Blog | Friday, February 24, 2017

Will big data replace your doctor?

It's a common complaint among doctors: “Who is making these health care policies? It couldn't be real doctors, could it?” As a doctor “in the trenches” I wonder who these people are who influence our lawmakers. I'm sure they care, but the ideas they come up with are so insane to those of us in the office and in the hospital, we sometimes wonder if they're just messing with us.

For example (indulge me for a minute; I'll get to the point), Blue Cross requires us to enter data about our patients into their system. One question we often have to answer is, “Does the patient have depression or a history of depression?” If I answer, “Yes” but the patient isn't depressed right now the computer will not let me complete the form. Blue Cross's answer to this? “Just say they don't have a history of depression.” (I did not follow this advice, as it seemed dishonest.)

Which brings me to the point. I was reading a piece in the Wall Street Journal the other day, a piece about healthcare, written by “a former hedge-fund manager.” His disjointed piece makes for painful reading, but his central argument seems to be that if we just had access to enough data, we could have “cheaper and higher-quality health care.”

I'm a big fan of science-based medicine, something that requires lots of good data, but the article seems to imply something more emergent, a “SkyNet“ of medicine, if you will. The author, Andy Kessler, says that, “The key is data. With more of it, accuracy gets better over time.” If we could only improve the interoperability of our electronic health records, we could save lives and money. What kind of data is he talking about?

Tons of data come with medical records. Then there are digital scales, Fitbit steps, WellnessFX blood tests, Apple iWatch data and 23andMe genetic test results. Eventually there will be daily commode sensors measuring blood sugar and prostate-specific antigen levels, among other things.

Leaving aside how disturbing this sounds, examples would be laughable if they weren't published in a paper widely read by policy makers.

Imagine all that data being crunched, in real time, by machines looking for patterns—which then put out a simple text message. “Your hemoglobin A1c (HbA1C) has spiked again. I thought we agreed to cut back on the linguine.”

This is not how it works. The practice of medicine relies on data on both large and small scales. On the large scale are things like clinical trials that use sophisticated statistics to help us make rational decisions. On the small scale are data points like our diabetic patients' HbA1C's , a number which estimates how well they are controlling their disease.

Large-scale data, the stuff we get from the medical literature, can tell us, for example, that most diabetics should keep their HbA1C below 7.0 to avoid the complications of their disease. What it cannot tell us is how to get Mrs. Smith to that number, or even if that number is right for her. And the number is only useful if checked at least three months apart. It's not a real-time thing.

But for the sake of argument, let's pretend that Kessler picked a better example. Sticking with diabetes, let's say Mrs. Smith is wearing a continuous glucose monitor, and that those numbers, along with her grocery receipts and her “sitting time” as measured by a fitness wearable, are being sent to SkyNet. And let's say this clever computer sends her a text saying, “Get moving and lay off the linguine.”

To be useful, this model assumes that Mrs. Smith doesn't realize that her diet choices are unwise, or that that if she does realize it, she's too stupid to change. Real people, it turns out, don't work this way. Diabetics work hard to control their disease, and it's not easy for medical, social, economic and dozens of other reasons.

Big data is unlikely to make a “real-time” difference in patients' lives. Where it can help us is in designing rational healthcare policy. But we already have much of the data we need to do that. We know we spend too much and have worse outcomes than other countries. We simply choose not to change. No amount of data will help us make better decisions if we don't really want to change.

Peter A. Lipson, ACP Member, is a practicing internist and teaching physician in Southeast Michigan. After graduating from Rush Medical College in Chicago, he completed his internal medicine residency at Northwestern Memorial Hospital. This post first appeared at his blog at Forbes. His blog, which has been around in various forms since 2007, offers "musings on the intersection of science, medicine, and culture." His writing focuses on the difference between science-based medicine and "everything else," but also speaks to the day-to-day practice of medicine, fatherhood, and whatever else migrates from his head to his keyboard.