Comparing the different sets of item-diagnostic criteria of the Coma Recovery Scale-Revised (CRS-R): a measurement-based approach driven by Rasch analysis

Objectives: (1) to replicate the assessment of the internal construct validity of the Coma Recovery Scale-Revised (CRS-R) within the Rasch Measurement Theory framework using a larger multicenter sample size; (2) to compare the different sets of item diagnostic criteria against the measurement ruler constructed from Rasch analysis to understand how those criteria relate to the overall level of persons' consciousness.

Design: Multicenter retrospective study.

Setting: Seven centers.

Participants: 380 inpatients with a disorder of consciousness (DOC) with one or more observations, for a total sample of 1460 observations.

Interventions: Not applicable.

Main outcome measure: CRS-R.

Results: We created two subsamples: a validation subsample of one randomized assessment per subject (N=380) and a confirmation subsample using the further available assessments per subject (N=347). The Rasch analyses, conducted on the validation subsample and replicated on the confirmation one, demonstrated adequate satisfaction of all the model's requirements, including monotonicity, unidimensionality, local independence, invariance (χ²_df=40.224; p=.020), and absence of significant Differential Item Functioning (DIF) across all person factors explored, including etiology. The reliability (Person Separation Index>.870) was sufficient for individual person measurement, with the distinction of five Distinct Levels of Performance Ability. The CRS-R rulers based upon the Rasch calibration allowed the visual comparison of the various sets of DOC diagnostic criteria available, suggesting the possibility of a further refinement of these criteria.

Conclusions: This study improved the results of a previous Rasch analysis published in 2013. It delivered a new stable Rasch calibration of the CRS-R within the largest multicenter sample size available to date and without any DIF by patient's etiology. The adopted measurement-based approach provided further insights into the diagnostic meaning of several score categories of the CRS-R, confirming previous findings and suggesting that 'automatic motor response' (item: motor function) and 'object recognition' (item: visual function) are likely to represent behavioral manifestations of MCS+ and eMCS, respectively.