I love a good mystery. And I’ve noticed one post-mortem analysis after another lately exploring the recently announced impending demise of Google Health, a service aimed at helping people manage their own health information. Why would as powerful an entity as Google abandon so major a project in such a seemingly promising field?
Google’s own explanation, in a blog post, is that the service just “didn’t catch on” as hoped. That rings true to me: Here I am, immersed in health care news, and I’d heard little about Google Health and nothing enticing enough to make me want to sign up. But the answers go deeper.
In MIT’s Technology Review, David Talbot recently wrote of Google Health that its passing reflects a broken medical system: “Experts say its untimely death is, in many ways, an extension of U.S. health-care providers’ failure to share data across institutions, or make it easy for patients to obtain it.” (Tech Review also has an interesting new article out here about a Cambridge company’s health-tracking platform for employers.)
I just spoke with Dr. Kenneth Mandl, an associate professor at Harvard Medical School, physician and researcher in the informatics program at Children’s Hospital Boston, and co-founder of the open-source project Indivo, the first electronic personal health record. Indivo, he says, began in 1998, inspired the Google Health model in 2006 and is still in wide use, including for patients at Children’s.
The concept behind personal health records is that each of us should have access to our own medical information, and the ability to share it where needed—with our doctors, our family, and with computer apps that can help care for us. Here is his detective work, lightly distilled:
Q: So, Ken, whodunit?
Basically, it was a combination of a couple of factors—the stagnant data flow in the health system and some failures in Google’s strategic execution.
The data about our health histories, medications and treatments, usually captured by our physicians on paper or, increasingly, in electronic health records, usually stay put—in physicians’ offices or hospitals–even though federal law entitles patients to copies of their electronic health records.
And Google was more dependent on real ‘data liquidity’ than I think they fully realized. Job number one for them, with their enormous resources, should have been to try to ensure that there were generalizable, standardized ways to move data around in the health system. They needed to dig in and do serious engineering with clinical systems. But they didn’t do that. They were hoping to sit on a high perch and to provide storage and applications for managing that data. But the data were not there.
So there’s real work to be done. Google was a little ahead of its time in a sense, but really, they should have spent more energy and dollars helping to generate data flows through policy reform and technological innovation.
Also, Google Health fell short as an “apps” platform. Contrast it with the iPhone platform — Google never really captured a serious community of third-party developers who could add value to data by generating apps that could run against their programming interface.
Google viewed itself as the primary innovator; if instead Google had distributed that innovation to third party developers, as Steve Jobs did, Google Health might have attracted a more excited community of developers and users. In a related research project I co-direct, www.smartplatforms.org, we are developing the necessary components to fully enable an “app store” for health.
Q: And as Google says, it just didn’t catch on? I certainly noticed lack of demand in myself.
Yes. One other reason for that was that they didn’t do very well in establishing a trust model. What would Google do with your health data once they had it? And do I trust a company to manage my health that does not even have a customer service number? Is search and advertising the revenue model I am looking for in my health records company?
But beyond these issues, which Google probably could have fixed if they wanted to, we must ask, ‘What does this say about the use of personal health records in general?’
Patient-centered health information technology will undoubtedly be a pillar in our health system. But we need to recognize that the user experience and the availability of data have to be the primary goals. The user experience has to be elegant and has to accommodate most levels of literacy and technological expertise. We are learning from Facebook and online health-oriented social networks that a social component to the experience may be a key ingredient.
In the Google Health environment, even the early adopters, even the technophiles who wanted to make this work, couldn’t really get it populated with data, and once they did there wasn’t much they could do with the data. So it’s about working both ends of the system. From the bottom, there need to be standardized, easy ways for getting data in. On the top there should be apps creating a seamless experience for the user. Once the patient has given permission, she should expend very little or no effort to have an app running that is improving her care by accessing and processing data in the personal health record from her doctor, from the hospital where she was just seen, or from the lab where she just got results.
To improve health care and reduce costs, we need intelligence, and for intelligence, we need data. Patients are going to need to play a big role in this. We need a channel to capture the patient voice in our health information systems. Only the patient knows how she is responding to her treatments, whether she is taking her medications, which drugs work better for her and which drugs make her feel better.
Q: In a utopian future, a decade or so from now, what would this look like?
Say I have diabetes. My doctor’s care plan is delivered through an app that helps me manage it. The system supports communication between me and all the members of my care team. My glucometer automatically feeds my personal health record data that is used by algorithms and apps making sure that my sugar control is on the money. My insulin pump is also sending data into the record. Every time I fill a prescription at a pharmacy, the system knows I’ve filled that prescription and have taken that med. The insulin bottle I’m using and the insulin pump also signal back that medication has been used. Now the patient, the computer, and the doctor know what is going on. Therapies can be adjusted. Lessons can be learned from the collective experience across millions of patients.
The subjective symptoms of my disease and the side effects of my medication are being surveyed actively for me. I can report these, and computers process that information between the time I report it and my doctor gets it. I’m sharing my experience with other patients like me online — through Facebook, or perhaps Google+ ! I’m learning from my peers and from information aggregated across my peers’ reports.
‘Patients should individually and collectively demand copies of their health data after encounters — copies that are in computable electronic format.’
The system unburdens doctors from having to acquire and curate that information. They get intelligently processed summaries and action items helping them to manage a panel of thousands of patients. As a result, I’ve got better adherence, better sugar control, and better outcomes. With my permission, the health system is tracking outcomes for me and across the whole population to improve and personalize therapies; I am therefore on the regimen — derived from the real-world experience of thousands or millions of patients — that works best for me.
Just think about IBM’s Watson and the leaps made in artificial intelligence in the last ten years. By ten years from now we’ll have figured out where it’s safe to take humans out of the loop, and be providing very low-cost management solutions for some portions of health care that requires a lot of effort now by humans.
Q: So how about right now? Is there any sort of minimum electronic health record that it’s reasonable to demand of our doctors?
Carey, actually, everything I just described is technologically possible now and much of it exists in testbeds. But it will take years to execute on the whole vision and see it widely adopted.
So what should we do now? With new strong federal incentives for physicians to use electronic health records, patients should individually and collectively demand copies of their health data after encounters — copies that are in computable electronic format. Not just pdfs, not just CDs, but the kind of technologies that we’re used to when we use Facebook and the iPhone.
There is absolutely no reason why the now 600 vendors of electronic medical records cannot begin to export standardized information. And my bias is that a very good place to begin would be to get the medication information correct.
It takes time for everyone to agree on standards for storing and sharing every possible kind of health information. In the meantime, let’s at least get medication information correct. It may surprise patients to learn that clinicians do not actually know with any certainty what meds they are taking. High quality medication data could flow to personal health records immediately from doctors’ offices, from pharmacies, and from patients, and start to form the basis for the first set of really useful apps for patients and doctors.
Readers, reactions? What kind of electronic interaction do you want with the health care system?