Building and validating trend-based multiple sclerosis case definitions: a population-based cohort study for...

Area of Science:

• Epidemiology
• Health Informatics

Background:

• Accurate identification of multiple sclerosis (MS) cases is crucial for epidemiological studies and healthcare management.
• Traditional case definitions may not fully capture disease trends or require extensive data.
• Developing dynamic, trend-based case definitions can improve classification accuracy and efficiency.

Purpose of the Study:

• To develop and validate model-based, trend-based case definitions for multiple sclerosis (MS).
• To apply dynamic classification to determine the average number of data years required for accurate case identification.

Main Methods:

• Retrospective cohort study using data from April 2004 to March 2022.
• Multivariate generalized linear mixed models were used to construct trend-based case definitions.
• Dynamic classification was employed to estimate mean classification time and compare accuracy metrics (sensitivity, specificity, PPV, NPV, PCC, F1-scores) against a deterministic definition.

Main Results:

• Trend-based and deterministic case definitions demonstrated high classification accuracy (PCC > 0.94) when using the full study period.
• Accuracy metrics, particularly sensitivity and PPV for trend-based definitions, were lower when fewer data years were used.
• Dynamic classification identified 5 years as the average trend needed, with reduced accuracy for both definition types when applied to shorter time windows.

Conclusions:

• Trend-based case definitions, when validated, offer comparable accuracy to deterministic methods and population-based clinician assessments.
• Classification accuracy for both trend-based and deterministic definitions is sensitive to the number of data years utilized.
• Dynamic classification is a viable method for determining optimal data duration for trend-based MS case definitions.