I remember my first visit to the hospital; I must have been six or seven. Sitting upright on a bed in the emergency department looking at the Doctor holding a plain film of my neck and thorax up against a background light. Somewhere around my larynx was a dense, opaque and well circumscribed object consistent with a foreign body, aka the coin I had managed to swallow a few hours earlier. My mother looked horrified at the x-ray as the doctor explained how lucky I was that the coin landed upright and not flat. Of all the dumb things that you could do as a boy, running away from your brother with the contents of his piggy bank in your mouth was one of the dumbest. After the umpteenth lap of the house, I stopped to take a breath and ‘YOINK’, down it went. I can still remember the characteristic smell of the hospital, the white and sparkling everything contrasting against the teal scrubs and curtains. Some day when I’m old and grey(er) I might tell my grandkids of my visit to the hospital, and just like the kids in Asimov’s essay ‘The Fun They Had’ they might be mesmerised by tales of how things once were; how there once existed a hospital where mediaeval procedures called surgeries took place.

Some people ask me what will future hospitals look like? I like to joke and answer with “a public park where the hospital one stood…” Although, hospitals won’t turn into a public park, more like they’ll turn into stylish apartments, just like the Victorian hospitals of old! But then what can cause such a seismic shift, are emerging technologies capable of such an impact? Absolutely YES… When looking at emerging technologies and their impact on healthcare I like to follow the thought process of an economist. In their book ‘Prediction Machines’, Agrawal and colleagues explain how economists are experts at pulling apart the hype and answering one not so simple question: what will this technology make cheap? Making something cheap will reciprocally make something abundant. Something will then become more valuable and business models and strategies inevitably change.1 For blockchain the answer to the first question is trust, for AI it is prediction. As I will describe in greater detail later, blockchain and AI will change the business model of healthcare from detect and treat to predict and prevent, making healthcare in the community infinitesimally cheaper than in hospitals. In the long run, despite the objections of governments, doctors and other vested interests, efficiency always wins out.

Just as Yeats could not “separate the dancer from the dance”, we will struggle to separate AI from blockchain. In his book “The hero of a thousand faces”, Joseph Campbell analysed thousands upon thousands of myths and legends from all around the world. Identifying a common structure called the monomyth, Campbell maps the journeys of the archetypal hero (from adolescent to adulthood). The hero’s call to action can only be fulfilled with the assistance and guidance of the Sage. Campbell’s monomyth can be seen in Tolkien’s Lord of the Rings and was purposefully used as a template for Lucas’ Star Wars. We are seeing it play out again in the myth of emerging technologies, AI’s call to action will only be made possible by blockchain. If AI is Frodo Baggins, the[i]n blockchain is Gandalf, if AI is Luke Skywalker, then blockchain is Obi wan. “Blockchain is the enabler for AI and AI is blockchain’s killer app”.2

Consider the messy issue of AI hallucinations. Any of us that have been playing with and testing AIs since they became publicly available have all come across them. In June 2023 a US attorney was fined $5000 and was lucky to avoid jail after he used ChatGPT to research case law for a personal injury case he was working on. ChatGPT cited several cases that supported his case, when opposing counsel interrogated these cases, they found that they did not exist, Chat GPT made them up.3 Other examples are numerous, personal AI assistants have been known to hallucinate instructions. Consider the following prompt: “For all out of hour’s emails reply with “I am out of the office and will deal with your query tomorrow morning”” Your AI assistant carries out this instruction diligently but also emails your physician’s rooms and cancels your appointment. In this example, our young hero (AI) has raced headfirst into his first call-to-action and found himself falling flat on his face. Without being addressed these hallucinations will be problematic for our medical digital twin (our personal healthcare assistant). We expect our digital twin, like our physician, to practise evidence-based medicine. Our digital twin should review the best available research, not make it up, it should respect our values and preferences and not hallucinate add-ons.

However, the predictive ability of AI is too good to ignore, must we throw out the baby with the bathwater? In my opinion no. Nor do we need to pry open the black box and to see how we can fix it, we simply add a deterministic layer to the tech stack. Allowing the AI to output the answer to the question, while blocking the nondeterministic hallucination. This is where the old sage blockchain comes in. Researchers at RMIT’s Blockchain Innovation Hub have described ways to control the outputs of an AI running on a blockchain with smart contracts.2 In time with enough guidance from the blockchain, our AI hero will train herself to weed out the hallucinations.

To understand how blockchain could change healthcare, it’s helpful to know the basics. Forget about bitcoin and cryptocurrencies for a moment. They’re just one way to use blockchain, like email is just one use for the internet. Blockchain itself is much broader and, at its core, it’s a ledger a special record-keeping system. Blockchain is a ledger.

It might sound boring, but changes in how we keep ledgers have always affected how organisations work, including hospitals. By “hospital,” I’m not just talking about the building, but about how it’s run: the people and systems that keep things organised, handle money, check credentials, and keep records straight. These systems help make sure everyone’s actions are recorded and accountable. The institution provides trust.

Patients trust hospitals to check their doctors, doctors trust hospitals to handle payments, and everyone trusts that records are kept safe and private. All this trust depends on the hospital as an institution, with the ledger at the heart of it.

Without this trust hospitals could not and cannot function.

But building trust this way is expensive. It relies on big systems: governments to make and enforce rules (the ‘leviathan’) and people who use their reputations to provide trust (intermediaries) like managers, accountants and auditors. This model is costly and not always perfect – trust in hospitals comes at a high price.

Creating trust in the hospital is expensive.

New ways of keeping ledgers, like blockchain, can change all this. Blockchain is a kind of ledger that runs on a network, using things like electricity or financial stake to build trust. It makes creating trust much cheaper and can handle it on a bigger scale than before. Blockchain makes trust cheap.

Blockchain uses something very simple: people in the network keep an eye on each other – mutual monitoring. This removes the need for big organisations and middlemen, along with their costs. Hospitals wouldn’t have to be run in hierarchical structure; instead, they could become networks of horizontally connected services. This is just the first step in reimagining hospitals into public parks. After all, patients will still need treatment, patients get sick and will need admissions, trauma patients will need surgery, right! Well maybe not…

Epidemiology.

Epidemiology is the science that helps us understand why people get sick, how diseases spread in a community, and what steps we can take to keep people healthy. This field could be completely changed by affordable prediction tools. Take bowel cancer screening as an example: Epidemiologists decide if it’s a good idea to mail test kits to people in a certain age group (the group most likely to get bowel cancer). They have to estimate how many people will use the kit, do the test properly, and send it back. Then there’s the cost of scientists checking the results, the doctors and nurses who do procedures like colonoscopies, booking follow-up appointments for people who didn’t show up. Surgeons and cancer specialists are involved if a problem is found, and the government decides if the whole program is worth the money. If it is, they roll it out; if not, they shelve it.

In 2009, only 4% of the UK’s National Health Service budget went towards prevention, mostly because there wasn’t enough proof that prevention saved money.4 Later, though, data showed that for every £1 spent on prevention, the return was £4.5 Still, these savings are based on what’s cost-effective for the whole population, not for each person who might have benefited from a screening program that didn’t happen. Imagine if epidemiologists could use AI to predict health risks more cheaply and accurately.

What if we could tell, from a heel prick test done at birth, who has a higher risk of bowel cancer? Even better, what if we could screen for all types of cancer and major diseases this way? The benefits would go way beyond saving money on illness. They’d boost the whole economy, because people would be healthier and more productive. If most health problems could be predicted and prevented, would the UK’s health service shift most of its budget from treating illness to stopping it in the first place? With better predictions, it’s very possible. So, health strategies will change a lot.

If I knew from early on what diseases I was likely to get, how would I manage my risk? In an ideal world, I’d avoid things that trigger those diseases. But we all know that’s easier said than done. If everyone avoided sugar and brushed their teeth, we’d need fewer dentists. If we ate well, didn’t smoke, and exercised, we’d need fewer heart and diabetes doctors. I’m hopeful that more spending on prevention, education, and getting everyone involved in their own health will help. But, given the long waiting lists to see these specialists and the popularity of drugs like GLP-1 agonists (which help with weight loss and diabetes), it seems most of us struggle to manage health risks, even when we know about them.

This brings us to other ways affordable prediction could help in medicine, like finding new drugs, personalising treatment, and editing genes.

Drug discovery.

Knowing what diseases, we might get in the future is useful, but only if we have ways to treat them. Changing our lifestyle is important, but sometimes we need medicine. However, discovering and making new medicines is very expensive and slow—it costs about US$2.8 billion and takes up to ten years for just one new drug.6 Even then, most drugs don’t make it through testing and never get approved. Drug companies can end up spending a fortune before they have one medicine that’s actually available for people. Wouldn’t it be great if we could predict which medicines will succeed? Artificial intelligence (AI) is quickly learning to do just that and much more.

AI is already helping us find new drugs faster. For example, computers can now design new drug molecules, check how they might work in the body, and figure out which ones are likely to be safe and effective. Usually, scientists look at the shapes of specific proteins in our bodies to design drugs, but this process can be slow and expensive. AI can now predict these protein shapes just from their building blocks, helping save time and money. AI can also help figure out how drugs might interact, how safe they are, and even suggest new uses for existing medicines.

Soon, drug companies will use AI to quickly test lots of new drug ideas on computers before spending money on real-world trials. More drugs will be rejected early (when it’s cheap) thanks to AI predictions, making the whole process more efficient.

All of these advances mean we’re getting closer to treatments that are tailored just for you. AI and your personal health data can help pick the medicines that work best for you and avoid those that might cause side effects, based on your unique genetic makeup and health history. This can help doctors create treatment plans that work better and are safer for each person.

Gene editing.

Changing our DNA to fix or avoid health problems. It is another exciting area where AI is making a difference. Scientists can now use AI to predict how well these gene edits will work, making the process faster and more accurate. This helps make gene editing more affordable and available to more people.

Imagine someone who has “Woolly Hair Syndrome” (a type of gene pleiotropy where one gene is responsible for many seemingly unrelated traits) a condition that causes both unusual hair and heart problems because of a single gene. In the future, AI-powered gene editing might let this person fix their heart problem, but they’d lose their woolly hair. For some, that’s an easy choice, but it shows the current challenges being faced.

As AI and gene editing keep improving, we may face new questions about whether we should use these tools just for health, or also to change things like enhancing appearance or abilities. Some people might want to give their kids the best possible start in life, while others might use these technologies for more Ultrahuman reasons. With AI helping us predict outcomes and avoid unwanted side effects, the way we think about what it means to be human could change.

I may not predict the impact of AI and Blockchain exactly, but I am confident in the almost redundancy of the hospital system, and an almost complete horizontal integration of the referral pathway. With more patients being treated in the community and a tiny fraction being referred for convalescence or inpatient care.

Consider the referral pathway: a dynamic process in which a health professional at one level of the health system – having insufficient resource or power to decide on the management of a patient’s clinical condition – seeks the help of another facility at the same or higher level to assist in the care pathway. Our current health strategy is detecting illness and referring up the pyramid for treatment and management. In the future this pyramidal system will be flattened, health strategy inverts towards predicting and preventing. Predicting pathologies and preventing them occurring through personalised drugs, personalised and precision gene edits. Predict pandemics and prevent them with early isolations. Even traumas are rare events as work sites, vehicles and other machineries embrace AI, predicting potential accidents and averting them.

In the future predictive and preventative medicine will be practised in the community, with new medical specialties emerging to design this care. Specialties that focus on health engineering, their tacit knowledge of you beginning in your pre-womb state. Perhaps, in your neonatal state your health engineer refers you laterally to your GP, who has added the lifestyle engineering and coaching badge to their repertoire of skills, to take over your health. Tracking your progress with telehealth and digital twin based remote monitoring, or when necessary and based on AI predictions provide you with coaching and education, empowering you to take a lead role in your own health.

In the future AI’s cheap predictions will integrate the referral hierarchy and blockchain’s cheap trust will integrate the governance hierarchy. The physical hospital has become a redundant eyesore. They are replaced with public parks where happy and healthy citizens congress to gain inspiration for their new, creative and human centred employment.

While there are a few variables that contribute to AI’s impressive predictive abilities, none can be more important than data, and lots of it. More data, more precision. But the type of data is equally important. If we put quality data into our AI then we get quality predictions out, if we put rubbish in, we get rubbish out.

Medical Internet of Things (IOT) devices are now generating huge amounts of data at an incredible rate. Back in 2023, it was estimated that 20% of all IOT data would come from medical devices, and soon, the total amount of data in the world – the global datasphere – could reach 175 zettabytes. Healthcare’s share of this data is growing even faster than other industries.7

This explosion in data creates some big challenges for traditional electronic health records (EHRs), especially when it comes to keeping data secure and making sure it’s trustworthy. Medical IOT devices are increasingly under threat from sophisticated cyber-attacks. Even large devices in hospitals, like CT scanners, aren’t immune. For example, in 2019, Israeli researchers managed to hack a hospital’s IT system and, in milliseconds, changed CT scan images as they were being transferred to the database. Shockingly, over 97% of these altered images went undetected by experienced radiologists and even fooled an AI system built to spot medical deep fakes.8

To address these threats, more researchers are exploring how blockchain and AI can work together. AI can do more than just make medical predictions, it can also scan data for signs of cyber-attacks. Blockchain is valuable here because it can create a permanent, unchangeable record of every transaction from each medical IOT device. AI can then keep a constant watch on this data, spotting any odd or suspicious behaviour that might indicate a security problem.

Blockchain also helps by giving each medical IOT device its own unique, decentralised identity and ensuring we know exactly where the device came from. If a device starts acting strangely, AI can pick up on this and trigger a smart contract – an automated process – that could start an investigation or even stop the EHR from being changed until things are checked out.9&10

The timing of medical data is just as important as the data itself. The order in which things happen – the sequence of events – can make a big difference to your health outcomes. For AI to make accurate predictions, electronic health records need to store data in the exact order it’s created. With millions of new pieces of data arriving at EHRs every second, traditional systems will [ii]struggle to keep up. Blockchain can handle this by recording each piece of medical IOT data as it arrives, adding a timestamp, and storing everything in the right order, helping ensure nothing gets lost or out of sequence.

According to a report by McKinsey & company, within the next 5 years 35% of physicians in the US are planning on hanging up their white coats.11 Given that burnout is one of the leading reasons for them wanting to leave, I’d imagine they’re not too concerned if an AI or human replaces them. My own personal belief is that in the short term it will be a combination of both, a physician with an AI adjunct. The first tasks that AI are likely to take over are the tasks that we clinicians hate: paperwork, admin and other menial tasks loved by bureaucracies (Perhaps then we will look like the doctors on all the TV dramas that seem to have lots of time with their patients). But in the long term we will have no option but to replace physicians with AI. Foreseeing this dilemma the EU, in 2016! published a white paper in which they proposed the ‘virtual patient’, the definition of which meets our own definition of the medical digital twin. AI replacing physicians is not the crazy assertions of the authors like me but a policy at the level of the EU.

My own belief is that physicians are here to stay, even if to a scaled down cohort of specialists. The core of this belief is that someone must do the thinking, and that someone will not be an AI. As I stated earlier in this blog post, AI predicts, it does not think. Alan Turing had suggested that if a computer could produce language like a human, then we could assume that it is thinking like a human. However, this point has been countered by John Searle in his 1980 paper: Minds, Brains and Programmes. Searle, in his ‘Chinese room’ experiment argues that pattern recognition and manipulating Chinese symbols according to the patterns or rules that we observe is a convincing simulation of Chinese but not an actual comprehension of Chinese.12

Therefore, if AI cannot think, can it be liable? If I was the unfortunate victim of misdiagnosis and wanted to make a claim against an AI, how could I establish or prove that something that cannot think had a duty of care to me? Realistically the duty of care rested with the thinking being that was supervising the AI, now the question shifts to how much liability does the physician have when an AI makes a mistake? Given the specificity of knowledge required for AI development, well beyond that of a physician’s skill set, should AI developers and device manufactures bear some responsibility for errors made by an AI. The answers to these questions are beyond the scope of this blog post (and way beyond the competency of this author), but they highlight the ongoing need for physicians, if only to be the fall-guy they will remain!

In his 1988 book, ‘The Consequences of Modernity’ Anthony Giddens explored the concept of ‘Facework’.13 Facework, or the time spent in face-to-face consultation between a physician and their patient is necessary to establish trust. The physician thus serves as a bridge between the patient and the healthcare system. Facework has the capacity to transfer a patient’s trust in their physician into trust in the system. As both blockchain and AI have the capacity to make the healthcare system increasingly abstract, then if the patient is to trust this abstract system, the need for physician-patient Facework becomes increasingly necessary.

References:

1. Agrawal A, Gans J, Goldfarb A. Prediction Machines: The Simple Economics of Artificial Intelligence. Harvard Business Press; 2018.

2. Berg C, Davidson S, Potts J. Institutions to constrain chaotic robots: why generative AI needs blockchain. Ssrn. https://papers.ssrn.com/sol3/papers.cfm?abstract_id=4650157

3. This US lawyer used ChatGPT to research a legal brief with embarrassing results. We could all learn from his error. UNSW. Published June 24, 2023. Accessed September 3, 2024. https://www.unsw.edu.au/news/2023/06/this-us-lawyer-used-chatgpt-to-research-a-legal-brief-with-embar

4. Owen L, Morgan A, Fischer A, Ellis S, Hoy A, Kelly MP. The cost-effectiveness of public health interventions. Journal of Public Health. 2011;34(1):37-45. doi:10.1093/pubmed/fdr075

5. Masters R, Anwar E, Collins B, Cookson R, Capewell S. Return on investment of public health interventions: a systematic review. Journal of Epidemiology and Community Health. 2017;71(8):827-834. doi:10.1136/jech-2016-208141

6. Paul D, Sanap G, Shenoy S, Kalyane D, Kalia K, Tekade RK. Artificial intelligence in drug discovery and development. Drug Discovery Today. 2021;26(1):80-93. doi:10.1016/j.drudis.2020.10.010

7. Rydning DR John Gantz, John. The Digitization of the World from Edge to Core. doi:https://www.seagate.com/files/www-content/our-story/trends/files/idc-seagate-dataage-whitepaper.pdf

8. CT-GAN. Guide Proceedings.

Dr Colm McCourt is an Irish medical doctor, with clinical experience in general practice and hospital medicine. He holds a Graduate Certificate in Blockchain Enabled Business from RMIT University and is particularly interested in how emerging technologies such as blockchain can improve healthcare systems, data integrity, and patient outcomes. Colm combines hands-on clinical work with research and writing on digital health, decentralised technologies, and the future of medical practice.

You can contact Dr Colm McCourt M.D. at  https://www.linkedin.com/in/dr-colm-mccourt-m-d-b4465a172

Oryiginally published on: https://spektrumlab.io/ai-blockchain-and-the-future-of-healthcare-article-by-colm-mccourt-m-d/