A Telehealth Tragedy

Hey guys: do you remember Covid? A formerly face-to-face service delivered in homes – ABA – was converted into a majority-telehealth service virtually overnight. During those first few weeks, we had a short but productive journal search, attempting to answer the question: can this even work?

The answer, thanks mostly to David Wacker and the great state of Iowa, is an unqualified yes. Telehealth, while not for everyone, can work. Long before Covid (2006 in fact), several dozen peer-reviewed articles confirmed the basics. 

But one article stuck out. It was an article that described a time when it didn’t work.

Schieltz et al. (2018) begins with an inspiring description of the work of co-author David Wacker, the aforementioned telehealth pioneer. Functional communication training delivered via telehealth can reduce the problem behavior of children by over 90%. But some studies hint at deeper problems: in two studies highlighted, FCT was effective in only about 60% of cases, and in other cases it appeared to work rather slowly. 

Our intrepid authors specifically call out the file drawer problem, wherein authors only attempt to publish positive results, and present their analysis of two failures.

A previous large study used functional analysis and a function-based FCT treatment, and most kids communicated their needs, reduced problem behavior by 80% or more, and increased compliance to parent demands. Not so for Liem and Sara.

Liem was an autistic boy nearing 3 years old, and he engaged in self-injury, aggression and property destruction. Sara was an autistic girl nearing 7 years old, and she engaged in aggression and property destruction. During their functional analysis, neither manded.

The authors conducted telehealth coaching sessions, during which the parent was expected to:

• Instruct the child to complete a specific amount of work,

• Require that the child mand after completing work (by pressing a switch) before giving reinforcement,

• Provide reinforcement,

• And return the child to work if problem behavior occurred during reinforcement

The consultants were expected to correct errors in a timely manner.

Sara’s challenging behavior occurred in the establishing operation (EO) condition, rather than the reinforcement condition. That is to say, she engaged in problem behavior when asked to complete a task, but not while she was playing. Therefore Sara’s problem behavior was socially mediated. But despite coaching the social mediator, Sara’s mom, the situation wasn’t improving.

Treatment fidelity measures found that while the consultant was correcting errors at 100% fidelity, Sara’s mother was implementing the treatment at approximately 45% fidelity.

At 45% fidelity, it is actually more likely that the “treatment” is making things worse. We can’t know exactly why Sara’s mom did not adhere to the treatment, but Sara’s behavior did not improve.

Liem, on the other hand, experienced treatment fidelity around 96%. In fact, his behavior initially improved significantly. But his challenging behavior began to occur more and more frequently in the reinforcement condition, until there was an equal amount of challenging behavior occurring at all times. What could be happening?

The authors hypothesize that his behavior began to be maintained by automatic reinforcement. It’s true that the function of behavior can change, and that socially mediated behaviors can become automatically maintained. For example, you try balut to escape social pressure, only to find it delicious, and find yourself fighting a severe balut addiction (this has never happened).

The authors conclude, rather bravely, that telehealth may not be the right service for either child. Sara’s mom can’t be blamed – perhaps an in-person service would be a better fit. Liem’s behavior stopped responding to treatment, and the authors (in 2018) were not aware of telehealth treatments for automatically maintained behaviors.

But the analysis of the failures made that clear. If not for the publication of this work, these two kids would have been two data points in a larger successful project – “16 of 18 kids improved” or something along those lines. Instead, we get a glimpse at how to improve treatment: by examining failure.