Construct Validity of the Three Motor-Reduced Subscales of the Developmental Test of Visual Perception - Adolescent and Adult (DTVP-A): A Rasch Analysis Model Evaluation
Brown, Ted, British Journal of Occupational Therapy
Visual perception constitutes the ability to identify, organise and prescribe meaning to the visual information received through the eyes (Hammill et al 1993, Scheiman 1997a, Grieve and Gnanasekaran 2008). This metaskill has been acknowledged to be closely aligned with complex cognitive operations that contextualise raw visual information in terms of one's pre-existing environmental views and other sensory experiences (Bouska et al 1990, Hendee 1997, Grieve and Gnanasekaran 2008). Visual perception is an essential aspect of one's abilty to 'mentally manipulate visual information as needed to solve problems' and 'take action in response to environmental demands' (Kurtz 2006, p33). This, in turn, promotes a person's overall functional performance capacity and allows him or her safely and independently to perform self-care tasks and participate in work or leisure-related activities (Erhardt and Duckman 1997, Scheiman 1997a, Cooke et al 2006, American Occupational Therapy Association [AOTA] 2008). Unfortunately, however, the complexity of the human visual perceptual system renders it 'vulnerable to many dysfunctional skills which can severly impact its effectiveness as an information processing tool' (Su et al 1995, p3).
As a consequence, therapists regularly assess visual perception, particularly when working with clients who present with pre-existing diagnoses commonly associated with visual perceptual problems (Zoltan 1996, Golisz and Toglia 2003, Phipps 2006, AOTA 2008, Brown and Rodger 2009). Diagnoses often presenting with visual perceptual dysfunction include multiple sclerosis, dementia, Parkinson's disease, cerebrovascular accident, brain tumour and acquired brain injury (Bouska et al 1990, Cockburn et al 1990, Neistadt 1990, Ogden et al 1990, Edmans et al 1991, York and Cermak 1995, Vleugels et al 2000, Mercier et al 2001, Glosser et al 2002, Kozeis et al 2007).
One visual perceptual test used by therapists to assess adult visual perception is the Developmental Test of Visual Perception--Adolescent and Adult (DTVP-A) (Reynolds et al 2002). The DTVP-A was developed and standardised in the United States and emerged from a revision and extension of the Developmental Test of Visual Perception 2nd edition (DTVP-2) (Hammill et al 1993). As noted in its manual, the 'DTVP-A is presented as, in essence, an age extension of the DTVP-2' (Reynolds et al 2002, p v). Given that health care professionals outside the United States often use the DTVP-A to assess the visual perception of adult-age clients, it is important to explore its validity when used in cross-cultural contexts (Leverett 2005, Brown et al 2008).
Current measurement theory argues that the development of instruments must begin with establishing their construct validity; that is, assessing how well the items of instruments, collectively and individually, capture a single cohesive theoretical concept that they seek to measure. In other words, the validity of an instrument rests on the ability of its items to reflect an underlying construct (Nunnally and Bernstein 1994). According to Rasch theorists, empirical validation of test quality requires (at least) that the items of an instrument exhibit four properties: (1) potential to yield interval level scaling (scalability); (2) unidimensionality; (3) lack of differential item functioning across different subjects and test occasions; and (4) hierarchical ordering (Doble and Fisher 1998). 'When these properties are demonstrated, then the items act together as a "ruler", indicating amounts of the construct that the scale seeks to measure' (Mallinson et al 1998, p222). This study, therefore, addresses this issue by assessing the construct validity of the DTVP-A's three motor-free subscales using the Rasch analysis model in an Australian setting (Bond and Fox 2007).
In the past, validity was defined as three separate types: content, criterion and construct, with criterion-related validity subdivided into concurrent and predictive validity (Nunnally and Bernstein 1994, Anastasi and Urbina 1997). …