Explainable machine learning-driven models for predicting Parkinson's disease and its prognosis: obesity patterns...

Area of Science:

• Neurology
• Public Health
• Data Science

Background:

• Parkinson's disease (PD) is a common neurodegenerative disorder.
• The impact of obesity on PD risk and outcomes is debated.

Purpose of the Study:

• Investigate associations between obesity patterns and PD risk.
• Examine the link between obesity patterns and all-cause mortality in PD patients.
• Develop machine learning (ML)-driven predictive and prognostic models for PD.

Main Methods:

• Utilized data from 51,394 adults (NHANES 1999-2018).
• Classified participants into four obesity patterns using BMI and waist circumference (WC).
• Employed multivariable logistic and Cox regression for risk and mortality analyses.
• Developed and validated ML nomograms for PD prediction and prognosis using AUCROCs and calibration curves.

Main Results:

• Compound obesity was associated with significantly increased PD risk (OR=1.83, P<0.001).
• Compound obesity correlated with reduced all-cause mortality in PD patients (HR=0.43, P=0.003).
• ML models achieved moderate performance for PD prediction (AUCROC=0.75) and prognosis (AUCROC=0.72).

Conclusions:

• Compound obesity is linked to higher PD risk but lower mortality in PD patients.
• Validated ML nomograms demonstrate robust predictive and prognostic capabilities for PD.
• Further longitudinal studies are needed to confirm these findings.