Anybody who followed this blog closely should not be surprised by my skepticism about Personal Health Records (PHR). A patient-managed record is a feel-good idea that unfortunately lacks the economic model to support itself.

Data quality is one of the factors underappreciated by people who expect that technology will be taken care by someone else. We expect the cars, planes and sewers to just work, why not medical records? Who cares about technical details when we feel entitled to having the data there when we need it?

ePatient Dave's experience blasted this nice illusion into pieces

So what happened and why this is so interesting? This all started about a month ago when Dave deBronkart (ePatient Dave), a well know patient blogger and technology enthusiast, decided to bite the bullet and get his medical records into Google Health. His records happened to be at Beth Israel Deaconess Medical Center (BIDMC), considered to be on the forefront of PHR adoption. A medical center, whose CIO, John Halamka is known as a national figure. If anything, you could expect BIDMC to be THE place where everything is figured out.

Yet, once Dave imported the records he found their accuracy (or lack thereof) nothing less than appalling. His detailed report (Imagine someone had been managing your data, and then you looked) as of today gathered 115 impassioned comments, most of them expressing shock, outrage and disbelief. Read his post and comments if you want details, but the basic problem was that his records were created based on the billing codes instead of clinical classifications. This resulted in widely exaggerated diagnoses, several bogus alerts and of course lack of actually useful information and data.

The only surprise to me is that people are surprised

These medical records have not been populated for the patient. Anything that gets written into the record goes there for a reason. Getting clinicians paid is one of them. Helping them reduce legal liability is another one. Minimizing their time spent on activities that do not result is getting paid (thus draining institutional resources) is right there. There is no incentive in the system to produce an accurate clinical record that could reduce reimbursement, increase demands from the patient and perhaps even help provide documentation for a lawsuit.

Of course, nobody would speak up against patient empowerment, but absent systemic incentives to provide accurate records, how can you expect them to happen. BIDMC has a huge incentive to become a national showcase for PHR implementation and generate positive PR, but this might not be a strong enough economic incentive to revamp record keeping. Even then, they would have to overcome the motivations of clinicians that can be hardly expected to generate extended documentation that does not get them paid. Instead they can and will focus on adjusting the documentation to maximize reimbursement.

The idea of PHR lacks economic alignment of incentives

Every now and then we hear comparisons between healthcare and banking. If banks could automate download of financial statements, the thinking goes, why not healthcare institutions. The trouble with the analogy is there is no separation between "financial" and "clinical" views of the data in banking. There is no incentive to manipulate it to maximize reimbursement. There is not much complexity in data and ways tangle it up. If data model is simple automation is a cinch.

Medical data structures are anything but simple

Many ways to describe and code the same condition and procedure. Reliance of human judgment for making these decisions. Financial and professional interests, riding on such determinations - including reimbursement and liability. The knack of this data to have limited shelf life when it is useful. Lack of incentives for accurate record-keeping and in turn lack of confidence in the data by its users - both clinicians and patients. You could have too little data which you might consider not useful enough. Or too much which you might not trust. Or you might trust it but regret it later. Is there anyone who has both means and motivations to work to keep PHRs accurate? If even BIDMC and Google are producing such dismal results after years of trying, my answer would be: HIGHLY UNLIKELY.

PHR industry is built on the premise of accurate and complete patient records. I do not see this happening beyond most basic applications.