Bio
Transcript
Bruce never calls single system design revolutionary, but it could be. Consider this: a group home for individuals with schizophrenia, a staff member desperate to improve residents’ physical activity, and a single stationary bike. With just a little data, the first two letters of the alphabet, and a bit of creativity, Bruce helped transform that scenario into a success story that not only improved lives but also debunked some myths about serious mental illness and exercise. Single system design isn’t flashy, but as Bruce will show us, it’s the kind of simple innovation that can help social workers evaluate practice in a way that’s rigorous, practical, and client-centered.
We start our conversation diving into why social workers should know about single system design, its core components, and the kinds of measures and assessments that work best and where to find them. We discuss real-world examples, the nuances of designs like AB, ABC, and beyond, and how to use single subject designs for complicated, multilayered problems. I wanted to know if they work across levels of practice—micro, mezzo, and macro—and whether they’re appropriate for evaluating evidence-based practice. When I told Bruce I had concerns about the generalizability of SSDs, he reminded me that the requirements for generalizability are almost never met, even in rigorous large scale studies. We ended with a discussion about publishing single system designs.
In Part 2, episode 144, Bruce and I talk about why social workers at the BSW, MSW and doctoral level should understand, and in the right situations, use experimental research design. He points out that social workers started publishing experimental design in the 1940s, around the same time as psychologists and sociologists. He distinguishes quasi-experimental from experimental design, and why the "natural experiment" of COVID doesn't really count as experimental design. In 2023, Dr. Thyer published Experimental Research Designs in Social Work: Theory and Applications through Columbia University Press.
And now, without further ado, on to episode 143 of the Social Work Podcast: The Power of One: Single System Design with Bruce Thyer, PhD, LCSW.
[00:04:14] Jonathan Singer: Bruce, thank you so much for being here on the Social Work Podcast.
[00:04:18] Bruce Thyer: Thank you for having me.
[00:04:20] Jonathan Singer: So, why should social workers know about single system design?
[00:04:28] Bruce Thyer: Well, I would think that every social worker would like to have some type of empirical evidence on the outcomes of their own practice. And I note that the NASW Code of Ethics says that we're ethically obligated to evaluate the outcomes of our own practice. And that applies to people at the BSW, MSW, and doctoral levels. Also, the Council on Social Work Education says that we need to teach students how to evaluate methods, how to use methods of evaluation for their own practice in the BSW and the MSW program.
[00:05:56] Jonathan Singer: So, it's both an ethical obligation and also in terms of our accreditation and sort of the educational standards, it's explicitly mentioned for MSW and BSW.
[00:05:07] Brucey Thyer: Very much so. And in fact, evaluation of practice is the only research skill that the Council on Social Work Education requires that we teach students to do. All the other research methods we teach them can be taught for the students to be able to consume them, to be able to read them and understand them.
[00:05:26] Jonathan Singer: So, it's really interesting to think about single system design as something that can sort of fulfill that ethical obligation to evaluate practice because a lot of what we talk to students about when they're going through their MSW or their BSW are these kind of off-the-shelf interventions like cognitive behavioral therapy or manualized treatments and things like that. Is single system design something that you could use with one of these off-the-shelf manualized interventions or is it something different?
[00:05:56] Bruce Thyer: I think that single system designs can be used for any type of intervention that's actually expected to change the lives of the people that we work with.
[00:06:05] Jonathan Singer: So, what are the components of a single system design?
[00:06:04] Bruce Thyer: Well, there's only four. The first is that we have to develop or locate some type of reliable and valid measure of client functioning. It can be a problem like a score in the Beck Depression Inventory. It can be a behavior like drinking too much. It can be a physiological variable like high blood pressure or perhaps being overweight. The second prerequisite is to ask the client to complete this measure—and you can have more than one—repeatedly over time. Ideally, you do that before you begin intervention or treatment, and then you continue that evaluation in the same way during intervention. You graph the data, and then you share the data with the client and make any inferences that you can visually detect. You don't need statistics to evaluate single subject designs. All you need is a pair of eyeballs and the ability to draw a conclusion: yes, they got better; it's obvious they didn't get better; or it's obvious that they got worse.
[00:07:13] Jonathan Singer: So, the measurement—it seems like a really important piece to this, right? Knowing what it is that you're measuring and being able to see change on those scores. Are there any types of measures or assessments that you think work better or that are sort of more appropriate for single system design?
[00:07:34] Bruce Thyer: I think that question needs to be tailored to the needs of the individual client. There's really only three things we can possibly measure when we're assessing client functioning. One is the client's self-report of their feelings or thoughts. A second is that we can measure their actual behavior. And the third is that we could conceivably measure some type of physiological aspect of their functioning, such as a drug urine test or something like that.
[00:08:04] Jonathan Singer: So, you have this ethical obligation. You want to do the best you can for your clients. Where, first of all, do you find measures to use? How does one find that?
[00:08:17] Bruce Thyer: Well, as you know, there's a great book called The Handbook of Measures for Social Work Research and Practice by David Springer, Kevin Corcoran, and Joel Fischer [note: The book title is slightly different and there are two volumes: Measures for Clinical Practice and Research: A Sourcebook: Volume 1: Couples, Families, and Children and Volume 2: Adults.] It's been around through multiple editions and contains literally hundreds of examples of high-quality, what we call rapid assessment instruments, that measure virtually everything under the sun—from individual behavior to families to couples to marriages. That's a good source that could be on the shelf, and I think every agency should get that book. Another way is we can use a literature search device like PsychInfo or Web of Science and type in a phrase like “depression,” and then qualify it with things like “rating” or “scale” or “measurement.” With a little bit of searching, you can find previously published measures that are reliable and valid for just about every kind of outcome we might think of in social work.
[00:09:12] Jonathan Singer: So, what's an example of a single system design. Can you describe something that illustrates the kind of problem that you might use single system design with or a situation and how it works?
[00:09:27] Bruce Thyer: Well, let's take somebody, oh, like myself. I was interested in losing weight. So, I measured my weight for a number of weeks on a weekly basis, or even more frequently than that. Then I began a formal weight reduction program and graphed the results. I was able to see that with the weight program I was on, I was actually losing weight over a number of months. Over time, I dropped about 30 pounds. That would be a type of design we would call an AB design. “A” refers to a baseline—pre-treatment measures of client functioning or behavior or some other aspect of their well-being. The “B” phase refers to the intervention phase. Hence, baseline is “A,” and “B” is the intervention phase.
[00:10:11] Jonathan Singer: And so, I love that you use the example of yourself and measuring your weight and then doing the intervention. For the AB design, are there different designs? Like, is there an ABC? Is there an ABCD? Like, can you keep going down the alphabet, or would it be repeating the AB? How does that work?
[00:10:31] Bruce Thyer: There's a variety of single system, or single subject, research designs that can be done. The AB is one of the simplest, but not quite the simplest. Then you can remove the intervention and graph this in what would be called an ABA design. This would be where you remove the treatment after it's been implemented for a while to see if the client regresses or goes back to their initial baseline functioning. This can be good in terms of making causal inferences about the effects of treatment. It's not particularly good clinically, except in a situation where the client continues to improve even though you've discontinued treatment. That suggests that your intervention, if it was causally responsible for the improvement, is continuing to have an effect over a longer period of time.
[00:11:14] Jonathan Singer: Which seems like it would be an ideal situation because you had an intervention and then you stopped, and the person continued to get better on some outcome, which is what we want for our clients because we’re not always going to be around to intervene.
[00:11:14] Bruce Thyer: Sure. And if you did, as an example, the ABA design and the client got worse or deteriorated, you could then reinstitute the initial treatment in the B phase and see if their functioning improved. So that would be what would be called an ABAB design. The idea is that if you see a systematic change from the first baseline to the first treatment phase, and then a reversal from the first treatment phase to the second baseline, and then a reinstatement of improvement from the second treatment phase to the second baseline, that’s very strong grounds for making a causal inference that the client got better because of the intervention you did. There are very few alternative explanations that could explain such a pattern apart from inferring that it was the treatment.
[00:12:18] Jonathan Singer: That’s really impressive. And as you’re saying that, I’m visualizing a chart, and I see the A, and then I see a line with the B, right? So, it sort of distinguishes the baseline from the intervention, and then you’ve got the intervention phase. Now, for the single system design, the ABAB—you know, with social work, I think one of the things that I often think about is the problems that people come to us with tend to be complicated, right? There are multiple layers to what goes on. For a single system design, is there a way to set it up so that you can look at multiple interventions at the same time, or is this something where you would intervene with one thing and then you would stop and go to something else? How does that work with a really complicated problem?
[00:13:04] Bruce Thyer: That’s a great question. There’s nothing to prevent you from using multiple interventions at the same time in the context of an AB design. It does make it difficult to determine what element of the intervention package that you used was responsible for the improvement. But clinically, we don’t care about that so much. We just want to see if we can gather evidence that the client really improved over time, whether that be with only one intervention or multiple interventions at the same time. Alternatively, you could do an AB design, then add another A phase. If they got worse, you could say, “Well, maybe they didn’t get that much better during the first B phase. I’m going to try a different treatment for my next phase.” You’d call this an ABAC design. For example, let’s say you’re treating somebody with PTSD, and you’re measuring PTSD on a reliable and valid scale, of which many exist. You could baseline their functioning over time, and then do, let’s say, progressive relaxation training. If they don’t get very much better but a little bit, you remove the relaxation training and they go back to the initial baseline. Then you think, “Well, I can do better than that. I’m going to use something a little more sophisticated, let’s say, EMDR.” For the C phase, you would try EMDR and see if the client got more improvement with EMDR than they did with progressive relaxation training. So, you can use it to evaluate a sequence of interventions and maybe even to compare the relative efficacy of one intervention over another.
[00:14:33] Jonathan Singer: I mean, that sounds very both practical and sophisticated. One of the things about single system design that I think is so fascinating—and you mentioned this—is that this is the kind of thing an everyday practitioner can and possibly should be using to evaluate practice.
[00:14:58] Bruce Thyer: I agree. But when I was in full-time practice, if I used a single system design to evaluate my work with one out of ten clients, I thought I was doing pretty good. So, I don’t want to imply in any way, shape, or form that practitioners should use these designs with every one of their clients. But sometimes the opportunity arises to make use of it, and it’s tremendously rewarding and clinically useful when it does.
[00:15:20] Jonathan Singer: Are there any problems or levels of practice, like micro, mezzo, or macro, that you think single system design isn’t really applicable to or isn’t really a good idea?
[00:15:35] Bruce Thyer: Not really. If you look in the literature, you can find that single system research designs have been used across all levels of social work practice. Evaluating clinical work with individuals, evaluating group work, evaluating community practice, and even in the evaluation of social welfare policy. In fact, single system research designs and a modification called time series designs are the primary methods by which we evaluate social welfare policies. To give you an example, let’s say a state does not require motorcycle riders to wear a helmet. You can get state statistics, which are already gathered, and look for the 12-month pattern of monthly fatalities before a new law was passed requiring helmet use for motorcycle riders. Then, you look at the pattern for the 12 months or so after the helmet law was introduced. You’d be able to see pretty clearly whether or not motorcycle rider fatalities were reduced following the introduction of a helmet law. That’s macro practice. That’s the evaluation of a health and social policy. It’s relatively easy to do and doesn’t cost much money. You don’t need grants to do single system design work.
[00:16:42] Jonathan Singer: I love that you gave a macro practice example because I think that so often when I think about single system design, I do think about the one client sitting in my office. I think, “Okay, now we’re going to do this and see how it changes.” Are there other examples of single system design that you think really paint a picture of how it can be used in different ways? I’m sure there are people listening who are thinking, “Oh, now that you say that, I’d like to try that.”
[00:17:19] Bruce Thyer: Well, I can give an example early on in my career when I was a consultant for a group home taking care of individuals with the diagnosis of schizophrenia. The activity therapist called me one day and said, “Bruce, I can’t get the residents to participate in an exercise program. They lead a pretty sedentary life. Most of them are heavily medicated and sort of lethargic throughout most of the day. I’ve been trying to get them to participate with me in a morning aerobics program that’s on TV, and they just won’t do it. Can you come up with something to get them to exercise with me?” I thought about it for a while and said, “Is it important they exercise with you or that they actually get aerobic exercise?” She thought about it and said, “Well, it’s more important they exercise, whether it’s with me or not.” So, I said, “Let’s try this. Let’s buy a used stationary bicycle with an odometer on it and put it in the TV room. Invite the residents to use it whenever they wish. Ask the staff to record the time in terms of mileage when people began to ride the bike and when they stopped. Keep track of their mileage daily for a week or two. Then, we’ll implement a reinforcement program to see if completing half a mile, a mile, or a mile and a half increases their daily mileage.” We did this with two residents, each diagnosed with chronic mental illness. During the first week, they didn’t ride the bike at all, to speak of. Then we applied the reinforcement program, and they rode significantly more compared to the first baseline. We then discontinued the reinforcement program and later reinstated it. By the end of the fourth week, one resident was riding six miles a day. Follow-up a couple of months later showed both residents were still riding considerably above the baseline level. That was a nice design, and I actually published it with the activity therapist.
[00:19:27] Jonathan Singer: Wow. So, there was this need to address aerobic activity. You had this AB design—or ABAB. Would that be an ABAB, now that I’m saying that?
[00:19:43] Bruce Thyer: Yes, it would be.
[00:19:45] Jonathan Singer: And then you were even able to publish it. So, journals are publishing single system designs?
[00:19:51] Bruce Thyer: Yeah, they’ve been published in almost all the major social work journals. They’ve also been published in the evaluation of virtually all models of social work practice—even psychoanalysis, Rogerian counseling, EMDR, behavior analysis, and family therapy. You can find examples of using these designs with almost any practice model. Even qualitative approaches like narrative therapy have been evaluated using single subject designs.
[00:20:20] Jonathan Singer: Everything that we’ve been talking about is a single system design. Is there ever a situation where you would have multiple single system designs, or would you use a single system design for multiple different people with the idea of understanding how something affects multiple people?
[00:20:47] Bruce Thyer: Sure. There’s a variation of single system designs called a multiple baseline across clients design. This is where you’d have two or more individuals with the same problem that you’re going to treat with the same type of therapy or intervention. You start baselining both at about the same time. After a week, you implement the intervention with the first client but continue baselining the second client. You’re looking to see if the first client improves when they get the intervention and if the second client continues with a stable baseline. Then you introduce the same intervention after a longer baseline with client number two and see if they improve only after the intervention is applied.
[00:21:42] Bruce Thyer: If you can do this with two people, that’s great. If you can do it with three, that’s even better. Seeing three successive coincidental shifts of improvement in client one, client two, and client three gives you pretty good grounds for causal inference—yes, this intervention caused improvement. But I want to emphasize that we don’t need to use single subject designs to demonstrate the causality of our interventions. I like to think of these designs as being useful to answer two simple questions. The first is: Did the client improve? We can answer that with an AB design. It’s relatively straightforward. The more sophisticated question is: Can we prove they got better because the intervention worked with them? That requires an extended design, like an ABAB or a multiple baseline design. So, yes, you can use these designs with multiple people at the same time. You can also use them in a community setting. For example, let’s say there’s a high rate of crime. You gather baseline statistics over a few months. Then a new type of policing—say, community policing—is introduced. You see if the crime rate goes down after the new method is implemented. Your unit of analysis could be the number of crimes per day, week, or month in an AB-type format. That’s how these designs can be applied to community practice.
[00:23:04] Jonathan Singer: That seems like it would be pretty useful in clinical practice. But as a clinician, is this valuable in other areas of social work?
[00:23:13] Bruce Thyer: Well sure and not just clinical practice but as I said more macro levels as well and this is very useful clinically and in all attempts when we're expecting to intervene at some level to see whether or not the situation has improved over time. Now this can also be good in clinical supervision. So when your supervisor comes to see you or you go see them and they say, "How's your client doing?" And you can say, "They're fine." Um, sometimes a supervisor like me would say, "Well, what evidence do you have? Where's your data?" And if you can pull out a nice AB design showing that the client's problem was high and stable, and then after you introduce the intervention, it fairly rapidly began to get better. That's a really impressive method that you can use in your own supervision. And in selected cases, some social workers take the extra step to write it up in the form of a formal publication or an article that they would submit to a professional journal. And if it gets accepted there, then you started making a contribution to generalized knowledge that other people can read about what you've done and build upon it and extend it. Maybe they can take your technique that worked with one client population and use it with a different type of client population. Um, and these designs are also very useful in terms of developing pilot data on evaluating let's say a novel treatment, a new intervention, a new therapy and rather than trying to get a big grant and spending a lot of money with randomized controlled experiments and things like that, collect a series of single subject designs showing fairly consistent positive effects of the intervention and then that can be used to submit with a research grant to justify a larger scale evaluation. It would be the height of folly to do a randomized control trial on a brand new intervention with no evidence at all besides antidote that it really works. But if you can collect a series of single system or single subject studies, then the grounds for doing a larger scale group study are greatly enhanced. And sometimes these designs can be used if a student is undertaking either an MSW capstone project or a DSW capstone project or a PhD doctoral dissertation. These designs are widely used at Florida State university, we've had six or seven students do single subject designs as their method of data analysis for their PhDs. My first doctoral student did her doctoral dissertation on teaching very young children to consistently use seat belts. This was done before there were laws mandating seat belt use. She trained the whole school. She got good data showing that seat belt was seat belt use was very low. And then when she did her intervention, which was a multicomponent package, it immediately went up to near 100% compliance and then when the program was discontinued after a few weeks she went back a few months later when the program had been halted that whole period of time and the seat belt use of the kids was still very very high so dissertation research DSW projects MSW projects all are grist for the mill for single system research designs.
[00:26:30] Jonathan Singer: And, as you're saying that, I'm thinking about one of the things that we've talked about in the profession for years - I had Danielle Parish come and talk about the process of evidence-based practice. You know, if we think about the steps of evidence-based practice there is that piece of it which says evaluate practices, so you need to evaluate practice. Is this something that you could use in the evaluation component of the process of evidence-based practice.
[00:26:59] Bruce Thyer: Yes, in the original book titled evidence-based medicine from which the model of evidence-based practice was derived from, they outline a five-step process of evidence-based practice. And the first is to come up with an answerable question dealing with what the information needs you have to help your client and come up with some type of intervention. The second step is to track down the available best evidence you can find. The third step is to critically evaluate it. The fourth step is to summarize the results with your clients and with their consent and if it's ethically suitable and you have the resources and so forth actually apply the intervention with your client and then the fifth step of evidence-based practice is to evaluate how well you did both clinically and in terms of implementing the model. So, evidence-based practice and single subject designs are completely consistent with each other and the very book evidence-based medicine has a whole chapter on using single subject designs to evaluate practices. The two go together very very well.
[00:28:04] Jonathan Singer: One of the things that you mentioned was publishing the results of your single system design and that one of the benefits is that somebody could read it and then say, "Oh, I I'm going to try this with my clients with my population." would you say that single system designs are generalizable? I know that that is an issue that we talk about when we talk about different types of research methods. And single system design is quantitative, right? And that that there are numbers, right? And we typically think of qualitative as not being generalizable. But as we're talking, I'm just wondering actually if if if folks consider single system design sort of to be the results to be generalizable.
[00:28:56] Bruce Thyer: Well, in my opinion, no single research finding is generalizable, whether it be single subject design or whether it be a group design. Because as you know, the typical way we infer generalizability from a research project is to have a representative sample drawn from a larger population and randomly drawn at that. That almost never happens. Not even in the biggest of experiments. You could have a thousand people, but it's extremely unlikely they're going to be randomly drawn from a larger population of interest. It's virtually impossible. So, the generalizability of group experiments is not a function of sample size, but of representativeness. And that almost is never possible to gather people r randomly from a larger population of interest. Now, single system designs, of course, obviously these people aren't selected randomly. And the way we get around this in single system design and in group research designs is we're trying to demonstrate generalizability through the ma miracle of replication. You do the study again and see if you get the same results. Then other people do the study again and see if these independent researchers get the same results. Then you do the study in different contexts and in different populations and so forth. And each successful replication of a positive finding enhances the confidence we have that yes, this intervention may really work. And that applies to both single system designs and group designs. Don't let anybody tell you that group designs have greater generalizability. They don't unless their sample was randomly drawn from a larger population of interest and that is virtually impossible to do.
[00:30:33] Jonathan Singer: Well, Bruce, this is, for me personally, this is very exciting to think about the possibilities for kind of rigorous evaluation of what you're doing with your clients and those clients of course could be communities, it could be… there could be, as you were talking about, policy issues and to think about them being publishable if that's something that people want.
[00:30:57] Bruce Thyer: Yeah, I would encourage people listening to this podcast if they're MSW or BSW students to contemplate trying to use these designs in your internship. if they're a doctoral student, think about trying to use them to evaluate your own client work or maybe do secondary analysis of existing agency data that can be formatted retrospectively in the form of a single subject design. And in some cases, I would encourage people who try this type of study to submit them to the many many journals that welcome these types of designs to help us build a knowledge base of our field. And I'd like to conclude with some quotes and see if their listeners can Guess who said these following three statements? Here's the first quote. Every treatment is an experiment. End of quote. Here's the next quote. In work with individuals, averages mean very little. End of quote. And then here's the long one. Quote, "Special efforts should be made to ascertain whether abnormal manifestations are increasing or decreasing in number and intensity as this often has a practical bearing on the management of the case. Now, I won't ask you if you know Who said this, Jonathan? But I will tell you that it was said by Mary Richmond in her classic 1917 book called Social Diagnosis.
[00:32:18] Jonathan Singer: So over a hundred years ago, somebody basically said, "Go do single system design."
[00:32:24] Bruce Thyer: Not just somebody, Jonathan, Mary Richmond, the founder of clinical social work.
[00:32:31] Jonathan Singer: Those are those are amazing quotes to end with. Thank you. So, thank you, Bruce so much for being here on the podcast and talking with us about single system design.
[00:32:42] Bruce Thyer: It's been my great pleasure. I appreciate being asked and I hope the listeners learned something.
~~~END~~~
References and Resources
- Research on Social Work Practice: https://journals.sagepub.com/home/rsw
- Thyer, B. A. (2023). Experimental Research Designs in Social Work: Theory and Applications. Columbia University Press.
- Thyer, B. A., & Pignotti, M. G. (2015). Science and Pseudoscience in Social Work Practice. Springer Publishing Company.
Singer, J. B. (Producer). (2025, January 12). #143 - The Power of One: Single System Design with Bruce Thyer, PhD, LCSW [Audio Podcast]. Social Work Podcast. Retrieved from https://www.socialworkpodcast.com/2025/01/Thyer1.html
No comments:
Post a Comment