257: Confirmation bias affects estimation of blood loss and amniotic fluid volume: a randomized simulation-based trial

Abstract

To determine whether confirmation bias affects the diagnostic accuracy of obstetrical providers. We hypothesized that the presence or absence of confirming or un-confirming data affects clinical diagnoses. We conducted a randomized simulation-based trial. Block-randomization was performed by type of service: physicians, nurses, and allied health professionals. In both scenarios, participants were randomized to one of two arms, the difference between them being the presence or absence of a confirming diagnostic clue. In the first scenario, providers were asked to estimate blood loss (EBL) in the presence or absence of reported hypotension. Two different blood loss (BL) volumes were used, 500 mL and 1500 mL, with crossover. In a second scenario, participants viewed the same four quadrants images of an amniotic cavity with normal amniotic fluid (AF) volume, and, after being told either that the patient was hypertensive or normotensive, were asked to describe the volume (dichotomized as either oligohydramnios or as “good” [amniotic fluid is low normal and above]). In the first scenario, there were 84 participants. After adjusting for job experience, there were significant main effects of both actual BL (p<0.001) and blood pressure (BP) status (p=0.024), with no significant interaction (p=0.781). While there was an overall difference between the BP status conditions, power consideration precluded a definitive statement regarding a BP status effect in either the 500 mL (p=0.080) or the 1500 mL (p=0.462) stratum (table 1). BL was underestimated in all four study conditions. In the AF scenario, there were 58 participants (nurses were excluded). After adjusting for job experience, there was a significant difference between the two groups (p=0.032). Providers in the chronic hypertension arm were more likely to estimate the AF as low in comparison to the non-hypertensive arm (table 2). Overall, BL is underestimated by obstetrical providers. In the presence of a confirming clinical clue (hypotension), this underestimation is less pronounced. Similarly, in the presence of a confirming clinical clue (chronic hypertension), amniotic fluid volume is more likely to be “underestimated” as oligohydramnios. Confirmation bias affects estimation of BL and AF. Providers need to be aware of the effect of confirmation bias on diagnosis since it may affect their subsequent management decisions.View Large Image Figure ViewerDownload Hi-res image Download (PPT)