Clinician Perceptions of Artificial Intelligence in Healthcare and Frameworks for Ensuring Safe Integration into Clinical Practice of the West African College of Physicians: A Multi-Country Mixed-Methods Study

Clinician Perceptions of Artificial Intelligence in Healthcare and Frameworks for Ensuring Safe Integration into Clinical Practice of the West African College of Physicians: A Multi-Country Mixed-Methods Study | 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 Clinician Perceptions of Artificial Intelligence in Healthcare and Frameworks for Ensuring Safe Integration into Clinical Practice of the West African College of Physicians: A Multi-Country Mixed-Methods Study Benjamin S. Chudi Uzochukwu, Yakubu Joel Cherima, Ugo Uwadiako Enebeli, and 7 more This is a preprint; it has not been peer reviewed by a journal. https://doi.org/ 10.21203/rs.3.rs-9066307/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 The integration of artificial intelligence (AI) into clinical practice holds significant promise for improving diagnostic accuracy, reducing medical errors, and enhancing healthcare efficiency, particularly in resource-constrained settings. However, successful adoption depends heavily on clinicians’ perceptions, trust, and concerns regarding autonomy, reliability, and safety. Empirical evidence on West African physicians’ views remains limited, despite unique regional challenges like infrastructure deficits and workforce shortages. This study aimed to assess clinicians’ perceptions of AI in healthcare, identify factors influencing willingness to adopt AI tools, and explore recommended frameworks for safe, accountable integration into clinical practice across West Africa. A cross-sectional survey of 136 physicians affiliated with the West African College of Physicians and 72 key informant interviews were conducted. While 85.3% agreed that AI could improve diagnostic accuracy and 83.1% believed it could reduce errors, 77.9% perceived AI as a threat to clinical autonomy, and 67.6% rated AI information as unreliable. Despite low prior AI experience (only 33.1% had used AI tools) and limited familiarity, 94.1% expressed willingness to use AI if proven effective. Trust in AI was the strongest predictor of adoption willingness (β = 0.48, p < 0.001), with prior use also significant (β = 0.21, p = 0.03). Younger clinicians (0–10 years’ experience) showed higher willingness than those with 20 + years (mean scores 4.5 vs. 4.1, p < 0.05). Qualitative findings highlighted AI’s potential as a cognitive partner for decision support, error reduction, and administrative relief but raised major concerns about over-reliance, skill erosion, workflow disruption, and accountability. Clinicians emphasised three core requirements for trust: transparency and explainability, local validation in similar populations, and clear governance with defined accountability mechanisms. West African clinicians recognise AI’s potential benefits but exhibit low current trust and significant autonomy concerns, driven by limited experience, perceived unreliability, and contextual barriers. Willingness to adopt was highly conditional on proven effectiveness. Safe integration of AI requires frameworks prioritising transparency, local validation, clinician-centred design, robust governance (preferably independent oversight), infrastructure investment, and AI literacy in medical training. These findings guide contextually appropriate AI policies and implementation strategies to enhance patient safety and care quality in West Africa. Clinical trial number : Not applicable. artificial intelligence clinician perceptions healthcare adoption trust in AI clinical autonomy West Africa safe integration AI governance transparency local validation BACKGROUND The use of artificial intelligence (AI) in clinical practice is a landmark in healthcare delivery. It improves the accuracy of diagnoses, the planning of treatments, the making of clinical decisions, and the overall performance of the health system. 1 , 2 AI is being used increasingly in many areas of healthcare around the world, including medical imaging, pathology, predictive analytics, and clinical decision support. Recent studies indicate that these applications can significantly enhance efficiency and patient outcomes. 3 , 4 However, as noted by some authors, the successful integration and utilisation of AI in healthcare will largely depend on doctors' perspectives, trust in AI, and its acceptance. 5 , 6 This is because health technologies are mostly used and controlled by doctors. Therefore, their willingness to use AI tools, trust in AI-generated recommendations, and understanding of the limitations of AI will all have a direct effect on whether these technologies are used, how they are integrated into clinical workflows, and whether they will ultimately lead to better patient care. 7 , 8 Resistance from clinicians in developed countries is as a result of concerns about trust, accountability and autonomy issues, and these are largely responsible for a good number of barriers to AI integration. 9 , 10 Adoption of AI in West African healthcare systems has its own difficulties that could influence how doctors see AI compared to systems with more resources. Health systems in the region encounter significant resource challenges, including insufficient healthcare personnel, inadequate diagnostic equipment, unreliable electrical supply, and unstable internet connections. 11 , 12 These contextual aspects create both imperatives for AI adoption, as a potential means for task-shifting and quality enhancement, and considerable implementation constraints that may influence physicians’ perceptions and interactions with AI technology. 13 , 14 While there is a growing body of international literature on AI adoption, empirical research regarding physician perspectives of AI in West African contexts is still limited. 15 , 16 It is important to understand these views. For example, health system characteristics affect how people adopt technology in different ways, which means that results from high-income nations may not be directly relevant to the West African context. 17 , 18 It is also important to know what West African doctors know and want and what information they require to create effective implementation strategies, training programmes, and governance frameworks. 19 , 20 They also need to be able to examine new AI tools and discover possible faults, especially those that will compromise patient safety. The level of knowledge and perspectives of AI systems determines these skills. 21 , 22 The concept of “AI assurance” - that is, the systematic processes and structures required to ensure that AI systems are safe, dependable, accountable, ethically sound, and compliant with therapeutic standards - has garnered significant global attention. 23 , 24 The criteria for effective AI assurance in resource-constrained environments remain insufficiently defined, and there is limited direction on how West African health systems should evaluate, regulate, and monitor AI tools both before and after implementation. 25 , 26 Clinicians are crucial partners in the development of these frameworks, as their own experiences and concerns must inform the establishment of governance systems that are both robust and practically effective. 27 , 28 This study filled in these gaps in the evidence by looking at what doctors in West Africa think about AI in healthcare settings, what factors affect its use, and what doctors recommend to make sure it is safely integrated. This aims to create strategies for using AI that are appropriate for the situation and will improve, not hurt, patient safety and quality of care. Research Questions : What do West African doctors think about using AI in healthcare? What factors make clinicians more likely to accept and use AI tools in West African healthcare settings? What do clinicians think are the frameworks that need to be in place to make sure that AI is used safely, reliably, and with accountability in clinical practice? How do clinician perceptions of AI differ across facility types (tertiary, secondary, and primary), medical specialties, and levels of experience? MATERIALS AND METHODS Study design and study area This cross-sectional mixed-methods study included a structured, self-administered questionnaire to assess the attitudes of West African professionals regarding artificial intelligence in healthcare and a qualitative exploration through key informant interviews (KII). The study was conducted in the West African subregion. Data was gathered from February 22 to February 28, 2026. Study Population and Sampling The target group comprised physicians affiliated with the West African College of Physicians (WACP), the principal institution for postgraduate medical education in the sub-region. The membership of this college covers both the Anglophone, Francophone, and Lusophone countries of the subregion. Most of the WACP fellows and members work in Nigeria; however, they can work in any West African country. This is because the college was started there and has a lot of students. A convenience sampling method was used with the college's current communication channels to find willing participants. Any physician listed in the WACP contact database during the study period was eligible for inclusion. The study sought to capture a comprehensive spectrum of physician perspectives from the sub-region, imposing no restrictions on specialty, years of experience, or practice environment. Sample size determination A mixed-methods approach was employed. For the quantitative phase, a cross-sectional survey was administered to physicians of the West African College of Physicians. The sample size was calculated on the WinPepi sample size calculator, version 11.65, 29 to estimate the primary outcome (e.g., 'positive perception of AI utility') with a 95% confidence level, a 5% acceptable difference, and a 10% non-response rate. Using an assumed proportion of 92% (0.92) positive perception of AI utility from a previous study, 30 the initial sample size was 126. This was rounded up to 136 for more precision. Data Collection For the survey, the questionnaire had four parts: demographic and professional information (age, gender, country of practice, practice setting (primary/secondary/tertiary facility), specialty, years of clinical experience, and previous exposure to AI (familiarity and experience)); perceptions of AI in clinical practice (perceived benefits, concerns about autonomy, trust in AI-generated information, and perceived reliability). The 5-point Likert scale was used for scoring, with 1 being “strongly disagree” and 5 being “strongly agree”. The 5-point Likert scale was also used to rate how willing people were to use AI (how ready they were to use AI tools in different clinical situations) and how they wanted AI to be monitored and regulated (the bodies they wanted to do this, with options like independent bodies, government agencies, professional colleges, and vendors). The questionnaire was pre-tested with 12 WACP-affiliated physicians (not included in the final sample) from diverse specialties and experience levels to assess clarity, completion time, and comprehension. Minor wording adjustments were made based on feedback. Internal consistency of multi-item Likert scales (perceptions of benefits, concerns, and willingness) was evaluated post-data collection using Cronbach's alpha, with values ranging from 0.78 to 0.89, indicating acceptable to good reliability. Content validity was ensured through iterative expert review by the research team (including senior clinicians and AI specialists) during questionnaire development, drawing on established domains from prior AI healthcare perception studies. To prevent multiple entries from the same participant, the ODK link was designed for single-use submission, with built-in validation rules and skip logic to minimize data-entry errors. The Open Data Kit (ODK) platform was used for data collection. This is an open-source program that enables data collection, both online and offline, in regions where resources are scarce. The ODK was chosen because it works well in West Africa, where internet availability is quite diverse from country to country, and it has been used successfully in health research in the past. The ODK-based questionnaire with a link was sent through the WACP communication methods, like email and WhatsApp. The questionnaire took 15 to 20 minutes to complete. No personally identifiable information was collected, ensuring the anonymity of participants. To gain informed consent electronically, the first page of questions had to be filled out and agreed upon before moving on. For the qualitative study, a key informant interview (KII) approach was employed that enabled the exploration of complex phenomena, including perceptions, concerns, and recommendations regarding AI governance, clinical integration, and medical education, that cannot be adequately captured through quantitative methods alone. There was a total of 72 key informants’ interviews. The respondents were key stakeholders in the West Africa health space whose professional roles and expertise positioned them to produce authoritative perspectives. Each interview lasted less than 60 minutes using a semi-structured interview guide. All interview responses were captured through the online data collection platform, which recorded verbatim text entries for each open-ended question. This approach eliminated transcription errors associated with audio recording and transcription while ensuring complete capture of participant responses. For quality assurance purposes, all entries were reviewed for completeness and clarity within 24 hours of submission. Data Analysis The survey data from ODK was downloaded as Excel and exported into SPSS version 29 (IBM Corp., Armonk, NY, USA). 31 The descriptive statistics (frequencies, percentages, means, and standard deviations) were computed for each variable. Cronbach’s alpha was used to determine if multi-item scales were consistent with each other. Values greater than 0.70 were considered acceptable. Initially, the relationships between demographic and experiential characteristics were determined, as well as the desire to employ AI using t-tests. The ANOVA was used for inferential analysis. Factors exhibiting significant correlations (p < 0.10) were included in a multivariate linear regression model to forecast the propensity to employ AI. The regression model included previous AI experience (binary), familiarity with AI (Likert scale), trust in AI information (Likert scale), perceived danger to autonomy (Likert scale), years of experience (categorical), and facility type (categorical). The cutoff for statistical significance was set at p < 0.05 (two-tailed). Missing data was minimal across survey items (< 1% per variable, primarily in optional demographic fields) and was handled using listwise deletion for all inferential analyses (assumed missing completely at random given the low rate and random pattern observed). The sample’s demographic distribution (e.g., predominant Nigerian representation) aligns broadly with WACP membership patterns. The thematic analysis of the qualitative data was conducted using the framework approach described by Braun and Clarke. 32 This method was selected for its flexibility and suitability for identifying, analysing, and reporting patterns within qualitative data, particularly when working with policy-relevant research questions. Analysis proceeded through several stages, namely, data familiarisation, initial coding, framework development and application of framework. The final coding framework was systematically applied to all 72 transcripts using qualitative data analysis software NVivo version 14. 33 RESULTS Quantitative Results Respondents’ Characteristics The survey link was distributed through the West African College of Physicians (WACP) email and WhatsApp channels to all eligible physicians between 22 and 28 February 2026. A total of 137 physicians accessed the link, resulting in 136 complete responses included in the analysis after the exclusion of one participant who declined consent. Table 1 presents the demographic characteristics of the sample. The majority of respondents were from Nigeria (84.6%), with smaller representations from Ghana (5.9%), Liberia (3.6%), Togo (2.9%), Niger (1.5%), and Côte d'Ivoire (1.5%). Gender distribution was balanced (51.5% female, 48.5% male). The largest age group was 35–44 years (30.1%), followed by 45–54 years (26.5%). One-third of respondents (33.0%) had over 20 years of clinical experience. Table 1 Demographic characteristics of survey respondents Demographic Variables Category Frequency Percent Country Nigeria 115 84.6 Ghana 8 5.9 Liberia 5 3.6 Togo 4 2.9 Niger 2 1.5 Cote d’Ivoire 2 1.5 Gender Female 70 51.5 Male 65 47.8 Prefer not to say 1 0.7 Age group 25–34 26 19.1 35–44 41 30.1 45–54 36 26.5 55–64 21 15.4 65+ 11 8.1 Prefer not to say 1 0.7 Clinical Experience (Years) 0–5 15 11.0 6–10 27 19.9 11–15 27 19.9 16–20 22 16.2 20+ 45 33.0 Clinician familiarity and prior AI experience Table 2 presents findings on West African clinicians' self-reported familiarity with AI and prior experience using AI clinical tools. While a combined 73.5% reported being “somewhat/moderately familiar” with AI, only 19.1% considered themselves “highly familiar” or “experts”. Critically, half of respondents (52.2%) had never used an AI clinical tool, and another 14.7% were unsure whether they had. Table 2 Clinician familiarity and experience with AI Variable Category Frequency Percent (%) Familiarity with AI Not familiar 10 7.4 Somewhat familiar 34 25.0 Moderately familiar 66 48.5 Highly familiar 18 13.2 Expert 8 5.9 Prior Use of AI Tool Yes 45 33.1 No 71 52.2 Not Sure 20 14.7 Core perceptions of AI in clinical practice Table 3 shows that 77.9 percent indicated that they agree or strongly agree that AI threatens their clinical autonomy. A smaller portion, 10.3 percent, remained neutral on this issue, while 11.8 percent expressed disagreement. As high as 94.1% agreed that they would use AI if proven effective. Only 4.4 percent were neutral on this question, and a mere 1.5 percent disagreed. Regarding the potential of AI to reduce medical errors, 83.1 percent of those surveyed agreed or strongly agreed that AI could serve this function. 11.0 percent took a neutral stance, and 5.9 percent disagreed. Finally, when considering whether AI could improve diagnostic accuracy, 85.3 percent of respondents expressed agreement. 8.8 percent were neutral, and again, 5.9 percent disagreed with the statement. Table 3 Core perceptions of AI in clinical practice Perception Statement Agree/Strongly Agree (%) Neutral (%) Disagree/Strongly Disagree (%) AI threatens my clinical autonomy 77.9 10.3 11.8 I would use AI if proven effective 94.1 4.4 1.5 AI can help reduce clinical errors 83.1 11.0 5.9 AI can improve diagnostic accuracy 85.3 8.8 5.9 Perceived reliability of AI information Table 4 presents data on the perceived reliability of AI information, showing both the frequency and percentage of responses across four categories. For the category “Very reliable”, 8 respondents reported this perception, representing 5.9 per cent of the total. For "Somewhat reliable”, 28 respondents are recorded, accounting for 20.6 percent. The table also groups responses indicating a negative perception under “Unreliable”, which totals 92 respondents, or 67.6 percent. This category is further divided into two subcategories: "Somewhat unreliable”, with 46 respondents and 33.8 percent, and "Very unreliable”, also with 46 respondents and 33.8 percent. Finally, 8 respondents are recorded in the "Neutral / Not sure" category, representing 5.9 per cent. Table 4 Perceived reliability of AI information Reliability Rating Frequency Percentage (%) Very Reliable 8 5.9 Somewhat Reliable 28 20.6 Somewhat Unreliable 46 33.8 Very Unreliable 46 33.8 Not Sure 8 5.9 Willingness to use AI by clinical experience level Table 5 compares willingness scores between clinicians with different levels of experience. The table demonstrates significant variation in adoption willingness by clinical experience. Younger/less experienced clinicians (0–10 years) show significantly higher openness to AI adoption (mean = 4.5) compared to their most senior colleagues with 20 + years’ experience (mean = 4.1), with the difference reaching statistical significance (p < 0.05). Table 5 Willingness to use AI by clinical experience level Experience level Mean willingness score (1–5 scale) Standard Deviation Statistical Significance 0–10 years experience 4.5 0.6 p < 0.05 20 + years experience 4.1 0.8 Significant difference Predictors of willingness to use AI Table 6 presents the results of multiple linear regression analysis identifying factors associated with willingness to use AI. Trust in AI emerges as the dominant predictor (β = 0.48, p < 0.001). The model explains 45% of the variance in willingness to use AI (R² = 0.45). Table 6 Regression analysis on predictors of willingness to use AI Predictor Variable Beta Coefficient (β) Significance (p-value) Interpretation Trust in AI (composite) 0.48 < 0.001 Strongest positive predictor Prior use of AI (Yes/No) 0.21 0.03 Significant positive predictor Threat to autonomy -0.15 0.07 Trend towards negative predictor Familiarity with AI 0.08 0.42 Not significant Age group -0.04 0.68 Not significant Facility type 0.02 0.82 Not significant Model R² = 0.45 (The model explains 45% of variance in willingness to use AI) Qualitative Results A total of 72 key informant interviews were completed as in Table 7 . The perspectives of senior clinicians, department heads, and hospital leadership were gathered through key informant interviews. Their voices painted a clear picture of a profession standing at the cusp of a technological shift, expressing both a powerful vision for AI’s potential and a deeply practical understanding of the risks and requirements for its safe integration into their world. Table 7 Categories of key informants Informant Category Number of Participants Technical Leads 15 AI Developers 9 Data Scientists 16 Program Directors 7 Policymakers 6 Ministry Officials 6 Senior Clinicians 3 Hospital Leadership 4 Department Heads 2 Medical Educators 4 Perceived Opportunities and Roles for AI in Clinical Practice The potential of AI as a cognitive partner is the most notable theme in the qualitative results. It is regarded by clinicians as a potent “second opinion” that can break through the clutter of complicated cases. A future in which AI steps in “at the decision support stage to surface missed differentials and reduce cognitive bias” was outlined by a senior clinician. This vision is of a tool that doesn't make decisions but sharpens the clinician's own, helping to ensure that in the pressure of a busy ward, nothing important is overlooked. This support extends to the very beginning and end of a patient’s journey through a department. Furthermore, in the complex choreography of patient flow, one respondent highlighted AI's role “in disposition to predict acuity and optimise patient flow” , a function that could significantly improve efficiency and resource allocation in overcrowded hospitals. A strong, almost physical need for AI to help with the heavy administrative work that leads to burnout is behind many of the answers. Clinicians express a desire to be liberated from paperwork to reclaim time for their patients. As one respondent put it, they hope for “automated documentation and monitoring” , while another framed the objective as simply “reducing administrative burden” , allowing the clinician to focus on being, first and foremost, a healer. Core Concerns and Perceived Risks The respondents’ primary fear was not that AI will take their jobs, but that it will fundamentally change and potentially damage how they think and practise. The most significant and consistently voiced concern is “ over-reliance and the erosion of clinical skills” , a phenomenon known as automation bias. A senior clinician articulated this danger with striking clarity, stating the risk is “significant where diagnosis is the purpose of interaction.” They painted a stark picture of a future where "clinical instincts can erode, pattern recognition skills atrophy, and the clinician effectively becomes a validator of the machine. The deepest fear embedded in this quote is that when the AI is wrong, as it inevitably will be, the human safety net will have vanished, and "no one catches it." This fear is compounded by concerns about workflow. Clinicians operate in environments that are already at capacity. Any new tool that adds friction will be resisted. One respondent highlighted the critical need for the AI to “fit invisibly into the clinician's existing workflow, surfacing insights at the right moment without demanding attention or extra clicks.” The spectre of a clunky, disruptive system is enough to erode trust before a single patient is seen. A bad experience, they warn, can lead to a complete and permanent loss of faith in technology. Underpinning all of this is a firm belief in the primacy of clinical judgement. There is a strong consensus that AI must remain a support tool and that the final decision must always rest with the human expert. As one respondent put it, establishing a non-negotiable principle: “Clinical judgement should always have the final word.” Frameworks for Safe Integration: Building Trust and Ensuring Safety When asked what it would take for them to trust and safely use AI, clinicians did not speak in abstract terms. They laid out a concrete, three-part framework for safe integration. First and foremost is transparency and explainability . A recommendation from a black box is useless. Trust, they say, comes from “transparency: the AI must show its reasoning, not just its conclusion.” Clinicians want to be able to look under the hood. They need to see the “key drivers” of a recommendation, the specific lab values, vital signs, or historical data points that led to the AI's suggestion, presented in the clear, clinical language they use every day. This leads directly to the second criterion: a proven track record and local validation . Trust is not given; it is earned through consistent, reliable performance in a familiar context. Clinicians need to see that the AI has “been tested in their hospitals” and has “a proven track record... validated in real clinical populations similar to the one being served.” An algorithm trained on data from Boston or London holds little weight in a clinic in Accra or Lagos until it has proven itself there. Finally, they want clear rules and accountability. According to a respondent, “In a field where responsibility is very important, adding a machine to the decision-making process could make it hard to hold people accountable.” When an AI-influenced choice goes wrong, everyone has to know who is responsible. This involves having “downtime protocols” for when the AI fails and a way for clinicians to have “structured reflection” when they don't agree with the AI. Implementation Realities and Barriers Translating these frameworks into reality will require confronting significant barriers on the ground. Hospital leaders and clinicians are brutally honest about the current state of readiness. Most respondents felt that infrastructure deficits are the biggest problem. One hospital leader gave their institution a low score of 4 out of 10, saying that “policy and infrastructure” were the primary shortcomings that made them feel unprepared for AI. According to one of the respondents, “Without reliable electricity, internet, and hardware, the most elegant AI solution is destined to fail.” Furthermore, there is a recognised need for new forms of governance and oversight. Respondents called for the creation of clear governance structures, such as a “technical committee that makes sure that what the AI is generating is conforming with known clinical standards” and an “ethical committee expert on AI”. These bodies would provide the necessary checks and balances. Ultimately, all these concerns circle back to the central challenge of workflow integration . Any new tool must be designed to integrate seamlessly with existing systems, particularly Electronic Medical Records (EMRs). Clinicians are deeply wary of solutions that, however clever, require extra steps in an environment already operating at capacity. The promise of AI, in their view, will only be realised if it can make their work easier, safer, and more effective, without ever getting in the way. DISCUSSION This study examined how West African doctors feel about using AI in healthcare in their practice. The results showed that there are four connected categories: (1) a lack of experience and trust issues, (2) potential and professional concern, (3) trust as the most important factor in their willingness to adopt AI, and (4) significant differences based on their clinical experience. These results suggest that for AI to be successfully integrated into West Africa, frameworks will need to address not only the technical and training needs but also the professional and psychological aspects of adopting AI technology. The Foundational Basis The primary challenge identified is the significant gap in experience with AI, as 73.5% of clinicians had somewhat/moderate familiarity with it. Previous experience influences technology acceptance, with prior AI usage correlating strongly with a willingness to adopt AI. 10,18 Familiarity with AI did not play a key role in the regression analysis, implying that knowing a little bit about AI without having confidence in it does not lead to a tendency to adopt it. Also, predictors show that there are significant, but not all, factors that affect adoption intentions. Qualitative interviews bring this experiential gap to life, revealing what clinicians fear most about engaging with AI without adequate preparation. A senior clinician articulated this danger with striking clarity, describing the risk of automation bias as “significant where diagnosis is the purpose of interaction.” They painted a harrowing picture of a future where “clinical instincts can erode, pattern recognition skills atrophy, and the clinician effectively becomes a validator of the machine.” The deepest fear embedded in this testimony is that when the AI is wrong, as it inevitably will be, the human safety net will have vanished, and “no one catches it.” When planning to integrate AI, this should include training and exposure strategies as well as plans to build trust and not just give information. This has been well noted by another author. 28 One department head in the current study also emphasised that successful implementation requires that clinicians “see that the system works on their patients” and “has been tested in their hospitals” before they can confidently integrate it into their decision-making. The Paradox of Trust Building on this experiential foundation, the most striking finding is what we term the “trust paradox”: clinicians simultaneously express strong belief in AI's potential to improve care (over 83% agreement on benefits) and overwhelming willingness to use AI if proven effective (94.1%), yet over two-thirds (67.6%) find AI information unreliable, and more than three-quarters (77.9%) perceive AI as threatening their clinical autonomy. Building on this experiential foundation, the most striking finding is what we term the "trust paradox": clinicians simultaneously express strong belief in AI's potential to improve care (over 83% agreement on benefits) and overwhelming willingness to use AI if proven effective (94.1%), yet over two-thirds (67.6%) find AI information unreliable, and more than three-quarters (77.9%) perceive AI as threatening their clinical autonomy. This strange pattern of high potential recognition, low trust, and high concern has been seen in other studies of healthcare technology adoption, 7,34 but it seems to be especially strong in this West African sample. The qualitative findings illuminate the professional values underpinning this paradox. Underpinning all clinician responses was a firm, non-negotiable belief in the primacy of clinical judgment. As one respondent unequivocally stated, establishing a foundational principle: "Clinical judgment should always have the final word." This is not resistance to technology; it is a deeply held professional commitment to patient safety and accountability. There are probably a lot of different things that are connected that are causing the lack of trust. First, there is not a lot of access to well-tested AI tools that work in the right context in West Africa, making it very difficult to appreciate how algorithms made in developed countries would be able to work in areas with limited healthcare resources like in developing countries. 12 Second, the fact that many AI systems are "black boxes" could be a problem in situations where doctors already don’t have much help with diagnosing and have to rely on their own clinical judgement. 21 , 22 Third, a general distrust of AI may come from past experiences with faulty equipment and failed digital health projects. 11 Within this paradox, the autonomy concern, which was voiced by 77.9% of respondents, deserves special attention. This finding is in line with what other researchers have found about professional autonomy being a core value in medicine and a common source of conflict when new technologies are introduced. 8 , 9 The high prevalence in this study, on the other hand, could also be due to certain aspects of clinical practice in West Africa, where autonomy is highly valued as a way to deal with limited resources and clinicians have to make a lot of decisions on their own with little help. The qualitative interviews revealed that clinicians are not seeking to exclude AI; they are demanding that AI respect and augment their professional role. One senior clinician eloquently described a future where AI intervenes "at the decision support stage to surface missed differentials and reduce cognitive bias.” The threat to autonomy only showed a trend toward significance in predicting willingness to use AI. The Importance of Trust According to the results, trust is the most important factor in deciding whether or not to adopt AI. This finding fits with theories that say trust is key to accepting technology in situations where there is uncertainty and possible risk. 35 , 36 In healthcare, where decisions can mean life or death, trusting AI systems is not just a nice thing to do; it's also the right thing to do. 6 The importance of trust helps to explain the seeming paradox above. Clinicians are willing to use AI even though they don't trust it right now because they want the tool to “prove itself effective,” which means they need proof that trust is justified. The 94.1% who would use AI if it worked were not blindly trusting; they were open to it, but only after it had been proven to work. This interpretation fits with the strong link between trust and willingness, as well as the fact that past experience makes people more willing, in part by making them trust more. On the other hand, the lack of trust is what keeps potential recognition from automatically leading to current acceptance. Clinicians can think that AI could make care better and also think that the tools they have now are not reliable because they had not seen any systems that they can trust yet. The implication is that trust must be built before or during technical deployment. Trust is not a byproduct of implementation; it is a requirement for safe and effective use. When asked what it would take for them to trust and safely use AI, they laid out a concrete, three-part framework. First and foremost is transparency and explainability. Trust, they explained, comes from “transparency: the AI must show its reasoning, not just its conclusion.” They must see the “key drivers” of a recommendation, the specific lab values, vital signs, or historical data points that generated the AI's suggestion, presented in the clear, clinical language they use daily. As one respondent emphasised, without this transparency, the AI cannot function as a true cognitive partner. Second is a proven track record through local validation, and they noted that trust is not granted; it is earned through consistent, reliable performance in familiar contexts. Clinicians need to see that the AI has “been tested in their hospitals” and has “a proven track record... validated in real clinical populations similar to the one being served.” According to one respondent, an algorithm optimised on data from a developed country like Britain will carry little weight at a health facility in Lagos or Lomé until the value has been demonstrated in those countries. This has been well noted by some authors. 37 , 38 Generational and Contextual Variations Generational disparities in the use of digital health technology indicate that clinicians with 0 to 10 years of experience are more inclined to embrace these advances compared to older colleagues. This is in line with studies that show that younger doctors who are used to technology are more likely to be open to new digital tools. 39 , 40 However, the readiness of senior clinicians to adopt AI indicates that generational issues are not the key determinants of technology acceptance. In the key informant interviews, many older doctors are open to AI if it meets their trust and autonomy needs. One respondent emphasised the critical requirement for AI to “fit invisibly into the clinician's existing workflow, surfacing insights at the right moment without demanding attention or extra clicks.” There were no significant differences between facility types in the regression analysis. This could mean that the way people think about AI is more influenced by their professional training and shared values than by the way they work right now, or that clinicians in different types of facilities have the same worries no matter what their specific situation is. However, the qualitative research reported that primary care clinicians face infrastructure challenges, unreliable electricity, limited internet connectivity, and hardware shortages that their tertiary-based colleagues do not, fundamentally shaping how AI would need to be deployed in their settings. One hospital leader candidly rated their institution's readiness as a stark 4 out of 10, bluntly citing “policy and infrastructure” as the main gaps. Without addressing these foundational barriers, the most elegant AI solution is destined to fail. What this means for safe AI integration frameworks Putting these results together, four related implications from both quantitative and qualitative data come to light for creating frameworks to make sure that AI is safely integrated into clinical practice in West Africa. Building trust must come before or go along with technical deployment. Because trust is the most important factor in predicting behaviour and there is currently a lack of trust, implementation strategies can't assume that trust will grow naturally through exposure. Instead, frameworks need to have clear ways to build trust, such as open communication about the development and validation of AI, pre-deployment studies in West African populations, clear descriptions of performance characteristics and limitations, and ongoing monitoring with results shared openly with clinicians. 24 The fact that 94.1% of clinicians said they would be willing to trust shows that there is an opportunity. Clinicians are willing to trust, but frameworks must earn that trust by being reliable and relevant to the local context with a clear accountability framework, as mentioned by Habli et al . 25 The fact that autonomy concerns only seem to predict willingness when trust is taken into account suggests that high-trust environments may help with autonomy concerns to some extent. However, this does not mean that autonomy should be ignored when designing systems. Clinicians need to trust the governance mechanisms. The preference of 40.4% of individuals for independent bodies to oversee AI indicates a lack of trust in both vendors and government regulation. Establishing unbiased governance structures is crucial for ensuring safety and performance. 26 Suggested structures include oversight committees with diverse stakeholders, transparent reporting systems for adverse events, and mechanisms allowing clinicians to challenge AI recommendations without fear of repercussions. Limitations of the study First, the convenience sampling method employed may not capture the perspectives of clinicians unaffiliated with WACP. Second, what people say about their own experience with and knowledge of AI may not match up with what they actually know. Third, the cross-sectional design only looks at people's opinions at one point in time, so it can't show how attitudes might change when AI is actually used. CONCLUSION The doctors appreciate the potential of AI and are willing to use it under certain conditions. However, losing their independence is a great worry for them; thus, they do not trust it enough. The results show that people are less resistant to AI because they don't believe it can do what it says it can do and more because they are worried about how reliable it is, how relevant it is to their local situation, and how it will affect their professional role. Therefore, to effectively implement AI in healthcare, frameworks must be developed that emphasise transparency, system autonomy, experiential learning, and trusted governance structures for clinicians. For policy and practice, it is recommended that independent groups for AI validation and monitoring at the various Federal/National Ministries of Health that can review AI tools before and after they are used be created across the West African sub-region; explicit guidelines on who is responsible when AI affects a patient be set up; investment in digital infrastructure be made; the West African College of Physicians be involved in the process of developing and testing AI from the start; complete training and continuing assistance to help with implementation of new AI be provided; and AI literacy be taught in medical colleges so that future doctors know what AI can and cannot do and how to utilise it correctly. Declarations Ethics approval and consent to participate: Ethical clearance for the study was obtained from the Ethical Clearance Committee of the University of Nigeria Teaching Hospital. Participation was voluntary, and confidentiality and protection against unauthorised access were maintained using unique anonymous codes, entering data on password-protected Android tablets, transmitting forms and storing them securely. The study was conducted in accordance with the ethical standards of the Declaration of Helsinki. Consent for publication: Not applicable. Competing interests: The authors declare no competing interests. Funding: The study did not receive any direct research funding. Author Contribution Data analysis and interpretation: BSCU, YJC, UUE, CCO, AU. Drafting the article: AO, BH, EME, SMY, KAU. Project supervision: BSCU. Reviewing and editing: All authors (BSCU, YJC, UUE, CCO, AU, AO, BH, EME, SMY, KAU) contributed critically to reviewing and editing multiple versions of the manuscript for important intellectual content. All authors read and approved the final version of the manuscript. Acknowledgement The authors would like to acknowledge all the West African College of Physicians for the platform provided to support this study. Data Availability The data is available from the corresponding author on reasonable request. References Topol EJ. High-performance medicine: the convergence of human and artificial intelligence. Nat Med. 2019;25(1):44–56. 10.1038/s41591-018-0300-7 . PubMed PMID: 30617339. Davenport T, Kalakota R. The potential for artificial intelligence in healthcare. Future Healthc J. 2019;6(2):98. 10.7861/futurehosp.6-2-94 . PubMed PMID: 31363513. Yin J, Ngiam KY, Teo HH. Role of Artificial Intelligence Applications in Real-Life Clinical Practice: Systematic Review. J Med Internet Res. 2021;23(4):e25759. doi:10.2196/25759 PubMed PMID: 33885365. Liu X, Faes L, Kale AU, Wagner SK, Fu DJ, Bruynseels A, et al. A comparison of deep learning performance against health-care professionals in detecting diseases from medical imaging: a systematic review and meta-analysis. Lancet Digit Health. 2019;1(6):e271–97. 10.1016/S2589-7500 . (19)30123-2 PubMed PMID: 33323251. Longoni C, Bonezzi A, Morewedge CK. Resistance to medical artificial intelligence is an attribute in a compensatory decision process: Response to pezzo and beckstead (2020). Judgm Decis Mak. 2020;15(3):446–8. 10.1017/s1930297500007233 . Asan O, Bayrak AE, Choudhury A. Artificial Intelligence and Human Trust in Healthcare: Focus on Clinicians. J Med Internet Res. 2020;22(6):e15154. doi:10.2196/15154 PubMed PMID: 32558657. Blease C, Kaptchuk TJ, Bernstein MH, Mandl KD, Halamka JD, Desroches CM. Artificial intelligence and the future of primary care: exploratory qualitative study of UK general practitioners’ views. J Med Internet Res. 2019;21(3):e12802. doi:10.2196/12802 PubMed PMID: 30892270. Pumplun L, Tauchert C, Heidt M. A New Organizational Chassis for Artificial Intelligence - Exploring Organizational Readiness Factors. In: Proceedings of the 27th European Conference on Information Systems (ECIS), Stockholm, 8–14 June 2019 [Internet]. Stockholm: European Conference on Information Systems; 2019 [cited 2026 Mar 7]. Available from: https://aisel.aisnet.org/ecis2019_rp/106/ Castagno S, Khalifa M. Perceptions of Artificial Intelligence Among Healthcare Staff: A Qualitative Survey Study. Front Artif Intell. 2020;3:578983. 10.3389/frai.2020.578983 . Abdullah R, Fakieh B. Health care employees’ perceptions of the use of artificial intelligence applications: Survey study. J Med Internet Res. 2020;22(5):e17620. doi:10.2196/17620 PubMed PMID: 32406857. Adebayo P, Oluwole OJ, Taiwo FT. COVID-19 and Teleneurology in Sub-Saharan Africa: Leveraging the Current Exigency. Front Public Health [Internet]. 2021 Jan 25 [cited 2026 Mar 7];8(574505):1. Available from: https://doi.org/10.3389/fpubh.2020.574505 Owoyemi A, Owoyemi J, Osiyemi A, Boyd A. Artificial Intelligence for Healthcare in Africa. Front Digit Health. 2020;2:6. 10.3389/fdgth.2020.00006 . PubMed PMID: 34713019. Zhou X, Jomhari N, Farid NDN, Xing Y. Physicians’ perceptions and attitudes on artificial intelligence in hospital management: a qualitative study. BMC Health Serv Res. 2025;26(1):14. 10.1186/s12913-025-13815-1 . PubMed PMID: 41310636. Heinrichs H, Kies A, Nagel SK, Kiessling F. Physicians’ Attitudes Toward Artificial Intelligence in Medicine: Mixed Methods Survey and Interview Study. J Med Internet Res. 2025;27(1):e74187. doi:10.2196/74187 PubMed PMID: 40857713. Alaran MA, Lawal SK, Jiya MH, Egya SA, Ahmed MM, Abdulsalam A, et al. Challenges and opportunities of artificial intelligence in African health space. Digit Health. 2025;11:20552076241305916. doi:10.1177/20552076241305915 PubMed PMID: 39839959. Henzler D, Schmidt S, Koçar A, Herdegen S, Lindinger GL, Maris MT, et al. Healthcare professionals’ perspectives on artificial intelligence in patient care: a systematic review of hindering and facilitating factors on different levels. BMC Health Serv Res. 2025;25(1):633. 10.1186/s12913-025-12664-2 . PubMed PMID: 40312413. Holden RJ, Karsh BT. The Technology Acceptance Model: Its past and its future in health care. J Biomed Inf. 2010;43(1):159–72. 10.1016/j.jbi.2009.07 . .002 PubMed PMID: 19615467. Venkatesh V, Thong JYL, Xu X, Venkatesh V, Thong JYL, Xu X. Unified Theory of Acceptance and Use of Technology: A Synthesis and the Road Ahead. J Assoc Inf Syst. 2016;17(5):1. 10.17705/1jais.00428 . Schoville RR, Titler MG, Computers. Inf Nurs. 2015;33(3):99–107. 10. 1097/CIN.0000000000000130 PubMed PMID: 25799235. Gagnon MP, Ngangue P, Payne-Gagnon J, Desmartis M. m-Health adoption by healthcare professionals: a systematic review. J Am Med Inform Assoc. 2016;23(1):212–20. 10.1093/jamia/ocv052 . PubMed PMID: 26078410. Price WN, Gerke S, Cohen IG. Potential Liability for Physicians Using Artificial Intelligence. JAMA. 2019;322(18):1765–6. 10.1001/jama.2019.15064 . PubMed PMID: 31584609. Challen R, Denny J, Pitt M, Gompels L, Edwards T, Tsaneva-Atanasova K. Artificial intelligence, bias and clinical safety. BMJ Qual Saf. 2019;28(3):231–7. 10.1136/bmjqs-2018-008370 . PubMed PMID: 30636200. Floridi L, Cowls J, Beltrametti M, Chatila R, Chazerand P, Dignum V, et al. AI4People-An Ethical Framework for a Good AI Society: Opportunities, Risks, Principles, and Recommendations. Minds Mach (Dordr). 2018;28(4):689–707. 10.1007/s11023-018-9482-5 . PubMed PMID: 30930541. Reddy S, Allan S, Coghlan S, Cooper P. A governance model for the application of AI in health care. J Am Med Inform Assoc. 2020;27(3):491–7. 10.1093/jamia/ocz192 . PubMed PMID: 31682262. Habli I, Lawton T, Porter Z. Artificial intelligence in health care: accountability and safety. Bull World Health Organ. 2020;98(4):256. 10.2471/BLT . .19.237487 PubMed PMID: 32284648. World Health Organization. Ethics and governance of artificial intelligence for health. Guidance on large multi-modal models [Internet]. Geneva: World Health Organization. 2024 [cited 2026 Mar 7]. 1–98 p. Available from: https://iris.who.int/server/api/core/bitstreams/e9e62c65-6045-481e-bd04-20e206bc5039/content Morley J, Machado CCV, Burr C, Cowls J, Joshi I, Taddeo M, et al. The ethics of AI in health care: A mapping review. Soc Sci Med. 2020;260:113172. 10.1016/j.socscimed.2020.113172 . PubMed PMID: 32702587. Scott IA, Carter SM, Coiera E. Exploring stakeholder attitudes towards AI in clinical practice. BMJ Health Care Inf. 2021;28(1):e100450. 10.1136/bmjhci-2021-100450 . PubMed PMID: 34887331. Abramson JH, WINPEPI (PEPI-for-Windows). Computer Programs for Epidemiologists. Epidemiol Perspect Innovations. 2004;1(1):6. 10.1186/1742-5573-1-6 . PubMed PMID: 15606913. van der Meijden SL, de Hond AAH, Thoral PJ, Steyerberg EW, Kant IMJ, Cinà G, et al. Intensive Care Unit Physicians’ Perspectives on Artificial Intelligence-Based Clinical Decision Support Tools: Preimplementation Survey Study. JMIR Hum Factors. 2023;10(1):e39114. doi:10.2196/39114 PubMed PMID: 36602843. IBM. IBM Support [Internet]. 2024 [cited 2026 Feb 7]. IBM SPSS Statistics 29. Available from: https://www.ibm.com/support/pages/downloading-ibm-spss-statistics-29 Braun V, Clarke V. One size fits all? What counts as quality practice in (reflexive) thematic analysis? Qual Res Psychol. 2021;18(3):328–52. 10.1080/14780887.2020.1769238 . Lumivero. Lumivero Community [Internet]. 2026 [cited 2026 Mar 7]. About NVivo (NVivo 14 Windows). Available from: https://community.lumivero.com/s/article/TRC-About-NVivo-NVivo-14-Windows?language=en_US Walter Z, Lopez MS. Physician acceptance of information technologies: Role of perceived threat to professional autonomy. Decis Support Syst. 2008;46(1):206–15. 10.1016/j.dss.2008.06.004 . Mayer RC, Davis JH, Schoorman DF. An Integrative Model of Organizational Trust. Academy of Management Review [Internet]. 1995 [cited 2026 Mar 7];20(3):709–34. Available from: https://www.makinggood.ac.nz/media/1270/mayeretal_1995_organizationaltrust.pdf Mcknight DH, Carter M, Thatcher JB, Clay PF. Trust in a specific technology: An investigation of its components and measures. ACM Trans Manag Inf Syst. 2011;2(2):12–32. 10.1145/1985347.1985353 . Zuhair V, Babar A, Ali R, Oduoye MO, Noor Z, Chris K, et al. Exploring the Impact of Artificial Intelligence on Global Health and Enhancing Healthcare in Developing Nations. J Prim Care Community Health. 2024;15:21501319241245850. doi:10.1177/21501319241245847 PubMed PMID: 38605668. Victor A. Artificial intelligence in global health: An unfair future for health in Sub-Saharan. Africa? Health Affairs Scholar. 2025;3(2):qxaf023. 10.1093/haschl/qxaf . 023 PubMed PMID: 39949826. de Grood C, Raissi A, Kwon Y, Santana MJ. Adoption of e-health technology by physicians: a scoping review. J Multidiscip Healthc. 2016;9:344. 10.2147. /JMDH.S103881 PubMed PMID: 27536128. Spil TAM, Schuring RW. E-health systems diffusion and use: The innovation, the user and the USE IT model. E-Health Systems Diffusion and Use: The Innovation, the User and the USE IT Model [Internet]. Hershey: Idea Group Publishing; 2005 [cited 2026 Mar 7]. 1–342 p. Available from: https://doi.org/10.4018/978-1-59140-423-1 doi:10.4018/978-1-59140-423-1. Additional Declarations No competing interests reported. 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-9066307","acceptedTermsAndConditions":true,"allowDirectSubmit":true,"archivedVersions":[],"articleType":"Research Article","associatedPublications":[],"authors":[{"id":602959826,"identity":"1f8fbcb8-81f4-4aad-bc6e-ac650bb8d19f","order_by":0,"name":"Benjamin S. Chudi Uzochukwu","email":"","orcid":"","institution":"University of Nigeria","correspondingAuthor":false,"prefix":"","firstName":"Benjamin","middleName":"S. Chudi","lastName":"Uzochukwu","suffix":""},{"id":602959827,"identity":"4ae56b50-4074-4159-ab9a-c5eb9883d507","order_by":1,"name":"Yakubu Joel Cherima","email":"","orcid":"","institution":"Center for Artificial Intelligence Accountability in Health","correspondingAuthor":false,"prefix":"","firstName":"Yakubu","middleName":"Joel","lastName":"Cherima","suffix":""},{"id":602959828,"identity":"c7df6f63-5e8a-49c7-a83d-7813d0b419de","order_by":2,"name":"Ugo Uwadiako Enebeli","email":"data:image/png;base64,iVBORw0KGgoAAAANSUhEUgAAAZAAAAAyAQMAAABI0h/eAAAABlBMVEX///8AAABVwtN+AAAACXBIWXMAAA7EAAAOxAGVKw4bAAAA50lEQVRIiWNgGAWjYDCCAzCGBAPjgYQKBgYDIrQwNkC1MBxIOEOyFsY2IrTwHW9+/uDjHps8funmAwcezjssb87efIDhR8U2nFokzxwzbJzxLK1Ycs6xhAOJ2w4b7uw5lsDYc+Y2Ti0GN3IYm3kOHE7ccCPHAKSFEcRgZmwjqOV/4v4b+R8OJM45bE+slgOJGyRyGA4kNkCsw6sF5JeZMw4kJ864c8zgQMKx9OQNZ44lHMTnF2CIPfjw4YBdYv/s5ocPf9RY22443nzwwY8K3FrQQTOYPEC0eiCoI0XxKBgFo2AUjBAAAIbyaz53HbGMAAAAAElFTkSuQmCC","orcid":"","institution":"Federal University of Technology Owerri","correspondingAuthor":true,"prefix":"","firstName":"Ugo","middleName":"Uwadiako","lastName":"Enebeli","suffix":""},{"id":602959829,"identity":"f005da61-2c95-428a-85a7-f28177f1b474","order_by":3,"name":"Chinyere Cecelia Okeke","email":"","orcid":"","institution":"University of Nigeria","correspondingAuthor":false,"prefix":"","firstName":"Chinyere","middleName":"Cecelia","lastName":"Okeke","suffix":""},{"id":602959830,"identity":"3cc4ab67-9d3b-4fd3-a06c-8033787ebda3","order_by":4,"name":"Adaora Chinelo Uzochukwu","email":"","orcid":"","institution":"University of Nigeria","correspondingAuthor":false,"prefix":"","firstName":"Adaora","middleName":"Chinelo","lastName":"Uzochukwu","suffix":""},{"id":602959831,"identity":"5b5a6f56-6f7e-445e-be2e-7f95b90fb499","order_by":5,"name":"Amobi Omoha","email":"","orcid":"","institution":"David Umahi Federal University of Health Sciences","correspondingAuthor":false,"prefix":"","firstName":"Amobi","middleName":"","lastName":"Omoha","suffix":""},{"id":602959832,"identity":"1498bf31-fe70-4963-844e-756f16ff87cb","order_by":6,"name":"Blessing Hassan","email":"","orcid":"","institution":"Center for Artificial Intelligence Accountability in Health","correspondingAuthor":false,"prefix":"","firstName":"Blessing","middleName":"","lastName":"Hassan","suffix":""},{"id":602959836,"identity":"ca8672f9-3b04-47f4-a1c6-52bc25978019","order_by":7,"name":"Emmanuel Majiyebo Eronu","email":"","orcid":"","institution":"University of Abuja","correspondingAuthor":false,"prefix":"","firstName":"Emmanuel","middleName":"Majiyebo","lastName":"Eronu","suffix":""},{"id":602959837,"identity":"66f836f4-7ef6-49a0-829e-96b0ca5ed90b","order_by":8,"name":"Shehu Mohammed Yusuf","email":"","orcid":"","institution":"Ahmadu Bello University","correspondingAuthor":false,"prefix":"","firstName":"Shehu","middleName":"Mohammed","lastName":"Yusuf","suffix":""},{"id":602959838,"identity":"7a8e0cc3-90e2-4b9c-85e2-31767c67c479","order_by":9,"name":"Kennedy Anenechukwu Uzochukwu","email":"","orcid":"","institution":"Center for Artificial Intelligence Accountability in Health","correspondingAuthor":false,"prefix":"","firstName":"Kennedy","middleName":"Anenechukwu","lastName":"Uzochukwu","suffix":""}],"badges":[],"createdAt":"2026-03-08 19:53:15","currentVersionCode":1,"declarations":"","doi":"10.21203/rs.3.rs-9066307/v1","doiUrl":"https://doi.org/10.21203/rs.3.rs-9066307/v1","draftVersion":[],"editorialEvents":[],"editorialNote":"","failedWorkflow":false,"files":[{"id":105916604,"identity":"4f32dda3-aee5-44d1-a6a4-a116d2bb653b","added_by":"auto","created_at":"2026-04-01 11:27:51","extension":"pdf","order_by":0,"title":"","display":"","copyAsset":false,"role":"manuscript-pdf","size":1293672,"visible":true,"origin":"","legend":"","description":"","filename":"manuscript.pdf","url":"https://assets-eu.researchsquare.com/files/rs-9066307/v1/117648d7-3267-4349-bc39-53d473f9587a.pdf"}],"financialInterests":"No competing interests reported.","formattedTitle":"Clinician Perceptions of Artificial Intelligence in Healthcare and Frameworks for Ensuring Safe Integration into Clinical Practice of the West African College of Physicians: A Multi-Country Mixed-Methods Study","fulltext":[{"header":"BACKGROUND","content":"\u003cp\u003eThe use of artificial intelligence (AI) in clinical practice is a landmark in healthcare delivery. It improves the accuracy of diagnoses, the planning of treatments, the making of clinical decisions, and the overall performance of the health system.\u003csup\u003e\u003cspan citationid=\"CR1\" class=\"CitationRef\"\u003e1\u003c/span\u003e,\u003cspan citationid=\"CR2\" class=\"CitationRef\"\u003e2\u003c/span\u003e\u003c/sup\u003e AI is being used increasingly in many areas of healthcare around the world, including medical imaging, pathology, predictive analytics, and clinical decision support. Recent studies indicate that these applications can significantly enhance efficiency and patient outcomes.\u003csup\u003e\u003cspan citationid=\"CR3\" class=\"CitationRef\"\u003e3\u003c/span\u003e,\u003cspan citationid=\"CR4\" class=\"CitationRef\"\u003e4\u003c/span\u003e\u003c/sup\u003e\u003c/p\u003e \u003cp\u003eHowever, as noted by some authors, the successful integration and utilisation of AI in healthcare will largely depend on doctors' perspectives, trust in AI, and its acceptance.\u003csup\u003e\u003cspan citationid=\"CR5\" class=\"CitationRef\"\u003e5\u003c/span\u003e,\u003cspan citationid=\"CR6\" class=\"CitationRef\"\u003e6\u003c/span\u003e\u003c/sup\u003e This is because health technologies are mostly used and controlled by doctors. Therefore, their willingness to use AI tools, trust in AI-generated recommendations, and understanding of the limitations of AI will all have a direct effect on whether these technologies are used, how they are integrated into clinical workflows, and whether they will ultimately lead to better patient care.\u003csup\u003e\u003cspan citationid=\"CR7\" class=\"CitationRef\"\u003e7\u003c/span\u003e,\u003cspan citationid=\"CR8\" class=\"CitationRef\"\u003e8\u003c/span\u003e\u003c/sup\u003e Resistance from clinicians in developed countries is as a result of concerns about trust, accountability and autonomy issues, and these are largely responsible for a good number of barriers to AI integration.\u003csup\u003e\u003cspan citationid=\"CR9\" class=\"CitationRef\"\u003e9\u003c/span\u003e,\u003cspan citationid=\"CR10\" class=\"CitationRef\"\u003e10\u003c/span\u003e\u003c/sup\u003e\u003c/p\u003e \u003cp\u003eAdoption of AI in West African healthcare systems has its own difficulties that could influence how doctors see AI compared to systems with more resources. Health systems in the region encounter significant resource challenges, including insufficient healthcare personnel, inadequate diagnostic equipment, unreliable electrical supply, and unstable internet connections.\u003csup\u003e\u003cspan citationid=\"CR11\" class=\"CitationRef\"\u003e11\u003c/span\u003e,\u003cspan citationid=\"CR12\" class=\"CitationRef\"\u003e12\u003c/span\u003e\u003c/sup\u003e These contextual aspects create both imperatives for AI adoption, as a potential means for task-shifting and quality enhancement, and considerable implementation constraints that may influence physicians\u0026rsquo; perceptions and interactions with AI technology.\u003csup\u003e\u003cspan citationid=\"CR13\" class=\"CitationRef\"\u003e13\u003c/span\u003e,\u003cspan citationid=\"CR14\" class=\"CitationRef\"\u003e14\u003c/span\u003e\u003c/sup\u003e\u003c/p\u003e \u003cp\u003eWhile there is a growing body of international literature on AI adoption, empirical research regarding physician perspectives of AI in West African contexts is still limited.\u003csup\u003e\u003cspan citationid=\"CR15\" class=\"CitationRef\"\u003e15\u003c/span\u003e,\u003cspan citationid=\"CR16\" class=\"CitationRef\"\u003e16\u003c/span\u003e\u003c/sup\u003e It is important to understand these views. For example, health system characteristics affect how people adopt technology in different ways, which means that results from high-income nations may not be directly relevant to the West African context.\u003csup\u003e\u003cspan citationid=\"CR17\" class=\"CitationRef\"\u003e17\u003c/span\u003e,\u003cspan citationid=\"CR18\" class=\"CitationRef\"\u003e18\u003c/span\u003e\u003c/sup\u003e It is also important to know what West African doctors know and want and what information they require to create effective implementation strategies, training programmes, and governance frameworks.\u003csup\u003e\u003cspan citationid=\"CR19\" class=\"CitationRef\"\u003e19\u003c/span\u003e,\u003cspan citationid=\"CR20\" class=\"CitationRef\"\u003e20\u003c/span\u003e\u003c/sup\u003e They also need to be able to examine new AI tools and discover possible faults, especially those that will compromise patient safety. The level of knowledge and perspectives of AI systems determines these skills.\u003csup\u003e\u003cspan citationid=\"CR21\" class=\"CitationRef\"\u003e21\u003c/span\u003e,\u003cspan citationid=\"CR22\" class=\"CitationRef\"\u003e22\u003c/span\u003e\u003c/sup\u003e\u003c/p\u003e \u003cp\u003eThe concept of \u0026ldquo;AI assurance\u0026rdquo; - that is, the systematic processes and structures required to ensure that AI systems are safe, dependable, accountable, ethically sound, and compliant with therapeutic standards - has garnered significant global attention.\u003csup\u003e\u003cspan citationid=\"CR23\" class=\"CitationRef\"\u003e23\u003c/span\u003e,\u003cspan citationid=\"CR24\" class=\"CitationRef\"\u003e24\u003c/span\u003e\u003c/sup\u003e The criteria for effective AI assurance in resource-constrained environments remain insufficiently defined, and there is limited direction on how West African health systems should evaluate, regulate, and monitor AI tools both before and after implementation.\u003csup\u003e\u003cspan citationid=\"CR25\" class=\"CitationRef\"\u003e25\u003c/span\u003e,\u003cspan citationid=\"CR26\" class=\"CitationRef\"\u003e26\u003c/span\u003e\u003c/sup\u003e Clinicians are crucial partners in the development of these frameworks, as their own experiences and concerns must inform the establishment of governance systems that are both robust and practically effective.\u003csup\u003e\u003cspan citationid=\"CR27\" class=\"CitationRef\"\u003e27\u003c/span\u003e,\u003cspan citationid=\"CR28\" class=\"CitationRef\"\u003e28\u003c/span\u003e\u003c/sup\u003e\u003c/p\u003e \u003cp\u003eThis study filled in these gaps in the evidence by looking at what doctors in West Africa think about AI in healthcare settings, what factors affect its use, and what doctors recommend to make sure it is safely integrated. This aims to create strategies for using AI that are appropriate for the situation and will improve, not hurt, patient safety and quality of care.\u003c/p\u003e \u003cp\u003e \u003cb\u003eResearch Questions\u003c/b\u003e:\u003c/p\u003e \u003cp\u003e \u003col\u003e \u003cspan\u003e \u003cli\u003e \u003cp\u003eWhat do West African doctors think about using AI in healthcare?\u003c/p\u003e \u003c/li\u003e \u003c/span\u003e \u003cspan\u003e \u003cli\u003e \u003cp\u003eWhat factors make clinicians more likely to accept and use AI tools in West African healthcare settings?\u003c/p\u003e \u003c/li\u003e \u003c/span\u003e \u003cspan\u003e \u003cli\u003e \u003cp\u003eWhat do clinicians think are the frameworks that need to be in place to make sure that AI is used safely, reliably, and with accountability in clinical practice?\u003c/p\u003e \u003c/li\u003e \u003c/span\u003e \u003cspan\u003e \u003cli\u003e \u003cp\u003eHow do clinician perceptions of AI differ across facility types (tertiary, secondary, and primary), medical specialties, and levels of experience?\u003c/p\u003e \u003c/li\u003e \u003c/span\u003e \u003c/ol\u003e \u003c/p\u003e"},{"header":"MATERIALS AND METHODS","content":"\u003cdiv id=\"Sec3\" class=\"Section2\"\u003e \u003ch2\u003eStudy design and study area\u003c/h2\u003e \u003cp\u003eThis cross-sectional mixed-methods study included a structured, self-administered questionnaire to assess the attitudes of West African professionals regarding artificial intelligence in healthcare and a qualitative exploration through key informant interviews (KII). The study was conducted in the West African subregion. Data was gathered from February 22 to February 28, 2026.\u003c/p\u003e \u003c/div\u003e\n\u003ch3\u003eStudy Population and Sampling\u003c/h3\u003e\n\u003cp\u003eThe target group comprised physicians affiliated with the West African College of Physicians (WACP), the principal institution for postgraduate medical education in the sub-region. The membership of this college covers both the Anglophone, Francophone, and Lusophone countries of the subregion. Most of the WACP fellows and members work in Nigeria; however, they can work in any West African country. This is because the college was started there and has a lot of students. A convenience sampling method was used with the college's current communication channels to find willing participants. Any physician listed in the WACP contact database during the study period was eligible for inclusion. The study sought to capture a comprehensive spectrum of physician perspectives from the sub-region, imposing no restrictions on specialty, years of experience, or practice environment.\u003c/p\u003e\n\u003ch3\u003eSample size determination\u003c/h3\u003e\n\u003cp\u003eA mixed-methods approach was employed. For the quantitative phase, a cross-sectional survey was administered to physicians of the West African College of Physicians. The sample size was calculated on the WinPepi sample size calculator, version 11.65,\u003csup\u003e29\u003c/sup\u003e to estimate the primary outcome (e.g., 'positive perception of AI utility') with a 95% confidence level, a 5% acceptable difference, and a 10% non-response rate. Using an assumed proportion of 92% (0.92) positive perception of AI utility from a previous study,\u003csup\u003e30\u003c/sup\u003e the initial sample size was 126. This was rounded up to 136 for more precision.\u003c/p\u003e\n\u003ch3\u003eData Collection\u003c/h3\u003e\n\u003cp\u003eFor the survey, the questionnaire had four parts: demographic and professional information (age, gender, country of practice, practice setting (primary/secondary/tertiary facility), specialty, years of clinical experience, and previous exposure to AI (familiarity and experience)); perceptions of AI in clinical practice (perceived benefits, concerns about autonomy, trust in AI-generated information, and perceived reliability). The 5-point Likert scale was used for scoring, with 1 being \u0026ldquo;strongly disagree\u0026rdquo; and 5 being \u0026ldquo;strongly agree\u0026rdquo;. The 5-point Likert scale was also used to rate how willing people were to use AI (how ready they were to use AI tools in different clinical situations) and how they wanted AI to be monitored and regulated (the bodies they wanted to do this, with options like independent bodies, government agencies, professional colleges, and vendors).\u003c/p\u003e \u003cp\u003eThe questionnaire was pre-tested with 12 WACP-affiliated physicians (not included in the final sample) from diverse specialties and experience levels to assess clarity, completion time, and comprehension. Minor wording adjustments were made based on feedback. Internal consistency of multi-item Likert scales (perceptions of benefits, concerns, and willingness) was evaluated post-data collection using Cronbach's alpha, with values ranging from 0.78 to 0.89, indicating acceptable to good reliability. Content validity was ensured through iterative expert review by the research team (including senior clinicians and AI specialists) during questionnaire development, drawing on established domains from prior AI healthcare perception studies. To prevent multiple entries from the same participant, the ODK link was designed for single-use submission, with built-in validation rules and skip logic to minimize data-entry errors.\u003c/p\u003e \u003cp\u003eThe Open Data Kit (ODK) platform was used for data collection. This is an open-source program that enables data collection, both online and offline, in regions where resources are scarce. The ODK was chosen because it works well in West Africa, where internet availability is quite diverse from country to country, and it has been used successfully in health research in the past. The ODK-based questionnaire with a link was sent through the WACP communication methods, like email and WhatsApp. The questionnaire took 15 to 20 minutes to complete. No personally identifiable information was collected, ensuring the anonymity of participants. To gain informed consent electronically, the first page of questions had to be filled out and agreed upon before moving on.\u003c/p\u003e \u003cp\u003eFor the qualitative study, a key informant interview (KII) approach was employed that enabled the exploration of complex phenomena, including perceptions, concerns, and recommendations regarding AI governance, clinical integration, and medical education, that cannot be adequately captured through quantitative methods alone. There was a total of 72 key informants\u0026rsquo; interviews. The respondents were key stakeholders in the West Africa health space whose professional roles and expertise positioned them to produce authoritative perspectives.\u003c/p\u003e \u003cp\u003e Each interview lasted less than 60 minutes using a semi-structured interview guide. All interview responses were captured through the online data collection platform, which recorded verbatim text entries for each open-ended question. This approach eliminated transcription errors associated with audio recording and transcription while ensuring complete capture of participant responses. For quality assurance purposes, all entries were reviewed for completeness and clarity within 24 hours of submission.\u003c/p\u003e \u003cdiv id=\"Sec7\" class=\"Section2\"\u003e \u003ch2\u003eData Analysis\u003c/h2\u003e \u003cp\u003eThe survey data from ODK was downloaded as Excel and exported into SPSS version 29 (IBM Corp., Armonk, NY, USA).\u003csup\u003e\u003cspan citationid=\"CR31\" class=\"CitationRef\"\u003e31\u003c/span\u003e\u003c/sup\u003e The descriptive statistics (frequencies, percentages, means, and standard deviations) were computed for each variable. Cronbach\u0026rsquo;s alpha was used to determine if multi-item scales were consistent with each other. Values greater than 0.70 were considered acceptable.\u003c/p\u003e \u003cp\u003eInitially, the relationships between demographic and experiential characteristics were determined, as well as the desire to employ AI using t-tests. The ANOVA was used for inferential analysis. Factors exhibiting significant correlations (p\u0026thinsp;\u0026lt;\u0026thinsp;0.10) were included in a multivariate linear regression model to forecast the propensity to employ AI. The regression model included previous AI experience (binary), familiarity with AI (Likert scale), trust in AI information (Likert scale), perceived danger to autonomy (Likert scale), years of experience (categorical), and facility type (categorical). The cutoff for statistical significance was set at p\u0026thinsp;\u0026lt;\u0026thinsp;0.05 (two-tailed). Missing data was minimal across survey items (\u0026lt;\u0026thinsp;1% per variable, primarily in optional demographic fields) and was handled using listwise deletion for all inferential analyses (assumed missing completely at random given the low rate and random pattern observed). The sample\u0026rsquo;s demographic distribution (e.g., predominant Nigerian representation) aligns broadly with WACP membership patterns.\u003c/p\u003e \u003cp\u003eThe thematic analysis of the qualitative data was conducted using the framework approach described by Braun and Clarke.\u003csup\u003e\u003cspan citationid=\"CR32\" class=\"CitationRef\"\u003e32\u003c/span\u003e\u003c/sup\u003e This method was selected for its flexibility and suitability for identifying, analysing, and reporting patterns within qualitative data, particularly when working with policy-relevant research questions. Analysis proceeded through several stages, namely, data familiarisation, initial coding, framework development and application of framework. The final coding framework was systematically applied to all 72 transcripts using qualitative data analysis software NVivo version 14.\u003csup\u003e33\u003c/sup\u003e\u003c/p\u003e \u003c/div\u003e"},{"header":"RESULTS","content":"\u003cdiv id=\"Sec9\" class=\"Section2\"\u003e \u003ch2\u003eQuantitative Results\u003c/h2\u003e \u003cdiv id=\"Sec10\" class=\"Section3\"\u003e \u003ch2\u003eRespondents\u0026rsquo; Characteristics\u003c/h2\u003e \u003cp\u003eThe survey link was distributed through the West African College of Physicians (WACP) email and WhatsApp channels to all eligible physicians between 22 and 28 February 2026. A total of 137 physicians accessed the link, resulting in 136 complete responses included in the analysis after the exclusion of one participant who declined consent. Table\u0026nbsp;\u003cspan refid=\"Tab1\" class=\"InternalRef\"\u003e1\u003c/span\u003e presents the demographic characteristics of the sample. The majority of respondents were from Nigeria (84.6%), with smaller representations from Ghana (5.9%), Liberia (3.6%), Togo (2.9%), Niger (1.5%), and C\u0026ocirc;te d'Ivoire (1.5%). Gender distribution was balanced (51.5% female, 48.5% male). The largest age group was 35\u0026ndash;44 years (30.1%), followed by 45\u0026ndash;54 years (26.5%). One-third of respondents (33.0%) had over 20 years of clinical experience.\u003c/p\u003e \u003cp\u003e \u003cdiv class=\"gridtable\"\u003e\u003ctable float=\"Yes\" id=\"Tab1\" border=\"1\"\u003e \u003ccaption language=\"En\"\u003e \u003cdiv class=\"CaptionNumber\"\u003eTable 1\u003c/div\u003e \u003cdiv class=\"CaptionContent\"\u003e \u003cp\u003eDemographic characteristics of survey respondents\u003c/p\u003e \u003c/div\u003e \u003c/caption\u003e \u003ccolgroup cols=\"4\"\u003e \u003cdiv align=\"left\" class=\"colspec\" colname=\"c1\" colnum=\"1\"\u003e\u003c/div\u003e \u003cdiv align=\"left\" class=\"colspec\" colname=\"c2\" colnum=\"2\"\u003e\u003c/div\u003e \u003cdiv align=\"char\" char=\".\" class=\"colspec\" colname=\"c3\" colnum=\"3\"\u003e\u003c/div\u003e \u003cdiv align=\"char\" char=\".\" class=\"colspec\" colname=\"c4\" colnum=\"4\"\u003e\u003c/div\u003e \u003cthead\u003e \u003ctr\u003e \u003cth align=\"left\" colname=\"c1\"\u003e \u003cp\u003eDemographic Variables\u003c/p\u003e \u003c/th\u003e \u003cth align=\"left\" colname=\"c2\"\u003e \u003cp\u003eCategory\u003c/p\u003e \u003c/th\u003e \u003cth align=\"left\" colname=\"c3\"\u003e \u003cp\u003eFrequency\u003c/p\u003e \u003c/th\u003e \u003cth align=\"left\" colname=\"c4\"\u003e \u003cp\u003ePercent\u003c/p\u003e \u003c/th\u003e \u003c/tr\u003e \u003c/thead\u003e \u003ctbody\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eCountry\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003eNigeria\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e \u003cp\u003e115\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e \u003cp\u003e84.6\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003eGhana\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e \u003cp\u003e8\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e \u003cp\u003e5.9\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003eLiberia\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e \u003cp\u003e5\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e \u003cp\u003e3.6\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003eTogo\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e \u003cp\u003e4\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e \u003cp\u003e2.9\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003eNiger\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e \u003cp\u003e2\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e \u003cp\u003e1.5\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003eCote d\u0026rsquo;Ivoire\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e \u003cp\u003e2\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e \u003cp\u003e1.5\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eGender\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003eFemale\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e \u003cp\u003e70\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e \u003cp\u003e51.5\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003eMale\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e \u003cp\u003e65\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e \u003cp\u003e47.8\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003ePrefer not to say\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e \u003cp\u003e1\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e \u003cp\u003e0.7\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eAge group\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e25\u0026ndash;34\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e \u003cp\u003e26\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e \u003cp\u003e19.1\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e35\u0026ndash;44\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e \u003cp\u003e41\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e \u003cp\u003e30.1\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e45\u0026ndash;54\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e \u003cp\u003e36\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e \u003cp\u003e26.5\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e55\u0026ndash;64\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e \u003cp\u003e21\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e \u003cp\u003e15.4\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e65+\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e \u003cp\u003e11\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e \u003cp\u003e8.1\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003ePrefer not to say\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e \u003cp\u003e1\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e \u003cp\u003e0.7\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eClinical Experience (Years)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e0\u0026ndash;5\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e \u003cp\u003e15\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e \u003cp\u003e11.0\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e6\u0026ndash;10\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e \u003cp\u003e27\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e \u003cp\u003e19.9\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e11\u0026ndash;15\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e \u003cp\u003e27\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e \u003cp\u003e19.9\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e16\u0026ndash;20\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e \u003cp\u003e22\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e \u003cp\u003e16.2\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e20+\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e \u003cp\u003e45\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e \u003cp\u003e33.0\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003c/tbody\u003e \u003c/colgroup\u003e \u003c/table\u003e\u003c/div\u003e \u003c/p\u003e \u003c/div\u003e \u003c/div\u003e \u003cdiv id=\"Sec11\" class=\"Section2\"\u003e \u003ch2\u003eClinician familiarity and prior AI experience\u003c/h2\u003e \u003cp\u003eTable\u0026nbsp;\u003cspan refid=\"Tab2\" class=\"InternalRef\"\u003e2\u003c/span\u003e presents findings on West African clinicians' self-reported familiarity with AI and prior experience using AI clinical tools. While a combined 73.5% reported being \u0026ldquo;somewhat/moderately familiar\u0026rdquo; with AI, only 19.1% considered themselves \u0026ldquo;highly familiar\u0026rdquo; or \u0026ldquo;experts\u0026rdquo;. Critically, half of respondents (52.2%) had never used an AI clinical tool, and another 14.7% were unsure whether they had.\u003c/p\u003e \u003cp\u003e \u003cdiv class=\"gridtable\"\u003e\u003ctable float=\"Yes\" id=\"Tab2\" border=\"1\"\u003e \u003ccaption language=\"En\"\u003e \u003cdiv class=\"CaptionNumber\"\u003eTable 2\u003c/div\u003e \u003cdiv class=\"CaptionContent\"\u003e \u003cp\u003eClinician familiarity and experience with AI\u003c/p\u003e \u003c/div\u003e \u003c/caption\u003e \u003ccolgroup cols=\"4\"\u003e \u003cdiv align=\"left\" class=\"colspec\" colname=\"c1\" colnum=\"1\"\u003e\u003c/div\u003e \u003cdiv align=\"left\" class=\"colspec\" colname=\"c2\" colnum=\"2\"\u003e\u003c/div\u003e \u003cdiv align=\"char\" char=\".\" class=\"colspec\" colname=\"c3\" colnum=\"3\"\u003e\u003c/div\u003e \u003cdiv align=\"char\" char=\".\" class=\"colspec\" colname=\"c4\" colnum=\"4\"\u003e\u003c/div\u003e \u003cthead\u003e \u003ctr\u003e \u003cth align=\"left\" colname=\"c1\"\u003e \u003cp\u003eVariable\u003c/p\u003e \u003c/th\u003e \u003cth align=\"left\" colname=\"c2\"\u003e \u003cp\u003eCategory\u003c/p\u003e \u003c/th\u003e \u003cth align=\"left\" colname=\"c3\"\u003e \u003cp\u003eFrequency\u003c/p\u003e \u003c/th\u003e \u003cth align=\"left\" colname=\"c4\"\u003e \u003cp\u003ePercent (%)\u003c/p\u003e \u003c/th\u003e \u003c/tr\u003e \u003c/thead\u003e \u003ctbody\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003e\u003cb\u003eFamiliarity with AI\u003c/b\u003e\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003eNot familiar\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e \u003cp\u003e10\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e \u003cp\u003e7.4\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003eSomewhat familiar\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e \u003cp\u003e34\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e \u003cp\u003e25.0\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003eModerately familiar\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e \u003cp\u003e66\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e \u003cp\u003e48.5\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003eHighly familiar\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e \u003cp\u003e18\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e \u003cp\u003e13.2\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003eExpert\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e \u003cp\u003e8\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e \u003cp\u003e5.9\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003e\u003cb\u003ePrior Use of AI Tool\u003c/b\u003e\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003eYes\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e \u003cp\u003e45\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e \u003cp\u003e33.1\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003eNo\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e \u003cp\u003e71\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e \u003cp\u003e52.2\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003eNot Sure\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e \u003cp\u003e20\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e \u003cp\u003e14.7\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003c/tbody\u003e \u003c/colgroup\u003e \u003c/table\u003e\u003c/div\u003e \u003c/p\u003e \u003c/div\u003e \u003cdiv id=\"Sec12\" class=\"Section2\"\u003e \u003ch2\u003eCore perceptions of AI in clinical practice\u003c/h2\u003e \u003cp\u003eTable\u0026nbsp;\u003cspan refid=\"Tab3\" class=\"InternalRef\"\u003e3\u003c/span\u003e shows that 77.9 percent indicated that they agree or strongly agree that AI threatens their clinical autonomy. A smaller portion, 10.3 percent, remained neutral on this issue, while 11.8 percent expressed disagreement. As high as 94.1% agreed that they would use AI if proven effective. Only 4.4 percent were neutral on this question, and a mere 1.5 percent disagreed. Regarding the potential of AI to reduce medical errors, 83.1 percent of those surveyed agreed or strongly agreed that AI could serve this function. 11.0 percent took a neutral stance, and 5.9 percent disagreed. Finally, when considering whether AI could improve diagnostic accuracy, 85.3 percent of respondents expressed agreement. 8.8 percent were neutral, and again, 5.9 percent disagreed with the statement.\u003c/p\u003e \u003cp\u003e \u003cdiv class=\"gridtable\"\u003e\u003ctable float=\"Yes\" id=\"Tab3\" border=\"1\"\u003e \u003ccaption language=\"En\"\u003e \u003cdiv class=\"CaptionNumber\"\u003eTable 3\u003c/div\u003e \u003cdiv class=\"CaptionContent\"\u003e \u003cp\u003eCore perceptions of AI in clinical practice\u003c/p\u003e \u003c/div\u003e \u003c/caption\u003e \u003ccolgroup cols=\"4\"\u003e \u003cdiv align=\"left\" class=\"colspec\" colname=\"c1\" colnum=\"1\"\u003e\u003c/div\u003e \u003cdiv align=\"char\" char=\".\" class=\"colspec\" colname=\"c2\" colnum=\"2\"\u003e\u003c/div\u003e \u003cdiv align=\"char\" char=\".\" class=\"colspec\" colname=\"c3\" colnum=\"3\"\u003e\u003c/div\u003e \u003cdiv align=\"char\" char=\".\" class=\"colspec\" colname=\"c4\" colnum=\"4\"\u003e\u003c/div\u003e \u003cthead\u003e \u003ctr\u003e \u003cth align=\"left\" colname=\"c1\"\u003e \u003cp\u003ePerception Statement\u003c/p\u003e \u003c/th\u003e \u003cth align=\"left\" colname=\"c2\"\u003e \u003cp\u003eAgree/Strongly Agree (%)\u003c/p\u003e \u003c/th\u003e \u003cth align=\"left\" colname=\"c3\"\u003e \u003cp\u003eNeutral (%)\u003c/p\u003e \u003c/th\u003e \u003cth align=\"left\" colname=\"c4\"\u003e \u003cp\u003eDisagree/Strongly Disagree (%)\u003c/p\u003e \u003c/th\u003e \u003c/tr\u003e \u003c/thead\u003e \u003ctbody\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eAI threatens my clinical autonomy\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c2\"\u003e \u003cp\u003e77.9\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e \u003cp\u003e10.3\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e \u003cp\u003e11.8\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eI would use AI if proven effective\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c2\"\u003e \u003cp\u003e94.1\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e \u003cp\u003e4.4\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e \u003cp\u003e1.5\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eAI can help reduce clinical errors\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c2\"\u003e \u003cp\u003e83.1\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e \u003cp\u003e11.0\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e \u003cp\u003e5.9\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eAI can improve diagnostic accuracy\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c2\"\u003e \u003cp\u003e85.3\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e \u003cp\u003e8.8\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e \u003cp\u003e5.9\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003c/tbody\u003e \u003c/colgroup\u003e \u003c/table\u003e\u003c/div\u003e \u003c/p\u003e \u003c/div\u003e \u003cdiv id=\"Sec13\" class=\"Section2\"\u003e \u003ch2\u003ePerceived reliability of AI information\u003c/h2\u003e \u003cp\u003eTable\u0026nbsp;\u003cspan refid=\"Tab4\" class=\"InternalRef\"\u003e4\u003c/span\u003e presents data on the perceived reliability of AI information, showing both the frequency and percentage of responses across four categories. For the category \u0026ldquo;Very reliable\u0026rdquo;, 8 respondents reported this perception, representing 5.9 per cent of the total. For \"Somewhat reliable\u0026rdquo;, 28 respondents are recorded, accounting for 20.6 percent. The table also groups responses indicating a negative perception under \u0026ldquo;Unreliable\u0026rdquo;, which totals 92 respondents, or 67.6 percent. This category is further divided into two subcategories: \"Somewhat unreliable\u0026rdquo;, with 46 respondents and 33.8 percent, and \"Very unreliable\u0026rdquo;, also with 46 respondents and 33.8 percent. Finally, 8 respondents are recorded in the \"Neutral / Not sure\" category, representing 5.9 per cent.\u003c/p\u003e \u003cp\u003e \u003cdiv class=\"gridtable\"\u003e\u003ctable float=\"Yes\" id=\"Tab4\" border=\"1\"\u003e \u003ccaption language=\"En\"\u003e \u003cdiv class=\"CaptionNumber\"\u003eTable 4\u003c/div\u003e \u003cdiv class=\"CaptionContent\"\u003e \u003cp\u003ePerceived reliability of AI information\u003c/p\u003e \u003c/div\u003e \u003c/caption\u003e \u003ccolgroup cols=\"3\"\u003e \u003cdiv align=\"left\" class=\"colspec\" colname=\"c1\" colnum=\"1\"\u003e\u003c/div\u003e \u003cdiv align=\"char\" char=\".\" class=\"colspec\" colname=\"c2\" colnum=\"2\"\u003e\u003c/div\u003e \u003cdiv align=\"char\" char=\".\" class=\"colspec\" colname=\"c3\" colnum=\"3\"\u003e\u003c/div\u003e \u003cthead\u003e \u003ctr\u003e \u003cth align=\"left\" colname=\"c1\"\u003e \u003cp\u003eReliability Rating\u003c/p\u003e \u003c/th\u003e \u003cth align=\"left\" colname=\"c2\"\u003e \u003cp\u003eFrequency\u003c/p\u003e \u003c/th\u003e \u003cth align=\"left\" colname=\"c3\"\u003e \u003cp\u003ePercentage (%)\u003c/p\u003e \u003c/th\u003e \u003c/tr\u003e \u003c/thead\u003e \u003ctbody\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eVery Reliable\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c2\"\u003e \u003cp\u003e8\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e \u003cp\u003e5.9\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eSomewhat Reliable\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c2\"\u003e \u003cp\u003e28\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e \u003cp\u003e20.6\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eSomewhat Unreliable\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c2\"\u003e \u003cp\u003e46\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e \u003cp\u003e33.8\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eVery Unreliable\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c2\"\u003e \u003cp\u003e46\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e \u003cp\u003e33.8\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eNot Sure\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c2\"\u003e \u003cp\u003e8\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e \u003cp\u003e5.9\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003c/tbody\u003e \u003c/colgroup\u003e \u003c/table\u003e\u003c/div\u003e \u003c/p\u003e \u003c/div\u003e \u003cdiv id=\"Sec14\" class=\"Section2\"\u003e \u003ch2\u003eWillingness to use AI by clinical experience level\u003c/h2\u003e \u003cp\u003eTable\u0026nbsp;\u003cspan refid=\"Tab5\" class=\"InternalRef\"\u003e5\u003c/span\u003e compares willingness scores between clinicians with different levels of experience. The table demonstrates significant variation in adoption willingness by clinical experience. Younger/less experienced clinicians (0\u0026ndash;10 years) show significantly higher openness to AI adoption (mean\u0026thinsp;=\u0026thinsp;4.5) compared to their most senior colleagues with 20\u0026thinsp;+\u0026thinsp;years\u0026rsquo; experience (mean\u0026thinsp;=\u0026thinsp;4.1), with the difference reaching statistical significance (p\u0026thinsp;\u0026lt;\u0026thinsp;0.05).\u003c/p\u003e \u003cp\u003e \u003cdiv class=\"gridtable\"\u003e\u003ctable float=\"Yes\" id=\"Tab5\" border=\"1\"\u003e \u003ccaption language=\"En\"\u003e \u003cdiv class=\"CaptionNumber\"\u003eTable 5\u003c/div\u003e \u003cdiv class=\"CaptionContent\"\u003e \u003cp\u003eWillingness to use AI by clinical experience level\u003c/p\u003e \u003c/div\u003e \u003c/caption\u003e \u003ccolgroup cols=\"4\"\u003e \u003cdiv align=\"left\" class=\"colspec\" colname=\"c1\" colnum=\"1\"\u003e\u003c/div\u003e \u003cdiv align=\"char\" char=\".\" class=\"colspec\" colname=\"c2\" colnum=\"2\"\u003e\u003c/div\u003e \u003cdiv align=\"char\" char=\".\" class=\"colspec\" colname=\"c3\" colnum=\"3\"\u003e\u003c/div\u003e \u003cdiv align=\"left\" class=\"colspec\" colname=\"c4\" colnum=\"4\"\u003e\u003c/div\u003e \u003cthead\u003e \u003ctr\u003e \u003cth align=\"left\" colname=\"c1\"\u003e \u003cp\u003eExperience level\u003c/p\u003e \u003c/th\u003e \u003cth align=\"left\" colname=\"c2\"\u003e \u003cp\u003eMean willingness score (1\u0026ndash;5 scale)\u003c/p\u003e \u003c/th\u003e \u003cth align=\"left\" colname=\"c3\"\u003e \u003cp\u003eStandard Deviation\u003c/p\u003e \u003c/th\u003e \u003cth align=\"left\" colname=\"c4\"\u003e \u003cp\u003eStatistical Significance\u003c/p\u003e \u003c/th\u003e \u003c/tr\u003e \u003c/thead\u003e \u003ctbody\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003e0\u0026ndash;10 years experience\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c2\"\u003e \u003cp\u003e4.5\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e \u003cp\u003e0.6\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003ep\u0026thinsp;\u0026lt;\u0026thinsp;0.05\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003e20\u0026thinsp;+\u0026thinsp;years experience\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c2\"\u003e \u003cp\u003e4.1\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e \u003cp\u003e0.8\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003eSignificant difference\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003c/tbody\u003e \u003c/colgroup\u003e \u003c/table\u003e\u003c/div\u003e \u003c/p\u003e \u003c/div\u003e \u003cdiv id=\"Sec15\" class=\"Section2\"\u003e \u003ch2\u003ePredictors of willingness to use AI\u003c/h2\u003e \u003cp\u003eTable\u0026nbsp;\u003cspan refid=\"Tab6\" class=\"InternalRef\"\u003e6\u003c/span\u003e presents the results of multiple linear regression analysis identifying factors associated with willingness to use AI. Trust in AI emerges as the dominant predictor (β\u0026thinsp;=\u0026thinsp;0.48, p\u0026thinsp;\u0026lt;\u0026thinsp;0.001). The model explains 45% of the variance in willingness to use AI (R\u0026sup2; = 0.45).\u003c/p\u003e \u003cp\u003e \u003cdiv class=\"gridtable\"\u003e\u003ctable float=\"Yes\" id=\"Tab6\" border=\"1\"\u003e \u003ccaption language=\"En\"\u003e \u003cdiv class=\"CaptionNumber\"\u003eTable 6\u003c/div\u003e \u003cdiv class=\"CaptionContent\"\u003e \u003cp\u003eRegression analysis on predictors of willingness to use AI\u003c/p\u003e \u003c/div\u003e \u003c/caption\u003e \u003ccolgroup cols=\"4\"\u003e \u003cdiv align=\"left\" class=\"colspec\" colname=\"c1\" colnum=\"1\"\u003e\u003c/div\u003e \u003cdiv align=\"char\" char=\".\" class=\"colspec\" colname=\"c2\" colnum=\"2\"\u003e\u003c/div\u003e \u003cdiv align=\"char\" char=\".\" class=\"colspec\" colname=\"c3\" colnum=\"3\"\u003e\u003c/div\u003e \u003cdiv align=\"left\" class=\"colspec\" colname=\"c4\" colnum=\"4\"\u003e\u003c/div\u003e \u003cthead\u003e \u003ctr\u003e \u003cth align=\"left\" colname=\"c1\"\u003e \u003cp\u003ePredictor Variable\u003c/p\u003e \u003c/th\u003e \u003cth align=\"left\" colname=\"c2\"\u003e \u003cp\u003eBeta Coefficient (β)\u003c/p\u003e \u003c/th\u003e \u003cth align=\"left\" colname=\"c3\"\u003e \u003cp\u003eSignificance (p-value)\u003c/p\u003e \u003c/th\u003e \u003cth align=\"left\" colname=\"c4\"\u003e \u003cp\u003eInterpretation\u003c/p\u003e \u003c/th\u003e \u003c/tr\u003e \u003c/thead\u003e \u003ctbody\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eTrust in AI (composite)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c2\"\u003e \u003cp\u003e0.48\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e \u003cp\u003e\u0026lt;\u0026thinsp;0.001\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003eStrongest positive predictor\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003ePrior use of AI (Yes/No)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c2\"\u003e \u003cp\u003e0.21\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e \u003cp\u003e0.03\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003eSignificant positive predictor\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eThreat to autonomy\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c2\"\u003e \u003cp\u003e-0.15\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e \u003cp\u003e0.07\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003eTrend towards negative predictor\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eFamiliarity with AI\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c2\"\u003e \u003cp\u003e0.08\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e \u003cp\u003e0.42\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003eNot significant\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eAge group\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c2\"\u003e \u003cp\u003e-0.04\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e \u003cp\u003e0.68\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003eNot significant\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eFacility type\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c2\"\u003e \u003cp\u003e0.02\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e \u003cp\u003e0.82\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003eNot significant\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003c/tbody\u003e \u003c/colgroup\u003e \u003c/table\u003e\u003c/div\u003e \u003c/p\u003e \u003cp\u003eModel R\u0026sup2; = 0.45 (The model explains 45% of variance in willingness to use AI)\u003c/p\u003e \u003c/div\u003e \u003cdiv id=\"Sec16\" class=\"Section2\"\u003e \u003ch2\u003eQualitative Results\u003c/h2\u003e \u003cp\u003eA total of 72 key informant interviews were completed as in Table\u0026nbsp;\u003cspan refid=\"Tab7\" class=\"InternalRef\"\u003e7\u003c/span\u003e. The perspectives of senior clinicians, department heads, and hospital leadership were gathered through key informant interviews. Their voices painted a clear picture of a profession standing at the cusp of a technological shift, expressing both a powerful vision for AI\u0026rsquo;s potential and a deeply practical understanding of the risks and requirements for its safe integration into their world.\u003c/p\u003e \u003cp\u003e \u003cdiv class=\"gridtable\"\u003e\u003ctable float=\"Yes\" id=\"Tab7\" border=\"1\"\u003e \u003ccaption language=\"En\"\u003e \u003cdiv class=\"CaptionNumber\"\u003eTable 7\u003c/div\u003e \u003cdiv class=\"CaptionContent\"\u003e \u003cp\u003eCategories of key informants\u003c/p\u003e \u003c/div\u003e \u003c/caption\u003e \u003ccolgroup cols=\"2\"\u003e \u003cdiv align=\"left\" class=\"colspec\" colname=\"c1\" colnum=\"1\"\u003e\u003c/div\u003e \u003cdiv align=\"char\" char=\".\" class=\"colspec\" colname=\"c2\" colnum=\"2\"\u003e\u003c/div\u003e \u003cthead\u003e \u003ctr\u003e \u003cth align=\"left\" colname=\"c1\"\u003e \u003cp\u003eInformant Category\u003c/p\u003e \u003c/th\u003e \u003cth align=\"left\" colname=\"c2\"\u003e \u003cp\u003eNumber of Participants\u003c/p\u003e \u003c/th\u003e \u003c/tr\u003e \u003c/thead\u003e \u003ctbody\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eTechnical Leads\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c2\"\u003e \u003cp\u003e15\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eAI Developers\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c2\"\u003e \u003cp\u003e9\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eData Scientists\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c2\"\u003e \u003cp\u003e16\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eProgram Directors\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c2\"\u003e \u003cp\u003e7\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003ePolicymakers\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c2\"\u003e \u003cp\u003e6\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eMinistry Officials\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c2\"\u003e \u003cp\u003e6\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eSenior Clinicians\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c2\"\u003e \u003cp\u003e3\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eHospital Leadership\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c2\"\u003e \u003cp\u003e4\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eDepartment Heads\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c2\"\u003e \u003cp\u003e2\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eMedical Educators\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c2\"\u003e \u003cp\u003e4\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003c/tbody\u003e \u003c/colgroup\u003e \u003c/table\u003e\u003c/div\u003e \u003c/p\u003e \u003c/div\u003e \u003cdiv id=\"Sec17\" class=\"Section2\"\u003e \u003ch2\u003ePerceived Opportunities and Roles for AI in Clinical Practice\u003c/h2\u003e \u003cp\u003eThe potential of AI as a cognitive partner is the most notable theme in the qualitative results. It is regarded by clinicians as a potent \u003cem\u003e\u0026ldquo;second opinion\u0026rdquo;\u003c/em\u003e that can break through the clutter of complicated cases. A future in which AI steps \u003cem\u003ein \u0026ldquo;at the decision support stage to surface missed differentials and reduce cognitive bias\u0026rdquo;\u003c/em\u003e was outlined by a senior clinician. This vision is of a tool that doesn't make decisions but sharpens the clinician's own, helping to ensure that in the pressure of a busy ward, nothing important is overlooked. This support extends to the very beginning and end of a patient\u0026rsquo;s journey through a department. Furthermore, in the complex choreography of patient flow, one respondent highlighted AI's role \u003cem\u003e\u0026ldquo;in disposition to predict acuity and optimise patient flow\u0026rdquo;\u003c/em\u003e, a function that could significantly improve efficiency and resource allocation in overcrowded hospitals.\u003c/p\u003e \u003cp\u003eA strong, almost physical need for AI to help with the heavy administrative work that leads to burnout is behind many of the answers. Clinicians express a desire to be liberated from paperwork to reclaim time for their patients. As one respondent put it, they hope for \u003cem\u003e\u0026ldquo;automated documentation and monitoring\u0026rdquo;\u003c/em\u003e, while another framed the objective as simply \u003cem\u003e\u0026ldquo;reducing administrative burden\u0026rdquo;\u003c/em\u003e, allowing the clinician to focus on being, first and foremost, a healer.\u003c/p\u003e \u003c/div\u003e \u003cdiv id=\"Sec18\" class=\"Section2\"\u003e \u003ch2\u003eCore Concerns and Perceived Risks\u003c/h2\u003e \u003cp\u003eThe respondents\u0026rsquo; primary fear was not that AI will take their jobs, but that it will fundamentally change and potentially damage how they think and practise. The most significant and consistently voiced concern is \u0026ldquo;\u003cem\u003eover-reliance and the erosion of clinical skills\u0026rdquo;\u003c/em\u003e, a phenomenon known as automation bias. A senior clinician articulated this danger with striking clarity, stating the risk is \u003cem\u003e\u0026ldquo;significant where diagnosis is the purpose of interaction.\u0026rdquo;\u003c/em\u003e They painted a stark picture of a future where \u003cem\u003e\"clinical instincts can erode, pattern recognition skills atrophy, and the clinician effectively becomes a validator of the machine.\u003c/em\u003e The deepest fear embedded in this quote is that when the AI is wrong, as it inevitably will be, the human safety net will have vanished, and \u003cem\u003e\"no one catches it.\"\u003c/em\u003e\u003c/p\u003e \u003cp\u003eThis fear is compounded by concerns about workflow. Clinicians operate in environments that are already at capacity. Any new tool that adds friction will be resisted. One respondent highlighted the critical need for the AI to \u003cem\u003e\u0026ldquo;fit invisibly into the clinician's existing workflow, surfacing insights at the right moment without demanding attention or extra clicks.\u0026rdquo;\u003c/em\u003e The spectre of a clunky, disruptive system is enough to erode trust before a single patient is seen. A bad experience, they warn, can lead to a complete and permanent loss of faith in technology.\u003c/p\u003e \u003cp\u003eUnderpinning all of this is a firm belief in the primacy of clinical judgement. There is a strong consensus that AI must remain a support tool and that the final decision must always rest with the human expert. As one respondent put it, establishing a non-negotiable principle: \u003cem\u003e\u0026ldquo;Clinical judgement should always have the final word.\u0026rdquo;\u003c/em\u003e\u003c/p\u003e \u003c/div\u003e \u003cdiv id=\"Sec19\" class=\"Section2\"\u003e \u003ch2\u003eFrameworks for Safe Integration: Building Trust and Ensuring Safety\u003c/h2\u003e \u003cp\u003eWhen asked what it would take for them to trust and safely use AI, clinicians did not speak in abstract terms. They laid out a concrete, three-part framework for safe integration.\u003c/p\u003e \u003cp\u003eFirst and foremost is \u003cem\u003etransparency and explainability\u003c/em\u003e. A recommendation from a black box is useless. Trust, they say, comes from \u003cem\u003e\u0026ldquo;transparency: the AI must show its reasoning, not just its conclusion.\u0026rdquo;\u003c/em\u003e Clinicians want to be able to look under the hood. They need to see the \u003cem\u003e\u0026ldquo;key drivers\u0026rdquo;\u003c/em\u003e of a recommendation, the specific lab values, vital signs, or historical data points that led to the AI's suggestion, presented in the clear, clinical language they use every day.\u003c/p\u003e \u003cp\u003eThis leads directly to the second criterion: a \u003cem\u003eproven track record and local validation\u003c/em\u003e. Trust is not given; it is earned through consistent, reliable performance in a familiar context. Clinicians need to see that the AI has \u003cem\u003e\u0026ldquo;been tested in their hospitals\u0026rdquo;\u003c/em\u003e and has \u003cem\u003e\u0026ldquo;a proven track record... validated in real clinical populations similar to the one being served.\u0026rdquo; An algorithm trained on data from Boston or London holds little weight in a clinic in Accra or Lagos until it has proven itself there.\u003c/em\u003e\u003c/p\u003e \u003cp\u003eFinally, they want clear rules and accountability. According to a respondent, \u0026ldquo;In a field where responsibility is very important, adding a machine to the decision-making process could make it hard to hold people accountable.\u0026rdquo; When an AI-influenced choice goes wrong, everyone has to know who is responsible. This involves having \u0026ldquo;downtime protocols\u0026rdquo; for when the AI fails and a way for clinicians to have \u0026ldquo;structured reflection\u0026rdquo; when they don't agree with the AI.\u003c/p\u003e \u003c/div\u003e \u003cdiv id=\"Sec20\" class=\"Section2\"\u003e \u003ch2\u003eImplementation Realities and Barriers\u003c/h2\u003e \u003cp\u003eTranslating these frameworks into reality will require confronting significant barriers on the ground. Hospital leaders and clinicians are brutally honest about the current state of readiness. Most respondents felt that infrastructure deficits are the biggest problem. One hospital leader gave their institution a low score of 4 out of 10, saying that \u003cem\u003e\u0026ldquo;policy and infrastructure\u0026rdquo;\u003c/em\u003e were the primary shortcomings that made them feel unprepared for AI. According to one of the respondents, \u0026ldquo;Without \u003cem\u003ereliable electricity, internet, and hardware, the most elegant AI solution is destined to fail.\u0026rdquo;\u003c/em\u003e\u003c/p\u003e \u003cp\u003eFurthermore, there is a recognised need for new forms of governance and oversight. Respondents called for the creation of clear governance structures, such as a \u003cem\u003e\u0026ldquo;technical committee that makes sure that what the AI is generating is conforming with known clinical standards\u0026rdquo; and an \u0026ldquo;ethical committee expert on AI\u0026rdquo;.\u003c/em\u003e These bodies would provide the necessary checks and balances.\u003c/p\u003e \u003cp\u003eUltimately, all these concerns circle back to the central challenge of \u003cem\u003eworkflow integration\u003c/em\u003e. Any new tool must be designed to integrate seamlessly with existing systems, particularly Electronic Medical Records (EMRs). Clinicians are deeply wary of solutions that, however clever, require extra steps in an environment already operating at capacity. The promise of AI, in their view, will only be realised if it can make their work easier, safer, and more effective, without ever getting in the way.\u003c/p\u003e \u003c/div\u003e"},{"header":"DISCUSSION","content":"\u003cp\u003eThis study examined how West African doctors feel about using AI in healthcare in their practice. The results showed that there are four connected categories: (1) a lack of experience and trust issues, (2) potential and professional concern, (3) trust as the most important factor in their willingness to adopt AI, and (4) significant differences based on their clinical experience. These results suggest that for AI to be successfully integrated into West Africa, frameworks will need to address not only the technical and training needs but also the professional and psychological aspects of adopting AI technology.\u003c/p\u003e \u003cdiv id=\"Sec22\" class=\"Section2\"\u003e \u003ch2\u003eThe Foundational Basis\u003c/h2\u003e \u003cp\u003eThe primary challenge identified is the significant gap in experience with AI, as 73.5% of clinicians had somewhat/moderate familiarity with it. Previous experience influences technology acceptance, with prior AI usage correlating strongly with a willingness to adopt AI.\u003csup\u003e10,18\u003c/sup\u003e Familiarity with AI did not play a key role in the regression analysis, implying that knowing a little bit about AI without having confidence in it does not lead to a tendency to adopt it. Also, predictors show that there are significant, but not all, factors that affect adoption intentions.\u003c/p\u003e \u003cp\u003eQualitative interviews bring this experiential gap to life, revealing what clinicians fear most about engaging with AI without adequate preparation. A senior clinician articulated this danger with striking clarity, describing the risk of automation bias as \u003cem\u003e\u0026ldquo;significant where diagnosis is the purpose of interaction.\u0026rdquo;\u003c/em\u003e They painted a harrowing picture of a future where \u003cem\u003e\u0026ldquo;clinical instincts can erode, pattern recognition skills atrophy, and the clinician effectively becomes a validator of the machine.\u0026rdquo;\u003c/em\u003e The deepest fear embedded in this testimony is that when the AI is wrong, as it inevitably will be, the human safety net will have vanished, and \u003cem\u003e\u0026ldquo;no one catches it.\u0026rdquo;\u003c/em\u003e\u003c/p\u003e \u003cp\u003eWhen planning to integrate AI, this should include training and exposure strategies as well as plans to build trust and not just give information. This has been well noted by another author.\u003csup\u003e\u003cspan citationid=\"CR28\" class=\"CitationRef\"\u003e28\u003c/span\u003e\u003c/sup\u003e One department head in the current study also emphasised that successful implementation requires that clinicians \u003cem\u003e\u0026ldquo;see that the system works on their patients\u0026rdquo;\u003c/em\u003e and \u003cem\u003e\u0026ldquo;has been tested in their hospitals\u0026rdquo;\u003c/em\u003e before they can confidently integrate it into their decision-making.\u003c/p\u003e \u003cdiv id=\"Sec23\" class=\"Section3\"\u003e \u003ch2\u003eThe Paradox of Trust\u003c/h2\u003e \u003cp\u003eBuilding on this experiential foundation, the most striking finding is what we term the \u0026ldquo;trust paradox\u0026rdquo;: clinicians simultaneously express strong belief in AI's potential to improve care (over 83% agreement on benefits) and overwhelming willingness to use AI if proven effective (94.1%), yet over two-thirds (67.6%) find AI information unreliable, and more than three-quarters (77.9%) perceive AI as threatening their clinical autonomy. Building on this experiential foundation, the most striking finding is what we term the \"trust paradox\": clinicians simultaneously express strong belief in AI's potential to improve care (over 83% agreement on benefits) and overwhelming willingness to use AI if proven effective (94.1%), yet over two-thirds (67.6%) find AI information unreliable, and more than three-quarters (77.9%) perceive AI as threatening their clinical autonomy. This strange pattern of high potential recognition, low trust, and high concern has been seen in other studies of healthcare technology adoption,\u003csup\u003e7,34\u003c/sup\u003e but it seems to be especially strong in this West African sample.\u003c/p\u003e \u003cp\u003eThe qualitative findings illuminate the professional values underpinning this paradox. Underpinning all clinician responses was a firm, non-negotiable belief in the primacy of clinical judgment. As one respondent unequivocally stated, establishing a foundational principle: \u003cem\u003e\"Clinical judgment should always have the final word.\"\u003c/em\u003e This is not resistance to technology; it is a deeply held professional commitment to patient safety and accountability.\u003c/p\u003e \u003cp\u003eThere are probably a lot of different things that are connected that are causing the lack of trust. First, there is not a lot of access to well-tested AI tools that work in the right context in West Africa, making it very difficult to appreciate how algorithms made in developed countries would be able to work in areas with limited healthcare resources like in developing countries.\u003csup\u003e\u003cspan citationid=\"CR12\" class=\"CitationRef\"\u003e12\u003c/span\u003e\u003c/sup\u003e Second, the fact that many AI systems are \"black boxes\" could be a problem in situations where doctors already don\u0026rsquo;t have much help with diagnosing and have to rely on their own clinical judgement.\u003csup\u003e\u003cspan citationid=\"CR21\" class=\"CitationRef\"\u003e21\u003c/span\u003e,\u003cspan citationid=\"CR22\" class=\"CitationRef\"\u003e22\u003c/span\u003e\u003c/sup\u003e Third, a general distrust of AI may come from past experiences with faulty equipment and failed digital health projects.\u003csup\u003e\u003cspan citationid=\"CR11\" class=\"CitationRef\"\u003e11\u003c/span\u003e\u003c/sup\u003e\u003c/p\u003e \u003cp\u003eWithin this paradox, the autonomy concern, which was voiced by 77.9% of respondents, deserves special attention. This finding is in line with what other researchers have found about professional autonomy being a core value in medicine and a common source of conflict when new technologies are introduced.\u003csup\u003e\u003cspan citationid=\"CR8\" class=\"CitationRef\"\u003e8\u003c/span\u003e,\u003cspan citationid=\"CR9\" class=\"CitationRef\"\u003e9\u003c/span\u003e\u003c/sup\u003e The high prevalence in this study, on the other hand, could also be due to certain aspects of clinical practice in West Africa, where autonomy is highly valued as a way to deal with limited resources and clinicians have to make a lot of decisions on their own with little help. The qualitative interviews revealed that clinicians are not seeking to exclude AI; they are demanding that AI respect and augment their professional role. One senior clinician eloquently described a future where AI intervenes \u003cem\u003e\"at the decision support stage to surface missed differentials and reduce cognitive bias.\u0026rdquo;\u003c/em\u003e The threat to autonomy only showed a trend toward significance in predicting willingness to use AI.\u003c/p\u003e \u003c/div\u003e \u003c/div\u003e \u003cdiv id=\"Sec24\" class=\"Section2\"\u003e \u003ch2\u003eThe Importance of Trust\u003c/h2\u003e \u003cp\u003eAccording to the results, trust is the most important factor in deciding whether or not to adopt AI. This finding fits with theories that say trust is key to accepting technology in situations where there is uncertainty and possible risk.\u003csup\u003e\u003cspan citationid=\"CR35\" class=\"CitationRef\"\u003e35\u003c/span\u003e,\u003cspan citationid=\"CR36\" class=\"CitationRef\"\u003e36\u003c/span\u003e\u003c/sup\u003e In healthcare, where decisions can mean life or death, trusting AI systems is not just a nice thing to do; it's also the right thing to do.\u003csup\u003e6\u003c/sup\u003e\u003c/p\u003e \u003cp\u003eThe importance of trust helps to explain the seeming paradox above. Clinicians are willing to use AI even though they don't trust it right now because they want the tool to \u0026ldquo;prove itself effective,\u0026rdquo; which means they need proof that trust is justified. The 94.1% who would use AI if it worked were not blindly trusting; they were open to it, but only after it had been proven to work. This interpretation fits with the strong link between trust and willingness, as well as the fact that past experience makes people more willing, in part by making them trust more. On the other hand, the lack of trust is what keeps potential recognition from automatically leading to current acceptance. Clinicians can think that AI could make care better and also think that the tools they have now are not reliable because they had not seen any systems that they can trust yet. The implication is that trust must be built before or during technical deployment. Trust is not a byproduct of implementation; it is a requirement for safe and effective use.\u003c/p\u003e \u003cp\u003eWhen asked what it would take for them to trust and safely use AI, they laid out a concrete, three-part framework. First and foremost is transparency and explainability. Trust, they explained, comes from \u003cem\u003e\u0026ldquo;transparency: the AI must show its reasoning, not just its conclusion.\u0026rdquo;\u003c/em\u003e They must see the \u003cem\u003e\u0026ldquo;key drivers\u0026rdquo;\u003c/em\u003e of a recommendation, the specific lab values, vital signs, or historical data points that generated the AI's suggestion, presented in the clear, clinical language they use daily. As one respondent emphasised, without this transparency, the AI cannot function as a true cognitive partner.\u003c/p\u003e \u003cp\u003e Second is a proven track record through local validation, and they noted that trust is not granted; it is earned through consistent, reliable performance in familiar contexts. Clinicians need to see that the AI has \u003cem\u003e\u0026ldquo;been tested in their hospitals\u0026rdquo;\u003c/em\u003e and has \u003cem\u003e\u0026ldquo;a proven track record... validated in real clinical populations similar to the one being served.\u0026rdquo;\u003c/em\u003e According to one respondent, an algorithm optimised on data from a developed country like Britain will carry little weight at a health facility in Lagos or Lom\u0026eacute; until the value has been demonstrated in those countries. This has been well noted by some authors.\u003csup\u003e\u003cspan citationid=\"CR37\" class=\"CitationRef\"\u003e37\u003c/span\u003e,\u003cspan citationid=\"CR38\" class=\"CitationRef\"\u003e38\u003c/span\u003e\u003c/sup\u003e\u003c/p\u003e \u003cdiv id=\"Sec25\" class=\"Section3\"\u003e \u003ch2\u003eGenerational and Contextual Variations\u003c/h2\u003e \u003cp\u003eGenerational disparities in the use of digital health technology indicate that clinicians with 0 to 10 years of experience are more inclined to embrace these advances compared to older colleagues. This is in line with studies that show that younger doctors who are used to technology are more likely to be open to new digital tools.\u003csup\u003e\u003cspan citationid=\"CR39\" class=\"CitationRef\"\u003e39\u003c/span\u003e,\u003cspan citationid=\"CR40\" class=\"CitationRef\"\u003e40\u003c/span\u003e\u003c/sup\u003e However, the readiness of senior clinicians to adopt AI indicates that generational issues are not the key determinants of technology acceptance.\u003c/p\u003e \u003cp\u003eIn the key informant interviews, many older doctors are open to AI if it meets their trust and autonomy needs. One respondent emphasised the critical requirement for AI to \u003cem\u003e\u0026ldquo;fit invisibly into the clinician's existing workflow, surfacing insights at the right moment without demanding attention or extra clicks.\u0026rdquo;\u003c/em\u003e\u003c/p\u003e \u003cp\u003eThere were no significant differences between facility types in the regression analysis. This could mean that the way people think about AI is more influenced by their professional training and shared values than by the way they work right now, or that clinicians in different types of facilities have the same worries no matter what their specific situation is. However, the qualitative research reported that primary care clinicians face infrastructure challenges, unreliable electricity, limited internet connectivity, and hardware shortages that their tertiary-based colleagues do not, fundamentally shaping how AI would need to be deployed in their settings. One hospital leader candidly rated their institution's readiness as a stark 4 out of 10, bluntly citing \u003cem\u003e\u0026ldquo;policy and infrastructure\u0026rdquo;\u003c/em\u003e as the main gaps. Without addressing these foundational barriers, the most elegant AI solution is destined to fail.\u003c/p\u003e \u003c/div\u003e \u003cdiv id=\"Sec26\" class=\"Section3\"\u003e \u003ch2\u003eWhat this means for safe AI integration frameworks\u003c/h2\u003e \u003cp\u003ePutting these results together, four related implications from both quantitative and qualitative data come to light for creating frameworks to make sure that AI is safely integrated into clinical practice in West Africa. Building trust must come before or go along with technical deployment. Because trust is the most important factor in predicting behaviour and there is currently a lack of trust, implementation strategies can't assume that trust will grow naturally through exposure. Instead, frameworks need to have clear ways to build trust, such as open communication about the development and validation of AI, pre-deployment studies in West African populations, clear descriptions of performance characteristics and limitations, and ongoing monitoring with results shared openly with clinicians.\u003csup\u003e\u003cspan citationid=\"CR24\" class=\"CitationRef\"\u003e24\u003c/span\u003e\u003c/sup\u003e The fact that 94.1% of clinicians said they would be willing to trust shows that there is an opportunity. Clinicians are willing to trust, but frameworks must earn that trust by being reliable and relevant to the local context with a clear accountability framework, as mentioned by Habli \u003cem\u003eet al\u003c/em\u003e.\u003csup\u003e25\u003c/sup\u003e\u003c/p\u003e \u003cp\u003eThe fact that autonomy concerns only seem to predict willingness when trust is taken into account suggests that high-trust environments may help with autonomy concerns to some extent. However, this does not mean that autonomy should be ignored when designing systems.\u003c/p\u003e \u003cp\u003eClinicians need to trust the governance mechanisms. The preference of 40.4% of individuals for independent bodies to oversee AI indicates a lack of trust in both vendors and government regulation. Establishing unbiased governance structures is crucial for ensuring safety and performance.\u003csup\u003e\u003cspan citationid=\"CR26\" class=\"CitationRef\"\u003e26\u003c/span\u003e\u003c/sup\u003e Suggested structures include oversight committees with diverse stakeholders, transparent reporting systems for adverse events, and mechanisms allowing clinicians to challenge AI recommendations without fear of repercussions.\u003c/p\u003e \u003c/div\u003e \u003cdiv id=\"Sec27\" class=\"Section3\"\u003e \u003ch2\u003eLimitations of the study\u003c/h2\u003e \u003cp\u003eFirst, the convenience sampling method employed may not capture the perspectives of clinicians unaffiliated with WACP. Second, what people say about their own experience with and knowledge of AI may not match up with what they actually know. Third, the cross-sectional design only looks at people's opinions at one point in time, so it can't show how attitudes might change when AI is actually used.\u003c/p\u003e \u003c/div\u003e \u003c/div\u003e"},{"header":"CONCLUSION","content":"\u003cp\u003eThe doctors appreciate the potential of AI and are willing to use it under certain conditions. However, losing their independence is a great worry for them; thus, they do not trust it enough. The results show that people are less resistant to AI because they don't believe it can do what it says it can do and more because they are worried about how reliable it is, how relevant it is to their local situation, and how it will affect their professional role. Therefore, to effectively implement AI in healthcare, frameworks must be developed that emphasise transparency, system autonomy, experiential learning, and trusted governance structures for clinicians.\u003c/p\u003e \u003cp\u003e For policy and practice, it is recommended that independent groups for AI validation and monitoring at the various Federal/National Ministries of Health that can review AI tools before and after they are used be created across the West African sub-region; explicit guidelines on who is responsible when AI affects a patient be set up; investment in digital infrastructure be made; the West African College of Physicians be involved in the process of developing and testing AI from the start; complete training and continuing assistance to help with implementation of new AI be provided; and AI literacy be taught in medical colleges so that future doctors know what AI can and cannot do and how to utilise it correctly.\u003c/p\u003e"},{"header":"Declarations","content":"\u003cp\u003e \u003cstrong\u003eEthics approval and consent to participate:\u003c/strong\u003e \u003cp\u003e Ethical clearance for the study was obtained from the Ethical Clearance Committee of the University of Nigeria Teaching Hospital. Participation was voluntary, and confidentiality and protection against unauthorised access were maintained using unique anonymous codes, entering data on password-protected Android tablets, transmitting forms and storing them securely. The study was conducted in accordance with the ethical standards of the Declaration of Helsinki.\u003c/p\u003e \u003c/p\u003e \u003cp\u003e \u003cstrong\u003eConsent for publication:\u003c/strong\u003e \u003cp\u003eNot applicable.\u003c/p\u003e \u003c/p\u003e \u003cp\u003e \u003cstrong\u003eCompeting interests:\u003c/strong\u003e \u003cp\u003eThe authors declare no competing interests.\u003c/p\u003e \u003c/p\u003e\u003ch2\u003eFunding:\u003c/h2\u003e \u003cp\u003eThe study did not receive any direct research funding.\u003c/p\u003e\u003ch2\u003eAuthor Contribution\u003c/h2\u003e\u003cp\u003eData analysis and interpretation: BSCU, YJC, UUE, CCO, AU. Drafting the article: AO, BH, EME, SMY, KAU. Project supervision: BSCU. Reviewing and editing: All authors (BSCU, YJC, UUE, CCO, AU, AO, BH, EME, SMY, KAU) contributed critically to reviewing and editing multiple versions of the manuscript for important intellectual content. All authors read and approved the final version of the manuscript.\u003c/p\u003e\u003ch2\u003eAcknowledgement\u003c/h2\u003e\u003cp\u003e The authors would like to acknowledge all the West African College of Physicians for the platform provided to support this study.\u003c/p\u003e\u003ch2\u003eData Availability\u003c/h2\u003e\u003cp\u003eThe data is available from the corresponding author on reasonable request.\u003c/p\u003e"},{"header":"References","content":"\u003col\u003e\u003cli\u003e\u003cspan\u003eTopol EJ. High-performance medicine: the convergence of human and artificial intelligence. Nat Med. 2019;25(1):44\u0026ndash;56. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003e10.1038/s41591-018-0300-7\u003c/span\u003e\u003cspan address=\"10.1038/s41591-018-0300-7\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e. PubMed PMID: 30617339.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eDavenport T, Kalakota R. The potential for artificial intelligence in healthcare. Future Healthc J. 2019;6(2):98. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003e10.7861/futurehosp.6-2-94\u003c/span\u003e\u003cspan address=\"10.7861/futurehosp.6-2-94\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e. PubMed PMID: 31363513.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eYin J, Ngiam KY, Teo HH. Role of Artificial Intelligence Applications in Real-Life Clinical Practice: Systematic Review. J Med Internet Res. 2021;23(4):e25759. doi:10.2196/25759 PubMed PMID: 33885365.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eLiu X, Faes L, Kale AU, Wagner SK, Fu DJ, Bruynseels A, et al. A comparison of deep learning performance against health-care professionals in detecting diseases from medical imaging: a systematic review and meta-analysis. Lancet Digit Health. 2019;1(6):e271\u0026ndash;97. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003e10.1016/S2589-7500\u003c/span\u003e\u003cspan address=\"10.1016/S2589-7500\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e. (19)30123-2 PubMed PMID: 33323251.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eLongoni C, Bonezzi A, Morewedge CK. Resistance to medical artificial intelligence is an attribute in a compensatory decision process: Response to pezzo and beckstead (2020). Judgm Decis Mak. 2020;15(3):446\u0026ndash;8. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003e10.1017/s1930297500007233\u003c/span\u003e\u003cspan address=\"10.1017/s1930297500007233\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eAsan O, Bayrak AE, Choudhury A. Artificial Intelligence and Human Trust in Healthcare: Focus on Clinicians. J Med Internet Res. 2020;22(6):e15154. doi:10.2196/15154 PubMed PMID: 32558657.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eBlease C, Kaptchuk TJ, Bernstein MH, Mandl KD, Halamka JD, Desroches CM. Artificial intelligence and the future of primary care: exploratory qualitative study of UK general practitioners\u0026rsquo; views. J Med Internet Res. 2019;21(3):e12802. doi:10.2196/12802 PubMed PMID: 30892270.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003ePumplun L, Tauchert C, Heidt M. A New Organizational Chassis for Artificial Intelligence - Exploring Organizational Readiness Factors. In: Proceedings of the 27th European Conference on Information Systems (ECIS), Stockholm, 8\u0026ndash;14 June 2019 [Internet]. Stockholm: European Conference on Information Systems; 2019 [cited 2026 Mar 7]. Available from: \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://aisel.aisnet.org/ecis2019_rp/106/\u003c/span\u003e\u003cspan address=\"https://aisel.aisnet.org/ecis2019_rp/106/\" targettype=\"URL\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eCastagno S, Khalifa M. Perceptions of Artificial Intelligence Among Healthcare Staff: A Qualitative Survey Study. Front Artif Intell. 2020;3:578983. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003e10.3389/frai.2020.578983\u003c/span\u003e\u003cspan address=\"10.3389/frai.2020.578983\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eAbdullah R, Fakieh B. Health care employees\u0026rsquo; perceptions of the use of artificial intelligence applications: Survey study. J Med Internet Res. 2020;22(5):e17620. doi:10.2196/17620 PubMed PMID: 32406857.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eAdebayo P, Oluwole OJ, Taiwo FT. COVID-19 and Teleneurology in Sub-Saharan Africa: Leveraging the Current Exigency. Front Public Health [Internet]. 2021 Jan 25 [cited 2026 Mar 7];8(574505):1. Available from: \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.3389/fpubh.2020.574505\u003c/span\u003e\u003cspan address=\"10.3389/fpubh.2020.574505\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eOwoyemi A, Owoyemi J, Osiyemi A, Boyd A. Artificial Intelligence for Healthcare in Africa. Front Digit Health. 2020;2:6. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003e10.3389/fdgth.2020.00006\u003c/span\u003e\u003cspan address=\"10.3389/fdgth.2020.00006\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e. PubMed PMID: 34713019.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eZhou X, Jomhari N, Farid NDN, Xing Y. Physicians\u0026rsquo; perceptions and attitudes on artificial intelligence in hospital management: a qualitative study. BMC Health Serv Res. 2025;26(1):14. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003e10.1186/s12913-025-13815-1\u003c/span\u003e\u003cspan address=\"10.1186/s12913-025-13815-1\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e. PubMed PMID: 41310636.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eHeinrichs H, Kies A, Nagel SK, Kiessling F. Physicians\u0026rsquo; Attitudes Toward Artificial Intelligence in Medicine: Mixed Methods Survey and Interview Study. J Med Internet Res. 2025;27(1):e74187. doi:10.2196/74187 PubMed PMID: 40857713.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eAlaran MA, Lawal SK, Jiya MH, Egya SA, Ahmed MM, Abdulsalam A, et al. Challenges and opportunities of artificial intelligence in African health space. Digit Health. 2025;11:20552076241305916. doi:10.1177/20552076241305915 PubMed PMID: 39839959.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eHenzler D, Schmidt S, Ko\u0026ccedil;ar A, Herdegen S, Lindinger GL, Maris MT, et al. Healthcare professionals\u0026rsquo; perspectives on artificial intelligence in patient care: a systematic review of hindering and facilitating factors on different levels. BMC Health Serv Res. 2025;25(1):633. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003e10.1186/s12913-025-12664-2\u003c/span\u003e\u003cspan address=\"10.1186/s12913-025-12664-2\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e. PubMed PMID: 40312413.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eHolden RJ, Karsh BT. The Technology Acceptance Model: Its past and its future in health care. J Biomed Inf. 2010;43(1):159\u0026ndash;72. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003e10.1016/j.jbi.2009.07\u003c/span\u003e\u003cspan address=\"10.1016/j.jbi.2009.07\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e. .002 PubMed PMID: 19615467.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eVenkatesh V, Thong JYL, Xu X, Venkatesh V, Thong JYL, Xu X. Unified Theory of Acceptance and Use of Technology: A Synthesis and the Road Ahead. J Assoc Inf Syst. 2016;17(5):1. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003e10.17705/1jais.00428\u003c/span\u003e\u003cspan address=\"10.17705/1jais.00428\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eSchoville RR, Titler MG, Computers. Inf Nurs. 2015;33(3):99\u0026ndash;107. 10. 1097/CIN.0000000000000130 PubMed PMID: 25799235.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eGagnon MP, Ngangue P, Payne-Gagnon J, Desmartis M. m-Health adoption by healthcare professionals: a systematic review. J Am Med Inform Assoc. 2016;23(1):212\u0026ndash;20. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003e10.1093/jamia/ocv052\u003c/span\u003e\u003cspan address=\"10.1093/jamia/ocv052\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e. PubMed PMID: 26078410.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003ePrice WN, Gerke S, Cohen IG. Potential Liability for Physicians Using Artificial Intelligence. JAMA. 2019;322(18):1765\u0026ndash;6. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003e10.1001/jama.2019.15064\u003c/span\u003e\u003cspan address=\"10.1001/jama.2019.15064\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e. PubMed PMID: 31584609.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eChallen R, Denny J, Pitt M, Gompels L, Edwards T, Tsaneva-Atanasova K. Artificial intelligence, bias and clinical safety. BMJ Qual Saf. 2019;28(3):231\u0026ndash;7. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003e10.1136/bmjqs-2018-008370\u003c/span\u003e\u003cspan address=\"10.1136/bmjqs-2018-008370\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e. PubMed PMID: 30636200.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eFloridi L, Cowls J, Beltrametti M, Chatila R, Chazerand P, Dignum V, et al. AI4People-An Ethical Framework for a Good AI Society: Opportunities, Risks, Principles, and Recommendations. Minds Mach (Dordr). 2018;28(4):689\u0026ndash;707. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003e10.1007/s11023-018-9482-5\u003c/span\u003e\u003cspan address=\"10.1007/s11023-018-9482-5\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e. PubMed PMID: 30930541.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eReddy S, Allan S, Coghlan S, Cooper P. A governance model for the application of AI in health care. J Am Med Inform Assoc. 2020;27(3):491\u0026ndash;7. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003e10.1093/jamia/ocz192\u003c/span\u003e\u003cspan address=\"10.1093/jamia/ocz192\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e. PubMed PMID: 31682262.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eHabli I, Lawton T, Porter Z. Artificial intelligence in health care: accountability and safety. Bull World Health Organ. 2020;98(4):256. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003e10.2471/BLT\u003c/span\u003e\u003cspan address=\"10.2471/BLT\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e. .19.237487 PubMed PMID: 32284648.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eWorld Health Organization. Ethics and governance of artificial intelligence for health. Guidance on large multi-modal models [Internet]. Geneva: World Health Organization. 2024 [cited 2026 Mar 7]. 1\u0026ndash;98 p. Available from: \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://iris.who.int/server/api/core/bitstreams/e9e62c65-6045-481e-bd04-20e206bc5039/content\u003c/span\u003e\u003cspan address=\"https://iris.who.int/server/api/core/bitstreams/e9e62c65-6045-481e-bd04-20e206bc5039/content\" targettype=\"URL\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eMorley J, Machado CCV, Burr C, Cowls J, Joshi I, Taddeo M, et al. The ethics of AI in health care: A mapping review. Soc Sci Med. 2020;260:113172. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003e10.1016/j.socscimed.2020.113172\u003c/span\u003e\u003cspan address=\"10.1016/j.socscimed.2020.113172\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e. PubMed PMID: 32702587.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eScott IA, Carter SM, Coiera E. Exploring stakeholder attitudes towards AI in clinical practice. BMJ Health Care Inf. 2021;28(1):e100450. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003e10.1136/bmjhci-2021-100450\u003c/span\u003e\u003cspan address=\"10.1136/bmjhci-2021-100450\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e. PubMed PMID: 34887331.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eAbramson JH, WINPEPI (PEPI-for-Windows). Computer Programs for Epidemiologists. Epidemiol Perspect Innovations. 2004;1(1):6. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003e10.1186/1742-5573-1-6\u003c/span\u003e\u003cspan address=\"10.1186/1742-5573-1-6\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e. PubMed PMID: 15606913.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003evan der Meijden SL, de Hond AAH, Thoral PJ, Steyerberg EW, Kant IMJ, Cin\u0026agrave; G, et al. Intensive Care Unit Physicians\u0026rsquo; Perspectives on Artificial Intelligence-Based Clinical Decision Support Tools: Preimplementation Survey Study. JMIR Hum Factors. 2023;10(1):e39114. doi:10.2196/39114 PubMed PMID: 36602843.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eIBM. IBM Support [Internet]. 2024 [cited 2026 Feb 7]. IBM SPSS Statistics 29. Available from: \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://www.ibm.com/support/pages/downloading-ibm-spss-statistics-29\u003c/span\u003e\u003cspan address=\"https://www.ibm.com/support/pages/downloading-ibm-spss-statistics-29\" targettype=\"URL\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eBraun V, Clarke V. One size fits all? What counts as quality practice in (reflexive) thematic analysis? Qual Res Psychol. 2021;18(3):328\u0026ndash;52. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003e10.1080/14780887.2020.1769238\u003c/span\u003e\u003cspan address=\"10.1080/14780887.2020.1769238\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eLumivero. Lumivero Community [Internet]. 2026 [cited 2026 Mar 7]. About NVivo (NVivo 14 Windows). Available from: \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://community.lumivero.com/s/article/TRC-About-NVivo-NVivo-14-Windows?language=en_US\u003c/span\u003e\u003cspan address=\"https://community.lumivero.com/s/article/TRC-About-NVivo-NVivo-14-Windows?language=en_US\" targettype=\"URL\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eWalter Z, Lopez MS. Physician acceptance of information technologies: Role of perceived threat to professional autonomy. Decis Support Syst. 2008;46(1):206\u0026ndash;15. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003e10.1016/j.dss.2008.06.004\u003c/span\u003e\u003cspan address=\"10.1016/j.dss.2008.06.004\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eMayer RC, Davis JH, Schoorman DF. An Integrative Model of Organizational Trust. Academy of Management Review [Internet]. 1995 [cited 2026 Mar 7];20(3):709\u0026ndash;34. Available from: \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://www.makinggood.ac.nz/media/1270/mayeretal_1995_organizationaltrust.pdf\u003c/span\u003e\u003cspan address=\"https://www.makinggood.ac.nz/media/1270/mayeretal_1995_organizationaltrust.pdf\" targettype=\"URL\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eMcknight DH, Carter M, Thatcher JB, Clay PF. Trust in a specific technology: An investigation of its components and measures. ACM Trans Manag Inf Syst. 2011;2(2):12\u0026ndash;32. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003e10.1145/1985347.1985353\u003c/span\u003e\u003cspan address=\"10.1145/1985347.1985353\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eZuhair V, Babar A, Ali R, Oduoye MO, Noor Z, Chris K, et al. Exploring the Impact of Artificial Intelligence on Global Health and Enhancing Healthcare in Developing Nations. J Prim Care Community Health. 2024;15:21501319241245850. doi:10.1177/21501319241245847 PubMed PMID: 38605668.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eVictor A. Artificial intelligence in global health: An unfair future for health in Sub-Saharan. Africa? Health Affairs Scholar. 2025;3(2):qxaf023. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003e10.1093/haschl/qxaf\u003c/span\u003e\u003cspan address=\"10.1093/haschl/qxaf\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e. 023 PubMed PMID: 39949826.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003ede Grood C, Raissi A, Kwon Y, Santana MJ. Adoption of e-health technology by physicians: a scoping review. J Multidiscip Healthc. 2016;9:344. 10.2147. /JMDH.S103881 PubMed PMID: 27536128.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eSpil TAM, Schuring RW. E-health systems diffusion and use: The innovation, the user and the USE IT model. E-Health Systems Diffusion and Use: The Innovation, the User and the USE IT Model [Internet]. Hershey: Idea Group Publishing; 2005 [cited 2026 Mar 7]. 1\u0026ndash;342 p. Available from: \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.4018/978-1-59140-423-1\u003c/span\u003e\u003cspan address=\"10.4018/978-1-59140-423-1\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e doi:10.4018/978-1-59140-423-1.\u003c/span\u003e\u003c/li\u003e\u003c/ol\u003e"}],"fulltextSource":"","fullText":"","funders":[],"hasAdminPriorityOnWorkflow":false,"hasManuscriptDocX":false,"hasOptedInToPreprint":true,"hasPassedJournalQc":"","hasAnyPriority":true,"hideJournal":true,"highlight":"","institution":"","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":"artificial intelligence, clinician perceptions, healthcare adoption, trust in AI, clinical autonomy, West Africa, safe integration, AI governance, transparency, local validation","lastPublishedDoi":"10.21203/rs.3.rs-9066307/v1","lastPublishedDoiUrl":"https://doi.org/10.21203/rs.3.rs-9066307/v1","license":{"name":"CC BY 4.0","url":"https://creativecommons.org/licenses/by/4.0/"},"manuscriptAbstract":"\u003cp\u003eThe integration of artificial intelligence (AI) into clinical practice holds significant promise for improving diagnostic accuracy, reducing medical errors, and enhancing healthcare efficiency, particularly in resource-constrained settings. However, successful adoption depends heavily on clinicians\u0026rsquo; perceptions, trust, and concerns regarding autonomy, reliability, and safety. Empirical evidence on West African physicians\u0026rsquo; views remains limited, despite unique regional challenges like infrastructure deficits and workforce shortages. This study aimed to assess clinicians\u0026rsquo; perceptions of AI in healthcare, identify factors influencing willingness to adopt AI tools, and explore recommended frameworks for safe, accountable integration into clinical practice across West Africa. A cross-sectional survey of 136 physicians affiliated with the West African College of Physicians and 72 key informant interviews were conducted. While 85.3% agreed that AI could improve diagnostic accuracy and 83.1% believed it could reduce errors, 77.9% perceived AI as a threat to clinical autonomy, and 67.6% rated AI information as unreliable. Despite low prior AI experience (only 33.1% had used AI tools) and limited familiarity, 94.1% expressed willingness to use AI if proven effective. Trust in AI was the strongest predictor of adoption willingness (β\u0026thinsp;=\u0026thinsp;0.48, p\u0026thinsp;\u0026lt;\u0026thinsp;0.001), with prior use also significant (β\u0026thinsp;=\u0026thinsp;0.21, p\u0026thinsp;=\u0026thinsp;0.03). Younger clinicians (0\u0026ndash;10 years\u0026rsquo; experience) showed higher willingness than those with 20\u0026thinsp;+\u0026thinsp;years (mean scores 4.5 vs. 4.1, p\u0026thinsp;\u0026lt;\u0026thinsp;0.05). Qualitative findings highlighted AI\u0026rsquo;s potential as a cognitive partner for decision support, error reduction, and administrative relief but raised major concerns about over-reliance, skill erosion, workflow disruption, and accountability. Clinicians emphasised three core requirements for trust: transparency and explainability, local validation in similar populations, and clear governance with defined accountability mechanisms. West African clinicians recognise AI\u0026rsquo;s potential benefits but exhibit low current trust and significant autonomy concerns, driven by limited experience, perceived unreliability, and contextual barriers. Willingness to adopt was highly conditional on proven effectiveness. Safe integration of AI requires frameworks prioritising transparency, local validation, clinician-centred design, robust governance (preferably independent oversight), infrastructure investment, and AI literacy in medical training. These findings guide contextually appropriate AI policies and implementation strategies to enhance patient safety and care quality in West Africa.\u003c/p\u003e \u003cp\u003e \u003cb\u003eClinical trial number\u003c/b\u003e: Not applicable.\u003c/p\u003e","manuscriptTitle":"Clinician Perceptions of Artificial Intelligence in Healthcare and Frameworks for Ensuring Safe Integration into Clinical Practice of the West African College of Physicians: A Multi-Country Mixed-Methods Study","msid":"","msnumber":"","nonDraftVersions":[{"code":1,"date":"2026-03-10 14:51:08","doi":"10.21203/rs.3.rs-9066307/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":"3b806182-9dce-42f8-89a6-1ff76adfe669","owner":[],"postedDate":"March 10th, 2026","published":true,"recentEditorialEvents":[],"rejectedJournal":[],"revision":"","amendment":"","status":"posted","subjectAreas":[],"tags":[],"updatedAt":"2026-04-01T11:26:22+00:00","versionOfRecord":[],"versionCreatedAt":"2026-03-10 14:51:08","video":"","vorDoi":"","vorDoiUrl":"","workflowStages":[]},"version":"v1","identity":"rs-9066307","journalConfig":"researchsquare"},"__N_SSP":true},"page":"/article/[identity]/[[...version]]","query":{"redirect":"/article/rs-9066307","identity":"rs-9066307","version":["v1"]},"buildId":"XKTyCvWXoU3ODBz1xrDgd","isFallback":false,"isExperimentalCompile":false,"dynamicIds":[84888],"gssp":true,"scriptLoader":[]}