Sonoa Health's plans to outdo Dr Google and take on the world
Melbourne-based Sonoa Health isn't the first digital health company in the world to proclaim it has the potential to take on the power of Dr Google or to become the Facebook of personalised healthcare – Google itself tried and failed with Google Health after all – and it probably won't be the last.
It is also not the first company to enter the consumer electronic health record market or to develop a site where live data from fitness and medical devices can be stored, graphed and analysed. Another giant in Apple is already investigating that.
And nor is it the first to work with computerised support and algorithms that can allow healthcare providers to triage patients or provide information and advice through an extensive library of medical literature. There are some major players in that field as well, including major health insurers.
What is a first is that Sonoa Health - now known as Health& - plans to bring all of those technologies and capabilities together into a single consumer-focused platform, with personal health record capabilities and APIs for apps and devices, underpinned by an intelligent machine reasoning system that can learn and produce individualised information, and which is also easy to use, can link to the PCEHR and can even tell you when and why you are due for that long-dreaded colonoscopy.
It might sound grand, or even grandiose, but Health& – which is also the name of its master product (pronounced 'health and') – has not only attracted some serious private investors but a seriously well-credentialed medical advisory board as well, including Monash University pro-vice chancellor and medical educator Leon Piterman, fertility specialist Gab Kovacs and the GP's GP, John Murtagh.
It already has some health literacy and educational materials that are available for professional or commercial use at its Health&education site, including some nifty animations that explain a health condition succinctly and attractively. Next month it will launch the first stage of its consumer portal Health&, which will include a content search engine, health FAQ service, and an opportunity to sign up early for a personal EHR launching later in the year.
In its early stages, the consumer portal will contain a content library, personal EHR, preventative health alerts, ‘find a facility’ service, health FAQ service, health news, medicines information, and the capability to upload live data from fitness devices like Fitbit or medical devices like glucometers and track measurements over time. Importantly too, it is also authorised to link with the national PCEHR.
The portal will also contain search functionality that Health& says will out-Google Dr Google and provide medically validated, easy to understand, reasoned answers to healthcare questions. The technology will be able to learn about who you are, your medical history and your current conditions, and provide individualised healthcare information. That, in turn, will hopefully eventually provide the opportunity to go after the holy grail of contemporary healthcare – prevention.
Health&'s plans are big and expensive, but its strategy has been well thought-out and the time seems increasingly ripe for what it plans to do. It currently has a team of over 50 doctors, communications experts,, writers, animators, IT experts and business strategists beavering away at its offices in the Melbourne CBD, getting set for its major product launch in November.
It even has a new CEO in former investment banker John Stewart who has done a bit of a Victor Kiam, liking the company so much he has invested in it. At the heart of the company, however, is founder and chairman Bob Biddle, a second-time entrepreneur who has invested a great deal of time and money, much of it his own, in an idea that he has been thinking about for several decades.
As Mr Biddle explains it, the idea for what Health& is trying to achieve has been kicking around for decades, when artificial intelligence began to be applied to many different industries. Mr Biddle himself made his name building a nurse employment agency called Origin Healthcare, during which he designed a complex artificial intelligence system that could match up the right nurse with the right patient and streamlined the placement of nurses from weeks to minutes.
Origin Healthcare had 30,000 people on its books and was the largest agency in the country when Mr Biddle sold the company in 2004 to Skilled Engineering. After overseeing the transition for a year or two, he then took some time out to seriously think about putting into action an idea he had been interested in for many years: how to solve the problems of an ageing population, experiencing increasing ill-health through preventable diseases, with fewer healthcare workers and scarcer resources.
“The only way we can solve this is to make people experts in their own health,” Mr Biddle said. “If people are responsible for themselves we can then say you own it, you're responsible for it, you're in charge, but the foundation of all healthcare is information. Without information, doctors can't diagnose or treat or give advice, and people can't participate in their own healthcare.
“The only way we can make people experts in their own health is through technology, but we have to invent something to do that, and that requires a machine language.”
Mr Biddle set out to build artificial intelligence into a system that could reflect or mimic how a doctor reasons through patient variables, and for this an expert system was required. This is a term first coined in the 1950s that describes how a systems engineer can harvest the knowledge and experience of an expert, put it into a programmable language and then build a knowledge database.
“It works when you've got mathematical predictability,” Mr Biddle said. “For example, you go to a bank for a loan and they take certain factors into account and quickly come up with a figure that they will lend you. This is where expert systems in the past have been used.
“But medicine is unpredictable – you have a certain condition and they treat you but quite often they don't know why it works. All they can say is that on the basis of the evidence, if we do this most times it works. So there is no real mathematical predictability and that is why there have been thousands of people working on this for decades, trying to crack this thing, and it's an impossible process.”
What he was looking for was a way to ask questions of a machine so that it could provide individualised answers based on known factors but that could also learn from the individual. For example, a person asks a medical question, the machine searches its expert database and comes up with an answer, but that answer is tailored to the individual based on factors such as age, gender and medical history – just like a doctor does.
One problem is the huge amount of information that would be required to build something capable of providing answers to the gamut of potential questions in medicine. At the time Mr Biddle began serious work on it, the International Compendium of Diseases (ICD) was at number nine and contained 21,400 different topics. ICD-10 is due shortly and that has 141,000 topics, he said.
Trying to put all of this vast knowledge into a programmable system seemed impossible, so Mr Biddle decided to work on one health category to see if it could be done. As he was friends with the late IVF pioneer Carl Wood, he decided to tackle fertility as a topic to see if artificial intelligence could be applied, fertility being what Mr Biddle describes as the most complex area in medicine as it involves two people.
“If we ask certain questions in fertility, we come up with some answers, but everything would have to have a particular response,” he said. “For example, if you are a 75-year-old female and you want to have a child, the answers would come back about surrogacy or adoption.
“But if you are a 35-year-old female and you want to have a child but can't get pregnant, then there would be further questions about height and weight, because obesity might create those issues. The next question might be do you smoke or drink, then some more technical things like do you have a sister or mother who has been diagnosed with endometriosis or polycystic ovary syndrome. Then there are questions like do you have a partner with a low sperm count or with a sports injury. Then it got into things like surrogacy or egg donors – it becomes really complicated.
“I got some specialists in fertility to work with me on this process and after 12 months we decided we couldn't crack it, it was too complicated. I looked at mathematics – Bayes' theorem, heuristic analysis, chaos theory – and in the end I couldn't do it. Carl had died at that stage and I got in touch with the guy who had replaced Carl and that was Professor Gab Kovacs. So we started working on this together. He got another professor from the US and another from the UK, and worked on this for two years. After two years we managed to crack this thing.”
Mr Biddle and his colleagues developed an app for fertility that has mainly been used in the US. It contained about 200 questions in total, and experience showed that people would be asked about 70 to 80 questions in total. That would then produce a report with up to 30 to 40 pages of information, which while valuable was way too big, Mr Biddle said.
However, the information contained was helping patients in that they were better following what their specialist recommended, and participating more in their healthcare. “We know that people who participate in their own healthcare are something like 300 per cent better off,” Mr Biddle said.
So the next step was to see if the knowledge base they had built for fertility could then be extended into other areas of healthcare. “I remember [Professor Kovacs] leaving the house one night saying we have something here but it is going to take us 42,800 years to get this done,” Mr Biddle said.
Further investigation, however, showed that out of the multitudes of topics, most people searched for a far smaller amount. If ICD-9 has over 21,000 topics, the vast majority of people only searched for 2500 or so.
Even so, that would still take many years to put together all of the possible questions and answers for each topic. Taking the 80/20 rule, however, showed that roughly 400 questions were searched for 80 per cent of the time, and 700 questions 90 per cent of the time. “That becomes an achievable process,” Mr Biddle said.
What he and his colleagues then came up with is now known as the knowledge acquisition report development system (KARDS), which has a patent pending. This system not only contains the knowledge in terms of the questions about topics that could be asked and the answers to them, but it has the ability to connect one answer to another, and to take into account known information such as demographics and previous medical history. It has, in effect, the ability to learn, and acts like a machine version of a doctor's knowledge.
“Imagine you have an electronic health record system, and it had all of this data about you,” Mr Biddle said. “It would have information about your medical history, your lifestyle, information from fitness devices, medical devices, mental health information, chronic conditions, medications – anything you can think of would go in here. What the PCEHR was supposed to do.
“Imagine all of that stuff being here, and it knows you completely, and then you say you'd like to know something about menopause. This would then ask you questions about menopause, but it would be tailored because this system begins to learn about you. This is what Health& is.”
Mr Biddle differentiates KARDS from other systems like decision trees, which are used by telephone triage operators and online symptom checkers to provide answers based on commonly asked questions. He says the knowledge base that Health& is building is not only extensive but is non-linear, in that the answers are based on what has been previously asked as well as the system's knowledge of your individual details.
“When you ask a question, we already know that you had rheumatic fever as a child or we know that you are on Viagra, which you don't normally tell a nurse about. If you have a certain allergy or a certain condition, then we'll prompt you to answer some questions about that. They will be incredibly different answers from everybody else.
“As you answer, this will be populated with all of that information so we now know a lot more about you. The information is all here and it is automated so it is looking at it in a millisecond. When we then add preventative activities, it comes up with a connection where one of these prevention things should happen – 'Happy birthday, Bob, you are 50, I'll arrange a colonoscopy for you'.”
What this in effect does is lead to the ability to do predictive and preemptive diagnosis, he said. “We might not be able to stop people getting a disease, but the earlier we get to them the better it will be. Predictive and preemptive diagnosis is the holy grail of big data, but it has to be intelligent big data. This is intelligent data looking at intelligent data.”
Mr Biddle and his team worked out early on that in order for consumers to be comfortable using the system and understanding the information it would generate, it would have to be provided in easy to understand formats. Doing a Google search on a particular topic or condition, for example, throws up anything from simplistic or incorrect factual information to some incredibly complex academic papers.
What his team of doctors, writers and animators have come up with is a large library of 40- to 60-second animated videos that explain conditions and diagnoses in easy to understand ways. These animations are being tailored to different languages and cultures and will also have a health promotion message that Health& hopes will lead to better preventative healthcare down the track.
The team has produced about 400 of these so far and is aiming for several thousand, accompanied by written reports on each topic that will be added to the Health& consumer portal but are also currently available on the Health&education website and will be offered to healthcare organisations on a commercial basis for patient education. The company is also planning to work with schools in Victoria to use them for health education.
The next step is getting live data into the system, which is where medical and fitness devices come in. Health&'s chief information officer, Prerak Joshi, said the company would partner with medical device manufacturers to build APIs allowing data to be uploaded to the consumer's individual portal, data that the system would be able to mine for information on the user and which can also be shared with healthcare practitioners.
There are also plans to link with the PCEHR. Health& has been registered with the Healthcare Identifiers (HI) Service as a healthcare provider organisation and received an HPI-O, meaning it will be able to access the user's PCEHR and drag in data that holds as well. The plan is to allow consumers to register for a PCEHR through the system too, with Health& working on developing an attractive user interface to the PCEHR that the system currently sorely lacks.
Mr Biddle is a firm fan of the PCEHR and believes the foundations for eHealth that NEHTA has laid down are incredibly valuable. “It is still early days (for the PCHER), but to get to where we are now, we are number one in the world,” he said. “If we can marry our system with public systems like the PCEHR, that would be remarkable.
“What that does, once we get the information, is we can then get to individualised disease prevention. Once you get to that, then we can enhance healthcare.”
The company is also working on gamification to add to the attractiveness of the platform and keep people coming back to use it. The underlying KARDS technology is being called a 'dynamic digital doctor' to make the concept understandable to consumers, and Mr Biddle's team of digital experts is working on the system's interface leading up to its launch in November.
The big question of course is how the system will pay for itself. Mr Biddle is clear that he doesn't want to go down the road of other sites like WebMD and use an advertising model, as he does not believe this does anything for a site's credibility. Instead, Health& will be offered on a subscription basis of about $100 per year per family or $60 per individual, with a free option also on offer.
Mr Biddle expects everyone to go for the free option – it will provide consumers with pretty much all of the functionality of the subscription model initially – if the consumer agrees to let the company use the data entered in a deidentified way.
“The people who are interested in that data are governments, who spend about 1.6 per cent of their healthcare budget on collecting data,” Mr Biddle said. “Governments are already paying hundreds of millions to get data from organisations like the Australian Institute of Health and Welfare, but their data collection is done every two or five years – this data is live. You will be able to see things changing in real time.” Deidentified data will also be used to support medical research efforts.
While the company is starting off in Australia, the bigger markets will be Asia and the Middle East, Mr Biddle said. He plans to make Health& an international concern, collecting and offering healthcare data to governments and private industry, backed by the credibility the company's links to local universities and research institutes brings.
According to the company's chief communications officer, Haley Price, the initial campaign will target four consumer groups: first-time mothers, Australian families, people with chronic disease and 'quantified selfers' or the tech savvy.
“There will be a trial component for advanced chronic disease management, which we anticipate that will require quite significant data to roll out,” Ms Price said. “Trials will be starting within three months of the launch and we have partnered with approved manufacturers of monitoring devices and we will be measuring the impact over time.”
What consumers will see is on the site is their own personalised dashboard, featuring their own EHR, and the dynamic digital doctor search function, which will contain general reports on medical conditions and diagnoses as well as the animations.
For Mr Stewart, making the company a going concern financially is the least of his difficulties. “The hard part is coming up with something that is truly globally innovative, that's like an iPhone,” he said. “The key thing for us to focus on at the moment – because the business will do fine once we are out there – is creating a product that is truly innovative, that consumers are going to absolutely love.
“Our key point of difference is not putting your information into a bucket. It is actually telling you what it means. It is correlating all of the information in that bucket and telling you what it means and what you should do about it.
“There are millions of people building buckets, and there are millions building stuff to pour into that bucket, but there are very few people building the analytics that tell you what does that mean.”
Posted in Australian eHealth