Can clinicians predict drop-out and treatment outcome?
How good are psychotherapists at making prognostic judgments? Is there a magical set of client characteristics that predicts prognosis? Smith and colleagues (2020) explored this to a limited extent in a Scottish primary care mental health team. The team provides clients with access to up to 10 sessions of cognitive behaviour therapy (CBT) or interpersonal therapy (ITP), guided self-help or antidepressants. At the end of the first assessment session, mental health clinicians were asked (a) whether the client was more than 50% likely to attend the next appointment with them and (b) whether the symptoms were likely to improve in 5 sessions or less. Clinicians thought 90% clients were more than likely to attend the next appointment and 77% clients would improve in 5 sessions or less. The first assessment session also collected demographics (age, gender, ethnicity, postcode and employment status), symptom measures (Patient Health Questionnaire [PHQ-9], Generalised Anxiety Disorder questionnaire [GAD-7], Clinical Outcomes in Routine Evaluation [CORE-10]) and two client complexity measures (more on this later).
During the period of the study, 576 clients were eligible for the study, of which 298 (52%) were recruited and 258 (45%) had sufficient data. Neither the symptom measures nor the complexity measures predicted treatment “drop-out” which – as best I can tell from the ambiguous explanation - was defined as not attending any of sessions 2 to 5 after attending session 1. However, being a current smoker was associated with “drop out” and clinicians were correct 80% of the time when they predicted clients would not return. It might be comforting to many readers to know that attrition was high: Only 62% of those who attended their assessment returned for a second session and only 15% completed 5 sessions!
Clinicians could not predict who would improve within 5 sessions. On average, the sample was in the moderate range for both anxiety and depressive symptoms at baseline. Of those who attended at least 2 sessions, 62% achieved reliable improvement on the PHQ-9. Of demographic variables, only being younger predicted greater likelihood of improvement; gender and antidepressant use did not. Baseline PHQ scores also predicted likelihood of improvement, with higher scores indicating greater likelihood improvement.
The fact that case complexity measures failed to predict improvement is interesting; it’s worth describing these measures in more detail. Here’s a list of what was included:
Childhood adversity (being bullied in the school or workplace, being sexually, physically or emotionally abused, being neglected, having child protection/social work involvement in childhood, being adopted or fostered, personal alcohol or drug use/dependence, parental alcohol or drug use/dependence, family history of mental illness, family history of suicide, previous personal history of anxiety or depression)
Current problems (childcare problems, money problems, relationship problems, family/friend problems, being a vulnerable adult, work-related stresses, unemployment and seeking work, bereavement/loss, grievance, active legal issues, housing problems, being a carer, children < 18 living at home)
Physical health problems that impact daily life
It’s always challenging to quantify complexity and there is an ongoing debate about the validity of just tallying a checklist of misfortunes to represent adversity or complexity. Nevertheless, this approach often works. So that readers might reflect on the similarity between their caseloads and this sample, this sample reported an average of 4 of the listed current problems and 3 of the listed past problems. Given that this level of adversity did not appear predictive of poorer treatment response, it may be worth considering whether your predictions for clients on your caseload are appropriately optimistic.
In drawing conclusions about therapist’s ability to judge prognosis, it should be noted that the therapists in this sample seemed to overestimate rather than underestimate the likelihood of improvement. Unless this leads therapists to neglect recruiting additional more promising services, a hopeful bias toward clients’ chances of success is arguably not a bad thing. However, hope and expectations are different. In my current supervision intervention study, I see many hard-working, conscientious therapists who mistake hope for expectation: they beat themselves up for not achieving 77% improvement rates when – for all therapists – the likelihood of improvement is more like 38% (that’s the proportion of all who were assessed that achieved reliable improvement in 5 sessions in this study). Most therapists in my experience are perfectionists: It’s fine to aim high, but go easy on yourself (and your clients) when you (unsurprisingly) don’t achieve those heights.
Take home: Clinical Implications
Drop out in routine primary mental health care is common: Only 62% of those who attended assessment attended session 2; and only 15% attended session 5.
When therapists expected drop out they were right 80% of the time
Therapists did not predict who would improve within 5 sessions; they expected more people (77% those assessed) to improve than actually did (38% those assessed)
Case complexity factors including past and current social and physical health problems did not predict improvement
Only being younger and having higher baseline depressive symptoms significantly predicted improvement: both predicted greater likelihood of improvement
For the original article, go to:
https://bmcpsychiatry.biomedcentral.com/articles/10.1186/s12888-020-02532-0
REFERENCES
Smith, M., Francq, B., McConnachie, A., Wetherall, K., Pelosi, A., & Morrison, J. (2020). Clinical judgement, case complexity and symptom scores as predictors of outcome in depression: an exploratory analysis. BMC Psychiatry, 20: 1-11.