Objective: To identify clinical and sociodemographic factors associated with cognitive decline (CD) in PD using a Random Forests model, classifying people with PD (PwP) into 3 cognitive trajectory groups.
Background: CD in PD varies widely among PwP, impacting disease management and quality of life (QoL). Some PwP maintain cognitive function, while others experience an accelerated decline. Understanding the factors influencing these trajectories is critical for early intervention and personalized treatment strategies.
Method: This observational longitudinal study analyzed data from PwP who underwent at least 2 clinical assessments, including MoCA measurements. Using k-means clustering classified cognitive trajectories as maintained cognition, expected decline, or accelerated decline. Key covariates included demographics, disease duration, diagnostic latency, socioeconomic factors, comorbidities, benzodiazepine use, and motor/QoL indicators (MDS-UPDRS parts 2–4, PDQ-39i). A Random Forest classifier was trained on 19 features. Model performance was evaluated using a confusion matrix, precision, recall, and F1-score, and feature importance analysis identified key predictors of CD.
Results:
A total of 345 PwP were included (56.4% male, 62.6 ± 12.7 years (y) at baseline). The follow-up period had a mean duration of 453.7 ± 367.6 days. Mean disease duration at baseline of 6.2 ± 5.1 y. The average number of years of education was 10.0 ± 5.1 y, with a delay in diagnosis of 2.3 ± 4.3 y. Distribution by cognitive progression was 48.9% in the accelerated decline and 36.9% in the expected decline. The Random Forest Classifier achieved an accuracy of 52.17% in predicting cognitive changes. Key predictive factors for an accelerated decline are displayed in [figure 1].
Conclusion: CD in PD is influenced by motor worsening, lower education, older age, poorer QoL, longer diagnostic delay, motor fluctuations, and declining socioeconomic status. Identifying these factors may aid in early intervention and personalized treatment strategies.
Feature importance in cognitive progression plot
To cite this abstract in AMA style:
AJ. Hernández-Medrano, A. Cervantes-Arriaga, MF. Medina-Pérez, D. Ulloa-Hernández, MF. Velasco-Delgado, GA. Martin-Mafud, AE. López-Lobato, D. Náfate-Wences, M. Rodríguez-Violante. Determinants of Cognitive Progression in Parkinson’s Disease: A Machine Learning Approach [abstract]. Mov Disord. 2025; 40 (suppl 1). https://www.mdsabstracts.org/abstract/determinants-of-cognitive-progression-in-parkinsons-disease-a-machine-learning-approach/. Accessed October 5, 2025.« Back to 2025 International Congress
MDS Abstracts - https://www.mdsabstracts.org/abstract/determinants-of-cognitive-progression-in-parkinsons-disease-a-machine-learning-approach/