What is really unfortunate and not communicated in this article is just how much Google's various star crossed forays into healthcare took oxygen from other fruitful and deserving efforts. It's an indictment of healthcare institutions as well for falling for it every time one of the tech megaliths promises the moon.
At ClearHealth we lost many, many potential projects, partners and dollars over the years to "we are going to see what google is doing", "we are in talks with google", etc.
The damage is very large but difficult to calculate.
I hear ya - at UNH if I had a dollar for every time I heard "We are in talks with google to do X" I'd be a very rich man. For years Google Glass in healthcare were the buzz words everyone wanted but could never put together. Google FHIR integration was always just over the horizon. Not surprised to hear their healthcare efforts are folding but like you I feel the disappointment.
Rule of thumb - ignore what Google does. Their lack of focus and proclivity to murder projects that are not runaway success without years of hard slogging has a solid track record.
I can't even write a simple comment about health tech without going on for multiple paragraphs. Truly, it is a wicked problem.
Have said this before, but healthcare is the Afghanistan of tech where giant empires go in thinking they can solve it and leave beaten years later wondering what happened. We're still about 5 years before a viable tech gets traction that could be useful in that space, and interestingly imo, one of the YC 2021 batch is the most well positioned to do it because they are doing authorization as a service (warrant.dev), which could meet the basic delegation need which is the pattern the whole health world is predicated on. We do collaboration pretty well with tech and we're just starting to make headway on identity and access control, but we haven't figured out workflow delegation and designation in any meaningful and open federated way, which is the necessary condition of the basic physician/nurse, physician/specialist, physician/hygenist, physician/orderly etc. workflow pattern.
What people don't get about healthcare is that it is not a vertical or a sector, it is a parallel planned economy of autonomous entities, not unlike higher education but more distributed. I have done a lot of consulting work in a very large single payer system over the years, and the reocurring pattern is that even government thinks it's an enterprise, or a collection of enterprises. It's really an archipeligo of federated diverse organizations that orbit doctors and other health care providers. When you think, "oh, I've got a document management system or search for health!" you don't unless it is document management or search for a hyper-federation (like a hypergraph). You don't build products for this, you build tools, and the "healthiness" of a given tech is about how it supports this complex and dynamic relationship pattern.
If I were google developing technologies for health, I'd be solving the basic hyperfederation problems. I hate to say it, but from a technology perspesctive, it's very likely a blockchain problem, where the health ontology (HL7,FHIR, etc) gets distributed via the consensus, and individual records would be graphs over its elements. Unfortunately, HL7 suffers from the same architecture complexity problem, where complex standards and rules yield stupid behavior in the field. This is to say the problem is health information techs all fail for the same reason, which is the basic problem of imposing central or top down design models on a stubbornly bottom up economy full of non-technologists.
> "healthcare is the Afghanistan of tech where giant empires go in thinking they can solve it and leave beaten years later wondering what happened."
Great metaphor. It's kinda like Education: if you come in thinking you're going to make money, all the existing players say "Hello, sailor!"
I joined Google in 2005. Even then, executives who were out of favor were sent to the "health care division" which was commonly understood as the one-way door out of Google.
> It's kinda like Education: if you come in thinking you're going to make money
Your statement can stop there.
There isn't some huge amount of money sloshing around in education when contrasted to healthcare.
In education, you're fighting over pennies. If you want money in education, you have to target the "tutoring" level which is boutique and caters to those with extra money.
Right. I knew a researcher at a well-known company who was developing software for kids. His attempts at getting outside funding were usually met with "This is great, I'd love to have this for my kids! But funding it? Let me get back to you on that."
Having worked on the government side of healthcare I can say all these systems support interoperability when money is on the line. That’s what tech giants don’t understand, every system is trying to create their own ecosystem and fighting each other in the process. The only way it works is when mandates have actual teeth, otherwise it’s its exactly as dysfunctional as intended.
Good points in here. If you want to read a good book that discusses all the different players in healthcare, i recommend “An American Sickness” by Elisabeth Rosenthal.
It does a great job pointing out the various players and the conflicting interests between them. Highly recommend for anyone looking to learn more.
This is a very salient comment. I wrote "Hacking Healthcare" and created ClearHealth/HealthCloud. If you are in any way looking for opportunities I'd love to chat with you.
That's because the poster probably thought that `tr` works like `sed`, replacing the string "spam" with the string "du" to produce their email address, du@50km.com.
That's not what `tr` does. It replaces each letter S-P-A-M with the letter in the same position in the string D-U. Because "du" is shorter than "spam," `tr` just reuses U for letters 3 and 4. So the mapping is S->D, P->U, A->U, M->U, making "spam@50km.com" into "duuu@50ku.cou," and the -s flag then compresses multiple repeated U's into one.
A few other lessons from this: 1) test your scripts, 2) be clever enough, 3) email obfuscation stopped working in 2000 for preventing spam, 4) handle spam at your receipt
What, elementary proficiency at the Linux command line of the kind that's exercised by boys playing with computers in middle school? I'd like to think the peak is somewhere higher.
> but we haven't figured out workflow delegation and designation in any meaningful and open federated way, which is the necessary condition of the basic physician/nurse, physician/specialist, physician/hygenist, physician/orderly etc. workflow pattern
This sounds like standard delegation and the facade pattern from capability security. Software lost a lot of time because we didn't go with security models built for delegation from the get go. People ended up rediscovering then when the web exploded and wanted to share things securely.
Your federated network is an interesting problem. Capabilities are suited to the core delegation and authorization aspects, but there's obviously a whole lot to be layered on top that is no doubt challenging.
>Have said this before, but healthcare is the Afghanistan of tech where giant empires go in thinking they can solve it and leave beaten years later wondering what happened.
Very rightly said.
However, that's also because of the traditional resistance of the healthcare to adopt (and adapt technology) to improve processes/ workflows. As well as onerous regulations.
Current blockchain technologies aren't scalable enough for storing clinical data. The volumes are just way too high. And blockchain doesn't really give us anything we need. There are only a limited number of payers and providers, and they can all trust each other to an extent based on legal agreements. Or at least they can trust a smaller set of organizations which act as centralized data aggregators and network hubs. For compliance reasons we also need the ability to permanently and completely delete patient records, which is problematic with blockchains that only allow appending blocks.
HL7 standards don't really dictate any particular architecture, they just define public APIs and interoperability formats. The data models are inherently complex because the problem domain is complex; attempts at simplification tend to cut off important use cases.
I'm suggesting blockchains for the data dictionaries and ontologies, catalogs, schemas, service billing codes, etc. because they need to be looked up and verified often, but written much less often. Clinical PHI doesn't need blockchains (other than for say, identity potentially), but I think the actual standards, APIs, and practices may benefit from it.
I still don't see the point. There is generally a single trusted source for code systems, ontologies, data standarda, etc. You can download local copies of anything that you need to frequently access.
Techbros are like moths to a candle when it comes to healthcare. There's a vast amount of money in it, a lot of the processes appear to be (or are) clunky and crufty, and you get to pretend you're just out to save the world while you make your first billion.
But the unfortunate reality is you can't just go "hack stuff", disrupt some paradigms, craft a few apps, throw in some AI, sprinkle some crypto around, code it all up in the cool language and technology stack of the month, and come up with something useful. It's long hard thankless grinding work that's frequently blocked by regulations, risk adverse institutions, interoperability with legacy systems, and people and business processes more so than simple matters of technology.
You also have to accept you frequently won't be the smartest person in the room when you're dealing with healthcare professionals, which is real kryptonite to a lot of move-fast-and-break-stuff types.
Turns out Google culturally has a hard time playing nice with other large partners? Color me shocked.
Minor story as perspective. Initiative was hatched to extend Google Maps Street View inside of buildings. High level meetings were held. No technical roadblocks.
Deal eventually killed, because Google was unwilling to provide their raw data to the buildings' owner, for their own uses.
You're hamstringing a feature on your own app because you don't want to ship the data you collect to the person you're collecting it from? mind boggles
Google’s Sidewalk Labs “smart city” initiative in Toronto was canceled for similar reasons. Google wanted to collect all the data, refused to share any of the data with the municipality and refused to take any measures to address the privacy concerns. Google’s arrogance and their refusal to work with stakeholders in a productive manner insured the death of the proposal.
There was an excellent article [1] in The Atlantic about the Books project. It was really a tragic situation which no one in Congress of either party is interested in fixing.
Full disclosure: I joined Google in 2005 when this was still an active project, and in fact walked around GWC-3 and saw the Yellow Badge people (great symbolism there) taking their breaks. These were contractors who actually scanned the books for Burger King-level wages, and whose badges wouldn't even let them into the micro-kitchens.
Anyhow, I don't know if the University of Michigan (early volunteer) was supposed to get the scans. Supposedly, the whole world would. That didn't happen and you can read [1] for details.
As of 2015, at least, not even a Google employee could get at them. I was told I could go to someone's desk and read the book on their computer if I really wanted to see it (this was for patent research).
The universities did get the scans. They're in the HathiTrust project. The public gets access to scans of out-of-copyright works. University faculty, students, and other researchers can get access to in-copyright works too.
For the general public HathiTrust is like Google Books with worse search and worse UI but more books readable in full. Sometimes I search Books and then go over to HathiTrust when I find that Books has unreasonably restricted viewing of some search hits.
The initial goal of Google Books was to make the scans publicly available to everyone, but then the copyright system got in the way. Blame our politicians, not Google, for that one.
Depends on Google's philosophy. Are features a means to an end or the end itself? It's not surprising that Google would decide not to do something if it doesn't serve their goals.
What were you doing to reach the API limit? Were you circumventing the terms of use?
Also, what do you mean, "it's my device. Don't rate limit me"? Do you own Nest's network? If not, why are you mixing owning a network with owing a device?
To be fair this is what happens when you hire entirely from the technical Ivy League. No one wants to do the last 10% that's really the last 80% of the work because it doesn't carry the same visibility.
For what it's worth, Google, or at least some subsegment, does seem to be trying to fix this issue. I had a somewhat offensive conversation with a recruiter recently where they told me they're looking for people (like me) with non-standard backgrounds who don't necessarily have the pedigree that the usual Google SWE hire has. I hope they also extend this thinking to moms with young children. I know a good many technical women who fell off the career ladder after having children and were never seem to be able to get a chance to get back on, ultimately taking jobs far beneath their abilities or just shifting industries altogether.
They’ll give interviews to most and downplay the difficulty of the interviews on social forums in order to keep the accept rate low. You have to be an intellectual elite in order to pass the interviews either way.
Thanks I was wondering too. I think it means “it happens partially” but that can’t be right as it doesn’t seem noteworthy. Although glad that GP isn’t breaking NDAs with that fortune teller statement.
The road to healthcare tech is littered with bodies of failed efforts from all kinds of companies. Seems like a particularly hard segment to break into. I'm not exactly sure what the biggest reason for that is. Regulations? Social relationships/politics? Difficult to understand processes and landscape? Resistance to change? Maybe a mix of all of these?
It's very similar to education. Neither health care nor education function anything like a "free market" so the normal business strategies don't apply.
And it's so many things -- it's regulation, it's slow-moving, it's hyper-local and non-uniform, it's privacy concerns, it's vested interests, it's interoperability, it's political.
Both health care and education have huge amounts of money poured into them -- the money spent on a child's education per-year is orders of magnitude greater than a Netflix or Spotify subscription. Same with your monthly health insurance costs. So despite all the challenges I listed above (and more), it still seems like if you can pull it off, it's a bet worth making. But that doesn't stop it from being really, really hard.
These are questions that are worth delving into more deeply. In my brief period at IBM I was adjacent to the Watson Health folks and the challenge, to my unsophisticated eye, was greed. Basically the "Health Care System" in the US appears to be completely corrupt[1]. Thus any proposed change the is presented based on how it will "Improve the system" is immediately suspect by those for whom the unimproved system is paying them a hefty paycheck. In the two (and be aware there were thousands so this is an anecodtal sized sample set) engagements I had visibility into with Watson Health teams, the one that went in with "This will improve patient outcomes" got shut out, and the one that went in with "This will increase your net margins" was welcomed with open arms.
I'm not sure how one unties that particular knot.
[1] Corruption here being that the primary mission is the collection of monies and health outcomes are a tolerated side effect of that.
> Thus any proposed change the is presented based on how it will "Improve the system" is immediately suspect by those for whom the unimproved system is paying them a hefty paycheck
Sadly, my observation is that this is the rule (and not the exception) in many industries.
"It must be remembered that there is nothing more difficult to plan, more doubtful of success, nor more dangerous to manage than a new system. For the initiator has the enmity of all who would profit by the preservation of the old institution and merely lukewarm defenders in those who gain by the new ones."
In my view it's not a problem specific to any industry or domain, but a fundamental problem of social organizations: the clash between individual (employee) interests and organization interests.
When I tell my wife about something like this conversation goes something like ...
Wife: Why would they do something like that
Me: You know why
Wife (sheepishly): Money
Me: ding, ding, ding
This is why the industry is gradually shifting from a fee-for-service model to a value-based care model. It doesn't make sense to pay providers for performing procedures. Instead they need to bear some financial risk and focus on producing positive outcomes.
The long term trend is towards consolidation into combined payer + provider organizations like Kaiser Permanente. All the large medical insurers are hiring or acquiring more doctors to bring member care in house and better align the financial incentives.
Outcomes are tied to margins in the end anyway. This was part of the ACA for improving readmission rates. Vertical integration works as well if you look at a place like Kaiser as well does that.
If Apple and Google are both ditching health projects simultaneously, I wonder if that suggests that they have advance warning of some change to the regulatory climate.
Apple never really tried to compete directly against Epic. There is only a very small overlap in the PHR space, which doesn't generate much revenue for either company.
I think the problem is that we try to simplify healthcare - like it is vertical. Healthcare is more like world on it own. The only way get into “healthcare world” is to attack a very specific niche but the niche is too small for big companies. So the companies decide to “boil the ocean” - which always fails.
All of the legal and compliance requirements make it extremely expensive to build new healthcare software. Vendors have to meet a long and growing list of mandatory requirements before they can even start writing new business or clinical functionality. Customers have been burned so many times by new vendors which failed to deliver that they're hesitant to take a risk on start-ups.
But I think there is a growing opportunity for innovation in the new SMART on FHIR app stores run by several major EHR vendors. They're copying the Apple / Google app store model by allowing ISVs to use their platform in exchange for a cut of the revenue. In theory you can write an app once, then deploy to any standards compliant EHR (reality is a little more complex). So the platform handles all the compliance stuff and you can focus on innovation. I expect we'll see a wave of start-ups targeting that platform over the next few years.
I've seen people do well by targeting specific processes around an illness. IMHO the landscape is so fragmented that a holistic approach would never make sense except as a government initiative.
>I'm not exactly sure what the biggest reason for that is
I suspect healthcare in the US (and probably elsewhere) has to deal with a lot of liability, and so processes that fly in other situations do not fare well in healthcare. Everything i has to be dotted and every t crossed to make sure you do not end up with any part of the culpability for an adverse healthcare outcome, as well as liabilities from laws such has HIPAA.
I had a colleague go work for Google health in a fairly high role, but I didn’t know them well enough to ask why. Seeing what they did, it was a lot of positive sounding, but vapid, content put out and I could never tell what they did.
I kept thinking, “who would work for them when they’ll just get bored of this as PR?”
I remember the 2006 era version of Google Health that also failed and the executive in charge went off to found his own health startup (which itself failed).
Hmm, I did some contracting work for a hospital system that was heavily invested in Google Health for the EHR systems. I think a lot of that work was not going to be easily moved to another platform. I wonder what happens to them now?
According to the article, it's not the products that are being shutdown, just the centralized organization structure around them. It's safe the expect that at least some of the projects will be terminated as well, but for the most part, it seems like this is just a removal of a (presumably redundant) layer of management.
Between Google's habit of handling data carelessly (mining data it shouldn't[1][2], keeping your data when you delete it[3], let others look into your data[4]), as a consumer I can't say I'm sad ; especially when you end up not having a choice, since the decision is made by your healthcare provider - and you might end up in their records without even knowing.
Not that others may necessarily be better, but they also don't have the same existing power.
I hope you people realize that Health data handling is a heavily regulated area. If you break the rules you will face penalties. Plus what kind of data expected to be shared? Your blood pressure or allergy reaction to penicillin? The real goal of the project is not to steal and trade your data but to help clinicians find consolidated data about you faster and spend more time to help the patient.
I mean if Google is serious they should buy Cerner and a few medical device manufacturers.
I don’t see federation of health records as all that interesting. Most people get comprehensive care from a major hospital.. I’ve taken records from one place to another and guess what, it’s basically a DVD of PDFs and images.
I don’t actually think it’s a real use case when you are in a network like that.
Doesn't look like it's actually being shut down. Just a re-org under Jeff Dean and some other leaders? It could have just been a leadership issue, but I don't want to suggest that Google has given up on Health altogether.
At ClearHealth we lost many, many potential projects, partners and dollars over the years to "we are going to see what google is doing", "we are in talks with google", etc.
The damage is very large but difficult to calculate.