key: cord-0878076-339it3op authors: Ronquillo, Charlene Esteban; Peltonen, Laura-Maria; Pruinelli, Lisiane; Chu, Charlene H.; Bakken, Suzanne; Beduschi, Ana; Cato, Kenrick; Hardiker, Nicholas; Junger, Alain; Michalowski, Martin; Nyrup, Rune; Rahimi, Samira; Reed, Donald Nigel; Salakoski, Tapio; Salanterä, Sanna; Walton, Nancy; Weber, Patrick; Wiegand, Thomas; Topaz, Maxim title: Artificial intelligence in nursing: Priorities and opportunities from an international invitational think-tank of the Nursing and Artificial Intelligence Leadership Collaborative date: 2021-09-01 journal: J Adv Nurs DOI: 10.1111/jan.14855 sha: 044879b96c9d85e7d4bfe0d520a5b7968c71d496 doc_id: 878076 cord_uid: 339it3op AIM: To develop a consensus paper on the central points of an international invitational think-tank on nursing and artificial intelligence (AI). METHODS: We established the Nursing and Artificial Intelligence Leadership (NAIL) Collaborative, comprising interdisciplinary experts in AI development, biomedical ethics, AI in primary care, AI legal aspects, philosophy of AI in health, nursing practice, implementation science, leaders in health informatics practice and international health informatics groups, a representative of patients and the public, and the Chair of the ITU/WHO Focus Group on Artificial Intelligence for Health. The NAIL Collaborative convened at a 3-day invitational think tank in autumn 2019. Activities included a pre-event survey, expert presentations and working sessions to identify priority areas for action, opportunities and recommendations to address these. In this paper, we summarize the key discussion points and notes from the aforementioned activities. IMPLICATIONS FOR NURSING: Nursing's limited current engagement with discourses on AI and health posts a risk that the profession is not part of the conversations that have potentially significant impacts on nursing practice. CONCLUSION: There are numerous gaps and a timely need for the nursing profession to be among the leaders and drivers of conversations around AI in health systems. IMPACT: We outline crucial gaps where focused effort is required for nursing to take a leadership role in shaping AI use in health systems. Three priorities were identified that need to be addressed in the near future: (a) Nurses must understand the relationship between the data they collect and AI technologies they use; (b) Nurses need to be meaningfully involved in all stages of AI: from development to implementation; and (c) There is a substantial untapped and an unexplored potential for nursing to contribute to the development of AI technologies for global health and humanitarian efforts. Artificial intelligence (AI) is defined as '… the science and engineering of making intelligent machines, especially intelligent computer programs' (McCarthy, 1956) . Increasingly sophisticated AI such as personalized advertisement and self-driving cars are revolutionizing a diverse range of professional sectors. In healthcare, AI is being adopted to aid healthcare professionals deliver high-quality care more efficiently and equitably. For example, AI can support less experienced healthcare professionals who may have fewer resources to still deliver high-quality care through learning from other's experiences (e.g. identification of rare disease symptoms through massive database searches) (Schaefer et al., 2020) . In the context of nursing, examples of applications of AI demonstrate the potential impact that the use of these technologies can have in nursing practice. For example, speech recognition technologies can speed up and enhance nursing documentation (Fratzke et al., 2014; Monica, 2018) and machine learning has been used to develop a tool to aid nurses in using standardized technologies, by automatically suggesting the most relevant terms to be used based on the text written by the nurse (Moen et al., 2020) . Other applications include text mining where AI technologies are being used to mine millions of nursing notes to identify patients with fall history (Topaz, Murga, Gaddis, et al., 2019) or drug and alcohol use disorders (Topaz, Murga, Bar-Bachar, et al., 2019) , to support care planning and patient risk detection. Similarly, machine learning, specifically deep learning, has been experimented to predict pain sensation and physical deterioration for acute critical conditions (Pruinelli et al., 2018; Pruinelli, Stai, et al., 2019; Pruinelli, Westra, et al., 2019) . In the near future, AI technology will be able to help nurses provide precise and individualized evidence-based care that meets patients' goals and priorities. AI technologies will also help nurses integrate different types of relevant data (e.g. environmental, genomic, health data, socio-demographics) strengthening nurses' capacity to provide multifaceted care. Moreover, a recent scoping review has highlighted that much of the research on AI in healthcare has focused on secondary and tertiary care, leaving still considerable opportunity to explore nurses' use of AI in primary care (Abbasgholizadeh-Rahimi et al., 2020) . From these examples, it is clear that nurses are not exempt from the proliferation of AI in healthcare systems, with AI often touted as tools that can transform the provision of health care and improve health outcomes (Clancy, 2020) . The dynamics between AI and nursing has yet to be critically interrogated. This is despite nurses being the largest group of healthcare professionals internationally (International Council of Nurses, 2017) , and, by sheer volume of the workforce, nurses likely being the healthcare professionals who are most exposed to new AI technologies. Recognizing the mixed and complex, albeit limited, perspectives about AI in nursing, the Students and Emerging Professionals Special Interest Group in the International Medical Informatics Association organized the first international invitational expert think-tank workshop of the Nursing and Artificial Intelligence Leadership Collaborative (NAIL), titled 'Artificial intelligence in nursing: social, ethical and legal implications'. The 3-day think-tank (October 23-25, 2019) held at the Brocher Foundation in Switzerland invited 19 interdisciplinary participants from Canada, Finland, Switzerland, the United States and the United Kingdom. The NAIL Collaborative comprises experts in AI development, AI implementation, nursing, and biomedical ethics, AI in primary health care, AI legal aspects, philosophy of AI in health, nursing practice, implementation science, high-level policymakers for healthcare institutions and international informatics groups, a representative of patients and the public, and Chair of the ITU/WHO Focus Group on Artificial Intelligence for Health. Activities included a pre-event survey to elicit attendees' initial perspectives of AI in nursing, presentations by all invited attendees on their areas of expertise as related to AI and/or nursing and working sessions with attendees, to delve into in-depth discussions. Implications Invitational Think-Tank In this paper, we summarize and highlight poignant points of discussion from the think-tank. These include central issues, priorities and key insights associated with AI technologies in nursing in the context of current discourses. We conclude the paper with actionable recommendations on issues related to the safe development, implementation and adoption of AI in nursing including the ethical, legal and social implications of AI technology. In nursing, advancements in AI technologies are often received with cautious excitement (Erikson & Salzmann-Erikson, 2016; Robert, 2019; Skiba, 2017) . On the one hand, the use of AI presents the potential for optimizing nursing care delivery by alleviating mundane and time-consuming and burdensome tasks that do not require specialized nursing skills or knowledge (e.g. managing hospital room logistics, calling housekeeping for cleaning and restocking room supplies) and freeing up time for nurses to spend on direct (versus indirect) patient care. On the other hand, the use of AI concurrently introduces the risk for unintended consequences that can have a potential negative impact on the nursing profession. AI technologies have the potential to propel nursing capabilities and enable nurses to provide more evidence-based and personalized care to their patients. AI technologies have the potential to support responsive and evidence-based nursing practice through the provision of cognitive insights and decision support, for example, through visualization of patient trends that can provide insights for both immediate patient care as well as long-term planning and management. Proponents of AI also point to the potential for AI to free-up time for healthcare professionals to dedicate in improving the relationships with patients (Topol, 2019) . Indeed, the time that can be freed up for nurses can be spent on fostering relational care, supporting nurses' ability to develop broader insights into the contexts of patients' health. Moreover, time that is freed up for nurses can be spent on engaging with recent research and supporting up-to-date knowledge of the evidence to support practice, activities that are among the most common to be put aside for lack of time and opportunity (Duncombe, 2018) . Better relationships with patients and up-to-date knowledge of the evidence, taken together, support nurses' ability to provide personalized care that considers a holistic view of patients. Along with the potential or positive outcomes, AI technologies can have unintended consequences that can have a potential negative impact on the nursing profession and on the main aims of nursing practice. For example, there exists the risk for AI to perpetuate or systematically embed existing human biases into systems (Benjamin, 2019) , such as a recent case where a clinical decision algorithm introduced racial bias by prioritizing care for less sick white patients over sicker Black patients in the United States (Obermeyer et al., 2019) . Beyond impacts on clinical and health outcomes, AI in nursing could also exacerbate the push towards market-driven goals of efficiency. There exists a very real potential to instead reallocate newly freed-up time towards increasing the volume of patients and tasks assigned to nurses. Hence efficiency goals (i.e. quantity of care) run the risk of eclipsing the opportunities that the use of AI in health systems are meant to create (i.e. quality of care). Such negative impacts are not inevitable. For instance, AI also has the potential to make visible and remove human bias and improve decision making (Leibert, 2018) , for example by discovering and quantifying the impact of taken for granted variables such as sex, gender, ethnicity, or race (while we recognize that race has no scientific meaning, experiences of racism have clear links to health outcomes), for which our understanding of impacts are emergent (Davenport & Kalakota, 2019) . Ensuring the best possible consequences from AI for nursing will depend on which values and priorities end up guiding the development of AI tools, and whether they implemented with an adequate understanding of both their potentials and limitations. Placed in nurses' hands, unintended consequences of using AI tools can be direct and serious, reflecting the same concerns discussed by O'Keefe- Mccarthy (2009) in their classical discussion of the mediating role of technology in the nurse-patient encounter and the subsequent effects on the moral agency of nurses. Given the potential magnitude of the impact of AI tools, there is an ethical imperative for nurses to have a minimum basic understanding of how these tools come to be developed, what informs them, and the implications of using such tools on their clinical judgement and practice. The responsibility of having a minimum understanding of AI that all nurses must develop is arguably no different from the requirement of nurses to have a basic understanding and competency in the use of any type of new technology or tool that they use in their practice. Notwithstanding these important implications of AI for the nursing profession, there is a growing, but still a limited critical discourse in the nursing literature (Brennan & Bakken, 2015; Linnen et al., 2019) . In the sphere of nursing education, addressing AI remains, largely, absent. Nursing curricula continue to struggle with incorporating basic nursing informatics competencies as part of basic nursing education (Ronquillo et al., 2017; Topaz et al., 2016) , which will become more worrisome given the growing interest in using AI tools in health systems. In other words, there is the potential that the challenges that nurses currently face regarding the effective use of and potential for leading innovations in health information technologies can be further compounded if a gap in AI knowledge is added to existing gaps in basic health informatics knowledge. The following represent a summary of the discussion points identified in the NAIL Collaborative think-tank discussions, framed as pressing priorities for the nursing profession. Each priority point is introduced with the identification of a current gap in understanding or use of AI in relation to nursing practice. For each identified gap, we propose strategies and opportunities--with implications for nursing practice, education, research and leadership-that can be pursued to ensure the appropriate and safe use of AI in nursing and enable the nursing profession to use AI tools to optimize health outcomes. Gap: Nurses are the group of healthcare professionals who generate the most data in health systems, as they complete the most documentation (Collins et al., 2018) . Nurses play an important role in collecting data that might be eventually used by AI tools, as evidenced by work that has linked the nature and patterns of nursing documentation practices with patients' mortality (Collins et al., 2013) . There nevertheless appears to be limited understanding of the link between nursing documentation and how these documents may be used for purposes beyond immediate clinical decision making, administrative reporting and keeping a legal record as taught in basic nursing education. While understanding these aspects of documentation has been sufficient to inform nursing practice in the past, we argue that nurses should also understand the relationship between their clinical documentation and AI. For one, understanding the nature and quality of data that are collected and documented as part of the nursing practice, can and do, directly inform AI tools. Also, AI-based clinical decision support has various levels of uncertainty that requires clinician interpretation (Shortliffe & Sepulveda, 2018) . When deciding to follow an AI-based recommendation, nurses serve as the last line of evaluation for the appropriateness of an intervention (Eisenhauer et al., 2007) . Moreover, a significant current challenge is that many nursing educational programmes-both in entry-level nursing education and continuing education of professional nurses-do not have enough expertise in teaching health informatics and AI technologies (Cummins et al., 2016; Mantas & Hasman, 2017) to effectively address this gap in AI understanding. To bridge the educational gap, there is a need to develop a curriculum with 'minimum AI in nursing competencies', a set of domains and concepts that all entry-level nurses should receive as part of their basic nursing education (Michalowski, 2019) . Some organizations, such as the American Association of Colleges of Nursing (AACN), are moving to a competency-based education with a technology domain crossing over all domains due to the current need for this topic in all levels of nursing education. Similar efforts concurrently need to be made to support the development of these competencies among practising nurses, as well as nurse leaders (Pruinelli et al., 2020) , where this material can be delivered through continuing education initiatives. Graduate nursing education also would benefit from the creation of opportunities for advanced AI education as well as the formation of sub-specializations in AI under health informatics programs. Specific recommendations are outlined in the summary Table 1 towards ensuring that a curriculum with 'minimum AI in nursing competencies' can be met, with the goal of having all nurses hold basic knowledge and competence related to AI use in nursing. Gap: Currently, nurses are often end-users of technologies that incorporate AI (e.g. advanced clinical decision support) rather than collaborators in development. As such, there are other calls for nursing: to take the driver's seat in determining which aspects of nursing care can be delegated and to be key actors in introducing AI technologies in health systems (Pepito & Locsin, 2019) . In a clinical context, the AI development lifecycle must start with a thorough understanding of the clinical question and clinical work-flows, as these ultimately shape the successful use and subsequent impact of these technologies on patient and organizational outcomes. AI development teams should be interdisciplinary, including nurses, to ensure that contributions of computer science and engineering members of teams are grounded in clinical realities of the provision of patient care. Nurses' contributions to all stages of the AI development lifecycle become crucial when recognizing the intertwining of the consequences accompanying the use of AI in nursing (both positive and negative) with the foundational underpinning of the nursing profession as being concerned with beneficence towards patients, communities, and populations, and advocacy for social justice (Paquin, 2011; Stievano & Tschudin, 2019; Wilmot, 2012) . Patient, family, and community advocacy and the promotion of person-centred care comprise foundational functions of the nurse. As such, nurses are uniquely positioned to propose how the impact of AI should be measured in terms of nursing and patient outcomes. It is through active participation in all aspects of the AI development lifecycle (Matinolli et al., 2019 ) that unique insights from nursing can contribute to the thoughtful development and use of AI that optimize potential benefits and minimize potential negative consequences for patients, communities, populations, healthcare systems and the nursing profession. Nurses need to be meaningfully (rather than tokenistically) involved and contribute as key members of AI development and implementation teams in health systems. While nursing can contribute in many ways across the AI development lifecycle, we have identified three potential distinct and important informant/communicator roles that can be contributed by nursing. These include: (a) delineating clinical problems; (b) serving as intermediaries between the clinical and technical spheres; and (c) incorporating features of relational practice (Dykes & Chu, 2020) . Nurses' expertise and deep familiarity with working closely with patients should be tapped into, to better delineate clinical problems that AI technologies aim to address. For example, when predictive algorithms are being developed from clinical data, nurses can contribute with practice-based perspectives to technical teams (often consisting of engineers, computer scientists, user interface design experts, etc.) to understand why some data elements are missing or incomplete (e.g. poor documentation of social risk factors) (Navathe et al., 2018) and offer potential strategies to address these shortcomings. Closely related is the potential for nurses to serve as key intermediaries between technical experts developing solutions and nurses as clinical end-users (Dykes & Chu, 2020) . These two groups speak very different professional languages and nurses educated in AI concepts are perfect for bridging this vocabulary gap. Finally, nursing expertise in relational practice (i.e. understanding and focus on the quality of human relationships) represents a unique strength to contribute to the AI development lifecycle. The primacy of nurse-patient relationship as a defining priority of nursing can contribute greatly to AI applications in robotics and elsewhere. Nurses can provide insight into the value of empathy and human touch, the role these concepts play in therapeutic relationships (Dobson et al., 2002; Kerr et al., 2019) , and the dynamics between AI technologies and human relationships that need to be considered throughout the AI development lifecycle. Gap: There is a limited recognition of the relationship between AI technologies and the nursing profession as related to the contribution towards global (and oftentimes national) health and humanitarian efforts. There are numerous movements focused on the use of 'AI for good' in the academic, non-profit and industry spheres (e.g. Google's AI for Social Good, Google AI, 2018; Microsoft's AI for Good, Microsoft, 2020; AI for Good Foundation, AI for Good Foundation, 2015) , AI for Good Global Summit (International Telecommunications Union, 2020), advocating for the use of AI to benefit humanity and address difficult social, economic, environmental, health and humanitarian challenges around the globe. Despite the potentially significant impact of AI technologies on nursing work, there remain to be efforts from nursing relating to the notion of using AI4GN, or the use of AI technologies in nursing to achieve a greater good for the profession and for populations. Efforts that recognize the contributions that fall in AI4GN can include leveraging the unique positionality of nurses in healthcare systems towards advocating for the inclusion of equity and social justice considerations in the development and implementation of AI technologies in health systems. Nurses are health professionals who spend the most time with patients and are often referred to as the most trusted profession (Colduvell-Gaines, 2019). Nurses are well situated to identify potential biases in data collection (e.g. decontextualized data that does not consider the impact of systemic structures on individuals) which can lead to the embedding of these biases in the AI tools developed. As well, nurses are ideally situated to identify ethical concerns relating to the implementation and use of AI tools (e.g. highlighting the problematic nature of using facial recognition tools) and instances that can exacerbate existing inequities and cause potential harm among particular groups and populations. For example, a recent study highlights the greater likelihood of digital data being collected and shared from children's use of apps when those children come from lower-education backgrounds (Zhao et al., 2020) . In the context of the healthcare system, this translates to an important facet of nursing education that needs to be developed and embedded as a routine component of a holistic nursing assessment and intervention. Namely, this comprises educating patients and families about digital literacy, digital privacy, laws and regulations on data collection and protection of digital health data and how these all relate to the AI tools that are used in healthcare provision. AI technologies will change the profession of nursing. AI technologies can serve as important tools to support the contribution of nurses towards higher level aims of evolving the nursing profession and improving population and global health. If nursing takes a proactive role in addressing these above-mentioned priorities, AI has the potential to enhance and extend nursing capabilities. In return, nursing has much to contribute to the development of AI systems that leverage nurses' strengths and expertise in relational practice and patient advocacy, towards the development of AI that considers patients with a more holistic view. It is important to note that all priority areas discussed in this paper are necessarily linked. They do not each sit on their own but inform a broad purposeful approach to empowering nurses in their active involvement in all aspects of AI in health care. We argue that nurses have a responsibility to know about the AI technology they use, as has been stated from an industry perspective (McGrow, 2019) . Moreover, there is a great opportunity for AI tools to support nurses' problem-solving abilities and identify solutions for optimizing care provision (Cato et al., 2020) . There is nevertheless a need for support from health systems stakeholders and high-level decision-makers to facilitate the ability of the nursing profession to address these identified priorities. The priorities presented in the paper are summarized in Table 1 , alongside a list of specific recommendations based on the strategies and opportunities outlined in this paper. Current works and future directions on application of machine learning in primary care AI for Good Foundation. AI for good foundation. AI for Good Foundation Race after technology: Abolitionist tools for the new Jim code Nursing needs big data and big data needs nursing Transforming clinical data into wisdom: Artificial intelligence implications for nurse leaders Artificial intelligence and nursing: The future is now Nurses rank most honest profession 17 years in a row Relationship between nursing documentation and patients' mortality Quantifying and visualizing nursing flowsheet documentation burden in acute and critical care Nursing informatics certification worldwide: History, pathway, roles, and motivation The potential for artificial intelligence in healthcare Touch in the care of people with profound and complex needs: A review of the literature A multi-institutional study of the perceived barriers and facilitators to implementing evidence-based practice Now more than ever, nurses need to be involved in technology design: Lessons from the COVID-19 pandemic Nurses' reported thinking during medication administration Future challenges of robotics and artificial intelligence in nursing: What can we learn from monsters in popular culture? Enhancing nursing practice by utilizing voice recognition for direct documentation Retrieved from https://ai.google/social-good International Council of Nurses. Who we are. International Council of Nurses Neurophysiology of human touch and eye gaze in therapeutic relationships and healing: A scoping review AI will improve healthcare and cut costs -If we get these 4 things right Ripe for disruption? Adopting nurse-led data science and artificial intelligence to predict and reduce hospital-acquired outcomes in the learning health system IMIA educational recommendations and nursing informatics Health and medical device development for fundamental care: Scoping review What is artificial intelligence Artificial intelligence: Essentials for nursing Artificial intelligence and nursing: The what, the why, and the how Brocher Workshop on Assisting nurses in care documentation: From automated sentence classification to coherent document structures with subject headings Using EHR voice recognition to improve clinical documentation, usability. Xtelligent Healthcare Media, LLC Hospital readmission and social risk factors identified from physician notes Dissecting racial bias in an algorithm used to manage the health of populations Technologically-mediated nursing care: The impact on moral agency Social justice advocacy in nursing: What is it? How do we get there? Creative Nursing Can nurses remain relevant in a technologically advanced future? Nursing management minimum data set: Cost-effective tool to demonstrate the value of nurse staffing in the big data science era An applied healthcare data science roadmap for nursing leaders: A workshop development, conceptualization, and application A holistic clustering methodology for liver transplantation survival A likelihood-based convolution approach to estimate event occurrences in large longitudinal incomplete clinical data A multi-dimensional general health status concept to predict liver transplant mortality How artificial intelligence is changing nursing Competency recommendations for advancing nursing informatics in the next decade: International survey results Theuse of machine learning in rare diseases: a scoping review Clinical decision support in the era of artificial intelligence Augmented intelligence and nursing The ICN code of ethics for nurses: A time for revision Extracting alcohol and substance abuse status from clinical notes: The added value of nursing data Mining fallrelated information in clinical notes: Comparison of rule-based and novel word embedding-based machine learning approaches Advancing nursing informatics in the next decade: Recommendations from an international survey The Topol review: preparing the healthcare workforce to deliver the digital future We acknowledge Haley DeForest and Rajbinder Nibber for their support in developing and organizing the think tank events. RN was supported by the Wellcome Trust (213660/Z/18/Z) and the Leverhulme Trust through the Leverhulme Centre for the Future of Intelligence. SAR receives salary support from a Research Scholar Junior 1 Career Development Award from the Fonds de Recherche du Québec-Santé (FRQS), and her research programme is supported by Natural Sciences and Engineering Research Council (NSERC) Discovery grant no. 2020-05246. KC was supported by Dr. Bakken's Center 'Reducing Health Disparities Through Informatics' NIH grant no. 5T32NR007969-14. Funding and hosting of the Artificial Intelligence for Nursing: Ethical, Legal and Social Implications think tank was provided by Fondation Brocher in Hermance, Switzerland. The Brocher foundation mission is to encourage research on the ethical, legal and social implications of new medical technologies. Its main activities are to host visiting researchers and to organize symposia, workshops and summer or winter academies. More information on the Brocher Foundation program is available at www.brocher.ch. The data that support the findings of this study are available from the corresponding author on reasonable request.