Feasibility Study Using Propensity Score Matching Methods for the Pseudo-Common Person Equating Requirement.

Area of Science:

• Health outcomes research
• Psychometrics
• Rehabilitation science

Background:

• The Rasch model's unidimensionality assumption is crucial for linking similar rehabilitation instruments.
• Propensity score (PS) matching is a statistical technique used to control for confounding variables.
• Equating methods aim to make scores from different tests comparable.

Purpose of the Study:

• To evaluate if propensity score (PS) matching supports the unidimensionality assumption of the Rasch model.
• To determine the suitability of PS matching for equating rehabilitation instruments across different groups.

Main Methods:

• Utilized data from 1,013 respondents of the 2009 Hispanic Established Populations for Epidemiologic Studies of the Elderly Frailty study.
• Employed a unidimensional item pool comprising 10 SF-36 physical function and nine activities of daily living items.
• Applied PS matching based on respondent functionality (high and low) and conducted exploratory factor analysis to assess unidimensionality in the matched sample.

Main Results:

• The matched sample exhibited two distinct measurement structures, indicating a failure to meet the unidimensionality assumption.
• Exploratory factor analysis revealed excellent model fit for these two structures, not a single dimension.
• Propensity score matching did not successfully mimic the common-person assumption required for equating.

Conclusions:

• Propensity score matching alone does not ensure the unidimensionality required for the Rasch model when linking rehabilitation instruments.
• The combination of PS matching and common-person equating may be inappropriate for comparing scores from different groups using rehabilitation instruments.
• Further research is needed to explore alternative methods for equating rehabilitation instruments across diverse populations.