Tuesday, January 17, 2017

A Little Knowledge Could Be a Dangerous Thing

One day soon, we'll have real-time or near real-time information about our health.  Not just how we are at the moment but also whether and for what we are at-risk.  I'm fairly certain about this.

I'm less certain that this will be necessarily a good thing.

Michael Snyder and wearables (Steve Fisch, Stanford.edu)
The topic has received a lot of press recently due to a study by Stanford researchers (Synder, et. al) in PLOS Biology, which concluded that:
these results indicate that the information provided by wearable sensors is physiologically meaningful and actionable. Wearable sensors are likely to play an important role in managing health.

The study collected nearly 2 billion measurements (!) on 60 participants, who wore up to 7 tracking devices.  It focused especially on identifying early signs of Lyme disease and inflammation, and risk for Type 2 diabetes, but the authors expect that its implications will go much further.

Dr. Snyder told Scientific American: "Too much of the time we spend time measuring people when they’re sick.  What we really want to understand is what does it mean to define a healthy state, then quickly identify deviations from that state."  The trouble will be that we won't always know what those deviations mean to our health.  It will take a long time to figure out what our baseline is, and when which deviation mean what.

Dr. Snyder further noted, "We have more sensors on our cars than we have on human beings," a situation he believes soon will change, as the current wave of mainly activity trackers evolve to more directly track health measures.

The reference to sensors in cars is valid.  A Reuters article profiled how insurers are betting big on sensors.  Thirty percent of North American auto insurers are using sensors in cars to track driving behavior of their insureds in order to more accurately price their policies, and this is expected to grow to 70% by 2020.  Health care and health insurance will quickly follow.
Tracking isn't just limited to wearables.  A new urine test can determine within five minutes how healthy your diet is, and an MIT-backed start-up is developing a "smart toilet device" that can measure, in its first iteration, glucose and hydration levels.

A new breathalyzer claims to be able to diagnosis 17 diseases with one breath, including several types of cancer and kidney disease.  The researchers hope to incorporate the technology into smartphones.

University of Cambridge
That would just add to the ever-growing capabilities of smartphones.  Just within the past few weeks there have been announcements about them tracking heartbeats, diagnosing malaria, diagnosing and managing respiratory diseases, identifying genetic conditions, even sequencing DNA.  

There seem to be no foreseeable limit on what we will be able to track and even diagnose with the various ubiquitous technologies that are being developed.  We'll need AI to sift through all the data that will be generated about us, and to synthesize it into actionable information.  If we think EHR alert fatigue is an issue now, just imagine what it will be like when we have billions or trillions more data on our health, and more of that information is directed towards us, not just to our physicians.

We just may not want to believe everything they tell us.

For example, a new study found that a third of patients who had been diagnosed by a physician as having asthma did not, in fact, have it.  The lead author diplomatically cautioned that: "It's impossible to say how many of these patients were originally misdiagnosed with asthma, and how many have asthma that is no longer active," but it is sobering that the "gold standard" of a physician diagnosis can be that fallible.  Why then should we believe a wearable or smartphone?

Of course, one could argue that wearables or other sensors would have picked up the diagnosis sooner and/or more definitively, and could better have determined when it was no longer active.

The problem is that a lot of what is considered the state-of-the-art in medical beliefs is subject to change.  Aaron Carroll recently urged that we view such beliefs with a "healthy skepticism."  He detailed several examples of where what we "knew" to be true turned out to be, well, not so much.

As he pointed out, "Sometimes it’s hard to separate what’s truly a medical certainty from what is merely solid scientific conjecture."  Sometimes even those certainties turn out to be not quite so certain, and sometimes "solid scientific conjectures" prove neither solid nor scientific (e.g., the appendix is important after all).

All this is going to make it hard for us to turn all that data we're going to be collecting into meaningful advice.

The Mayo Clinic recently published The Promise and Perils of Precision Medicine. warning that so-called precision medicine, based on genetic testing, may not be all that precise.  It can lead to misdiagnosis, as well as unnecessary or even harmful treatment.  As the article concluded: "Although the technological advances in genetic sequencing have been exponential, our ability to interpret the results has not kept pace."

For "genetic sequencing," we could equally substitute a host of other new types of data.  For example, full body imaging was supposed to provide peace of mind, catching cancer and other issues sooner, but is more likely to result in unnecessary tests and procedures than in helping.

Our ability to be more precise does not mean we'll always be more accurate.

We will track more about our health.  The data will eventually tell us more about our health than we know now.  In the meantime, though, we're going to have to take what it says with a rather large grain of salt, rather than always rushing into action.  That will not be easy.

If tracking can help teach us to listen better to our body and to take appropriate action only when necessary, that's great.  If we end up relying on it to manage our health, though, then we've taken one more step away from our health, and from ourselves.

What I hope most is that all that data are training wheels rather than crutches.

No comments:

Post a Comment