Handling missing values in the MDS-UPDRS.

Area of Science:

• Neurology
• Clinical Assessment
• Biostatistics

Background:

• The Movement Disorder Society Unified Parkinson's Disease Rating Scale (MDS-UPDRS) is crucial for Parkinson's disease assessment.
• Handling missing data in the MDS-UPDRS is essential for accurate scoring and interpretation.
• Existing imputation strategies require validation for scale-specific data.

Purpose of the Study:

• To determine the maximum number of permissible missing values for each MDS-UPDRS part to ensure valid total scores.
• To establish data-driven thresholds for imputing missing MDS-UPDRS scores or rejecting incomplete ratings.
• To provide evidence-based guidelines for handling missing data across different Hoehn and Yahr stages.

Main Methods:

• Utilized two large MDS-UPDRS datasets to simulate missing item scores (consistent and random deletions).
• Employed Lin's Concordance Correlation Coefficient (CCC) to compare complete scores with prorated scores.
• Identified the maximum missing values retaining a CCC > 0.95 as the validity threshold.
• Validated findings using a confirmatory sample from the MDS-UPDRS international translation program.

Main Results:

• For consistent missing items across all Hoehn and Yahr stages: 1 in Part I, 1 in Part II, 3 in Part III, 0 in Part IV.
• For random missing items across all Hoehn and Yahr stages: 1 in Part I, 2 in Part II, 7 in Part III, 0 in Part IV.
• All identified thresholds were confirmed in the validation sample.

Conclusions:

• Defines specific, actionable thresholds for missing data in the MDS-UPDRS, enabling valid surrogate score creation.
• Provides scientific justification for accepting or rejecting partially completed MDS-UPDRS ratings based on missing item counts.
• Enhances the reliability and consistency of MDS-UPDRS assessments in clinical research and practice.