At Bugando Medical Centre, a 1,000-bed zonal referral hospital in Mwanza, Tanzania, a survey of 320 healthcare professionals has found that most clinicians hold a moderate, rather than deep, understanding of artificial intelligence, and that fewer than half have ever used an AI tool in their day-to-day work.
The research team recruited participants through stratified random sampling to ensure representation from departments across the hospital, calculating a minimum sample of 323 using the Kish-Leslie formula with a 95 percent confidence interval, a 5 percent margin of error, and an estimated prevalence of 30 percent drawn from prior work. Interns and volunteer health professionals were excluded. In the end, 320 of 323 consented subjects completed the validated, self-administered questionnaire, a response rate of 99.1 percent. The instrument, adapted from previously published knowledge-attitude-practice studies on AI in healthcare, was pre-tested among 30 healthcare professionals at Sekou Toure Regional Hospital, where it showed good internal consistency, with Cronbach’s alpha values of 0.82 for the knowledge section, 0.85 for attitudes, and 0.79 for practice. Knowledge was scored across ten items and categorized as good at seven or more correct answers, moderate at four to six, and poor below four.
The demographic profile of the respondents skews toward junior clinical staff. Physicians made up 44.5 percent of the sample and nurses 32.8 percent. Most participants held a bachelor’s degree, 70.3 percent in total, and the largest share, 62.1 percent, reported between one and five years of work experience. That composition matters when interpreting the findings, because the analysis linked education level, years of experience, and professional role to whether someone had actually put AI tools to use, with all three associations statistically significant at p < 0.05.
On knowledge, the headline number is that 70 percent of respondents demonstrated moderate understanding of AI. The underlying item-level results suggest the familiarity is real but uneven. Roughly 98 percent could define AI as a computer program used to mimic human behavior, and 88.5 percent knew that machine learning is the most common AI technique, while far fewer, 10.5 percent, identified natural language processing in that role. Understanding of AI’s clinical functions was stronger still: 91.4 percent knew that a primary function of AI in medical systems is to assist healthcare workers with patient diagnosis, 96.1 percent recognized its role in analyzing patient data and suggesting treatment plans, and 87.9 percent understood that AI can help predict disease outcomes and reduce human error in diagnosis. What the survey did not find was a large cadre of clinicians with advanced technical mastery, a gap the authors attribute, by comparison with high-income settings, to differences in access to structured AI training, technological infrastructure, and exposure to digital health tools.
Attitudes, by contrast, were warmly positive. More than two-thirds of respondents, 214 of 320 or 66.9 percent, agreed that it is very important to integrate AI with medical practice, particularly diagnostics and treatment planning. Majorities endorsed specific expected benefits: 76.5 percent strongly agreed that AI would reduce waiting times, 56.8 percent saw it reducing healthcare costs, and 27.9 percent strongly agreed it enhances diagnostic accuracy. Yet the optimism came with a striking caveat. A majority, 161 respondents or 50.3 percent, comprising mostly physicians, specialists, and surgeons, definitely agreed that AI could replace doctors in the future. That belief, sitting alongside enthusiasm for adoption, captures the ambivalence many clinicians report about automation in medicine.
Practice lagged well behind both knowledge and attitude. Only 131 respondents, or 40.9 percent, had ever practically applied an AI tool. Within that group, utilization was heavily concentrated in specific roles: 14 of 17 specialists surveyed, or 94.1 percent, and 15 of 16 medical technologists, or 93.7 percent, had used AI, and all of those who had agreed it made their tasks easier to complete. Physicians fell far lower, with 55 of 142, or 38.7 percent, reporting prior use. On the other side of the ledger, 215 respondents, or 67.2 percent, had never applied AI in any task, including 65 of 105 nurses, or 61.9 percent, 13 of 105 pharmacists, or 12.5 percent, and 121 doctors, or 85.2 percent of that group. The specialists’ and medical technologists’ high rates, the authors suggest, likely reflect greater exposure to digital diagnostic tools in radiology and laboratory services.
Willingness to adopt AI in the future was considerably broader than current use. Of those asked, 318 respondents, or 67.7 percent, said they were ready to practically apply AI going forward, including 88 physicians, or 62.3 percent of that group, and 77 nurses and pharmacists, or 73.5 percent. Meanwhile, 28 medical doctors, or 26.5 percent, expressed no opinion on whether they would work with AI in the future. The bivariate analysis reported in the study found that education level, years of experience, and professional role each had a significant relationship with current practice and utilization of AI tools in medical diagnosis and treatment planning, reinforcing the picture of a workforce whose adoption patterns track exposure and academic background.
The authors are candid about the study’s limits. It was conducted at a single tertiary hospital, which may restrict generalizability to other Tanzanian healthcare settings. Responses were self-reported and therefore vulnerable to recall and social desirability bias. And the cross-sectional design captures a single point in time, which means the significant associations it reports between demographics and AI utilization cannot be read as causal relationships. Within those constraints, the survey still offers baseline evidence about clinician preparedness in a setting where such data have been scarce, and it situates Bugando’s findings alongside comparable work: the moderate-knowledge pattern echoes studies in Syria and the United Kingdom showing basic awareness without deep technical understanding, while the favorable attitudes align with surveys in India and Kenya reporting optimism about AI’s potential to enhance diagnostic accuracy and efficiency.
The gap the study exposes is not one of enthusiasm but of infrastructure and training. The authors point to limited availability of AI-enabled systems, the absence of institutional policies, and insufficient hands-on training as the structural barriers that keep awareness from translating into implementation, a pattern they note has been reported in other low-resource settings. Their recommendations follow directly: institutions should prioritize comprehensive training programs that demystify AI technologies and highlight their potential to enhance diagnostic accuracy and streamline administrative processes, and should foster collaboration between AI developers and healthcare professionals to build user-friendly, context-specific solutions suited to clinical environments. In low- and middle-income countries, AI has been framed as an opportunity to offset shortages of skilled health workers, high disease burden, and limited access to specialized services, but implementation remains constrained by inadequate infrastructure, limited technical expertise, and insufficient training. For Bugando and hospitals like it, the survey’s message is that readiness exists on the attitude side; converting it into practice will require deliberate capacity building, infrastructure strengthening, and policy development rather than waiting for clinicians to bridge the gap on their own.
Subject of Research: Knowledge, attitudes, and practices regarding artificial intelligence in medical diagnosis and treatment planning among healthcare professionals at a Tanzanian tertiary hospital
Article Title: Knowledge, attitude, and practice on the use of artificial intelligence in medical diagnosis and treatment planning among healthcare professionals in a tertiary hospital in Tanzania
Article References: Mhyellah, J. H., Rweyendera, A., Gwahula, G., Vedastus, Y., Mkama, M., Welema, E., Kiritta, R. F., & Jaka, H. (2026). Knowledge, attitude, and practice on the use of artificial intelligence in medical diagnosis and treatment planning among healthcare professionals in a tertiary hospital in Tanzania. BMC Artificial Intelligence, 2(1), Article 9. https://doi.org/10.1186/s44398-026-00028-7
Image Credits: AI Generated
DOI: 10.1186/s44398-026-00028-7
Keywords: artificial intelligence, healthcare professionals, Tanzania, medical diagnosis, treatment planning, cross-sectional study, knowledge attitude practice, Bugando Medical Centre, clinical adoption, low-resource settings, health workforce, digital health
News Source: Ophelia Keating. (October 29, 2026). Most Tanzanian Clinicians Know a Little About AI but Few Actually Use It. Scienmag.
