Healthcare leaders’ attitudes and perceptions on the use of artificial intelligence and artificial intelligence enabled tools in healthcare settings

Introduction

Artificial intelligence (AI) is a field of study that combines computer science and linguistics to create intelligent systems capable of analyzing, summarizing, and making predictions with large data sets. AI is a rapidly evolving technology being used across several industries including transportation, finance, and healthcare. The use of AI in healthcare is being studied in clinical practice, medical research and publishing, global and public health, and healthcare administration (1). AI can potentially improve and transform healthcare delivery in all healthcare settings. Studies have shown that AI can be useful in medical diagnosis and treatment planning, support clinical decision-making, and improve patient engagement (2).

The successful adoption and implementation of AI in healthcare settings are affected by multiple interconnected factors influenced by key stakeholder groups (3). Effective adoption and implementation of AI in healthcare settings need to consider the perspectives of the key stakeholder groups. AI has the potential to advance and improve healthcare processes across all settings through its ability to analyze incredibly large datasets and identify patterns. The use of AI and AI-enabled tools in healthcare settings is rapidly increasing and is predicted to have a significant impact on healthcare clinicians, healthcare leaders and managers, patients, and caregivers (4).

AI and AI-enabled tools such as chatbot, chat generative pre-trained transformer (ChatGPT), and clinical algorithms are increasingly being used in healthcare settings to assist with clinician documentation, patient communication, scheduling, and providing referrals (4). Natural language processing (NLP) can be used to analyze clinical notes on patients, transcribe conversations, and prepare reports based on physician examinations (5). This can reduce the amount of time providers are spending on administrative tasks and allow them more time to spend with patients. Rule-based expert systems used in electronic health records (EHR) have been employed in clinical decision-support tools to assist providers with diagnosis and treatment planning (5). These tools are expanding providers’ abilities to provide personalized and precision medicine tailored to patients’ unique conditions.

Studies show that AI used in diagnostic imaging can assist with greater recognition of clinical disease features that can aid in detecting and diagnosing malignant tumors (5). This can lead to more accurate diagnoses (6). Robotics in surgery has improved surgeons’ ability to see, perform minimally invasive procedures, make precise incisions, and stitch incisions (5). In clinical research, AI algorithms assist researchers with constructing cohorts in costly clinical trials, thereby reducing costs (5).

Rationale and knowledge gap

The literature on using AI and AI-enabled tools in healthcare is increasing (7,8). The majority of research in the literature on stakeholders’ perspectives regarding the use of AI in healthcare has focused primarily on understanding physicians’ views. There is a notable gap in research exploring the perspectives of other key stakeholders, such as healthcare leaders and managers (3,7,8). There is an increased need to explore healthcare leaders’ and managers’ attitudes and perceptions towards using AI and AI-enabled tools in healthcare settings. These key stakeholders play a crucial role in the successful adoption and implementation of AI and AI-powered tools in healthcare settings, making it essential to understand and consider their perspectives. This research seeks to fill a significant gap in existing literature by focusing on the perspectives of healthcare professionals in both clinical and administrative roles. The dissemination of these findings can assist healthcare organizations as they progress in integrating AI into both administrative operations and patient care processes.

Research objectives

This qualitative research study examined the attitudes and perceptions of healthcare leaders and managers recruited from various healthcare settings through semi-structured interviews. The aim was to explore: (I) AI application in the organization (clinical and administrative processes); (II) leaders’ perspective on AI; (III) staff engagement in the implementation process and facilitators that enable adoption; and (IV) the challenges faced when implementing AI in healthcare settings. We present this article in accordance with the COREQ reporting checklist (available at https://jmai.amegroups.com/article/view/10.21037/jmai-24-392/rc).

Methods

Participants and study design

In order to recruit participants for this qualitative descriptive study, the researchers shared details about the study with their professional networks and leaders in healthcare organizations across the Midwest United States. Notably, most of these leaders are members of the American College of Healthcare Executives, the Minnesota Organization of Nurse Leaders, and the Society of Hospital Medicine. Research participants were selected because of their experiences and roles as clinical and/or administrative leaders in healthcare settings. To be eligible to participate, participants had to be (I) practicing in leadership position in healthcare settings for more than 2 years; (II) communicate in English language both verbal and written; and (III) willing to participate in a Zoom interview with the research team. The research participants were recruited from a variety of healthcare organizations including hospitals, hospice care, and long-term care organizations located in Midwest United States.

The study was conducted in accordance with the Declaration of Helsinki (as revised in 2013). This study was approved by the Institutional Review Board at Minnesota State University Moorhead (MSUM) (No. 2109683-1) and informed consent was taken from all individual participants. A purposive sampling strategy was used to recruit participants from various healthcare settings (hospitals, long-term care organizations, and hospice). This sampling approach allows researchers to identify and recruit participants based on specific knowledge, experience, and ability to participate in the project. An email describing the purpose of the study was sent to 30 potential participants who met the selection criteria. A follow-up email was sent after a period of 14 days if response was not received. Researchers also included information about voluntary participation and total estimated time to complete the interviews in the initial email. The interview guide included open-ended questions that aimed to explore: (I) AI application in the organization (clinical and administrative processes); (II) leaders’ perspective on AI; (III) staff engagement in the implementation process and facilitators that enable adoption; and (IV) the challenges faced when implementing AI in healthcare settings (see Appendix 1). Interviews were conducted between October 2023 and April 2024. Data collection through interviews continued until saturation was reached. In order to collect demographic information about participants, a separate questionnaire was prepared with items such as gender identity, race, ethnicity, age, nature of current position, years of experience, highest academic degree, and type of academic degree. Braun and Clarke’s thematic framework was selected to analyze the collected data.

Researchers as instrument in research

Prior studies suggest that qualitative studies should include information about the research team’s role with the topic under study (9). It is important to note that the research team for this project comprised of two doctoral prepared lead investigators who had both clinical/practice and academic experience/preparation in information systems and AI. Other team members included students who had experience in either research, administration, and/or academia in the field of healthcare management and leadership. The team members held regular meetings while completing the research project.

Data analysis

Braun and Clarke’s [2006] framework was used to conduct data analysis for this qualitative research study (10,11). In order to familiarize themselves with the data, the team reviewed transcripts numerous times and then initiated the process of open coding to develop initial codes. Researchers utilized Microsoft Excel while working on data analysis. Color coding was utilized to highlight similar (or nearly similar) thoughts/ideas in the transcribed data. More than 40 codes were generated at this step of data analysis which were later condensed to 15 codes as research team continued their work on data analysis. Once codes were finalized, efforts were made to identify common themes. Data was revisited to ensure that themes reflected the information collected during the study. Once this important step was completed, all the codes were organized into themes that provided important information relevant to the study. These themes were reviewed and revised more than once to ensure alignment with the collected data. As the final step in the data analysis process, themes were further refined, and redundancies were removed (if detected at this step). It is important to note that inductive thematic analysis was utilized where data collected during the study served as the initial point to engage with the meaning.

Rigor of the research project

The four dimensions namely: credibility, dependability, confirmability, and applicability were utilized to maintain rigor of the study. Credibility of the overall research process including data collection was maintained by conducting few pilot Zoom interviews, giving participants opportunity to ask any clarifying questions, providing background information, and obtaining participant consent prior to the beginning of study. It is also important to note that only the interviewers and participants were present during the interviews, which were audio recorded. These Zoom sessions were subsequently transcribed. Use of semi-structured interview guide also allowed the research team to request additional information while participants were answering questions. The section researchers as instrument in research further describes how their practice and academic background helped in enhancing credibility of research process. Furthermore, peer debriefing sessions during data analysis allowed for better understanding of the final themes (10,12).

From dependability standpoint, being the subject matter expert, researchers conducted thorough literature review and worked on preparing a draft of study procedures which was submitted to the Office of Research Ethics at the university. Open ended questions on the interview guide allowed participants to express themselves freely during the interview process. Upon completion of the interviews, data was transcribed and reviewed twice by one of the lead researchers and students for accuracy. Transcripts were also sent to participants for confirmation. Data analysis started once confirmation was received from participants. Regular updates on coding, identification of codes, patterns, themes, and sub-themes were sent to the entire team. These updates and consensus building process while conducting analysis helped in enhancing rigor of the study (10,12).

As noted above, meetings and regular updates allowed researchers to incorporate different viewpoints during data analysis process. Investigator triangulation was achieved by discussions and adoption of mutually acceptable research procedures while conducting the study. In addition to transcripts, notes taken during the interviews helped in validating the data that was collected and helped the researchers to establish conformability of the process (10,12).

The applicability of the results of this project to healthcare institutions that are in the process of implementing AI and AI-enabled tools at their settings depends on the degree to which there is similarity between these institutions (work environments) (10-13). It should be noted that use of participants’ quotes while explaining themes allowed researchers to demonstrate the transferability of research findings. Discussions between research team helped in refining themes in the final data analysis process (12).

Results

A purposive sampling strategy was utilized to recruit participants, as this allowed the researcher team to work with individuals who had experience in usage of AI in healthcare settings. Twenty-four healthcare leaders (response rate of 80%) with an average of 15 years of experience in AI projects were chosen to participate in the study. The median age of the participants was 36 years and ranged from 27 to 56 years. The majority (58.82%) of participants were females. All the interviewees were White (100%) and 94.12% were non-Hispanic or non-Latino. Additionally, 52.94% of participants worked in non-clinical/administrative roles, 17.65% worked in clinical positions, and 29.41% held both clinical and administrative roles. Approximately 47% of participants had masters’ degrees and 30% of the participants had doctoral degrees. Four themes (outlined below) emerged from collected data.

AI can play a huge role in clinical and administrative processes

Participants described that AI could help with both clinical and administrative processes in healthcare organizations. AI was seen as a useful tool primarily because of the following reasons: (I) processes a lot of data faster than what manual efforts can do, and (II) with large volumes of data it can help in the identification of patterns which in turn can result in improvement of health outcomes for patients. To support this claim, one of the participants indicated:

“So, first of all, there’s a lot more data in healthcare than you may find in other industries. I think that the data contains very useful information that can literally help improve people’s lives or save those lives. So, the outcome’s also more significant than you may find.”

Participants indicated that AI has the potential to improve diagnostic methods, treatment planning, and examining the effectiveness of different treatment options/modalities. Being able to combine an organization’s internal records/data with external research and data sources can provide assistance with the evaluation of treatment options and better planning, especially as providers work on clinically complex cases. In addition to positive clinical outcomes, the majority of the participants indicated that AI can be leveraged to build better processes to streamline operations in human resources, supply chain, patient scheduling, and other activities that may help in more personalized care to patients and families. One of the key areas that was highlighted in several interviews addressed the use of AI to improve healthcare reimbursement and payment methods. One of the participants indicated:

“There are payment issues because of lack of documents, incorrect coding, limited knowledge about claim filing, and processing, and we can use AI to correct these inefficiencies.”

This was further supported by another participant, who highlighted how missing or incomplete information (financial data) can lead to incorrect payments. With significant healthcare finance experience, this participant also reflected on how faulty processes and lack of training can lead to more than a billion-dollar deficit which makes it extremely hard for rural healthcare facilities to survive in current economic conditions.

Plethora of issues surrounding AI implementation

Nearly all the participants who were interviewed indicated that current system and policies have not kept up with the recent advances in the field of AI. There is a clear need for creating policies and procedures that need to hold professionals accountable when incorrect decisions are made.

For example, one of the research participants noted “we need a whole set of new rules and laws to guide AI implementation and usage. Plus, there is a need to look at a set of regulations to see what will be done in cases of inappropriate diagnosis or treatment.”

Corroborating the claim another participant said, “inaccuracy of data, potential biases, and any shortcuts due to a large number of patients need to be critically examined from policy point of view.” Lack of education and inability to understand the importance of AI and AI-enabled tools (what is in it for me) were mentioned numerous times by healthcare leaders who participated in the study. While employees had some understanding of AI and how it could be used, there was a need to significantly invest in resources to support adoption of AI. Healthcare leaders are often required to function with limited funding (especially in rural areas); therefore, struggle to allocate resources to train employees was a major challenge and was cited by several participants.

A participant who has been working in healthcare organizations for over a decade indicated “it becomes difficult at times for us to invest in some of the new leading technologies out there and I kind of see AI as part of that. Unfortunately, we do not have resources to do so.”

Another participant with extensive background in rural healthcare indicated “I think AI adoption will require significant investment in training and education and obviously that’s going to be a really big piece moving forward for us likely. I am not sure we will be able to make that decision anytime soon.”

Similar thoughts were expressed by a long-term care facility administrator who mentioned, “we don’t have the ability to invest significant amounts of dollars in time and resources of our teams into new technologies just simply because we’re stretched pretty thin the way it is today.” Another key consideration that could pose major challenges in AI implementation is extremely limited options to share patient data within an organization and between different institutions. The need for change in health laws and/or reasonable modifications in privacy laws to protect patients was highlighted, especially when real patient data is utilized for advanced analytics. More specifically, Health Insurance Portability and Accountability Act (HIPAA) puts several safeguards in place to protect the privacy of patients. It is important to update these regulations on items such as notification of privacy practices, transparency, and safe transfer of health information without fear of external threats. This is particularly important because patient safety is paramount, and it is the providers’ utmost responsibility to protect patients and families against any potential problem.

For instance, a healthcare administrator added “you continue to see more increased call for regulation of AI oversight review. Health care is a very personal private matter to everyone. And we need to respect that and keep that moving forward while introducing these new solutions such as AI. HIPAA laws need to be re-examined to see how we will adapt to the rapidly changing IT environment.” Furthermore, participants also noted use of extensive data systems that are used (by AI) to reach a particular diagnosis. The large amount of data that is needed to create complex AI algorithms come from a particular population. However, these systems may not produce a correct diagnosis when used with underserved and underrepresented patient groups.

Structural support, process support, and buy in from employees for effective AI integration

Participants felt the need to build strong information technology infrastructure with highly skilled professionals who may help with large-scale implementation of AI and AI-enabled tools in healthcare settings. In addition to these professionals, implementation should be carried out in collaboration with researchers, evaluators, clinicians, and people who are responsible for daily operations in healthcare organizations. Efforts should be made to create education programs to enhance people’s comfort level with technology. Celebration of small wins and getting employees (at all levels) and patients on-board may help in building momentum for change.

This was supported by the participant who indicated “it just depends on people’s comfort level with technology, too, which you would need to have some kind of support for those types of people that aren’t as quick to grasp onto it, like, how it can benefit them and how to set up the different AI tools that you have, how you would use them, things like that.”

In order to build support for AI integration, another participant added “… the objective is getting the naysayers or the resistant people on board with them with the implementation. Create committees or working groups whatever we want to call a group of people that you have bestowed some kind of like official advising capacity on and inviting some of the naysayers to that group as well as the AI implementation pro-people. But make sure there’s more pro-AI people than anti-AI people, so that when they’re like voting or everything, then the pro-AI people can help kind of like convince others. And anti-AI people when you’re in a group setting like that and you’re talking about these issues. you tend towards the group.”

One of the challenges that was mentioned numerous times during the interviews was the problem associated with cyberattacks, release of data on the dark web, resistance to change, and how to create buy-in from employees. For instance, one participant noted, “the whole cyberattacks, coupled with we do have some long tenured people that really don’t like change … You know some of them change gradually … And so those are the people that we want at the table to make sure that they’re aware on the front end of where we’re going and have input on the speed with which we make changes.” There was a strong need to build systems to protect such problems from happening, especially issues that could compromise patient privacy.

The need to build processes to demonstrate how AI could benefit clinicians and administrators was highlighted during interviews/discussions. More specifically participants mentioned that these initiatives take a large amount of time (which is hard to spare given the seriousness of their jobs), so one needs to see clear benefits of how implementation of AI-enabled systems will help in building processes that have the potential to solve problems or make decision making easier for them.

Strategic management approach to implement AI

Data analysis clearly indicated that healthcare leaders need to be aware of both facilitators and barriers that may have an impact on AI adoption in their respective organizations. The need to acknowledge fear of the unknown, lack of knowledge/education, and the importance of an effective change management approach was highlighted in multiple interviews.

One participant mentioned, “it is important to build an environment in which people show enthusiasm towards use of AI. I work in a small facility and colleagues are already burnt out, there is no time for anything new. We do realize that we need education on AI and we have to encourage healthcare professionals (clinical and administrative) to use AI without compromising quality of care and patient privacy.”

For clinicians and administrators to understand and implement AI in daily rounding of work, there needs to be a better understanding of current processes, challenges, complexity of the work environment, and roles and responsibilities of team members in the plan of care. It is also important to involve team members from different disciplines and ensure any concerns are addressed. For instance, one of the participants noted:

“If you’re talking to them and involving them, they feel more respected and more heard in a more likely to like, go in that direction if they think that you’ve considered what they think are the major concerns, and come up with, you know, on a plan to address them, or some way that they’re going to be accommodated.”

In addition, this participant further highlighted the importance of acknowledgement, support, and celebrating small or big wins in AI adoption process.

“Celebrate wins and provide acknowledgement. You’re implementing AI to get these, you know business benefits out of it. Say, it’s a reduced cost. So then, once you get those small wins. For example, we saved $1,000 this week. So, I will oftentimes give praise to team members and be like it was instrumental how a team member did a fantastic job getting me access to this thing. It was like, really, without that, we couldn’t have saved this $1,000. So, thank you so much to the team member. … Give them public praise, for like successes with the AI, because it feels nice to be praised for something winning. And then they’re like, oh, okay, actually, this was a positive outcome for our group.”

Discussion

The findings of this study point out significant considerations and implications for practice for healthcare leaders, managers, AI developers, implementers, and facilitators. AI can play a huge role in clinical and administrative processes. Healthcare leaders, managers, implementers, and facilitators need to effectively communicate and accentuate the benefits that AI can bring to healthcare settings for patients, staff, and providers. Patient outcomes can be improved. Better processes can be identified to maximize scarce resources and ensure appropriate reimbursements. While there are projects/published manuscripts that emphasize importance of AI adoption in healthcare, it is important to note that these studies have primarily focused on health professionals who have a clinical role. This study provides an insight into the perspectives of leaders who are generally underrepresented stakeholders such as healthcare leaders and managers (3).

This research suggests that the incorporation of AI in healthcare can lead to high-quality, safe, efficient, and timely care for patients, which can result in significant cost savings for healthcare organizations and providers. AI is an important tool for healthcare leaders and managers to fully understand in order to implement in their organizations. This also aligns with previously published work where researchers have noted that AI applications can lead to reduction in workflow inefficiencies and administrative burdens (7).

There is a plethora of issues that need to be addressed with using AI in healthcare. Prior research shows individuals who were not direct users or recipients of AI tended to have more guarded views compared to direct users or recipients of AI. All stakeholders expressed concern over AI that lacks human oversight in design, development, or implementation with expectations that regulatory standards must be developed and that the benefits of AI cannot be guaranteed (7). This project clearly suggests that there needs to be involvement from all levels (of employees) to enhance AI adoption. This project also highlights the need for policies and procedures to regulate the safe transfer and use of patient data and ensure patient privacy laws are upheld. There needs to be structural support, process support, and buy-in from employees for effective AI integration. Leaders must help employees understand the benefits of AI and reduce fears that employees may have about using AI and AI-enabled tools. Investments must be made in training and education for healthcare workers and providers that provide ongoing support throughout the change and implementation phases. Healthcare leaders, managers, and AI implementers should utilize a strategic management approach to successfully implement AI. AI utilization must be integrated into organizational strategic plans and processes as it will continue to shape and redefine healthcare. It is important to consider both the facilitators and barriers as healthcare organizations embark on their journey of AI adoption in a variety of healthcare settings.

This qualitative research study examined healthcare leaders’ and managers’ attitudes and perspectives on using AI and AI-enabled tools in healthcare and identified facilitating factors that contributed to effective AI adoption and factors that hindered AI adoption in healthcare settings. This research contributes to the growing data available that can aid in the implementation and adoption of AI in healthcare settings.

Limitations

For this study, participants were recruited from various healthcare settings, with efforts made to include individuals with at least 2 years of experience in leadership roles. While this sampling approach offers certain benefits, it is important to acknowledge the potential for researcher bias, particularly in comparison to other sampling methods that may reduce such biases. This could affect the generalizability of the findings. All participants were from healthcare organizations based in the Midwest, U.S. While the results can inform AI adoption in similar institutions, they may not be directly applicable to organizations outside the U.S. or in a global context.

Conclusions

AI combines computer science and linguistics to create intelligent systems capable of analyzing, summarizing, and making predictions with large data sets. If AI and AI-enabled tools are successfully implemented, they can improve healthcare process efficiency, improve patient outcomes, and reduce healthcare costs. This study revealed that healthcare leaders and managers have a critical perspective on the use of AI in healthcare settings, and AI can play a significant role in clinical and administrative processes. For AI to be successfully adopted and implemented, there is a need for structural support and resources to be developed and a need for the regulations and patient privacy issues to be addressed with using AI in healthcare. This research can be useful for healthcare leaders, managers, AI developers, implementers, and facilitators to understand the current state of healthcare leaders’ perspectives toward using AI in healthcare, so the barriers and dependent expectations can be adequately addressed and managed.

Acknowledgments

None.

Footnote

Reporting Checklist: The authors have completed the COREQ reporting checklist. Available at https://jmai.amegroups.com/article/view/10.21037/jmai-24-392/rc

Data Sharing Statement: Available at https://jmai.amegroups.com/article/view/10.21037/jmai-24-392/dss

Peer Review File: Available at https://jmai.amegroups.com/article/view/10.21037/jmai-24-392/prf

Funding: None.

Conflicts of Interest: All authors have completed the ICMJE uniform disclosure form (available at https://jmai.amegroups.com/article/view/10.21037/jmai-24-392/coif). The authors have no conflicts of interest to declare.

Ethical Statement: The authors are accountable for all aspects of the work in ensuring that questions related to the accuracy or integrity of any part of the work are appropriately investigated and resolved. The study was conducted in accordance with the Declaration of Helsinki (as revised in 2013). This study was approved by the Institutional Review Board at Minnesota State University Moorhead (MSUM) (No. 2109683-1) and informed consent was taken from all individual participants.

Open Access Statement: This is an Open Access article distributed in accordance with the Creative Commons Attribution-NonCommercial-NoDerivs 4.0 International License (CC BY-NC-ND 4.0), which permits the non-commercial replication and distribution of the article with the strict proviso that no changes or edits are made and the original work is properly cited (including links to both the formal publication through the relevant DOI and the license). See: https://creativecommons.org/licenses/by-nc-nd/4.0/.

References

1. Badal K, Lee CM, Esserman LJ. Guiding principles for the responsible development of artificial intelligence tools for healthcare. Commun Med (Lond) 2023;3:47. [Crossref] [PubMed]
2. Alowais SA, Alghamdi SS, Alsuhebany N, et al. Revolutionizing healthcare: the role of artificial intelligence in clinical practice. BMC Med Educ 2023;23:689. [Crossref] [PubMed]
3. Hogg HDJ, Al-Zubaidy MTechnology Enhanced Macular Services Study Reference Group, et al. Stakeholder Perspectives of Clinical Artificial Intelligence Implementation: Systematic Review of Qualitative Evidence. J Med Internet Res 2023;25:e39742. [Crossref] [PubMed]
4. Bohr A, Memarzadeh K. The rise of artificial intelligence in healthcare applications. In: Bohr A, Memarzadeh K. editors. Artificial Intelligence in healthcare. Cambridge: Academic Press; 2020:25-60.
5. Davenport T, Kalakota R. The potential for artificial intelligence in healthcare. Future Healthc J 2019;6:94-8. [Crossref] [PubMed]
6. PwC. No longer science fiction, AI and robotics are transforming healthcare. 2024. Available online: https://www.pwc.com/gx/en/industries/healthcare/publications/ai-robotics-new-health/transforming-healthcare.html
7. Scott IA, Carter SM, Coiera E. Exploring stakeholder attitudes towards AI in clinical practice. BMJ Health Care Inform 2021;28:e100450. [Crossref] [PubMed]
8. Terry AL, Kueper JK, Beleno R, et al. Is primary health care ready for artificial intelligence? What do primary health care stakeholders say? BMC Med Inform Decis Mak 2022;22:237. [Crossref] [PubMed]
9. Morrow SL. Quality and trustworthiness in qualitative research in counseling psychology. J Couns Psychol 2005;52:250. [Crossref]
10. Singh J, Eisenschenk T. A Thematic Analysis of the Attitudes and Perceptions of Faculty Towards Inclusion of Interprofessional Education in Healthcare Curriculum. Int J Health Sci Educ 2021;8:1. [Crossref]
11. Braun V, Clarke V. Using thematic analysis in psychology. Qual Res Psychol 2006;3:77-101. [Crossref]
12. Forero R, Nahidi S, De Costa J, et al. Application of four-dimension criteria to assess rigour of qualitative research in emergency medicine. BMC Health Serv Res 2018;18:120. [Crossref] [PubMed]
13. Lincoln YS, Guba EG. Naturalistic inquiry. London: Sage Publications; 1985.