and when/if the terrain or environment was challenging. Some participants also mentioned predictive modeling, how powerful chatbots are in personalizing care by answering health-related questions. Additionally, AI could be used to rapidly triage numerous queries from health service users to prioritize questions that require urgent responses. Participants highlighted that these questions and answers could then be used to develop datasets such as ultrasound imaging.\u0026nbsp;\u003c/p\u003e\n\u003cp\u003e\u003cem\u003eWe try to develop predictive models to deliver care more efficiently and to deliver the care really to the women who need it. So, there are some services we deliver as a blanket care to everybody but if we identify certain risk factors or certain special needs, then we think, predictive modeling through machine learning or through AI, can help us identify women or children who need specialized care, then we could deliver that care to only these clients which makes the service delivery more targeted because we do not need to provide these generalized messages. We would only provide messages that are relevant to that client.\u003c/em\u003e (KII, TZ)\u003c/p\u003e\n\u003cp\u003eIn this quote, participants suggest predicting particular health concerns mothers are likely to suffer from so that they deliver services and information that is tailored to patient needs, rather than providing the same general service and information to everyone. This helps in improving patient-centered care and in resource allocation.\u003c/p\u003e\n\u003cp\u003e\u003cem\u003e\u0026ldquo;with the \u003cu\u003eQ and A functionality\u003c/u\u003e; what we used to do is answer questions in the order in which they arrived. If\u0026hellip; we got\u0026hellip;. in a day [more than] 90 questions that were not urgent, and number 91 was something that needed immediate attention, we would have to answer the questions 1 to 90 before we even realize that 91 was something that needed immediate attention.\u0026rdquo;\u0026nbsp;\u003c/em\u003e(KII, Kenya)\u003c/p\u003e\n\u003cp\u003eMany participants in the scoping workshops as well as in the interviews and FGD emphasized that they felt that use of AI was not to replace current roles or individuals but to improve efficiency and enhance numerous medical services. \u0026nbsp;Those who are working in AI and those that felt and expressed a window of opportunity for uses of AI are likely be the early adopters of AI.\u0026nbsp;\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003e\u003cu\u003eBenefits of AI\u003c/u\u003e\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eGeneral benefits that participants stressed included using AI to fill gaps particularly when skilled and experienced health personnel were not continuously available, especially in rural and hard to reach areas. It was mentioned that AI can compete with the best human skills and could, in principle, eliminate human error, bias and corruption, and therefore should support equity as well as reducing time and resources spent on care. \u0026nbsp;Participants felt that with these advantages, we could save money and reduce workload, while increasing coverage and quality of health services. \u0026nbsp;Many felt that AI can increase accurate health information and improve collaboration between providers.\u0026nbsp;\u003c/p\u003e\n\u003cp\u003e\u003cem\u003eAI is competing with the best human skills and so it is able to get you more accurate support in terms of diagnosis and is able to do it more quickly.. . . . , in terms of being able to support with diagnosis in a timely manner which high accuracy, AI is quite instrumental, and it is low cost especially if it can come as open source. Bias too is eliminated. . . not totally eliminated. Biases are always there but when we give a lot of our data to AI algorithms, let\u0026rsquo;s say Nigeria\u0026rsquo;s or South Africa\u0026rsquo;s data, they tend to, over time, be able to predict more accurately our needs regarding to patients\u0026rsquo; diagnosis\u003c/em\u003e\u003cem\u003e.\u0026nbsp;\u003c/em\u003e(KII, Nigeria)\u003c/p\u003e\n\u003cp\u003eWhen asked about which populations may benefit the most from an increase in use of AI, it was highlighted that high literacy populations because AI innovations often use technology through text. Additionally, rural populations who may have limited access to conventional medical services, adolescents who adopt technology quickly and fishermen who are highly mobile so may not easily access health facilities. Though one participant from Nigeria stated that all populations will benefit from AI.\u003c/p\u003e\n\u003cp\u003e\u003cem\u003efishermen, as they are so migratory because today they are here, tomorrow they are there\u003c/em\u003e (CAB FGD)\u003c/p\u003e\n\u003cp\u003eContent areas that KII participants reported benefitting most from AI included maternal and newborn health as well as sexual and reproductive health. \u0026nbsp;Within MSRH, specific areas revolved around monitoring of health conditions including, pregnancy, blood sugar for gestational diabetes, post-natal monitoring and pregnancy complications.\u0026nbsp;\u003c/p\u003e\n\u003cp\u003e\u003cem\u003eThere are times when the baby is having a temperature. It may not necessarily need to go to hospital. . . .It (AI) can give you some form of first aid that can do (in the moment).\u0026nbsp;\u003c/em\u003e(KII, Nigeria).\u003c/p\u003e\n\u003cp\u003e\u003cem\u003eI think one of the things that AI can help (with) at community level is to develop software that can tell this mother about the dangers of the baby\u003c/em\u003e (CAB FGD)\u003c/p\u003e\n\u003cp\u003eSome growth areas that participants noted for further development would be in using images to build databases to use natural language processing for later transfer into voice. \u0026nbsp;\u003c/p\u003e\n\u003cp\u003e\u003cem\u003eWhat I discovered about that ultrasound is\u0026hellip; I saw that they take pictures and if that picture is well scanned, you can create a whole dataset. . . . If you know certain computations of how they show up in those pictures, they can be able to train a machine learning model that uses deep learning \u0026hellip; Not by looking at only images and deep learning but let us look at natural language processing\u003c/em\u003e\u003cem\u003e.\u0026nbsp;\u003c/em\u003e(KII, Uganda)\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003e\u003cu\u003eCOMMUNICATION CHANNELS are a key element in relation to adoption of new innovations.\u0026nbsp;\u003c/u\u003e\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eAn overarching, though not a surprising result, was the lack of understanding and awareness of AI among health workers and the general community. From health workers worried about job security to patients new to digital tools, the importance of user education was a key finding.\u0026nbsp;\u003c/p\u003e\n\u003cp\u003eEnsuring AI meets people\u0026rsquo;s needs also requires capacity building and education. A scoping workshop speaker, who works closely with frontline health workers, pointed out that many providers and patients currently have limited understanding of AI. To prevent AI tools from sitting unused or being misused, investment in training is vital. \u0026ldquo;\u003cem\u003eEven the best AI tool is useless if health workers don\u0026rsquo;t know how to use it or don\u0026rsquo;t trust it\u003c/em\u003e.\u0026rdquo; (scoping workshop)\u003c/p\u003e\n\u003cp\u003eAddressing fear around trust in AI would necessitate building confidence through education and transparent processes. Encouraging partnerships and enforcing collaboration through donor requirements may improve outcomes. One strategy discussed in a scoping workshop was to incorporate AI literacy into medical and nursing education, as well as offering continuous training for existing staff where AI systems would be introduced.\u0026nbsp;\u003c/p\u003e\n\u003cp\u003eOnce health workers decide to adopt AI then clear user guidance is necessary to mitigate harm from inaccurate queries or incomplete data. Health workers should learn not just how to operate AI-driven devices or apps, but also how to interpret AI outputs critically and integrate them with clinical judgment.\u003c/p\u003e\n\u003cp\u003eOne participant noted the significance of training on \u0026ldquo;\u003cem\u003ehow to prompt AI\u0026hellip;so if you get incorrect information, you are going to go with that wrong information\u003c/em\u003e.\u0026rdquo; (scoping workshop)\u003c/p\u003e\n\u003cp\u003eLikewise, community health educators could help familiarize the public with new AI tools and services (for example, teaching expectant mothers how to interact with an AI-driven messaging service that sends them prenatal care advice). The goal of these capacity-building efforts is to ensure that AI becomes a help rather than a hindrance in the workflow, and that communities feel empowered rather than intimidated by new technology. Other suggestions in the FGD included\u0026nbsp;educating the general public about what AI is and what AI\u003cem\u003e\u0026nbsp;\u003cstrong\u003eis not\u003c/strong\u003e.\u0026nbsp;\u003c/em\u003e\u003c/p\u003e\n\u003cp\u003eThe CAB members also suggested using AI in an integration process for both traditional and biomedical health care systems. Traditional birth attendants (TBAs) are highly valued in some communities in sub-Saharan Africa (SSA) and may be an important communication channel for AI in MSRH. For example, including traditional birth attendants into the awareness building so that they understand better which high risk health issues they should refer mothers to health facilities for.\u0026nbsp;\u003c/p\u003e\n\u003cp\u003e\u003cem\u003eI think the best way is to involve the TBAs, first, to educate them and show them what they can do and what they cannot do. That can give them contacts which they can refer to . . . it\u0026rsquo;s [their strength] is about the etiquette; it\u0026rsquo;s about the customer care that they show\u003c/em\u003e (CAB FGD)\u003c/p\u003e\n\u003cp\u003eEnsuring a cultural fit especially with respect to linguistic barriers was a key point highlighted in the scoping workshops. Recognizing that MSRH is deeply intertwined with cultural norms and sensitivities is paramount. AI systems, particularly in diverse contexts like in Africa where open discussions about sex are often taboo, all potential solutions must undergo thorough cultural reviews. The communication style, information delivery, and even the platform\u0026rsquo;s interface should be sensitive to local customs and communication patterns. Creating a sense of trust and comfort is essential, as demonstrated by the strategy of designing chatbots with a friendly and approachable persona. \u003cem\u003e\u0026ldquo;Young people often lack trusted sources for sexual health information. An AI chatbot available on their phone 24/7 can answer questions accurately and anonymously, which is a huge step forward.\u0026rdquo;\u003c/em\u003e\u003cem\u003e\u0026nbsp;\u003c/em\u003e\u003c/p\u003e\n\u003cp\u003eMany participants cited rural populations with limited English proficiency as an example for ensuring localized AI solutions. One scoping workshop speaker asked, \u0026ldquo;\u003cem\u003eIs it possible for me to deposit my Ateso [local language in Uganda] somewhere so that it can be utilized\u0026hellip;\u003c/em\u003e?\u0026rdquo; (scoping workshop) This underscores the urgent need for AI-driven tools that incorporate local languages and dialects, ensuring that critical MSRH information is accessible to all.\u0026nbsp;\u003c/p\u003e\n\u003cp\u003eCo-creation and inclusive design\u0026nbsp;were mentioned as key strategies to tackling the language and culture concern. Several contributors advocated for a co-creation process involving health workers, community members, and technology developers. \u0026ldquo;\u003cem\u003eIt\u0026rsquo;s important to have a co-creation process in the way that AI is developed,\u003c/em\u003e\u0026rdquo; (scoping workshop) as one attendee mentioned.\u0026nbsp;\u003c/p\u003e\n\u003cp\u003eAs noted in a scoping workshop, this cultural tailoring is vital when implementing AI in Africa, where stigma around sexual health influences care-seeking behaviors. \u0026ldquo;\u003cem\u003e\u0026hellip;ensuring it\u0026rsquo;s relatable to the daily experiences of these (specific) communities [is a key to successful uptake].\u0026rdquo;\u003c/em\u003e (scoping workshop)\u003c/p\u003e\n\u003cp\u003eSuch collaboration ensures user-friendly designs that accurately reflect community knowledge, practices and ethical priorities. Additionally, when communities feel a sense of ownership over a tool, they are more likely to trust it and use it.\u0026nbsp;\u003c/p\u003e\n\u003cp\u003eIn addition to the language and cultural concerns, several scoping workshop participants also touched on the importance of interdisciplinary collaboration to breaking down silos between technology developers and healthcare practitioners, social scientists, and the intended beneficiaries of the technology. One scoping workshop participant emphasized: \u0026ldquo;\u003cem\u003eWe learned that when engineers sit with midwives and doctors, they come up with much more practical solutions. Neither can do it alone \u0026mdash; it has to be a joint effort.\u003c/em\u003e\u0026rdquo;\u0026nbsp;\u003c/p\u003e\n\u003cp\u003eThrough collaboration, AI solutions are more likely to address real-world problems in a feasible way. One illustrative example from the discussion in a scoping workshop was a pilot project mentioned by one scoping workshop participant, where an AI tool for predicting postpartum hemorrhage risk was developed with direct input from obstetricians: the clinicians specified what risk factors they saw as red flags, and the engineers used those insights to train a model that aligned with clinical intuition. Such collaborative models help ensure the resulting AI is not a \u0026ldquo;black box\u0026rdquo; but something clinicians feel connected to. \u0026nbsp;Participants in scoping workshops, FGDs and KIIs were all asked about priorities for research and development. Some suggestions highlighted the importance of communication channels. CAB members signaled the area of how to incorporate AI in building awareness for high-risk health issues at the community level to prevent emergency situations. \u0026nbsp;Another CAB member suggested not to focus only on the mother, but to include other family members such as the male partner in pregnancy-related awareness raising. \u0026nbsp;\u0026nbsp;\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eTIME, IMPLEMENTATION AND ACTUALIZATION in AI solutions:\u0026nbsp;\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eChallenges that impact time from innovation to diffusion\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eParticipants\u0026rsquo; concerns revolved around six main areas including: ethics and regulation, cost, specifically internet accessibility costs and sustainability, trust in new innovations, access including population literacy, data and bias as well as cultural and religious barriers. \u0026nbsp;One participant mentioned that like current use of Google, there could be a risk of increase in self-medication, if people use AI alone to inform their medical choices.\u003c/p\u003e\n\u003cp\u003e\u003cem\u003ein situations where\u0026nbsp;you are supposed to go and see your doctor, you are relying on what your chatbot or your application is telling you. You misuse it. . . . It may lead to abuse of information, abuse of access to information you are supposed to use positively but people tend to abuse it\u003c/em\u003e. (KI Nigeria)\u003c/p\u003e\n\u003cp\u003e\u003cem\u003eRegulation and Ethical Concerns\u003c/em\u003e: Marginalized populations, including women in patriarchal societies, could face barriers to benefiting from AI. \u0026nbsp;Confidentiality breaches were mentioned as a concern with use of AI. Regulatory frameworks have not caught up to the pace of innovations in many SSA countries.\u003c/p\u003e\n\u003cp\u003eAI expert participants described some of the complexities ranging from data literacy and technology access bottlenecks.\u003c/p\u003e\n\u003cp\u003e\u003cem\u003eData and Bias Issues:\u003c/em\u003e In addition, both community and expert participants highlighted that AI models can face limited data sources in low-research languages and regions, compounded by biases in existing datasets. This can lead to inequitable outcomes, especially in marginalized communities.\u003c/p\u003e\n\u003cp\u003e\u003cem\u003eAccess to health data is not straightforward because of the sensitivity of the information; privacy is very important. Stakeholders do not always understand the importance of AI tools and hence are skeptical about having health data shared with developers. Lack of a central repository for data resulting in segmented data. There is need for data, otherwise AI cannot exist. Data acquisition is a challenge. Data entrants may not fully understand the concept of unique identifiers and may choose to use traceable data e.g. phone numbers as a unique ID. . \u0026nbsp;Medical workers and data entrants may not understand how to encrypt data; lack of knowledge on data encryption may limit data sharing where it would otherwise be possible. We must be careful what data we train AI on because what it is fed is what it gives. Not all data is good for AI. In many cases the data being used was not initially collected for AI use, so there may be a lot of missing data elements or assumptions. Data acquisition takes a long time and causes a lot of frustration to developers, especially because it involves a lot of stakeholder engagement. (KII Uganda)\u003c/em\u003e\u003c/p\u003e\n\u003cp\u003e\u003cem\u003eCost, including connectivity and Infrastructure:\u003c/em\u003e All groups of participants noted that limited internet access and inconsistent electricity in low-income regions could lead to either low uptake or low capacity to sustain these projects.\u003c/p\u003e\n\u003cp\u003e\u003cem\u003ecan we sustain it? Do we have the people we are going to work with? Are they well versed with the technology\u003c/em\u003e (CAB FGD)\u003c/p\u003e\n\u003cp\u003e\u003cem\u003eTrust and Literacy:\u003c/em\u003e It is common across the literature that initiation of new innovations often instill hesitancy. This was found in our data as well that adopting AI instill some fear of change, concerns about job loss, and low education levels, specifically literacy levels and more specifically in areas and populations with low data literacy among some population groups including community health workers. Some participants mentioned concern that their community would not know how to use AI correctly, thus could endanger their health instead of benefitting. A lack of collaboration and data sharing between stakeholders hinders progress.\u0026nbsp;\u003c/p\u003e\n\u003cp\u003e\u003cem\u003eCultural and Religious Barriers:\u003c/em\u003e Some communities resist technological innovations due to religious beliefs or cultural norms. Additionally, the belief in the importance of human-to-human connection and the risk of losing that with increased use of AI.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003e\u0026ldquo;\u003c/strong\u003e\u003cem\u003esomeone can misinterpret the guidelines or the information that AI is using to maybe give treatment or diagnosis and come up with a decision that will affect the patient in due course\u0026rdquo;\u003c/em\u003e\u003cstrong\u003e\u0026nbsp;. . .\u0026nbsp;\u003c/strong\u003e\u003cem\u003ecertain people believe that if I go to the health center and I don\u0026rsquo;t find Stella, I might not be treated well, I would rather go to someone I know is not even a health person but just talking to her or him might cure me\u003cstrong\u003e.\u0026rdquo;\u0026nbsp;\u003c/strong\u003e\u003c/em\u003e\u003cstrong\u003e(CAB FGD)\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eSOCIAL SYSTEM\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eGenerally, within the Diffusion of Innovations Theory, one considers the social system as including social norms, cultural values, social networks, political factors like government policies, ethics, regulations, and infrastructure.\u003c/p\u003e\n\u003cp\u003eOur data suggested that policy makers need to think ahead for sustainability to ensure capacity building. There were multiple examples highlighted of knowledge, experience, and attitude gaps that will hinder sustainability.\u003c/p\u003e\n\u003cp\u003e\u003cem\u003e(for sustainability) we need to find ways to engage the opinion leaders at a community level early on, . . . , we need to explore the use of experts (to educate opinion leaders)\u003c/em\u003e (CAB FGD)\u003c/p\u003e\n\u003cp\u003e\u003cem\u003ewe need to have a systems thinking approach however hard it is\u003c/em\u003e (CAB FGD)\u003c/p\u003e\n\u003cp\u003eAll groups of participants stressed that tailored engagement and inclusive strategies would be necessary to overcome barriers. \u0026nbsp;Addressing challenges requires a holistic approach, combining technical solutions that work with societal, educational, and policy interventions to strengthen the diffusion of any of the AI maternal health innovations within the social system.\u003c/p\u003e\n\u003cp\u003eA key point highlighted in a scoping workshop was the strategy of continuous evaluation and feedback. One speaker noted that implementing AI in MSRH should be an iterative process: \u0026ldquo;\u003cem\u003eWe should deploy AI carefully and study its impact. Are mothers actually healthier? Are providers happier? We need to gather feedback and be ready to improve the tools.\u003c/em\u003e\u0026rdquo; In practice, this would imply using traditional clinical evidence methods such as randomized controlled trials for efficiacy compared to humans alone, as well as monitoring outcomes once an AI system is rolled out (for example, tracking if an AI triage system actually reduced waiting times at clinics or improved maternal mortality rates in an area). Getting feedback from users would allow developers and health authorities to refine the system or even decide to scale it up or down. This approach would, if working well, ensure that AI remains aligned with the people\u0026rsquo;s needs over time, adapting to new information or changing circumstances in close to real time.\u0026nbsp;\u003c/p\u003e\n\u003cp\u003eWith regards to the ethical concerns, participants mentioned strengthening ethical protocols by ensuring data protection and combating institutional biases. It was emphasized that it is key to ensure that there is always a human in the loop, this is important to the general population in addition to the specifics for MSRH.\u0026nbsp;\u003c/p\u003e\n\u003cp\u003eTo alleviate infrastructural issues such as the cost of or lack of internet, participants suggested that future models would necessitate AI models that can function offline and remain efficient under resource constraints. In the FGD, some reflections on sustainability revolved around how use of cultural opinion leaders, scientific experts, and other stakeholders enable a longer term vision and can contextualize AI use for MSRH in SSA.\u0026nbsp;\u003c/p\u003e\n\u003cp\u003eParticipants in the scoping workshop cautioned that using AI cannot instantly remove all the challenges in our MSRH services, it will only provide one piece of the complex puzzle.\u003c/p\u003e\n\u003cp\u003e\u003cem\u003e\u0026ldquo;We need AI precisely because we need to do better for mothers and young people. But AI is not a silver bullet for all our health system problems,\u0026rdquo; the participant cautioned. (scoping workshop)\u003c/em\u003e\u003c/p\u003e"},{"header":"DISCUSSION","content":"\u003cp\u003eAcross the participant groups and data collection methods, we found a high level of diversity in the experiences and perceptions of AI in MSRH in sub-Saharan Africa. \u0026nbsp;In high income countries AI in MSRH is predicted to touch all areas of health care from disease prediction to health promotion in the very near future.\u003csup\u003e16\u003c/sup\u003e Our data clearly highlighted that stakeholders agree that AI is needed in MSRH because it offers benefits though powerful tools to improve outcomes and equity. Our findings showed the importance of AI innovations as a support tool to strengthen MSRH healthcare delivery including: improving stock management and triage, increasing MSRH awareness, enhancing diagnosis, guiding treatments, and enabling personalized care. However, our findings suggest that this same level of confidence in AI for MSRH among early adopters has not been reached to allow for a wide diffusion and for the potential of AI in MSRH among the diversity of African stakeholders.\u0026nbsp;\u003c/p\u003e\n\u003cp\u003eA global systematic review of health care professional experience by Ayorinde et al identified health care professionals’ understanding of AI applications, level of trust and confidence in AI tools and judging the value added by AI as main areas raised in their study.\u003csup\u003e17\u003c/sup\u003e In Africa, there have been several quantitative interview studies of health worker perceptions including trainee radiologists .\u003csup\u003e18,19\u003c/sup\u003e A comprehensive mixed methods study by Asiedu with health workers and general public across Africa revealed more positive attitudes towards AI from general public compared to more cautious health workers who identified trust, ethics and systemic barriers to integration. For MSRH specific work, a review of midwife perceptions revealed eight studies highlighting potential for improved quality of care, particularly in perinatal and neonatal settings, with barriers to integration of ethical concerns and hesitation among midwives, due to low levels of digital health literacy. We have not found any previous comprehensive qualitative studies exploring AI in MSRH.\u003c/p\u003e\n\u003cp\u003eParticipants in our study, were steadfast in emphasizing the core principle that AI should augment, not replace human healthcare. As other studies have highlighted lack of awareness of health workers about the opportunities of health AI; health workers are thirsty for knowledge about how they can use AI, and how to make sure it is used safely and appropriately. A novel finding from our study is the emphasis on co-creation of AI tools with health workers, traditional health workers and communities as a way to ensure that the tools benefit those most in need and are safe.\u0026nbsp;\u003c/p\u003e\n\u003cp\u003eOur data reinforces previous experience with health technology adoption, that it must go hand-in-hand with strengthening healthcare infrastructure and workforce to effectively use the innovations. The consensus across our data, as well as others, cautions that AI must be implemented thoughtfully and carefully alongside improvements in capacity, regulation and infrastructure.\u003csup\u003e8,20\u003c/sup\u003e Rigorous evaluation of new AI tools pre and during implementation was highlighted by participants as an important process that would support trust in the tools.\u0026nbsp;\u003c/p\u003e\n\u003cp\u003eEnsuring adequate and non-biased data was a concern that was less prevalent in health worker facing studies; this may be due to the inclusion of AI experts in this study\u003csup\u003e18,21,22\u003c/sup\u003e This is particularly important in MSRH, as women are often missing in research data, formal medical records and other socio-economic datasets. Notably in our study and in others were the challenges related to ethics and regulation.\u003csup\u003e23\u003c/sup\u003e\u0026nbsp;\u003c/p\u003e\n\u003cp\u003eAI innovations have been highlighted as a potential to revolutionize maternal and other health care in Africa, yet large scale studies are limited across the continent.\u003csup\u003e8,24\u003c/sup\u003e Our study found that content area specialists believe there are diverse AI innovations that could improve, widen the reach and increase equity in maternal, sexual and reproductive health yet, their diffusion depends on a number of factors related to all four key pillars of the theory, the innovation itself that is built on complex infrastructure, the communication channels that depend on accessibility to technology, the speed with which access is available for new innovative projects and the social system comprised of cultural and language, challenges. We referred to the diffusion of innovations theory in our analysis of findings to help frame our results. We found that diffusion of AI innovations has been hampered by awareness in the general public of specific AI innovations, complexity of many of the innovations, trialability, observability and in the realm of ethics and regulations.\u003csup\u003e25\u003c/sup\u003e\u003c/p\u003e\n\u003cp\u003eThe strengths of this study are the triangulation of three different settings, key informant interviews with international African AI and MSRH experts, FGD with a community advisory board, and observations at a scoping workshop. This allows for a wide cross-sectional view of different perceptions of AI. The potential weakness of this study is that there was a large gap in time between KII-FGD and the scoping workshop, and we were initially expecting results to be very different and require separate reporting; however during analysis the themes were more similar than expected so the data were combined.\u003c/p\u003e\n\u003cp\u003eOverall, the consensus of our participants suggested that AI has great potential to tackle the urgent MSRH challenges that disproportionality affect women and adolescent girls.\u0026nbsp;\u003c/p\u003e\n\u003cp\u003eAI for MSRH has potential to improve quality and efficiency of MSRH health care delivery in Africa and to widen the reach of health services. For the AI innovations to be disseminated equitably, increase in awareness of the characteristics of AI innovations, their layered complexities, capacity to understand and use them fairly and improving infrastructure will all be critical. Co-creation of AI tools with communities, heath workers and technologists and rigorous evaluation may help to create safe, trustworthy and easily adopted AI tools for MSRH.\u0026nbsp;\u003c/p\u003e"},{"header":"References","content":"\u003col\u003e\n \u003cli\u003eWHO. \u003cem\u003eWorld Health Organization Global adult estimates\u003c/em\u003e, \u0026lt;https://www.who.int/reproductivehealth/topics/rtis/STIs-Estimates.pdf\u0026gt; (2020).\u003c/li\u003e\n \u003cli\u003eWHO. \u003cem\u003eSexual and Reproductive Health\u003c/em\u003e, \u0026lt; https://www.who.int/reproductivehealth/congenital-syphilis-estimates/en/\u0026gt; (2022).\u003c/li\u003e\n \u003cli\u003eAdejumo, O. A., Malee, K. M., Ryscavage, P., Hunter, S. J. \u0026amp; Taiwo, B. O. Contemporary issues on the epidemiology and antiretroviral adherence of HIV-infected adolescents in sub-Saharan Africa: a narrative review. \u003cem\u003eJ Int AIDS Soc\u003c/em\u003e \u003cstrong\u003e18\u003c/strong\u003e, 20049 (2015). https://doi.org/10.7448/IAS.18.1.20049\u003c/li\u003e\n \u003cli\u003eIdele, P.\u003cem\u003e\u0026nbsp;et al.\u003c/em\u003e Epidemiology of HIV and AIDS among adolescents: current status, inequities, and data gaps. \u003cem\u003eJ Acquir Immune Defic Syndr\u003c/em\u003e \u003cstrong\u003e66 Suppl 2\u003c/strong\u003e, S144-153 (2014). https://doi.org/10.1097/QAI.0000000000000176\u003c/li\u003e\n \u003cli\u003eMbadugha, N. \u003cem\u003eBig Data and AI in Africa: Why the Future of the Continent Is Artificially Intelligent and Digitally Enabled\u003c/em\u003e, \u0026lt;https://nextbillion.net/big-data-ai-africa-artificial-intelligence-digital/\u0026gt; (2021).\u003c/li\u003e\n \u003cli\u003eBolarinwa, O., Adebisi, Y. A., Ajayi, K. V. \u0026amp; Boutahar, R. Sexual and reproductive health rights in the era of artificial intelligence. \u003cem\u003eLancet\u003c/em\u003e \u003cstrong\u003e404\u003c/strong\u003e, 120-121 (2024). https://doi.org/10.1016/S0140-6736(24)00696-2\u003c/li\u003e\n \u003cli\u003eBolarinwa, O. A., Mohammed, A. \u0026amp; Igharo, V. The role of artificial intelligence in transforming maternity services in Africa: prospects and challenges. \u003cem\u003eTher Adv Reprod Health\u003c/em\u003e \u003cstrong\u003e18\u003c/strong\u003e, 26334941241288587 (2024). https://doi.org/10.1177/26334941241288587\u003c/li\u003e\n \u003cli\u003eTownsend, B. A.\u003cem\u003e\u0026nbsp;et al.\u003c/em\u003e Mapping the regulatory landscape of AI in healthcare in Africa. \u003cem\u003eFront Pharmacol\u003c/em\u003e \u003cstrong\u003e14\u003c/strong\u003e, 1214422 (2023). https://doi.org/10.3389/fphar.2023.1214422\u003c/li\u003e\n \u003cli\u003eValente, T. W. Diffusion of innovations. \u003cem\u003eGenet Med\u003c/em\u003e \u003cstrong\u003e5\u003c/strong\u003e, 69 (2003). https://doi.org/10.1097/01.GIM.0000061743.67794.C4\u003c/li\u003e\n \u003cli\u003eValente, T. W., Dyal, S. R., Chu, K. H., Wipfli, H. \u0026amp; Fujimoto, K. Diffusion of innovations theory applied to global tobacco control treaty ratification. \u003cem\u003eSoc Sci Med\u003c/em\u003e \u003cstrong\u003e145\u003c/strong\u003e, 89-97 (2015). https://doi.org/10.1016/j.socscimed.2015.10.001\u003c/li\u003e\n \u003cli\u003eValente, T. W. \u0026amp; Fosados, R. Diffusion of innovations and network segmentation: the part played by people in promoting health. \u003cem\u003eSex Transm Dis\u003c/em\u003e \u003cstrong\u003e33\u003c/strong\u003e, S23-31 (2006). https://doi.org/10.1097/01.olq.0000221018.32533.6d\u003c/li\u003e\n \u003cli\u003eHASH. \u003cem\u003eBaseline Quantitative Online Survey\u0026nbsp;\u003c/em\u003e\u0026lt;https://hash.theacademy.co.ug/wp-content/uploads/2022/11/Survey-results_Baseline-Consultation-About-the-Landscape-of-AI-in-Africa.pdf\u0026gt; (2024).\u003c/li\u003e\n \u003cli\u003eCollaco, N.\u003cem\u003e\u0026nbsp;et al.\u003c/em\u003e Using the Framework Method for the Analysis of Qualitative Dyadic Data in Health Research. \u003cem\u003eQual Health Res\u003c/em\u003e \u003cstrong\u003e31\u003c/strong\u003e, 1555-1564 (2021). https://doi.org/10.1177/10497323211011599\u003c/li\u003e\n \u003cli\u003eGale, N. K., Heath, G., Cameron, E., Rashid, S. \u0026amp; Redwood, S. Using the framework method for the analysis of qualitative data in multi-disciplinary health research. \u003cem\u003eBMC Med Res Methodol\u003c/em\u003e \u003cstrong\u003e13\u003c/strong\u003e, 117 (2013). https://doi.org/10.1186/1471-2288-13-117\u003c/li\u003e\n \u003cli\u003eKlingberg, S., Stalmeijer, R. E. \u0026amp; Varpio, L. Using framework analysis methods for qualitative research: AMEE Guide No. 164. \u003cem\u003eMed Teach\u003c/em\u003e \u003cstrong\u003e46\u003c/strong\u003e, 603-610 (2024). https://doi.org/10.1080/0142159X.2023.2259073\u003c/li\u003e\n \u003cli\u003eWilkins, A. \u003cem\u003eConcerns raised over AI trained on 57 million NHS medical records\u003c/em\u003e, \u0026lt;https://www.newscientist.com/article/2479302-concerns-raised-over-ai-trained-on-57-million-nhs-medical-records/\u0026gt; (2025).\u003c/li\u003e\n \u003cli\u003eAyorinde, A.\u003cem\u003e\u0026nbsp;et al.\u003c/em\u003e Health Care Professionals\u0026apos; Experience of Using AI: Systematic Review With Narrative Synthesis. \u003cem\u003eJ Med Internet Res\u003c/em\u003e \u003cstrong\u003e26\u003c/strong\u003e, e55766 (2024). https://doi.org/10.2196/55766\u003c/li\u003e\n \u003cli\u003eNciki, A. I. \u0026amp; Hlabangana, L. T. Perceptions and attitudes towards AI among trainee and qualified radiologists at selected South African training hospitals. \u003cem\u003eSA J Radiol\u003c/em\u003e \u003cstrong\u003e29\u003c/strong\u003e, 3026 (2025). https://doi.org/10.4102/sajr.v29i1.3026\u003c/li\u003e\n \u003cli\u003eAsiedu M, H. I., Dieng A, Kauer K, Ahmed T, Ofori F, Chan C, Pfohl S, Rostamzadeh N, Heller K. in \u003cem\u003eACM Conference on Fairness, Accountability, and Transparency\u003c/em\u003e (ACM, New York, NY, USA, Athens, Greece, 2025).\u003c/li\u003e\n \u003cli\u003eReid, M. J. A.\u003cem\u003e\u0026nbsp;et al.\u003c/em\u003e Announcing the Lancet Global Health Commission on artificial intelligence (AI) and HIV: leveraging AI for equitable and sustainable impact. \u003cem\u003eLancet Glob Health\u003c/em\u003e \u003cstrong\u003e13\u003c/strong\u003e, e611-e612 (2025). https://doi.org/10.1016/S2214-109X(25)00049-X\u003c/li\u003e\n \u003cli\u003eGiaxi, P., Vivilaki, V., Sarella, A. \u0026amp; Gourounti, K. Artificial Intelligence in Midwifery: A Scoping Review of Current Applications, Future Prospects, and Midwives\u0026apos; Perspectives. \u003cem\u003eHealthcare (Basel)\u003c/em\u003e \u003cstrong\u003e13\u003c/strong\u003e (2025). https://doi.org/10.3390/healthcare13080942\u003c/li\u003e\n \u003cli\u003eGiaxi, P., Vivilaki, V., Sarella, A., Harizopoulou, V. \u0026amp; Gourounti, K. Artificial Intelligence and Machine Learning: An Updated Systematic Review of Their Role in Obstetrics and Midwifery. \u003cem\u003eCureus\u003c/em\u003e \u003cstrong\u003e17\u003c/strong\u003e, e80394 (2025). https://doi.org/10.7759/cureus.80394\u003c/li\u003e\n \u003cli\u003eFjeld, J., Achten, N., Hilligoss, H., Nagy, A., and Srikumar, M. Principled Artificial Intelligence: Mapping Consensus in Ethical and Rights-Based Approaches to Principles for AI. (2020).\u003c/li\u003e\n \u003cli\u003eChemisto, M., Gutu, TJL, Kalinaki, K, Bosco, DM, Egau, P, Fred, K, Oloya, IT, Rashid, K. (Nairobi, Kenya, 2023).\u003c/li\u003e\n \u003cli\u003ePham, T. Ethical and legal considerations in healthcare AI: innovation and policy for safe and fair use. \u003cem\u003eR Soc Open Sci\u003c/em\u003e \u003cstrong\u003e12\u003c/strong\u003e, 241873 (2025). https://doi.org/10.1098/rsos.241873\u003c/li\u003e\n\u003c/ol\u003e"}],"fulltextSource":"","fullText":"","funders":[],"hasAdminPriorityOnWorkflow":false,"hasManuscriptDocX":true,"hasOptedInToPreprint":true,"hasPassedJournalQc":"","hasAnyPriority":true,"hideJournal":true,"highlight":"","institution":"Infectious Diseases Institute, Kampala, Uganda","isAcceptedByJournal":false,"isAuthorSuppliedPdf":false,"isDeskRejected":"","isHiddenFromSearch":false,"isInQc":false,"isInWorkflow":false,"isPdf":false,"isPdfUpToDate":true,"isWithdrawnOrRetracted":false,"journal":{"display":true,"email":"[email protected]","identity":"researchsquare","isNatureJournal":false,"hasQc":true,"allowDirectSubmit":true,"externalIdentity":"","sideBox":"","snPcode":"","submissionUrl":"/submission","title":"Research Square","twitterHandle":"researchsquare","acdcEnabled":true,"dfaEnabled":false,"editorialSystem":"","reportingPortfolio":"","inReviewEnabled":false,"inReviewRevisionsEnabled":true},"keywords":"","lastPublishedDoi":"10.21203/rs.3.rs-7463408/v1","lastPublishedDoiUrl":"https://doi.org/10.21203/rs.3.rs-7463408/v1","license":{"name":"CC BY 4.0","url":"https://creativecommons.org/licenses/by/4.0/"},"manuscriptAbstract":"\u003cp\u003eArtificial intelligence has the potential to transform healthcare in low- and middle-income countries, where access to quality care remains limited. Maternal, sexual, and reproductive health (MSRH) outcomes are especially poor due to resource shortages, financial barriers, and geographic inequities. With thoughtful implementation, AI could help address these gaps through innovations in diagnostics, health education chatbots, and telemedicine. However, responsible use is essential to ensure AI reduces\u0026mdash;rather than exacerbates\u0026mdash;health disparities between high- and low-income regions. Our study examines the perceptions, uses, benefits, and challenges of AI in MSRH among medical professionals, community members, and AI experts, guided by the Diffusion of Innovations Theory. Our findings will inform the development of a continent-wide AI hub for MSRH, highlighting barriers and opportunities for improving health care access. We aim to support policymakers, researchers, and implementers in using AI to promote equitable maternal and sexual and reproductive healthcare delivery across Africa.\u003c/p\u003e","manuscriptTitle":"A Qualitative Inquiry Exploring Perceptions of Artificial Intelligence to Improve Outcomes in Maternal, Sexual and Reproductive Health in Sub-Saharan Africa","msid":"","msnumber":"","nonDraftVersions":[{"code":1,"date":"2025-09-03 06:09:14","doi":"10.21203/rs.3.rs-7463408/v1","editorialEvents":[{"type":"communityComments","content":0}],"status":"published","journal":{"display":true,"email":"[email protected]","identity":"researchsquare","isNatureJournal":false,"hasQc":true,"allowDirectSubmit":true,"externalIdentity":"","sideBox":"","snPcode":"","submissionUrl":"/submission","title":"Research Square","twitterHandle":"researchsquare","acdcEnabled":true,"dfaEnabled":false,"editorialSystem":"","reportingPortfolio":"","inReviewEnabled":false,"inReviewRevisionsEnabled":true}}],"origin":"","ownerIdentity":"e0d9c113-b7e2-41ee-8523-cb7ebcaac1f2","owner":[],"postedDate":"September 3rd, 2025","published":true,"recentEditorialEvents":[],"rejectedJournal":[],"revision":"","amendment":"","status":"posted","subjectAreas":[{"id":53955574,"name":"Artificial Intelligence and Machine Learning"}],"tags":[],"updatedAt":"2025-09-03T06:09:15+00:00","versionOfRecord":[],"versionCreatedAt":"2025-09-03 06:09:14","video":"","vorDoi":"","vorDoiUrl":"","workflowStages":[]},"version":"v1","identity":"rs-7463408","journalConfig":"researchsquare"},"__N_SSP":true},"page":"/article/[identity]/[[...version]]","query":{"redirect":"/article/rs-7463408","identity":"rs-7463408","version":["v1"]},"buildId":"8U1c8b4HqxoKbykW_rLl7","isFallback":false,"isExperimentalCompile":false,"dynamicIds":[84888],"gssp":true,"scriptLoader":[]}