When it comes to digital health technology, how do we bridge the gap between what is possible and what is actually happening?
The digital health revolution has been a long time coming.
Once it arrives, we are told, everything will run at peak efficiency, for less money, with higher quality care, greater transparency and immaculate interoperability.
The digital data trinity – electronic medical records, artificial intelligence and big data – will, futurists claim, metamorphose healthcare.
But this sparkling vision grates uncomfortably against the reality; the dull dialling of fax machines can still be heard in most waiting rooms around the country. Frustrated, sick patients still cannot access their most basic medical records remotely online.
Experts came together this month at the Medical Technology Association of Australia annual conference in Sydney to discuss what is being done to close the expectation gap between what technology can do, and what it is actually doing right now.
Australia has fallen behind much of the Western world in terms of hospital digitisation; while the US has around 2000 “digital” hospitals, Australia currently only has three.
But significant work was under way to push many more hospitals into the digital age, Richard Royle, the national digital health leader at PricewaterhouseCoopers Australia, said.
Australia’s first fully integrated digital hospital, St Stephen’s Hospital in Hervey Bay, Queensland, went live in 2014.
Since then, two additional hospitals have reached the highest levels of digitisation (called HIMSS Level 6): the Princess Alexandra Hospital in Brisbane and The Royal Children’s Hospital in Melbourne.
Mr Royle was the man behind the pioneering project at Hervey Bay.
As the former executive director of UnitingCare Health, Mr Royle transformed the Queensland facility using $47 million in funding from the former Gillard government.
Gone were the laborious, time-sapping and inaccurate paper records at this “hospital of the future”. Instead of clinicians at St Stephen’s having to write down the heart rate, blood pressure and temperature on the patient chart, data was automatically fed from monitoring systems into the electronic medical record.
The system nudged clinicians towards best practice by displaying standardised clinical decision-making pathways, Mr Royle said.
Medications used at St Stephen’s were packaged in individual, barcoded dosages and fully tracked to reduce waste and medical errors.
Doctors could also receive alerts about their patients in real-time, including information on allergies, medications, abnormal vitals and test results.
“The ability for the clinician to look on his or her laptop at home for the live vital signs, to be able to review and change the medications, to be able to talk directly with the staff who are treating the patients, that’s gold for the doctors,” Mr Royle said.
While Australian hospitals weren’t exactly embracing a digital future yet, there was significant traction occurring, particularly in NSW. But getting digital projects off the ground required engagement at all levels – from patients, to clinicians, up through management and inside parliament, he said.
A major obstacle was the cost. At around $50,000 to $70,000 per hospital bed, the price of digitisation remained steep. Mustering the political will to fund these projects was difficult, as the investment would only start produce efficiencies well into the following election cycle.
The government was also nervous about disasters in digital health because these could be very high profile and they hurt people, Mr Royle said.
The US was so successful in rolling out digital health systems in hospitals because it created incentives for reform through its “Meaningful Use” regime. This was a set of criteria, included in former president Barack Obama’s Affordable Care Act 2009, which required providers to show they were using certified electronic health record technology in ways that could be objectively demonstrated to improve quality, safety, efficiency, and to reduce healthcare disparities.
Mr Royle has advocated in the past for a similar scheme to be introduced in Australia to increase the uptake of electronic medical records.
Digitisation also required leadership at the highest levels of a health organisation. “You do need a strong CEO to drive this change because it is so fundamental to the way you deliver care,” Mr Royle said.
This leadership was sorely lacking at the hospital level in Australia. But Mr Royle said a network of digital health leaders around the country were looking into encouraging professional development. Health was an “ultra-conservative industry”, and clinicians would respond more favourably if the direct benefits of the technology were clear from the outset.
Digital health would be “pushing uphill” unless the staff had a positive attitude towards change and were engaged throughout the process.
“Change management is 101 here,” Mr Royle said. “But we’ve shown it can be done, and it can be done successfully. There is proof in the pudding.”
Artifical intelligence
Another compelling driver behind pushing forward with a digital future is the explosion of medical literature that a clinician must now read to stay up to date – requiring, by some estimates, as much as 29 hours a day.
Supercomputers such as IBM’s Watson, which could digest millions of research papers in minutes, are an obvious solution. “Augmented intelligence and Watson-type capability enables [clinicians] to have that wide breadth,” Annette Hicks, a senior adviser at IBM Australia, said.
Watson could aggregate data from structured and unstructured texts, it could reason, and it could be trained to support real-world clinical decisions. “It’s a very useful tool,” Ms Hicks said. “We’ve obviously targeted areas such as oncology. We’ve also done life sciences and we are starting to go into the world of imaging and aged care. But I think this is just the beginning of artificial intelligence in healthcare.”
However, the capabilities and potential of such technology is also creating roadblocks. While Watson has been touted as the ultimate diagnostic tool, its actual performance has disappointed some.
“Lately, much of the press for Watson has been bad,” MIT Technology Review reported in June.
The company suffered a hit when its partnership with M.D. Anderson Cancer Center fell apart earlier this year. IBM’s stock price fluctuated and some analysts started to publicly question the value of Watson.
Thanks to an overly enthusiastic marketing campaign, many had expected Watson to be radically changing clinical practice by now.
But to learn how to correctly recognise a disease, Watson needs specialists to feed it high quality data. Accruing that data takes time, money and collaboration with clinicians.
Where those prerequisites exist, machines generally surpass clinicians in diagnostics.
For instance, a machine trained by the CSIRO to detect diabetic retinopathy could already make a more accurate diagnosis than an ophthalmologist, Dr Rob Grenfell, the director of CSIRO Health and Biosecurity, said.
The system works by analysing an image of the patient’s retina. Within 30 seconds the clinician is informed as to the severity of the disease and whether a referral is necessary.
Medicare item numbers were launched last year to support screening for diabetic retinopathy in general practice, which affects almost all patients with type 1 diabetes, and more than 60% of those with type 2 diabetes, within 20 years of diagnosis.
“It is very challenging to a clinician to hear those words: ‘far more accurate than an ophthalmologist’,” Dr Grenfell said.
“‘You are telling me that I did 20 years training to become a specialist and the machine can do it better?’ Sorry, yes. You need to actually modify your role.”
Dermatologists might also be in for a shock. Another tool, currently in the clinical validation stage, used a vast library of images and a smartphone to analyse skin lesions, he said.
AI-powered “chat bots” for mental-health conditions weren’t far away either.
“The psychologists will get upset and say, ‘How could a computer do this?’
“Imagine this scenario: you wake up in the morning and you get told by your special friend, ‘Hey Rob, I notice you’re a little bit anxious this morning. Do you want to actually go through that?’ And it actually takes you through some staged, graduated ways of dealing with your anxiety through the day.”
This kind of machine learning also raised new regulatory challenges, Mr Grenfell said.
“Think about my mental-health chat bot. When that goes out and it’s learning about Rob’s characteristics, it’s redesigning and relearning and it’s not the same in a year’s time. So, you can see where this stuff really messes with regulators.”
Artificial intelligence had been around for a long time, Dr Grenfell said. And history showed that doctors got very stroppy every time clinical activities were outsourced to a computer.
“We saw the introduction of supporting ECG machines and as junior doctors,” he said. “We used to revel in finding the machine wrong and justifying our existence.”
At this time, junior doctors were required to understand the electrophysiology of the heart and the inner workings of an ECG machine (information that is mostly redundant today).
“It took time for people to accept an automatic ECG would give them a very accurate result,” Dr Grenfell said.
Dislike of AI among healthcare workers is, in part, driven by fear of losing funding or employment. No-one likes their mistakes being recorded and watched either.
But the biggest pitfalls for AI were the little, daily frustrations associated with using new technology, Cathie Reid, digital adviser at ICON Group, said.
“Clinicians are busy, busy people,” she said. “If you have to give them six new passwords and six new systems to log into, no matter what the patient benefit is, they just don’t have the time and the ability to do that.”
Watson for Oncology, IBM’s cancer AI product, made its debut in Australia in June through a partnership with ICON Group.
ICON planned to combine the technology with telehealth to keep cancer patients closer to home in rural and remote regions. Where oncologists weren’t thick on the ground, digital health could connect patients with specialists, Ms Reid said.
“[This way patients can] work through that initial part of their diagnosis and their treatment planning while they are close to home with the support of their friends and family around them, and often their trusted GP that they may have known since they were a child,” she said.
Around 50 healthcare organisations around the world use the Watson for Oncology platform. The system was trained by oncologists at Memorial Sloan Kettering Cancer Center in New York, and continues to learn through its interactions with clinicians. It can adapt to the clinical protocols across different geographic areas.
Relying solely on one’s capacity to read journal articles and attend conferences instead of using AI tools would simply be unacceptable from a medicolegal perspective in the future, Ms Reid predicted.
“There are vast quantities of data that no human can actually assess. In 10 years, clinicians who are not using augmented intelligence will be the dinosaurs and the ones who are outside the norm and the medicolegal system will reflect that.”
Big data in the US
Experts would readily acknowledge that the Australian healthcare system was far ahead that of US on a number of metrics, Dr Dora Hughes, the former health adviser to former US president Barack Obama, said.
The US always placed last on The Commonwealth Fund’s international comparison of healthcare systems. “But I think, in some ways, that has helped to spur a lot of activity in the US,” she said.
When it comes to collecting and using big data, the US is even a few steps ahead of Australia.
“The US in some ways got a kick in the pants, if you will, about the state of our healthcare system. And perhaps that helped to jumpstart some of these efforts to both collect the data and to better analyse the data.”
The US has a greater uptake of electronic medical records, and has comparison websites across almost every sector, including hospitals, nursing homes and physicians.
“So that is a very aggressive area, and one that has been supported to different degrees across political parties and across the different congress and administrations,” Dr Hughes said, who is now a senior medicine and health policy advisor in Sidley’s Government Strategies practice team.
And while the Trump administration has attempted to demolish Obamacare, digital health has largely escaped the President’s attention.
Landmark endeavours, such as the All of Us Research Program, which aims to collect genetic and other information from one million volunteers, are still operating despite proposed budget cuts to the National Institutes of Health.
Americans had a very conflicted attitude towards data privacy, Dr Hughes said.
“We have a very active and vigilant privacy set of stakeholder groups, and at times it can be fairly torturous to obtain data.”
The US strengthened its privacy requirements in 2009 as part of its push for the widescale adoption of electronic medical records. “It was a huge deal,” she said.
“And so that gave some solace to those that are concerned about privacy.”
Under the Meaningful Use provisions, patients became entitled to access their records, at substantial cost to the healthcare provider. However, few patients that have gone back after the initial sign up to retrieve their health data. The records must be either too onerous to access or not useful enough for patients, Dr Hughes said. “It is an area that remains hugely concerning to providers,” she said.
The US was very much awake to the potential of digital health, she said. It was now about finding the “carrots and sticks” to compel better performance using data.
“The leadership is there, the engagement is there. And we are pulling our providers along, some kicking and screaming, others quite willingly, but we are in a much better space.”
