Clinical Pathways for Diagnosing Neurocognitive Disorders: Insights From Process Mining a Memory Clinic Cohort.

Objectives: Process mining (PM) has only recently been used in medicine. Its implementation in the dementia field could be valuable, considering the epidemiologic breadth of the condition and its economic implications. This proof-of-concept study aims to apply PM in the context of dementia to provide a realistic picture of patients' diagnostic pathways in a memory clinic.

Design: Retrospective observational study.

Setting and participants: A total of 539 medical charts were reviewed to obtain sociodemographic data and type and timing of diagnostic evaluations (eg, clinical or neuropsychological visits, imaging scans, and fluid biomarker analyses).

Methods: We used a PM approach to create a process map from the clinical events and visualize the most common diagnostic pathways in the total cohort and subcohort of patients. PM techniques represent the temporal and dynamic sequence of clinical events in the patients' journeys, overcoming the traditional frequency analyses focused only on aggregate statistics and event distributions.

Results: The results showed that the diagnosis was typically reached during the third clinical visit, following the results of instrumental examinations (ie, morphologic imaging, routine blood and neuropsychological examinations) and a first-line diagnostic biomarker. In mild cognitive impairment (MCI) and mild dementia (DEM) subcohorts, cerebrospinal fluid analyses are the most frequently used first-line biomarkers to ascertain a suspicion of Alzheimer disease (23%). Differential PM analyses revealed that the DEM subcohort underwent morphologic imaging before accessing the memory clinic more often than the MCI subcohort (P < .05).

Conclusions and implications: This preliminary use of PM algorithms in memory clinics shows promising results in visualizing and measuring real-world diagnostic pathways.