Realising the benefits of artificial intelligence for nursing practice

Introduction

Artificial intelligence (AI) is an exciting and rapidly expanding area in healthcare that develops and applies advanced computing techniques, such as algorithms, to simulate aspects of human intelligence. Such techniques process large amounts of digital data, including:

• Written text;
• Audio;
• Images;
• Video.

From these data, AI tools can help predict the likelihood of a particular outcome (Russell and Norvig, 2021). These predictive algorithms are the foundation of many digital tools, such as chatbots and robots, that interact in a human-like way (Dwivedi et al, 2023). Other technologies using AI algorithms include:

• Social media platforms that recommend digital content based on a person’s previous actions;
• Internet search engines;
• Document-writing software that checks spelling and grammar and suggests improvements;
• Virtual assistants.

As such, AI is an established tool that is used in many sectors to enhance communication, support learning and improve decision making.

There are many definitions of AI and how it works. Samoili et al’s (2020) report, which was published by the Publications Office of the EU and based on a review of several definitions, defines it as: “software (and possibly also hardware) systems designed by humans that, given a complex goal, act in the physical or digital dimension by perceiving their environment through data acquisition, interpreting the collected structured or unstructured data, reasoning on the knowledge, or processing the information, derived from this data and deciding the best action(s) to take to achieve the given goal”. AI overlaps with other areas, such as data science, text mining and statistics, all of which have advanced computing techniques in common.

Although AI may seem new, its development began in the 1950s. Inspired by breakthroughs that Alan Turing and others made in encryption and computing, people began to explore how to create machines that could think and act like humans. For several decades there was lots of experimentation and dead ends in AI development but, in the early 1990s, two important areas in AI emerged:

• Machine learning (ML) – a set of statistical techniques applied to digital data to learn patterns or trends that can help predict a result or solve a problem (Russell and Norvig, 2021);
• Natural language processing (NLP) – this uses several computing techniques to break down the meaning and structure of spoken words and written text to understand human language. NLP also uses ML techniques.

Some of the most widely used AI techniques, many of which have applications in healthcare, are shown in Box 1, classified by the two categories noted above.

AI in healthcare

In the UK, a national strategy on AI was launched in 2021 to support the creation of new AI tools and systems and ensure that health and social care benefit from these advanced computer techniques (Office for Artificial Intelligence, 2021). NHS England also launched an AI Lab to foster collaboration on AI in healthcare and bring together initiatives that address barriers to developing and applying AI in different health and care settings (NHS England Transformation Directorate, nd). The AI Lab works with government, NHS trusts, industry, and the public to develop resources including reports on AI regulation, ethics, workforce development, and an AI roadmap to ensure safety and ease of use.

A 2019 report from NHSX highlighted many real-world case studies where AI tools and systems were being developed in the UK for areas of healthcare such as radiology, genomics and mental health (NHS X, 2019). A partnership of seven NHS trusts in the East Midlands is working with two AI companies to develop, test, and roll out AI tools for breast cancer screening. One AI tool uses deep learning techniques to understand mammographs and act as an independent reader in double screening programmes. Another AI tool aims to optimise clinic scheduling and staff resourcing to ensure breast cancer screening services are planned and delivered efficiently and effectively. King’s College London and South London and Maudsley NHS Foundation Trust have developed an open-source AI tool called Cogstack. This uses NLP and other AI techniques to improve the speed and accuracy of clinical coding and has been deployed successfully in outpatient clinics, helping save money and release staff for more complex tasks.

“Nurses need to develop an understanding of AI capabilities and applications in healthcare relevant to their clinical practice”

AI in nursing

Nurses started exploring AI in the early 1990s when informatics as a speciality area started to become popular. In the US, Rose Harvey used a neural network to develop a prototype computer system to improve the nursing diagnosis process (Harvey, 1993). Since then, other nurses have explored how to apply AI techniques to digital datasets to try to improve patient care. Im and Chee (2011) developed a decision support system that uses an AI technique called fuzzy logic to help nurses make better decisions around pain management for cancer patients.

A recent review by O’Connor et al (2023) summarised 140 research studies on AI with applications in nursing and midwifery. Most were published recently with an exponential increase in the last five years. The majority of studies were hospital based and used ML techniques to analyse data from the electronic health record to predict a range of patient outcomes or identify variables affecting outcome prediction (Table 1). A few studies examined how AI applications could improve nursing administration and management, such as nurse staffing and burnout. Some studies focused on nursing education using AI to predict student attrition, programme completion and graduation.

Benefits of AI

AI could bring a number of benefits to nurses. ML algorithms can be used to build predictive models to help nurses identify patients at risk of many kinds of physical, mental and social health problems. For example, a scoping review by O’Connor et al (2022) found 14 studies that applied AI techniques to falls data to develop prediction models that more accurately identify older people at risk of falling in hospital and community settings. In the future, these prediction models can be integrated into the electronic health record to send nurses a digital notification when patients are in the high-risk category. Such AI applications will support clinical decision making, enabling nurses to be more proactive in preventative care.

AI applications can also improve nurses’ diagnostic processes. For instance, Jain et al (2021) evaluated an AI tool for diagnosing skin conditions in primary care. When compared to the traditional approach of medical notes and skin conditions image review, they found the AI tool improved the diagnostic outcome.

AI applications may enhance the management and organisation of hospital wards and nursing services in the community. An et al (2021) used several algorithms to organise patients admitted to an intensive care unit (ICU) based on their disease severity and care needs. This computerised approach was designed to help nurse managers allocate ICU nurses with the right expertise to care for the patient.

Recent AI applications, such as ChatGPT, based on a type of AI model called a large language model, could be applied in healthcare. Madden et al (2023) suggested using these AI tools to analyse free text entries in electronic health records from doctors, nurses and other professionals to generate real-time summaries of patient care. This may be helpful in busy areas to support a range of tasks, such as clinical handover, patient discharge and patient education among others. The researchers used ChatGPT-4 to analyse unstructured medical notes in intensive care and found it produced concise summaries and answered queries. However, it also produced some false information and the authors highlighted data privacy and security risks when using the chatbot, as there are drawbacks to AI tools (Madden et al, 2023).

“Given their knowledge and skill set, nurses are integral to teams developing and applying AI algorithms to health data”

Limitations of AI

AI has several limitations of which nurses need to be aware. AI algorithms and predictive methods are only as good as the quality of the data they are developed on. If a health dataset has missing variables or some patients are overrepresented or underrepresented, the AI tool may give inaccurate or biased results. This problem is called algorithmic bias (O’Connor and Booth, 2022). For example, a research group in the US identified racial bias in an algorithm used by a health insurance company (Obermeyer et al, 2019), which could disadvantage certain patients. AI-based digital tools may also be developed on older datasets that may not help predict future health problems.

Another issue with AI is that it is not clear how some algorithms work. This is known as ‘black box AI’, which could mean the results produced by an AI tool may not be accurate and reliable (Wadden, 2022). Therefore, nurses and other professionals need to interpret the recommendations of any AI tool to ensure clinical accountability, patient safety and prevent any legal issues arising from overreliance on AI systems. To inform decision making, the recommendations of any AI application should be combined with clinical and managerial expertise, and patient values and preference (O’Connor et al, 2023). The World Health Organization (2021) has also highlighted several ethical issues with using AI in healthcare and developed key ethical principles for its governance and use (Box 2), which can guide nurses.

Several barriers can occur when introducing AI in healthcare. Many nurses lack knowledge and skills in AI (Booth et al, 2021). This may reduce how quickly and well AI tools are developed and applied in patient care. Some clinicians are concerned that AI may replace their jobs or that AI tools will substitute clinical decision making (Castagno and Khalifa, 2020). AI applications, such as robots used in healthcare settings, lack empathy and other human emotions that can influence decision making and patient care (Stokes and Palmer, 2020).

Finally, the cost involved in developing and applying ML, NLP and other AI techniques should be considered as they are costly and may not lead to many benefits for patients, nurses and other healthcare professionals, or the health service.

“If a health dataset has missing variables or some patients are overrepresented or underrepresented the AI tool may give inaccurate or biased results”

How nurses can engage with AI

Nurses wishing to make the best use of digital data to improve care could get involved in applying AI algorithms and creating AI-based technologies that support patients and staff in hospital and community settings. Given their knowledge and skill set, nurses are integral to teams developing and applying AI algorithms to health data. O’Connor (2022) recommended educating nurses on AI to give them an understanding of existing and possible AI applications, along with the benefits, limitations and risks of these advanced computer techniques.

There are increasing numbers of university postgraduate courses on AI to learn the principles of how it works. This would allow nurse involvement in multidisciplinary teams to help develop and deploy AI applications in healthcare. Nurses interested in further developing their AI knowledge and skills could take courses on ML, NLP and how to write programming code in Python and R so they can create their own software algorithms. Massive open online courses (MOOCs), YouTube channels, and videos on AI are also available to understand how specific algorithms work.

To help develop an AI tool, nurses can reach out to analytics teams in healthcare services. Computer scientists in universities with ML and NLP expertise could also be consulted for advice on algorithm selection and application (Topaz et al, 2016).

Commercial software companies specialising in AI can also help develop new technologies for specific clinical application. AI applications aimed at patients or carers could be co-designed with them to help tailor the application to their needs (Blakey et al, 2020). Clinical safety officers from the NHS Digital Clinical Safety Team can also help nurses follow the right information governance processes, data standards and other regulations for introducing new technologies like AI or integrating AI into existing computer systems and digital tools.

Introducing a new AI tool into clinical practice can be challenging and requires good management support. A chief nursing informatics officer can help with developing or purchasing an AI tool for nursing practice. Other leaders, such as the chief nursing officer or chief information officer, need to oversee and approve the introduction of new technology. Funding and other resources are also required to purchase AI software or develop a new AI tool, integrate it into existing workflows, IT systems and organisational processes, and to train staff how to use it (Ronquillo et al, 2021).

The benefits to patients and staff must be clearly communicated to senior management and all staff affected by the new AI tool. This type of complex change can be time consuming and requires multidisciplinary expertise and commitment to introduce a new digital innovation into clinical practice.

Evaluating the effect of an AI application on patient and staff outcomes is crucial, preferably through randomised trials, to demonstrate it is clinically and cost effective and safe to use (O’Connor et al, 2023). Although obtaining funding is highly competitive, nurses with the right skill set and project team are eligible to apply for funding through government funding agencies, such as the National Institute for Health Research, the Medical Research Council, and Innovate UK to scientifically evaluate AI applications. This can help strengthen the evidence base for this new technological trend so, if effective, nurses and other health professionals can use it to improve patient care.

“AI applications can improve nurses’ diagnostic processes”

Conclusion

Al is now commonplace in many healthcare settings. It can help us understand some of the daily problems patients, nurses and others in healthcare face by analysing large digital health datasets using algorithms and other computing techniques. Nurses need to develop an understanding of AI capabilities and applications in healthcare relevant to their clinical practice. They should seek opportunities to get involved and subsequently to lead AI initiatives in healthcare. This will help ensure the development of AI-based technologies that address the needs of the nursing profession as well as benefit patient care and the delivery of health services.