Data & AI: keeping the NHS out of critical care?
Mon 21 May 2018 | Leo Craig
With waiting lists growing and resources stretched to breaking point, the prognosis doesn’t look too promising for the nation’s healthcare sector. Leo Craig, General Manager of Riello UPS, outlines how a dose of data and artificial intelligence (AI) has the potential to be the perfect prescription that helps the NHS tackle the medical needs of the future.
We’re all well aware of the perfect storm facing the NHS. The population is growing and people are living for longer, often with chronic diseases that expensive new treatments help to manage. Resources are stretched, both in terms of funding and the personnel needed to deliver vital care. And patients are more demanding than ever, less likely to accept just medical opinion and more willing to explore alternatives and non-traditional care.
The NHS budget is already £120 billion a year, a figure that will top £123 billion by 2020. But the health service also deals with 1 million patients every 36 hours. Around 18 million Brits currently live with a chronic condition, more than 3.5 million alone have diabetes. A quarter of the population is obese, a figure which is steadily growing at a similar rate to the nation’s waistline. There are already more than 11 million over 65s and by 2040 a quarter of the UK will be in this age range.
In truth, we could easily fill this entire article with statistics such as these, but we’re sure you get the picture – a health service creaking under unprecedented and growing demands.
The increased use of data, robotics, and artificial intelligence has the potential to truly transform healthcare in the UK, both on the frontline in operating theatres and A&E departments and behind the scenes reducing the burden of administrative work.
A major report earlier this year by the think-tank Reform revealed 1-in-7 of us already use wearable fitness trackers. That’s prized data from millions of people that could be analysed to potentially help detect, diagnose, treat, or manage a myriad of medical conditions. Add to that electronic medical records, remote monitoring, and information captured from other sensors or apps and the possibilities are enormous.
This shift will undoubtedly provide huge opportunities for the data centre sector, as all this additional processing is reliant on extra capacity. But a high-risk area such as healthcare also poses unique challenges for managers and operators, namely safety, security, and ethical concerns – it’s an environment where a ‘mistake’ can have profound consequences, and in some cases literally be a matter of life or death.
The (virtual) doctor will see you now
The use of robotics in medicine is nothing new. One of the most famous examples is the da Vinci Surgical System, found in thousands of hospitals worldwide and assisting in more than 200,000 operations a year.
The Ultromics AI boosts the accuracy of diagnosis from 80% to above 90% and has the potential to save up to half of NHS’s annual pathology spend of £2.2 billion
But across every sphere of healthcare, from disease detection to end of life care, from the truly miraculous to the relatively mundane, the power of data and machine learning is making its mark.
AI-powered chatbots have been trialled on non-emergency cases to help reduce pressures on the NHS. People enter symptoms into the app, the AI consults a vast medical database and users get tailored responses depending on the information they’ve entered. Wearables and connected devices are playing a huge role in disease management for people with chronic conditions including cardiovascular disease or cancer. Sensors enable doctors to better monitor and detect changes in condition, while at the same time triggering timely reminders for patients to take vital medication.
AI currently being trialled even has potential to diagnose heart disease more accurately. Traditionally, cardiologists study echocardiogram scans to look for irregular heartbeats, but because of the qualitative element of the diagnosis, there’s a 20% margin of error – every year around 12,000 patients out of 60,000 are either sent home when at risk or undergo unnecessary surgery.
However, the Ultromics AI developed by a team at John Radcliffe Hospital in Oxford extracts 80,000 data points from a single echocardiogram scan and detects details doctors can’t even see. This boosts the accuracy of diagnosis from 80% to above 90% and has the potential to save up to half of NHS’s annual pathology spend of £2.2 billion.
While we might still lag behind other European countries, it appears a previously sceptical public is slowly coming around to the idea of increased AI and robotics in healthcare. Research from PWC finds 39% of us would be willing for a computer or robot to diagnose medical conditions and even recommend treatments. More than a third (36%) of people have no issue with a robot carrying out minor, non-invasive surgery on them, while 26% are happy for machines to perform major, invasive procedures.
Increasing automation could also have a positive impact on health administration too, slashing dreaded (but often necessary) ‘red tape’. The British Medical Association states a trainee doctor will spend 15% of their time on admin work, while the Royal College of Nursing estimates a fifth of nursing time is taken by non-essential paperwork. Think of all their tasks automated ‘virtual assistants’ could carry out instead – appointment booking, sending out reminders, composing patient letters, and more.
Data centre dilemmas: Storage, security, space
An uninterruptible power supply (UPS) already plays a pivotal role in healthcare. A clean, continuous flow of electricity is essential in the operating theatre, laboratories, or pharmaceutical research facilities where even the most minor disruption can result in disaster.
A data centre in as critical an environment as the medical sector should be designed to meet at least Tier III standard, providing N+1 redundancy
And successfully harnessing the data from all these medical devices and apps, not to mention patient medical records and latest advances in health research and technology, will rely on significantly increased data centre storage capacity backed up by reliable power protection.
NHS Digital, the health service’s technology arm, issued guidance early this year that the public cloud is a safe place for care providers to store confidential patient data, provided certain data security conditions are met. So the extra capacity required is likely to be provided by a mixture of cloud and on-site data centre storage. Cloud storage obviously offers the advantages of scalability (the scope to add additional memory as and when required), access (data is theoretically available to any organisation in the NHS), and initial cost (no need for space-consuming physical infrastructure), although there are obvious concerns about security and data protection.
A data centre in as critical an environment as the medical sector should be designed to meet at least Tier III standard, providing N+1 redundancy so that each component needed to support the IT processing environment can be shut down or maintained without the entire system going into shutdown.
It won’t eliminate the single point of failure that a Tier IV level data centre will, but it also won’t require the doubling up of infrastructure that a truly fault-tolerant facility will require, an important cost consideration in the public sector.
Robust data centre design also needs to be complemented with similarly strong critical power protection. Again, any UPS’s installed should be configured to N+1 redundancy to reduce the risk of system failure.
The moves to modular UPS make this easier to do without compromising on performance. Modular UPS units are ideal for medical data centres as they can be closely matched to the site’s power requirements while still delivering the necessary redundancy. They run more efficiently across all power loads so reduce energy consumption, take up less space, and all modules are ‘hot swappable’ so there’s no disruption to power protection if any individual components experience failure.
Of course, the medical sector poses other unique challenges to data storage too. Despite some recent improvements, much of the health service is still reliant on paper-based systems, while NHS ICT systems are notoriously outdated – only last year a third of all health trusts in England were hit by the WannaCry ransomware attack, while it was previously discovered that thousands of computers were still running versions of the Windows XP operating system.
NHS Digital is also part-way through replacing its previous N3 wide area network with its new Health and Social Care Network (HSCN). This transition is due to be completed by March next year, with the shift from a single, centralised system to this ‘network of networks’ aimed at enabling health and care organisations across England to access and share information more reliably and efficiently.
Data centres operating in the healthcare sector, or keen to take advantage of what will inevitably be an increasingly lucrative market, will need to ask themselves the question ‘are we HSCN compliant?’, and if the answer is no, what steps must they take to be so.
The importance of a reliable data centre infrastructure was demonstrated back in January when a technical fault at both of NHS Wales’ main datacentres in Blaenavon and Cardiff Bay led to nationwide network outage that prevented doctors and healthcare workers from accessing patient records, medical notes, and even staff email services. It appears lessons still need to be learned.
In certain circumstances, AI is better than doctors at diagnosing certain conditions
Data in healthcare – the current diagnosis
It’s clear embracing artificial intelligence and machine learning in the NHS could have a hugely positive impact on not just the physical well-being of the nation, but potentially the public purse-strings too.
The NHS has been criticised in the past for being slow to embrace new technologies, although in a field where decisions can literally mean the difference between life and death, such caution is perhaps inevitable.
But the appetite for change seems inescapable. NHS England is planning to invest even more of its £120 billion-plus budget into AI, with the organisation’s then outgoing National Medical Director Professor Sir Bruce Keogh claiming last autumn “…in certain circumstances, AI is better than doctors at diagnosing certain conditions.”
Simon Stevens, Chief Executive of NHS England, is similarly supportive, believing big data and machine learning can lead to significant advances in radiology, pathology, and dermatology, to name just three medical specialisms.
To fully harness this potential though, the datacentre and critical power protection industries need to pass their own ‘medical tests’ with flying colours. We’ve touched on the essential factors data centre managers need to consider in terms of redundancy, security, and complying with network standards, and how a robust UPS has a crucial part to play in the design of such facilities.
But as healthcare becomes more reliant on robotic equipment, the same principles must also apply there too. If, for instance, a robot is performing complex surgical procedures, it needs to be supported with the suitable power protection – a single UPS wouldn’t completely guarantee the critical continuous electricity required to ensure the operation isn’t impacted by power fluctuations. Hospitals need to install UPS systems with at the very least N+1 redundancy along with dependable backup generators.
Next week we travel from the hospital wards and GP practices of the medical profession to the training grounds and stadiums of elite sport. We’ll explore how big data and analytics are playing an increasingly influential part on and off the pitch, and examine the role datacentres have in determining which competitors eventually emerge as champions.