Journal of Pediatric Psychology Advance Access originally published online on November 13, 2007
Journal of Pediatric Psychology 2008 33(1):16-21; doi:10.1093/jpepsy/jsm101
| ||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
Editorial: Journal of Pediatric Psychology Statement of Purpose: Section on Single-Subject Studies
1University of Kansas Medical Center and 2Cincinnati Children's Hospital Medical Center
All correspondence concerning this article should be addressed to Michael A. Rapoff, PhD, Department of Pediatrics, University of Kansas Medical Center, 3901 Rainbow Blvd., Kansas City, KS 66160-7330, USA. E-mail: mrapoff{at}kumc.edu.
 |
Abstract |
|---|
 Top Abstract Statement of PurposeBackgroundReversal (withdrawal) DesignMultiple Baseline DesignChanging Criterion DesignCautions and CaveatsLimitationsConclusionsAcknowledgmentsReferences |
|---|
Objective The purpose of this editorial is to introduce a new special section on single-subject designs that is designed to encourage the submission of papers to the Journal of Pediatric Psychology that utilizes single-subject methodology to investigate the efficacy of treatments for improving the health, mental health, and quality of life for children and adolescents. Methods and Results We review the case for using single subject designs in pediatric psychology research and provide examples from the literature of the three most commonly used designs: the reversal or withdrawal, multiple baseline, and changing criterion. We also describe limitations of single subject designs. Conclusions We hope this new special section will re-ignite enthusiasm among clinicians and researchers for using single subject designs to advance our field, provide pilot data for larger trials, and/or further document the efficacy of psychological and medical interventions with various populations.
|
Statement of Purpose |
|---|
|
TopAbstractStatement of PurposeBackgroundReversal (withdrawal) DesignMultiple Baseline DesignChanging Criterion DesignCautions and CaveatsLimitationsConclusionsAcknowledgmentsReferences |
|---|
Although randomized, between-groups, controlled clinical trials (RCTs) are considered the "gold standard" for experimentally evaluating the efficacy of treatments, there is a long tradition in medicine and psychology of investigating the effects of interventions at the individual level using single-subject designs (Allen, Friman, & Sanger, 1992
; Barlow & Hersen, 1984; Janosky, 2005; Kazdin, 1982). Single-subject designs offer a number of advantages over traditional group designs: (a) they provide flexibility in the choice of independent variables and allowance for changes in these over the course of a study (if something is not working, an intervention can be modified and introduced as a new condition); (b) they accommodate for small sample sizes (appropriate for studying rare conditions or smaller sample sizes available at any one site); (c) they are appropriate when there are ethical objections to withholding treatment; (d) they are better at exposing individual variability in outcome measures; (e) they produce results that are more easily understood by clinicians (who work at the level of individual patients); (f) they have greater potential for attracting busy clinicians to do clinical research; and (g) they are recognized as legitimate designs that can help to establish empirically validated treatments and evidence-based practices (Barlow & Hersen, 1984; Spirito, 1999).
The purpose of this special section on single-subject designs is to encourage the submission of papers to the Journal of Pediatric Psychology (JPP) that utilize single-subject methodology to investigate the efficacy of treatments for improving the health, mental health, and quality of life for children and adolescents. We recognize that single-subject designs have been published in JPP, starting with an article published in the first volume of the journal, which used a multiple baseline design across subjects to demonstrate the effects of overcorrection on self-injurious behavior (Barnard, Christophersen, & Wolf, 1976) and most recently, an article describing the combined use of a multiple baseline across settings (meals) and changing criterion design to evaluate the effects of a behavioral and nutrition intervention for children with cystic fibrosis (Power et al., 2006). We want to continue to promote this use of single-subject design methodology in pediatric psychology research.
|
Background |
|---|
|
TopAbstractStatement of PurposeBackgroundReversal (withdrawal) DesignMultiple Baseline DesignChanging Criterion DesignCautions and CaveatsLimitationsConclusionsAcknowledgmentsReferences |
|---|
Beginning in the 1960s and driven primarily by the explosion of research in applied behavior analysis which ensued, single-subject designs became a viable option for researchers interested in intensively studying single or small numbers of individuals in their schools, homes, workplace, or in medical settings. Each of the single-subject designs reviewed subsequently share five common characteristics (Allen et al., 1992
): (a) they involve repeated measurements over time; (b) they require a well-established baseline (analogous to a no-treatment control condition of group designs); (c) they require stability in the dependent variable (defined as a consistent range of variability, a lack of trends, or a trend in the opposite direction of the desired effect) within each phase and particularly when a phase change is made; (d) they require replication across individuals or dependent variables to address threats to internal and external validity; and (e) ideally, they require changing only one variable at a time for experimental control (the caveat being that treatment "packages" containing several variables can be evaluated and then dismantled to identify key variables using single-subject designs).
Although there are many single-subject design options and variants (Allen et al., 1992; Barlow & Hersen, 1984; Kazdin, 1982), we will describe the three most common options: Reversal or withdrawal design, the multiple baseline design, and the changing criterion design, illustrated with figures from published studies (and a bit self-servingly from the first author's research on medical adherence). Other design options, such as the multi-element or alternating treatment design (Hains & Baer, 1989) and the multiple probe variation of the multiple baseline design (Horner & Baer, 1978) will not be presented here, but authors are encouraged to consult references for when these designs might be useful (e.g., using the multi-element design to evaluate the relative effectiveness of two or more treatments or conditions).
|
Reversal (withdrawal) Design |
|---|
|
TopAbstractStatement of PurposeBackgroundReversal (withdrawal) DesignMultiple Baseline DesignChanging Criterion DesignCautions and CaveatsLimitationsConclusionsAcknowledgmentsReferences |
|---|
The basic variant of this design is typically referred to as "ABA," where A = baseline (no treatment or usual care) and B = treatment to be evaluated. The logic of this design is that it controls for the effects of treatment and subsequent removal, thus demonstrating that changes in the dependent variable occurs only when the treatment is introduced and return to baseline levels (or nearly so) when treatment is withdrawn. Other variants of this design include "ABAB" (which ends on treatment and provides two treatment evaluations relative to baseline) and "ABCB" (where "C" is another treatment or placebo condition).
Figure 1 (Rapoff, Purviance, & Lindsley, 1988) illustrates the use of a reversal design to demonstrate the effects of two interventions (relative to baseline and withdrawal conditions), designed to improve adherence to an antiinflammatory medication for a 14-year-old male with polyarticular juvenile rheumatoid arthritis (JRA). The "simplified regimen" condition was not planned. The treating rheumatologist decided to reduce the medication dosing schedule from four to three times a day and we took advantage of this by including this change as a treatment condition, as simplifying regimens can improve adherence (Rapoff, 1999). The token system condition was planned and the patient earned points for taking medications daily and could exchange them for special privileges. Thus, this was an ABCAC variant of the reversal design (with A = baseline and intervention withdrawal; B = simplifying the regimen; and C = token system) and included maintenance and follow-up phases to determine if improvements in adherence were maintained after the token system was discontinued.
|
As can be seen in Fig. 1, this study clearly supports the effects of the token system in improving adherence to medications for JRA for this patient and illustrates the flexibility of single-subject designs, as we took advantage of the unplanned simplification of the patient's regimens to assess its effect on adherence.
|
Multiple Baseline Design |
|---|
|
TopAbstractStatement of PurposeBackgroundReversal (withdrawal) DesignMultiple Baseline DesignChanging Criterion DesignCautions and CaveatsLimitationsConclusionsAcknowledgmentsReferences |
|---|
This design option requires multiple (usually three or more) baselines across behaviors, subjects, or settings followed by the staggered introduction across time of a treatment condition. The logic of this design is that the effects of treatment are shown when change occurs after introduction of the treatment, while concurrent (untreated) behaviors (or person or settings) remain unchanged or constant. When treatment is introduced successively for each behavior (or each person or setting), the effects of treatments are replicated. Multiple baselines are particularly useful when treatments cannot or should not be withdrawn. However, they are considered weaker than reversal designs because the controlling effects of treatment are not confirmed by introduction and removal of the treatment. Also, in clinical settings, it might not be possible to have concurrent baselines that control for historical events. For example, Isaacs, Embry, and Baer (1982
) trained five families to modify their children's behavior in the home. Participants were not available to begin the study at the same time, so nonconcurrent baselines across families were employed, while retaining the staggered introduction of treatment by having differing lengths of baselines for each family. Although the authors noted that the nonconcurrent baseline approach was weaker than the traditional concurrent baseline one, this has remained a viable option, particularly in clinical settings where participants can be recruited as they become available or seek treatment.
Figure 2 (Pieper, Rapoff, Purviance, & Lindsley, 1989) illustrates the use of a multiple baseline across subjects design to evaluate the effects of a combined educational and behavioral intervention on adherence to prednisone in the treatment of pediatric rheumatic diseases. Patients 1 and 2 (African-American females, age 17 and 18 years) were diagnosed with systemic lupus erythematosus (SLE) and Patient 3 (Hispanic female, age 11 years) was diagnosed with dermatomyositis (DM). For this study, "acceptable" adherence levels (shaded area on Fig. 2) were defined as patients taking between 80% and 120% of their medication as determined by pill counts. This upper limit of 120% was established because patients were over-medicating as well as under-medicating and the lower limit of 80% is the minimum level of adherence necessary to effectively treat according to the pediatric rheumatologist. As shown in Fig. 2, relatively simple educational and behavioral strategies were effective in reducing over- as well as under-medicating with an ethnically diverse sample treated for two different pediatric rheumatic diseases.
|
|
Changing Criterion Design |
|---|
|
TopAbstractStatement of PurposeBackgroundReversal (withdrawal) DesignMultiple Baseline DesignChanging Criterion DesignCautions and CaveatsLimitationsConclusionsAcknowledgmentsReferences |
|---|
This design incorporates repeated B (treatment) phases over time following an initial A (baseline) phase. After the first baseline condition, treatment is carried out until a preset criterion is reached (e.g., exercising three times per week) and stability is achieved. Then, a more stringent criterion is set (e.g., exercising five times per week) and treatment is applied until this new level is met and stabilized. The changing criterion design is most useful when the investigator expects a close correspondence between the criterion and behavior over the course of an intervention phase. Also, for some dependent variables, setting a goal or criterion for change makes sense from a clinical perspective (e.g., reducing caloric intake to levels necessary for weight loss). The major problem that can occur with this design is that the criterion for change is not met or there are unstable effects. This problem can be addressed by returning briefly to a less-stringent criterion and then proceeding in a step-wise fashion to more stringent levels (Barlow & Hersen, 1984
).
Figure 3 (Powers et al., 2006) illustrates the use of a changing criterion design to evaluate the effects of a behavioral and nutrition intervention for increasing caloric intake in children with cystic fibrosis (they also used a multiple baseline across settings or meals). The data in Fig. 3 are averaged daily total caloric intake across the nine participants. After pretreatment, a criterion was set for snacks only and then in successive phases the criterion was increased and other meals were added across the phases. As can be seen in Fig. 3, changes in caloric intake reached or exceeded criterion levels for most days, with a drop off during the 3- and 12-month follow-up phases (although the levels at follow-up were still in the upper end of the range for recommended dietary allowance for children with cystic fibrosis). Data from this study are part of an ongoing RCT with participants assigned to the behavioral and nutrition intervention or a usual care control condition. The nine participants include four from the behavioral and nutrition group and five of six from the control group who elected to receive the intervention following completion of the clinical trial. Thus, the Powers et al. (2006) study is instructive in how to incorporate single-subject methodology within a randomized controlled clinical trial.
|
|
Cautions and Caveats |
|---|
|
TopAbstractStatement of PurposeBackgroundReversal (withdrawal) DesignMultiple Baseline DesignChanging Criterion DesignCautions and CaveatsLimitationsConclusionsAcknowledgmentsReferences |
|---|
Single-subject methodology is not consistently taught in most graduate programs. We would therefore encourage researchers to review the classic texts on these designs (Barlow & Hersen, 1984
; Kazdin, 1982). In reviewing for JPP, single-subject papers are typically rejected because they failed to show experimental control. That is, the label of a single-subject design was used to describe the study, but the principles of single-subject methodology were not followed. A common pitfall is changing conditions based upon time and not the subject's behavior or the dependent variable. It is important to keep a subject in a given condition, whether it is baseline or intervention, until the dependent variable is stable. Thus, conditions cannot be changed at a predetermined time. For example, if planning a multiple baseline design across three children, the length of the baseline for the second and third child cannot be set ahead of time. Intervention cannot occur for the second child until the effect of the intervention on the dependent variable of the first child is demonstrated and stable.
Graphing single-subject designs also require attention to phase changes and how to represent these. For example, data points across phases should not be connected by a line. In 1996, the Journal of Applied Behavior Analysis (JABA, 1996, 29 (1), 115) published pictorial guidelines for graphing single-subject data that included instructions on labeling phase changes and creating legends (JABA, 1996, p. 115). Recommendations include use of vertical dashed lines to separate conditions, placing the labels for phases above the data and centered for the phase and ensuring that the labels are descriptive of the phase, placing the subject descriptor (alias) in small type, in a box in the right hand corner of the graph, and where more than one set of data are presented in a figure, placing the labeling for each line of data directly next to it with an arrow pointed to the data being referred to. Authors wishing to submit single-subject designs are encouraged to follow these guidelines in preparing graphs for submission to JPP. In addition, tips for using Microsoft ExcelTM to create graphs have also been published and are a valuable resource for preparing graphs that meet the requirements (Carr & Burkholder, 1998).
Single-subject designs are often confused with case studies. Case studies are often presented in an AB format, where A = baseline and B = intervention, and therefore experimental control can not be established as with single-subject designs. However, well designed case studies are encouraged to be submitted to JPP to describe new treatments and new clinical phenomena, to develop hypotheses, to illustrate methodological issues, and to address difficult diagnostic issues (see Drotar, La Greca, Lemanek, & Kazak, 1995 for guidelines).
|
Limitations |
|---|
|
TopAbstractStatement of PurposeBackgroundReversal (withdrawal) DesignMultiple Baseline DesignChanging Criterion DesignCautions and CaveatsLimitationsConclusionsAcknowledgmentsReferences |
|---|
There are limitations to single-subject methodology of which researchers need to be aware and acknowledge in the discussion of the results. External validity is the most salient limitation of single-subject designs. That is, how generalizable are the results from the few subjects studied to the larger population of interest? Thus, it needs to be acknowledged that the results are most applicable to children who are most like the research participants. Systematic replication of the intervention effects across more subjects and subjects with greater diversity allow broader statements of generalizability to be made. Certainly, an ABAB design showing a similar positive effect has greater generalizability if done with four or five children rather than with only one child. Internal validity is also important to address. That is, one must conduct the study in a manner that allows inference that the change in the dependent variable was due to the manipulation of the independent variable and not to extraneous events. Thus, choosing the best design for the dependent variable and insuring that the phases (baseline and intervention) are adequate to demonstrate control is paramount. It is also important to operationalize both the intervention and the dependent variable so that the study could be replicated by others. The dependent measure should be measured serially over time and be reliable.
|
Conclusions |
|---|
|
TopAbstractStatement of PurposeBackgroundReversal (withdrawal) DesignMultiple Baseline DesignChanging Criterion DesignCautions and CaveatsLimitationsConclusionsAcknowledgmentsReferences |
|---|
Pediatric psychology has a rich history of using case studies and single-subject methodology to demonstrate the effectiveness of our interventions on unique, difficult to treat, and rare conditions. For example, Logan Wright's pioneering work on tracheotomy weaning (Wright, Nunnery, Eichel, & Scott, 1968
) using case studies set a precedent for utilizing "small n" studies to demonstrate the contribution of pediatric psychologists in medical settings (Willis, 2000). Single-subject methodology was instrumental in demonstrating the value of distraction for coping with procedural pain (Jay, Elliott, Ozolins, Olson, & Pruitt, 1985), when these interventions were being developed and the data generated by these studies led to several clinical trials by other investigators in the field. There continues to be a need to develop and demonstrate the contribution psychology makes in health care settings with various disorders, especially rare or rarely studied disorders, and single-subject designs continue to represent a methodologically sound first step in this process. By conceptualizing each patient as their own single-subject design and collecting outcome measures in a systematic manner, single-subject methodology is an excellent way to meet the clinical and research expectations that many medical settings have of their clinical faculty. There are likely many innovative treatments being implemented by pediatric psychologists that would advance our field, provide important pilot data for larger trials, and/or further document the importance of psychological and medical interventions with various populations. We would like to encourage our colleagues to incorporate single-subject methodology into their clinical work as well as in their design of pilot interventions and to submit this work to JPP.
|
Acknowledgments |
|---|
|
TopAbstractStatement of PurposeBackgroundReversal (withdrawal) DesignMultiple Baseline DesignChanging Criterion DesignCautions and CaveatsLimitationsConclusionsAcknowledgmentsReferences |
|---|
We thank Dr John Belmont, Professor Emeritus of Pediatrics at the University of Kansas Medical Center, for redrawing Figures 1–3
. Conflict of interest: None declared.
Received September 26, 2007; revision received September 26, 2007; accepted September 26, 2007
|
References |
|---|
|
TopAbstractStatement of PurposeBackgroundReversal (withdrawal) DesignMultiple Baseline DesignChanging Criterion DesignCautions and CaveatsLimitationsConclusionsAcknowledgmentsReferences |
|---|
Allen KD, Friman PC, Sanger WG. Small n research designs in reproductive toxicology. Reproductive Toxicology (1992) 6:115–121.[CrossRef][Web of Science][Medline]
Barlow DH, Hersen M. Single case experimental designs (1984) 2nd. New York: Pergamon Press.
Barnard JD, Christophersen ER, Wolf MM. Parent-mediated treatment of children's self-injurious behavior using overcorrection. Journal of Pediatric Psychology (1976) 1:56–61.
Carr JE, Burkholder EO. Creating single-subject design graphs with Microsoft excelTM. Journal of Applied Behavior Analysis (1998) 31:245–251.[CrossRef][Web of Science]
Drotar D, La Greca AM, Lemanek K, Kazak A. Case reports in pediatric psychology: Uses and guidelines for authors and reviewers. Journal of Pediatric Psychology (1995) 20:549–565.
Hains A, Baer DM. Interaction effects in multielement designs: Inevitable, desirable, and ignorable. Journal of Applied Behavior Analysis (1989) 22:57–69.[CrossRef][Web of Science][Medline]
Horner RD, Baer DM. Multiple-probe technique: A variation of the multiple baseline. Journal of Applied Behavior Analysis (1978) 11:189–196.[CrossRef][Web of Science][Medline]
Isaacs CD, Embry LH, Baer DM. Training family therapists: An experimental analysis. Journal of Applied Behavior Analysis (1982) 15:505–520.[CrossRef][Web of Science][Medline]
Janosky JE. Use of the single subject design for practice based primary care research. Postgraduate Medical Journal (2005) 81:549–551.
Jay SM, Elliott CH, Ozolins M, Olson RA, Pruitt SD. Behavioral management of children's distress during painful medical procedures. Behavior Research and Therapy (1985) 23:513–520.[CrossRef][Web of Science][Medline]
Kazdin AE. Single-case research designs: Methods for clinical and applied setting (1982) New York: Oxford University Press.
Manuscript preparation checklist. Journal of Applied Behavior Analysis (1996) 29:115.
Pieper KB, Rapoff MA, Purviance MR, Lindsley CB. Improving compliance with prednisone therapy in pediatric patients with rheumatic disease. Arthritis Care & Research (1989) 2:132–135.[CrossRef]
Powers SW, Piazza-Waggoner C, Jones JS, Ferguson KS, Daines C, Acton JD. Examining clinical trial results with single-subject analysis: An example involving behavioral and nutrition treatment for young children with cystic fibrosis. Journal of Pediatric Psychology (2006) 31:574–581.
Rapoff MA. Adherence to pediatric medical regimens (1999) New York: Kluwer/Plenum.
Rapoff MA, Purviance MR, Lindsley CB. Improving medication compliance for juvenile rheumatoid arthritis and its effect on clinical outcome: A single-subject analysis. Arthritis Care & Research (1988) 1:12–16.[CrossRef]
Spirito A. Introduction (to special series of papers on empirically supported treatments in pediatric psychology). Journal of Pediatric Psychology (1999) 24:87–90.[Web of Science]
Willis DJ. In memoriam: Logan Wright, Jr. Ph.D. Journal of Pediatric Psychology (2000) 25:359–361.
Wright L, Nunnery A, Eichel B, Scott R. Application of conditioning principles to problems of tracheotomy addiction in children. Journal of Consulting and Clinical Psychology (1968) 32:603–606.[CrossRef][Web of Science][Medline]
CiteULike 
Connotea 
Del.icio.us What's this?
This article has been cited by other articles:
|
|
 |
|
  M. A Rapoff Editorial: Assessing and Enhancing Clinical Significance/Social Validity of Intervention Research in Pediatric Psychology J. Pediatr. Psychol., November 13, 2009; (2009) jsp102v1. [Full Text] [PDF]
|
|
|
|
|
|
D. Drotar Editorial: Case Studies and Series: A Call for Action and Invitation for Submissions J. Pediatr. Psychol., September 1, 2009; 34(8): 795 - 802. [Full Text] [PDF]
|
|
|
|
|
|
M. M Graves, M. C. Roberts, M. Rapoff, and A. Boyer The Efficacy of Adherence Interventions for Chronically Ill Children: A Meta-Analytic Review J. Pediatr. Psychol., August 26, 2009; (2009) jsp072v1. [Abstract] [Full Text] [PDF]
|
|
|
|
|
|
R. S. Bernard, L. L. Cohen, and K. Moffett A Token Economy for Exercise Adherence in Pediatric Cystic Fibrosis: A Single-Subject Analysis J. Pediatr. Psychol., May 1, 2009; 34(4): 354 - 365. [Abstract] [Full Text] [PDF]
|
|
| ||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||