Framing Bias – It’s All in How You Present It
September 26, 2019
Written by Nicole McCoin
There were significant differences in the differential diagnoses generated depending on the framing of a presentation.
Why does this matter?
Framing, in this sense, referred to how the information in a case was presented or emphasized, but the actual data presented was the same. Changing the framing between two different presentations did not include adding information in one scenario and omitting it in another. The data stayed the same. This study also did not include variable descriptors such as “intoxicated”, “disheveled”, or “anxious,” for example, in one presentation versus another, which may also have introduced bias. Framing bias affects patient care every day, and our awareness of it is important. For example, the way a resident presents a case often determines what differential is generated and what tests are ordered. Unless the attending does a completely fresh history and exam with an open mind, they may pursue the differential generated from the resident’s presentation. This could also be seen in triage notes or during sign-out. Framing bias is everywhere, which is scary.
We’ve been framed – an everyday occurrence….
The authors of this study used two different presentations to study this phenomenon – one framed for a pulmonary embolism (PE) case and the other framed for an interstitial lung disease (ILD) case. The cases framed one way or the other were randomly presented to 69 physicians, 23 registrars and 46 consultants. Focusing on the PE case, 100% of participants who read the presentation framed toward PE listed PE in their differential, with 95% stating they would order a CTPA or D-dimer. However, only 50% of participants in the group who read the presentation framed away from PE had PE in their differential diagnosis, and just 30% stated they would order CTPA or D-dimer. Clinicians had 79 times greater odds of having PE in the differential when reading the presentation framed towards PE compared with framing away from PE (OR 79.0; 95%CI, 4.4 – 1402.8). Level of training did not mitigate the impact of framing bias. The actual case presentations framed toward and away from PE are below if you’re interested. This study had several limitations. There were only 69 participants. Plus, it was a bit artificial; in actual practice, the physician would have the luxury of examining the patient and asking more questions at various points in time. However, it emphasizes the very real effect of framing bias and our need as clinicians to be cognizant of this any time data is shared from one provider to another.
Framing of clinical information affects physicians’ diagnostic accuracy. Emerg Med J. 2019 Aug 8. pii: emermed-2019-208409. doi: 10.1136/emermed-2019-208409. [Epub ahead of print]
Open in Read by QxMD