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

PURPOSE: Recently developed minimal important change (MIC) estimation methods recover the mean individual MIC in a sample. These methods are the adjusted predictive modeling (APM) method and the longitudinal confirmatory factor analysis (LCFA) method. Both methods require LCFA of patient-reported outcome measure (PROM) data. In the APM-method, LCFA is used to estimate the reliability of the transition ratings, whereas in the LCFA-method, LCFA is used to estimate the latent MIC. However, LCFA cannot be performed if the PROM is a single item measure (SIM). Adding an auxiliary variable, that is correlated with the PROM, to the LCFA-model may be a solution. We developed three different LCFA-models in which an auxiliary variable is included. In this simulation study, we assessed the performance of the APM- and LCFA-methods to recover the true MIC of an SIM. We applied both methods to a real dataset in which the SIM was a numeric rating scale for pain.

METHODS: We simulated 15,552 samples, varying in 11 parameters, and estimated the APM-based and LCFA-based MICs.

RESULTS: The APM-method performed well, except if the proportion improved was high or low, and the present state bias (PSB) was high. The LCFA-method performed well, irrespective of the proportion improved and the PSB. In the real data, the LCFA-based MIC was 17 (on a 100-point scale), whereas the estimated APM-based MIC was 4 points higher, probably due to a high proportion improved and a high PSB.

CONCLUSION: The MIC of an SIM can be accurately estimated using an auxiliary PROM.