Abstract
Abstract
Background
The Forgotten Joint Score (FJS) is a 12-item patient reported outcome measurement instrument. It was developed with classical test theory, without testing assumptions such as unidimensionality (all items reflect one underlying factor), appropriate weighting of each item, no differential item function (DIF, different groups answer the same way), and monotonicity (people with higher function have higher score). We applied item response theory (IRT) to improve the validity of FJS to contemporary standards to optimise it for ongoing use.
Research Questions
Does the FJS reflect one latent trait? Can an IRT model be fitted to the FJS to provide interval-scaled measurement?
Methodology
Participants undergoing primary total knee replacement provided pre-operative and post-operative (6-months) responses for FJS. An exploratory factor analysis (EFA), confirmatory factor analysis (CFA) and Mokken analysis were conducted. The data were fitted to a graded response model (GRM).
Results
1288 response patterns were analysed. EFA showed a one factor model (all 12 items load to one underlying trait). CFA demonstrated excellent model fit (X2 <0.001, Tucker Lewis Index=0.96, Comparative Fit Index=0.96). Items did not have equal weighting. The FJS demonstrated good monotonicity with no differential item functioning by sex, age, or body mass index.
Conclusions
The FJS meets key validity assumptions supporting its use in clinical practice and research. The IRT-adapted FJS provides continuous measurements with greater granularity including individual measurement error. This adapted score has advantages over traditional FJS scoring, being interval scaled (using GRM) and can be retrospectively applied to existing response sets.
