Rasch Analysis of the Mayo-Portland Adaptability Inventory for Posthospital Brain Injury Rehabilitation

James F Malec; Daniel M Logan; Claire McGrath; Devan Parrott; G Joseph Walters

doi:10.1097/HTR.0000000000000935

Rasch Analysis of the Mayo-Portland Adaptability Inventory for Posthospital Brain Injury Rehabilitation

J Head Trauma Rehabil. 2024 Feb 27. doi: 10.1097/HTR.0000000000000935. Online ahead of print.

Authors

James F Malec¹, Daniel M Logan, Claire McGrath, Devan Parrott, G Joseph Walters

Affiliation

¹ Author Affiliations: Foundation to Advance Brain Rehabilitation (FABR), and Physical Medicine and Rehabilitation, Indiana University School of Medicine, Indianapolis (Drs Malec and Parrott); Mayo Clinic, Rochester, Minnesota (Dr Malec); On With Life, Ankeny, Iowa (Messrs Logan and Walters); and Bancroft NeuroRehab, Cherry Hill, New Jersey (Dr McGrath).

PMID: 38453628
DOI: 10.1097/HTR.0000000000000935

Abstract

Objectives: (1) Reexamine the item structure and reliability of the Mayo-Portland Adaptability Inventory (4th ed; MPAI-4) through Rasch analysis of admission and discharge scores for a large sample of adults with acquired brain injury (ABI) who participated in various types of posthospital brain injury rehabilitation (PHBIR) programs; (2) compare differential item functioning (DIF) for traumatic brain injury (TBI), stroke and other ABI; and (3) explore the viability of more specific subscales in addition to the established indices.

Setting: Data from Residential Neurobehavioral, Residential Neurorehabilitation, Home and Community, Day Treatment, and Outpatient rehabilitation programs serving individuals with ABI.

Participants: A total of 2154 individuals with TBI, stroke, or other ABI.

Design: Retrospective analysis of de-identified admission and discharge data from the Foundation to Advance Brain Rehabilitation (FABR) consortium database.

Main measure: MPAI-4.

Results: After adjusting 4 misfitting items and eliminating 20 misfitting persons, the MPAI-4 demonstrated real person reliability/separation = 0.93/3.52 and real item reliability/separation = 1.00/24.02. Independent Rasch analyses by diagnostic category found similar reliabilities and separations. Residual item correlations and principal component analysis of residuals (PCAR) indicated areas of local dependence arranged hierarchically reflecting the full-scale item hierarchy and providing the basis for 3 new subscales of Physical Abilities, Cognitive Abilities, and Autonomy. DIF across diagnostic categories revealed differences in item elevations characteristic of typical patients in each category. Measure means and SDs were very similar across categories.

Conclusions: MPAI-4 items demonstrate very good person and item reliabilities for individuals with TBI, stroke, and other ABI at a level that supports individual evaluation. Variations in item calibrations across diagnostic categories reflect the differential characteristics of typical patients within categories. The entire measure provides an overall assessment of common sequalae of ABI, and standard indices used in combination with newly derived subscales provide more specific assessments of rehabilitation needs for treatment planning.