Single-case designs

Transcript

SoheeA school psychologist has one student on her mind. A third grader who leaves his seat over and over during reading time. She has a plan to help. But she also wants to know, honestly, whether her plan is what makes the difference.

RyanAnd there's no control group. There's one child.

SoheeThis is EPPP Prep, the podcast. I'm Sohee.

RyanAnd I'm Ryan. Today, single-case designs: how you can run a real experiment with one person, or a handful, and what the exam wants you to know about each design.

SoheeOur main source is an open-access review by Lobo and colleagues on single-case design, analysis and quality assessment. It's written for rehabilitation researchers, and it notes these designs have become increasingly popular in educational and psychological research.

RyanFirst, names. You'll see these called n of 1 studies, or single-subject studies. The review prefers the term single-case, because the older names suggest only one participant. In fact, the strongest single-case studies commonly include more than one participant.

SoheeAnd the first trap: a single-case design is not a case study.

RyanRight. The review explains that a typical case study or case report has no purposeful manipulation of an independent variable, and doesn't necessarily use repeated measures. Case reports are usually narrative. Single-case results are presented as numbers or graphs.

SoheeSo what makes a single-case design an experiment? The outcome, the dependent variable, is measured repeatedly over time, across different levels of the intervention, the independent variable.

RyanThose levels are called phases. One phase serves as the baseline, and so each participant serves as his or her own control.

SoheeAnd that's the elegant part. The review points out that because each person is their own comparison, it controls for things like age, gender, cognition and home environment. Those characteristics stay the same from phase to phase. The intervention is what changes.

RyanThe review also contrasts this with large randomized trials. Those trials report group means and standard deviations, which may not represent the true performance of any one participant. Single-case results can show whether the intervention changed things for the individual.

SoheeAnd practically, the review says single-case studies may need fewer resources than large trials, and can be done in settings and with populations where large numbers of participants aren't available.

RyanIt names the situations where they're particularly useful: when resources are limited, when the condition being studied is rare, and when you're testing a novel or expensive intervention.

SoheeLet's go design by design. The simplest is the AB design. A is a baseline phase, measured repeatedly. B is the intervention phase, also measured repeatedly.

RyanThe review recommends a stable baseline before switching, when that's possible, and a minimum of five data points in each phase.

SoheeBack to our student. She counts how often he leaves his seat each reading period for a stretch, then starts her plan and keeps counting.

RyanAnd if the counts drop, is that proof? The review says no. There's no replication in a basic AB design, so it's weak at establishing cause. The change could come from other factors, like maturation, experience, learning or practice.

SoheeWhich is why we strengthen it. The reversal design, also called a withdrawal design. A, then B, then back to A. Or better, A, B, A, B.

RyanThe logic is wonderful. If the behavior improves when the intervention starts, worsens when it's removed, and improves again when it returns, it's very hard to blame maturation or chance.

SoheeThe review says at least four phases, ABAB, allows a stronger determination of a causal relationship, because the relationship shows up at three different points of change: A to B, B back to A, and A to B again.

RyanAnd that ABAB pattern is also the minimum the review cites for meeting quality standards: three replications. Four or more are highly recommended.

SoheeNow the two conditions on reversal designs, because both appear on exams. First, it's recommended when the intervention doesn't have carry-over effects.

RyanSecond, ethics. Before using it, researchers must decide it's safe and ethical to withdraw the intervention, especially when the intervention is effective and necessary.

SoheeImagine an intervention that stops a child from hurting himself. You would not want to take it away to prove a point.

RyanSo what do you do when you can't, or shouldn't, go back to baseline? Enter the multiple baseline design.

SoheeThe review says that when an intervention permanently changes a person's ability, a return to baseline isn't feasible, and reversal designs aren't appropriate. Multiple baseline designs are useful in those situations.

RyanThe key feature is staggered introduction of the intervention across time. In the review's description, each participant is assigned to one of at least three conditions that differ in how long the baseline lasts.

SoheeSo participant one starts the intervention after a short baseline, participant two after a longer one, participant three after a longer one still.

RyanAnd here's why it works. The review says the staggered start separates intervention effects from maturation, experience, learning and practice. If each person improves right when their own intervention begins, and not before, the intervention is the most likely explanation.

SoheeThe review's example is a medication study for muscle stiffness in five adult men with spinal cord injury. Each was randomly assigned to between five and nine baseline data points on a placebo before switching to the medication, and both participants and assessors were blind to the condition.

RyanAnd the results suggested the medication might not be a universal treatment choice for everyone with that kind of stiffness. Single-case designs can show you that kind of person-by-person variation.

SoheeThe review adds that replication can also be shown across settings, tasks, or service providers, not only across people. So a multiple baseline can be staggered across settings or tasks, too.

RyanOkay, pause and think. A clinician teaches a client to read sight words. Once learned, the skill isn't going to disappear when instruction stops. Which design fits better, a reversal or a multiple baseline?

SoheeMultiple baseline. A reversal design needs behavior to return toward baseline when the intervention is withdrawn. A learned skill won't, so you stagger the start across participants, settings or sets of words instead.

RyanOne more design. The alternating treatment design. After a baseline, the person is exposed to different intervention conditions, alternated in close proximity, for equal intervals.

SoheeThe review gives two examples. One alternated an exercise condition and a control condition to look at disruptive behavior at school. The other compared commercially available and custom-made video prompts for multi-step cooking tasks in participants with autism.

RyanIn that study, participants independently completed more steps with the custom-made prompts.

SoheeThe weakness the review flags is carry-over: the effects of one intervention influencing the outcomes measured for another.

RyanNow, how are these analyzed? Traditionally by looking at the graph. The review describes visual analysis: within each phase you look at level, trend and stability. Between phases you look at how immediate the change was, how consistent the pattern is, and how much the data overlap.

SoheeAnd then quantitative methods can estimate the size of the effect. The review describes visual analysis first, then quantitative analysis of the magnitude if there's evidence of an effect.

RyanLast piece. Randomization. The review says it should be built in when possible, to improve internal validity. For example, randomizing the order of phases, or randomizing when the intervention starts.

SoheePause and think, second one. A researcher runs an AB design with one adult, and anxiety ratings drop after the intervention begins. What's the strongest criticism?

RyanNo replication. With one A phase and one B phase, the drop could come from other factors like maturation, practice, or something else changing at the same time. Withdrawing and reintroducing, or staggering across people, would make the case far stronger.

SoheeSo how does the exam test this? Usually a short description, and you name the design, or identify why one design fits better than another.

RyanThe tempting wrong answers: calling a single-case design a case study, picking a reversal design for a skill that can't be unlearned or a treatment that's unsafe to withdraw, and treating a basic AB design as strong causal evidence.

SoheeLet's recap. Single-case designs measure the outcome repeatedly across phases, and each person serves as their own control.

RyanAB is the simplest and the weakest, because there's no replication.

SoheeReversal designs, especially ABAB, show the effect turning on, off and on again, but need an intervention without carry-over that's safe and ethical to withdraw.

RyanMultiple baseline staggers the start across people, settings or tasks, so nothing needs to be withdrawn.

SoheeAlternating treatments compare interventions rapidly, with carry-over as the main risk. And visual analysis looks at level, trend, stability, immediacy, consistency and overlap.

RyanI love this topic because it's so close to everyday practice. Every time you track one client's progress and adjust, you're thinking like a single-case researcher.

SoheeSo you already have more intuition here than you might think. Try a few design questions, and sketch the graph in your head as you read the stem. Thanks for listening to EPPP Prep.

Loading the read-along timings…