Uso da reflexão deliberada como estratégia para reduzir o viés de confirmação diagnóstica em residentes de ortopedia

Abstract

Introduction: Clinical reason as well as other decision-making processes seems to be subject to cognitive bias, among which validation bias stands out. Strategies, which stimulate deep thinking on decision-making in a structured manner, can help to reduce these biases and their use in orthopaedic trauma resolution cases need deeper evaluation. Objectives: To verify whether deliberate deep thinking reduces validation bias and increases diagnostic accuracy among orthopaedic residents in resolution of written clinical cases. Methods: An experimental study comparing diagnostic accuracy in resolution of eight clinical cases among a group that used deliberate deep thinking (GR) and a control group that reflected freely (GC). To induce validation bias, in all cases a referral diagnosis was mentioned, and for half of the cases this diagnosis was correct and for the other half, incorrect. Results: 55 third year orthopaedic residents participated in the study, 27 of whom were assigned to GR and 28 to GC. Regarding diagnostic accuracy, GR has shown higher scores than those encountered in GC, in clinical cases where the referral presented the correct diagnosis (62.0 ± 20.1 vs. 49.1 ± 21.0 respectively; p=0.021). In cases with incorrect referral diagnosis, the diagnostic accuracy was similar amongst residents of GR and those of GC (39.8 ± 24.3 vs. 44.6 ± 26.7 respectively; p=0.662). By the response analysis in both groups concerning cases containing incorrect referral diagnosis, a validation bias was observed in 26% of the initially provided diagnoses and 19.5% on final diagnoses. GR residents showed a reduction of validation of incorrect referral diagnosis, comparing the diagnosis given in a non-analytical manner with the final diagnosis, provided after deliberate deep thinking (25.9 ± 17.7 vs. 17.6 ±18.1, respectively; Cohen d: 0.46; p=0.003). In GC, the diagnosis incorrect validation reduction compared with the non-analytical diagnosis provided after free deep thinking was not statistically significant. Conclusion: GR showed greater diagnostic accuracy than GC in correct referral cases resolution. Validation bias presence was verified in clinical cases with incorrect referrals, and deliberate deep thinking contributed to reducing this bias. Despite the reduction of validation bias, the diagnostic accuracy of the GR residents was similar to that of the GC in resolution of this set of cases