Abstract
Objective
This study aims to assess the level of readiness among physicians at University of Health Sciences Türkiye, İzmir City Hospital for the adoption of medical artificial intelligence (AI) technologies.
Methods
Participants’ readiness levels were assessed with the medical artificial-intelligence readiness scale devised by Karaca et al. University of Health Sciences Türkiye, İzmir City Hospital employs 1.867 physicians. Using Baş’s (2006) sample-size formula with a ±0.05 margin of error and a 95% confidence level, the minimum required sample was calculated as 319, and 320 physicians ultimately completed the questionnaire. The 22-item scale was subjected to exploratory and confirmatory factor analysis (EFA). The initial solution explained 85.432% of the total variance, with excellent sampling adequacy (Kaiser-Meyer-Olkin=0.964) and a highly significant Bartlett’s test of sphericity (χ²=9.376.445, p<0.001). Inspection of the pattern matrix revealed substantial cross-loadings on three items; these items were removed, and the EFA was rerun on the refined item set.
Results
Statistical analyses showed no significant variation in physicians’ medical-AI readiness (MAIR) across age, sex, or marital status, either for the composite score or for any of the sub-dimensions (p>0.05). Years in practice influenced only the third factor, Foresight, with a significant difference emerging there (p<0.05) but not on the remaining dimensions. Departmental affiliation, by contrast, proved important: except for the ethics sub-scale, all dimensions -and the overall MAIR score- differed significantly among departments (p<0.05). The grand-mean MAIR score was 3.11 on a five-point scale. Thus, physicians’ readiness levels lie slightly above the midpoint, reflecting a generally positive yet essentially ambivalent attitude toward medical AI. The same “marginally above neutral” pattern applies to each individual sub-dimension.
Conclusion
The analysis reveals that physicians adopt a moderately positive stance toward AI, yet they exhibit a pronounced shortfall in the technical knowledge and practical competence required for its effective implementation.
Introduction
In recent years, health-care systems have been among the domains most profoundly affected by technological integration. Rapid population growth, the escalating prevalence of chronic diseases, shortages of health professionals, and persistent concerns over patient safety now necessitate the delivery of care that is more effective, efficient, and universally accessible. One of the principal drivers of this ongoing transformation is, unquestionably, artificial intelligence (AI) technology(1).
AI applications are currently spearheading an extensive process of change and transformation within health and medical domains(2). This technological evolution promises numerous benefits-chief among them greater sector-wide efficiency, improved patient-care services, and a reduction in clinicians’ workload(2, 3). Far from simply easing that workload, AI systems actively support professional judgment across a broad spectrum of functions, ranging from clinical decision‐support systems and patient-monitoring tools to image-processing technologies and digital triage platforms(4). Yet the ultimate success of these innovations hinges on whether the health-care personnel who will use them are cognitively, affectively, and ethically prepared for such sweeping change.
Because physicians occupy a central position in the provision of health‐care services, it is crucial that they interact directly with emerging technologies and be fully prepared to employ them in their decision-making processes(5). Accordingly, the present study has been undertaken to evaluate physicians’ readiness for AI. Its specific objective is to measure the level of readiness among hospital-based physicians with respect to the adoption of AI technologies in clinical practice.
AI: Conceptual and Clinical Perspectives
AI is a set of cognitive algorithms that enables machines to develop human-like capacities for thinking, learning, decision-making, and problem-solving(6). Broadly speaking, AI is defined as the technologies that allow computers to emulate human intelligence. In other words, it describes the reasoning and learning processes of information-technology systems that behave as if endowed with human intellect(7). Yet another formulation defines AI as a system’s ability to interpret data accurately, draw inferences from those data, and-through flexible adaptation-use those inferences to achieve specified goals and execute designated tasks(8). Recent definitions go further, positing the complete transfer of the knowledge stored in the human brain to machines. It is now claimed that a machine can perform cognitive functions traditionally associated with the human mind-perception, inference, learning, environmental interaction, problem-solving, and decision-making-and can even display creativity beyond these capabilities(9).
In the health sector, AI-most notably through machine learning, deep learning, and radiological image-analysis techniques-helps orchestrate the entire treatment pathway by supporting tasks such as disease diagnosis, radiography, pathology, electronic record keeping, risk prediction, patient monitoring, and personalised therapy(6, 10). Beyond these capabilities, AI technologies confer additional advantages, including higher diagnostic accuracy, more practical and effective treatment planning, and easier patient access to care. By harnessing information on patients’ medical histories and drug reactions, AI algorithms allow physicians to design treatment plans more efficiently and to deliver the required interventions in a timely manner(11).
The coronavirus disease-2019 pandemic has made the role of AI in health-care systems even more conspicuous, particularly through its capacity to lighten patient loads, generate data-driven predictions, and supply robust decision-support tools(12). At the same time, the expanding use of AI in the health sector has prompted fresh debates over system explainability, legal liability, ethical limits, and the distribution of professional responsibilities(13, 14).
Today, significant strides are being made toward the integration of routine medical practice with AI technologies(15). Physicians are expected to become adept users who can employ these tools on a broad scale, critically analyse their outputs, and develop a deeper grasp of the underlying algorithms(1). By devising solutions to a wide range of clinical problems and simplifying workflows, medical AI applications hold revolutionary potential in health care-potential that will likely accelerate their incorporation into everyday clinical practice.
Readiness: Conceptual Framework
The concept of readiness refers to the extent to which individuals are mentally, cognitively, emotionally, and socially equipped before undertaking a task. It has also been framed as “the cognitive-emotional disposition to consciously accept, embrace, or reject a specific plan intended to alter the status quo”(16). Although the term has long been used in educational research, it is now widely examined in the contexts of digital transformation and organisational alignment. For health-care professionals, readiness involves far more than possessing relevant knowledge; it also includes an openness to change, the ability to operate new systems, and the capacity to evaluate those systems within ethical and legal frameworks(17).
Successful adoption and implementation of any technology demand a high degree of user acceptance that is calibrated to the specific needs of those who will operate the system(18). In the context of technologies such as AI, readiness should be construed not merely as cognitive awareness but equally as professional competence and digital literacy. The requisite level of technological readiness in an individual provides the essential foundation for learning about-and meaningfully engaging with-the technology(19).
Materials and Methods
Aim and Significance
Given physicians’ pivotal role in health-care delivery, it is essential that they engage directly with emerging technologies and be adequately prepared to employ them in their decision-making processes(5). The present research, entitled “assessment of physicians’ readiness for AI,” seeks to determine whether hospital-based physicians possess a sufficient level of readiness for AI technologies in clinical care. Beyond this overarching aim, the study also examines whether AI readiness varies according to physicians’ age, gender, marital status, department, and years of professional experience.
A review of the domestic literature shows that prior investigations have focused on emergency medical personnel as well as medical-school and nursing students. Notable examples include Boillat et al.'s(20) survey study, “Readiness to Adopt AI among Medical Doctors and Students”, and AlZaabi et al.'s(21) work, “Are Physicians and Medical Students Ready for AI Applications in Health Care?”, both of which compared doctors with students. By contrast, the present study is the first to concentrate exclusively on physicians’ medical-AI readiness.
Research Hypotheses
Drawing on the findings reported in the literature, the study tests the following main and subsidiary hypotheses:
H1 Physicians’ total medical AI-readiness (MAIR) differs significantly by age.
• H1a The cognitive factor differs by age.
• H1b The skill factor differs by age.
• H1c The foresight factor differs by age.
• H1d The ethics factor differs by age.
H2 Physicians’ total AI-readiness differs significantly by sex.
• H2a The cognitive factor differs by sex.
• H2b The skill factor differs by sex.
• H2c The foresight factor differs by sex.
• H2d The ethics factor differs by sex.
H3 Physicians’ total AI-readiness differs significantly by marital status.
• H3a The cognitive factor differs by marital status.
• H3b The skill factor differs by marital status.
• H3c The foresight factor differs by marital status.
• H3d The ethics factor differs by marital status.
H4 Physicians’ total AI-readiness differs significantly by years of professional experience.
• H4a The cognitive factor differs by years of experience.
• H4b The skill factor differs by years of experience.
• H4c The foresight factor differs by years of experience.
• H4d The ethics factor differs by years of experience.
H5 Physicians’ total AI-readiness differs significantly by department.
• H5a The cognitive factor differs by department.
• H5b The skill factor differs by department.
• H5c The foresight factor differs by department.
• H5d The ethics factor differs by department.
Population and Sample
The study was deliberately situated within the health sector-an arena of paramount importance to human well-being-and focused on physicians, whose contributions are pivotal to disease diagnosis and treatment. The target population therefore consisted of all physicians employed at University of Health Sciences Türkiye, İzmir City Hospital, where 1.867 doctors are currently on staff. Applying Baş's(22) formula, the minimum required sample size was calculated as 319, assuming a ±0.05 margin of error and a 95% confidence level. During the online data-collection phase, 320 physicians completed the survey, thereby surpassing the threshold for adequacy.
Ethical Approval
Ethical approval for the study was granted by the University of Health Sciences Türkiye, İzmir City Hospital Social Research Ethics Committee (decision no: 2025/193, dated: 18 April 2025), and data collection commenced only after this clearance had been obtained.
Statistical Analysis
The survey instrument comprised two sections. The first contained seven items eliciting participants’ demographic characteristics. The second employed the medical AI readiness scale developed by Karaca et al.(1) which consists of 22 items grouped into four sub-dimensions-cognitive, skills, foresight, and ethical factors. Data were therefore gathered with a 22-item questionnaire formatted on a five-point Likert scale.
The questionnaire was created online (see link) and circulated to all potential respondents via WhatsApp messaging groups. Alongside demographic information, participants rated every statement on a scale from 1 (“strongly disagree”) to 5 (“strongly agree”) and submitted their responses electronically. Because the entire physician workforce at University of Health Sciences Türkiye, İzmir City Hospital was targeted and the population size was substantial, web-based data collection was deemed most practical.
All data were analysed with appropriate statistical software. Demographic variables were summarised by frequency and cross-tabulation analyses, and overall reliability was assessed by Cronbach’s (1951) alpha. Construct validity and dimensionality were investigated through principal-components analysis with Varimax rotation-initially via exploratory factor analysis (EFA) and subsequently via confirmatory factor analysis (CFA). Hypotheses involving two groups were tested with independent-samples t-tests, whereas comparisons across more than two groups were conducted using One-Way Analysis of Variance (ANOVA). When ANOVA indicated a statistically significant difference, post-hoc multiple comparisons were performed with Tukey’s test to pinpoint specific group differences.
Statistical Thresholds and Study Design
All analyses were performed at the 95% confidence level, and results were deemed statistically significant when p<0.05. The five-point Likert means were interpreted as follows:
This cross-sectional, descriptive field study was conducted online between May and June 2025 and reached 320 physicians.
Results
Demographic Characteristics
A summary of participants’ descriptive statistics is presented in Table 1.
Of the physicians surveyed, 39.7% (n=127) were aged 20-30 years, 22.2% (n=71) were 31-40 years, 23.8% (n=76) were 41-50 years, 10.3% (n=33) were 51-60 years and 4.1% (n=13) were over 61 years. Women accounted for 51.9% (n=166) of respondents, men for 48.1% (n=154). With respect to marital status, 54.1% (n=173) of participants were married, while 45.9% (n=147) were single. In terms of academic rank, 49.4% (n=158) were resident physicians, 34.1% (n=109) were specialists, and 16.6% (n=53) held associate- or full-professor titles. Departmental distribution showed that 57.5% (n=184) worked in internal-medicine disciplines, 38.4% (n=123) in surgical disciplines, 2.5% (n=8) in basic medical sciences and 1.6% (n=5) in health-sciences units. Regarding professional seniority, 53.8% (n=172) had 1-10 years of service, 19.7% (n=63) had 11-20 years and 26.6% (n=85) had 21 years or more. Finally, 90.6% (n=290) reported having at least one social-media account, whereas 9.4% (n=30) did not.
Findings from the Factor-structure and Reliability Analyses
Scale reliability was assessed with Cronbach’s (1951) alpha, while the factor structure and construct validity were evaluated first by EFA and subsequently by CFA. Physicians’ overall MAIR scores were then calculated via descriptive (mean) statistics.
An initial EFA on the 22-item instrument yielded excellent sampling adequacy (KMO=0.964) and a highly significant Bartlett’s test of sphericity (χ2=9.376.445, sig.=0.000). Close inspection revealed cross-loadings on three items; after removing those items, the EFA was repeated. The consolidated outcomes of the final EFA, the CFA fit indices, the sub-scale means and the reliability coefficients are summarised in Table 2.
As Table 2 shows, the scale and all four sub-dimensions display very high internal consistency EFA revealed a four-factor solution, and the subsequent CFA demonstrated that this structure provides the best overall fit: every fit index lies within acceptable limits(19, 23-28), and the CFA factor loadings are uniformly strong. Hence, all further analyses were conducted on the basis of these four factors: Cognitive-6 items, Skill-5 items, Foresight-6 items, Ethics-2 items. The grand-mean score for the scale was 3.11, indicating that, on average, physicians were undecided about their readiness for medical AI-a pattern that held across each sub-dimension as well.
Hypothesis Tests
Age
An ANOVA was performed to determine whether physicians’ total MAIR differs significantly across age groups. The results are summarised in Table 3.
Gender
A series of independent-samples t-tests compared female and male respondents on each readiness dimension and on the overall MAIR score. Descriptive statistics and test results are presented in Table 4.
Independent-samples t-tests showed no statistically significant differences between female and male physicians on the total MAIR score or on any of the four sub-dimensions (p>0.05 for every comparison). Consequently, H2 was rejected.
Marital Status
Whether readiness varies by marital status was examined with another independent-samples t-test. The descriptive statistics and test results are displayed in Table 5.
Physicians’ overall AI-readiness does not differ by marital status in any dimension (p>0.05), so H3 is rejected.
Years in Practice
A One-Way ANOVA tested whether readiness varies across three seniority bands (1-10 y, 11-20 y, ≥21 y). Descriptive statistics and results appear in Table 6.
Only the foresight dimension shows a significant tenure-related difference: physicians with 1-10 years in practice are more optimistic than those with 11-20 years (p<0.05). No significant contrasts appear in the cognitive, skill, ethics, or composite MAIR scores, so H4 is supported solely for the foresight factor.
Department
Whether physicians’ levels of Medical AI Readiness (MAIR) differ according to their department of employment was examined using One-Way Analysis of Variance (ANOVA), and the findings are presented in Table 7.
To test H₅, we ran a ANOVA on total MAIR and each sub-dimension across the hospital’s four departmental clusters (internal medicine, surgical sciences, basic medical sciences, health sciences).
Post-hoc Analysis (Tukey)
Homogeneous subsets are indicated by letter codes, with the mean difference between significantly different pairs shown in parentheses within the same column. The ANOVA revealed that-with the exception of the ethics factor-all sub-dimensions and the overall MAIR score varied significantly across departments (p<0.05). Accordingly, H5 is partially supported. Follow-up comparisons showed that: Physicians working in basic medical sciences achieved higher scores than their colleagues in internal medicine and surgical sciences on the total scale and on both the cognitive and skill factors. They also outperformed physicians in surgical sciences on the foresight factor. No departmental differences emerged for the ethics factor.
Discussion
Identifying physicians’ MAIR via a web-based survey-and examining how that readiness varies across demographic strata-helps forecast both the likely pace of technology adoption and its eventual impact on diagnostic and therapeutic workflows. Because AI systems now permeate virtually every stage of health-care delivery, such insight is indispensable. The present analyses revealed no statistically significant differences (p>0.05) in overall readiness or any sub-dimension with respect to age, sex or marital status. These findings echo those of Çankaya(29), who likewise observed no demographic variation in either total MAIR scores or their sub-scales among emergency-service personnel. By contrast, AlZaabi et al.(21) reported significant discrepancies when comparing physicians with medical students-suggesting that mixed or trainee-inclusive samples may yield patterns that do not apply to practising doctors alone.
Length of professional experience influenced only the foresight dimension of readiness (p<0.05); no differences emerged for the other sub-scales. Hence, H4 was partially supported. Post-hoc analysis showed that physicians with 1-10 years of service were significantly more optimistic than those with 11-20 years. Çankaya's(29), study of emergency-service staff found no tenure-related differences in either the overall scale or its sub-dimensions, whereas AlZaabi et al.(21) did report significant experience effects.
Analyses showed that, with the sole exception of the ethics sub-scale, every readiness dimension-and the total MAIR score-varied significantly by physicians’ departmental affiliation (p<0.05); thus, H₅ is partially supported. Post-hoc Tukey comparisons reveal that physicians based in basic medical sciences score more favourably than their ınternal medicine and surgical sciences colleagues on the overall scale as well as on the cognitive and skill factors. They also outperform surgical-sciences physicians on the foresight factor. No departmental differences emerged for ethics. By contrast, Çankaya's(29), study of emergency-service staff detected no department-related variation in either the composite scale or its sub-dimensions.
The overall mean score for the scale was 3.11, indicating that participants’ readiness for medical AI hovers just above the midpoint and can best be characterised as ambivalent. The same ambivalence holds across all four sub-dimensions. These findings align with several international studies(6, 12), which likewise report mildly positive-yet still uncertain-attitudes among physicians. Although clinicians view AI favourably, gaps in conceptual understanding and hands-on technical training appear to hinder seamless adoption. Accordingly, we recommend embedding core content on AI, machine learning and ethical data use into both undergraduate and residency curricula. In parallel, national guidance that clarifies the legal framework surrounding medical AI is essential to ensure that technological advances proceed in harmony with health-policy objectives.
Study Limitations
Several constraints should be acknowledged. First, the investigation was limited to physicians working at University of Health Sciences Türkiye, İzmir City Hospital; future studies could widen the sample to encompass all hospitals in İzmir and include other health-care professionals in addition to physicians. Second, the demographic section of the questionnaire was restricted to a narrow set of variables-sex, age, marital status, department, years in practice, academic title, and social-media use-thereby excluding potentially relevant factors. Finally, owing to the large target population and the heavy workload within the hospital, data were collected online rather than through face-to-face administration.
Conclusion
This study offered a multidimensional assessment of physicians’ readiness for medical-AI technologies at University of Health Sciences Türkiye, İzmir City Hospital in Türkiye. The analyses show that clinicians hold a moderately positive attitude toward AI, yet they exhibit clear deficits in technical knowledge and hands-on competence. Although their awareness of ethical and legal issues is slightly higher than in other domains, that knowledge remains largely theoretical and has not yet translated into the practical skills needed to evaluate, select or integrate AI systems effectively. In broad terms, the present results are consistent with much of the international literature, even if a few discrepancies emerge across individual studies.
This shortfall can undermine both the effective use of management- and clinical-decision-support systems and the quality of physician–patient–technology communication. A lack of familiarity with algorithmic logic, data types, model-training workflows and system limitations may also erode clinicians’ trust in AI-based tools. In this light, technological adaptation must be treated not merely as the installation of new devices but as a broader cognitive and cultural transformation. To raise physicians’ AI readiness, we recommend the following:
1. Curricular integration – embed core content on algorithm design, machine learning and data ethics in undergraduate and specialty-training syllabi.
2. Continuous professional development – offer regular digital-literacy workshops that focus on hands-on use of AI platforms.
3. Specialty-specific guidance – develop branch-tailored clinical AI guidelines to help physicians select and evaluate tools relevant to their fields.
4. Legal and ethical frameworks – establish national regulations that clarify accountability, data governance and malpractice boundaries for medical AI.
5. Collaborative decision-support models – integrate AI modules into existing clinical-decision workflows so that algorithms and physicians function as partners rather than substitutes.
6. Digital health-communication training – equip clinicians with strategies for explaining AI-assisted care to patients in clear, accessible language.
Treating AI adoption as a composite of technical proficiency, ethical competence and cultural change will position health professionals-and the systems they serve-to realise the full potential of AI in clinical practice.
