Validation of the Contextual Assessment Inventory for problem behavior.(Report)

Problem behavior is a major barrier to successful community integration for people with developmental disabilities. Recently, there has been increased interest in identifying contextual factors involving setting events and discriminative stimuli that impact the display of problem behavior. The authors previously developed the Contextual Assessment Inventory and evaluated it for efficiency, comprehensiveness, comprehensibility, and reliability. This study further evaluated this inventory with respect to convergent and predictive validity. Convergent validity was examined for 17 participants with developmental disabilities through a review of community residence log entries that included a record of the contextual events associated with each episode of problem behavior. Predictive validity was evaluated for a subset of 5 participants through direct observation of contextual events. Results indicated that the inventory had both convergent and predictive validity. Implications for extending contextual assessment and using such information to develop intervention strategies are explored.

Keywords: autism; problem behavior; functional assessment; setting events; antecedent control

**********

People with developmental disabilities often display severe problem behavior such as aggression, property destruction, and self-injury. These behaviors have been associated with limited community integration, higher rates of institutionalization, social rejection, and lowered self-esteem (e.g., Bruininks, Hill, & Morreau, 1988; Koegel, Koegel, & Dunlap, 1996) and with increased stress in family members, direct care staff, and teachers (Lucyshyn, Dunlap, & Albin, 2002). Therefore, much effort has been expended in assessing and treating such behavior.

Following an extensive meta-analysis of the published literature, we concluded that functional assessment is critical for the design of effective interventions (Carr et al., 1999). Functional assessment has been the primary method of determining the factors that maintain problem behavior, factors that include getting attention from others (Durand, Crimmins, Caulfield, & Taylor, 1989), avoiding or escaping from aversive situations (Carr, Newsom, & Binkoff, 1980), obtaining tangible items (Durand & Crimmins, 1988), avoiding social interaction (Taylor & Carr, 1992), and generating sensory reinforcement (Favell, McGimsey, & Schell, 1982).

Several methods of assessment, including experimental manipulation, direct observation, and interviews/checklists have been employed to help identify the functions of problem behavior (Cart et al., 1994; Desrochers, Hile, & Williams-Moseley, 1997; O'Neill et al., 1997; O'Neill, Homer, Albin, Storey, & Sprague, 1990). However, research suggests that it is not always possible to identify the functions of problem behavior (McGill, 1999), and even when it is possible, problem behavior is typically influenced by context (Kennedy & Meyer, 1996; McGill, 1999; McGill, Teer, Rye, & Hughes, 2003; O'Reilly, 1997). That is, function does not exist in a vacuum; it is embedded in and influenced by specific contexts that involve physical, activity/routine, social, and biological factors. Identifying these contexts may help generate additional opportunities for intervention.

Problem behavior is thought to be associated with two classes of contextual variables: discriminative stimuli and setting events (Carr & Smith, 1995). A discriminative stimulus is an event in whose presence a response is reliably reinforced (Skinner, 1938). Because the stimulus predicts reinforcement, contingent on performance of the response, the response is more likely to be exhibited in the future whenever the discriminative stimulus is presented. A setting event is a variable that momentarily alters ongoing stimulus-response relationships (Bijou & Baer, 1961, 1978; Kantor, 1959). In other words, an individual's response to a particular discriminative stimulus may differ depending on the presence or absence of the setting event. For example, an individual who is ill may frequently display self-injurious behavior when given a difficult task to complete and, yet, when well, successfully complete that same task without displaying self-injurious behavior. For this individual, physical illness is considered a setting event because it alters the association between a discriminative stimulus (task presentation) and a response (self-injury).

The assessment of context makes use of the same general strategies employed in the assessment of function, namely, experimental manipulation, direct observation, and interviews/checklists. Experimental manipulation (Dunlap et al., 1994; Dunlap, Kern-Dunlap, Clarke, & Robbins, 1991; Iwata, Dorsey, Slifer Bauman, & Richman, 1982) involves the systematic alteration of contextual events hypothesized to contribute to problem behavior in order to identify cause--effect relationships. As such, it provides the most compelling evidence regarding the influence of contextual factors on the occurrence of problem behavior. Although experimental manipulation provides the best evidence for causality, there are many drawbacks to this method (Desrochers et al., 1997). First, it is often time-consuming. Second, many contextual variables are difficult, if not impossible, to manipulate (e.g., menstrual discomfort). Third, even if manipulation of contextual variables is possible, it requires expertise on the part of the clinician to successfully carry out. Unfortunately, many clinicians do not have the necessary background or training to conduct such assessments. Because of inadequate training, school personnel, in particular, often lack the specific skills and expertise to implement assessment strategies in an expert manner (Hendrickson, Gable, Conroy, Fox, & Smith, 1999). Finally, due to the severity of the problem behavior that some individuals display, it may be unethical to intentionally manipulate contexts that are associated with destructive behaviors.

Direct observation is a second method used to assess context. Carr et al. (1994) elaborated on the use of one common variation of this method for assessing the antecedents and consequences related to problem behavior, namely, ABC assessment (an abbreviation for antecedents-behaviors-consequences). Using this procedure, one records each episode of problem behavior along with the interpersonal context (social antecedents) and the reactions of others (social consequences) to the behavior. Aggregating data over a designated period of time allows the clinician the opportunity to identify patterns of contextual events (social antecedents) associated with behavior. Overall, the advantage of direct observation is that it avoids the ethical and logistic issues associated with experimental manipulation. However, because manipulation of context is not a feature of this type of assessment procedure, causality cannot be inferred from the results. That is, the results are purely correlational in nature. Another drawback is that if the frequency of problem behavior is low, then a lengthy period of direct observation may be required to determine the relationship between context and problem behavior. Unfortunately, many clinicians do not have the time or the resources needed for this type of contextual assessment.

Interviews and checklists have been developed as an alternative third method to the time-and labor-intensive assessment methods just described. In illustration, the Setting Event Checklist (Gardner, Cole, Davidson, & Karan, 1986) and the Functional Assessment Interview (O'Neill et al., 1990, 1997) were each developed as an attempt to provide an efficient, user-friendly method of assessing contextual variables. These methods of contextual assessment are faster than either experimentation or direct observation, and they require less expertise on the part of the clinician. However, because the Functional Assessment Interview is open ended, it is meant to provide a general framework for building interventions rather than a specific listing of contextual factors. In contrast, the Setting Event Checklist involves a specific listing of contextual factors, but the list was developed in reference to particular individuals and was not designed as a comprehensive summary of factors potentially relevant to all individuals.

In response to some of the issues just noted, we developed the Contextual Assessment Inventor. (CAI; McAtee, Carr, & Schulte, 2004) as a way of incorporating and extending the benefits associated with existing interviews and checklists. The CAI was designed to be an efficient, comprehensive, and comprehensible way to rapidly identify generic classes of contextual variables associated with problem behavior in individuals with developmental disabilities.

The development phase of the CAI followed the guidelines commonly used by researchers in survey development (DeVellis, 1991). Initially, an item pool was generated through a review of the literature on setting events and discriminative stimuli and a review of items contained in existing assessment instruments such as the Functional Assessment Interview (O'Neill et al., 1990) and the Setting Event Checklist (Gardner et al., 1986). Subcategories of items were identified and independently coded by two researchers into four generic classes: social/cultural contexts, task/activity contexts, physical contexts, and biological contexts. These four generic classes constituted the major subcategories that defined the CAI. The process just described generated a total of 93 contextual items. However, 13 of these items were intentionally open ended to give raters an opportunity to add information not captured by the other 80 items. For example, a specific item for the social/cultural subcategory was "ongoing difficulty communicating wants or needs." In contrast, an open-ended item for the social/cultural subcategory was "please list any other social/cultural factors that you believe may be setting events or discriminative stimuli for the person you work with." The exclusive focus of the present study was on validity responses to the 80 specific items rather than responses to the 13 open-ended items because 98.7% of the contexts nominated by the raters in the original study referred to 1 or more of the 80 specific items. It should be noted as well that each of the …