February 08, 2009
Medical Records Substitute For Randomized Trials?

Obviously analysis of medical records for millions of people can not substitute for randomized trials for new treatments that are not yet in use. However, some UPenn researchers claim they have demonstrated that at least in some cases different treatments can be compared by examining computerized medical records.

PHILADELPHIA – For years controversy has surrounded whether electronic medical records (EMR) would lead to increased patient safety, cut medical errors, and reduce healthcare costs. Now, researchers at the University of Pennsylvania School of Medicine have discovered a way to get another bonus from the implementation of electronic medical records: testing the efficacy of treatments for disease. 

Often illnesses have multiple potential treatments. In many cases the best treatment is not known. But randomized medical trials are expensive and pose ethical problems as well. If we only had more electronic data about courses of treatment and outcomes in theory it might be possible to tease out of the data which sorts of treatment work best for different categories of patients.

An obvious bias with medical records is that doctors likely used characteristics of patients in determining which treatment to give them. But as the sum total of all research-accessible medical records covers an increasing length of time even that problem is surmountable by comparing records from time periods before and after treatments came into widespread use. Also, countries, regions, and individual doctors differ in terms of preferred treatments. These preferences on the part of individual doctors serve to partially randomize patient treatment choices.

What is surprising to me is that this press release claims that their use of electronic medical record (EMR) databases in this manner is a first of its kind. Is that true?

In the first study of its kind, Richard Tannen, M.D., Professor of Medicine at the University of Pennsylvania School of Medicine, led a team of researchers to find out if patient data, as captured by EMR databases, could be used to obtain vital information as effectively as randomized clinical trials, when evaluating drug therapies. The study appeared online last week in the British Medical Journal.

“Our findings show that if you do studies using EMR databases and you conduct analyses using new biostatistical methods we developed, we get results that are valid,” Tannen says. “That’s the real message of our paper — this can work.”

I expect medical records to become much more useful due to several factors: 1) longer numbers of years tracked for each person; 2) More medical tests per person; 3) eventually detailed DNA sequencing per person; and 4) eventually home sensors will continually collect information about each person from sinks, toilets, cameras, and even sensors on bedstands that'll monitor breathing and other biosigns. Adverse drug reactions will be compared against DNA sequencing results. Bedstand monitors will catch early signs of sleep apnea and insomnia.

While I view most fiscal stimulus program elements as wastes the plan to boost electronic medical records collection will pay back with higher quality care and an acceleration of the rate of advance of biomedical research.

In January 2009, President Barack Obama unveiled plans to implement electronic medical records nationwide within five years, arguing that such a plan was crucial in the fight against rising health care costs. Of the nearly $900 billion in Obama’s planned stimulus package currently before the United States Senate, $20 billion is proposed for electronic health records.

Once we get home medical sensors and medical sensors embedded in our bodies the business model for medical care will change drastically. Rather than going into a doctor to describe your symptoms so that the doc can order tests the continual streams of sensor readings will flow into expert systems running on web servers. If you start feeling ill you'll be able to call up a web page where you can describe your symptoms. The system might recommend a sensor pill to swallow or that you spit into a home microfluidic sensor device to collect more information. Then you'll get referred by the expert system software to a doctor for treatment.

Share |      Randall Parker, 2009 February 08 03:44 PM  Policy Medical

Tj Green said at February 9, 2009 3:41 PM:

Cheap accurate DNA sequencing, microfluidic sensors, medical databases. When this much information comes together something amazing happens.

Xenophon Hendrix said at February 9, 2009 6:13 PM:

I want Larry Niven's autodoc.

Quite a while ago--perhaps a couple decades--I read that expert systems then were either as good as or nearly as good as doctors at diagnosis. (I can't remember which.) I suspect, but don't assert, that they are probably better, nowadays.

Getting error-prone humans as much out of the loop as possible will probably save lives. So will keeping people away from disease-filled hospitals as much as we can.

Nick G said at February 10, 2009 9:41 AM:

"Quite a while ago--perhaps a couple decades--I read that expert systems then were either as good as or nearly as good as doctors at diagnosis. "

They were better than the average doctor, not as good as the best.

Half of all doctors are below average...

Post a comment
Name (not anon or anonymous):
Email Address:
Remember info?

Go Read More Posts On FuturePundit
Site Traffic Info
The contents of this site are copyright ©