Identification of perception gaps between physicians and patients with neurological diseases and the prediction of these gaps using machine learning

Abstract Understanding perception and communication gaps between patients with neurological diseases and their treating physicians is essential for optimizing patient-centered care. The GAP-AI study aimed to identify these gaps in a cohort of patients with Parkinson’s disease, multiple sclerosis, or epilepsy. This single-center observational study involved patients (N = 197) and their treating physicians (N = 12) answering questionnaires (18-item Patient Satisfaction Questionnaire Short Form, 9-item Shared Decision Making Questionnaire for patients and physicians, Barthel Index, and 36-item Short Form subdomains) over two clinic visits. The primary outcome was the difference between pairwise items in the questionnaires (perception gap). Perception gaps, albeit minimal, were identified for patient satisfaction, shared decision-making, activities of daily living, and quality of life. Attributes that significantly influenced perception gaps included physician’s age, years of experience/holding a neurologist qualification, disease area, and the number of patients treated, with experienced physicians tending to provide more rigorous evaluations than their patients’ self-assessments. Multiple machine learning algorithms were used to develop predictive models based on study data. The k-nearest neighbors algorithm demonstrated the best performance in predicting a patient–physician perception gap. Insights from our study highlight the potential to recognize, predict, and ultimately address these gaps, thus enhancing clinical practice by increasing the level of understanding between patients and their physicians.