Review article
Efficacy of behavioral interventions for reducing problem behavior in persons with autism: An updated quantitative synthesis of single-subject research

https://doi.org/10.1016/j.ridd.2014.06.017Get rights and content

Highlights

  • Behavioral interventions for reducing problem behavior were studied.

  • 213 studies and 358 persons with autism were included in the statistical analyses.

  • Behavioral interventions were effective in reducing problem behavior.

  • Interventions were equally effective regardless of type of problem behavior targeted.

  • Treatment and experimental characteristics were statistically significant moderators.

Abstract

Problem or challenging behaviors are highly prevalent among persons with autism and bring along major risks for the individual with autism and his/her family. In order to reduce the problem behavior, several behavioral interventions are used. We conducted a quantitative synthesis of single-subject studies to examine the efficacy of behavioral interventions for reducing problem behavior in persons with autism. Two hundred and thirteen studies representing 358 persons with autism met the inclusion criteria and were included in the statistical analyses. Overall, we found that behavioral interventions were on average effective in reducing problem behavior in individuals with autism, but some interventions were significantly more effective than others. The results further showed that the use of positive (nonaversive) behavioral interventions was increasing over time. The behavioral interventions were on average equally effective regardless of the type of problem behavior that was targeted. Interventions preceded by a functional analysis reduced problem behavior significantly more than interventions not preceded by a functional analysis. Finally, treatment and experimental characteristics, but not participant characteristics, were statistically significant moderators of the behavioral treatment effectiveness.

Introduction

Problem behaviors such as aggressive, stereotyped, and self-injurious behavior are highly prevalent among persons with autism (e.g., Matson and LoVullo, 2008, Murphy et al., 2009). The problem behaviors bring along major risks for the individual with autism and his/her family with regard to their physical, emotional, and social well-being, and can accordingly reduce their quality of life (e.g., Walsh, Mulder, & Tudor, 2013). In order to reduce problem behavior in persons with autism, several (cognitive-)behavioral interventions are used, such as differential reinforcement of other behavior (DRO), differential reinforcement of incompatible behavior (DRI), differential reinforcement of alternative behavior (DRA), antecedent control, antecedent exercise, noncontingent reinforcement, social stories, picture exchange communication system (PECS) interventions, and mindfulness-based interventions.

Many studies published in the domain of behavioral intervention research for reducing problem behavior among persons with autism are single-subject studies. In order to synthesize the results of these studies and to study which variables are moderating the effectiveness of the behavioral interventions, meta-level research is needed. Accordingly, Campbell (2003) conducted a quantitative synthesis of single-subject studies published between 1966 and 1998 on the efficacy of behavioral interventions for reducing problem behavior in persons with autism. In the meantime many more studies were published in this domain (cf. Matson & LoVullo, 2009), and an update of the meta-analysis of Campbell (2003) was warranted. The present article provides a double update of this meta-analysis: one from a methodological perspective and one from a temporal perspective.

First, we applied a methodological update. Campbell (2003) calculated three single-subject nonparametric statistics for estimating the effects of the behavioral treatments: the percentage of nonoverlapping data (PND; Scruggs, Mastropieri, & Casto, 1987), the percentage of zero data (PZD; Scotti, Evans, Meyer, & Walker, 1991), and the mean baseline reduction (MBLR; Kahng, Iwata, & Lewin, 2002). However, in the meantime new single-subject nonparametric statistics have been developed that avoid some of the drawbacks of the earlier developed statistics (e.g., see Heyvaert et al., in press, Parker and Brossart, 2003, and Parker, Vannest, & Davis, 2011, for overviews). Therefore we included the percentage of data points exceeding the median of baseline phase (PEM; Ma, 2006) and the percentage of all nonoverlapping data (PAND; Parker, Hagan-Burke, & Vannest, 2007) in our update. In contrast with some other recently developed nonparametric statistics, PEM and PAND have comparable ease of use to PND, PZD, and MBLR (cf. Parker et al., 2011): All five effect sizes can easily be calculated by practitioners.

PND is the most frequently used effect size index across single-subject syntheses in the field of disability research (cf. Maggin, O’Keeffe, & Johnson, 2011). In order to meet PND's main drawback (i.e., the deficient performance in the presence of data outliers in the baseline phase) Ma (2006) developed PEM as an alternative effect size for summarizing results of single-subject studies: Whereas PND takes into account the one most extreme value of the baseline phase, and refers to the percentage of data points in the treatment phase that exceeds this most extreme value, PEM takes into account the median value of the baseline phase. PEM leans very close to PND in its calculation and interpretation. PAND was also developed to meet the drawbacks of PND, but conceptually deviates a bit more from it. The main advantages of PAND over PND are: (1) PAND uses all data from the baseline and intervention phases, avoiding the criticism leveled at PND for overemphasis on one unreliable data point; and (2) PAND can be translated to Pearson's Phi and Phi2, and because Phi and Phi2 have known sampling distributions, p values are available, statistical power can be estimated, and confidence intervals can be included to indicate effect size reliability (Parker et al., 2007). Accordingly, for the present study we calculated PND, PZD, MBLR, PEM, and PAND for estimating the effects of behavioral interventions for reducing problem behavior in persons with autism. We compared these five nonparametric statistics with one another and examined to what extent they agreed in the analysis of the same data set. Answers to such questions are needed for scientist-practitioners to confidently use nonparametric statistics in the analysis of single-subject data (Parker & Brossart, 2003).

Second, for the temporal update we included single-subject studies published between 1999 and 2012 in our meta-analysis. Analogous to what Campbell (2003) did for the studies published between 1966 and 1998, we summarized single-subject studies published between 1999 and 2012: We studied the overall efficacy of behavioral interventions in reducing problem behavior in individuals with autism, examined whether some behavioral interventions were more effective than others, and investigated which variables, if any, moderated the overall efficacy of the behavioral interventions. Furthermore, we compared the two sets of studies and examined whether there were differences in the use of behavioral interventions and their effectiveness in reducing problem behavior in individuals with autism.

Accordingly, the research questions addressed in the current review were: (1) What is the overall efficacy of behavioral interventions in reducing problem behavior in individuals with autism; (2) Are some behavioral interventions more effective than others in reducing certain types of problem behavior in individuals with autism; (3) Do participant, treatment, or experimental variables influence the overall efficacy of behavioral interventions; (4) Are there any differences between the three older effect sizes (i.e., PND, PZD, and MBLR; Campbell, 2003) and the two more recently developed effect sizes (i.e., PEM and PAND) regarding treatment efficacy and moderating variables; and (5) Are there any differences between the single-subject studies published between 1966 and 1998 (Campbell, 2003) and the studies published between 1999 and 2012 regarding the use of behavioral interventions and their effectiveness in reducing problem behavior in individuals with autism?

Section snippets

Inclusion and exclusion criteria

We aimed at reviewing single-subject studies on behavioral interventions for reducing problem behavior in people with autism. The inclusion criteria were defined in the same way as Campbell (2003) did. First, the review included studies about participants diagnosed with autistic disorder. An article was included if at least one participant was diagnosed with autism. When articles included multiple individuals, only those participants diagnosed with autism were included in the review.

Results

Two hundred and thirteen studies representing 358 persons with autism met the inclusion criteria and were included in the statistical analyses. Detailed information about the participants, treatments, and experimental studies is presented in Table 1, Table 2, Table 3 respectively.

To answer the first research question we examined the overall efficacy of the behavioral interventions. Across all participants the averages were 74.9%, 44.7%, 70.2%, 91.4% and 91.9% for PND, PZD, MBLR, PEM, and PAND

Discussion

The present study aimed to answer five questions: (1) What is the overall efficacy of behavioral interventions in reducing problem behavior in individuals with autism; (2) Are some behavioral interventions more effective than others in reducing certain types of problem behavior in individuals with autism; (3) Do participant, treatment, or experimental variables influence the overall efficacy of behavioral interventions; (4) Are there any differences in the conclusions for the five calculated

References (28)

  • M. Heyvaert et al.

    A meta-analysis of intervention effects on challenging behaviour among persons with intellectual disabilities

    Journal of Intellectual Disability Research

    (2010)
  • M. Heyvaert et al.

    Systematic review of restraint interventions for challenging behaviour among persons with intellectual disabilities: Focus on effectiveness in single-case experiments

    Journal of Applied Research in Intellectual Disabilities.

    (2014)
  • M. Heyvaert et al.

    Randomization and data-analysis items in quality standards for single-case experimental studies

    Journal of Special Education

    (2014)
  • S. Kahng et al.

    The impact of functional analysis on the treatment of self-injurious behavior

  • Cited by (82)

    View all citing articles on Scopus
    View full text