Estimating the minimal important change of single-item measures using the adjusted predictive modeling method or the...

Area of Science:

• Health Research
• Psychometrics
• Biostatistics

Background:

• Minimal important change (MIC) estimation is crucial for interpreting patient-reported outcome measures (PROMs).
• Existing methods like Adjusted Predictive Modeling (APM) and Longitudinal Confirmatory Factor Analysis (LCFA) require specific data structures.
• LCFA cannot be directly applied to single-item measures (SIMs), limiting MIC estimation for these common PROMs.

Purpose of the Study:

• To evaluate the performance of APM and LCFA methods in recovering the true MIC of SIMs.
• To assess the feasibility of using an auxiliary variable within LCFA models to overcome SIM limitations.
• To compare the accuracy of APM-based and LCFA-based MIC estimations for SIMs.

Main Methods:

• A simulation study involving 15,552 samples with 11 varying parameters was conducted.
• Three LCFA models incorporating an auxiliary variable were developed.
• Both APM and LCFA methods were applied to simulated data and a real-world pain rating scale dataset.

Main Results:

• The LCFA-method demonstrated robust performance across different proportions of improvement and levels of present state bias (PSB).
• The APM-method's performance was compromised by high or low proportions of improvement and high PSB.
• In real data, the LCFA-based MIC was 17, while the APM-based MIC was 4 points higher.

Conclusions:

• The study confirms that MIC for SIMs can be accurately estimated by incorporating an auxiliary PROM into LCFA models.
• The LCFA-method with an auxiliary variable is a reliable approach for MIC estimation of SIMs.
• The findings suggest LCFA-based MIC estimation is more stable than APM-based estimation for SIMs, especially under challenging data conditions.