University of ManchesterLucy Pollyanna GoldsmithPhD MedicineCausal modelling of the therapeutic alliance in randomized controlled trials of complex interventions in mental health2015This thesis addresses one of the oldest debates in psychological therapies: whether the therapeutic alliance has a causal effect on outcome. This was previously unresolved due to the limitations of prior methodologies. The statistical profile of two methods (multiple imputation and structural equation modelling full information maximum likelihood) to deal with missing data in mediators including interactions were investigated using simulation studies. The simulation studies revealed that the methods worked well with simple models, but SEM FIML was biased for models including interactions and some multiple imputation models were biased for simulation studies including interactions and a categorical variable.The causal effect of the therapeutic alliance on outcome in mental health is investigated through analyses of the SoCRATES and FINE randomized controlled trials. This thesis finds that the therapeutic alliance has a causal effect in therapies for schizophrenia, but not in therapies for chronic fatigue syndrome. Further research is needed to investigate the reliability of these findings and to investigate the effects of therapeutic alliance in different disorders. Theoretical work is required in psychology to explore why the therapeutic alliance has a causal effect in some disorders but is irrelevant to the outcome in others.