In this video blog, Adam Lipworth, MD, discusses highlights from his lecture, Reducing Diagnostic Error in Dermatology: A Bayesian Approach, on July 27, 2024, during the inaugural Elevate-Derm Summer Alliance Conference at the Westin Copley Place in Boston, Massachusetts.

In this video

Dr. Lipworth is quick to reassure clinicians that his talk "is not about the math." Bayes' theorem is a formula, but he argues you never need to run the equation to use it at the bedside. The core idea: to judge whether a patient has a given disease, you take everything you know coming into the encounter and add each new piece of independent data in a weighted fashion — weighted by how reliable it is. He frames every step as a test. The history is a test, the physical exam is a test, and lab work is a test, each moving you from a pre-test probability to an updated post-test probability. As he puts it, "there's no such thing as 0%" and "no such thing as 100%" — probabilities simply shift as new independent data arrives.

On bias, Dr. Lipworth calls out the danger of confirmation bias and "double counting your history" — expecting cellulitis, then seeing cellulitis partly because you expected it. Dermatology's advantage, he notes, is the morphologic exam: "we don't need an x-ray to look inside somebody," so clinicians should keep the exam independent of the incoming history, and admit and describe a lesion when they don't know what it is rather than force it to fit. He and Richardson highlight collaboration and Grand Rounds as ways to gather independent assessments — a "wisdom of the crowds" that, when each observer isn't influenced by the others, lands closer to the right diagnosis.

  • You don't need to do the math to use Bayes' theorem — the practical skill is updating a diagnosis's probability as each new piece of independent data comes in.
  • Treat every step as a test: history, physical exam, and labs each move you from a pre-test to a post-test probability, weighted by the reliability of the data.
  • Keep the morphologic exam independent of the history to avoid confirmation bias and "double counting your history."
  • When you don't recognize a lesion, admit it and describe it — don't let it be biased by what you think it should be.
  • Inheriting a colleague's patient carries anchoring bias, but a fresh set of eyes can also "reset the clock"; independent assessments and collaboration yield better care.