Predicting Quality of Life in Parkinson's Disease: A Machine Learning Approach Employing Common Clinical Variables

Area of Science:

• Neurology
• Medical Informatics
• Machine Learning

Background:

• Parkinson's Disease (PD) significantly affects health-related quality of life (HRQoL).
• Assessing HRQoL in PD is challenging due to subjective patient experiences.
• The Parkinson's Disease Questionnaire-39 (PDQ-39) is a common assessment tool.

Purpose of the Study:

• Develop a machine learning (ML) model to predict HRQoL outcomes in Parkinson's Disease.
• Utilize accessible clinical data for predictive modeling.
• Employ explainable ML to identify key factors influencing HRQoL in PD patients.

Main Methods:

• Analysis of data from the Parkinson's Real-world Impact Assessment study (PRISM).
• Inclusion of 627 complete patient observations from six European countries.
• Development of an ensemble ML model with a 90% training and 10% validation split.

Main Results:

• The ML model showed strong performance on the training set (R²=0.75).
• Moderate predictive accuracy was observed on the validation set (R²=0.36).
• Identified key predictors including age at diagnosis, country, dementia, and patient age.

Conclusions:

• A robust ML model can predict the impact of Parkinson's Disease on quality of life using clinical variables.
• ML holds potential for enhancing clinical decision-making and patient care in PD.
• Future research should focus on improving model generalizability and applicability.