Modelling Trajectories of Symptom Change in Feedback-informed Psychotherapy for Adults and Youth

Abstract

Psychotherapy has long been recognised as an effective treatment to alleviate distress, promote mental health, and enhance functioning for both adults and youth. However, it is not equally effective for all clients, particularly in routine care which has lower effectiveness rates than highly standardised research trials. One promising method of enhancing the effectiveness of psychotherapy is feedback: the process of routinely assessing clients’ self-reported symptoms throughout psychotherapy and providing the treating psychotherapist with feedback about these outcomes. To date, several reviews and meta-analyses have found strong evidence that feedback increases rates of improvement and decreases rates of deterioration for clients in adult psychotherapy – particularly for clients who are not progressing as expected and are at risk of deterioration. However, there has been limited investigation into the effectiveness of feedback in youth psychotherapy, despite often higher rates of deterioration for youth. The first of the three studies in this thesis examined whether feedback using the Youth-Outcome Questionnaire (Burlingame, Wells & Lambert, 1996) could also improve outcomes for youth clients in routine care in a psychology training clinic. A benchmarking design compared the effectiveness results to published benchmarks for both adults and youth. Adult results were also examined to determine whether feedback had been implemented with enough fidelity to affect outcomes. Psychotherapy with a feedback-informed approach was more effective than treatmentas- usual benchmarks, with 50% of adults and 64% of youth significantly improving after psychotherapy. Rates of adult improvement were similar to the feedback benchmark, although there was a slightly higher rate of deterioration in the current sample. There were no feedback benchmarks for youth using the Y-OQ, making this sample the first benchmark for future studies. Having established that training clinics can achieve similar outcomes to those reported in the feedback literature, the second and third studies examined whether current feedback procedures could be enhanced by considering individualised trajectories of treatment response. One aspect considered to be core to effective feedback is the comparison of an individual’s progress with an expected treatment response curve. However, this expected treatment response is calculated from group averages and assumes that all clients respond the same way to treatment. Comparing an individual client’s trajectory to a group average invokes the ecological fallacy, as group averages do not accurately represent the individuals comprising the group if there is heterogeneity within the group. Indeed, there has been increasing evidence that clients respond differently to psychotherapy, and that this variability is clinically meaningful. Using separate adult (Study 2) and youth (Study 3) samples, individualised trajectories of client symptom change were modelled using the Bayesian statistical paradigm. Both studies found that individualised trajectories were highly heterogenous, varying on model shape, and rate, direction and magnitude of change. For both adults and youths, the most common trajectory shapes were constant and linear, highlighting that rate of change is often consistent throughout psychotherapy. However, clients bestfitting a constant model tended to have poorer outcomes (i.e., less improvement) and attended fewer sessions than clients best-fitting other models shapes. Next, both studies used an updating approach to simulate real-time data collection to demonstrate how individualised modelling could be incorporated into feedback procedures to predict a client’s future change from their past change. Trajectory projections were highly specific and tailored to the individual based on their past response to psychotherapy. For adults, model shape changed over time for approximately 50% of the clients, but for youth, model shape change was more common with approximately 85% of the clients who remained in psychotherapy for at least ten sessions experiencing at least one change in their model choice. Case examples were presented to illustrate how the model predictions could be interpreted. This individualised modelling approach could be incorporated into already existing software, such as the OQ-Analyst (OQ Measures LLC, 2007). This thesis has uncovered the limitations inherent in applying group averages to individual clients. A shift in the paradigm is therefore necessary to better understand changes that occur during psychotherapy. This individualised modelling approach could improve feedback to psychotherapists, and also result in enhanced patient-focused research as psychotherapists could generate and test hypotheses about clients’ future change to uncover the mechanisms of change for each client. This would allow psychotherapy to be better tailored to each client’s unique response to treatment.