Dissociation is a ubiquitous construct in modern psychopathology. DSM-5 defines dissociation as “disruption of and/or discontinuity in the normal integration of consciousness, memory, identity, emotion, perception, body representation, motor control, and behavior” (
1). The corresponding phenomena cover a range from relatively common experiences, such as being completely absorbed by a book or movie, to severe states, such as not recognizing oneself in the mirror (
2). More common experiences have often been linked to mild forms of absorption, that is, focusing on one aspect of experiences and blocking others (
3). More severe dissociative experiences are reflected in the DSM-5 subtypes of dissociative disorders: dissociative amnesia describes the inability to recall autobiographical information; depersonalization/derealization disorders comprise experiences of feeling disconnected or estranged from one’s body, thoughts, or emotions and/or perceiving one’s surroundings as foggy, surreal, or visually distorted (
1).
Beyond the disorders primarily characterized by dissociation, “transient, stress-related severe dissociative symptoms” serve as a criterion for borderline personality disorder (
1), and a dissociative subtype of posttraumatic stress disorder (PTSD) was introduced in DSM-5 (
4). Less noted but equally important research has shown that dissociative features also seem to play a role in the pathology of many other mental disorders, such as schizophrenia (
5), eating disorders (
6), panic disorders (
7), affective disorders (
8,
9), and obsessive-compulsive disorder (OCD) (
10).
Dissociative symptoms in mental disorders are of high clinical relevance. They have been linked to maladaptive functioning and symptom severity in some disorders, such as executive functioning in borderline personality disorder (
11), neuropsychological performance in depression (
9), number of binge episodes in eating disorders (
6), alexithymia in panic disorders (
12), and anxiety and depression in OCD (
13). Apart from a higher burden of illness, patients may also benefit less from psychotherapeutic interventions. Several studies have indicated that dissociative symptoms can serve as a predictor for nonresponse in psychotherapeutic treatments of PTSD (
14–
16), OCD (
17–
19), and panic disorders (
20).
Transdiagnostically, the experience of dissociative symptoms has been linked to acute or chronic stress (
21). Neurobiological findings suggest that dissociative phenomena are likely to disrupt information processing, learning, and memory on various levels (
22). Dissociation has been further linked to physiological processes such as sleep (
23) and fluid intake (
24), as well as to personality variables, such as fantasy proneness and suggestibility (
25). On a cognitive-emotional level, dissociation may be a learned automatic response to reduce or avoid aversive emotional states (
26,
27). As a secondary process, the experience of dissociation can induce stress itself because it not only disrupts neurocognitive functioning, but can also be perceived as losing control (
28). Recurrent dissociation may therefore reduce the individual’s confidence in reality monitoring ability, perceived control, and sense of self (
29,
30), which in turn may result in a higher burden of disease.
The transdiagnostic evaluation of those mechanisms is impeded by the fact that neurobiological studies have been mostly conducted in populations of patients who had experienced various traumas, often chronically and/or early in life (e.g.,
31). Although the statistical association was found to be rather small in some studies (
27), several studies have pointed to a strong association between trauma and dissociation (
32–
36). Thus, the experience of trauma does not seem to be a
conditio sine qua non for pathological dissociation. Studies covering a broader range of mental disorders could shed light on common mechanisms and enhance the development of transdiagnostic treatment modules to deal with dissociative symptoms. The meta-analysis we present here aims to stimulate this line of research by providing an overview of the occurrence of dissociative symptoms across mental disorders.
By far, the most commonly used psychometric instrument for the assessment of dissociative experience is the Dissociative Experiences Scale (DES) (
2). The DES is a self-rating instrument comprising 28 items that build on the assumption of a “dissociative continuum” ranging from mild normative to severe pathological dissociation. Subjects are asked to make slashes on 100-mm lines to indicate where they fall on a continuum for questions on experiences of amnesia, absorption, depersonalization, and derealization—for example, “Some people have the experience of driving a car and suddenly realizing that they don’t remember what has happened during all or part of the trip. Mark the line to show what percentage of time this happens to you” (
2, p. 733). As the scoring procedure of the continuous scale was time consuming, a revised version of the scale, the Dissociative Experiences Scale–II (DES-II) (
37) uses an 11-point Likert scale ranging from 0 to 100.
Studies on the psychometric properties of the scale have shown high validity and reliability for both versions, both in clinical and nonclinical populations (
38–
42). The first, and so far the only, comprehensive meta-analysis on the DES, by van Ijzendoorn and Schüngel (
43), conducted in 1996, shows a mean Cronbach’s alpha of 0.93 in 16 studies, a high predictive validity concerning dissociative disorders and PTSD, as well as a high convergent validity with alternative measures of dissociation (mean Cohen’s d=1.82; N=5,916). While initial studies (e.g.,
44) found a three-factor structure with the factors amnesia, absorption, and depersonalization/derealization, the factorial structure of the DES remains controversial (
41,
42,
45,
46).
Considering the high number of original publications on the DES (N>2,000), few meta-analyses have been conducted. One meta-analysis on schizophrenia showed a large effect size comparing dissociation scores of patients (N=293) and healthy subjects (N=474) (g=−0.86, 95% CI=−1.13, −0.60), with trauma history being a potential mediator (
5). Scalabrini et al. (
47) compared the dissociation scores in borderline personality disorder with those in other mental disorders and found significantly elevated dissociative symptoms in patients with borderline personality disorder compared with patients with all other disorders (N=2,035; d=0.54, p<0.01) but lower levels of dissociation than in patients with PTSD (d=−0.50, p<0.01) and dissociative disorders (d=−0.35, p<0.05). As noted, the only comprehensive meta-analysis, by van Ijzendoorn and Schüngel (
43), was published about 20 years ago and included 85 individual studies with about 6,000 patients. As expected, the highest scores for dissociation were found for dissociative disorders (mean=35.3), followed by PTSD (mean=32.6), affective disorders (mean=19.4), schizophrenia (mean=19.1), personality disorders (mean=16.6), eating disorders (mean=14.5), and anxiety disorders (mean=10.2). Comparison scores were calculated for healthy samples (mean=11.57) and students (mean=14.27). The authors conclude that “against the background of potential comorbidity and undiscovered dissociation, the means for normals and nondissociative patients were remarkably similar” (
43, p. 372).
Since the meta-analysis by van Ijzendoorn and Schüngel (
43), dissociation has been studied in a range of mental disorders that had not been included, such as OCD (
10) and substance abuse (
48). Other research has shown that dissociation plays a role in diseases like panic disorders (
7,
31), which showed surprisingly low mean dissociation scores in that first analysis. The goal of our meta-analysis is thus to provide an evidence base for the prevalence and distribution of dissociation in adults suffering from mental disorders.
Method
Study Selection
We searched the following databases for primary studies through November 2016: PubMed, PsycINFO, Web of Science, and Academic Search Premier. Our search strategy aimed at articles using the DES or the German version of the scale (FDS) (
49,
50) in adults with mental disorders. Although there are formal differences between versions I and II of the DES (visual analogue scale versus Likert-type scale, both ranging from 0 to 100), differences in the results for the two versions have been shown to be negligible (
51). Therefore, we decided not to differentiate between the versions of the scale. We developed the search strategy for PsycINFO (“dissoc* exper* scale” OR “FDS”) and adapted it for the other databases. We reviewed relevant review articles and related systematic reviews to identify studies that were missed in the database searches. If full text was not retrievable from online databases or university libraries, we contacted the corresponding authors. There were no language or publication date restrictions.
Two trained investigators independently screened titles and abstracts for relevance. In the full-text screening, the following inclusion criteria were imposed: 1) studying a population with mental disorders diagnosed according to ICD (
52) or DSM; 2) reporting the sample size and the mean score and standard deviation on the DES, or sufficient information to calculate them; and 3) specification of psychometric properties for translations of non-English versions of the DES. Data were extracted by two independent raters using a standard form and systematically screened for full agreement between raters. Every disagreement was resolved by discussion within the review team. The protocol for this meta-analysis is available in PROSPERO (the “International prospective register of systematic reviews”) and can be accessed at
http://www.crd.york.ac.uk/prospero/display_record.asp?ID=CRD42015020731.
Data Synthesis
Diagnostic group, mean and standard deviation of the dissociation score, and number of participants were extracted from the primary studies. For each diagnostic group, the random-effects model described in DerSimonian and Laird (
53) was used to calculate a group-specific mean and the 95% confidence interval. This approach allows for the integration of data from intrinsically heterogeneous populations that result, for example, from the use of different diagnostic systems. To quantify heterogeneity of the dissociation scores between studies, we used I
2 (
54)—an index, based on chi-square statistics and degrees of freedom, that was recommended for Cochrane Reviews (
55). Because only descriptive data on dissociation scores were included in the analysis, the risk of bias in the primary studies was assumed to be unlikely and therefore was not assessed. Data synthesis was conducted with R, version 3.2.4 (
56), using the metafor package (
57).
Results
The search in the electronic databases yielded 1,907 different articles (
Figure 1). After exclusion of 660 articles during title or abstract screening, 1,247 articles were retrieved for full-text screening, of which 1,026 were subsequently excluded; reasons for exclusion are listed in
Figure 1. Across all diagnostic groups, we included 216 articles with a total of 15,219 individuals.
To calculate meta-analytic statistics, the original studies were grouped according to the DSM diagnosis described in the articles. For some diagnoses, this procedure revealed specific subcategories of DSM chapters (e.g., gambling disorders). For some categories, only articles reporting on broader categories or entire DSM chapters (e.g., bipolar disorders) were found. To avoid the confounding influence of diagnostic specification, we included articles reporting on subcategories in both the relevant subcategory as well as the corresponding broader category. Articles that reported on more than one diagnostic group were included in every category the authors reported dissociation scores on. In cases of co-occurring disorders, we included the individuals in both categories. We included all subcategories in which at least four studies reported data, regardless of whether this subcategory of disorders is still included in DSM-5. In the final step, we excluded five studies because there were not enough studies for each diagnosis: one study each on kleptomania (
58) and pathological Internet use (
59) and three studies on mixed personality disorders (
60).
Diagnostic categories, number of individual studies, and number of individual patients as well as statistics are listed in
Table 1. A graphical illustration of the results is presented in
Figure 2. Forest plots of each diagnostic category are included in the
data supplement that accompanies the online edition of this article.
The highest dissociation scores were found for dissociative identity disorders, with a mean score of 48.7 (95% CI=46.4, 50.9), based on 29 publications with 1,313 patients (
Figure 3; the full reference list of included studies can be found in the
online data supplement).
Scores for posttraumatic stress disorder were the second highest, with a mean score of 28.6 (95% CI=25.6, 31.5), based on 33 publications with 2,106 patients (
Figure 4).
Scores for borderline personality disorder were third largest, with a mean score of 27.9 (95% CI=25.3, 30.6), based on 27 publications and 1,705 individual patients (
Figure 5). Scores for other mental disorders were distributed among (in descending order) conversion disorder (mean=25.6), somatic symptoms disorder (mean=18.8), substance-related and addictive disorders (gambling disorder, mean=19.9; alcohol use disorder, mean=19.7; other substance-related disorders, mean=17.7), feeding and eating disorders (mean=18.6), schizophrenia (mean=17.8), OCD (mean=15.3), depressive disorders (mean=15.3), anxiety disorders (mean=15.2), and bipolar and related disorders (mean=14.8).
Only three categories yielded enough studies to analyze dissociation subfactors: borderline personality disorder, dissociative disorders, and schizophrenia. Patients suffering from borderline personality disorder and schizophrenia had the highest scores for absorption, and patients with dissociative disorders had the highest scores for depersonalization/derealization (see Table S1 in the online data supplement for details).
Heterogeneity, as assessed by I
2, was >70% in all analyses, except for somatic symptom disorders (I
2=16.1%); the highest heterogeneity was observed in gambling disorders (I
2=98.5%) (see
Table 1).
Discussion
This is the second meta-analysis of dissociation scores in a broad variety of psychiatric disorders. While the first meta-analysis was published more than 20 years ago (
43) and comprised 85 individual studies, our meta-analysis reports on 216 individual studies with more than 15,000 individuals with mental disorders. The largest dissociation scores were found for dissociative disorders, followed by PTSD, borderline personality disorder, and conversion disorder, and the lower range of scores included substance-related and addictive disorders, feeding and eating disorders, schizophrenia, anxiety disorder, OCD, and affective disorders.
Our data confirm some but not all of the results reported in the earlier meta-analysis. Confirming results were found regarding dissociative disorders showing the highest overall dissociation scores. In their analysis, van Ijzendoorn and Schüngel (
43) reported mean dissociation scores of 45.6 for multiple personality disorder (now called dissociative identity disorder), 41.1 for unspecified dissociative disorders, and 35.3 for the category of dissociative disorder not otherwise specified. In our study we differentiated between dissociative identity disorder and depersonalization/derealization disorder, as listed in DSM-5 (
1). Although the existence of dissociative identity disorder has been discussed controversially (e.g.,
61), our result of a mean dissociation score of 48.7 in a total of 1,313 individuals with this diagnosis indicates very high levels of dissociative experience in this diagnostic group. Interestingly, depersonalization/derealization disorder yielded numerically lower DES scores than PTSD and borderline personality disorder. This may be due to the fact that depersonalization/derealization disorder does not cover the entire spectrum of dissociative symptoms, therefore leading to lower overall dissociation scores.
Dissociation scores in PTSD and schizophrenia in our analysis were close to those reported by van Ijzendoorn and Schüngel (
43), although their study included only one study on schizophrenia (compared with 17 here) and four studies on PTSD (compared with 33 here). Scores in PTSD were the second highest in our analysis, reflecting the importance of dissociation in relation to PTSD (
62), for which a dissociative subtype was introduced in DSM-5. Although dissociative symptoms are less pronounced in schizophrenia (mean score, 17.8), they have been studied intensively in this disorder because of similarities in the description of dissociative phenomena and psychotic symptoms (
63). Empirical studies have yielded varying correlations between schizophrenia and different aspects of dissociation, with depersonalization/derealization showing the strongest relation (
5). Several authors have emphasized the relevance of depersonalization as a mediator between childhood trauma and hallucinatory experiences, thus acting as a risk factor for schizophrenia (e.g.,
64). It is hypothesized that depersonalization may facilitate a person’s attribution of their own thoughts to external sources (
65), and a trauma-dissociation subgroup within schizophrenia has been proposed (
66).
Our data differ from the earlier meta-analysis (
43) with respect to eating disorders, anxiety disorders, and affective disorders. While eating disorders and anxiety disorders show considerably higher mean dissociation scores in our analysis than in the earlier one (18.6 compared with 14.5 for eating disorders; 15.2 compared with 10.2 for anxiety disorders), we found lower scores for affective disorders (15.3 for depressive disorders and 14.8 for bipolar disorders compared with 19.4 for affective disorders in the earlier analysis). Recent research points to differential relations between dissociation and symptoms of these disorders. In anorexia nervosa, where the highest mean dissociation scores were found (mean score, 24.1), symptoms of depersonalization in the form of body schema distortions have been investigated (
67). In bulimia nervosa, dissociative qualities of amnesia, timelessness, and involuntariness seem to play a role in bingeing behavior and severity (
6,
68). In anxiety disorders, experiences of depersonalization/derealization have often been described in relationship with panic attacks, although the sequence of incidence is not clear: dissociation might trigger panic attacks—for example, via the fear of losing control—but concomitant symptoms of panic attacks, such as hyperarousal or hyperarousal, might also produce dissociation (
29). In depressive disorders, the research on mechanisms of dissociation is impeded by a strong overlap between depressive symptoms such as emotional numbing, feelings of detachment, and restricted emotional responsiveness (
69), as well as by shared covariates, such as sleep quality and distortions in autobiographic memory (
23,
70).
Our analysis is the first to report systematically retrieved mean dissociation scores for borderline personality disorder, somatic symptom disorder, conversion disorder, substance-related and addictive disorders, and OCD. Borderline personality disorder showed dissociation scores similar to those of PTSD in 27 studies (mean score, 27.9). Furthermore, our study confirmed the significance of dissociative symptoms in borderline personality disorder, which has been acknowledged by adding dissociative experiences as part of one of the nine criteria for borderline personality disorder in DSM-IV (
71). Although classified as a personality disorder, borderline personality disorder is closely associated with traumatic stress. Rates of adverse childhood experiences have been consistently demonstrated to be higher than 50% (
72). Independent of trauma experience and comorbid diagnoses, almost all patients with borderline personality disorder report identity confusion, unexplained mood changes, and depersonalization (
73).
“Somatic symptom and related disorders” is a new category in DSM-5 (
1) and comprises a broad spectrum of disorders, including somatic symptom disorder (formerly known as somatoform disorders), illness anxiety disorders, conversion disorder (functional neurological symptom disorder), and factitious disorder. Notably, conversion disorder is part of the dissociative spectrum in ICD-10 (
52), and dissociation scores were in a range similar to those of other dissociative and trauma-related disorders in our meta-analysis.
The high mean dissociation scores for addictive disorders—19.9 for gambling disorder, 19.7 for alcohol use disorder, and 17.7 for other substance-related disorders—may be partly related to comorbidities with PTSD, borderline personality disorder, and dissociative disorders (
74–
76). General findings regarding the link between dissociation and substance abuse have been inconsistent but suggest lower scores in samples without comorbid disorders (
77–
79). The mean dissociation score of 15.3 for OCD falls within the lower range of dissociative symptoms. Nevertheless, dissociation has gained increasing attention in this area of research. On a symptomatic level, dissociative amnesia has been related to checking compulsion (
80). This effect does not seem to be linked to poorer memory or reality monitoring performance but rather to a reduced confidence in these abilities (
81).
Recent population-based studies show mean dissociation scores in the general population of 8 in a Finnish sample (N=2,001) (
82) and 10 in a Portuguese sample (N=224) (
83). In their meta-analysis, van Ijzendoorn and Schüngel (
43) report a mean score of 11.6 for healthy subjects. Those numbers appear to be considerably lower than all mean dissociation scores calculated for mental disorders in our analysis. Van Ijzendoorn and Schüngel’s conclusion that “the means for normals and nondissociative patients were remarkably similar” (
43, p. 372) does not seem to be supported by our results. The variety of mental disorders ranging between 15 and 25 in dissociation scores clearly speaks for dissociative experience as an unspecific and ubiquitous psychopathological phenomenon. From a clinical perspective, this finding underlines the importance of careful evaluation of dissociative symptoms, and not only in patients with dissociative or trauma-related disorders.
Our study has several limitations. Although the overall number of included studies was quite large, the number of studies and subjects per diagnostic category varied substantially. We only included categories with at least four individual studies, but categories varied between four and 66 studies and between 187 and 2,860 subjects. As we had no a priori hypothesis to explain heterogeneity, we did not carry out subgroup analysis. Heterogeneity may be rooted in different factors, including heterogeneity of the diagnostic entity, diagnostic shifts over time, and differences between individual studies, for example, with respect to diagnostic procedures, gender distribution, and the countries of origin. Most likely, comorbidity and trauma experiences also influence dissociation scores and should be systematically considered in future studies. Finally, we note that the DES is a self-rating instrument and that certain dissociative features may be over- or underrepresented in comparison to observer-based ratings (
84).
In summary, our meta-analysis confirms the prevalence of dissociative symptoms not only in dissociative disorders, posttraumatic stress disorder, and borderline personality disorder, but in nearly all mental disorders. Research on the distinct diagnostic categories suggests a variety of mechanisms linking dissociative experiences to a higher burden of illness and detrimental effects on treatment. An evaluation of dissociation should therefore be part of every careful psychopathological assessment, and future studies should engage a transdiagnostic perspective to enhance the development of treatment modules to deal with dissociative symptoms.