Linguistic predictors of symptom change in an internet-based cognitive behavioural intervention for prolonged grief symptoms

This study investigates linguistic predictors of reduction in prolonged grief symptoms following a writing intervention in an internet-based cognitive behavioural therapy for people bereaved by cancer. Data stem from a randomized control clinical trial with 70 people. The Linguistic Inquiry and Word Count program was used to analyse patient language. Absolute change scores and reliable change index were used to calculate reduction in grief symptoms and clinical significant change. Best subset regression and Mann-Whitney U tests were conducted. A higher reduction of prolonged grief symptoms was correlated with more social words in the first module (beta = -.22, p = .042), less risk (beta = .33, p = .002) and body words (beta = .22, p = .048) in the second module and more time words in the third module (beta = -.26, p = .018). Patients with clinically significant change showed a higher median in function words in the first module (p = .019), a lower median in risk words in the second module (p = .019) and a higher median in assent words in the last module (p = .014) compared to patients without clinically significant change. Findings suggest that it may be beneficial for therapists to encourage a more detailed description of patients' relationship with their deceased relative during the first module, a change in perspective during the second module and a summary of past, present and future aspects at the end of therapy. Future studies should include mediation analyses to allow causal attribution of the studied effects.