Biais cognitifs dans le diagnostic de maladie d’Alzheimer

Abstract

Objectives: Cognitive biases (CB), an important source of medical errors, have been investigated in many clinical specialties. They have never been demonstrated in the field of neurocognitive disorders and Alzheimer’s disease, where the diagnostic accuracy is complex, particularly regarding their differential diagnoses. This study aims to establish the presence of CB in this field, their frequency and to understand the context favouring their appearance, as well as to propose protective strategies (i.e., “debiasing”). Method: 15 clinicians in memory consultation (neuropsychologists, doctors, nurses) took part in interviews in 2020, including a structured interview, a discussion based on a list of CBs referenced in medicine and a reflective writing exercise relating to an experience involving a bias in the diagnosis of Alzheimer’s disease (AD). Among these biases where fixing on the patient’s initial presentation and not adjusting one’s diagnostic hypothesis despite the presence of new elements (“anchoring bias”), being influenced by the clinical context or the way in which the information is transmitted (“frame effect”), looking for clinical elements that validate the diagnostic hypothesis and not those that would redirect it (“confirmation bias”), looking for typical manifestations of a disease and missing atypical clinical pictures (“representativeness bias”) and being influenced by the patient’s psychiatric context (“psychosis error”). The frequency of CBs was analysed semi-quantitatively and then compared between professions and age groups. The narratives were analysed qualitatively to identify contextual factors and possible protective strategies. Results: The most frequently cited CBs in the interviews were confirmation bias (86.67%), representativeness bias (80%), psychosis error (60%), frame effect (46.67%) and anchoring bias (46.67%). Differences were noted between the subgroups. Neuropsychologists more often noted the multiple alternatives CB, while doctors more frequently mentioned the overconfidence and availability CB. Juniors frequently described multiple alternatives CB, psychosis error and confirmation CB, while seniors more frequently noted confidence error and outcome bias. Participants mentioned patient, caregiver, diagnostic procedure and clinical picture elements to explain their biases. They cited several protective strategies including a systematic approach to diagnosis, regular monitoring of their progress, interdisciplinarity and frequent supervision. Conclusion: CBs are found in Memory Centres as everywhere else in medicine and are influenced by age and profession. Protective factors to prevent CBs can be integrated into the diagnostic and follow-up process.