Doctorate in Clinical Psychology: Main Research Portfolio

: 1) Efficacy of CBT-based Self-Help Interventions for Psychosis: A Systematic Review and Meta-Analysis; 2) Description of referral patterns in a Cancer Psychology Service in terms of age, gender, ethnicity, and tumour site compared with overall confirmed cancer diagnoses; 3) Clinical and Demographic Predictors of Paranoia Scores following a Structured Digital Task: A Secondary Data Analysis using Data from a Randomised Controlled Feasibility Study.

Abstract

Background

Cognitive Bias Modification trials often demonstrated improvements in both intervention and active control arms. Identifying who benefits regardless of treatment allocation can be helpful to understand which patients are likely to respond to engaging in a structured digital task in atrial.

Methods

A secondary data analysis of a two-arm feasibility RCT (N = 63) of Cognitive Bias Modification for Paranoia (CBM-pa) versus active control was conducted, pooling both trials arms. The primary outcome was the Paranoia Scale (PS) measured at baseline, post-intervention, 1- and3-month follow up. A linear mixed effects model with a random intercept for participant and time as a categorical variable (with four levels; baseline, post-intervention, 1- and 3-monthfollow up) tested time x predictor interactions for baseline cognitive flexibility, negative symptoms, age, and gender. Age of onset of distressing paranoia did not improve model fit and was excluded. The final analytic sample was N = 57.

Results

Fixed effects explained 23% of variance (marginal R² = .237); the full model including random effects explained 79.2% (conditional R² = .792). statistically significant effects were detected for negative symptoms at T1, T2 and T3 (B = 0.52 to 0.84; 95% CIs [0.03, 1.31]; semi-partialR2= .004 to .014), cognitive flexibility at T2 and T3 ((B = 0.57 to 0.59; 95% CIs [0.15, 0.99];semi-partial R2 = .005 to .009) and age at T3 (B = 0.60; 95% CI [0.26, 0.93]; semi-partial R^2= .017)very small (semi-partial R2.000 – 0.017). The unique amount of variance explained by each predictor x time interaction was very small (.004–.017; = 0.4–1.7% unique variance). All remaining interactions were null, with 95% CIs overlapping zero and semi-partial R2<= .004.

Conclusions

The results were extremely limited due to inadequate statistical power, and all results must be interpreted cautiously. Across both trial arms, baseline negative symptoms, cognitive flexibility, and age showed statistically detectable but tiny associations with change in paranoia. Given the small, underpowered feasibility sample and multiple tests, these findings are exploratory.

Date of Award	10 Sept 2025
Original language	English
Awarding Institution	University of Bath
Supervisor	Pamela Jacobsen (Supervisor)