• Stephen M Haley, Wendy J Coster, Patricia L Andres, Larry H Ludlow, Pengsheg Ni, Tamara L Y Bond, Samuel J Sinclair, and Alan M Jette.
• Research and Training Center on Measuring Rehabilitation Outcomes, Center for Rehabilitation Effectiveness, Sargent College of Health and Rehabilitation Sciences, Boston University, Boston, Massachusetts 02215, USA. smhaley@bu.edu
• Med Care. 2004 Jan 1; 42 (1 Suppl): I49-61.

BackgroundEfforts to evaluate the effectiveness of a broad range of postacute care services have been hindered by the lack of conceptually sound and comprehensive measures of outcomes. It is critical to determine a common underlying structure before employing current methods of item equating across outcome instruments for future item banking and computer-adaptive testing applications.ObjectiveTo investigate the factor structure, reliability, and scale properties of items underlying the Activity domains of the International Classification of Functioning, Disability and Health (ICF) for use in postacute care outcome measurement.MethodsWe developed a 41-item Activity Measure for Postacute Care (AM-PAC) that assessed an individual's execution of discrete daily tasks in his or her own environment across major content domains as defined by the ICF. We evaluated the reliability and discriminant validity of the prototype AM-PAC in 477 individuals in active rehabilitation programs across 4 rehabilitation settings using factor analyses, tests of item scaling, internal consistency reliability analyses, Rasch item response theory modeling, residual component analysis, and modified parallel analysis.ResultsResults from an initial exploratory factor analysis produced 3 distinct, interpretable factors that accounted for 72% of the variance: Applied Cognition (44%), Personal Care & Instrumental Activities (19%), and Physical & Movement Activities (9%); these 3 activity factors were verified by a confirmatory factor analysis. Scaling assumptions were met for each factor in the total sample and across diagnostic groups. Internal consistency reliability was high for the total sample (Cronbach alpha = 0.92 to 0.94), and for specific diagnostic groups (Cronbach alpha = 0.90 to 0.95). Rasch scaling, residual factor, differential item functioning, and modified parallel analyses supported the unidimensionality and goodness of fit of each unique activity domain.ConclusionsThis 3-factor model of the AM-PAC can form the conceptual basis for common-item equating and computer-adaptive applications, leading to a comprehensive system of outcome instruments for postacute care settings.

32 keywords...

hide…