The latent difference score model is a viable alternative to arithmetic difference score-based anchor-driven minimal...

Area of Science:

• Biostatistics
• Clinical Research
• Health Outcomes

Background:

• Traditional methods for estimating Minimal Important Change (MIC) from Patient-Reported Outcome Measures (PROMs) are susceptible to measurement error.
• Latent Data Structure Models (LDSM) present a robust alternative for estimating MIC while accounting for this inherent error.
• A framework for calculating anchor-based MICs is proposed, addressing multi-step considerations.

Purpose of the Study:

• To illustrate the application of Latent Data Structure Models (LDSM) for estimating Minimal Important Change (MIC).
• To compare LDSM-based MIC estimates against traditional methods in the context of knee osteoarthritis.
• To evaluate the performance of LDSM in controlling for measurement error in PROMs.

Main Methods:

• Utilized data from the KASTPain multi-center randomized clinical trial involving 364 participants with knee osteoarthritis post-arthroplasty.
• Employed a structural equation modeling framework to implement Latent Data Structure Models (LDSM).
• Estimated MICs for three common PROMs assessing pain and functional status.

Main Results:

• Nine MIC estimates were generated using LDSM, crossing three PROMs with three post-intervention measurement occasions.
• Receiver operating curve (ROC) analysis demonstrated high predictive accuracy, with a median area under the curve (AUC) of 0.88 (range: 0.77-0.90).
• LDSM effectively controlled for PROM measurement error in MIC estimation.

Conclusions:

• Latent variable modeling of PROM change offers a method to exclude measurement error.
• This approach may provide significant advantages over traditional methods for estimating Minimal Important Change (MIC).
• LDSM is a valuable tool for accurate health outcomes research, particularly in chronic conditions like knee osteoarthritis.