The Daily Dose • Tuesday, May 1

Biases in Education and Research and Their Impact in Medicine

By Dr. Michael D. McDonald, from the IARS 2018 Annual Meeting*

Dr. Edward Nemergut used this review session to serve as a brief overview on a few of the many biases that influence the care we provide, education we provide, and how we analyze the literature. Biases he specifically touched upon included the Dunning-Kruger effect, publication bias, survivorship bias, fundamental attribution error, hindsight bias, and intentional bias.

Dr. Nemergut began his discussion with the July effect, a validated effect which demonstrates increased mortality in patients treated in July. There have been a few hypotheses generated to explain this effect, however by far the most common is the new resident hypothesis, which theorizes that there is a decrease in quality of care as green providers are given increased clinical responsibilities. Dr. Nemergut used this example to describe the Dunning-Kruger effect, which is a cognitive bias where people of low ability suffer from illusory superiority where they mistakenly assess their ability as greater than it is. This is not to describe arrogance, but instead ignorance where young providers fail to recognize lack of skill and their extent of inadequacy. Validation studies of this effect also have demonstrated that those with greatest ability tend to underestimate their score.

Next, Dr. Nemergut shifted gears to publication bias, which occurs when the outcome of an experiment or research study strongly influences the decision whether or not it is published. This is more likely to occur when the results of a study are positive and in journals with high impact factors. The typical story for publication bias is when a large randomized study, often funded by industry, demonstrates efficacy and smaller future studies that fail to demonstrate efficacy remain unpublished because they are thought to be underpowered or poorly designed. An analysis of studies investigating the efficacy of Prozac demonstrated that 94% of published trials demonstrated efficacy while only 51% of unpublished trials demonstrated efficacy. In reviewing meta-analyses and review papers, look for analysis for publication bias through funnel plots. But take note, nearly half of these papers fail to investigate the possibility of publication bias and thus may be misleading.

Dr. Nemergut then moved to discuss survivorship bias, which is the error of concentrating on the people or things, which have made it past some selection process and overlooking those that did not, typically because of their lack of visibility. This is most commonly seen with retrospective investigations in medicine.

Dr. Nemergut also reviewed Fundamental attribution error. Attribution theory dictates that we all attempt to explain behavior in others by attributing a cause to the behavior, either through internal or external factors. Fundamental attribution error is our tendency to explain someone’s behavior as the result of internal factors, coupled with our tendency to explain our own behavior as a result of external factors. For instance, attributing one’s own high exam score to hard work (intrinsic) and someone else’s low score to laziness. This is common in medicine as we all have a tendency to attribute errors and unanticipated bad outcomes to shortcomings (internal) of other providers.

Dr. Nemergut then briefly introduced hindsight bias, or the tendency to see a prior event as having been predictable, despite there having been little or no objective basis for predicting it. Hindsight bias is often implicated in medical malpractice and much work has been done toward reducing this bias in expert witnesses.

Dr. Nemergut finished by introducing intentional biases, or opinions held despite data indicating otherwise. Examples include strategic ignorance, reality denial and self-signaling.

Other biases Dr. Nemergut did not discuss but pointed out, include commercial support, anchoring bias, gambler’s fallacy, and posterior probability error. There are many biases that influence the care we provide, education we provide, and how we analyze the literature. However, by being aware of these biases we can work on minimizing their effect of negatively impacting our practice.

*Coverage from the Review Course Lecture, Biases in Education and Research and Their Impact on Patient Safety