Artificial intelligence in healthcare: medical technology or technology medical?

The fourth industrial revolution, beginning in the early 21st century, is a concept which asserts that emerging technologies will reshape all aspects of human society [1]. One of these, artificial intelligence (AI), has been described as the most profound technology humanity will ever work on [2]. Whilst AI has already had great impact in the world at large, it is yet to have the same influence in healthcare [3-5]. However, anaesthesia, critical care, peri-operative medicine and pain management are at an inflection point where our clinical practice is soon likely to feel the effects of AI [6-8]. The world's first regulatory approval for an AI-based medical device was granted by the US Food and Drug Administration (FDA) in 1995 [9]. Since then, the FDA has approved a further 949 AI-based medical devices (at the time of writing) [9]. Only eight of these have been approved by an anaesthesia panel, illustrating that our specialty lags others, such as radiology (723) and cardiology (98). Most approvals (729/949, 76.8%) occurred after the start of 2020, which demonstrates the increasing pace of development in this field. This recent proliferation of AI is also reflected in anaesthesia publications, as shown by a recent scoping review of AI in ultrasound-guided regional anaesthesia [10]. Given the rapid increase of activity in this nascent field, with impending clinical impact there is an urgent need to establish greater medical leadership in healthcare AI to ensure appropriate development, evaluation and implementation in clinical practice [11]. Here we discuss the current status quo, learnings from our experience in healthcare technology (Table 1), and present the case for greater integration of physicians in medical technology fields. Learning 1. Digital health companies fail to employ physicians with a voice in product development at their peril Companies often form with strong commercial and technology leadership, resulting in a development/engineering culture 'back of house' and a sales culture 'front of house'. With little clinical leadership, companies may produce products which do not fit the clinical need and/or are unable to scale due to a failure to meet user interface, experience and functionality expectations. Learning 2. Clinical digital transformation is a change management exercise, not an IT project If the value proposition speaks to the executive or manager, rather than the physicians, adoption and deployment will face recurring challenges. If the clinical community has little or no voice in the transformation process, it can take longer, be more costly and be less effective. Learning 3. Healthcare can learn from other industries Technology and specialty profession synergy are abundant in other areas. Learning 4. It's not a conflict for physicians to take a role in commercial digital health entities Fewer privacy walls exist between health and human service sectors in Nordic countries (e.g. Finland). Health service providers have ready access to data regarding housing and socioeconomic status, as these are key determinants of health and can influence policy/treatment decisions. Learning 5. Tech is now less technical The highly technical nature of software development has previously been a two-way barrier to physician engagement in healthcare technology. 'Techies don't get healthcare' and 'physicians don't understand programming' principles or constraints. This is changing; the advent of low-code and no-code technologies increasingly allows physicians to engage whilst focusing on what they know best (e.g. functionality, clinical workflow, interface design). Learning 6. Physicians need new skills to fully participate in the digital health sector Medical training prepares physicians well for clinical practice but places less emphasis on skills required for a career in healthcare management. Lack of fluency and experience in business, technology and management increases the challenges for physicians when participating at the cross-functional executive tier. We must equip physicians with the skills and experience to lead. In this time of rapid technological expansion and growth, investment is gravitating towards AI [12], such that resources in this sector far surpass those available to academia. For example, UK research funding for anaesthesia-related work is typically < £100,000 (US$ 127,000, €120,000) per grant. In contrast, Google's (Mountain View, CA, USA) acquisition of DeepMind cost in the region of £400 million (US$500 million, €475 million) [13]. Understandably, technology and industry leaders do not want to relinquish control of their resources unnecessarily by inviting physicians to instruct them on how to allocate their time and capital. This would be akin to a surgeon surrendering the care of their patient to the radiologist interpreting the imaging or the anaesthetist providing peri-operative care. However, without active clinical participation in the development of medical devices, technology-led organisations may ask the wrong questions, focus their efforts on developing a solution that addresses a problem of lesser importance or fail to fully understand how that solution best addresses the problem and is implemented [5]. For example, airway assessment in nurse-led pre-operative assessment clinics [14, 15] or the need to predict and manage acute postoperative pain more effectively [16] are current challenges that physicians will have unique insights into; incorporating clinical considerations into any AI-based solution will be essential to ensure clinical utility and improvement in patient-centred outcomes. However, physicians have busy schedules with multiple competing demands for their time and effort. Unless the individual is employed by an academic institute or has funding to protect non-clinical time, clinical priorities often supersede. In addition, technology does not interest all physicians and many are not motivated to accelerate its development. Furthermore, the skills learnt during medical training are often markedly different to those required in medical technology (Table 1). As a result, a substantial contribution to this field is not possible by all physicians. We have also encountered an undercurrent of reticence (and perhaps even fear) in the clinical community in relation to remunerated industry work. Physicians may believe they can only be impartial and focus solely on patient benefit if not paid for non-clinical and non-academic activities, sometimes viewing industry conflict of interests through a negative lens. However, conflicts of interest (both personal and professional) arise in numerous aspects of our routine clinical practice, and knowledge gained in other areas may result in additional insights which can lead to patient benefits (Table 1). Potential conflicts must be managed actively and declared on any publicly-facing work, but it is often easier to define and state paid commercial work rather than nuances of clinical and academic practice, where conflicts of interest may not even be fully recognised. Professional experience has led us to conclude that, with the prevailing arrangement of technology leadership, medical technology companies encounter numerous pitfalls (Table 1). Of fundamental importance is that such commercial organisations may produce a device which struggles for product fit [5]. In addition, companies may place excess emphasis on achieving technical steps (e.g. regulatory approval) and deprioritise clinical engagement or acceptance. This is analogous to the challenges seen with engaging pharmaceutical companies and pharmacology researchers to focus on clinically important outcomes (e.g. remimazolam and target-controlled infusion models [17]). The clinical viewpoint, and that of the patient, also helps ensure that studies target clinically important differences in outcome measures (e.g. in postoperative pain [18] or quality of recovery [19]) and that these measures are important to patients. In the absence of clinically important outcomes, companies may deliver a product to market that garners little clinical interest. Without clinical guidance, greater collaboration between industry and academia can lead to its own challenges. Universities and academics may not have the expertise or inclination to commercialise ideas, and metrics such as journal publication and conference presentation may take precedence. Progress in academia is often not at the speed required by industry, which can require rapid return on investment. Ultimately, physicians are rarely able to deliver technology solutions to a professional standard; AI-based medical devices must therefore be developed with significant technology input. We also believe that commercial involvement is imperative as this sector brings a motivation to achieve the end goal in order to realise a financial return (and hence attract future investment). Furthermore, industry boasts an abundance of expertise required to realise the implementation of a medical device in clinical practice that can be lacking in clinical and academic fields (e.g. knowledge of product management, regulatory approval and marketing processes). We propose that physicians assume leadership positions which contribute to identifying the areas requiring development of AI-based solutions, overseeing the clinical aspects of development, and steering the evaluation and implementation of these devices (Table 1). Anaesthetists are particularly well-placed to adopt such a role as they work with patients and other healthcare professionals in many areas of clinical practice, regularly interface with technology and often assume leadership positions in the clinical setting. Furthermore, physicians with an interest in AI should feel comfortable pursuing this course, as a medical career of ≥ 30 years allows time to gather expertise in both domains. Clearly, technology experts will continue to contribute to all these areas and will undoubtedly deliver the technical component of the solution. If technology specialists are the heart of medical AI, physicians can be the brain – guiding where activities and resources are focused. Physicians know the problems encountered in clinical practice and can help industry identify opportunities. Their understanding of clinical process means AI-based devices can be designed to ensure the solution fits the problem, within the natural clinical workflow. When this needs to be disrupted, physicians are likely to be the best professional group to effect change in clinical practice. Evaluation of medical devices – for regulatory approval studies and to generate data which support claims – may also be best led by physicians, to make sure regulation/medical device evaluation is clinically-centred and that these assessments are consistent with clinically-relevant standards [20]. Finally, physicians can address the clinical audience more easily, by publication in journals, presentation at conferences and through their professional networks, thereby enabling companies to reach the desired audience. It is important that medical bodies overseeing standards of practice and professional membership societies are involved actively in this field. This can mitigate undue influence by polarised opinion of individuals, who may have a specific clinical interest and divert industry efforts down a suboptimal route for the wider clinical community. The involvement of such organisations also carries greater authority in the clinical domain and can help to engage higher level organisations or activity (e.g. influencing government and policy). Medical professional bodies and societies can generate consensus, set standards and produce position statements, as well as involving patients and hospital/organisational managers (Table 2). This may help direct academic and industry effort to high priority areas, as determined by the wider body of medical opinion. This approach can also enable greater co-ordination of action between different geopolitical regions and promote interdisciplinary collaboration. Physicians should not view paid industry work as compromising their clinical credibility or commitment to patients. Academia should be careful not to view industry relationships as a means to secure financial investment. Equally, industry should not consider physicians and academia purely as consultants to call on when the need is felt – industry must view them as leaders who influence strategy and company direction, and attract them to such positions. Through medical leadership positions and career structures in companies, physicians can contribute to research and development; product management; oversee studies (to gain data for regulatory approval studies and support marketing claims); manage medical affairs; and enable integration with the clinical community. Similarly, medical organisations could create positions for industry experts, such as industry members/representatives of specialist societies and learned bodies. In such roles, industry experts can advise on the technical and commercial feasibility of ideas, help prioritise activities and facilitate the realisation of potential opportunities. Physicians are optimally placed to bridge the gap between clinical practice, academia and industry, to create better engagement on all sides. Part of this is to establish a better means of communication and information transfer between these domains. For example, few industry individuals are likely to read this article and few in the clinical readership will hold positions/influence in medical device companies. However, as publication is an important medium of communication for clinical and academic communities, it is imperative that we engage industry partners. Involving industry as a key stakeholder in conferences (rather than simply view from the sidelines through sponsorship) is also likely to be of interest, particularly for an AI-centred meeting. We believe the clinical community has been too passive in this field, which has resulted in industry leadership – 'technology medical'. We must change this relationship and join our industry colleagues to guide its development – 'medical technology'. The Association of Anaesthetists has established an AI Workstream to promote greater collaboration between the two sectors. This can be beneficial for all involved, including the patients who have potentially the most to gain with improvements in healthcare technology. Just as the fourth industrial revolution has blurred the boundaries of physical, digital and biological worlds [8], we must foster greater integration of clinical, academic and industry sectors to achieve the optimal implementation of AI in anaesthesia, critical care, pain and peri-operative medicine. Importantly, these efforts must be guided by medical leadership, to achieve the best outcomes for physicians, commercial entities and patients. JB is an employee of GE Healthcare, a medical technology company producing medical devices. He has previously served as a Senior Clinical Advisor for Intelligent Ultrasound, receiving research funding and honoraria, and acted as a consultant for AutonomUS. He is a co-opted member of the Association of Anaesthetists Heritage, Research and Innovation Committee; the Safety, Standards, Environment and Sustainability Committee; and the Association AI Workstream. SK is the Chief Medical Officer (Australia and New Zealand) of Microsoft and a board director at Innowell. MW is the Editor-in-Chief of Anaesthesia, a member of the Association of Anaesthetists Heritage, Research and Innovation Committee and a Board Member. No external funding or other competing interests declared.