“Healthcare is not like banking,” says Dr. Anushka Patchava, Expert Adviser to the United Nations. “Rather it’s a more emotional business. Artificial intelligence in healthcare implies the use of…non-human systems, like computers, performing tasks such as decision-making, without a human touch. With healthcare, we aren’t there yet. Although artificial intelligence can solve some of the problems, for example within the supply chain– augmented intelligence is more applicable clinically.” As for data, Patchava explained why distributed ledger is an exciting solution, promoting data fluidity across systems, networks and services.
Patchava, who is speaking at the upcoming Tech Leaders Summit in London on 12 September, puts emphasis on the human factor. “In banking,” she said, “it’s very much about being low touch, high tech, and that’s acceptable because people want that. They seek convenience; access as and when they want it and as quickly and effectively as possible. Consumer experience is scored on first call resolution. They almost don’t want to have to deal with a human, it’s time-consuming and often a distractor, rather than enabler. Whereas in healthcare it’s actually the flip of that. Consumers are in a different mental mindset. The sick patient seeks empathy and reassurance. In fact, it is the trigger of these two feelings that has been proven to help healing. The psychology, physiology and pathology of healthcare consumers (or patients as we know them) is ever complex, and to date, strongly benefits from human interaction.”
Healthtech valuations reach five year high
Dr. Anushka Patchava: We don’t want to completely replace doctors and healthcare workers; we need to find synergistic partnerships
It is a truth universally acknowledged, however (and as Jane Austin once said), that globally, health systems face several challenges. Ageing populations, the rise of non-communicable diseases such as obesity, diabetes, cardiovascular disease and spiralling consumer expectations all create an enormous burden on health care providers. Traditionally, large scale, wide-reach adoption of new technologies within health systems and amongst health professionals has been a huge barrier. If applied well, emerging technologies can help take away some of the strain, shifting care from reactive to proactive, and from treatment to prevention. Critically for these technologies to succeed, they need to be embedded within the workflow, driving productivity without at the same time sacrificing the human touch.’
Take artificial intelligence in healthcare. “We don’t want to completely replace doctors and healthcare workers; we need to find synergistic partnerships, where it’s neither high touch/low touch or low tech/high touch, it’s tech and touch working in combination to deliver the best outcomes for both physician and patient.”
How AI is revolutionising healthcare: 10 use cases of artificial intelligence in healthcare
The growth of artificial intelligence is evidential. Although we might see it, AI is truly changing our lives directly or indirectly, starting from its application in voice assistants such as Siri, Google Assistance, and Alexa to large scale applications in the supply chain, retail, manufacturing, enterprise mobility, autonomous cars, and more. Despite its progress in other industries and sectors, AI has genuinely made a difference in healthcare and affected thousands of people and made their lives better
She gives an example: Medic Bleep, a secure, real-time service – a healthcare ‘Whatsapp’, created to replace the old style, now out-of-date pagers, whilst remaining sensitive to security and privacy regulation. In the right hands of doctors, nurses, hospital managers and porters, Medic Bleep enables a efficient communication flow, of events and information as it happens. Results from a recent time-motion analysis revealed nurses saved an average of 21 minutes per shift whereas doctors saved an average of 48 minutes per shift – highly impactful when you think about what that equates to in a week, a month and a year. Furthermore, technologies like this can drive improved work prioritisation, easier collaboration, and through provision of an auditable record reduce errors and adverse incidents. Who would say no to safer and more efficient care delivery?
She envisions the Medic Bleep technology extending to emergency teams across multiple healthcare networks —not just a single hospital — and even onto social care “so that healthcare professionals can have a real-time view of health services and patients, who require care across the continuum, don’t have to worry about their results being transferred from doctor to social worker.’
‘We know Matt Hancock (the UK’s Secretary of State for Health and Social Care) has been really pushing the agenda to convert hospital pagers – and Medic Bleep provides a powerful solution, perpetuating real-time communication (and data flow) across health, and one day perhaps even social and emergency, care services.’
What’s the biggest barrier to AI adoption in healthcare?
The challenge of Artificial Intelligence in healthcare and why we call it augmented intelligence
Dr. Anushka Patchava: Real, true, artificial intelligence is almost always, at least in my lifetime, going to be non-existent in clinical care
Not so long ago, IBM’s AI engine, Watson, was being heralded as a panacea to healthcare. But “Watson has failed as a business unit,” says Patchava, “and I think the reason why the technology hasn’t achieved it’s potential and failed to impress is primarily because they didn’t understand the problem they were trying to solve. It’s hard to illustrate the complexity that exists within healthcare. There are so many stakeholders, touchpoints, disease areas and treatment options to name a few. Furthermore, if you think of each healthcare pain-point; it can exist at an individual level, at a population level, and/or at a system level – it’s a very intricate network.’
“Real, true, artificial intelligence is almost always, at least in my lifetime, going to be non-existent in clinical care. Although there are several areas where machine learning could be deployed to improve system processes, the unknowability of how output is derived from input and the overreliance on clinical decision support systems remain unsolved ethical dilemmas, slowing the replacement of humans in healthcare. For example, if I have a breast mammogram, although perhaps 6 or 7 times out of 10, AI may be able to detect breast cancer accurately, if my scan fell into the 3 or 4 out of 10 deemed ‘ambigious’ I would most certainly want a human to take a look, consult me, and reassure me of my diagnosis. Recent advances in AI algorithms have narrowed the gap between computers and human experts in detecting breast cancer, however we hold computers to a different standard. Although systems may, satisfy Turing’s test (the ability to exhibit intelligent behaviour equivalent to, or indistinguishable from that of a human), would patients (consumers) accept this? I believe if computers are to replace humans in healthcare, this substitution would have to prove improved accuracy, not merely equivalence.’
How IoT – and IoMT – is changing the face of medicine and healthcare today
Artificial intelligence and healthcare – where is there value?
“The computer is never going to know a human entirely. Algorithms and code cannot truly replicate human conscious and subconscious behaviour, attitudes, ideas and expectations. However, one area where machine learning is significantly evolving is genomics – the study of a complete set of genes, an organism’s make-up’. ‘The use of artificial intelligence to make sequencing and analysing DNA faster, cheaper, more accurate and more available could have significant implications on the way we deliver healthcare’. Through genomics we can achieve a deeper understanding of an organism’s behaviour – for example what diseases they are susceptible to, and use this to make decisions about their care. Earlier we spoke about the shift from treatment to prevention, and genomics forms the missing piece to both parts. Pharmacogenomics can help us understand how an individual may react to certain drugs, driving personalisation of therapy. In turn, it can also help prepare for the future. For example, if someone is at high risk of diabetes, but is not yet diabetic, we can offer them a greater level of targeted interventions to reduce risk and/or prevent catastrophic events – overall reducing cost, and improving outcomes.’
“I think genomics is going to push the future of medicine whereby we can deliver tailored and personalised services, however I also think genomics is never going to be the full answer. I’d still want a doctor, or at least healthcare professional to interpret my results, counsel me through what to expect, and formulate a treatment/action plan taking into account my concerns and preferences. As doctors, we must remember, the patient is one that is sat in front of you, not just the algorithm or piece of code.’
The technical potential for automation differs significantly across industries. “The financial services industry has been far superior in capturing and utilising our data, in person and more recently digitally. By doing so, they have been able to identify patterns based on for example, income earned, and deploy predictive analytics to forecast consumer behaviour. In healthcare we are slightly further behind, with somewhat still disparate and fragmented data sets peppered across health systems. To be effective, not only do we need to bring together existing data sets at an individual and population level, but we also need to obtain a a far deeper, more holistic picture of a person’s wellness, not just sickness. This is where new technologies such as wearables and connected devices can help fill some of the gaps – providing healthcare professionals and systems with larger, broader data sources, which in turn can improve algorithms and predictability. Only then can we shift healthcare from ‘sick care’ to ‘health care’.’
Digital integration of healthcare means empowering patients and clinicians
Enter DLT for healthcare
In most cases, if you hear the word blockchain, you immediately think of bitcoin, and some may even affiliate this to the dark web, money laundering, scams, gambling and such likes. So, let’s decouple this – move away from calling it blockchain and consider the potential of the underlying technology distributed ledger technology, or DLT.
‘We’ve talked about obtaining more data – to help decision making. However a key challenge in healthcare, and health systems globally is data fluidity. Barriers, regulation, policy and processes prevent data flow not only within the system, but also between’
“Right now, you go to your GP. Guess what? Your GP doesn’t know you were admitted into hospital last week and has not received your recent test results. You have to then call the clinic to find out where these are. If you’re lucky, after ten minutes of holding and a further twenty minutes of conversation you manage to have your results emailed, more likely faxed to your local GP practice. And now, you have to make another appointment because by doing this, your 10-minute appointment time is up. Now imagine if you had more complex needs, suppose you needed social care on discharge – and the information hasn’t reached social care. So you either never get your carer at home, causing you much difficulty and inconvenience or worse yet, you end up back in hospital because you are unable to cope.’ You can see almost immediately how the lack of data fluidity can have serious impact, to patient outcomes, to healthcare professional decision-making and to system costs.’
“Electronic health records were never designed to manage multi-institutional life time medical records. As it stands, patients scatter their health data across various organisations, pharmacies, fitness centres, hospitals, clinics as their health journey evolves from one institutes data silo into another. What we need to solve for, across healthcare professionals, technology companies, data scientists and managers, is how do we perpetuate data flow? How do we enable systems to talk to one another, transferring accurate interpretable data between themselves without adding a huge cost to already diminishing budgets?’. This is where DLT may have potential.
“At the moment? With current systems, even though we may capture data, this could be held in different electronic health records – with no communicating layer in between. Without data flow, no one person has a full holistic picture of their patient.’
“So although the intention is there, to use data from wide and varied sources including pharmacogenomics, wearables, Alexa, Google Home – to drive better healthcare decision making and outcomes – we still need to crack how we effectively allow the right people access to this data, at the right time.’
“Unless we have access to the holistic data set, all that data in one place, our ability to analyse and draw meaningful insights is restricted. How can we accurately train computers (machine learning) without providing the full picture?’
Robotic process automation in healthcare will lead to better patient outcomes
Robotic process automation in healthcare: Software bots represent a powerful digital workforce that can work tirelessly in the background supporting doctors and administrators, says Henry Xie, is the CEO and founder at Simple Fractal
The DLT stone-age
While organisations like MedicalChain are deploying distributed ledger technology to partially solve for interoperability issues, , providing immutable trust and distributed secure access to patient longtitudinal health data, she feels that as it exists currently, the technology is still in the ‘dial-up’ internet phase; a nod to the time when logging onto the internet involved a modem connected to your telephone circuit, dialling onto the internet via the telephone at local call rate. It was slow, and terribly frustrating.
Eventually, with iteration and evolution she sees DLT as something that is just there, existing in the infrastructure across sectors — like the carburettor in the car. People might say ‘I’ve got a new car’, they might say ‘ I’ve got a BMW car’ what they don’t say is ‘I’ve got a BMW car and then tell you about the engine type, power, operating mechanism etc.’.
“There are still several challenges to overcome. Firstly, we’ve needed to go through almost a conscious de-coupling of DLT from blockchain in order to initiate conversations about DLT in the first place. We are that point right now, across forward thinking health systems, whereby there is interest peppered with enthusiasm from early adopters. However, we can’t replace the entire legacy infrastructure of a hospital that’s existed over decades? And not without significant investment (which in this country hospitals just don’t have) and then causing severe disruption to operations in the process.
Laying the foundations to future-proof our NHS and power UK HealthTech
Following Matt Hancock’s recent app-focused policy paper on the future of health, Afshin Attari, Director of Public Sector at Exponential-e, argues why it’s no use championing new technologies such as video consultations or curative chatbots if the right infrastructure isn’t in place to support it
“So the challenge is identifying where and how do you use DLT amongst the legacy infrastructure to improve efficiency of already existing processes.
“We’ve come to the philosophy that DLT in it’s first instance could be regarded as a ‘plug and play’ technology. One that is recognised by the existing system and is able to replace part of the existing system without jeopardising integrity, functionality, or reliability of the existing system.”
Dr. Anushka Patchava: Through identifying a small pain-point, and solving for this, they’ve done really well proving if we find the right use cases for DLT, it doesn’t have to replace the entire hospital infrastructure
She cites the example of Dovetail Lab, “They, working together with Surgical Consent, have pioneered a shared decision-making platform of which a paperless, personalised consent form can be used to consent for elective surgery, based on ‘distributed ledger technology.’ The vision here is to create a fully consented database of personal preferences, risks and benefits associated with different treatments, decisions made, actions taken and patient reported experience and outcomes. Over time, there is potential for this to capture more than just consent – interoperating with the existing patient record – and contributing to machine learning to ensure each decision is the right one for that patient.
“Through identifying a small pain-point, and solving for this, they’ve done really well proving if we find the right use cases for DLT, it doesn’t have to replace the entire hospital infrastructure, but can operate in conjunction with all the other layers in the hospital infrastructure – a better business case to take to the board. Through small use cases, and incremental improvements we can drive adoption; and only through these deployments can we test, understand, iterate and improve the technology for purpose.”