Uncertainty is one of the most difficult concepts to master in the art of medicine.
The prevailing paradigm, as portrayed on TV and feared by medical students, is that there must be some question to ask, or some exam finding to auscultate or palpate, or some lab test that could be ordered that will pinpoint the correct diagnosis and guide treatment. This mythos separates modern physicians from oracles reading entrails and shamans practicing magic.
It is a useful paradigm when it works. It fails rather frequently. An example is the teenage girl with chronic functional abdominal pain whose repeated blood tests and imaging have revealed no pathology. The paradigm of finding the root cause, choosing an appropriate treatment to remedy that cause, and then having the child return to school needs to be replaced by an alternative paradigm of managing symptoms, returning to the activities, and coping until the problem remits. In many cases, chronic pain won’t go away until regular activities have been resumed.
There is a third paradigm in the diagnostic regimen. It applies to an ever-increasing group of disorders. These disorders don’t have a pathognomonic finding or a gold standard test. Physicians have only a statistical probability that the patient has or will in the future get the disease. They must employ clinical judgment to weigh the risks and benefits of treatment.
Incomplete Kawasaki disease is one such diagnosis. A 5-year-old girl presents with 6 days of fever, chapped lips during the winter, and a rash. She is fussy. She has mild sterile pyuria, and her C-reactive protein level is elevated. Should she get intravenous immunoglobulin (IVIG)?
My approach is to explain to the parents that it is possible for me to both say, "I don’t think she has Kawasaki’s" and "I think we should treat." I find most parents, given a few extra minutes, grasp the general idea of number needed to treat [NNT] and number needed to harm [NNH], although I don’t use the jargon. This isn’t the first way in which most people approach decision making, but with explanation they can comprehend how, if there is a 10% risk of serious consequence and no perfect test to guide us, it makes sense to treat "just in case."
The concept is easier to understand if it is a test. ED doctors frequently state, "I don’t think it is broken, but we’ll get the x-ray to be sure." Concerns over radiation have reduced the number of head CTs performed for minor head bumps.
Although I find the general public can grasp the basics, they will depend on the physician to provide expertise in balancing the risks and consequences. Bayesian decision making is still an advanced concept for most medical students. When I assess the risks of sequelae from Kawasaki syndrome, I consider the risk that the patient really has Kawasaki times the risk of developing coronary changes times the risk those changes will progress to aneurysms times the risk aneurysms won’t heal spontaneously if ignored times the risk of an event happening because of the aneurysms times the risk that that event will be serious/catastrophic. It is a small number. This is analogous to the Drake equationin the search for extraterrestrial intelligence.
My first career experience with a patient receiving IVIG was a code blue featuring anaphylaxis, syncope, and apnea. My most recent was a patient who had 3 hours of excruciating headache during the infusion. The American Heart Association diagnostic criteria for Kawasakiemphasize high sensitivity, but they don’t adequately describe the NNT nor do they quantify the harms from overtreatment.
There is a bias to treat, even when the risk of adverse effects is greater than the risk of the disease. Factors include fear of malpractice, perceived culpability for errors of omission vs. commission, and economic gain. One of the most common error-producing biases I see in medicine is the response to a referral from an outlying hospital. I’ve worked both places. Tertiary centers find any possible reason to embark on diagnostic tests (particularly lucrative imaging) and treatment in order to justify or rationalize the transfer/admission and to keep the referring doctor happy and looking good in the patient’s/parent’s eyes. In low-risk, high-consequence Bayesian decision making, managing bias may be more important than improving the accuracy of the risk assessment.
Similar diagnostic and treatment dilemmas will occur even more frequently in the near future as genomic screening and exome sequencing become more common. Obstetrics has dealt with this for several years. Mothers are now frequently being confronted with low-specificity (positive predictive value of less than 5%) testing during pregnancy with triple and quad screens. Counseling for BRCA1 testing is the prototype for adult screening. In November 2013, the Food and Drug Administration cracked down on unregulated direct-to-consumer genomic screening. Pediatrics will soon see a large influx of this type of testing in the work-up of failure to thrive and developmental delay. Explaining these scenarios to parents will be a key, acquired professional skill.