Amazon, Apple, and Google are all having another go at e-Health. But we have been here before: remember Microsoft HealthVault? It’s still around, and still hasn’t taken off. Google Health went live in 2008, but was retired at end of 2011, due to ‘lack of adoption’.
Fast forward to 2018, and we see Apple, Amazon, Google and Uber making new e-Health plays. Initially each corporation will probably work from its strengths – retail delivery for Amazon, booking for Uber, data for Google, and devices for Apple. In the US at least, some of them will build their own healthcare providers for the workforce, and use the environment as a place to work on next-generation health IT solutions. Some challenges, particularly in the devices area, will undoubtedly see progress – there is no doubt that the tech giants do some things really well.
But I remain sceptical about overall success.
The reason is that big tech doesn’t understand the e-Health problem space. They appear to think it is a computing problem, like managing bank accounts or friend networks or SEO, and a question of just applying better technology. Things they don’t understand include:
- privacy: the interests of consumers with respect to data in the healthcare space are complicated by the tension of needing open access to health data among current carers and data partners (e.g. pathology labs testing your blood), but strong privacy outside the care loop, including potentially from family, employers, and commercial data users. Not to mention national legislation generally banning the siting of patient data outside the country of residence. The big tech companies don’t have a terribly strong grip on this kind of privacy. Certainly Google would have its work cut out for it to convince healthcare professionals, and Alexa and Siri are enough to make many dubious of claims to data protection by Amazon and Apple.
- patient records cross enterprise boundaries: the tech giants will no doubt realise that patient data crosses health provider enterprise boundaries, but in their efforts to each be the e-Health solution of the future, they will lock patient data into their own walled gardens, creating a new version of the same problem.
- ethical data use: as the saying goes in tech, if you’re getting it free, then you are the product. Google is the best known for secondary use of data, although not many people really know exactly what they do with our data – but mining our every purchase, movement, and what we type in browsers is how they earn their vast wealth. Would we trust them to treat our health data differently?
- semantics: all the tech companies think they can solve everything with corpus mining, machine learning, trained AI algorithms and other brute-force methods of extracting intelligence from noise. Because they are so addicted to brute force computing methods, they don’t think they need to understand the semantics of any domain. Yet anyone who has used Google translate for 5 minutes will know how far these methods are from anything resembling intelligence or fidelity. Consequently, they don’t even employ the right kind of people to help them understand and construct the kinds of solutions that might help. I therefore doubt they’ll ever learn what an EHR really is, or even get to grips with the sheer scale of content semantics, terminologies or ontologies in the health domain.
- clinical community: solving the semantics problem, and numerous other challenges in the e-Health space absolutely requires a close relationship with clinical professionals. It’s the only way to find out about clinical processes, information governance, and how exceptional situations are managed. Big tech is starting at ground zero here.
The problem in e-Health is that you have to create coherent, long-term, patient-centric information based on 10’s or 100’s of thousands of domain information elements, underpinned by terminologies and ontologies, and make the resulting records outlast all applications, OSs, DBs and other technology. This is hard to do, because the information is created during complex processes full of exceptions to rules, and routinely crossing enterprise boundaries.
Solving it requires design, and part of that design is to provide a way to formalise the semantics of the domain into artefacts that can be used at runtime in the workplace. But these artefacts can only be built by domain professionals who understand their information and workflow. And that requires advanced tools and model-based engineering.
I predict the most likely outcome of the current gamble(s) to be:
- patients will have some healthcare and wellness data stuck inside Amazon or Google data siloes and still have acute episode data still stuck inside today’s hospital EMR siloes.
- competing booking systems, creating the same problem when trying to find a movie – is it on Netflix, HBO, Blinkbox, or where? Did I buy that service?
- a data quality and coherence nightmare – data will be dumped into your record by shiny devices, but since big tech doesn’t get semantics, they will have failed to design a meaningful health record, and the data won’t integrate longitudinally, nor will BPs from two devices be recorded the same way.
- lack of content standardisation, due to a failure to appreciate the need to model it, or to build the tools to do that job – there will be a dozen ways BP is represented. Given that we know of O(10k) information elements types that are needed, the size of this problem will be enormous, and querying will work poorly.
- lack of adoption: due to a failure to understand that modelling of semantics can only come from domain professionals, the projects will hit a wall once they have dealt with devices (which big tech understands well enough) and try to move onto clinical recording in general – there will be no people or infrastructure to do the modelling.
I expect at least 5 years of flailing around before those running the e-Health programmes inside the big tech corps get even an inkling of the category of problems they need to start dealing with. At which point they will need to start hiring some people who actually know a thing or two. And then start all over again. Or maybe not…
Fabulous insights brought together no neatly. Thank you Tom.