Response shift in patients with multiple sclerosis: an application of three statistical techniques

Area of Science:

• Psychometrics
• Statistical Modeling
• Health Outcomes Research

Background:

• Response shifts, including recalibration, reprioritization, and reconceptualization, can impact patient-reported outcomes.
• Accurate detection of these shifts is crucial for interpreting longitudinal health data.
• Existing statistical techniques for detecting response shifts require comparative validation.

Purpose of the Study:

• To evaluate and compare the performance of three statistical techniques for detecting response shifts.
• To validate cross-method findings using a single patient sample.

Main Methods:

• Employed Structural Equation Modeling (SEM), Latent Trajectory Analysis (LTA), and Recursive Partitioning and Regression Tree (RPART) modeling.
• Utilized a sample of 3,008 patients from the North American Research Committee on Multiple Sclerosis (NARCOMS) Registry.
• Assessed patient-reported outcomes including disease-specific Performance Scales, Patient-Derived Disease Steps, and the SF-12v2 health survey.

Main Results:

• All three statistical methods detected small response shift effect sizes.
• SEM identified recalibration response shift.
• RPART demonstrated patterns consistent with all three types of response shift, while LTA detected response shift in less than 1% of the sample without distinguishing types.

Conclusions:

• The study provides a comparative overview of statistical techniques for response shift detection.
• Findings highlight the strengths and limitations of each method regarding operationalization, interpretability, and assumptions.
• Further research directions for refining response shift detection methodologies are discussed.