You Are Not an Algorithm: Why Online Self-Diagnoses Fail

Share if you find this post helpful:

Annoying self-diagnosis cartoon

How many people search for health information online?  According to Google, the world’s two billion internet users spend 21% of their online time performing searches, and the single most popular search is for health information.  In fact, eight in ten internet users look online for health information, making it the third most popular online activity just behind email and general search.  That’s 1.6 BILLION people trying to find information or a solution for their health problem online.

This has become such a pervasive and compelling need that Google recently announced a major collaboration with the Mayo Clinic to improve the value and usefulness of online health searches.  Google product manager Prem Ramaswami said that a newly updated ‘Knowledge Graph’ will reconfigure Google’s search algorithm with internal and publicly available databases to display relevant information at the top of each user’s search results page.

The display will include information about how prevalent the condition is, typical symptoms with medical illustrations, and most effective treatments.  A rather amusing example was provided for the search term ‘rabies’, which will display a picture of a raccoon next to an arm with a bite, and comments such as “very rare” and “medically treatable by a doctor or professional.”

Let’s hope this example is a just a starting point.  Online health consumers are increasingly frustrated and disappointed by their search experience and results, with many complaining of overly inflated and inaccurate diagnoses—often wrongly indicating a fatal or life-threatening condition.  Online health boards and social media sites are rife with stories about people searching simple symptoms such as ‘headache’, only to return as many results for brain tumors as for benign causes such as caffeine withdrawal.

Doctors are deeply skeptical and irritated by patients who search online for a self-diagnosis, and have even coined a term for such people, calling them “cyberchondriacs.”  This practice has led newly informed and emboldened patients to argue with their doctor about the proper diagnosis and treatment when they go into the physician’s office for an examination.  Equally worrying are those patients who bypass the doctor’s office entirely and use their online search results as the definitive authority for diagnosing and treating their condition.

A whole online cottage industry has arisen in recent years to satisfy this insatiable demand for self-help health diagnoses.  WebMd.com, diagnose-me.com, medhelp.com, zipnosis.com, etc.—there is no shortage of sites seeking to help individuals find a quicker diagnosis and solution for their health problem.  Even many bricks-and-mortar hospitals, most notably the Mayo Clinic, have reconfigured their websites in an attempt to help health consumers filter and prioritize their conditions.

The problem with all of these sites—as with many doctors—is that their diagnosis algorithm is based on quickly matching typical disease prototypes and symptoms.  Online sites use this method because of the limitations of the data provided; doctors employ this method because of the time limitations associated with most office examinations.  But there are two fundamental flaws with this logic. First, many presenting symptoms are associated with equally many potential diseases, and therefore it is difficult using this diagnosis approach to narrow down and confirm the true underlying cause.  ‘Abdominal pain’ is just as likely to return a diagnosis of indigestion as appendicitis, not to mention a raft of other possibilities ranging from irritable bowel syndrome to diverticulosis.  The other inherent flaw in this system is that symptoms are grouped into common disease prototype ‘buckets’.   Heart attacks, for instance, often initially feel like indigestion and therefore ‘abdominal discomfort’ can easily be associated with a common—and incorrect—serious condition.

For these reasons, an accurate diagnosis (and therefore proper treatment) is hard to come by from an online search or ‘expert system’.  In many if not most cases, each set of unique symptoms has its own unique triggering cause.  In other words, not every presenting symptom is associated with a typical disease prototype.  To truly get to the bottom of a new health condition requires a thorough and careful description of the symptoms—what type, where on the body they are presenting, when expressed—and an equally thorough and rigorous exploration and analysis of the possible causes.

Complicating the matter further, research has shown that 75-80% of all illnesses are caused by external or environmental issues, as opposed to known medical ‘disease’ causes.   No online symptom checker or medical diagnosis tool has the capability to identify and filter what you’ve been doing differently, the external contaminants you may have been exposed to, or the places you’ve visited recently that may have contributed to your illness.  Only good, thorough detective work can figure that out, and only with a full and complete picture of your illness history and the associated external events that contributed to your illness.

I’ll be very surprised if the day ever comes when computers can replace a thorough and proper custom investigation of someone’s illness to deliver a truly reliable and accurate diagnosis.

By Reid Jenner

Comments

  1. Madison

    I self diagnosed myself with Ehlers-Danlos Syndrome. The doctor did not diagnose because I asked him about it though… I just thought I had it and he came up with EDS before I could ask.

    1. Madison, thanks for sharing your experience. I find it amazing how easily doctors can miss sometimes the obvious diagnosis, but when we remind them of a possible cause how quick they are to confirm it. What is the doctor really adding to this equation?

Leave a Reply

Your email address will not be published.