Trauma‐Related Compulsions in Obsessive‐Compulsive Disorder: Functional Insights From Factor Analysis

ABSTRACT

Resisting obsessive‐compulsive disorder (OCD) compulsions is a key predictor of exposure and response prevention (ERP) effectiveness, yet trauma‐exposed individuals with OCD endorse more severe compulsions and are more likely to be labelled ‘treatment resistant.’ Among trauma‐exposed individuals, compulsions are theorised to function to prevent future trauma and cope with trauma‐related responses. Recent findings suggest that certain compulsions in the context of trauma exposure differentially impact treatment outcomes, yet missing from extant literature is a larger empirically derived conceptualisation for how trauma may impact the underlying function of compulsions. In a sample of 662 adult patients enroled in intensive OCD treatment (40.3% trauma‐exposed), factor analyses and structural equation models provided insight into the potential functionality connecting clusters of compulsions and their impact on treatment response. A unique Absolution factor observed only in the trauma‐exposed group suggested a stronger trauma association with compulsions intended to absolve guilt, obligation, or punishment compared to individuals without trauma. Longer lengths of stay were found with sexual abuse through Cleanliness and Certainty Seeking compulsions, and bullying through Certainty Seeking compulsions. Relationships were observed between other trauma types and compulsion clusters, though these did not impact treatment response. Findings offer support that some compulsions may serve trauma‐related functions related to preventing and/or coping with trauma. Trauma‐related Cleanliness and Certainty Seeking compulsions are important targets for treatment given their propensity to impact treatment response.

1. Introduction

Obsessive‐compulsive disorder (OCD) is characterised by intrusive, disturbing, unwanted thoughts images and urges (obsessions) and repetitive, excessive behaviours or mental acts (compulsions; APA ²⁰¹³). Compulsions are considered both a hallmark symptom of OCD as well as a proposed maintenance factor of the disorder (Foa 2010). Accordingly, exposure and response prevention (ERP), the gold‐standard behavioural treatment for OCD (Hezel and Simpson 2019), emphasises the role of preventing engagement in compulsions when urges arise and, indeed, response prevention is the main predictor of ERP treatment response (Wheaton et al. 2016). However, trauma‐exposed individuals endorse more severe compulsions (Miller and Brock 2017; Pinciotti and Fisher ²⁰²²) and are more likely to be labelled as having ‘treatment resistant OCD’ in ERP treatment settings (Gershuny et al. 2008; Semiz et al. ²⁰¹⁴), suggesting that trauma exposure may complicate the delivery and effectiveness of ERP.

While several studies support trauma as an etiological factor in the onset of OCD for some individuals (Murayama et al. 2020; Pinciotti and Fisher ²⁰²²; Reifels et al. ²⁰¹⁹), more crucially related to conceptualisation and treatment is how trauma exposure may continue to influence OCD long after the traumatic event has passed. Trauma‐exposed individuals with OCD often perceive a connection between their OCD and traumatic experience (Pinciotti and Fisher 2022; Pinciotti, Wadsworth, et al. ²⁰²⁵), and the association between OCD and trauma appears strongest among survivors of interpersonal traumas (Miller and Brock 2017; Ojalehto et al. ²⁰²³; Pinciotti, Riemann, and Wetterneck ²⁰²¹). Endorsement of common symptom dimensions across trauma types (Barzilay et al. 2019; Cromer et al. ²⁰⁰⁷; Ojalehto et al. ²⁰²³; Pinciotti and Fisher ²⁰²²; Pinciotti, Riemann, and Wetterneck ²⁰²¹) and specific compulsions among trauma survivors (Pinciotti, Horvath, et al., 2025) offers support to the notion that the function of compulsions can be impacted by trauma. Namely, it has been proposed that OCD symptoms may develop or worsen after trauma exposure to manage trauma‐related symptoms, providing trauma survivors a sense of control, predictability, and certainty in the wake of an unpredictable personal catastrophe (Pinciotti 2023; Pinciotti, Riemann, and Wetterneck ²⁰²¹). Trauma‐exposed individuals with OCD may engage in compulsions that aim to prevent future trauma (e.g., planning an escape route to prepare for a mass shooting); provide certainty about the past trauma (e.g., mentally reviewing to ‘figure out why’ the trauma happened) or future potential trauma (e.g., a sexual abuse survivor checking oneself for signs of being a paedophile); or cope with overwhelming trauma‐related symptoms (e.g., ritualised counting to distract from intrusive trauma memories).

These trauma‐related compulsions have been described at length in published case examples (e.g., Gershuny et al. ²⁰⁰³; Pinciotti ²⁰²³; Pitman ¹⁹⁹³), yet very little empirical research has examined the influence of trauma on OCD compulsions. Recently, a novel mixed method approach was used to identify the most distressing and functionally impairing compulsions in a sample of patients in intensive OCD treatment (Pinciotti et al. 2023). The study extracted documented compulsions from patient charts that were collaboratively identified by clinicians and patients during treatment as primary targets for response prevention as part of the programme's treatment as usual protocol (termed ‘bans’ in the treatment programme). These compulsions, coded as present/absent, were subsequently examined for patterns of co‐occurrence and dissimilarity, providing context to the larger conceptualisation of these behaviours. Of the 62 compulsions extracted from patient charts, cleaning and handwashing compulsions exhibited the greatest degree of co‐occurrence while also having the highest degree of dissimilarity from other compulsions, suggesting that these compulsions tend to occur in tandem and are conceptually unique from other OCD compulsions. Counting, use of barriers (e.g., using a sleeve to open the door), isolation, list‐making, and rewriting all exhibited conceptual relatedness through a high propensity for co‐occurrence, while checking compulsions evidenced the highest dissimilarity to other OCD compulsions, reflecting perhaps its lack of content specificity and likelihood to present indiscriminately across content dimensions. With respect to the impact of these compulsions on intensive ERP treatment, patients with just right compulsions had markedly worse treatment outcomes compared to those who did not.

Given the unique trauma‐related functions proposed to motivate compulsions among trauma‐exposed individuals with OCD, it is likely that the clustering (and therefore conceptualisation) of compulsions differs between trauma‐exposed and non‐trauma‐exposed patients. Indeed, the authors suggested that the previously found association between trauma and just right symptoms (Pinciotti, Riemann, and Wetterneck ²⁰²¹) may be a potential driver for why treatment outcomes were worse among patients with these compulsions (Pinciotti et al. 2023). A secondary analysis of these data provided some preliminary context for the influence of trauma on compulsions (Pinciotti, Horvath, et al., 2025). Reassurance, rumination, and pulling compulsions were found to be more common among trauma‐exposed patients, and self‐assurance compulsions were found to be less common. Interestingly, however, whereas reassurance‐seeking compulsions were associated with worse treatment outcomes for trauma‐exposed patients in this sample, self‐assurance compulsions were associated with better treatment outcomes for trauma‐exposed patients.

While findings from Pinciotti, Horvath, et al., ²⁰²⁵ offer some cues as to trauma‐related differences in the prevalence of compulsions and the moderating effect of trauma on treatment outcomes, these outcomes were only compared for the compulsions that significantly differed in prevalence, excluding compulsions that may not exhibit greater prevalence among trauma survivors but may, when present, be more likely to be trauma‐related. Checking compulsions, for example, were highly prevalent across the total sample (45.5%) regardless of trauma exposure, reflecting its centrality to OCD overall and likelihood to exist across different OCD contexts (Pinciotti et al. 2023). Yet compulsive checking has been observed among trauma exposed individuals with and without OCD (Barzilay et al. 2019; Cromer et al. ²⁰⁰⁷; Pinciotti ²⁰²³; Pinciotti and Fisher ²⁰²²; Pinciotti, Riemann, and Wetterneck ²⁰²¹; Tuerk et al. ²⁰⁰⁹), suggesting a unique underlying trauma‐related function compared to checking outside the context of trauma exposure. Similarly, the prevalence of people pleasing behaviours such as over‐apologising and over‐explaining did not significantly differ between trauma‐exposed and non‐trauma exposed patients (Pinciotti, Wadsworth, et al. ²⁰²⁵), yet among survivors of interpersonal trauma, clinical observations suggest that these compulsions function to prevent future revictimisation by mitigating conflict that could turn violent. Thus, examining trauma‐related differences at only the individual compulsion‐level informs just part of the conceptualisation for how trauma exposure can influence OCD presentation.

Missing still from extant research is a larger conceptualisation for how trauma exposure impacts the underlying function of compulsions. Factor analysis is a statistical approach that provides insight into the underlying latent constructs connecting phenomena that co‐occur. Examining differences in the factor analysis‐derived clustering of compulsions between trauma‐exposed and non‐trauma exposed individuals with OCD can inform the larger conceptualisation of how these compulsions relate to one another, and how their function may differ as a result of trauma exposure. As the clustering of compulsions within a latent factor is likely to reflect a common function or motivation for the behaviour (e.g., handwashing and cleaning compulsions that mutually function to remove contaminants), differences in clustering between individuals with and without a history of trauma could suggest that the experience of trauma alters the expression of and connections between OCD compulsions—theoretically by modifying the underlying function of engaging in the compulsion. Accordingly, the present study offers a tertiary analysis of the sample described in these methodologically novel studies (Pinciotti et al. 2023; Pinciotti, Horvath, et al., ²⁰²⁵) to determine how the clustering of OCD compulsions differs as a function of trauma exposure, and how trauma‐derived compulsion clusters might uniquely relate to intensive treatment outcomes.

2. Method

2.1. Participants

This project was reviewed and approved by the Rogers Behavioral Health System Institutional Review Board. Data from this study included only patients who provided informed consent to have their data used for research. Included in the sample were all 622‐research consented, adult patients diagnosed with OCD receiving specialised treatment in residential, partial hospitalisation (PHP), or intensive outpatient (IOP) treatment programs for OCD, anxiety, and mood at Rogers Behavioral Health in Oconomowoc, Wisconsin between July 29^th, 2015, and October 1^st, 2021. As it is common for patients to have multiple admits (e.g., stepping up or stepping down in level of care, readmitting to a level of care, etc.), data used in the present study represents each participant's first admit at their highest level of care.

Participants ranged in age from 18 to 70 years (M = 29.42, SD = 11.61) and were 0.6% American Indian/Alaskan Native (n = 4), 2.4% Asian (n = 15), 2.1% Black (n = 13), 0.5% Native Hawaiian/Pacific Islander (n = 3), 86.0% White (n = 535), 1.8% multiple (n = 11), 0.1% Other (n = 1), and 6.4% refused or were not recorded (n = 40). An additional 3.9% identified their ethnicity as Hispanic or Latin/e (n = 24). Participants were 52.4% female (n = 326) and 47.6% male (n = 296) and, of the 347 for whom gender identity information was available, identified as 45.5% cisgender man (n = 158), 51.0% cisgender woman (n = 177), and 3.5% transgender/nonconforming (n = 12).

2.2. Measures

2.2.1. Compulsions

As described in Pinciotti et al., ²⁰²³ and Pinciotti, Horvath, et al., ²⁰²⁵, compulsions are referred to in the treatment programme as ‘bans’ and response prevention for each compulsion (i.e., number of resists and submits to each compulsion urge) is tracked in a small ‘ban book’ by the patient on a momentary basis each day. These compulsions represent the behaviours that are most frequent and functionally impairing for the patient at the time of treatment, collaboratively agreed upon by clinicians and patients at the start of treatment, and added or amended throughout treatment as needed based on clinical observations (e.g., newly identified compulsion once insight increases or the behaviour is observed by treatment team).

A total of 5540 compulsions were extracted from patient charts and reviewed and coded by the first author. Because the aim of the original study was to identify commonly occurring clusters of compulsions, compulsions were relabelled and/or recoded conservatively and only when discrepancies were clearly semantic (e.g., barriers and using barriers) or were idiosyncratic to a specific patient yet clearly indicative of a more general term for the compulsion (e.g., calling or emailing pastors or past professors and reassurance seeking; Pinciotti et al. ²⁰²³). Compulsions that appeared to be likely related but potentially distinct (e.g., counting [possibly related to not just right experiences] and good/bad numbers [possibly related to superstition]) were left as distinct compulsions. This coding scheme was subsequently reviewed by the third author of the original study who served as clinical supervisor for the treatment programs sampled. Both authors involved in coding were closely involved in the treatment programme, worked closely with the treatment providers whose documentation of ‘bans’ were extracted, and were familiar with the common terminology used to describe behaviours, which also informed coding decisions. Discrepancies in coding decisions, typically related to whether to lump or split categories, were discussed between first and third author of the original study until a consensus was achieved. A total of 62 distinct compulsions were coded.

2.2.2. Traumatic/Stressful Life Event Exposure

Trauma exposure was coded based on responses to a social services assessment conducted by each patient's licenced therapist as part of treatment as usual. As part of this interview, therapists ask each patient: ‘Have you ever experienced a serious traumatic event?’ and documented responses as none, within the last 6 months, or more than 6 months ago. Positive responses were collapsed as 1 = trauma exposed and 0 = not trauma exposed. A series of yes/no questions subsequently ask about specific interpersonal traumatic or stressful life experiences with being threatened or bullied, emotionally abused, physically abused, forced to engage in unwanted sexual activity, and being the victim of neglect or exploitation of any kind (e.g., sexual, financial, or labour). As reported in Pinciotti, Horvath, et al., ²⁰²⁵, 39.7% (n = 247) endorsed experiencing any type of trauma, including 42.9% who were physically threatened or bullied (n = 174), 30.3% who experienced emotional abuse (n = 183), 15.6% who experienced physical abuse (n = 84), 13.6% who experienced sexual abuse (n = 72), and 8.1% who experienced neglect or exploitation (n = 17).

2.2.3. Length of Stay

Length of stay was defined as the number of active treatment days. For residential treatment, this includes 7 days/weekly; for PHP, this includes 5 days/weekly; and for IOP, this includes 4 days/weekly.

2.2.4. Yale Brown Obsessive‐Compulsive Scale–Self‐Report (Y‐BOCS‐SR; Goodman et al. ¹⁹⁸⁹)

The Y‐BOCS‐SR is a 10‐item self‐report measure of OCD severity. Participants rate each item on a 5‐point Likert scale, with higher numbers indicating greater severity. Total scores range from 0–40, with a cutoff of 22 indicating moderately severe OCD (Cervin et al. 2022). The Y‐BOCS‐SR has strong internal consistency and test‐retest reliability (Steketee et al. 1996); in the current study, internal consistency was likewise strong (α = 0.86).

2.3. Procedure

Patients seeking treatment at Rogers Behavioral Health first completed a telephone interview assessing symptoms of OCD and anxiety. Licenced psychiatrists and/or psychologists with training and expertise in OCD and anxiety reviewed all potential admissions and determined appropriateness for OCD/anxiety specialty treatment as well as level of care. Upon admission to the programme, all patients completed a diagnostic evaluation with a licenced psychiatrist to confirm the diagnosis of OCD and any co‐occurring diagnoses based on the DSM‐5, as well as battery of self‐report assessments. Within 3 days of admit to residential and 5 days of admit to PHP or IOP, patients meet with their therapist to complete the social services assessment. During this timeframe, patients also meet with their behaviour specialist who is responsible for constructing and implementing the exposure hierarchy and collaboratively agreeing upon the compulsions that will be considered ‘bans’ throughout treatment. A battery of self‐report assessments is completed at regular timepoints throughout treatment and, relevant to the present study, again at treatment discharge.

The primary mode of treatment provided at all levels of care is cognitive behavioural therapy with ERP. Additional therapeutic interventions include dialectical behaviour therapy, cognitive restructuring, and recreational therapy. Ancillary treatments (e.g., behavioural activation) may be provided secondarily on an as‐needed basis. The IOP/PHPs provide multidisciplinary care that includes individual, group, and family therapy, and medication management by psychiatrists. IOP involves 3 hours of treatment 4 days a week and PHP involves 6 hours of treatment 5 days a week. The residential treatment programme provides longer‐term, 8 hours of daily multidisciplinary care that includes individual, group, and family therapy, medication management by psychiatrists, medical support by nursing staff, and dietary support as needed.

2.4. Analytic Plan

Consistent with Pinciotti et al. (2023), compulsions prevalent in less than 5% of the total sample were removed, leaving 26 compulsions for analyses. ¹ Exploratory factor analysis (EFA) and principal component analysis (PCA) were conducted first, on the set of 26 compulsions. ² Two methods were used to select discovery groups for the factor structure of the compulsions. First, the group of participants that had experienced a trauma was used as the discovery sample, and second, a k‐fold method was used to split the full sample into ‘k’ random groups, each of which was used as the discovery sample in turn. The k‐fold method allows for a comparison and contrast of the trauma‐derived model with a model affected by different biases, as the sample subset used for model discovery is selected differently (theory for the trauma model and randomly for the k‐fold). Thus, overlap between the two models is less likely to be due to bias and should be more replicable in future samples. ^{3 , 4}

These models were then reduced by assessing the Cronbach's alpha of each factor and removing indicators that reduced or did not contribute to fit (r.drop) and whether the standardised alpha would be improved if they were removed, until no more improvements could be made through item reduction. Models were also reduced based on inter‐item and item‐factor correlations. Factors were then named, and small alterations were made based on theoretical connections between variables. ⁵ Reduced models were then used in structural equation models (SEM) evaluating prediction of length of stay and Y‐BOCS‐SR scores at discharge by trauma through these factors (see Figures 1 and 2). Chi square differences tests were conducted to evaluate differences in prevalence of compulsion factors between those with and without trauma exposure and across specific traumatic and stressful life event types. Sensitivity analyses were conducted with specific, rather than general, traumatic or stressful life event indicators. These analyses were exploratory given a dearth of research to form specific hypotheses. All analyses were performed in R version 4.5.0 (2025‐04‐11, ucrt). CFA and SEM were conducted using the lavaan package (v0.6.19). EFA and PCA were carried out with the psych package (v2.5.3). Data cleaning and manipulation using the tidyverse suite (v2.0.0).

FIGURE 1.

CFA factor structures for trauma (left) and k‐fold derived (right) models. Standardised factor loadings. p < 0.1, *p < 0.05, **p < 0.01, ***p < 0.001.

FIGURE 2.

SEM structural paths to latent factor structures. Intensive outpatient programme (IOP) is reference category for residential and partial hospitalisation programme (PHP) levels of care. Significant paths coloured in green, p < 0.1. *p < 0.05, **p < 0.01, ***p < 0.001.

3. Results

3.1. EFA and PCA

3.1.1. Trauma

In the initial Trauma model, factor 1 was comprised of six items (rewriting, rereading, list making, repeating, just right, ordering/arranging); factor 2 was comprised of three items (cleaning, handwashing, and barriers); factor 3 was comprised of seven items (researching, rumination, avoidance, self‐assurance, reassurance, worrying, and confessing), factor 4 was comprised of five items (pulling, picking, somatic movement, touch/tap/rub, and counting), factors 5 and 6 were comprised of 2 items each (over‐apologising and people‐pleasing; praying and neutralising, respectively), and factor 7 was indicated by only a single indicator (mental review) and was subsequently moved to factor 3. Item reduction was subsequently conducted. For factor 1, this led to dropping ordering/arranging (r.drop = 0.23), just right (r.drop = 0.35), and repeating (r.drop = 0.32). Remaining variables had an r.drop ≥ 0.61, and the final standardised alpha = 0.84. For factor 3, this led to dropping mental review (r.drop = 0.12) and self‐assurance (r.drop = 0.17), at which point r.drop ≥ 0.17 and the standardised alpha = 0.46 with no improvements—although worrying was ultimately dropped due to concerns of it being too vague or general. For factor 4, counting was dropped (r.drop = 0.15) over pulling, given that pulling had a greater theoretical attachment to somatic movement and touch/tap/rub, at which point r.drop ≥ 0.16, std. alpha = 0.45. Picking was then re‐added due to its theoretical connection with pulling, at which point, r.drop ≥ 0.18 and std. alpha = 0.45. Notably, although there was some suggestion that picking and pulling and somatic movement and touch/tap/rub, respectively, may load onto their own factors, models where these factors were included produced negative variance estimates, and so all four were loaded onto the same factor. Factors 5 and 6 were not reduced as each contained only two indicators.

After item reduction, the chosen model consisted of six factors: Factor 1 (Perfecting; AVE = 0.65, CR = 0.84) consisted of rewriting, rereading, and list making; factor 2 (Cleanliness; AVE = 0.63, CR = 0.82) consisted of cleaning, handwashing, and barriers; factor 3 (Certainty Seeking; AVE = 0.13, CR = 0.47) consisted of researching, rumination, avoidance, reassurance, and confessing; factor 4 (Body Focus; AVE = 0.17, CR = 0.44) consisted of pulling, picking, somatic movement, touch/tap/rub; and factors 5 (Impression Control; AVE = 0.29, CR = 0.44) and 6 (Absolution; AVE = 0.26, CR = 0.33) consisted of two compulsions each (over‐apologising and people‐pleasing; praying and neutralising, respectively). This model had acceptable fit: χ²(174) = 377.43 (χ²/df = 2.17), AGFI = 0.93, robust (R) CFI = 0.91, R‐TLI = 0.89, R‐RMSEA = 0.043(CI 90% [0.037, 0.049]), SRMR = 0.051. In this CFA, the first loading of each factor was fixed for identification and not tested for significance. Among the estimated loadings, all were significant after Benjamini‐Hochberg correction (BH‐adjusted p < 0.05) except for somatic movement and touch/tap/rub (Body Focus factor), people pleasing (Impression Control factor), and neutralising (Absolution factor), which were not significant. These factors (present/absent) did not differ in prevalence between trauma‐exposed and non‐exposed patients (see Table 1).

TABLE 1.

Prevalence of trauma model‐ and K‐Fold model‐derived factors across trauma types.

Trauma model	Trauma		Threat/Bullied		Emotional abuse		Physical abuse		Sexual abuse		Neglect/Exploit
	% Yes (n)	% No (n)	% Yes (n)	% No (n)	% Yes (n)	% No (n)	% Yes (n)	% No (n)	% Yes (n)	% No (n)	% Yes (n)	% No (n)
Perfecting	15.7% (42)	14.9% (59)	9.2% (19)	13.7% (33)	12.7% (26)	16.1% (71)	6.3% (6)**	16.4% (79)**	14.3% (12)	14.8% (72)	5.3% (1)	7.4% (17)
Cleanliness	31.5% (84)	36.0% (143)	27.1% (56)*	35.3% (85)*	31.2% (64)	35.6% (157)	30.5% (29)	33.7% (163)	39.3% (33)	31.3% (152)	26.3% (5)	25.1% (58)
Certainty seeking	83.5% (223)	85.1% (338)	75.8% (157)	81.7% (197)	82.0% (168)	85.3% (376)	81.1% (77)	84.7% (409)	85.7% (72)	83.5% (405)	63.2% (12)	72.3% (167)
Body focus	20.2% (54)	19.6% (78)	15.9% (33)	17.4% (42)	20.0% (41)	20.2% (89)	15.8% (15)	21.5% (104)	27.4% (23)	19.4% (94)	15.8% (3)	15.2% (35)
Impression control	13.1% (35)	10.3% (41)	13.5% (28)	8.3% (20)	10.2% (21)	12.2% (54)	13.7% (13)	12.0% (58)	20.2% (17)**	10.7% (52)**	10.5% (2)	11.3% (26)
Absolution	15.4% (41)	16.1% (64)	12.1% (25)	15.4% (37)	14.6% (30)	16.1% (71)	12.6% (12)	15.1% (74)	9.5% (8)	16.1% (78)	15.8% (3)	13.4% (31)
K‐fold model
Perfecting	15.7% (42)	14.9% (59)	9.2% (19)	13.7% (33)	12.7% (26)	16.1% (71)	6.3% (6)**	16.4% (79)**	14.3% (12)	14.8% (72)	5.3% (1)	7.4% (17)
Cleanliness	31.5% (84)	36.0% (143)	27.1% (56)*	35.3% (85)*	31.2% (64)	35.6% (157)	30.5% (29)	33.7% (163)	39.3% (33)	31.3% (152)	26.3% (5)	25.1% (58)
Certainty seeking	83.5% (223)	85.1% (338)	75.8% (157)	81.7% (197)	82.0% (168)	85.3% (376)	81.1% (77)	84.7% (409)	85.7% (72)	83.5% (405)	63.2% (12)	72.3% (167)
BFRB	15.0% (40)	13.1% (52)	13.0% (27)	11.6% (28)	15.6% (32)	13.2% (58)	10.5% (10)	15.1% (73)	22.6% (19)*	13.0% (63)*	15.8% (3)	12.6% (29)
Movement	8.6% (23)	8.3% (33)	5.3% (11)	7.9% (19)	7.3% (15)	9.3% (41)	9.5% (9)	8.7% (42)	9.5% (8)	8.7% (42)	0.0% (0)	4.3% (10)
Impression control	13.1% (35)	10.3% (41)	13.5% (28)	8.3% (20)	10.2% (21)	12.2% (54)	13.7% (13)	12.0% (58)	20.2% (17)**	10.7% (52)**	10.5% (2)	11.3% (26)

Abbreviation: BFRB, body‐focused repetitive behaviors.

^* p < 0.05.

^** p ≤ 0.01.

3.1.2. K‐Fold

In the initial K‐fold model, factor 1 was comprised of four items (rewriting, rereading, list making, and ordering/arranging); factor 2 was comprised of three items (cleaning, handwashing, barriers); factor 3 was comprised of four items (pulling, somatic movement, touch/tap/rub, and counting); factor 4 was comprised of six items (researching, rumination, self‐assurance, reassurance, worrying, and confessing); factor 5 was comprised of three items (over‐apologising, people‐pleasing, and praying); and factor 6 was comprised of six items (repeating, just right, picking, mental review, avoidance, and neutralising). Item reduction was subsequently conducted. For factor 1, this led to dropping ordering/arranging (r.drop = 0.28), resulting in r.drops ≥ 0.61, std. alpha = 0.83. Factor 2 was not adjusted. Factor 3 was adjusted similarly to the trauma model. For factor 4, self‐assurance was dropped from factor 4 (r.drop = 0.15), resulting in r.drops ≥ 0.17, and a std. alpha = 0.43, but worrying was maintained as this model was viewed as the more ‘empirically derived’ model and its removal reduced fit. Praying was removed from factor 5 (r.drop = 0.09), resulting in r.drops = 0.28, std. alpha = 0.44. Initially, no changes were made to factor 6 except to drop picking to add it to factor 3, however factor 6 was judged to be lacking in any clear pattern, so changes were made to align it to a clearer trend by dropping avoidance and adding praying, resulting in increased model fit.

After item reduction, the chosen model consisted of 6 factors: Factor 1 (Perfecting; AVE = 0.65, CR = 0.84) consisted of rewriting, rereading, and list making; factor 2 (Cleanliness; AVE = 0.63, CR = 0.82) consisted of cleaning, handwashing, and barriers; factor 3 (Certainty Seeking; AVE = 0.13, CR = 0.47) consisted of researching, rumination, avoidance, reassurance, self‐assurance, worrying, and confessing; factor 4 (Body Repetitive Focused Behaviours [BFRB]; AVE = 0.29, CR = 0.45) consisted of pulling and picking; factor 5 (Movement; AVE = 0.48, CR = 0.62) consisted of somatic movement and touch/tap/rub; and factor 6 (Impression Control; AVE = 0.28, CR = 0.44) consisted of over‐apologising and people‐pleasing. The resulting model had good fit, χ²(137) = 242.27 (χ²/df = 1.77), AGFI = 0.95, R‐CFI = 0.95, R‐TLI = 0.94, R‐RMSEA = 0.034 (CI 90% [0.027, 0.041]), and SRMR = 0.047. All estimated factor loadings were statistically significant (BH‐adjusted p < 0.05), except for picking (BFRB factor) and touch/tap/rub (Movement factor), which were not significant. Loadings fixed for identification were not tested for significance.

3.2. SEM

Trauma ‐ Factor Loadings (Table 2). For the CFA model, average standardised factor loadings were generally around 0.5 (M = 0.50, SD = 0.25). Average loadings were highest for Perfecting (M = 0.80, SD = 0.14), Cleanliness (M = 0.77, SD = 0.24) and Impression Control (M = 0.53, SD = 0.07), but lower for Certainty Seeking (M = 0.33, SD = 0.16), Body Focus (M = 0.40, SD = 0.13), and Absolution (M = 0.42, SD = 0.41), identifying these as targets for improvement. Although most of the latent variables were well‐identified, picking was the only significant indicator for Body Focus. Fit statistics indicated acceptable, though not ideal, model fit according to common benchmarks (Xia and Yang 2019). The trauma SEM model met criteria for RMSEA and SRMR—R‐RMSEA = 0.034 (90% CI: 0.029–0.039) and SRMR = 0.044—while incremental fit indices were acceptable but slightly below more stringent thresholds (R‐CFI = 0.92; R‐TLI = 0.90). Other indices were χ²(282) = 520.34, χ²/df = 1.85, and AGFI = 0.97. These factors (present/absent) did not differ in prevalence between trauma‐exposed and non‐exposed patients (see Table 1).

TABLE 2.

Factor loadings, trauma model.

Compulsion	Perfecting	Cleanliness	Certainty seeking	Body focus	Impression control	Absolution
	λ (SE)	λ (SE)	λ (SE)	λ (SE)	λ (SE)	λ (SE)
Rewriting	0.95
Rereading	0.69 (0.07)***
List making	0.75 (0.11)***
Cleaning		0.89
Handwashing		0.92 (0.05)***
Barriers		0.49 (0.03)***
Researching			0.26
Rumination			0.32 (0.46)***
Avoidance			0.40 (0.79)**
Self‐assurance			0.25 (0.52)*
Reassurance			0.67 (1.09)***
Worrying			0.20 (0.20)**
Confessing			0.24 (0.29)**
Pulling				0.56
Picking				0.44 (0.45)*
Somatic movement				0.25 (1.20)
Touch/tap/rub				0.36 (1.50)
Over‐apologising					0.59
People pleasing					0.48 (0.54)
Praying						0.71
Neutralising						0.14 (0.42)

Note: p‐values adjusted for multiple testing using benjamini‐hochberg false‐discovery rate (BH‐FDR, Benjamini and Hochberg ¹⁹⁹⁵). Items without SE had unstandardised loadings fixed 1.

^* p < 0.05.

^** p < 0.01.

^*** p < 0.001.

Trauma—Regressions (Table 3). None of the latent variables derived from the trauma‐exposed sample predicted Y‐BOCS‐SR scores at discharge. Likewise, trauma exposure was not predictive of the latent variables representing compulsion clusters, nor was trauma predictive of either Y‐BOCS‐SR or length of stay through compulsions, directly or indirectly through compulsion clusters. However, Cleanliness was associated with length of stay, β = 0.11, SE = 0.06, p = 0.04, indicating that patients with compulsions related to ‘Cleanliness’ had treatment stays that were ∼ 3 days longer per standard deviation increase of ‘Cleanliness’.

TABLE 3.

Regressions, trauma model.

Predictors	Y‐BOCS‐SR (D)	LOS	Perfecting	Cleanliness	Certainty seeking	Body focus	Impression control	Absolution
	β (SE)	β (SE)	β (SE)	β (SE)	β (SE)	β (SE)	β (SE)	β (SE)
Trauma	0.03 (0.06)	−0.03 (0.07)	−0.01 (0.02)	−0.08 (0.06)	0.1 (0.01)	0.09 (0.02)	0.08 (0.02)	0.05 (0.02)
Perfecting	−0.02 (0.13)	0.03 (0.11)
Cleanliness	0.01 (0.05)	0.1 (0.06)*
Certainty seeking	−0.13 (0.79)	0.13 (0.94)
Body focus	0.08 (0.48)	0.02 (0.55)
Impression control	−0.01 (0.41)	0.01 (0.44)
Absolution	0.18 (1.19)	0.11 (1.33)
Residential	0.13 (2.33)*	0.41 (13.48)***
PHP	0.11 (2.21)*	0.02 (0.92)
LOS	0.02 (0.3)	—
Y‐BOCS‐SR baseline	0.38 (8.98)***	0.14 (3.8)***
R 2	0.194	0.247	< 0.001	0.006	0.010	0.008	0.007	0.004

Note: †p < 0.1, p‐values adjusted for multiple testing using BH‐FDR.

Abbreviations: (D), measured at discharge; LOS, length of stay.

*p < 0.05. **p < 0.01. ***p < 0.001.

K‐Fold ‐ Factor Loadings (Table 4). Standardised factor loadings were slightly larger in the k‐fold CFA, than for the trauma model (M = 0.55, SD = 0.25). Average loadings were above 0.5 for Perfecting (M = 0.80, SD = 0.14), Cleanliness (M = 0.76, SD = 0.24), Movement (M = 0.66, SD = 0.32), BFRB (M = 0.54, SD = 0.04), and Impression Control (M = 0.53, SD = 0.04) but lower for Certainty Seeking (M = 0.33, SD = 0.16). Fit for the k‐fold SEM model was good, with χ²(233) = 371.63 (χ²/df = 1.59), AGFI = 0.98, R‐CFI = 0.95; R‐TLI = 0.94; R‐RMSEA = 0.028, 90% CI[0.022, 0.034]; SRMR = 0.042. These factors (present/absent) did not differ in prevalence between trauma‐exposed and non‐exposed patients (see Table 1).

TABLE 4.

Factor loadings, K‐fold model.

Compulsion	Perfecting	Cleanliness	Certainty seeking	BFRB	Movement	Impression control
	λ (SE)	λ (SE)	λ (SE)	λ (SE)	λ (SE)	λ (SE)
Rewriting	0.95
Rereading	0.69 (0.07)***
List making	0.75 (0.11)***
Cleaning		0.89
Handwashing		0.92 (0.05)***
Barriers		0.49 (0.03)***
Researching			0.26
Rumination			0.31 (0.45)***
Avoidance			0.39 (0.79)*
Reassurance			0.68 (1.09)**
Self‐assurance			0.25 (0.53)**
Worrying			0.20 (0.21)**
Confessing			0.22 (0.27)**
Pulling				0.57
Picking				0.51 (0.86)
Somatic movement					0.43
Touch/tap/rub					0.88 (1.99)
Over‐apologising						0.50
People‐pleasing						0.56 (0.42)*

Note: p‐values adjusted for multiple testing using Benjamini‐Hochberg false‐discovery rate (BH‐FDR). Items without SE had unstandardised loadings fixed 1.

^* p < 0.05.

^** p < 0.01.

^*** p < 0.001.

K‐Fold ‐ Regressions (Table 5). Neither the latent variables derived from the subsamples nor trauma exposure predicted Y‐BOCS‐SR scores at discharge after controlling for Y‐BOCS‐SR scores at baseline, level of care, and length of stay. However, Cleanliness, β = 0.11, SE = 0.05, p = 0.02, and Certainty Seeking, β = 0.16, SE = 0.079, p = 0.04, predicted length of stay, indicating that each standard deviation (SD) increase in Cleanliness‐related compulsions was associated with a ∼3 days increase in length of stay, and each SD increase in Certainty Seeking compulsions was associated with ∼4 days increase in length of stay. As in the trauma‐derived model, trauma did not predict either length of stay or Y‐BOCS‐SR through any compulsion cluster.

TABLE 5.

Regressions, K‐fold model.

Predictors	Y‐BOCS‐SR (D)	LOS	Perfecting	Cleanliness	Certainty seeking	BFRB	Movement	Impression control
	β (SE)	β (SE)	β (SE)	β (SE)	β (SE)	β (SE)	β (SE)	β (SE)
Trauma	0.03 (0.06)	−0.03 (0.07)	−0.01 (0.02)	−0.08 (0.06)	0.10 (0.01)	0.10 (0.02)	0.01 (0.02)	0.08 (0.02)
Perfecting	−0.01 (0.12)	0.03 (0.11)
Cleanliness	0.02 (0.05)	0.11 (0.05)*
Certainty seeking	−0.08 (0.59)	0.15 (0.77)*
BFRB	0.05 (0.36)	0.02 (0.32)
Movement	−0.01 (0.35)	−0.03 (0.33)
Impression control	0.03 (0.33)	0.03 (0.36)
Residential	0.13 (0.11)*	0.41 (0.07)***
PHP	0.12 (0.12)*	0.03 (0.08)
LOS	0.04 (0.04)	—
Y‐BOCS‐SR baseline	0.38 (0.05)***	0.14 (0.05)***
R 2	0.183	0.245	< 0.001	0.006	0.010	0.009	3.20e–04	0.007

Note: †p < 0.1, p‐values adjusted for multiple testing using BH‐FDR.

Abbreviations: (D), measured at discharge; LOS, ength of stay.

*p < 0.05. **p < 0.01. ***p < 0.001.

3.3. Sensitivity Analyses—Interpersonal Traumatic and Stressful Life Events

3.3.1. Trauma‐Derived Model

Cleanliness was more common among patients who had not been threatened or bullied compared to those who had been threatened or bullied (χ ² = 3.49, p = 0.04, Φ = −0.09); Perfecting was more common among patients who had not been physically abused compared to those who had been physically abused (χ ² = 6.38, p = 0.01, Φ = −0.10); Impression Control was more common among patients who had been sexually abused compared to those who had not been sexually abused (χ ² = 6.08, p = 0.01, Φ = 0.10), and somewhat more common among patients who had been threatened or bullied compared to those who had not been threatened or bullied, although statistical significance was not reached (χ ² = 3.18, p = 0.05, Φ = 0.08). No differences were observed in prevalence of compulsion factors (present/absent) between patients with and without a history of emotional abuse or neglect or exploitation (see Table 1).

As in the primary models, trauma overall did not predict outcomes, but, in this model, bullying predicted less Perfecting (β = −0.10, SE = 0.02, p = 0.013) and Cleanliness (β = −0.10, SE = 0.06, p = 0.013), and physical abuse predicted less Perfecting (β = −0.08, SE = 0.02, p = 0.002). No significant indirect or total effects of bullying or physical abuse on treatment outcomes were observed via these compulsion pathways (all indirect effects and total effects: |β| < 0.03, p > 0.22), suggesting that, despite differences in prevalence for these compulsions among individuals with histories of being bullied/threatened or physically abused, there was no direct or indirect impact of these compulsions on treatment outcomes. For sexual abuse, the indirect effect on length of stay via Cleanliness compulsions was significant (indirect path: β = 0.24, SE = 0.17, p = 0.008). The total effect of sexual abuse via Certainty Seeking compulsions was also significant (β = 0.26, SE = 0.78, p = 0.011). Conversely, the direct effect of sexual abuse on length of stay was not significant (β ≈ 0, p > 0.05). These results indicate that sexual abuse impacts treatment duration primarily through its effects on Cleanliness and Certainty Seeking compulsions, rather than through a direct pathway. Thus, findings suggest that it is not a history of sexual abuse itself that contributes to longer length of treatment stay but rather the likelihood of sexual abuse to influence specific compulsive behaviours.

3.3.2. K‐Fold Derived Model

Of the unique factors, BFRB was more common among patients who had been sexually abused (χ ² = 5.38, p = 0.02, Φ = 0.10). No other differences across trauma types or from the Trauma model were found (see Table 1). As in the trauma‐derived model, bullying predicted less Perfecting (β = −0.10, SE = 0.02, p = 0.013) and Cleanliness (β = −0.10, SE = 0.06, p = 0.013), and physical abuse predicted less Perfecting (β = −0.08, SE = 0.02, p = 0.002). In addition, neglect predicted less Movement (β = −0.06, SE = 0.02, p = 0.004), and, as in the primary models, engagement in Cleanliness compulsions predicted longer length of stay (β = 0.11, SE = 0.05, p = 0.009).

Consistent with the trauma‐derived model, engagement in Cleanliness compulsions predicted longer length of stay (β = −0.11, SE = 0.05, p = 0.03). Further, the total effect of sexual abuse on length of stay, combining direct and indirect pathways through Cleanliness compulsions, was significant (total effect: β = 0.25, SE = 0.17, p = 0.01), as was the total effect through Certainty Seeking compulsions (β = 0.13, SE = 0.74, p = 0.05). The total effect of bullying through Certainty Seeking compulsions also predicted longer length of stay (β = 0.14, SE = 0.78, p = 0.02). However, direct effects of sexual abuse and bullying on length of stay remained non‐significant, suggesting again that it is not the trauma type alone that directly increases treatment length but rather the likelihood of these traumas to influence certain compulsive behaviours. All other indirect pathways tested were non‐significant.

Thus, findings from the primary models were replicated in the sensitivity analyses, except for the main effect of Cleanliness compulsions on length of stay in the trauma‐derived model, which may have been accounted for by the total effects of sexual abuse through Cleanliness. Moreover, several associations emerged only when examining specific interpersonal trauma categories, indicating that different trauma types are differentially related to specific compulsions and treatment response, with some effects dependent on compulsion type.

4. Discussion

The present study sought to inform conceptualisation of compulsions related to trauma, potentially by offering insight into the underlying, latent functionality of compulsions in a sample of trauma‐exposed and non‐trauma exposed patients receiving intensive OCD treatment.

4.1. Trauma‐Related Compulsions

First, the factor structure of compulsions was examined to inform how certain compulsions may cluster together as a function of trauma. Exploratory factor and principal components analyses found the best fit for a six‐factor structure in both trauma‐derived and k‐fold models. These two models largely replicated one another with respect to Perfecting, Cleanliness, Certainty Seeking, and Impression Control factors, suggesting high reliability of these factors regardless of trauma exposure. However, Certainty Seeking in the k‐fold model also included compulsions of self‐assurance and worrying, whereas these compulsions were not included in the Certainty Seeking factor among trauma‐exposed patients. In an earlier analysis of this data (Pinciotti, Horvath, et al., 2025), self‐assurance compulsions were less prevalent among trauma‐exposed patients yet, when present, were associated with better treatment outcomes for trauma‐exposed patients; the same pattern was not found for non‐trauma exposed patients. Taken all together, these data suggest that self‐assurance behaviours may be less associated with a certainty seeking motivation among trauma survivors, perhaps because self‐assurance may adaptively counteract negative trauma‐related beliefs, facilitate appropriate engagement during exposures, and improve trauma‐exposed patients' ability to accurately differentiate threat versus safety cues (Pinciotti, Horvath, et al., 2025). More research on the functionality of self‐assurance among trauma‐exposed individuals with OCD is needed to support these preliminary claims.

Differences in body‐focused compulsions also emerged. In the trauma‐derived model, Body Focus consisted of pulling, picking, somatic movement, and touch/tap/rub, whereas in the k‐fold model, this factor (termed BFRBs) consisted only of pulling and picking. This difference suggests a unique symmetry/just‐right flavour to Body Focus compulsions among trauma‐exposed patients—reflecting prior work linking symmetry symptoms and trauma exposure (Pinciotti, Riemann, and Wetterneck ²⁰²¹)—whereas among those without trauma, a more classic BFRB presentation is observed.

Perhaps most notable was the emergence of an Absolution factor in the trauma‐derived model consisting of praying and neutralising, compulsions that intend to ‘undo’ a perceived wrong, whereas a Movement factor consisting of somatic movement and touch/tab/rub emerged in the k‐fold model. These discrepancies in factor structures suggest that trauma exposure may be more strongly associated with compulsions intended to absolve guilt, obligation, or punishment, perhaps associated with self‐blame about the cause or consequences of the traumatic event. Guilt is theorised to involve four main cognitive components: (1) perceived responsibility for the event, (2) perceived insufficient justification for actions taken, (3) perceived values violation, and (4) perceived (yet distorted) beliefs about the foreseeability and preventability of bad outcomes (Kubany and Ralston 2006).

The emergence of an Absolution factor among trauma‐exposed patients may reflect these cognitions related to the traumatic event, manifesting as compulsions intended to assuage them, while also relating to maladaptive responses to obsessions themselves. Individuals with OCD often experience inflated responsibility to control unwanted intrusive thoughts, particularly those that are at odds with their morals and values, and through repeated negative reinforcement, develop an inflated belief in one's ability to foresee and prevent future bad outcomes (Salkovskis 1985, 1999; Salkovskis et al. ¹⁹⁹⁶). Moreover, inflated responsibility to control thoughts is associated with scrupulosity symptoms (Nelson et al. 2006), which are also more common among individuals who perceive that their OCD was caused by trauma (Pinciotti and Fisher 2022), and may further explain the loading of praying compulsions onto the Absolution factor in the present sample. Thus, trauma‐exposed patients may endure a ‘double dose’ of guilt and responsibility related to both their trauma and their OCD symptoms, contributing to a greater perceived need to obtain absolution for trauma‐ and OCD‐related wrongdoings. BFRB and Movement‐related compulsions, conversely, may be more strongly associated with OCD in general, as evidenced by their emergence only in the k‐fold model but not in the trauma‐derived model.

4.2. Event‐Specific Compulsions

Sexual abuse was linked to a greater prevalence of Impression Control and BFRB compulsions. Disconnection and rejection schemas are common among individuals with sexual abuse histories (Tezel et al. 2015), suggesting that individuals with these experiences may be more likely to expect that others will hurt them, that needed protection will not be provided by caregivers and loved ones, that relationships will end imminently or unpredictably, and believe that they are inherently flawed, unlovable, and ‘different’ from others. Over‐apologising and people pleasing behaviours may function among sexual abuse survivors to prevent harm and abandonment from others or may stem from an inherent belief that one's own needs are less important than the needs of others. Regarding BFRBs, individuals with trichotillomania (hair pulling) and excoriation (skin picking) disorders report greater severity of childhood trauma (Lochner et al. 2002; Özten et al. ²⁰¹⁵), and 86% of female patients with trichotillomania in one study described a history of violence concurrent with the onset of their hair pulling symptoms (Boughn and Holdom 2003). The close onset between these experiences and behaviours suggests that hair pulling and skin picking might emerge as a means of coping with acute interpersonal trauma.

Cleanliness compulsions were somewhat more common among individuals with a history of sexual abuse (39% vs. 31%), however this was not a statistically significant difference. The well‐established relationship between sexual trauma and contamination is less clear in clinical samples of individuals with OCD (e.g., Pinciotti and Fisher ²⁰²²). Findings from the present study contribute to a growing literature showing a stronger effect of sexual abuse on Cleanliness in community samples than in clinical samples, likely a result of the centrality and prevalence of contamination and washing symptoms in OCD broadly (e.g., Pinto et al. ²⁰⁰⁸), which may obscure the unique association between sexual abuse and Cleanliness behaviours when evaluating group‐level differences. Nonetheless, although the present study did not directly ask about trauma‐relatedness, one might assume that individuals with sexual abuse and Cleanliness have a greater likelihood of these compulsions stemming from this traumatic experience and/or serving a trauma‐related function (e.g., washing away the sexual violation). Several negative associations were found between trauma types and compulsions as well, with a history of being threatened or bullied associated with less Cleanliness and physical abuse associated with less Perfecting.

4.3. Sensitivity Analyses

None of the latent variables derived from the trauma‐exposed sample predicted Y‐BOCS‐SR scores at discharge or length of stay, mirroring earlier findings from this data that trauma exposure did not itself predict treatment outcomes (Pinciotti, Horvath, et al., 2025). It appears that, despite trauma‐exposed individuals having more severe and stable OCD symptoms over time (Tibi et al. 2020) and being more likely to be labelled ‘treatment resistant’ (Gershuny et al. 2008; Semiz et al. ²⁰¹⁴), treatment provided during this time at Rogers Behavioral Health does not differ in effectiveness for those who do and do not report a trauma history. Trauma exposure may not adversely impact treamtent outcomes during a single episode of care but may still be associated with greater risk for symptom relapse, perhaps aligning prior research regarding the stability and chronicity of OCD symptoms over time among trauma‐exposed individuals (e.g., Tibi et al. ²⁰²⁰).

Notably, our trauma categorisation relied upon clinician documentation, did not utilise a psychometrically validated measure of trauma history, and probing for specific events included interpersonal traumas only. This limits generalisability because the open‐ended term ‘trauma’ may not resonate with specific experiences, such as unacknowledged rape (Wilson and Miller 2016) or indirectly experienced events (Pinciotti, Riemann, and Wetterneck ²⁰²¹). Indeed, the prevalence of documented trauma exposure in the current sample (40.3%) is lower than what would be expected in intensive levels of care based on what has been previously reported in similar settings using psychometrically validated assessments including a wider variety of traumatic events (e.g., Gershuny et al. ²⁰⁰⁸). Further, measurement limitations impede the ability to confirm that events coded as traumatic would indeed meet diagnostic criteria for a trauamtic event. Of note, several mediation effects were found in sensitivity models that were not found in the primary models, suggesting that examination of specific traumatic and stressful life event types—described below—adds critical nuance to these relationships that are otherwise harder to detect at the broader level.

4.4. Sexual Abuse and Cleanliness

In both base trauma‐derived and k‐fold sensitivity models, length of stay was consistently predicted by the total effect and indirect effect of sexual abuse through Cleanliness. This consistent finding provides stronger evidence that sexual abuse may predispose individuals with OCD to develop compulsions related to cleanliness, which may increase the amount of time needed in treatment to achieve the same benefit as peers without these experiences and compulsions. This finding provides a clinical extension to existing research on contamination and disgust following sexual victimisation in non‐clinical samples (Badour et al. 2012, 2014; Barzilay et al. ²⁰¹⁹; Fairbrother and Rachman ²⁰⁰⁴; Pinciotti et al. ²⁰²²). Memory cues specific to sexual abuse can trigger urges to engage in washing behaviours (Fairbrother and Rachman 2004), so it may be harder to resist these compulsions if they are seen as necessary to cleanse one's essence or physical body from the sexual violation. In addition, these Cleanliness compulsions may be likely to coincide with posttraumatic stress symptoms such as intrusive memories which may make symptoms more severe and treatment more complex (Fergus and Bardeen 2016).

The weaker association between sexual trauma and contamination among clinical compared to community samples, described above, may also explain why, in the current study, only the total effect (not the direct or indirect effect independently) was a significant predictor of length of stay. Contamination concerns are characterised by inhibitory intolerance of uncertainty, suggesting that individuals with these concerns may be more likely to feel ‘paralysed’ by uncertainty due to increased dependence on overt immediate compulsions to reduce anxiety (Pinciotti, Riemann, and Abramowitz ²⁰²¹). Individuals who engage in Cleanliness compulsions require more time in treatment to obtain the same reduction in Y‐BOCS‐SR as others, perhaps because of greater difficulty resisting the urge to engage in immediate compulsions (Wheaton et al. 2016), and these compulsions may be even more difficult to resist when linked to sexual violations.

5. Sexual Abuse, Bullying, and Certainty Seeking

Broadly, compulsions associated with Certainty Seeking were also associated with longer lengths of treatment stay, perhaps mirroring previous research that persistent intolerance of uncertainty has an adverse impact on treatment outcome (Pinciotti et al. 2020). Mediation analyses suggest that sexual abuse and bullying may uniquely predispose individuals with OCD to develop these compulsions, thereby increasing the amount of time needed in treatment to achieve an adequate reduction of OCD symptoms. In the current sample, Certainty Seeking consisted of compulsions involving repetitive negative thinking (i.e., rumination and worrying), repetitive internal and external certainty seeking compulsions (i.e., researching, reassurance, self‐assurance, and confessing), and overt avoidance behaviours. Thus, the common overarching theme is repetitive engagement in overt and covert compulsions to achieve absolute certainty about feared outcomes. Survivors of sexual abuse and bullying may rely on Certainty Seeking compulsions to provide a sense of predictability and control with the ultimate goal of preventing revictimisation, thus making these compulsions more difficult to resist and requiring a greater dose of treatment to attain comparable treatment outcomes.

Importantly, just right compulsions were dropped from the trauma sample because, although they initially loaded onto Perfecting in both models, they did not correlate well with rereading, rewriting, and list‐making, and ultimately reduced the internal reliability of that factor. It is possible that just right compulsions may be correlated with, but more general than, other Perfecting compulsions, leading to more error in its association with Perfecting and a smaller loading. Rereading, rewriting, and list‐making notably all share a literacy component, whereas just right is a broader description of a range of behaviours including symmetry and completeness‐driven actions and, most commonly in this sample, behaviours vaguely documented in charting as ‘just right behaviours’ which are harder to parse apart in the absence of additional qualitative information. Thus, it should be noted that trauma and just right could have been predictive of treatment responsiveness if included, and future research should continue to examine these relationships more precisely.

5.1. Limitations

The novel approach to identifying and coding idiosyncratic compulsions represents both a strength and weakness of this study. The approach is a strength because it provides a more nuanced, precise, and comprehensive reporting of compulsions based on clinically observed behavioural data not constrained by the limits of existing measures and predetermined checklists, offering a more thorough and naturalistic understanding of the most debilitating compulsions for each patient. The approach is limited, however, because the language used to describe compulsions may not generalise to other programs or settings, making exact replication of the coding structure and associated findings a challenge. Despite the novel approach to identifying compulsions, they did, crucially, evidence moderate to high validity with the Y‐BOCS‐SR checklist (Pinciotti et al. 2023) when compulsions did overlap, suggesting that the human subjectivity inherent in labelling and coding decisions did not significantly interfere with reliability. Nonetheless, to reduce patient burden about the number of ‘bans’ they were to track each day, these documented compulsions reflected only those most distressing and functionally impairing at the time of treatment and do not reflect the full range of compulsions experienced by each patient.

Additional limitations include reliance on a largely homogenous group and on self‐report outcomes data which may reduce accurate estimation of patient severity. There is a possibility that some information is missing due to the lack of standardised assessment procedures. Indeed, the single item measure of trauma exposure lacks validation with existing measures (e.g., Life Events Checklist for DSM‐5; Weathers et al. ²⁰¹³) and may not have captured all potential trauma types. The follow‐up trauma questions focused only on interpersonal traumas and did not encompass as many indirectly experienced traumas or any non‐interpersonal traumas. While interpersonal traumas appear to be most robustly related to OCD (Miller and Brock 2017; Ojalehto et al. ²⁰²³; Pinciotti, Riemann, and Wetterneck ²⁰²¹), this investigation warrants extension among survivors of non‐interpersonal traumas that could influence the occurrence and clustering of compulsions (e.g., checking compulsions following a house fire). The number of patients who had experienced certain trauma experiences (physical and sexual abuse, neglect) was quite low and may reduce reliability of findings related to those trauma experiences. Several effects were present only when examining specific traumatic and stressful life event types, so future research with a larger and more reliably derived trauma sample is needed to replicate findings and ensure generalisability.

With respect to limitations to the statistical approach, compulsions present in less than 5% of the sample were dropped to improve reliability of our analyses. However, these compulsions could be theoretically relevant to traumatic experiences, even if rare, such as compulsions related to caretaking (1.4%), distraction (3.1%), and preparing (3.6%). Exclusion of these compulsions, while statistically appropriate, reduces the generalisability of findings to other trauma samples regarding the full range of compulsions that may be relevant to trauma. Although the k‐fold derived models are discovered on no more than 50% of the sample, because they are selected based on their fitness with the whole sample, they may not generalise as much to other samples as well as the trauma‐derived model might. Further, due to data limitations, the number of folds that could be made in the k‐fold method was low, which increases the power of the discovery sample but decreases the number of internal replications possible. Notably, however, both folds produced by the k = 2 method produced the same model after item reduction, indicating decent internal replicability.

By offering insight into the underlying (latent) structure of observed data, factor analysis is a data‐driven method for conceptualising why a set of compulsions may be related. In the present study, we inferred that the latent construct underlying these compulsion clusters implies something about the functionality of each compulsion set, yet functionality was not directly measured. While the present study suggests that Absolution and Body Focus compulsions may be particularly relevant for trauma‐exposed individuals with OCD, future research should seek to replicate and extend findings by asking explicitly about the function underlying these behaviours to corroborate our proposed conceptualisation. Moreover, fit of the trauma‐derived model approached conventional benchmarks, perhaps reflective of clinical heterogeneity and the unique functions of compulsions for each patient. Finally, CFA was conducted on the same set of participants in which the model was discovered, inflating the likelihood of good model fit. This limitation is somewhat addressed in that the k‐fold model discovered and tested models in different sub‐samples, unlike traditional split‐sample designs, and, in the trauma derived model, by the trauma‐exposed portion of the sample being the minority (40.3%) of the sample but is still notable and will only be truly ameliorated by replicating these models in future independent populations.

Funding

Dr. Pinciotti receives research support from the Texas Child Mental Health Care Consortium and fees to be a consultant and workshop presenter with Jenna Overbaugh LLC, The Knowledge Tree, and OCD Training School. Dr. Cervin receives research support from the Swedish Research Council for Health, Working Life and Welfare, the Kavli Foundation, the Lindhaga Foundation, Stiftelsen Clas Grochinskys Minnesfond, the Crown Princess Lovisa's Association, Region Skåne, Fonden för Psykisk Hälsa, and Skåne University Hospital's Foundations and Donations; and financial compensation from Springer for editorial work outside of the submitted work.

Ethics Statement

Ethical approval was granted by the Rogers Behavioral Health Institutional Review Board.

Conflicts of Interest

The authors declare no conflicts of interest.

Pinciotti, Caitlin M. , Thorsson Max, Horvath Gregor, and Cervin Matti. 2026. “Trauma‐Related Compulsions in Obsessive‐Compulsive Disorder: Functional Insights From Factor Analysis,” Stress and Health: e70154. 10.1002/smi.70154.

Data Availability Statement

Due to data sharing restrictions at Rogers Behavioral Health, data sharing is not available.