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

Abstract

<title>Abstract</title> 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 &lt; 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 &lt; 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. <bold>Clinical trial number</bold> : Not applicable.

Ähnliche Arbeiten