Data quality evaluation for observational multiple sclerosis registries.

Area of Science:

• Epidemiology
• Data Science
• Neurology

Background:

• High-quality observational data is crucial for real-world research in multiple sclerosis (MS).
• Standardized methods are needed for reproducible evaluation of data quality in large patient cohorts.
• MSBase, a global MS cohort, requires robust data quality assessment.

Purpose of the Study:

• To develop and implement standardized scores for evaluating data quality, density, and generalizability in the MSBase cohort.
• To provide an objective framework for assessing the reliability of real-world MS data.
• To inform the design of future observational studies using MS data.

Main Methods:

• Developed three scores: error rate, data density score, and generalizability score.
• Utilized data from 35,869 patients in the MSBase cohort (as of November 2015).
• Evaluated data completeness, accuracy, consistency, and representativeness of demographic and treatment information.

Main Results:

• Median error rate was 3 per 100 patient-years.
• Generalizability score confirmed sample representativeness of MS epidemiology.
• Moderate correlation between density and generalizability scores (ρ=0.58); weak correlation with error rate (ρ=-0.32 to -0.33).
• Generalizability score strongly correlated with center size (ρ=0.79).

Conclusions:

• The developed scores enable objective, reproducible evaluation of observational MS data quality.
• These metrics enhance the reliability and interpretability of real-world evidence in MS research.
• The scores will influence the design and analysis of future multiple sclerosis studies.