Digital nursing 1: exploring the benefits and risks of artificial intelligence

Introduction

The ministerial forward to Joshi and Morley’s (2019) report, published by NHSX, provided one key reason to be excited about artificial intelligence (AI) in healthcare: “put simply, this technology can make the NHS even better at what it does: treating and caring for people”. AI is exciting, but what is it, exactly? This first in a series of articles about digital healthcare will discuss the benefits and risks of AI.

Increasingly, nurses are encouraged to lead and shape emerging digital technologies, to ensure the changes made are fit for purpose and ward against unintended consequences, such as increased workloads, dehumanised care and the exclusion of already marginalised groups of people. The areas that could be improved, or even transformed by AI, according to Joshi and Morley’s (2019) report, include:

• Diagnostics – AI image recognition tools (driven by data) can support clinicians’ judgement by reading images such as mammograms, brain scans and eye scans. Diagnosis could become faster and treatments more accurate;
• Planning and management of NHS resources – AI can make complex calculations, based on huge amounts of data, that make it possible to predict, as an example, how much blood plasma a hospital will need or the likelihood of people missing their outpatients’ appointments;
• Medicines – AI comparisons of drug compounds, for example, can achieve breakthroughs in weeks that would take human researchers years.

Despite this, it is difficult to approach, let alone lead on, technological advancements if nurses do not understand them, and AI may feel too big to grasp at times.

There are many definitions in the ever-growing literature released about AI. Joshi and Morley’s (2019) report suggested that one of the most useful definitions in the field of healthcare is also the oldest; they explained that it dates from a research project in 1955 and stated that AI is “the science of making machines do things that would require intelligence if done by people”.

A publication by The King’s Fund – namely, Mistry (2020) – gave a more detailed, but still straightforward explanation, stating that AI “is an umbrella term encompassing a number of different approaches where software replicates functions that have, until recently, been synonymous with human intelligence. This includes a wide spectrum of abilities such as visually identifying and classifying objects [and] converting speech to text and text to speech”.

The origins of AI go back decades, so why are we hearing so much about it now? One reason is that recent developments in applied mathematics and computer science have made computers much better at reading patterns in large amounts of complex data, releasing AI’s potential (Mistry, 2020). The possibilities of AI are being further expanded by machine learning, which has been defined by Mistry (2020) as “a type of artificial intelligence that enables computers to learn without being explicitly programmed, meaning they can teach themselves to change when exposed to new data”.

AI in nursing

Most health staff still lack direct experience with AI technologies, as is highlighted in Nix et al’s (2022) report, developed by NHS AI Lab and Health Education England (Box 1). However, the increasing use of AI technologies in nursing, such as providing information for advanced clinical decision support, is thought to be inevitable (Booth et al, 2021; Robert, 2019).

AI-enabled decision support systems potentially provide numerous benefits; for example, they have already dramatically improved the detection of sepsis (Horng et al, 2017). However, there are also risks, as AI is only as good as its data. Nix et al’s (2022) report warns that confidence in artificial intelligence (AI) is not always desirable when using it for clinical decision making, and nurses need to recognise when to balance it with other sources of clinical information (Box 2).

AI systems that evolve themselves may reflect or reinforce societal biases (for example, racial biases) and other inequities present in the data (Obermeyer et al, 2019; Gianfrancesco et al, 2018). It is important for nurses to be involved in innovations such as AI to make sure they are developing systems in line with ethical frameworks and to advocate for patient involvement (Booth et al, 2021). There is also a risk that AI systems that perform extremely well in controlled conditions will be less impressive in the real world, and there are unanswered questions about their safety and cost effectiveness in healthcare settings (Maguire et al, 2021).

The NHS is already using AI and machine learning, at a population level, to help identify older people in local areas who are at risk of frailty and adverse health outcomes; one example of this is the Electronic Frailty Index, which draws on data that is routinely recorded by GP practices (NHS England, 2017).

Predictive analytics in electronic patient records should also, increasingly, help doctors and nurses to diagnose and treat the individual patient in front of them (HEE, 2019). AI has also played a key role in informing the government’s response to the coronavirus pandemic: the launch of the NHS Covid-19 Data Store by NHSX has aided the analysis of vast amounts of data to:

• Reveal how the virus is spreading;
• Ascertain how the NHS is coping;
• Suggest the most effective current and future interventions (Maguire et al, 2021; Gould et al, 2020).

AI is also central to the government’s new digital health and social care plan (Department of Health and Social Care and NHS England, 2022). This includes using AI to develop “new diagnostics capacity to enable image-sharing and clinical decision support… These technologies support testing close to home, streamlining of pathways, triaging of waiting lists, faster diagnosis and levelling up under-served areas”

Robotics in nursing

With the development of smaller and more-sophisticated electronic components, robots embedded with AI will likely become more widely used in healthcare. (Mistry, 2020; HEE, 2019). The highly influential Topol review predicted that robots would become the “hardware” for AI, performing manual and cognitive tasks, and freeing up healthcare staff to spend more time doing things that are “uniquely human”, such as interacting with patients (HEE, 2019).

This picture is complicated by the fact that robots have also been developed to provide social and emotional support to people, arguably blurring the line between machines and humans. Examples currently in use include:

• Sophia, a companion robot for older people;
• Miko 2, a robot for children that can respond to emotions;
• Paro, an animal therapy robot (Robert, 2019).

A common theme in the literature on AI and robotics in healthcare is the expectation that patients, as well as staff, will receive multiple benefits from the introduction of intelligent machines, with improvements in early diagnosis, and the accuracy and safety of care (Mistry, 2020; HEE, 2019). Unlike human healthcare workers, robots never get bored or tired, are unaffected by hazards in clinical settings, such as X-ray radiation, and can endlessly repeat tasks that require precision without a drop in performance (Mistry, 2020). The potential is there for robotics to help with everything from moving patients to surgical procedures that are beyond the capabilities of surgeons (Mistry, 2020).

With the input of nurses, robots are also being developed to take on some nursing functions, including:

• Ambulation support;
• Vital-signs measurement;
• Medication administration;
• Infectious disease protocols (Robert, 2019).

What AI means for nursing

Developments do not mean that nurses are about to be replaced by intelligent machines, but they do suggest that nursing, as it is currently understood, will change. A 2019 study by former American Nurses Association executive vice president Nancy Robert suggested that the arrival of telehealth and smart robots in people’s homes will see nursing evolve into more of a coaching role, guiding patients to improve their health and providing continuity of care, but still being physically present at the bedside when it really matters (Robert, 2019). A nursing dean quoted in the study said they could not imagine ever choosing a robot over a human to care for them if they were dying, and stated that: “Nuances in human behaviour will keep nurses on the front line of care” (Robert, 2019).

It is hoped that AI and robotics will work together as assistants to nurses by, on the one hand, supporting advanced clinical decisions and, on the other, automating basic tasks that are time consuming but could be performed by someone – or something – else. In this vision of an AI-enabled future, machines free up nurses professionally to use their education, skills and experience (Robert, 2019). A barrier to nurses using the technology to fulfil their potential could be other healthcare disciplines’ resistance to nurses practising at the top of their licence (Robert, 2019).

There is also the risk that, as AI tools become more widely used in healthcare, they will influence how nurses’ practise, without nurses having the opportunity to influence them. A 2021 study on how the nursing profession should adapt for a digital future called for an “immediate inquiry” into the influence of AI on nursing practice for the next 10 years and beyond (Booth et al, 2021). The authors pointed out that the increased use of AI is bringing with it new policy, regulatory, legal and ethical issues; they called on the nursing profession to:

• Investigate the risks and opportunities;
• Examine its own role;
• Develop frameworks and guidelines to steer nursing practice.

Robert (2019) suggests that nurses have a responsibility to ask about the data used to train AI systems they use, and ensure they have been checked for bias.

As outlined in Box 3, nurses will need training and support to feel confident in, and overcome the barriers to, using AI tools – which will only work properly if the ageing technology infrastructure of the NHS improves (Joshi and Morley, 2019). The excitement about AI is justified but it is important not to get dazzled by the “hype” (Joshi and Morley, 2019).