Nurses’ knowledge and attitudes toward artificial intelligence in Antigua hospital

Artificial intelligence is reshaping modern healthcare at a remarkable pace, from diagnostic algorithms that read medical images to predictive models that flag deteriorating patients hours before a crisis. Yet the professionals who spend the most time at the bedside, registered nurses, have often been overlooked in conversations about AI adoption. A new study conducted at a national hospital in Antigua and Barbuda now offers one of the first detailed portraits of how nurses in the English-speaking Caribbean understand and feel about this technology, and the results reveal a striking paradox: most nurses know very little about AI in their own practice, yet most believe it will fundamentally transform their profession and want it built into their training.

The research, led by Kadian Henry of The University of the West Indies, Five Islands Campus, together with colleagues at the university’s Cave Hill Campus in Barbados, surveyed registered nurses at Sir Lester Bird Medical Centre, the country’s national referral hospital. All 179 registered nurses employed at the institution were invited to participate, and 86 completed the survey, a response rate of 48.1 percent. The participants had a mean age of 37.6 years, representing a workforce in the middle of their careers, and the study was approved by the University of the West Indies Five Islands Campus Research Ethics Committee and the hospital’s Institutional Review Board in accordance with the Antigua and Barbuda Medical Council Code of Ethics.

The methodological design combined descriptive statistics with formal psychometric evaluation. Participants answered demographic questions and items probing their knowledge of AI, the channels through which they encountered AI-related information, and their views on the specific competencies nurses should possess in an AI-enabled workplace. Attitudes were measured with five-point Likert-type items, and the researchers assessed the internal consistency of the attitude scale using Cronbach’s alpha, a statistical measure of how coherently a set of items taps a single underlying construct. The 18 attitude items produced an alpha of 0.891, a value conventionally considered excellent, which suggests the scale reliably captured the nurses’ overall disposition toward AI. To examine associations between background characteristics and specific attitudes, the team used ordinal logistic regression, a technique suited to ordered response categories such as disagree, neutral and agree. Differences in perceived competency requirements across groups of nursing students, practising nurses and nursing faculty were tested with Cochran’s Q and McNemar tests, paired-comparison procedures appropriate for repeated or matched categorical data.

The headline finding on knowledge was sobering. Only 34.9 percent of respondents indicated that they were aware of AI’s application in nursing practice, meaning roughly two out of three nurses at the hospital did not recognise that artificial intelligence already intersects with their daily work. Just as telling were the sources from which nurses reported learning about speech-to-text recognition applications, one of the most tangible examples of AI entering clinical documentation. Television and radio led the list, cited by 81.4 percent of respondents, followed closely by social media at 80.2 percent. Formal channels lagged far behind: only 27.9 percent named educational institutions and a mere 16.3 percent named the workplace. In other words, nurses are learning about transformative clinical technology primarily from entertainment media and social platforms rather than from the professional institutions responsible for preparing them for practice, a pattern with obvious implications for the accuracy and depth of their understanding.

Despite this knowledge gap, the nurses’ attitudes were strikingly positive and, in some respects, visionary. A majority, 52.9 percent, agreed or strongly agreed that AI would revolutionise nursing. Large proportions endorsed specific clinical applications as useful: 67 percent favoured automated visit documentation, 56.0 percent supported AI-generated intervention recommendations, 55.3 percent saw value in automated problem list generation, 50.6 percent each approved of AI identifying appropriate care interventions and recommending care coordination strategies, and 50.5 percent valued AI-assisted assessment of social determinants of health, the environmental and socioeconomic factors that shape patient outcomes. These endorsements suggest that once concrete use cases are described, nurses can readily see how algorithmic tools might relieve documentation burdens and augment clinical decision-making rather than simply abstract away from their lived work.

Perhaps the most reassuring result concerns job security, a theme that dominates public anxiety about automation. When asked whether AI would soon replace nurses, 64.3 percent disagreed or strongly disagreed. This confidence appears well grounded in the technical realities of nursing, a profession whose core activities, physical assessment, emotional support, advocacy, ethical judgment and hands-on care, remain among the most difficult tasks to automate. Interestingly, the nurses were more ambivalent about the experiential dimension of the technology: 63.5 percent were neutral on whether AI would make nursing more exciting, suggesting neither enthusiasm for a technological future nor dread of it, but a wait-and-see posture. Meanwhile, 61.2 percent agreed or strongly agreed that nursing education should include information on AI applications, signalling a clear appetite for formal training.

The regression analysis uncovered a meaningful education effect. Nurses holding diploma or associate degrees were substantially less likely than those with bachelor’s or master’s degrees to believe that automated identification of care interventions would be useful, with an odds ratio of 0.27 and a 95 percent confidence interval of 0.11 to 0.64. Because this confidence interval excludes 1.0, the association is statistically significant at conventional thresholds, and the magnitude is considerable: degree-qualified nurses had roughly a quarter to a third of the odds of their lower-credentialed colleagues of endorsing this particular application. The finding points to a gradient of technological acceptance that tracks educational exposure, and it raises equity concerns for workforces in which diploma-level preparation remains common. If AI literacy and acceptance concentrate among the most highly educated nurses, hospitals risk creating a two-tier environment in which enthusiasm and competence for digital tools are unevenly distributed across shifts and units.

The competency analysis added a normative dimension to the picture. Respondents were asked whether different groups, nursing students, practising nurses and nursing faculty, should be competent in various technical domains of AI. Expectations were higher for practising nurses and faculty than for students, a difference that reached statistical significance at p < 0.016. This pattern is intuitively sensible: faculty are expected to teach emerging content and practising nurses to apply it, while students are still acquiring fundamentals. Yet it also implies a clear institutional obligation. If practising nurses and educators are expected to be competent in AI, then hospitals and nursing schools must provide the continuing education, faculty development and infrastructure that make such competence achievable, otherwise the expectation becomes an unfunded mandate placed on clinicians already stretched by staffing pressures.

The Antigua findings resonate with, and extend, a growing international literature on nursing and AI. Studies conducted in North America, Europe and Asia have similarly reported moderate awareness, generally favourable attitudes and strong support for integrating AI content into nursing curricula, but the Caribbean context adds distinctive weight. Small island developing states often face acute health workforce migration, limited continuing education infrastructure and uneven access to digital health technologies. In such settings, the gap between informal information channels and formal professional training may be even wider than in resource-rich health systems, and the cost of leaving AI literacy to television and social media is correspondingly higher. The authors argue that targeted AI education is needed, and this study supplies the baseline evidence on which such programmes can be built, establishing where knowledge currently stands and which attitudes can be leveraged in curriculum design.

What would effective targeted education look like? The study’s findings offer concrete guidance. Because attitudes are broadly favourable but knowledge is thin, curricula should move quickly from motivational content to practical, case-based instruction covering real applications such as automated documentation, clinical decision support and predictive risk scoring, exactly the use cases the nurses rated most useful. Because nurses with lower academic credentials show less acceptance of some applications, training should be tiered and accessible rather than assuming a graduate-level starting point. And because faculty are expected to be competent, investment in educator preparation must come first, since a curriculum on AI delivered by instructors unfamiliar with the technology will not survive contact with a sceptical classroom. The nurses’ neutrality about whether AI will make their work more exciting also suggests that education should emphasise not only skills but realistic framing, showing how algorithmic tools can return time to the bedside rather than adding another layer of screen-based work.

The study arrives at a moment when healthcare systems worldwide are deciding how fast and how far to integrate AI into clinical workflows, and its central message is deceptively simple: the nursing workforce in at least one Caribbean nation is not resistant to artificial intelligence, it is simply uninformed about it, and that distinction matters enormously. Resistance requires persuasion; uninformed readiness requires only good teaching. With strong internal consistency in its attitude measurement, statistically grounded associations between education and acceptance, and a clear mandate from the respondents themselves, this research provides a template that other small island and resource-constrained health systems can adapt. As AI tools continue their advance into documentation, decision support and care coordination, the nurses of Antigua have signalled that they are willing partners in that transformation, provided someone finally teaches them what the technology actually is.