What is the relationship between dropout from therapy and outcome? Do clients who drop out experience better or worse outcomes than those who complete? And when comparing treatments, is the rate of dropout from one treatment relative to another predictive of the relative effectiveness of those treatments for clients who complete? We highlight a study that set out to provide some answers.
Dropout: a sign that clients aren’t getting what they want or need from therapy and aren’t able to tell us, or that they have got what they want and have forgotten to let us know? On which side of this divide you sit is likely to influence how relaxed you are about dropout.
Personally, I’m far from relaxed about any kind of dropout. Even if my clients have got what they need from therapy, I want to know. If they have got what they need, and I don’t know, then I want to know why I don’t know.
All my experience as a service manager, and more lately as a consultant working in this space, tells me that more often than not, unilateral disengagement from therapy is a sign that something’s not working as it might. I also happen to believe, that with proper contracting, it should be possible to largely avoid satisfied clients leaving without telling us.
What, however, does the evidence say?
What does the evidence say?
I’ve written about dropout before. My first blog on the topic looked at dropout in general terms, including some of the available literature and benchmark data on the subject. In the second, I outlined a simple framework to monitor and reflect on our own levels of dropout, whether at an individual or service level. In the third, I put my own practice under the microscope and looked at my own dropout data over two years.
Joshua Swift & Roger Greenberg [i] suggested a reasonable estimate of dropout rates across therapies would be about 20%. In my experience dropout rates vary enormously across services. I’ve estimated, for example, that IAPT dropout rates for clients that are recorded as receiving one or more treatment sessions are close to 40%. In an earlier study of primary care, dropout rates were estimated to be between 22.5 and 52.4%.
The client has got what they needed from the therapy process and reached a good enough level of benefit not to need more. In not telling us they are exercising their right as autonomous individuals and we must respect that.
As for the relationship between therapy outcomes and dropout, the evidence seems to be rather mixed. [ii] Some studies report more psychological symptoms, less symptom relief and lower satisfaction among clients that dropout. By contrast, others report symptom improvement, fewer symptoms relative to completers, improvement following dropout, and satisfaction with the services received.
With all this in mind, a recently published study set out to illuminate this area a little further.
The current study
The study in question seeks to shed further light on the relationship between dropout and outcome. In particular, it sought to answer three key questions:
Do clients who drop out of therapy start with more or less symptom distress than those who complete?
Do clients who drop out experience better or worse outcomes than clients who complete?
When comparing two treatments, is the rate of dropout from one treatment relative to another predictive of the relative effectiveness of those treatments for clients who complete?
Three separate meta-analytic studies were conducted to try and provide answers to these questions. Brief details of each study and their findings are outlined below. For each individual study that met the inclusion criteria, dropout rates were calculated as the proportion of the total dropped out over the total sample at the start of therapy. A range of variables that might influence findings were also recorded that included: the type of dropout; the number of sessions and the number before dropout occurred; and therapy modality, as well as a number of therapist and client characteristics.
Study 1: Initial level of symptom distress of dropouts
The aim of this meta-analysis was to investigate the pre-therapy level of symptom severity or distress of clients who drop out against those who remain in therapy. It included 76 studies in all.
It was found that clients who dropped out had greater severity of symptoms prior to starting treatment than those who completed. While the effect size was small (d= –0.14), it was nonetheless reliably greater than zero. Notably, it was found that:
Younger clients were reliably associated with more severe distress among dropouts relative to completers
Studies with a greater proportion of male clients were reliably associated with more severe distress among dropouts relative to completers
Studies examining eating disorder treatments showed the highest level of dropout v. completer distress effect size, followed by treatments which examined trauma.
Otherwise, the dropout-completer pre-therapy distress effect size did not reliably vary as a result of any other treatment, therapist, client or study factor at post-therapy or follow up.
Study 2: Post-therapy outcomes of clients that drop out
The aim of the second meta-analysis was to establish whether clients who drop out experience better or worse outcomes than clients that complete, and it comprised data from 43 individual studies.
Overall, the analysis found that clients who dropped out were more distressed following therapy than clients that completed. The differences here were larger than for the first analysis (d= -0.56), which is a medium sized effect. Overall, effect sizes across the studies were statistically significantly similar, however, two variables were found to moderate differences:
The source of the outcome measure accounted for significant differences with the client completed outcome measure effect size being (d= – 0.52) and that for assessor rated outcome measures being (d= -0.98).
There was a positive association between year of publication dropout-completer effect sizes such that effect sizes have become larger in publications over time.
Otherwise, the dropout-completer effect size did not reliably vary as a function of any other treatment, therapist, client or study factors.
Study 3: Do dropout rates between treatments predict their relative effectiveness for completers?
The aim of this meta-analysis was to determine whether the rate of dropout from one treatment relative to another is predictive of the relative effectiveness of those treatments for clients who complete. Put another way, do treatments with more dropouts also have completers with worse outcomes?
In all, 34 studies were included in this analysis.
To explore this question the relative rates of dropout and effectiveness were derived from studies that compared two or more treatments. Also explored were the relative rates of dropout for different dropout types, as well as possible moderators of the dropout-outcome association.
To allow for exploration of dropout rates across different types of dropout, the dropout data in each study was coded to one of six dropout types (see adjacent panel).
The analysis found a clear association between relative dropout and relative outcome that tended to be negative. Greater dropout from a treatment was associated with poorer outcome in that treatment for the completers, when compared to another treatment.
Volitional dropout: Clients chose to end therapy early because of discontent with the therapeutic experience.
Non- volitional dropout: Clients reported extra-therapeutic barriers to their continued attendance.
Withdrawal dropout: A decision was made to end treatment or refer to another service.
Improvement dropout: Study stated that the client was not counted in the list of completers but had improved and no longer wanted or needed the service.
Unknown dropout: Other dropout including where the client had failed to return for additional therapy sessions without further contact.
Overall dropout: The total of the other five types of dropout.
Greater dropout for volitional reasons was reliably associated with poorer outcomes for the treatment.
Where treatments were shorter in length, greater overall dropout was reliably associated with worse outcomes, whereas in longer treatments, the association between overall dropout and poorer outcomes was not reliable. The same pattern was observed for volitional dropout.
No reliable association was found between greater dropout and poorer outcomes depending on treatment length where dropout was due to nonvolitional, treatment withdrawal, improvement, or unknown reasons.
Otherwise, the dropout-outcome effect did not reliably vary as a function of any other treatment, therapist, client or study factors.
What can we take from all this?
From this series of meta-analyses, it would appear, first, that clients that drop out start the therapy journey with a higher level of distress. In particular, younger clients, male clients and those in eating disorder and trauma treatments are reliably associated with higher levels of pre-therapy distress.
Second, it would also appear that not only do clients that drop out start therapy with higher levels of distress, their therapy ends with higher levels of distress than clients that complete. Third, therapies with a higher level of dropout, in particular those of a shorter length, are associated with correspondingly poorer outcomes.
Any narrative overlay needs to be tentative, but if I were to try and make sense of these themes, I’d start with a speculation that clients with higher levels of distress may find it harder to engage with the process from the outset. Therapy can require hard work and focus, and not all clients may be in the right psychological space at the outset.
Unless we can help them to engage successfully with the process, it is quite probable that, early on, they will begin to question how much therapy can help them. If they are not feeling some early sense of benefit, including a reduction in their level of distress, it is more likely that they will disengage. This is entirely consistent with research which shows that one of the most reliable predictors of overall improvement is improvement in the early stages of therapy.
We may also need to determine and manage clients’ expectations about what therapy may involve. As demonstrated in a study published in 2011, [iii] client dropout can be considerably reduced if clients are informed of the likely ‘dose’ of therapy they may require.
With all those points in mind, these are my suggestions for minimising dropout and maximising outcomes:
Hold an awareness of the potential for higher levels of pre-therapy distress among particular groups of clients, especially younger clients and men, and those in eating disorder and trauma interventions.
Monitor pre-therapy levels of distress of clients using a standard measure to determine, as far as possible, patterns of dropout across particular client group.
Discuss clients expectations of what therapy may involve and offer an evidence-based perspective of the possible trajectory of change.
Maintain meticulous attention to the three elements of the alliance (goals, tasks and bond), especially in the earliest stages of therapy.
Monitor clients’ progress and pay particular attention to those clients who appear not to be making progress and may be at greater risk of dropping out.
Routinely record dropout type, using a taxonomy such as the one used in the third study above, or the standard CORE taxonomy of ending types.
[i] Swift JK & Greenberg RP. Premature Termination in Psychotherapy: Strategies for Engaging Clients and Improving Outcomes. 2012. APA, Washington.
[ii] Reich CM & Berman JS. 2018. Are psychotherapies with more dropouts less effective? Psychotherapy Research, 2018 https://doi.org/10.1080/10503307.2018.1534018
[iii] Decreasing treatment dropout by addressing expectations for treatment length. Joshua Swift & Jennifer Callaghan. Psychotherapy Research · March 2011
We thrive on feedback, so please let us know what you think about what you’ve read in the comment section below. Only the name you use to identify yourself will be shown publicly. Thank you!