Artificial Intelligence: Friend or Foe of Healthcare Professionals?
 
Although the concept of Artificial Intelligence (AI) has a seventy-year history, only recently have its synergistic and generative capabilities reached a performance level suitable for real-world applications. Spanning data mining, complex statistical modelling, deep machine learning, computer vision systems and predictive natural language, AI now has the capacity to structure, analyse and present information in a manner beginning to rival human capabilities, including that of healthcare professionals. Here, AI tantalisingly offers enormous scope for improving speed, accuracy, and cost-effectiveness across everything from disease prevention to drug discovery, and in the face of booming demand, it will revolutionise diagnosis, treatment, and recovery as well as how healthcare is delivered.
 
AI’s potential application spans the entire range of medical specialisms, promising to become a professional’s best friend throughout training and in practice. AI is already used in a range of focussed areas. For example, streamlining workflows and optimising staff and resource scheduling allows professionals to focus on more complex aspects of patient care. Software such as ‘Notatio Copilot’ can listen to consultations, extract clinical information, and generate consistent, digitised notes instantly available via mobile phone. In the future, this will expand to include making comparisons across related data, reaching evidenced conclusions, and suggesting further investigations or treatments. Other examples include the rapid analysis of imaging data (X-rays, MRI, CT scans etc.) or pathology samples (examining tissue or other disease indicators under a microscope). AI enables earlier detection of many conditions including breast, skin, and prostate cancers, eye, heart, and neurological diseases, foetal complications, endocrine disorders and stroke. This aids faster treatment and significantly improves the patient experience and clinical outcome (Kumar et al 2023). AI trained on 130,000 images of over 2,000 skin diseases proved as effective as a dermatologist in detecting malignancy (Esteva et al, 2017). Trials are underway for AI-assisted operations, helping plan and execute a wider range of procedures, reducing recovery time and costs (Moglia et al, 2021). AI is helping epidemiologists better understand and predict population-level health trends, disease risks, and the efficacy of interventions. It can forecast drug or vaccine potential, uncover novel uses for existing drugs, and increase the efficiency of clinical trials, notably being deployed in genomic sequencing of SARS-Cov-2 (Abubaker et al, 2022). AI is also aiding the interrogation of genomic data to identify disease markers and predict treatment responses, heralding personalised medicine optimised for an individual patient (NHS England 2023). The UK government is drawing upon the NHS’s unique data repository spanning the health of over sixty million people and investing millions in digital transformation (including AI) across health and social care (Community Practitioner 2023). Several NHS Trusts already collaborate with research organisations, pharmaceutical companies, and technology firms (including The Alan Turing Institute, DeepMind, Sensyne, and IBM). Telemedicine, in which health is monitored remotely via wearable devices such as ‘Doc@Home’, is widely used for geriatric, cardiac, and mental health patients and in ‘virtual wards’ (NHS England, 2023). AI-enabled processing of this data improves healthcare accuracy, efficiency and accessibility aiding provision in remote locations or where regular clinics would be difficult.
 
Clearly, AI is a healthcare tool with a widening capability to directly improve working practices and patient care. In that respect, it is a friend not just to healthcare staff but to the wider population. Yet it is important not to get carried away. Whilst there is undoubtedly monumental potential, there is also the prospect of significant downsides which will directly and indirectly impact healthcare professionals, and it is crucial we consider and take steps to mitigate these. The most obvious limitation of AI in healthcare is the availability, reliability, and security of the sensitive data with which AI is trained and upon which algorithms are derived. Data input into these applications must represent the population on which the AI will be deployed. Unfortunately, much existing data is incomplete or biased in some way, such as the underacknowledged differences in presentation between male and female cardiac patients (St Pierre et al, 2022) or images of skin disorders being mostly on white skin (we must ensure AI reliably detects disease across all skin tones). Trust is a cornerstone of medicine, so biases and inequalities must be eliminated, not enshrined or exacerbated (Kizilcec et al, 2024). Additionally, the sheer scale and complexity of the algorithms powering machine learning make understanding precisely how AI reaches a given result very difficult to disentangle. Doctors not only must be aware of the technical nuances in interpreting AI output, but also need to be able to explain them to patients so they understand their prognosis and can meaningfully consent to treatment. Consent also extends to the implications of AI for patient data, specifically how it will be validated and used, whether the right to opt out will continue to be feasible, and the implications for patient privacy when anonymised data could be reconstructed across multiple data sources to re-identify an individual. Alongside this, commercial algorithms are closely guarded secrets and it remains to be seen how this will impact understanding, acceptance, and legal liability for errors. Equally, some patients may enthusiastically grasp the broad concept of AI without appreciating its limitations, developing unrealistic expectations and subsequent dissatisfaction with the healthcare they receive.
 
AI also raises longer-term professional fears, including redundancy. Medicine’s extended, arduous, and expensive training is founded upon an expectation of employability that reflects both this and the ongoing challenges the job entails. What may start as a useful tool may become a standard against which professionals are monitored, and if AI can do the job round the clock without error or fatigue, how can a human compete? The flip side is the potential for over-reliance on automated output, resulting in a gradual erosion of first-hand critical thinking and ethical decision-making. A computer spitting out the answer is likely to diminish a perspective otherwise honed over years of experience. Also, if AI handles less-challenging patients, what does that mean for those whose shift comprises only the most difficult, cognitively- and emotionally-draining cases? Burnout is a key driver of the healthcare staffing crisis worldwide, so AI must empower personnel by supporting professional autonomy and satisfaction, not add to existing burdens (Chen et al, 2023). This links closely to accountability and ethical concerns. Where AI is involved in challenging, life-altering decisions, what should be done if there is a significant discrepancy between the healthcare professional’s clinical judgement and the AI’s recommendation? Where does moral, legal, and regulatory responsibility lie? Who ultimately controls AI and how can they be held to account for their creation?
 
At the same time, the cost of developing, training, and deploying AI in healthcare will be staggering. While many such applications will eventually lead to greater efficiencies, the up-front costs will take decades to recover and, in the meantime, costs will be borne by taxpayers or through higher healthcare or insurance charges. Financing healthcare is a key component of the current crisis, and aging populations with more complex health needs will only fuel further rationing of care, exacerbating existing digital, financial, and other inequalities of access. Some would argue that the millions invested in AI now would be better spent repairing dilapidated hospitals, reducing waiting times, eliminating barriers to healthcare, and addressing staff attrition. Perhaps we should get patients out of corridors, and healthcare staff on a salary commensurate with the job they do before we boost the profits of technology companies. Others counter that investment in AI in the areas of prevention, screening and early diagnosis is exactly what will reduce pressures throughout healthcare and achieve a far better outcome for patients, staff, and society.
 
So, is AI a foe? In the short-term, I believe not, as these tools complement and augment rather than replace the professional. We are in a honeymoon period, playing with possibilities across myriad medical arenas and the future appears one of boundless potential. However, AI’s longer-term prognosis is much more uncertain. Computing advances will continue to be rapid and ultimately transformational - the AI we marvel at today will not be the AI we experience in twenty- or thirty-years’ time. Our grandchildren’s medics will not be trained or practice medicine in the same way ours do today, but there will still be doctors, nurses, paramedics, and myriad other professionals striving to give us the best possible healthcare. These professions will necessarily adapt and evolve in response to AI and any other impacts the future holds, but the obstacles that must be overcome are significant.
 
There is however one area where AI can only scratch the surface - our humanity. Empathy, caring, and compassion lie at the heart of the practitioner-patient relationship and have the power to instil courage, inspire trust, and banish isolation. Despite its advantages, beneath AI’s carefully-constructed veneer lies a machine. Deep down we will always know that and so continue to value human contact (Ruggeri, 2024). Equally, once AI’s novelty wears off and its inherent negatives have emerged, like Pandora we may not be able to put them back in the box. So, the challenge for us now is to establish ethical and regulatory frameworks that balance leveraging AI’s benefits while addressing its challenges and ensuring the well-being of professionals and patients alike. AI should be regarded as a co-pilot rather than an auto-pilot, retaining human oversight and patient contact as sacrosanct. Regulation must also remain flexible enough to cope with the unpredictable course AI is likely to take, and recognise that it will not develop in isolation but will integrate exponentially with other developing medical innovations such as nanotechnology and robotics. But that is a whole other essay…
 
 
1,592 words (excluding title and references).
 
 
References
 
Abubaker B.S., Ibrahim N.K., Abubaker B.H., Hashem A.R., 2022. Covid-19 and Artificial Intelligence: Genome sequencing, drug development and vaccine discovery. J Infect Public Health. February;15(2) Pp289-296. Available at www.pubmed.ncbi.nlm.nih.gov/35078755 (accessed 15 March 2024).

Chen, V., Liao, Q.V., Vaughan, J.W., Bansal, G., 2023. Understanding the Role of Human Intuition on Reliance in Human-AI Decision-Making with Explanations Proceedings of the ACM on Human-Computer Interaction, Volume 7 Issue CSCW2 Article No: 370, Pp 1–32. 04 October 2023. Available at www.dl.acm.org/doi/10.1145/3610219 (acessed 16 March 2024).

Community Practitioner, 2023. Artificial Intelligence in Healthcare: Friend or Foe? 13th November 2023. Available at www.communitypractitioner.co.uk/artificial-intelligence-in-healthcare-friend-or-foe (accessed 12 March 2024).

Cowan, H., 2023. AI in Medicine: Friend or Foe? Readers Digest Health. Published 18th October 2023, available at www.readersdigest.co.uk/health/ai-in-medicine (accessed 12 March 2024).

Esteva, A., Kuprel, B., Novoa, R.et al., 2017. Dermatologist-level classification of skin cancer with deep neural networks, Nature 542, P115–118.Published 25 January. Available at www.nature.com/articles/nature21056 (accessed 16 March 2024).

Kizilcec R.F, Shung D.L., Sung, J.J.Y., 2024. Human-machine interaction: AI-assisted medicine, instead of AI-driven medicine. From Chapter 10 in Artificial Intelligence in Medicine, ed. Sung J.J.Y., Stewart C., Academic Press, Pages 131-140. Available at www.sciencedirect.com/science/article/abs/pii/B9780323950688000108 (accessed 20 March 2024).

Kumar, Y., Koul, A., Singla, R. et al., 2023. Artificial intelligence in disease diagnosis: a systematic literature review, synthesizing framework and future research agenda. Journal of Ambient Intelligence and Human Computing 14, Pp 8459–8486. Published 13 January. Available at www.link.springer.com/article/10.1007/s12652-021-03612-z#Abs1 (accessed 24 March 2024).

NHS England, 2023. Artificial Intelligence (AI) and Machine Learning. Version 1.1, 14 June. Available at www.england.nhs.uk/long-read/artificial-intelligence-ai-and-machine-learning (accessed 18 March 2024).

Ruggeri, A., 2024 Artificial Kindness, New Scientist, 9 March, Pp 33-35.
St Pierre S.R., Peirlinck M., Kuhl E., 2022. Sex Matters: A Comprehensive Comparison of Female and Male Hearts. Frontiers in Physiology, 22 March. Available at www.frontiersin.org/journals/physiology/articles/10.3389/fphys.2022.831179/full (accessed 17 March 2024).