Evaluation of diagnostic concordance between algorithms for Parkinson’s disease dementia

Background: Significant number of patients with Parkinson’s disease (PD) gradually progress to Parkinson’s disease dementia (PDD). Recent proposal for updating current diagnostic criteria for PDD recommends alternative screening tests and broader functional assessments. Objective: To evaluate the diagnostic concordance among algorithms for PDD based on Level I (i.e., screening) criteria and to assess their predictive validity for Level II (i.e., neuropsychological battery) diagnosis. Methods: A cross-sectional retrospective analysis of 190 patients with PD who underwent a comprehensive neuropsychological assessment. A total of 68 diagnostic algorithms were operationalized using combinations of scores derived from the Mini-Mental State Examination (MMSE) and the Montreal Cognitive Assessment (MoCA). Functional impairment was based either on the Functional Assessment Questionnaire (FAQ) item 9 or the FAQ total score. Diagnostic concordance was evaluated using Cohen’s κ. Predictive validity for Level II classification was assessed using projection predictive variable selection. Results: Estimated PDD rates ranged from 2.1% up to 16.8%. Concordance was moderate to high among algorithms using the same functional impairment definition (κ FAQ total = 0.75, κ FAQ 9 = 0.86) but substantially lower when functional impairment definitions differed (κ = 0.43). A parsimonious screening model combining MoCA Five Words and MMSE Sevens adequately approximated Level II classification and yielded a prevalence-adjustable heuristic decision rule. Conclusions: Diagnostic outcomes for PDD are sensitive to the choice of cognitive and functional instruments suggesting that different algorithms may capture partially distinct constructs. Minimal screening using selected items can approximate Level II PDD diagnosis, but this simplification entails a trade-off between sensitivity and specificity.