PHYSICIANS' FIRST IMPRESSIONS IN THE DIAGNOSIS OF EARLY CANCERS

Purpose: Physicians' initial hypotheses ("first impressions") are believed to exert strong influence on the final diagnosis, however, the evidence is only anecdotal. This study aimed to investigate the impact of physicians’ first impressions on subsequent information gathering and diagnosis.

Method: We constructed six detailed patient scenarios. Cancer was a possible diagnosis in three of the scenarios, while the other three served as “decoys”. Data collection took place remotely over the Internet. Ninety UK family physicians participated and diagnosed all the scenarios. They saw the same introductory information about each patient and their main health complaint. After this, they could ask questions about the patient, until they were ready to diagnose. A researcher on simultaneous phone communication answered their questions. In two of the cancer scenarios, participants were asked to think aloud. We were interested in their initial verbalizations (first impressions), after they read the introductory patient information but before asking any questions. We coded these as either acknowledging the possibility of cancer or not.

Result: First impressions were strongly associated with final diagnosis: the odds of giving cancer as either the working diagnosis or in the differential were reduced by 74% if cancer was not acknowledged as a possibility from the start: OR 0.26 [0.14 to 0.47] (P<0.001). Physicians with more years in practice were less likely to acknowledge the possibility of cancer at first (OR 0.93 [0.90 to 0.97], P<0.001) and to diagnose it at the end: OR 0.97 [0.95 to 0.99] (P=0.02).

   To measure the extent to which the impact of first impressions on final diagnosis operated via information search, we constructed a mediation model with the number of cancer-related questions as the mediator. The bootstrapped unstandardized indirect effect was -0.10 [-0.17 to -0.04] and statistically significant, explaining 28% of the total effect.

Conclusion: This study provides empirical evidence of the powerful impact of first impressions on diagnosis of early cancers without “alarm” symptoms. We recommend that attempts to improve diagnosis of early cancers should target the early stages of the diagnostic process. The largest part of the influence of first impressions on final diagnosis could not be explained by the extent of information search. We therefore suggest that the remaining influence of first impressions operates via the interpretation of the information gathered.