Examining Predictors of Different ABA Treatments: A Systematic Review

Abstract

In the recent literature, there is a broad consensus on the effectiveness of Applied Behavior Analysis interventions for autism spectrum disorder (ASD). Despite their proven efficacy, research in clinical settings shows that these treatments are not equally effective for all children and the issue of which intervention should be chosen for an individual remains a common dilemma. The current work systematically reviewed studies on predictors and moderators of response to different types of evidence-based treatment for children with ASD. Specifically, our goal was to critically review the relationships between pre-treatment child characteristics and specific treatment outcomes, covering different aspects of functioning (i.e., social, communicative, adaptive, cognitive, motor, global functioning, play, and symptom severity). Our results questioned the binomial "better functioning-better outcome", emphasizing the complex interplay between pre-treatment child characteristics and treatment outcomes. However, some pre-treatment variables seem to act as prerequisites for a specific treatment, and the issue of "what works for whom and why" remains challenging. Future research should focus on the definition of evidence-based decision-making models that capture those individual factors through which a specific intervention will exert its effects.

Keywords: applied behavior analysis; autism; predictor; treatment outcome.

Figure 1

Adapted from the PRISMA flow diagram [24], showing the steps of the selection process undergone by the studies in this review.

Figure 2

Panel (a) shows the effects between predictors (in column) and dependent variables (in line). Full rectangles (black for EIBI studies, grey for NDBI studies) indicate that a significant predictive effect has been found for that dependent variable. Empty rectangles (with black outlines for EIBI studies and grey outlines for NDBI studies) indicate that the effect was not significant for that dependent variable. Panel (b) shows responders’ profile for each treatment approach, EIBI on the top and NDBI at the bottom. Solid lines: Hedval et al. [35]; Rhombus texture: Remington et al. [44]; Dotted texture: Fossum et al. [31]; Horizontal lines: Laister et al. [36]; Vertical lines: Sherer and Schreibman [37]; Diagonal lines: Sinai-Gavrilov et al. [27].

Figure 2

Panel (a) shows the effects between predictors (in column) and dependent variables (in line). Full rectangles (black for EIBI studies, grey for NDBI studies) indicate that a significant predictive effect has been found for that dependent variable. Empty rectangles (with black outlines for EIBI studies and grey outlines for NDBI studies) indicate that the effect was not significant for that dependent variable. Panel (b) shows responders’ profile for each treatment approach, EIBI on the top and NDBI at the bottom. Solid lines: Hedval et al. [35]; Rhombus texture: Remington et al. [44]; Dotted texture: Fossum et al. [31]; Horizontal lines: Laister et al. [36]; Vertical lines: Sherer and Schreibman [37]; Diagonal lines: Sinai-Gavrilov et al. [27].