Generating and interpreting evidence from psychotherapy: An examination of measurement models, missing cases, and classification methods
thesisposted on 28.03.2022, 17:51 authored by Eyal Karin
The primary aim of this thesis is to explore and identify some of the statistical assumptions that underpin the measurement, statistical analysis, and interpretation of the effects of psychotherapy for anxiety and depression. The identification of these statistical assumption are then used to reflect on the suitability of common statistical techniques that underpin quantitative psychotherapy research and treatment evaluation. A series of five studies are presented, exploring the different statistical assumptions that underpin the measurement of symptom change through treatment (Studies 1 and 2), the handling of missing cases (Studies 3 and 4), and the classification of symptom outcomes into categories that represent the individual impact of treatment (Studies 5). The clinical datasets employed in these studies are comprised from samples of participants enrolled in randomised controlled trials (n>820) or patients enrolled in routine care (n>6700), who receive internet-delivered cognitive behavioural therapy (iCBT) for anxiety and depression. The iCBT context is used as an exemplar psychotherapy context, with highly protocolised procedures which reduces measurement variance due to therapy type or therapist. The results of these studies, identify several statistical assumptions that seem to generalise across psychotherapy data; being the proportional reduction of symptom change, the assumption of missing at random that is conditional on treatment adherence, and the occurrence of proportional symptom change that is non-specific to treatment. These results also indicate that the use of conventional methods for reporting treatment efficacy, including Cohen's d effect size and the Reliable Change Index (RCI), and for statistically adjusting for data missing from clinical trials,including the missing completely at random assumption (MCAR), may result in error in evaluation and interpretation. Each of the five studies also point to the benefits of selecting alternative statistical methods that better fit the context of psychotherapy data, reduce measurement error, and increase the ability to interpret clinical change with increased validity. As a body of work, the thesis seeks to point to a set of methods that strike a balance between the competing priorities of researchers to select methods that fit the specific nuances of psychotherapy data, and the selection of methods that enable the comparison and generalisability of psychotherapy outcomes across different contexts (e.g. different symptom scales and treatment types). The research of this thesis is explored through both statistical and clinical viewpoints but is primarily written and directed for clinical researchers and a clinical audience. Implications for the broader field of mental health research are also discussed.