Differential constellations of dissociative symptoms and their association with childhood trauma – a latent profile analysis

ABSTRACT

Background: While several studies documented a positive correlation between childhood maltreatment severity and dissociation severity, it is currently unknown whether specific dissociative symptoms cluster together among individuals with childhood trauma histories ranging from none to severe.

Objective: We aimed to explore symptom constellations across the whole spectrum of dissociative processing from patients with severe dissociative disorders to healthy controls and relate these to maltreatment severity and sociodemographic characteristics.

Methods: We employed latent profile analysis to explore symptom profiles based on five subscales, measuring absorption, depersonalization, derealization, somatoform and identity alteration, based on the 20 items of the German short version of the Dissociative Experiences Scale-II (Fragebogen zu Dissoziativen Symptomen-20) in a large aggregate sample (n = 3,128) overrepresenting patients with trauma-related disorders. We then related these profiles to maltreatment severity as measured by the five subscales of the Childhood Trauma Questionnaire as well as sociodemographic characteristics.

Results: Based on the five FDS subscales, six clusters differentiated by symptom severity, but not symptom constellations, were identified. Somatoform dissociation varied in accordance with the remaining symptom clusters. The cluster with the highest overall symptom severity entailed nearly all subjects diagnosed with Dissociative Identity Disorder and was characterized by extreme levels of childhood maltreatment. Both abuse and neglect were predictive of cluster membership throughout.

Conclusions: The higher the severity of dissociative processing in a cluster, the more subjects reported high severity and multiplicity of childhood maltreatment. However, some subjects remain resilient to the development of dissociative processing although they experience extreme childhood maltreatment.

HIGHLIGHTS

• Dissociative symptoms, including identity alterations, are closely related to the severity of experienced childhood abuse.

• Somatoform dissociation occurs on all levels of overall dissociation severity.

• Some subjects with a history extreme childhood maltreatment do not develop dissociative symptoms, while some subjects with extreme dissociative symptoms do not report any childhood maltreatment.

Antecedentes: Aunque varios estudios han documentado una correlación positiva entre la severidad del maltrato infantil y la severidad de la disociación, actualmente se desconoce si los síntomas disociativos específicos se agrupan entre individuos con historias de trauma infantil que van desde ninguno a severo.

Objetivo: Nos propusimos explorar las constelaciones de síntomas en todo el espectro del procesamiento disociativo de pacientes con trastornos disociativos graves y controles sanos, y relacionarlas con la gravedad del maltrato y las características sociodemográficas.

Métodos: Se empleó el análisis de perfil latente para explorar los perfiles de síntomas basados en cinco subescalas, que miden la absorción, despersonalización, desrealización, somatomorfos y alteración de la identidad, basados en los 20 ítems de la versión corta alemana de la Escala de Experiencias Disociativas-II (Fragebogen zu Dissoziativen Symptomen-20) en una gran muestra agregada (n = 3.128) sobrerrepresentada por pacientes con trastornos relacionados con el trauma. Luego relacionamos estos perfiles con la gravedad del maltrato, medida por las cinco subescalas del Cuestionario de Trauma Infantil, así como con las características sociodemográficas.

Resultados: Sobre la base de las cinco subescalas de la FDS, se identificaron seis clústeres diferenciados por la gravedad de los síntomas, pero no por constelaciones de síntomas. La disociación somatomorfa varió de acuerdo con los grupos de síntomas restantes. El clúster con la mayor gravedad sintomática global incluía a casi todos los sujetos diagnosticados con Trastorno de Identidad Disociativo y se caracterizaba por niveles extremos de maltrato en la infancia. Tanto el maltrato como la negligencia fueron predictivos de la pertenencia al clúster en todo momento.

Conclusiones: Cuanto mayor era la severidad del procesamiento disociativo en un clúster, más sujetos reportaban una alta severidad y multiplicidad de maltrato infantil. Sin embargo, algunos sujetos permanecen resilientes al desarrollo del procesamiento disociativo, aunque experimenten maltrato infantil extremo.

KEYWORDS:

• Dissociation
• childhood maltreatment
• amnesia
• depersonalization
• derealization
• dissociative identity disorder

PALABRAS CLAVE:

• Disociación
• maltrato infantil
• amnesia
• despersonalización
• desrealización
• trastorno de identidad disociativo

1. Introduction

Dissociative experiences occur in the majority of mental disorders (Lyssenko et al., 2018) and can best be conceptualized as a transdiagnostic phenomenon. However, just like most mental disorders (Carr et al., 2013; Dvir et al., 2014), the frequency of their occurrence is associated with the extent to which individuals have suffered traumatic events, especially in childhood (Rafiq et al., 2018; Vonderlin et al., 2018). Convergently, the average dissociation severity per person, as measured with the most commonly employed assessment instrument for dissociation Dissociative Experiences Scale (DES-II; Carlson & Putnam, 1993), is found to be highest in patients suffering from trauma-related disorders, such as posttraumatic stress disorder (PTSD; see White et al., 2022 for meta-analytic evidence for the dissociative subtype, e.g. as identified by latent class analyses), borderline personality disorder (BPD) and dissociative disorders (DD), especially dissociative identity disorder (DID; Lyssenko et al., 2018). Because most studies exclusively employed the DES-II total severity score to investigate the association with traumatic experiences, little attention has been paid to whether specific dissociation symptom constellations are associated with specific forms of childhood trauma.

A recent meta-analysis indicated that all forms of childhood trauma, including both accidental trauma and interpersonal trauma such as neglect and abuse experiences, are associated with overall dissociation severity among patients with severe mental disorders (Rafiq et al., 2018). Similarly, another meta-analysis (Vonderlin et al., 2018) indicated that both abuse and neglect experiences are associated with overall dissociation in the general population and in patients with less severe mental disorders, with strongest associations for sexual and physical abuse. In addition, earlier age of onset, longer duration of abuse, and abuse committed by a parent were associated with substantially higher dissociation scores. Multiplicity of childhood trauma, i.e. exposure to more than one form of maltreatment, has equally been associated with increased dissociation severity (Schalinski et al., 2016).

However, associations between childhood trauma experiences and the DES-II total score might not be very elucidating, given that this scale should not be assumed to be unidimensional: most studies indicated a two-factor, three-factor, or four-factor structure both within and across different languages and Western cultures (Amdur & Liberzon, 1996; Darves-Bornoz et al., 1999; Espirito Santo & Abreu, 2009; Larøi et al., 2013; Lipsanen et al., 2003; Mazzotti et al., 2016; Ross et al., 1991; Ruiz et al., 2008; Schimmenti, 2016; Soffer-Dudek et al., 2015; Stockdale et al., 2002). Doubts regarding the unidimensional nature of dissociative experiences as measured with the DES-II are further supported by a correlation network analysis, which yielded a representation as a network of several layers of dissociative symptoms that interact with each other (Schimmenti & Sar, 2019). In addition, somatoform symptoms of dissociation, which are not covered by the DES-II, have also been associated with childhood trauma (Cheung et al., 2023; Maaranen et al., 2004; Mueller-Pfeiffer et al., 2010).

Given the evidence that dissociative experiences might best not be conceptualized as a unidimensional construct, it is of interest to know whether specific subsets of dissociation symptoms preferentially appear together, yielding differential symptom profiles that individuals may show, and whether these subsets show a particularly strong association with specific childhood trauma experiences. The fact that this is currently unknown is surprising in itself. Dissociative symptoms in the realm of identity fragmentation and the ensuing dissociative amnesia are thought to be limited to DID and dissociative disorders not otherwise specified, which in turn are thought to stem specifically from very severe childhood abuse, especially sexual abuse (Putnam et al., 1986; Ross, 1991; Schroder et al., 2018; Wabnitz et al., 2013). To our knowledge, only one previous study (Millman et al., 2021) studied such profiles of dissociative symptoms, but this study exclusively included subjects diagnosed with depersonalization-derealization disorder. In this study, subscale scores on three different dissociation questionnaires as well as an anxiety questionnaire were entered into a latent class analysis. Besides three classes differentiated solely by severity, two classes emerged that were similar in overall severity, but differed regarding their symptom constellation, with one class scoring much higher on symptoms of identity fragmentation such as amnesia, and the other class scoring much higher on somatoform symptoms. However, as depersonalization-derealization disorder is the one dissociative disorder that frequently shows very weak associations with childhood trauma (Baker et al., 2003; Michal & Beutel, 2009; Sierk et al., 2018), this class solution might not generalize to other patient populations.

1.1. Aims of the study

As it is currently unknown whether specific dissociative symptoms cluster together in individuals with differential childhood trauma histories, we opted to employ latent profile analysis (see e.g. Oberski, 2016) to explore symptom profiles based on five FDS subscales, measuring absorption, depersonalization, derealization, somatoform dissociation as well as identity alteration based on the 20 items of the German short version of the Dissociative Experiences Scale-II (Fragebogen zu Dissoziativen Symptomen-20) across a large aggregate sample representing the whole spectrum from patients with severe mental disorders to healthy controls. A latent profile analysis reveals clusters of individuals with distinct symptom profiles. In addition, we hypothesized that dissociative symptoms related to identity fragmentation would be associated with (1) more severe childhood abuse, especially sexual and physical abuse, than symptoms of absorption, depersonalization, or derealization and (2) a higher prevalence of trauma-related disorders, especially dissociative disorders (Rodewald et al., 2011).

2. Methods

We present aggregated data (n = 3,128) derived from six separate samples of German-speaking participants. Two samples (Samples 1 & 2) were assessed anonymously online and four samples consisted of patients treated for mental disorders assessed in their respective treatment facilities (Samples 3–6). These samples were selected to ensure (1) a broad representation across the psychopathology continuum including healthy subjects as well as (2) an oversampling of patients with trauma-related and dissociative disorders for sufficient statistical power to detect subgroup-specific symptom patterns.

We included one sample that was population-representative with regard to sex, age, and education (Sample 1: n = 573). The majority of this sample (n = 463) was never diagnosed with a mental disorder. Three samples consisted predominantly of patients treated for disorders other than trauma-related disorders (Sample 4: n = 208; Sample 5: n = 1602; Sample 6: n = 532). Two samples predominantly consisted of patients treated for trauma-related disorders, including dissociative disorders (Sample 2: n = 116, Sample 3: n = 97).

Each study contained several additional instruments, which are not covered here. Participants provided sociodemographic information regarding their age, gender, level of education, and current employment status. Clinical diagnoses of mental disorders were either reported by the participants themselves (Samples 1 & 2) or by their clinicians (Samples 3–6). As such, the clinical diagnoses were not independently confirmed by the researchers. Sociodemographic data were assessed in slightly different ways in the six studies and were harmonized post-hoc for their joint analysis. For details on the samples and harmonization, we refer to the supplementary material (SM1, Table S1, and SM2). No values were imputed. Studies were approved by local ethics boards (see supplement for details) and participants were informed about the proceedings of the respective study and consented to study participation.

2.1. Assessment instruments

The Fragebogen zu Dissoziativen Symptomen (FDS-20) was employed to assess dissociation severity (Spitzer et al., 2004). This German questionnaire consists of 20 items, of which 15 were translated from the Dissociative Experiences Scale-II (DES-II, Carlson & Putnam, 1993). In its development, 16 items covering somatoform dissociation were added to the full DES-II. Subsequently, the short-form instrument with 20 items was developed, which contains five somatoform dissociation items not included in the original DES-II scale (Spitzer et al., 2015). These five items were treated as one subscale. In accordance with expert ratings (provided in Mazzotti et al., 2016) and the classification in the Structured Clinical Interview for Dissociative Disorders (Steinberg et al., 1991), the remaining 15 items were structured into four subscales (see Supplement): absorption, depersonalization, derealization, and identity alterations (as a measure of identity fragmentation). All items were rated on an 11-point scale from 0 percent (never) to 100 percent (always), indicating how often the symptom occurred. The five subscale scores were computed as the mean of the respective items . The total questionnaire showed good internal consistency in the current sample (α = .94). The subscales all showed acceptable/good internal consistency (Absorption: α = .77; Depersonalization: α = .83; Derealization: α = .77; Somatoform: α = .85, Identity: α = .76). The latent profile analysis was executed on the five subscales and a total mean score was computed for the sample characterization.

The Childhood Trauma Questionnaire (CTQ; Bernstein et al., 2003; German version by Wingenfeld et al., 2010) comprises 25 items, which assess the frequency of instances of emotional abuse (EA), physical abuse (PA), sexual abuse (SA), emotional neglect (EN), and physical neglect (PN) in childhood on a 5-point Likert scale ranging from 1 ‘Never true’ to 5 ‘Very often true’. (An additional three items assess denial, and since they are beyond our research interests these are excluded from the analyses here.) Summing the five items per subscale yields a subscale score (possible score range 5–25), indicating the severity of the form of maltreatment involved. The sum of the five subscales gives the CTQ total score, providing an overall indication of severity of maltreatment. The subscales and total scale were computed for the sample characterization. Subscale scores were unavailable for one participant on the emotional abuse subscale and for two participants on the sexual abuse subscale. The subscales have shown good internal consistency with exception of the subscale ‘physical neglect’ (Wingenfeld et al., 2010), which was replicated with the current data (PN: α = .49, all other subscales between α = .90 – .97; Total scale α = .87; based on n = 1,526, no data available for Sample 5). For the interpretation of differences between the clusters, we employed the established cut-off scores (Walker et al., 1999; see supplement for details) for the ‘clinical range’. To further characterize the clusters we computed multiplicity scores by summing the number of subscales on which an individual scored above the respective cut-off.

2.2. Participants

Across the six samples, participants were on average 38.42 years old (SD: 13.20; Range: 17–76) and predominantly female (2,038 females (65.2%), 1,088 males (34.8%), no data available for 2 subjects; see Table S1 for sample-specific data). The largest subgroup of participants were single at the time of assessment (n = 1469, 47.2%), while n = 1130 subjects (36.3%) were in a committed relationship, n = 376 (12.1%) were separated or divorced, and n = 134 (4.3%) widowed. For n = 18 individuals (0.6%) no information regarding their relationship status was available. On average, participants reported having been diagnosed with 1.55 mental disorders (SD: 1.26; Range: 0–9). Of the full sample, n = 322 (10.4%) individuals reported suffering from PTSD, n = 611 (19.5%) from borderline personality disorder (BPD), and n = 133 (5.1%) from a dissociative disorder, of which n = 70 (2.2%) specified having been diagnosed with dissociative identity disorder (DID). At the time of assessment, 1,637 subjects (76.3%) stated that they were fit to work, while 509 subjects were declared unfit (23.7%; information not available for 982 subjects).

2.3. Statistical analyses

Descriptive statistics were computed for the total sum score on the FDS-20, and the five subscales of the CTQ. Bivariate correlations between the FDS subscales absorption, depersonalization, derealization, and identity alterations and each of the CTQ subscales emotional abuse (EA), physical abuse (PA), and sexual abuse (SA) were calculated.

To identify groups of individuals that show similar symptom constellations on the FDS-20 and relate them to the different forms of childhood maltreatment as measured by the CTQ and individual characteristics, latent profile analysis (LPA; Becker & Yang, 1998; McCutcheon, 1987) using a three-step approach (Vermunt, 2010) were employed. The LPA was performed with Latent GOLD 6.0 (Vermunt & Magidson, 2021). Step 1 involved identifying the (latent) clusters underlying the five FDS subscales, while Step 2 and 3 involved relating the identified clusters to the CTQ subscales and individual characteristics.

In Step 1, the scores on the five FDS subscales were modelled with a multivariate normal mixture model. Herewith, it is assumed that each individual belongs to a cluster and that the scores on the five FDS subscales are multivariate normally distributed per cluster. The latter means that each cluster is fully described by the cluster means of the five FDS subscales, and their variance-covariance matrix. Two types of models were fit, namely one with a diagonal variance-covariance matrix (i.e. assuming uncorrelated FDS subscales per cluster), and one with a free diagonal variance-covariance matrix per cluster. For both types, 15 models with number of clusters from one till 15 were fitted, using 500 starts to reduce the probability of ending in a local minimum. The number of clusters were selected using the Bootstrap Likelihood Ratio Test (BLRT), which compares model fit between k-1 and k cluster models. If the BLRT indicated 15 clusters or more, the model with the lowest Bayesian Information Criterion (BIC) would be selected. If this would be the model with 15 clusters, then a scree plot of the BICs would be inspected to select the model with relatively low BIC with a limited number of clusters. For the type of model given a number of clusters, the one with the lowest BIC was selected. The BLRT and BIC proved to be the best indicators for the number of clusters in an LPA (Nylund et al., 2007).

In Step 2, the posterior probability for each individual belonging to each cluster was estimated, which allows for a more precise estimation of the relationship between the cluster membership and the predictors as it takes the uncertainty regarding the cluster membership into account. In Step 3, these posterior probabilities were related to several potential predictors: the dichotomized CTQ subscale scores and the multiplicity score based on the clinical cut-offs as well as individual characteristics that we presume to be meaningfully related to the latent cluster, namely sex and clinical diagnoses of trauma-related disorders (i.e. BPD, DID, PTSD). This allowed us to test whether the percentage of individuals diagnosed with DID is higher in the cluster(s) characterized by severe identity alteration than the remaining clusters (Hypothesis 2). We used a significance level of p < .05 / k(k-1)/2 (with k being the number of selected clusters) to correct for multiple testing per predictor throughout. To assess Hypothesis 1, we tested whether each of the three CTQ abuse subscales correlates stronger with the FDS subscale identity alterations than with each of the three the subscales absorption, depersonalization, or derealization, using nine one-sided Z-tests for the difference between two correlations (correcting for multiple hypothesis testing by taking a significance level of p < .05/9 = .006).

3. Results

3.1. Descriptive statistics

Regarding dissociation severity, the FDS-20 mean scale score showed a mean of 16.81 (SD = 17.05, Range: 0–100). The mean FDS-20 correlated significantly with the CTQ total score (r = .41, p < .001). The majority of the participants (n = 2,149, 68.7%) scored above the cut-off on at least one subscale of the CTQ, indicating that they were exposed to some type of childhood maltreatment (for sample-specific data please see Table S1). The mean CTQ total score was M = 53.11 (SD = 19.93, Range: 25–125; Subscale scores: M_EA = 12.18, SD_EA = 6.25; M_PA = 8.29, SD_PA = 4.85; M_SA = 7.99, SD_SA = 5.61; M_EN = 14.68, SD_EN = 5.99; M_PN = 9.97, SD_PN = 3.82).

3.2. Clusters derived from FDS subscales

In Step 1 of the LPA, we identified the clusters, that is the groups of individuals that show similar symptom constellations on the FDS-20. The BLRT and BIC indicated 15 or more than 15 clusters. The scree plot of the BIC values indicated six clusters for both model types, where the model with a free diagonal variance-covariance matrix per cluster had the lowest BIC (see Table S2). This model was thus selected. For each cluster, the cluster size, the mean FDS subscale scores (see Table 1) and the within-cluster variance-covariances (see Table S3) were estimated. The cluster sizes differ only slightly between the six clusters, with the percentage of individuals within each cluster ranging from 13.4% to 19.7%. For ease of reading, we present the clusters from most severe to least severe dissociation (see Figure 1), as indicated by the estimated FDS subscale scores (i.e. Cluster 5, 4, 1, 3, 6, 2, respectively). We included statistical comparisons to the next lower cluster in the text, the remaining statistics can be found in Table S4.

Figure 1. Means (with 95% confidence intervals) per Cluster and Subscale.

Table 1. Per cluster, the means of the FDS subscales, and the proportions for predictor variables.

	1	2	3	4	5	6	M
Cluster size (%)	16.2	17.2	16.8	16.7	19.7	13.4
Absorption M (SD)	24.08 (12.88)	3.86 (4.57)	12.35 (9.39)	26.93 (14.55)	48.81 (19.01)	9.58 (6.75)	20.93
Depersonalization M (SD)	9.43 (6.76)	0.00 (0.37)	0.00 (0.38)	23.80 (13.38)	47.21 (22.87)	5.52 (3.16)	14.33
Derealization M (SD)	10.28 (8.19)	0.00 (0.36)	4.16 (4.58)	24.61 (14.49)	44.69 (22.09)	4.88 (4.87)	14.77
Somatoform M (SD)	15.13 (11.24)	0.54 (1.16)	5.47 (5.93)	13.64 (10.04)	38.74 (22.07)	3.51 (3.51)	12.84
Identity Alteration M (SD)	17.07 (11.81)	1.33 (2.10)	6.77 (6.34)	15.90 (10.78)	41.37 (21.77)	4.54 (4.06)	14.50
FDS Scale Score M (SD)	14.45 (5.25)	1.12 (1.18)	5.68 (2.87)	21.60 (7.64)	45.74 (14.45)	5.51 (2.84)	16.81
Sex (% female)	66.70	55.00	57.60	70.50	76.50	63.70	65.20
PTSD (% with diagnosis) within cluster	9.62	4.29	4.35	11.19	23.47	5.81
PTSD (% with diagnosis) across sample	15.13	7.17	7.08	18.17	44.92	7.53	10.30
BPD (% with diagnosis) within cluster	16.38	6.53	8.31	34.48	35.27	12.35
BPD (% with diagnosis) across sample	13.58	5.76	7.12	29.51	35.58	8.44	19.50
DID (% with diagnosis) within cluster	0.21	0.00	0.16	0.05	10.71	0.52
DID (% with diagnosis) across sample	1.48	0.00	1.18	0.37	93.89	3.09	2.20
CTQ_EA (% above cut-off)	59.70	32.34	44.85	69.47	80.07	46.35	56.37
CTQ_PA (% above cut-off)	36.74	23.73	27.29	39.95	59.04	24.97	36.28
CTQ_SA (% above cut-off)	25.77	14.51	14.78	31.68	53.14	18.12	27.35
CTQ_EN (% above cut-off)	48.15	40.84	44.85	55.75	67.97	43.51	50.90
CTQ_PN (% above cut-off)	64.42	59.51	61.05	74.73	85.16	59.79	68.20
No subscale above cutoff (%)	17.94	28.40	23.93	11.88	4.60	23.75	17.88
Multiplicity (2 or more subscales above cut-off (%))	66.84	51.46	58.02	74.81	88.00	56.26	66.84
Multiplicity (all 5 subscales above cut-off) %	10.32	5.24	5.90	13.89	29.48	6.12	10.32
Multiplicity M (SD)	2.34 (1.62)	1.71 (1.51)	1.93 (1.52)	2.70 (1.57)	3.47 (1.51)	1.93 (1.55)	2.39

Cluster 5, comprising 19.7% of the sample, shows extremely frequent dissociation across the board (see Figure 1). That is, this cluster reaches the highest mean score on every single FDS subscale (vs. Cluster 4: all p < .001; Cohen’s d: 1.08–1.48), with an estimated mean FDS scale score of M = 45.7. Therefore, we denote this cluster as ‘Extreme Dissociation’.

Cluster 4 is characterized by frequent dissociation, with an emphasis on derealization and depersonalization (vs. Cluster 1: both p < .001; Cohen’s d: 1.22–1.36). The mean values for the subscales measuring absorption, somatoform dissociation and identity alteration, on the other hand, do not significantly differ from the next lower cluster (Cluster 1; Wald(1) = 2.95, p = .085, Cohen’s d: 0.21; Wald(1) = 0.98, p = .320, Cohen’s d: −0.14; Wald(1) = 0.49, p = .480, Cohen’s d: −0.10, respectively). Therefore, we denote this cluster as ‘Frequent Dissociation’.

Cluster 1 exhibits elevated dissociation scores with significantly higher mean values for all five subscales than the three remaining clusters (all p < .001, Cohen’s d: 0.74–2.09). Therefore, we denote this cluster as ‘Elevated Dissociation’.

The three remaining clusters (3,6, and 2) all fall into the low dissociation range with mean FDS scale scores below M = 6.0 (see Figure 1). Nevertheless, they differ significantly from each other on all subscales (all p < .001, Cohen’s d: 0.34–2.46), with the exception of Cluster 3 and Cluster 6 not showing significantly different levels of derealization (Wald(1) = 3.46, p = .063, Cohen’s d: −0.15) and Cluster 3 and Cluster 2 not differing significantly regarding depersonalization severity (Wald(1) < 0.01, p = 1.0, Cohen’s d: 0). We denote these clusters as ‘Low Dissociators’.

3.3. Relating the latent clusters to the CTQ subscale scores and individual demographic and clinical characteristics

The individual posterior probabilities of cluster membership were related to the CTQ scores above cut-off, and the demographic and clinical characteristic (see Table 1), and their differences between all clusters were tested pairwise (see Table S4). Based on these results we characterize the clusters as follows:

The Extreme Dissociation cluster (Cluster 5) with the highest severity of identity alteration also entails the highest percentage of participants who were formally diagnosed with DID (10.7%, p < .001, supporting Hypothesis 2). Across the whole sample, nearly all subjects with a diagnosis of DID (93.8%) are allocated to this cluster. This Extreme Dissociation cluster also contains the highest percentage of subjects diagnosed with PTSD (23.5%; p < .001) and nearly half of all subjects with a diagnosis of PTDS (44.9%) are in this cluster. In addition, a higher percentage of individuals diagnosed with BPD were allocated to this cluster (35.3%) than all other clusters with the exception of Cluster 4 (Wald (1) = 0.05, p = .82). This cluster also has the highest percentage of female participants (77.1%; significantly higher than all other clusters with the exception of Cluster 4; Wald(1) = 5.03, p = .025).

Regarding childhood maltreatment, the Extreme Dissociation cluster has the highest percentage of subjects scoring above cut-off on every single CTQ subscale (ranging from 53.1% to 85.2%, all p < .001). Consequently, this cluster also exhibits the highest level of multiplicity (88%; all p < .001), with 95.4% scoring in the clinical range for more than one subscale and 29.5% scoring in the clinical range on all five subscales concomitantly (see Table 1). On average, subjects in this cluster reported values above the cut-off on 3.5 of the five CTQ subscales.

However, across all subjects, the FDS subscale identity alteration did not exhibit a closer association with the severity of abuse than the FDS subscales absorption, depersonalization, or derealization. All bivariate correlations were between r = .216 and r = .388 (see Table S5 for details). The pairwise comparisons of the correlation coefficients between the CTQ abuse subscales and the respective FDS subscales did not indicate any significant differences (all tests p > .02, see Table S5, not supporting Hypothesis 1). A small minority (4.6%) of subjects in the Extreme Dissociation cluster did not report any significant childhood maltreatment, meaning they scored below the clinical cut-off on all five subscales.

The Frequent Dissociation cluster (Cluster 4) comprises significantly more subjects diagnosed with BPD (vs. Cluster 1, Wald(1) = 23.28, p > .001): across the whole sample, nearly a third of the individuals diagnosed with BPD are in this cluster (29.5%). However, it does not entail more subjects diagnosed with DID (vs. Cluster 1, Wald(1) = 0.69, p = .40) or PTSD (vs. Cluster 1, Wald(1) = =.36, p = .54) than the next cluster. The Frequent Dissociation cluster is further characterized by very frequent childhood maltreatment, exhibiting comparable values as the next cluster (see Table S4 for details). Only 11.9% did not score above the clinical cut-off on any of the CTQ subscales and multiplicity was high with subjects in this cluster scoring on average on 2.7 subscales above cut-off.

The Elevated Dissociation cluster (Cluster 1) entails more individuals diagnosed with BPD (vs. Cluster 3, Wald(1) = 11.45, p < .001), but is comparable to the next cluster with respect to prevalences for DID (Wald(1) = 0.01, p = .890) and PTSD (Wald(1) = 8.15, p = .004). The individuals allocated to this cluster reported significantly more frequent emotional and sexual abuse in the clinical range (both p < .001), but not neglect (both p > .320) or physical abuse (p = .004) than the next cluster (see Table S4 for details).

The three Low Dissociator clusters (Clusters 3, 6, 2) also exhibit the lowest prevalence rates for all three trauma-related disorders, with comparable percentages of individuals diagnosed with PTSD in all three clusters (all p > .050) and slightly more individuals diagnosed with DID and BPD in Cluster 6 compared to Cluster 2 (Wald(1) = 27.17, p < .001 and Wald(1) = 8.62, p = .003, respectively). Regarding childhood maltreatment, the proportions of individuals scoring above the clinical cut-offs did not differ significantly between these three clusters for all maltreatment types (all p > .150), with the exception of emotional abuse (Cluster 2 exhibits significantly lower values than the other two clusters, both p < .001, see Table S4). These clusters still each entailed a sizeable subgroup reporting maltreatment above the threshold (lowest percentages in Cluster 2: emotional (32.7%), physical (24.0%) and sexual (14.7%) abuse, emotional (41.2%) and physical (59.7%) neglect), indicating that individuals resilient to the development of dissociation are allocated to all three clusters. Overall, multiplicity was comparable between these three clusters (all p > .022) with averages ranging from 1.7–1.9 subscales above cut-off.

4. Discussion

The current analysis provides data on dissociative symptoms in a diverse sample consisting of subjects never diagnosed with a mental disorder as well as in- and outpatients treated for different mental disorders. The sampling successfully overrepresented patients suffering from trauma-related disorders (e.g. 10.4% were diagnosed with PTSD instead of the expected 2.3%; Jacobi et al., 2014) and subjects exposed to interpersonal childhood trauma (CTQ sum score M = 53.1 instead of the population-wide expected M = 36.0; Iffland et al., 2013). The majority of the participants (68.7%) scored above the cut-off on at least one subscale of the CTQ as compared to 33.9% in the population-representative sample. Convergently, average dissociation scores were also substantially higher than in the general population and in patients with mental disorders in general, but lower than in patients with trauma-related disorders (Spitzer et al., 2015). Clusters mostly differed regarding the overall severity of dissociation, with few distinctive symptom constellations emerging:

4.1. Identity fragmentation

The overwhelming majority (93.8%) of subjects diagnosed with DID was allocated to the Extreme Dissociation cluster. Convergently, this cluster also scored significantly higher on the identity alteration subscale than all other clusters, with the Frequent Dissociation cluster only showing somewhat elevated, but not high levels of identity alteration, supporting our second hypothesis. This is in line with the result of a recent meta-analysis indicating that symptom severity of identity alterations differentiates well between subjects with and without a dissociative disorder (Mychailyszyn et al., 2021). As already reported by previous studies, the occurrence of severe symptoms of identity alteration was associated with extreme childhood maltreatment and very high levels of multiplicity, that is the co-occurrence of more than one form of abuse and neglect (Dorahy et al., 2014; Webermann et al., 2014). On average, subjects allocated to this cluster score above the cut-off on 3.5 subscales, meaning they experienced maltreatment in the form of combined abuse and neglect. However, we could not confirm a stronger association between the level of abuse (emotional, physical, or sexual) and the severity of symptoms of identity fragmentation versus symptoms of absorption, depersonalization, or derealization as stated in our Hypothesis 1.

4.2. Depersonalization and derealization

The second-most severe cluster (the Frequent Dissociation cluster) was characterized mainly by symptoms of derealization and depersonalization, but did not differ significantly from the next cluster (the Elevated Dissociation cluster) regarding absorption, somatoform dissociation and identity alteration. This cluster might contain subjects that would otherwise have been characterized as exhibiting the dissociative subtype of PTSD as identified in previous latent class/profile analyses (White et al., 2022). However, as we do not have data on this subtype differentiation, this remains speculative. It also entails a higher percentage of subjects diagnosed with BPD, which in conjunction might point in the direction of a preponderance of subjects suffering from emotion dysregulation (van Dijke et al., 2018).

As expected, the majority of individuals diagnosed with a trauma-related disorder were allocated to the two most severe dissociation clusters. But while the individuals diagnosed with DID were overwhelmingly allocated to the Extreme Dissociation cluster, the subjects diagnosed with BPD or PTSD were mainly spread over the two most severe clusters. In short, the diagnostic labels do not clearly differentiate between the clusters, which might be due to the fact that we allowed multiple diagnostic labels per person and comorbidity in this sample was high.

4.3. Somatoform dissociation

Somatoform symptoms co-occurred with the remaining dissociation symptoms on all levels of severity. This is interesting as it indicates that the occurrence of somatoform symptoms neither seems to be dependent on the overall severity of dissociation nor childhood maltreatment. As previous studies indicated that somatoform dissociation also occurs more often in several groups of patients with somatic disorders such as migraines (Schalinski et al., 2016) and fibromyalgia (Bohn et al., 2013) as well as mental disorders not typically seen as directly trauma-related such as eating disorders (Palmisano et al., 2018), the addition of items assessing somatoform dissociation to the screening instrument for dissociation seems useful. Interestingly, some preliminary evidence exists that points to a role of emotional maltreatment (Brown et al., 2005) in the advent of somatoform symptoms, especially when combined with low family functioning (Cheung et al., 2023).

4.4. Association with severity of childhood trauma

Across clusters, the severity of childhood maltreatment follows a clear pattern: more severe dissociation was associated with more severe maltreatment. Few specific associations with maltreatment types emerged, indicating that all types of maltreatment are associated with the occurrence of dissociative processing. Interestingly, the cluster exhibiting the most severe dissociation symptoms (the Extreme Dissociation cluster) also included a small minority (4.6%) of subjects who did not report childhood maltreatment severe enough to exceed any clinical cut-off. The same is true for 11.8% of the Frequent Dissociation cluster. This points in the direction of a potential etiology other than exposure to childhood abuse and neglect such as exposure to trauma types not covered by the CTQ like peer bullying, trauma exposure during adulthood, or trauma-unrelated factors. This finding should be carefully replicated in an independent sample as it might also inform the ongoing debate about other potential etiological factors, including neurobiological aberrations that might act as a diathesis (Daniels et al., 2015; Lotfinia et al., 2020; Roydeva & Reinders, 2021). In line with recent meta-analyses (Rafiq et al., 2018; Vonderlin et al., 2018), we did not identify a coherent differentiation regarding the impact of abuse versus neglect experiences. Hence, neglect can be considered to have comparable associations with dissociation severity and symptom combination as abuse.

4.5. Association with multiplicity of childhood trauma

In the current sample, multiplicity was the norm rather than the exception. That is, the clusters showed average numbers of CTQ subscales above the cut-off between 1.7 and 3.5, with the an average of 2.4 subscales in the clinical range. Multiplicity also differentiated well between the different levels of dissociation severity, with the cluster characterized by the most severe levels of dissociation also showing pervasive multiplicity of maltreatment.

4.6. Resilience regarding the development of dissociative processing

The three clusters characterized by low dissociation frequencies (Clusters 3, 6, 2) each contained at least 76.1% of subjects who reported at least one form of maltreatment. These three clusters also each contained sizeable subgroup of subjects who experienced extreme childhood maltreatment, scoring above cut-off on all five CTQ subscales (between 5.2% and 6.1%). This speaks towards the capacity to remain resilient to the development of dissociative processing even in the face of extreme maltreatment. Thus, while there is a strong association between childhood trauma and dissociation on average, this confirms that even multiplicity of maltreatment does not necessarily lead to dissociation.

4.7. Strengths and limitations

The main strength of this study is the size and composition of its sample. The overinclusion of individuals suffering from trauma-related disorders combined with healthy subjects and clinical samples with diagnoses outside the trauma spectrum provided a good basis to characterize the clusters. The main limitation of this study is the reliance on clinician- and patient-reported clinical diagnoses, which could not be confirmed independently. As such, the characterization of the clusters with regards to clinical diagnoses should be considered preliminary. In a similar vein, the cluster solution per se awaits independent replication. As all our analyses are purely based on self-report data, the typical limitations apply, including concerns of over- (Merckelbach et al. 2017) and under-reporting (Beutler et al. 2020) of dissociative symptoms. As many of the participants filled out the questionnaires unsupervised, their adherence to the instructions or understanding of the items remain unknown. In addition, the usual concerns regarding the validity of ratings for items assessing amnesia apply, as ‘amnesia for the amnesia’ is a commonly observed phenomenon in individuals suffering from severe dissociative disorders. These concerns might then also apply to the ratings of childhood maltreatment provided by subjects suffering from amnestic periods during their upbringing. In addition, the chosen questionnaire instruments come with their own limitations, such as the CTQ neither assessing victimization by peers nor more subtle forms of maltreatment by adults that might still impact the development of healthy attachment.

5. Conclusion

Across clusters, the overall severity of dissociative processing is closely associated with the severity of the reported childhood maltreatment, so that more severe maltreatment is generally associated with more severe dissociation. However, there are notable exceptions: A relevant percentage of subjects exposed to even multiple maltreatments seems to be resilient to the development of dissociative symptoms. Conversely, a small minority of subjects experiencing severe dissociation do not report clinical levels of childhood maltreatment. We believe these results, in line with previous studies, can further inform the discussion regarding putative causal mechanisms inthe development of dissociative processing given the clear dose–response relationship with childhood trauma severity evident in the data. This dose–response relationship also holds for symptoms of fragmentation typical for Dissociative Identity Disorder. It might therefore be advisable to include a measure of dissociation severity in the routine clinical assessment of clients reporting exposure to childhood trauma.

Acknowledgements

This study was supported by the EU Rosalind-Franklin Fellowship Programme to J.K. Daniels.

Disclosure statement

No potential conflict of interest was reported by the author(s).

Data availability statement

The data that support the findings of this study are not publicly available due to their containing information that could compromise the privacy of research participants, but are available from the corresponding author [JKD] upon reasonable request.