The Relationship Between Dissociation and Symptoms of Psychosis

A Meta-analysis

Eleanor Longden; Alison Branitsky; Andrew Moskowitz; Katherine Berry; Sandra Bucci; Filippo Varese


Schizophr Bull. 2020;46(5):1104-1113. 

In This Article


Search Strategy

This review was conducted and reported according to PRISMA[20] and its protocol registered on Prospero (CRD42017058214; for PRISMA guidelines, see Supplementary Table S1). MEDLINE, PsycINFO, PubMed, and Scopus were systematically searched using strings for dissociative and psychotic symptoms: (dissociat* OR depersonali* OR dereali* OR absorption OR multiple personalit*) AND (psychosis OR psychotic OR schizophren* OR hallucinat* OR voices OR visions OR delusion* OR paranoi* OR cognitive disorgani* OR positive symptoms OR negative symptoms OR first-rank symptoms OR Schneiderian). A hand-search of references and citations from eligible articles was also performed in order to identify additional studies. Primary searches were completed in April 2017 and updated in January 2020. Articles were subsequently assessed for eligibility based on screening of titles, abstracts, and full texts and only retained for review with consensus agreement from at least three reviewers. Details of the search and screening procedure are presented in Figure 1.

Figure 1.

Flow diagram of systematic search.

Inclusion and Validity

Studies meeting the following criteria were included for review: (1) published in English in a peer-reviewed journal, (2) use of self-report measures of dissociation, (3) use of self-report measures of psychotic symptoms, and (4) use of quantitative methods to report on the association between dissociative experiences and psychotic symptoms. Studies meeting these criteria were subsequently excluded if: (1) they were presented in a conference abstract or single case study format; (2) global scores of psychosis measures were provided rather than separate measures of positive and/or negative symptoms; (3) the study did not report sufficient statistical information to estimate effect sizes; and (4) there were overlapping participant samples. When multiple reports considered overlapping samples, we only included the report which provided a more precise estimate of the effects (ie, considered a larger sample size) or contained more complete statistical information to estimate relevant effects. No restrictions were placed on age or diagnostic status of participants, study design, or study start date.

Study quality was evaluated using relevant items from the Effective Public Health Practice Project tool[21] (EPHPP). This is an instrument used to evaluate health research on the basis of study design and methodology, sample selection, and analytic methods and has demonstrated validity[22] and inter-rater reliability.[23] Studies were assessed by A.B. Oversight was provided by F.V., E.L., and S.B., with any queries or disagreements over scoring decisions resolved amongst these authors.

Data Extraction

Data extraction was conducted by the first two authors and systematically checked for accuracy by F.V. Information extracted from the primary studies was recorded on a standardized form and included general characteristics (eg, county, publication year), design, sample characteristics (eg, age, gender, diagnostic status), measures used to assess psychotic symptoms and dissociative experiences, the specific symptoms and subtypes of dissociation considered in the article, and statistical information to compute relevant effect sizes.

Effect Size Computation and Statistical Analysis

Analyses were conducted using CMA v.2. Results were pooled using a random-effects meta-analysis. Pearson's r correlation coefficient was used as the primary metric, as effect sizes of the r-family were most commonly reported in the literature. When studies reported statistical information consistent with other families of effects (eg, d-family and binary effects), these were converted to effect sizes of the r family using computations methods outlined by Borenstein et al.[24] To ensure that different study designs (ie, between-group and correlational) did not impact the findings, a subgroup analysis was conducted to contrast the magnitude of the aggregated correlational and between-group effects extracted from the primary studies. In all analyses, heterogeneity was assessed using the Q-test and the I 2 statistic. Publication bias was also assessed through visual inspection of funnel plots and Egger's tests. When evidence of publication or other selection bias was evident, analyses were followed with the trim-and-fill method to estimate the influence of potentially missing studies on summary effects.

The following analytic approach was taken. First, we summarized effects considering the relationship between dissociation and global symptom cluster measures (ie, positive, negative, and disorganization symptoms). These analyses were first conducted by including studies which reported effects for total symptom cluster measures (eg, PANSS positive symptom scores) as well as aggregated effect estimates in the case of studies that provided statistical information pertaining to specific symptoms within that cluster (ie, aggregated effects for studies reporting multiple correlation coefficients between dissociation and positive symptom measures, eg, hallucinations and delusions). These analyses were followed-up through multiple subgroup analyses, including: (1) comparison of clinical and nonclinical studies and (2) sensitivity analyses focusing on total symptom cluster measures (when sufficient numbers of studies were available). Secondly, for each cluster of symptoms, multiple meta-analyses were conducted to examine the association between dissociative experiences and specific symptoms of psychosis. Whenever the number of studies allowed it, we also explored the association between various psychosis symptoms and symptom clusters and specific subtypes of dissociation assessed by the Dissociative Experiences Scale[25] (DES-II: ie, absorption, depersonalization/derealization, and amnesia).