University of Twente Student Theses


Predicting early indicators of dropout in online therapy for problem drinkers : using LIWC to analyse email contact between client and counsellor

Hazel, T.S. van den (2020) Predicting early indicators of dropout in online therapy for problem drinkers : using LIWC to analyse email contact between client and counsellor.

[img] PDF
Abstract:Mental disorders such as Alcohol Use Disorders (AUD), including alcohol abuse and dependence, are highly prevalent in the Netherlands. Problem drinking affects the mental well-being of those suffering from it and often cause other psychological issues, mainly anxiety and depression disorders. Online therapy treatments designed for problem drinkers are effective on average, though do not consider individual differences, and are characterised by high dropout rates. Moreover, premature quitting the intervention often results in worsening or chronic symptoms. It is important to note that there exists a literature gap in predicting dropout in online interventions using linguistic predictors, as non-linguistic predictors for dropout such as gender, age and education have been researched before. Therefore, the purpose of this research was to identify what factors can predict dropout in online therapy, by means of using non-linguistic and linguistic factors. It is important to identify those factors; therapists are able to tailor their treatment in order to facilitate completion for their clients, which is essential for better therapy and long-term outcome. Method: A sample (n = 990) of clients was taken from the existing intervention found on developed by Tactus. Only clients who followed the intensive program were included in the sample.. The client data was first anonymised, then used to read conversations between the client and counsellor to label clients as dropout or completer. Clients were labelled as dropout when they did not complete all assignments. An explorative approach was used to investigate whether linguistic and non-linguistic factors could predict for dropout. Two logistic regression tests were performed to predict for dropout with demographic characteristics and LIWC categories. Results: Predictive factors for dropout were being male, younger of age, lower education, smoking, and a higher baseline of alcohol intake per week. A number of categories of LIWC were found to predict for completion: 3rd person singular, impersonal pronouns, common adverbs, positive emotion, negative emotion, male references, tentative, biological processes, affiliation, focus past, leisure, and informal words. Conclusion: It is possible to predict for completion and dropout using linguistic factors and might give more insight on dropout in online therapy in combination with non-linguistic predictors which have been researched before. A main limitation of this study was that only the average of the first four mails were used. Notably, linguistic and the non-linguistic predictors alike had small differences between the groups of dropout and completers and small predictive values. However, with no literature existing on using linguistic factors to predict for dropout in web-based treatments, more research is recommended. It could be useful to look at the first few mails separately, and include more non-linguistic factors regarding other psychological problems to broaden the scope of the current study.
Item Type:Essay (Bachelor)
Faculty:BMS: Behavioural, Management and Social Sciences
Subject:77 psychology
Programme:Psychology BSc (56604)
Link to this item:
Export this item as:BibTeX
HTML Citation
Reference Manager


Repository Staff Only: item control page