Google Health is dead, HealthVault Indivo win

Recently, Google announce that the Google Health PHR will be retiring.

I posted the announcement to the Society for Participatory Medicine mailing list, and there has been alot of discussion about this, there. There are several issues that lots of people do not seem to understand, and some implications of this that have been missed.

Losers: Google Health Users

Let me be perfectly clear. If you trusted Google Health with your healthcare data you are screwed, unless the Microsoft HealthVault team rescues you. Even then, you are likely screwed anyways.

The whole point of Google Health was that it was more than a mere store of your XML patient data. It was a network of providers, like pharmacies, drug companies, non-profits and countless other service providers who added value to your health record.

I love the Direct Project but it only makes health data mobile, it is another matter altogether to make your health data semantically useful again. There is no way that the HealthVault team will be able to replicate 100% of the value that Google Health was providing based merely on the XML output from Google Health. As service providers and patients themselves added to their Google Health record, it made those records more complex than HealthVault, or any other PHR system, can easily understand. This is called Lossy Data Conversion.

The patients who will lose the most useful data, are those patients who leveraged Google Health the most. The more you invested in Google Health, the more screwed you are now. Of course, the other group who is going to be really screwed are the people who do not pay attention to announcements like this at all, (and ignore or filter email warnings) who will try to find data they stored in Google Health four years from now, only to discover that the deadline for data download had passed, and their data is gone. Ironically, the people who this is most likely to happen to are older people, who are not terribly tech savvy -and- who might have stored data in Google Health precisely so that they could ensure it would be available as they aged. Again the more they invested, (seemed like a good idea at the time), the more screwed they are today. Not good.

Probably the most important lesson to take away from all of this is that trusting proprietary health software vendors or services with critical health data is a bad idea. But sadly, that will not likely be a lesson learned here.

Losers: PHR vendors

Second, there is the implications for  PHR vendors. It does not look good for you. Google is in a unique position as a company. It is capable of making money giving away very valuable services, because it makes more money on advertising when someone merely uses the site. The business plan on Google Health was, essentially

“Lets spend a few 10’s of millions on this, and then make it back because a few million people will click on ads, after leaving Google Health to do a web search of some kind.”

Your average PHR company cannot make that kind of play. Most companies do not have a way to translate mass visitors into dollars. That is what makes the gmail service work for Google. Enough people click on ads through the service to pay for the entire thing for everyone. Google specifically admitted this problem in the post above with:

But we haven’t found a way to translate that limited usage into widespread adoption in the daily health routines of millions of people. That’s why we’ve made the difficult decision to discontinue the Google Health service

So if you are trying to start a PHR business and you cannot afford to give away a product you spent millions developing for years… this spells trouble. Google Health and Microsoft Healthvault together spelled the end of the still languishing dot-com bubble PHR services. I cannot imagine an investor in their right mind who would touch this space with a business model anything like Google Health.

Here is the basic takeaway from Google Health PHR:

People are not willing to use a good stand-alone PHR, even if it is free.

That word  “stand-alone” is critical.

Losers: Me

I have been wondering, as a right this, if I should be a winner or a loser on this one. I get to say “I told you so” to everyone I warned not to invest in a proprietary platform… which is fun. But I also now have to almost entirely re-write a chapter in my new book.

I (along with intrepid David Uhlman) am writing the first book on Health IT for O’Reilly media, called “Getting to Meaningful Use and Beyond“. I wrote the chapter on patient-facing software, and I featured Google Health extensively. After all, it was relevant, last week. I felt reassured after I asked Roni Zeiger a month or two ago if Google Health would survive? After all, I had heard rumors. He told me not to listen to gossip and I left feeling like my chapter would be published intact.

So much for meeting my deadline.

Winners: Direct Project

As Google Health dies it is giving a ringing endorsement to the Direct Project (of which I am a contributor). Hopefully this will raise some awareness regarding Direct as the foundation for the first generation of the Health Internet.

Winners: Microsoft HealthVault

Most of the industry pundits, like myself, have recognized for years that the “build the platform” business model that worked for Facebook and Itunes, was not going to work for Personal Health Records. Why? No “killer app”. Itunes+Ipod was the killer app for the Iphone platform, for Xbox it was the original Halo. For Facebook it was your ‘wall’, or perhaps (shudder) FarmVille.

The killer app for a PHR is dead simple: Healthcare. The two most widely used and successful PHR deployments in the country are the Kaiser Permanente and the VA’s My Healthevet. Why? You can get a message to your doctor through them, and receive replies back. They are a component of your actual healthcare. You do not have to type data into them, its just there. If you want to schedule an appointment or view your lab results you can do that. If you want to renew a prescription, the PHR can help. In short, the PHR is a workhorse in your actual healthcare process.

The Microsoft Healthvault team gets this. That is why they have been working on the Direct Project for months. They know that the Direct Project is the only way that they can have their PHR connect to -all- doctors the same way that Kaiser and the VA connect to their doctors.

Moreover, HealthVault has the only working mass-scale Direct beta in deployment: It is very likely that the only place you will be able to transfer Google Health records will be directly into HealthVault, for the foreseeable future.

HealthVault just became the 800 gorilla in the space.

My only question is why didn’t the Google leadership see the strategic significance of the Direct Project? They were obviously aware of it technically, and they usually do a good job translating technical understanding into strategic understanding.

Seems pretty simple to me. PHR usage is high -only- in systems where you can communicate with your healthcare provider in various ways. Google was disappointed by how few people were using their PHR. Direct is the only chance in hell that you have to reach every healthcare provider in the United States in the next five years. When you put it like that, Microsoft’s strategy seems pretty obvious… why didn’t the Google leadership catch on? Probably the Direct opportunity was too little too late for the internal political process at Google.

Winners: Indivo X

Indivo X has almost all of the same benefits as HealthVault (they are little behind on the Direct implementation and beta deployment), but if you actually want to avoid a repeat of the Google Health fiasco, this is the way to go. If you import your Google Health record into an Indivo X instance, you are not locked-in again.

Indivo X is Open Source, you can run your own instance if you want..

From now on, people will regard Indivo X as the safe option for PHR deployment, and rightly so, it is the only safe option. Until I can convince Sean and the rest of the HealthVault team to go full kimono, Indivo X is by far the most mature Open Source option available.

Why I do not think Google Health will, or should go Open Source

If Google drops code for Google Health, thats cool and I would take a look…  but I am not going to hold my breath.

Its pretty simple; Indivo was Open Source and available before Google Health launched. Some people believe that Google Health, like Dossia, is actually a long-ago fork of Indivo.

Indivo has moved on to bigger and better things. Indivo X, the current version of Indivo already has substantial functionality that Google Health is missing. It is already a mature codebase, with a community, and is generally operating openly as an Open Source project should. The Indivo project is not perfect, but they have steam.

Steam, motion, community, these are the things that make the Open Source garden grow.

Google Health would not actually help the Open Source community that much. We already have a better PHR project, and anything coming out of Google would compete for developers and attention with Indivo X.

Even if they wanted to, I am not sure that Google could usefully Open Source the whole Google Health codebase. Google projects often run on Googles custom, and proprietary database and network services. It is entirely possible that Google  Health would be useless without that back end.

What -would- help is for Google Health to release any components that Indivo X is missing. If they have an interesting Blue Button parser (which I happen to know they do) for instance, or some generalizable code for managing CCRs (that CCR-in-a-feed thing was a nice trick for instance…) then those components would be very useful.

Moreover, any components that would help people to parse their own Google Health data would be very welcome.

Probably the most important thing that they can do is license their API under several Open Source licenses. This way, Indivo X and HealthVault would be able to write a bridge that would allow currently existing Google partners to interface with Indivo X, without re-writing code. That would be pretty cool.

Microsoft may allow FOSS implementations of HealthVault API

Sean Nolan has announced that Microsoft has placed the HealthVault API specification, under the Microsoft Community Promise (CP) at the time of the writing, this page has not been updated to list the HealthVault API, but the text is provided in the specification download. This may allow for FOSS implementations of HealthVault.

The Microsoft CP is not the same as Microsofts Open Specification Promise (OSP). That is problematic because the Open Specification Promise is already doubted by the larger FOSS community,  and the CP seems even more limiting.

Most notably the CP is different from the OS (from the CP FAQ):

The CP requires that implementations conform to all of required parts of the mandatory portions of the specification. Also, in specified cases (such as where the specifications have uses that exceed those needed to achieve the interoperability needs for which the release under the CP is being made), the CP may have special terms concerning what kinds of implementations are covered.


The CP applies only if the implementation conforms fully to required portions of the specification. Partial implementations are not covered.

The CP for Healthvault does have special terms (from the specification download)

Community Promise Restrictions on the Field of Use for the HealthVault Service Specification

HealthVault Service Specification is intended to support personalized healthcare. This technology is designed to be used by individuals to manage their health information, and is not intended to be provider-centric or health enterprise-centric.

That is a problem for projects like Tolven, which is a combined PHR/EHR system. If the PHR component of Tolven were to implement the HealthVault API,would the CP still be ineffect? In Tolvens architecture, the PHR and EHR are based on the same database. While Tolvens PHR is patient centric, the EHR is user centric.

Further it is not defined, as far as I can tell, what a ‘full’ vs. ‘partial’  implementation means. I could create an implementation of the web service calls for HealthVault over a long weekend, by stubbing everything. Now, my system would be a complete implementation from the protocol perspective, every call made to HealthVault would also work on my system, but nothing my system did would have any meaning. It would be much harder to implement things so that they conformed to the specification and actually worked (as opposed to merely appearing to). We might call an implementation that had no stubs and instead contained attempts at real working parts a ‘robust implementation’. But even a robust implementation would have bugs. It would typically work just like HealthVault, but sometimes it would behave differently, mysteriously.

Those ‘differences’ are what programmers like me might call a  ‘bug’. But would an implementation with bugs fall under the Microsoft CP?

More importantly, who decides if an implementation is buggy? HealthVault is still labelled as ‘beta’ and it is entirely possible that if HealthVault and a FOSS implementation of the HealthVault API worked differently, it would be because HealthVault had moved past its own specification.

At this point, I cannot recommend that anyone implement this API. There are too many unanswered questions here.

Having said that, Microsoft is  obviously trying to open up with HealthVault. I hope to convice them that the OSP is a better vehicle, but this is a step in the right direction. So far Google has released no information on the rules for re-implementing the Google Health API. At this point, Microsoft is (surprisingly) more open than Google.