When the doctor is biased

recent commentary in the New York Times eloquently addressed the phenomenon of psychiatric diagnoses biasing clinicians and resulting in poor care. One concern that arose in the comments was that this might increase the likelihood that patients would withhold crucial diagnostic and therapeutic information, to their detriment. 

This issue of hiding information to prevent bias is very real. I cared for an attorney who developed a bothersome movement disorder that stumped both me and the neurologist who saw him.  After 8 months and much testing with no answer, I told him we were stumped, and mentioned in passing the fact that the neurologist and I had both seen this particular problem only in patients who had received significant doses of some older antipsychotic medications. At that point he said: "Can I tell you something and have you NOT put it in my chart?"  He had been diagnosed with schizophrenia during college, had had a bad decade, in and out of treatment, and had ultimately learned to cope with (recognize and ignore) his ongoing hallucinations without medications. (He pointed out that: "We all have thoughts we know we have to ignore, right?"). He was quite legitimately concerned that his history of mental illness and his ongoing active hallucinations could impact his health care, his employment, and license to practice law. 

Hiding information is dangerous, but cognitive errors are numerous and inescapable. Inmedicine they are also potentially hazardous. Kahneman (Thinking Fast and Slow, Judgement Under Uncertainty), Ariely (Predictably Irrational), Heffernan (Willful Blindness), Piatelli-Palmarini (Inevitable Illusions), Ubel (Critical Decisions), Brafman (Sway), Chabris (The Invisible Gorilla), Burton (On Being Certain), Frankfurter (Bullshit),  Groopman (How Doctors Think, Your Medical Mind), Woloshin (Know Your Chances) and countless others have described this in both the professional and lay literatures.

If information is simultaneously necessary and hazardous, what should we do? Some of the above authors talk specifically about this:  Ubel, Kahneman, Ariely, Groopman.

Recognize it. Perhaps giving it a catchy name would make it easier to remember and talk about. 

Study it. There is lots of work on this in the behavioral literature, but it would be useful to study and publish specific information in the medical literature. The study showing frighteningly different rates of diagnosis of pulmonary embolus on pulmonary arteriograms depending on whether the accompanying history is high-risk or low-risk comes to mind.

Teach it. To clinicians AND to patients. When I precept family practice residents in the outpatient world, I harp on a series of questions that one has to ask:

  • How sure am I (of the diagnosis, of the treatment)?
  • What data have I discarded or discounted to be this sure? (Not have I discarded info. We always discount or discard information. Always.)
  • How will know if I am right?
  • What else could it be (could I do)? (Plans B-Z)
  • Do I need to actively consider other choices now, or can I wait?
  • If I wait, what am I waiting for?
  • What do I look for to alert me that I am wrong? (Not the same question as how will I know I am right? The framing is important.)
  • Have I included the patient (who is both the canary AND the mine) in this discussion?
  • Have I documented this in the note in a way that a colleague or the patient can use the note if I am not available? 

Talk about it. With and among patients and with clinicians. 

Lets change focus a bit and get more specific with a few thought experiments, which I think will illustrate that not all syncopal episodes warrant the same response. How intensely should one pursue the above questions in each of the following circumstances? 

  • A 19 year old competitive collegiate distance runner on no medications who has two syncopal episodes in the several hours after an unusually hard work-out during a heat wave, seen in the office 5 days later, feeling fine with normal vital signs, and has a copy of his ED evaluation with normal lab and EKG other than mild dehydration, labeled as vasovagal syncope secondary to dehydration. What if he had a brief seizure with the second episode? What if the seizure lasted 90 seconds and he was incontinent? What if his hobby is scuba diving?
  • A 43 year old sedentary female diabetic on oral agents for her sugar, on two medications for her blood pressure (136/74 today), who had two syncopal episodes after participating in a charity walk to raise money on a brutally hot day. She has a copy of an ED evaluation with normal lab and EKG other than mild dehydration, labeled as vasovagal syncope secondary to dehydration. What if her BP today is 102/52 compared to baseline of 155/70? What if her baseline EKG shows left bundle branch block? What if she had atypical chest pain and a negative cardiac stress test two months ago? Two years ago? What if her husband is an attorney? What if her daughter is an interventional cardiologist. (OMG!)
  • A 65 year old male smoker with a fib who had two syncopal episodes in association with a GI illness with vomiting and diarrhea and poor oral intake, on 9 medications including three for blood pressure, with a history of falls from orthostatic drops in blood pressure, who often forgets to take his coumadin. He was seen in the ED and told he had syncope because he was dehydrated. What if he has a right facial droop? What if he lives alone? 

My point is that context is extremely important to making a correct diagnosis. We haveknown this for a long time


I submit that the questions are ALWAYS appropriate, but that the effort invested should not always be the same. When I was a medical student, there was a sign in the ED doctor's lounge that said: Treat 'em and street 'em. The explicit goal was to get sick patients on the floor and everyone else out the door as quickly as possible. I don't know how much the ED culture has changed, but I can say with certitude that the pressure to maximize wRVUs makes this a serious problem in the PCP office. The overwhelming majority of 25 year old non-smoking healthy patients who present with two days of sore throat, swollen glands and temperature of 101 will have a self-limiting viral etiology, some will have strep, and a smaller number will have mono, which will not be diagnosable yet. The too typical quick and dangerous approach is to ask as few questions as possible (or perhaps have someone else ask them) to avoid having to deal with the answers, do the strep test, treat the positives  and tell the rest to go home and call if they don't get better. This works most of the time. It is consistent with evidence based medical guidelines. It is legally safe as long as one documents the instructions to call if not better. But it is not necessarily optimal care, as gonoccoccal pharyngitis, leukemia, diphtheria, Kawasaki's disease, acute HIV, peritonsillar abscess, Chlamydia, Mycoplasma, Lemierre's, Stevens Johnson syndrome, epiglottitis and agranulocytosis are in the differential. How many of these should be tested for, mentioned, or discussed? And I'm only talking about a sore throat, not chest pain or trouble breathing or blood in the stools or weight loss.

Both the clinician and the patient need to understand that every diagnosis is provisional and every treatment is experimental and that both should always be considered part of a PDSA cycle.


Links to more on this topic::