Adoption, perceived value, and barriers to generative artificial intelligence in emergency medicine: A cross-sectional survey of Latin American clinicians

Abstract

Background:

Generative artificial intelligence (AI) tools are increasingly used by healthcare professionals, yet empirical data on adoption among Latin American emergency medicine clinicians remain limited.

Methods:

A voluntary, anonymous, cross-sectional online survey was administered to emergency medicine professionals across Latin America (April–October 2025). The Spanish-language instrument, pilot-tested before deployment, assessed AI familiarity, use patterns, perceived value, risks, governance, and training needs. Descriptive statistics and Chi-square tests were used for analysis.

Results:

A total of 123 respondents from 16 countries were analyzed. Most practiced in public or mixed healthcare settings (71.3%) and had over 10 years of experience (61.3%). Familiarity with generative AI was moderate to high (76.9%), and professional use was common (78.3%), with ChatGPT the dominant platform (81.0%). Adoption patterns were homogeneous across countries, with no significant associations between country of origin and familiarity, frequency of use, or platform preference. Primary applications included research and literature review (66.1%), medical education (59.1%), and clinical documentation (40.9%). Nearly half reported high perceived utility (49.6%), and 51.7% noted improvements in productivity. Major concerns included misinformation, erosion of clinical judgment (56.2% each), algorithmic bias (53.7%), and legal liability (46.3%). Institutional AI governance was nearly absent, with only 7.4% reporting existing policies. Despite this, 88.4% expressed interest in formal AI training and 65.6% supported integration into clinical guidelines.

Conclusions:

Latin American emergency clinicians are rapidly adopting generative AI with perceived benefits across multiple domains, but almost entirely without governance frameworks or structured training, highlighting an urgent need for regional policy and education infrastructure.

INTRODUCTION

Generative artificial intelligence (AI) has rapidly evolved from an academic concept to a tool widely tested in clinical settings.[1,2] Large language models (LLMs) such as ChatGPT have enabled novel applications in clinical documentation, decision support, and educational content generation, with demonstrated potential to improve workflow efficiency and clinician productivity.[3,4] However, significant questions remain regarding the reliability, safety, bias, and responsible clinical integration of these systems.[5,6,7,8,9]

In emergency medicine, a specialty characterized by high-acuity, time-sensitive decisions, high patient volumes, and substantial documentation burden, AI has been explored for its potential in data synthesis, workflow optimization, and clinical decision support.[10,11] Despite this promise, concerns persist regarding model hallucinations (i.e., the fabrication of plausible but inaccurate content), bias in training data, interpretability, and clear accountability frameworks for real-world deployment.

Survey data on clinician adoption and perceptions of generative AI are growing but remain concentrated in high-income settings. A survey of primary care physicians in the United Kingdom found that while generative AI was perceived as potentially workload-reducing, limited training remained a central concern.[12,13] A 2024 survey of UK general practitioners reported that approximately one in five had used generative AI tools in clinical practice, with a majority expecting positive productivity impacts.[13] In the United States, the American Medical Association reported that physician AI use more than doubled between 2023 and 2024, with 66% of physicians using AI in at least one professional context by 2024.[14] A national convenience-sample survey of American College of Emergency Physicians (ACEP) members found that emergency physicians were cautiously optimistic about AI, reporting adoption for noninstitutional tasks in approximately one-third of cases, with concerns about data privacy, bias, and inconsistent institutional policies reflecting themes similar to those identified in our dataset.[15] These findings highlight a global pattern of accelerating informal adoption in advance of formal governance and training, a pattern that appears particularly pronounced in lower-resource settings.

The adoption of AI in emergency medicine within Latin America reflects a complex landscape shaped by high patient volumes, limited specialist availability, substantial administrative burdens, variable digital infrastructure, fragmented regulatory environments, and uneven access to continuing professional education. In this context, generative AI tools may function as pragmatic compensatory tools, enabling clinicians to access information, synthesize evidence, and manage documentation more efficiently. However, informal adoption may be occurring ahead of institutional policies, training frameworks, or ethical oversight.

Understanding how emergency medicine professionals in Latin America are engaging with generative AI, including their use patterns, perceived value, recognized risks, and training needs, is essential for designing responsible implementation strategies aligned with both clinical realities and governance requirements. This study aims to describe current patterns of use, perceived benefits and risks, institutional readiness, and educational needs related to generative AI among emergency medicine professionals in Latin America, providing an empirical foundation for policy development and educational intervention.

METHODS

Study design and setting

We conducted a voluntary, anonymous, cross-sectional online survey targeting physicians and healthcare professionals practicing emergency medicine in Latin America. The study was approved by the Institutional Review Board (IRB) (determination date: 9 January 2025) of Universidad Nacional Pedro Henríquez Ureña (UNPHU), Santo Domingo, Dominican Republic, with a determination of minimal-to-no risk and exemption from full review. The survey was administered from April 2025 to October 2025, with data analysis conducted from October to December 2025.

Sampling and survey dissemination

Participants were recruited through a convenience sampling strategy utilizing multiple distribution channels: (1) professional emergency medicine societies and associations in Latin America; (2) institutional academic and hospital networks; (3) professional social media platforms and specialty-specific online groups; and (4) peer-to-peer referral via participating clinicians. No incentives were offered. Participation was entirely voluntary. The survey was administered in Spanish, the primary professional language across the target region, to maximize accessibility and response validity.

Survey instrument

The survey instrument [Table 1] was developed collaboratively by the authorship team to capture the domains most relevant to generative AI adoption in emergency medicine: demographic and professional characteristics, institutional context, digital infrastructure access, AI familiarity and use patterns, perceived clinical value, confidence in AI outputs, productivity impact, perceived risks and barriers, institutional governance, and training preferences. The instrument included Likert-scale items (1–5), multiple-choice questions, and open-text fields.

Table 1.

Survey instrument: Generative artificial intelligence in emergency medicine

Domains		Survey item		Response options
Demographics		Country where you practice emergency medicine		Free text
Institutional context		Type of institution where you currently work		Public Private Mixed Other
Professional experience		Years of experience in emergency medicine		<5 5–10 >10
Digital access		Access to the Internet and digital devices at work		Yes No
AI familiarity		Level of familiarity with ChatGPT or other generative AI tools		1 (none)–5 (very high)
AI use		Have you used generative AI tools in your professional practice?		Yes No
Use cases		Areas where you have used or considered AI useful		Clinical notes/summaries; Diagnostic support; Translation/patient communication; Educational material; Research and literature review; Radiologic interpretation; Other
Platform used		Main AI platform used		ChatGPT; Copilot; Gemini; Grok; Claude; Perplexity AI; DeepSeek; Other
Knowledge source		Main source of knowledge about generative AI		Colleagues; Social media; Articles; Courses/workshops; Other
Perceived value		Perceived value of generative AI in emergency medicine		1 (very low)–5 (very high)
Confidence		Level of confidence in AI-generated information		1 (very low)–5 (very high)
Productivity impact		Perceived impact on productivity and quality of work		1 (very low)–5 (very high)
Clinical integration		Would you accept integrating AI into official clinical protocols or guidelines?		Yes No Not sure
Perceived risks		Main risks associated with AI use in clinical practice		Misinformation; Algorithmic bias; Loss of clinical judgment; Data confidentiality; Legal liability; Other
Barriers		Main barriers to using generative AI in your work		Lack of training; Limited devices/Internet; Institutional restrictions; Cultural resistance; Cost; Legal concerns; Other
Language limitations		Do language or medical terminology limitations affect your use of AI?		Yes No
Institutional policy		Does your institution have clear policies on AI use in health care?		Yes No Don’t know
Training interest		Interest in receiving training on AI use in emergency medicine		Yes No Maybe
Training format		Preferred training format		Virtual In-person Hybrid Other
Open response		Describe a situation where generative AI has been useful or could have been useful in your practice		Open text

Instrument developed and administered in Spanish; translated here for publication. Pilot-tested with five emergency clinicians before full deployment. AI: Artificial intelligence

Before full deployment, the instrument underwent pilot testing with a convenience sample of five emergency medicine clinicians not included in the final study population. Pilot testing assessed item clarity, response burden, and face validity. Minor wording adjustments were made based on pilot feedback before final deployment. No formal psychometric validation was conducted, which is acknowledged as a limitation.

Statistical analysis

All data were exported from the online survey platform and analyzed using descriptive and inferential statistical methods. Categorical variables are reported as frequencies and percentages, calculated based on total valid responses per item. Continuous and ordinal variables (including Likert scales) are described using frequencies and proportions across response categories.

Associations between categorical variables, specifically, between country of origin and AI familiarity, frequency of use, platform preference, and institutional policy status, were assessed using Pearson Chi-square tests. Due to the distribution of responses across country categories, countries with fewer than five respondents were grouped into a residual category for these analyses. Statistical significance was set at P < 0.05. Qualitative open-text responses were reviewed independently by two authors and grouped into thematic categories using an inductive approach to identify recurring patterns of use and perception. All analyses were performed using standard statistical software.

Ethical considerations

All participants provided voluntary electronic informed consent before completing the survey. No personally identifiable information was collected. The study was conducted in full accordance with the Declaration of Helsinki.

RESULTS

Participant characteristics

A total of 123 responses were analyzed, representing healthcare professionals from 16 countries across Latin America. The largest proportions of respondents practiced in the Dominican Republic (24.8%), Mexico (approximately 19.5%), Colombia (14.2%), and Bolivia (6.2%). Most participants worked in public (42.6%) or mixed public–private (28.7%) healthcare settings. Professional experience was substantial: 61.3% reported more than 10 years of experience in emergency medicine, 26.1% reported 5–10 years, and 12.6% fewer than 5 years. Access to the Internet and digital devices for professional use was nearly universal (97.5%). Full participant characteristics are presented in Table 2.

Table 2.

Participant characteristics (n=123)

Variables		n (%)
Country of practice
Dominican Republic		28 (24.8)
Mexico		24 (19.5)
Colombia		16 (14.2)
Bolivia		7 (6.2)
Peru		5 (4.4)
Ecuador		4 (3.5)
Venezuela		3 (2.7)
Uruguay		3 (2.7)
El Salvador		2 (1.8)
Nicaragua		2 (1.8)
Panama		2 (1.8)
Argentina		2 (1.8)
Other (Spain, Brazil, Guatemala, and others)		Remaining
Type of Institution
Public		52 (42.6)
Private		33 (27.0)
Mixed		35 (28.7)
Other		2 (1.7)
Years of experience
<5		15 (12.6)
5–10		31 (26.1)
>10		73 (61.3)
Access to the Internet/devices at work
Yes		118 (97.5)
No		3 (2.5)

Generative artificial intelligence familiarity and use patterns

Most respondents reported moderate-to-high familiarity with generative AI tools [Table 3 and Figure 1]. Specifically, 76.9% reported a familiarity level of 3 or higher on a 5-point scale. More than three-quarters of respondents (78.3%) had used generative AI tools in professional practice. ChatGPT was the most commonly used platform (81.0%), followed by Gemini (31.9%), Copilot (24.1%), and Perplexity AI (11.2%). Country of origin was not significantly associated with AI familiarity (P = 0.41), frequency of use (P = 0.38), or platform preference (P = 0.52), suggesting relatively homogeneous adoption patterns across the region.

Table 3.

Generative AI platforms used by respondents (n=123)

Platforms		n (%)
ChatGPT		94 (81.0)
Gemini		37 (31.9)
Copilot		28 (24.1)
Perplexity AI		13 (11.2)
Grok		10 (8.6)
Claude		3 (2.6)
DeepSeek		3 (2.6)
Others/none		<2 each

AI: Artificial intelligence

Figure 1.

Familiarity with generative artificial intelligence tools among Latin American emergency medicine professionals (n = 121). Scale: 1 = Not at all familiar; 5 = Very familiar

Clinical and professional use cases

Table 4 and Figure 2 detail the domains in which respondents reported using or considering generative AI. The most frequently reported application was research and literature review (66.1%), followed by educational material generation (59.1%), clinical documentation and summarization (40.9%), diagnostic reasoning support (38.3%), and patient communication or translation (32.2%). Radiologic interpretation support was reported by 16.5% of respondents. The breadth of use cases reflects the versatility of generative AI tools across clinical, educational, and administrative domains.

Table 4.

Areas of generative artificial intelligence use reported by respondents (n=123)

Area of use		n (%)
Research and literature review		76 (66.1)
Educational material generation		68 (59.1)
Clinical notes and summaries		47 (40.9)
Diagnostic reasoning support		44 (38.3)
Patient communication/translation		37 (32.2)
Radiologic interpretation support		19 (16.5)
Administrative tasks		<3 (<3)
None		1 (0.9)

Respondents could select multiple areas; percentages exceed 100%

Figure 2.

Perceived value of generative artificial intelligence in emergency medicine (n = 118). Scale: 1 = very low value; 5 = very high value

Perceived value, confidence, and productivity impact

Half of the respondents rated the overall usefulness of generative AI as high (scores of 4 or 5 out of 5), and 51.7% reported a positive impact on productivity and work quality [Table 5 and Figure 2]. Confidence in AI-generated outputs was more cautious: 39.5% of respondents rated their confidence as 3 out of 5, suggesting that while clinicians find generative AI useful, they maintain appropriate critical appraisal of its outputs.

Table 5.

Perceived value, confidence, and productivity impact (n per item varies)

Measures		1 - Very Low		2		3 - Moderate		4		5 - Very high
Perceived value (n=118)		10 (8.5)		11 (9.3)		38 (32.2)		39 (33.1)		20 (16.9)
Confidence in AI outputs (n=119)		7 (5.9)		16 (13.4)		47 (39.5)		43 (36.1)		6 (5.0)
Impact on productivity/quality (n=120)		7 (5.8)		14 (11.7)		37 (30.8)		47 (39.2)		15 (12.5)

AI: Artificial intelligence

Perceived risks and barriers to adoption

The most frequently cited perceived risks were misinformation or errors and erosion of clinical judgment (both 56.2%), followed by algorithmic bias (53.7%), legal liability (46.3%), and data confidentiality concerns (33.9%) [Figure 3 and Table 6]. These concerns are consistent with risks reported in the broader AI in healthcare literature.

Figure 3.

Perceived impact of generative artificial intelligence on work quality and productivity (n = 120). Scale: 1 = no impact; 5 = very high impact

Table 6.

Perceived risks and barriers to adoption (n=123)

Variables		n (%)
Perceived risks
Misinformation/errors		68 (56.2)
Erosion of clinical judgment		68 (56.2)
Algorithmic bias		65 (53.7)
Legal liability		56 (46.3)
Data confidentiality		41 (33.9)
Barriers to adoption
Lack of formal training		64 (54.2)
Legal concerns		47 (39.8)
Cultural resistance to new technologies		45 (38.1)
Restrictive/absent institutional policies		41 (34.7)
Implementation costs		28 (23.7)
Limited Internet/devices		26 (22.0)

Structural barriers to broader adoption included lack of formal training (54.2%), legal concerns (39.8%), cultural resistance to new technologies (38.1%), restrictive or absent institutional policies (34.7%), and implementation costs (23.7%).

Institutional governance

Governance of AI in healthcare appeared markedly underdeveloped in the surveyed sample. Only 7.4% of respondents reported that their institution had clear policies governing AI use, while 77.0% indicated no institutional guidance existed, and 15.6% were uncertain. Institutional policy status was not significantly associated with country of origin (P = 0.47), suggesting that this governance gap is uniformly distributed across the region [Table 7].

Table 7.

Institutional governance and training preferences (n=123)

Variables		n (%)
Institution has AI policy
Yes		9 (7.4)
No		94 (77.0)
Don’t know		19 (15.6)
Interested in AI training
Yes		107 (88.4)
Maybe		11 (9.1)
No		3 (2.5)
Preferred training format
Virtual		85 (69.7)
Hybrid		32 (26.2)
In-person		4 (3.3)
Other		1 (0.8)
Support AI integration into clinical guidelines
Yes		80 (65.6)
Not sure		36 (29.5)
No		6 (4.9)

Training interest and preferences

Interest in formal AI training was very high (88.4%), with a strong preference for virtual (69.7%) or hybrid (26.2%) formats. Importantly, 65.6% of respondents supported integration of generative AI into clinical protocols or official guidelines, indicating readiness for structured, regulated adoption once appropriate safeguards are in place.

DISCUSSION

This study provides a regional snapshot of how emergency medicine clinicians across 16 Latin American countries are engaging with generative AI. The central finding is that rapid, largely informal adoption is occurring across diverse professional and institutional contexts, driven by perceived gains in efficiency, education, and cognitive support, but in the near-total absence of institutional governance, formal training, or standardized oversight frameworks.

The adoption rate observed in this study (78.3%) is notably high by international standards. In the United States, the ACEP national convenience-sample survey found emergency physicians reporting noninstitutional AI use in approximately one-third of cases, with attitudes generally optimistic but governance concerns similarly prevalent.[15] In the United Kingdom, a 2025 longitudinal survey of general practitioners found that approximately 25% reported using generative AI tools in clinical practice, up from 20% in 2024, with documentation and differential diagnosis generation as primary use cases, and near-universal agreement (79%) that increased training was needed.[13,16,17] The considerably higher adoption rate observed in our Latin American sample may reflect the acute information and documentation pressures facing emergency clinicians in resource-constrained settings, where generative AI may serve a compensatory function for limited access to subspecialists, current medical literature, or administrative support. Alternatively, response bias toward more digitally engaged clinicians may partly explain this difference, a limitation acknowledged below.

Consistent with findings from the United States and Europe, the most prevalent applications of generative AI in this dataset were research and literature synthesis, educational content generation, and clinical documentation, areas where LLMs demonstrably reduce time burden without requiring highly specialized medical judgment. Diagnostic reasoning support, reported by 38.3% of respondents, represents a higher-stakes application that warrants closer scrutiny and validates the concern, expressed by over half of respondents, about erosion of clinical judgment. This coexistence of enthusiasm and caution reflects a mature pragmatic understanding of the technology’s capabilities and limitations.

The near-universal absence of institutional governance (92.6% of respondents reporting no clear AI policy) is perhaps the most policy-relevant finding of this study and aligns with patterns observed globally. The ACEP survey similarly found inconsistent institutional policies as a key barrier in U.S. emergency departments.[15] In the UK, despite a more developed national regulatory environment, governance gaps in clinical AI use persist at the institutional level. In the Latin American context, this vacuum places legal, ethical, and professional responsibility entirely on individual clinicians, a particularly precarious position given the documented risks of AI misinformation and hallucination in high-acuity settings.[10]

The strong demand for formal training (88.4%) represents a clear and actionable signal for professional societies, academic institutions, and ministries of health. Training needs extend beyond basic tool literacy and should address AI model limitations, prompt engineering, bias recognition, data governance, documentation accountability, and ethical frameworks for human-in-the-loop decision-making. The strong preference for virtual (69.7%) and hybrid (26.2%) formats reflects the geographic distribution and time constraints of emergency clinicians across the region and should inform curriculum delivery strategies.

From a policy perspective, the current landscape presents a classic tension between bottom-up innovation and top-down regulation. Rather than prohibition, which risks driving use underground while depriving clinicians of potentially valuable tools, regulatory strategies should focus on enabling safe, transparent, and accountable use. A staged implementation model could guide this transition: (1) a recognition phase that formally acknowledges existing informal use and defines its scope; (2) a capacity-building phase providing structured education and competency frameworks; (3) a governance phase developing institutional and national policy; (4) a pilot implementation phase with monitored, auditable use cases; and (5) an evaluation and scaling phase with continuous assessment of safety, quality, and equity outcomes.

Latin America’s heterogeneous digital infrastructure and regulatory landscape mean that national and regional approaches will need to be context-sensitive. Regional professional bodies could play a convening role by issuing consensus guidance, minimum standards, and ethical principles tailored to emergency care. Such guidance should emphasize human-in-the-loop oversight, advisory – not prescriptive – status of AI outputs, transparent documentation of AI-assisted reasoning, and mechanisms for error reporting and learning. Equity considerations are also essential: as generative AI becomes embedded in clinical workflows, its benefits must not accrue only to well-resourced institutions. These recommendations, together with the responsible actors and supporting evidence, are summarized in Table 8.

Table 8.

Key policy recommendations derived from study findings

Domains		Recommendation		Responsible actor		Evidence base
Governance		Develop baseline institutional AI policies defining acceptable use, data privacy standards, documentation requirements, and accountability mechanisms		Hospitals and health ministries		7.4% institutional policy coverage; 92.6% governance gap
Workforce training		Design and deploy virtual/hybrid AI literacy curricula for emergency clinicians covering model limitations, bias recognition, prompt literacy, ethics, and accountability		Professional societies and academic institutions		88.4% demand for training; 69.7% preference for virtual format
Clinical guidelines		Develop consensus-based guidance for acceptable and nonacceptable clinical AI applications in emergency settings, emphasizing advisory – not prescriptive – status of AI outputs		Emergency medicine societies and national bodies		65.6% support for AI integration into clinical protocols; 56.2% misinformation concern
Undergraduate and GME curriculum		Embed AI literacy competencies into medical education and residency training programs, normalizing safe, evidence-based use as a professional standard		Medical schools and residency programs		54.2% barrier: Lack of formal training
Regional coordination		Convene multicountry expert consensus process (e.g., Delphi) to define minimum standards for AI use in emergency care, including governance, training, and evaluation metrics		PAHO and regional emergency medicine bodies		Adoption homogeneous across 16 countries; no significant country-level governance differences
Patient safety oversight		Establish mechanisms for monitoring, reporting, and learning from AI-related errors or near-misses in clinical settings, aligned with existing patient safety frameworks		Health ministries and accreditation bodies		56.2% misinformation risk; 53.7% algorithmic bias concern

AI: Artificial intelligence, PAHO: Pan American Health Organization, GME: Graduate Medical Education

Comparison with international evidence

The patterns identified in this study align substantively with evidence from higher-income settings, while revealing important contextual differences. The primary use cases (documentation, research and education), the dominant platform (ChatGPT), the coexistence of enthusiasm and risk awareness, the demand for training, and the governance vacuum are consistent across the ACEP U.S. survey, the UK GP longitudinal surveys, and this Latin American dataset.[13,15,16,17] The key distinction is scale: adoption rates appear higher in our sample, governance infrastructure is more absent, and the resource constraints motivating adoption are more acute. This suggests that the transition from informal to structured AI use may need to occur more urgently, and with more contextually tailored support, in Latin America than in settings with more developed digital health infrastructure.

Limitations

This study has several limitations that should be considered when interpreting the findings. First, the survey relied on self-reported data and is subject to social desirability and recall bias. Second, the convenience sampling strategy, utilizing professional networks and social media, likely overrepresents digitally engaged, urban, and institutionally affiliated clinicians, potentially inflating adoption and familiarity estimates relative to the broader population of emergency medicine professionals in Latin America. Third, the sample, while geographically diverse across 16 countries, is not statistically representative of any individual country or the region as a whole, and the small per-country sample sizes limited country-level subgroup analysis. Fourth, the cross-sectional design precludes causal inference and longitudinal assessment of adoption trends. Fifth, while the instrument underwent pilot testing, formal psychometric validation was not performed, and the survey was not adapted from a previously validated instrument. Finally, the study did not assess objective clinical outcomes, actual AI output quality, or patient safety metrics associated with AI use.

CONCLUSIONS

Generative AI tools are already being widely adopted by emergency medicine professionals in Latin America, primarily for education, research, clinical documentation, and diagnostic reasoning support. Clinicians perceive substantial benefits but also recognize significant ethical, legal, and safety risks. This adoption is occurring almost entirely in the absence of institutional governance, structured training, or formal oversight, a pattern consistent with, but more pronounced than, trends observed in the United States and Europe.

These findings underscore an urgent need for regionally relevant frameworks that promote responsible, equitable, and safe integration of generative AI into emergency care. Investment in workforce education, institutional governance, and policy development, delivered through professional societies, academic institutions, and ministries of health, will be essential to realizing the benefits of generative AI while safeguarding clinical judgment, patient safety, and professional accountability. This study provides an empirical foundation for those efforts.

Research Quality and Ethics Statement

This study was conducted in accordance with the ethical principles of the Declaration of Helsinki. Ethical review was performed by the Institutional Review Board of Universidad Nacional Pedro Henríquez Ureña (UNPHU), Santo Domingo, Dominican Republic, which determined the study to be less than minimal risk and exempt from full board review (determination date: 9 January 2025). Participation was voluntary and anonymous, and electronic informed consent was obtained from all participants before survey completion. The survey instrument underwent pilot testing prior to deployment, and the study was designed, conducted, analyzed, and reported in accordance with accepted standards for cross-sectional survey research.

Availability of data and materials

The datasets generated and analyzed during the current study are available from the corresponding author upon reasonable request. Due to the anonymous nature of the survey and the need to protect participant confidentiality, the full dataset is not publicly available. Aggregated data supporting the findings are included within the article.

Authors’ contributions

AAB conceived the study and supervised the research project. AAB, CA, and BD contributed to the study design and development of the survey instrument. Data collection and analysis were performed collaboratively by all authors. AAB drafted the initial manuscript. CA and BD contributed to data interpretation and critically revised the manuscript for intellectual content. All authors reviewed and approved the final version.

Conflicts of interest

There are no conflicts of interest.

Funding Statement

Nil.