Estimating anchor-based minimal important change using longitudinal confirmatory factor analysis.

Area of Science:

• Psychometrics
• Health Outcomes Research
• Statistical Modeling

Background:

• The minimal important change (MIC) quantifies the smallest change in a patient-reported outcome measure (PROM) that holds significance for patients.
• Accurate estimation of MIC is crucial for interpreting clinical trial results and guiding patient care.
• Existing methods for MIC estimation have limitations, necessitating the development of more efficient and reliable approaches.

Purpose of the Study:

• To introduce and evaluate a novel method for estimating MIC using longitudinal confirmatory factor analysis (LCFA).
• To compare the performance of the LCFA-based MIC estimation method against a longitudinal item response theory (LIRT) approach.
• To investigate the impact of sample size on the accuracy and precision of LCFA-based MIC estimates.

Main Methods:

• Simulated 108 datasets with varying characteristics to estimate MIC using both LCFA and LIRT methods.
• Applied LCFA and LIRT methods to real-world PROMIS Pain Behavior data from 909 patients.
• Conducted further simulations with 3888 samples across different sizes (125-1000) to assess sample size effects on LCFA-based MIC.

Main Results:

• LCFA demonstrated comparable accuracy to LIRT in MIC recovery but was over 50 times faster computationally.
• In PROMIS Pain Behavior data, LCFA yielded an MIC of 2.85, while LIRT yielded 2.60.
• Simulations indicated that smaller sample sizes (below 125) reduced the precision of LCFA-based MIC estimates and increased non-convergence rates.

Conclusions:

• LCFA provides an accurate and computationally efficient method for estimating the minimal important change (MIC).
• The LCFA method is robust, but sample sizes exceeding 125 are preferable for enhanced precision and model stability.
• This study validates LCFA as a valuable tool for MIC estimation in patient-reported outcome research.