Enriching Data Can Impoverish Reality


Medical historian Stanley Reiser wasn’t kidding when he entitled his best-known book, “Medicine and the Reign of Technology.”  To a large degree, technology has taken over medicine.  I am not talking primarily about our increased reliance on such technologies as advanced imaging equipment or assistive procedural devices.  In such cases, the technology remains largely a tool, and the wielder remains basically in charge.  I am talking about a far more pervasive and insidious form of technology whose very name tells a good part of the tale – health information technology.

Many physicians and other health professionals find health information systems clunky, perverse, and intrusive, but their problems go far deeper.  Underneath unwieldiness lies the temptation that we begin relying on such indicators to such an extent that we stop attending to our internal resources.  Consider the case of the patient said in his admission note to be “status post BKA” – below the knee amputation – but who turns out on rounds to have ten toes.  What happened?  DKA – diabetic ketoacidosis was mis-transcribed into the medical record as BKA, and the error simply propagated like a virus.

At stake is what we mean by knowledge.  Is what we know defined by our own experience – what we have seen, heard, felt, and perhaps even intuited in the presence of the patient?  Or do we instead rely on what is represented on a computer screen?  Which is a more likely occasion for us to exclaim, “That can’t be right!” – when what the computer screen indicates does not comport with what we have observed of the patient, or when what we have observed in the presence of the patient does not jibe with what the computer is telling us? 

Consider anesthesiology.  No one would dispute that the ability to monitor inspired oxygen concentration, end-tidal carbon dioxide tension, pulse oximetry, blood pressure, temperature, and electrocardiogram have improved patient management and safety in the operating room.  Yet in our zeal to attend to instruments, we often fail to attend to a still better indicator of each patient’s condition –skin color.  We naturally want to see the instruments, but we must never overlook the importance of adequate exposure and illumination to assess color, and even more importantly, to actually look at the patient.

Increasingly, technology obscures our view of the patient.  Instead of the patient right in front of us, we see the vital sign, or the laboratory value, or the radiologic image, a problem that is as acute in education as patient care.  Instead of encountering the human being who is applying to medical school or residency, how often do we see a standardized test score?  In the name of lowering costs and enhancing objectivity and fairness, we accept a quantitative score that often tells us little more than how well the candidate is likely perform on the next standardized test.

There is so much that such scores omit.  They tell us nothing about dedication, resilience, creativity, innovation, courage, and wisdom.  They do not tell us what kind of human being we are going to be dealing with and what that person is likely to contribute to patients, colleagues, and community.  Even worse, by failing to provide any insight about such traits of character, they implicitly undermine our sense of their importance.  If we devote less attention than we should to something of real importance, or ignore it altogether, we end up with an unbalanced view of what matters most.

Likewise, in patient care we often find ourselves trusting in the numbers above all else, as though that which cannot be counted somehow does not count.  Regardless what the p-value may be, we assume that quantitative indicators are inherently more trustworthy than qualitative ones, even when this is clearly not the case.  Whether a TSH value is 4.0 or 0.4 mIU/L, we still need to encounter patients and assess how they are doing.  Whether a board score is 260 or 210, we still need to talk with students to determine what they are capable of, who they are, and how well they would fit in a program.

When we are not careful, our tools transform us into tools.  Instead of using them, they use us.  In fact, they not only use us, they also transform us.  What is most worth attending to?  A standardized test score or how candidates handle failure, bring out the best in others, and innovate?  The question is not simply what we rely on most in practice but what we look to in theory.  Too often, such quantitative scoring systems have merely fostered an arms race or bidding war, driving up scores but doing little to foster meaningful enhancements regarding who candidates are and what they are capable of.

When applied to machine learning and artificial intelligence, the risks of a data-driven model of medicine are magnified.  Looking at the complex reality of a patient strictly through the lens of data reveals some aspects but omits far more.  We know, in a way that artificial intelligence cannot, that medical records are like photographs – they may reveal a great deal about what we include in the frame, but even more lies beyond its margins.  What we don’t see is often as important as what we do – and without stepping back from time to time to take in the larger context, we often don’t really see at all.

Our very language has changed, and with it our thoughts, feelings, and aspirations.  Our zest to count too often blinds us to what really counts most.  As Abraham Maslow suggested 50 years ago, when our only tool is a hammer, we begin seeing everything as nails.  When our tool is a health information system with an insatiable appetite for data, we produce more and more data.  But what if our world contains not only nails but screws and bolts and a host of other fittings that a hammer would only destroy?  And what if healthcare, however data rich, contains a reality far richer than mere data can ever reveal?

Richard Gunderman is chancellor’s professor at Indiana University School of Medicine.

Leave a Reply