Do they really care? Specificity of social support issues in hoarding disorder and obsessive–compulsive disorder

Abstract

Objectives

Unmet interpersonal needs may play a role in excessive emotional attachments to objects for people with hoarding disorder (HD). Previous research indicates that social support (but not attachment difficulties) may be specific to HD. The study aimed to evaluate social networks and support in HD relative to clinical controls with obsessive–compulsive disorder (OCD) and healthy controls (HC). The secondary aim was to explore the extent of loneliness and thwarted belongingness. Potential mechanisms for deficits in social support were also considered.

Design

A cross-sectional between-groups design was used to compare scores on measures in those with HD (n = 37); OCD (n = 31); and HCs (n = 45).

Methods

Participants completed a structured clinical interview by telephone (to assign diagnostic categories) followed by online questionnaires.

Results

Whilst individuals with HD and OCD both report smaller social networks than HC, lower levels of perceived social support appear to be specific to HD. The HD group also showed higher levels of loneliness and thwarted belonging compared to OCD and HC. No differences were found between groups for perceived criticism or trauma.

Conclusions

The results support previous findings of lower levels of self-reported social support within HD. Loneliness and thwarted belongingness also appear significantly elevated within HD compared with OCD and HC. Further research is required to explore the nature of felt support and belonging, direction of effect and to identify potential mechanisms. Clinical implications include advocating and promoting support systems (both personal supporters and professionals) for individuals with HD.

Practitioner Points

• A major part of the impact of hoarding on the person affected is the way it leads to social isolation and loneliness.
• The present study shows similarly low levels of social contact in OCD and hoarding, but those experiencing hoarding feel much less supported by those around them.
• Overall, there was evidence of social disconnection in those with hoarding relative both to community controls and people experiencing OCD.
• Ways of enhancing positive experiences of social support are likely to be helpful for those with hoarding issues and may require special attention to the way such support is perceived.

INTRODUCTION

Hoarding disorder (HD) is a condition defined by a persistent difficulty with over-acquiring and failure to discard possessions, typically due to a strong perceived need to save items and/or distress associated with discarding (American Psychiatric Association, 2013). It is estimated that HD affects 2.5% of the population in developed countries (Postlethwaite et al., 2019). Symptoms of HD typically develop before the age of 20 (Tolin et al., 2010). Difficulties commonly increase in severity over time, with clinically significant levels most likely to manifest in middle to late adulthood (Dozier et al., 2016) and with clinical presentation often occurring in later life (Thew & Salkovskis, 2016). Whilst previously categorized under the umbrella of obsessive–compulsive disorder (OCD) or obsessive–compulsive personality disorder (OCPD), HD is now classified as a distinct psychiatric diagnosis (American Psychiatric Association, 2013).

For individuals with HD, the most evident difficulties focus on the accumulation of belongings in ways that result in severe clutter and the obstruction of living areas to the point of being unusable. The extent of hoarding is linked to levels of distress and impairment (Timpano et al., 2013). Extreme clutter can lead to a range of related problems including unsanitary living conditions, high rates of evictions and fire hazards (Frost et al., 2008). It can adversely impact and threaten the health and safety of the individual, those around them and surrounding community (Frost et al., 2000, 2008). Compared to community controls, individuals with HD are more likely to have a physical health condition or comorbid mental health disorder and less likely to be married (Nordsletten et al., 2013). Greater clutter and hoarding symptomatology is also associated with a lower frequency of friends and family visiting the home (Davidson et al., 2020), possibly indicating links between HD and levels of social support.

Social support is broadly defined as the assistance and protection given to individuals and may refer to tangible help (e.g., financial aid) and intangible help (e.g., emotional assistance) (Langford et al., 1997). Social support can play a substantial role in the maintenance of mental well-being and in the prevention of and recovery from mental health difficulties (Mezzina et al., 2006; Wang et al., 2018). Langford et al. (1997) describe three antecedents to social support: a social network (a structure of people around the individual), social embeddedness (the connectedness to those in their network) and social climate (whether there is an atmosphere of helpfulness and protection). It is possible that individuals may have large networks but do not feel connected to or helped by them. Individuals may experience social isolation (an objective lack of interactions with others or the wider community) and loneliness (a subjective feeling of the absence of a social network or a companion), both of which are associated with poorer mental health outcomes (Leigh-Hunt et al., 2017). Further, Joiner's interpersonal theory states that the presence of two interpersonal constructs, thwarted belongingness (TB; ‘I am alone’) and perceived burdensomeness (PB; ‘I am a burden’) increases suicide risk (Van Orden et al., 2010).

According to cognitive-behavioural models of HD, abnormal object attachment may be an attempt to compensate for unmet interpersonal needs and/or attachment difficulties (Frost & Hartl, 1996). It may be that those with HD develop abnormal attachments to possessions as a compensatory source of connectedness, with inanimate objects being experienced as more reliably providing comfort and safety (Yap & Grisham, 2019). However, a recent systematic review found that whilst interpersonal attachment and negative early family environment (e.g., parental warmth) showed some association with greater HD symptomatology, this was associated with broader clinical symptomatology rather than specific to HD (Chia et al., 2021). Indeed, Barton et al. (2021) found that individuals with HD did not differ from OCD controls in terms of impaired parental bonding, attachment insecurity and loneliness; both groups were impaired relative to community controls. However, these authors did find that whilst individuals with HD and OCD had comparably lower numbers of close relationships than community controls, individuals with HD reported feeling significantly less supported.

Chen et al. (2022) found in a community sample recruited online that HD symptoms were correlated with decreased social support and reduced social motivation, even after controlling for depressive symptoms. Another community sample found that loneliness and object attachment were positively associated with HD symptoms (Yap & Grisham, 2019). Poor social support may also form a barrier to recovery as it was associated with increased likelihood of dropping out of treatment in older adults with HD (Weiss et al., 2020).

Reduced social support may be particularly apparent in HD due to the nature of extreme clutter and associated health and safety risks. Interpersonal conflict with family, neighbours and housing providers is common, with stigma surrounding accumulation of possessions potentially contributing to defensiveness and conflict (Bratiotis et al., 2019). Reduced support may be associated with alienation from family members who, typically out of concern, try clear the homes of individuals with HD (Kim et al., 2001; Tolin et al., 2008). Yet individuals with HD report that the clearing of clutter can lead to feelings such as loss and violation so may damage relationships (Taylor et al., 2019). There may also be a negative feedback loop between reduced social support and increased clutter as described by the compensatory model of HD (Yap & Grisham, 2019). This suggests that possessions are an attempt to compensate for interpersonal needs such as loneliness, but this may unintentionally perpetuate interpersonal problems and further increase isolation, object attachment and possessions.

The evidence base indicates that individuals with HD may experience reduced social support due to poorer quality relationships and/or experiences of relationships as threatening or dangerous and therefore the quality of relationships and trauma are important possible mechanisms to investigate further in future research. It is possible that family environments may be a mechanism for differences in support. Expressed emotion (EE) refers to the quality of a family context and caregiving relationship (Kuipers, 1979). High EE is linked to poorer outcomes in mental health conditions including OCD, depression and schizophrenia (Bebbington & Kuipers, 1994; Chambless & Steketee, 1999; Hooley & Teasdale, 1989). Indeed, family members of individuals with HD often report high levels of frustration, conflict, hostility, relationship impact (including rejection and breakdown) and loss of ‘normal’ family life (Büscher et al., 2014; Park et al., 2014; Tolin et al., 2008; Wilbram et al., 2008). Family conflict and low family competence (defined as poor cohesion, communication and co-operative problem-solving) is associated with increased object attachment and HD symptoms (Davidson et al., 2020).

This may also link to wider social discourses and stigma, as HD was associated with a more negative public perception than conditions such as OCD (Chasson et al., 2018). Individuals with HD were assigned more blame for their condition compared to those with severe mental illness and OCD. Further, participants who had a friend or loved one with HD reported greater disdain and blame towards individuals with HD. Self-criticism and shame have been positively associated with HD symptoms (Chou et al., 2018) suggesting that perceived criticism may feature in HD.

It is also possible that individuals with HD utilize less social support as they experience relationships as more threatening. Individuals with HD symptoms often report increased rates of physical assault and sexual trauma (Przeworski et al., 2014; Shaw et al., 2016) and interpersonal difficulties, and violence has been associated with symptom onset or exacerbation (Tolin et al., 2010). Individuals with HD report greater exposure to a range of traumatic and stressful life events compared with individuals with OCD (without HD) and non-clinical controls (Landau et al., 2011).

As social support, loneliness and TB represent related but distinct constructs (Berkman et al., 2000; Van Orden et al., 2012) their role within clinical populations including HD should be investigated. As there is evidence for greater levels of criticism and trauma amongst individuals with HD, these could be potential mechanisms.

Aims and hypotheses

In previous work, individuals with HD reported a comparably reduced quantity of social support as individuals with OCD, and there was an indication that in HD only there was also a reduction in the perceived quality of support (Barton et al., 2021). This current study aimed to replicate and extend that finding, seeking to understand the role of present-day relationships. The choice of OCD as a control group relates to the comparable impact of the disorder and the similar patterns of comorbidity, allowing the identification of the extent to which any differences found could be considered to be HD-specific factors.

This study will include a more comprehensive assessment of available support, including the extent of social networks (size of network, length known and frequency of contact) and support (emotional support and aid). Measures of loneliness and TB were added to evaluate the extended impact of feeling unsupported, and two potential mechanisms were considered (perceived criticism and self-reported trauma), as candidate variables for feeling unsupported.

1. Primary hypothesis: Those in HD and OCD groups will both have smaller social networks relative to HC, however, HD will report lower levels of perceived support compared to OCD and HC.
2. Secondary hypothesis: Individuals with HD will report greater loneliness and thwarted belongingness than those with OCD and HC.
3. Exploratory variables (possible mechanisms) were perceived criticism and trauma

METHOD

Design

The study had a cross-sectional, between-group design (HD, OCD and HC). Data collection took place from August 2021–April 2022. Ethical approval was obtained from the Central University Research Ethics Committee (R74797/RE001).

Sample size

Power analysis was conducted using G*Power (Faul et al., 2007) to determine sample size. For a three-group independent measures ANOVA (with 80% power, α = .05), 111 participants (37 per group) were required to detect a medium effect size of .4. This was informed by previous findings (Barton et al., 2021).

Participants

Participants were recruited from the general population. Inclusion criteria included adults (≥18 years with no upper limit) based in the United Kingdom. Participants were excluded if they were unable to speak English or provide informed consent. Participants in the HC group were excluded if they meet criteria for obsessive, compulsive or hoarding difficulties (assessed through SCID and measures).

Procedure

Service user involvement

Key study documentation and measures were piloted with six individuals representing each of the three groups to consider clarity, ease of understanding and participant burden. Individuals were recruited via email to individuals signed-up to be service user representatives within the University department and through personal contacts. All participants were offered reimbursement for their time in accordance with University guidelines. Overall, feedback indicated that the survey was clear and accessible. Following feedback clarifying points were made, for example, specifying if participants were being asked how much they felt supported in general or regarding specific difficulties. Given feedback around the length of the questionnaires and ease of use, the 2-item PCM and OCI-short form were selected.

Recruitment and procedure

Participants were recruited through voluntary organizations (e.g., Hoarding Disorders UK and OCD UK) and through social media (e.g., Twitter). The HC group were recruited by social media and snowball sampling via individuals in the clinical groups to help recruit participants with similar demographics. Individuals who had previously participated in research at the University and who had consented to be contacted about future studies that might be of interest were emailed study details.

Participation included a telephone call that entailed obtaining oral consent and completing the HD and OCD subsections of the Structured Clinical Interview for DSM-5 (SCID; First et al., 2015) to determine group membership (details below). SCID interviews were conducted by VE, a final year Clinical Psychologist, who was trained and supervised by the other authors who themselves have extensive experience of its use. Individuals were assigned a participant identification number and either emailed a link (to Qualtrics software) or sent a paper copy (as requested) of self-report demographic and psychometric measures taking approximately 30–40 minutes. Participants completed a consent form at the start of the questionnaire and a debrief page was provided at the end.

Measures

Diagnostic screening was conducted using the structured clinical interview for DSM-5 (SCID; First et al., 2015), a semi-structured interview for making major DSM-5 Axis I diagnoses. The HD and OCD subsections provided a priori criteria to determine group assignment (further supported by minimum required scores decided a priori on OCD and HD measures below).

The following measures were used to assess sample characteristics to ensure groups were comparable in terms of non-specific but likely relevant variables (i.e., mood and impairment) but differed on specific psychopathology measures (i.e., OCD and HD): the Patient Health Questionnaire-9 (PHQ-9; Kroenke et al., 2001), Generalized Anxiety Disorder-7 (GAD-7; Spitzer et al., 2006), Work and Social Adjustment Scale (WSAS; Mundt et al., 2002), Savings-Inventory Revised (SIR; Frost et al., 2004), Obsessive Compulsive Inventory-Revised (OCI-R) and Beliefs About Hoarding Questionnaire Revised (BAH; Gordon et al., 2013).

Measures of social support, loneliness and belonging included the Revised Norbeck Social Support Questionnaire (NSSQ-R; Norbeck et al., 1981), Medical Outcomes Study Social Support Survey (MOS-SSS; Sherbourne & Stewart, 1996), UCLA Loneliness Scale Version 3 (UCLA-3; Russell, 1996) and Interpersonal Needs Questionnaire-Revised (INQ-R; Van Orden et al., 2012). To explore possible mechanisms, the Perceived Criticism Measure (PCM; Hooley & Teasdale, 1989; Masland & Hooley, 2015) and Life Events Checklist for DSM-5 (LEC-5; Weathers et al., 2013) were selected. See Table 1 for further details of the measures and their psychometric properties.

TABLE 1. Measures and their psychometric properties.

Domain	Measure
Diagnostic screening	Structured Clinical Interview for DSM-5 (SCID; First et al., 2015): A semi-structured interview for making major DSM-5 Axis I diagnoses. It provides criteria to allow judgements on whether participants met the criteria for HD and OCD subsections.
Psychopathology	The Patient Health Questionnaire-9 (PHQ-9; Kroenke et al., 2001): A widely used, valid and reliable 9-item questionnaire to measure depression. The PHQ9 shows excellent internal reliability (Cronbach's α = .92) and test–retest reliability (Kroenke et al., 2001). Within this study, the PHQ-9 also showed excellent internal reliability (Cronbach's α = .92).
	Generalized Anxiety Disorder-7 (GAD-7; Spitzer et al., 2006): A widely used 7-item questionnaire using a four-point scale to measure the severity of anxiety symptoms over the past two weeks. The GAD-7 demonstrates excellent internal consistency (Cronbach's α = .94) and good test–retest reliability (interclass correlation = .83; Spitzer et al., 2006). Within this study, the GAD-7 showed high internal reliability (Cronbach's α = .91).
	Work and Social Adjustment Scale (WSAS; Mundt et al., 2002): A 5-item measure of an individual's ability to carry out day-to-day tasks on an 8-point scale. It is a simple, reliable, and valid way of measuring impaired functioning. The WSAS demonstrates good-excellent internal consistency (Cronbach's α = .70–.94) and acceptable test–retest reliability (interclass correlation = .73; Mundt et al., 2002). Within this study the WSAS showed good internal reliability (Cronbach's α = .84).
	Savings-Inventory Revised (SIR; Frost et al., 2004): A 23-item measure of HD with three subscales: difficulty discarding, clutter and acquisition. The SI-R demonstrates excellent internal consistency (Cronbach's α = .94) and good test–retest reliability (r = .86; Frost et al., 2004). Within this study the SIR showed excellent internal reliability (Cronbach's α = .98).
	Obsessive Compulsive Inventory-Revised (OCI-R; Foa et al., 2002): A short version of the Obsessive Compulsive Inventory (Foa et al., 1998) containing 18 items and 6 subscales. Test–retest reliability has been demonstrated for OCD patients (r = .74–.91) and non-anxious controls (r = .58–.87). It has good diagnostic ability for OCD (Abramowitz & Deacon, 2006). The OCI-R total demonstrates excellent internal consistency (Cronbach's α = .90; Foa et al., 2002). In the current study, OCI-R total demonstrated excellent internal reliability (Cronbach's α = .92).
	Beliefs About Hoarding Questionnaire Revised (BAH; Gordon et al., 2013): A 28-item measure to assess beliefs and experiences characteristic of HD incorporated as a descriptive measure of the extent of differences in specific HD beliefs between groups. It includes subscales for HD motivated by harm avoidance, material deprivation, and attachment disturbance. The BAH demonstrates excellent internal consistency (Cronbach's α = .96) and good test–retest reliability for individuals with HD (r = .83; Gordon et al., 2013). In the current study, BAH demonstrated excellent internal reliability (Cronbach's α = .97).
Social support	Revised Norbeck Social Support Questionnaire (NSSQ-R; Norbeck et al., 1981): Participants list significant people in their life and rate them on various categories. Total network includes (a) number of supporters, (b) duration of relationship and (c) frequency of contact. Total support comprises (a) affect (how much they feel liked/loved and respected/admired), (b) affirmation (how much they feel agreed with and can confide in supporters), and (c) aid (immediate and longer-term help). It shows excellent test–retest reliability over one-week period (r = .86–.92; Norbeck et al., 1981). As recommended, averaged scores using the NSSQ-R with the use of n/a response options were used (Gigliotti & Samuels, 2020). In the current study, NSSQ-R network subscale demonstrated excellent internal reliability (Cronbach's α = .97) as did the support subscale (Cronbach's α = .98).
	Medical Outcomes Study Social Support Survey (MOS-SSS; Sherbourne & Stewart, 1996) A 20-item measure of social support. Items are divided into four subscales: perceived adequacy of tangible, emotional and affectionate support as well as positive social interaction. Each item is rated on a 5-point Likert scale to indicate how often the respondent receives support. One additional question ascertains the respondents' number of close relatives and friends. The MOS-SSS demonstrates good internal consistency (α = .97), and good test–retest reliability over a one-year period (r = .78). In the current study, the MOSS-SSS demonstrated excellent internal reliability (Cronbach's α = .97; Sherbourne & Stewart, 1996).
Loneliness	UCLA Loneliness Scale Version 3 (UCLA-3; Russell, 1996): A 20-item measure related to experiences of loneliness where participants rate their feelings using a 4-point Likert scale. It has a good internal reliability and convergent validity with other measures of loneliness. It demonstrates good internal consistency (α = .89–.94) and test–retest reliability over a one-year period (r = .73; Russell, 1996). In the current study, the scale demonstrated excellent internal reliability (Cronbach's α = .96).
Thwarted belongingness and perceived burdensomeness	Interpersonal Needs Questionnaire-Revised (INQ-R; Van Orden et al., 2012): A 15-item measures of the extent to which individuals believe their need to belong is met (i.e., thwarted belongingness; INQ-R-TB) and the extent to which they perceive themselves to be a burden on others (i.e., perceived burdensomeness; INQ-R-PB). Good internal reliability has been established for TB items (Cronbach's α = .85) and PB items (Cronbach's α = .89; Van Orden et al., 2008). The current study demonstrated excellent internal reliability for TB items (Cronbach's α = .94) and PB items (Cronbach's α = .91).
Perceived criticism	The Perceived Criticism Measure (PCM; Hooley & Teasdale, 1989; Masland & Hooley, 2015): Participants rate each important person to them on a scale from 1 (not at all critical) to 10 (very critical) and the extent they get upset when the important person criticizes them from 1 (not at all) to 10 (extremely). Perceived criticism has been identified as the most important element of EE (Hooley & Parker, 2006). The PCM is a straightforward measure to administer with good concurrent and predicative validity (Masland & Hooley, 2015).The test–retest reliability for the criticism portion of the PCM was shown to be moderately high (r = .75) over approximately five months (Hooley & Teasdale, 1989). In the current study, the scale demonstrated excellent internal reliability for perceived criticism (Cronbach's α = .97) and for perceived upset (Cronbach's α = .98).
Trauma	Life Events Checklist for DSM-5 (LEC-5; Weathers et al., 2013): A self-report measure of traumatic events. It assesses exposure to 16 events known to potentially result in PTSD or distress with an additional item assessing any other extraordinarily stressful event not captured by the previous items. The scale demonstrates good test–retest reliability over a one-week period (r = .82). For the purposes of this study, LEC-5 total will refer to the total of the following subscales: happened to me, witnessed it, learnt about it happening to close family member or friend, part of my job.

Data analysis

Statistical analyses were performed using IBM SPSS (version 28) and interpreted with reference to .05 significance levels. There were no missing data; most participants (118; 98%) completed the questionnaires online, which required all items to be completed.

Although the main allocation to group was through the SCID interview, measures of psychopathology were examined to ensure that individuals were appropriately assigned to diagnostic groups. This was achieved by looking at the spread of scores between and within groups (i.e., to ensure those assigned to HD and OCD groups had sufficiently high number of associated traits, that controls were not showing clinical levels of HD and OCD and that there was no overlap between groups). Data were screened for deviations from normality. Chi-square analysis was used to determine differences between the groups on categorical demographic variables. Continuous data were assessed for assumptions of normality. Analysis of variance (ANOVA) was employed to compare means across the three groups. Where within-subjects effects showed evidence of intercorrelation, this was adjusted using a Greenhouse–Geisser (G–G) correction. Where indicated, multiple comparisons used the LSD test; where the Levene test indicated problems with homogeneity of variance, Dunnets T3 was used instead.

RESULTS

Sample characteristics

A total of 132 individuals completed the telephone screening interview and 120 completed questionnaires. Six individuals who completed measures met criteria for both OCD and HD and were not included within the analyses. One individual initially assigned to the OCD group was excluded from analysis due to a low OCI-R score (scoring 11). Therefore, data analysis was on 113 participants: HD = 37, OCD = 31 and HC = 45.

Demographic and descriptive variables

All participants were aged between 23 and 84 years (M = 50.38, SD = 16.54). Participants were predominantly female (79.6%), identified as Caucasian (89.4%) with secondary school education or higher (100%). Fifty-five participants were married or living as a couple (48.7%) and fifty-six were employed (49.6%). Supporting Information: Appendix S1.

There was a significant main effect of group on age, F_(2,110) = 15.53, p < .001. Multiple comparisons indicated that the OCD group was significantly younger than HD and HC (which did not significantly differ). A Chi-square test was performed to assess the association between group and gender. No significant association was observed, ${\chi }_{\left(4\right)}^2$ = .018, p > .05. There was no association between group and employment, ${\chi }_{\left(2\right)}^2$ = 4.82, p > .05 or being married/ living as a couple, ${\chi }_{\left(4\right)}^2$ = 2.94, p > .05. There was a significant association between the group and education, ${\chi }_{\left(4\right)}^2$ = 17.90, p = .001. Individuals in OCD and HC groups generally had higher education. There were group differences in past and current mental health support (see Table 2).

TABLE 2. Characteristics by group.

	HD (n = 37)		OCD (n = 31)		HC (n = 45)		Main effect
	Mean	SD	Mean	SD	Mean	SD
Age	58.76^a	10.19	38.87^b	11.05	51.42^a	19.37	F(2,110) = 15.53, p < .001
	N	%	N	%	N	%
Gender							${\chi }_{\left(4\right)}^2$ = .018, p = .991
Female	30	81.1	24	77.4	36	80.0
Male	7	18.9	6	19.4	9	20.0
Other	0	0	0	0	1†	.9
Ethnicity							‡
Caucasian	33	89.2	28	90.3	40	88.9
Other	4	10.8	3	9.7	5	11.1
Education							${\chi }_{\left(4\right)}^2$ = 17.90 p = .001
Secondary school	17	45.9	8	25.8	5	11.1
University degree	15	40.5	10	32.3	17	37.8
Postgraduate degree	5	13.5	13	41.9	23	51.1
Occupational status							${\chi }_{\left(2\right)}^2$ = 4.82, p = .090
Employed	14	37.8	20	64.5	22	48.9
Other	23	51.1	11	35.5	23	62.2
Civic status							${\chi }_{\left(4\right)}^2$ = 2.94, p = .237
Married/living as a couple	14	37.8	18	58.1	23	51.1
Other	23	62.2	13	41.9	22	48.9
Past mental health support							‡
Yes	33	89.2	29	93.5	25	55.6
No	4	10.8	2	6.5	20	44.4
Currently receiving professional mental health support							${\chi }_{\left(2\right)}^2$ = 16.82, p < .001
Yes	13	35.1	18	58.1	6	13.3
No	24	64.9	13	41.9	39	86.7
Currently receiving talking therapy							${\chi }_{\left(2\right)}^2$ = 18.10, p < .001
Yes	8	21.6	15	48.39	3	6.6
No	29	78.37	16	51.6	42	93.3
Currently taking medication for mental health							${\chi }_{\left(2\right)}^2$ = 16.10 p < .001
Yes	20	54.1	16	51.6	7	15.6
No	17	45.9	15	48.4	38	84.4

• Note: Figures that share a superscript are not significantly different on multiple comparisons, those that do not share a subscript are statistically different.
• ^† Item excluded to complete Chi-square analysis.
• ^‡ Cell count too small to complete Chi-squared analysis.

As shown in Table 3, there were group differences on measures of depression, anxiety, and general functioning (WSAS). Clinical groups did not differ and scored significantly higher than HC.

TABLE 3. Measures of psychopathology (with multiple comparisons used).

	HD (n = 37)		OCD (n = 31)		HC (n = 45)
	Mean	SD	Mean	SD	Mean	SD	Main effect of group
PHQ-9	10.76^a	6.36	10.03^a	8.09	3.44^b	4.07	F(2,110) = 17.63, p < .001
GAD-7	7.76^a	5.11	10.58^a	6.11	3.22^b	3.69	F(2,110) = 21.76, p < .001
WSAS	21.24^a	10.05	18.29^a	10.96	5.00^b	6.64	F(2,110) = 36.92 p < .001
OCI-R total (-hoarding)	21.38^a	9.42	31.74^b	14.25	6.64^c	6.76	F(2,110) = 58.72, p < .001
OCI-R: washing	1.16^a	1.84	5.58^b	4.80	.47^a	1.04	F(2,110) = 33.63, p < .001
OCI-R: checking	2.84^a	2.57	5.25^b	3.75	1.07^c	1.67	F(2,110) = 22.69, p < .001
OCI-R: ordering	4.16^a	2.90	5.65^a	4.27	1.80^b	2.61	F(2,110) = 13.74, p < .001
OCI-R: obsessing	3.81^a	2.91	8.74^b	2.61	1.5^c	1.93	F(2,110) = 78.43, p < .001
OCI-R: neutralizing	.97^a	1.44	3.35^b	3.70	.49^a	1.39	F(2,110) = 15.70, p < .001
OCI-R: hoarding	8.43^a	2.51	3.16^b	3.25	1.27^c	1.64	F(2,110) = 89.96, p < .001
SIR total	60.92^a	15.60	23.81^b	19.47	12.31^c	10.70	F(2,110) = 110.15, G–G p < .001
SIR: clutter	24.70^b	8.10	7.42^a	7.81	4.47^a	6.39	F(2,110) = 84.22, G–G p < .001
SIR: DD	20.24^a	8.10	9.06^b	7.51	4.18^c	4.58	F(2,110) = 81.46, G–G p < .001
SIR: EA	15.97^a	5.63	7.32^b	6.30	3.67^c	3.278	F(2,110) = 61.91, G–G p < .001
BAH-HA	3.28^a	3.28	3.28^a	3.12	1.41^b	.74	F(2,110) = 13.48, p < .001
BAH-MD	5.62^a	2.26	3.19^b	2.70	1.67^c	.99	F(2,110) = 39.23, p < .001
BAH-AD	5.21^a	2.51	3.57^b	2.62	1.98^c	2.51	F(2,110) = 23.24, p < .001

• Note: Figures that share superscript are not significantly different on multiple comparisons, those that do not share a subscript are statistically different.
• Abbreviations: BAH-AD, attachment disturbance; BAH-HA, harm avoidance subscale; BAH-MD, material deprivation; GG, Greenhouse Geisser adjustment applied; SIR-DD, difficulty discarding; SIR-EA, excessive acquisition.

As the OCI-R contained a hoarding subscale, a mixed-model ANOVA for the other five subscales only was conducted in order to reduce the risk of artificially inflating scores on obsessionality in the hoarding group: 3 (diagnostic group) × 5 (subscale). There was a significant main effect of diagnostic group on the overall level of obsessional symptoms, F_(1,110) = 345.24, p < .001. There was a significant main effect of subscale, F_(4,440) = 30.81, p < .001. These effects were modified by a significant group × subscale interaction, F_(8,440) = 6.30, p < .001. A follow-up simple main effects analysis was therefore carried out on each scale (see Table 3), indicating that with the exception of the ordering subscale, OCD > HD (p > .05). The separate one-way ANOVA between groups on the OCI-R hoarding subscale indicated that HD showed highest levels followed by OCD then HC.

A mixed-model ANOVA for the SIR subscales was carried out: 3 (diagnostic group) × 3 (subscale). There was a significant main effect of diagnostic group, F_(1,110) = 504.69, p < .001 and subscale, F_(1.70,186.4) = 16.19, G–G p < .001. These effects were modified by a group × subscale interaction, F_(3.4,186.4) = 12.13, G–G p < .001. Simple main effects analysis indicated that HD had significantly higher clutter compared to OCD and HC (with no significant difference between OCD and HC). There were significant differences between groups for difficulty discarding and excessive acquisition, with HD reporting the greatest difficulties followed by OCD then HC.

A mixed-model ANOVA for the BAH subscales was carried out: 3 (diagnostic group) x3 (subscale). There was a significant main effect of diagnostic group, F_(2,110) = 28.58, p < .001 and subscale, F_(1,110) = 39.50, p < .001. These effects were modified by a group x subscale interaction, F_(2,110) = 11.49, p < .001. A follow-up simple main effects analysis was carried out on each scale. HD and OCD showed equal and significantly higher levels of harm avoidance than HC. There was a significant difference between all groups on material deprivation and attachment disturbance, with HD reporting the greatest difficulties followed by OCD then HC.

Primary outcomes

Social network

A one-way ANOVA for the total network subscale of the NSSQ-R was conducted to see if there was a difference in size of social network between each group (Table 4). There was a significant main effect of group, F_(2,110) = 6.39, p = .002. Planned comparisons revealed that, as predicted, HD and OCD had significantly similar and smaller social networks than HC (Figure 1). Analysis of subscales revealed that HD and OCD both had similar and reduced numbers of supporters and had known them less time than HC. HD had significantly less contact with supporters than the HC group (with no significant difference between OCD and HD or HC). Further details about differences between groups on subscales within the total network subscale (i.e., number of supporters, length known and frequency of contact) are displayed in Table 4.

TABLE 4. Outcome measures.

	HD (n = 37)		OCD (n = 31)		HC (n = 45)		Main effect of group
	Mean	SD	Mean	SD	Mean	SD
NSSQ-R- total network	83.05^a	53.38	87.23^a	48.60	114.02^b	60.31	F(2,110) = 3.82, p = .025
No. supporters	8.97^a	5.94	9.16^a	5.25	12.09 ^b	6.67	F(2,110) = 3.38, p = .038
Length known	41.03^a	27.20	41.68^a	24.78	57.29 ^b	30.84	F(2,110) = 4.38, p = .015
Freq. of contact	33.05^a	20.87	36.39^ab	19.03	44.64^b	23.35	F(2,110) = 3.9, p = .045
NSSQ-R- total support	14.34^a	4.44	17.08^b	3.40	17.30^b	4.05	F(2,110) = 6.39, p = .002
Emotional	10.60^a	3.04	11.76 ^ab	2.59	12.29^b	2.68	F(2,110) = 3.82, p = .025
Aid	3.73^a	1.86	5.31^b	1.53	5.01^b	1.73	F(2,110) = 6.39, p < .001
MOSS-SSS – supporters	4.32^a	3.20	5.97^a	4.45	8.80 ^b	5.79	F(2,110) = 9.62, p < .001
MOSS-SSS – total support	37.73^a	16.47	54.03^b	18.70	54.84^b	18.56	F(2,110) = 10.91, G–G p < .001
Emotional	17.24^a	7.77	23.10^b	8.42	23.91^b	7.37	F(2,110) = 8.32, G–G p < .001
Tangible	6.68^a	5.03	11.16^b	4.71	10.44^b	5.33	F(2,110) = 8.16, G–G p < .001
Affection	5.78^a	3.64	8.87^b	3.19	9.00^b	3.86	F(2,110) = 9.54, G–G p < .001
Positive	6.19^a	2.95	8.19^b	3.49	8.69^b	3.25	F(2,110) = 6.54, G–G p = .002
UCLA-3	34.65^a	11.60	27.77^b	14.05	19.73^c	12.31	F(2,110) = 4.37, p < .001
Perceived criticism –criticism	2.41^a	2.48	1.74^a	2.49	1.56^a	2.42	F(2,110) = 1.28, p = .281
Perceived criticism –upset	2.74^ab	1.75	3.42^b	1.70	2.17^a	1.68	F(2,109) = 4.95, p = .009
INQ-R-PB	8.70^a	7.87	7.13^a	8.58	1.33^b	3.29	F(2,110) = 13.90, p < .001
INQ-R-TB	29.22^a	13.06	20.68^b	13.38	14.04^c	12.61	F(2,110) = 12.50, p < .001
LEC – happened to me	1.59^a	1.36	1.10^a	1.22	1.59^a	1.36	F(2,110) = 3.03, p = .052

• Note: Figures that share superscript are not significantly different on multiple comparisons; those that do not share a subscript are statistically different.

FIGURE 1

Social network and social support by group as measured by the NSSQ-R. Note: Error bars represent standard errors.

Perceived social support

A one-way ANOVA for the total support subscale of the NSSQ-R was conducted to see if there was a difference in degree of perceived social support between each group. This showed a main effect of group on perceived support, F_(2,110) = 6.39, p = .002. Planned comparisons showed that HD reported less support than OCD and HC (which showed no significant difference from each other). Analysis of subscales showed a significant difference in emotional support between HD and HC groups (with no significant difference between OCD and HD or HC). HD reported receiving significantly less aid than OCD and HC (which showed no significant difference from each other). Further details about between-group differences on subscales within the total support subscale (i.e., emotional support and aid) are displayed in Table 4.

As this study aimed to replicate findings by (Barton et al., 2021) which utilized the MOSS-SSS, this was also analysed here. There was a significant main effect of group, F_(2,110) = 9.62, G–G p < .001. Planned comparisons showed that HD and OCD did not differ in terms of close friends and relatives and both were significantly lower than HC. A one-way ANOVA showed a significant difference in support between groups, F_(2,110) = 10.91, p < .001. Planned comparisons found that HD reported significantly less support than OCD and HC groups (which did not significantly differ from each other). This finding was consistent between groups for all subscales (i.e., emotional, tangible, affection and positive support subscales as detailed Table 4).

Secondary outcomes

Loneliness

A one-way ANOVA was conducted to investigate whether there was a difference in loneliness between groups measured by the UCLA-3. This showed a significant main effect of group, F_(2,110) = 14.37, p < .001. Planned comparisons revealed statistically significant differences between all groups. As predicted, HD individuals were the loneliest, followed by OCD then HC (Figure 2).

FIGURE 2

Reported loneliness by group as measured by the UCLA-3. Note: Error bars represent standard errors.

Thwarted belongingness

A one-way ANOVA was conducted to see if there was a difference in thwarted belongingness between groups on the TB subscale of the INQ-R. There was a significant main effect on group, F_(2,110) = 12.50, p < .001. Post-hoc analysis showed a statistically significant difference in TB between all groups. HD reported greatest TB, followed by OCD, then HC.

As PB was measured as part of the INQ-R, the PB subscale of the INQ-R was also analysed. A one-way ANOVA was also conducted to see if there was a difference in PB between group. There was a significant main effect of group, F_(2,110) = 13.90, p < .001. Post-hoc analysis showed that HD and OCD had statistically similar and higher PB than HC.

Exploratory analyses

Perceived criticism

A one-way ANOVA was conducted to see if there was a difference in perceived criticism between groups. There was no significant main effect of perceived criticism (p = .281). The measures also included a separate question asking how upset individuals felt when criticized. There was a significant main effect of upset on group, F_(2,109) = 4.95, p = .009. Post-hoc analysis showed that the OCD group reported significantly greater upset when criticized compared with HD and HC (with no significant difference between HD and HC).

Trauma

A one-way ANOVA was conducted to see if there was a difference in trauma between each group as measured by the LEC-5. There was no significant main effect on total trauma (p = .060) or ‘happened to me’ subscale between groups (p = .052). Therefore, no further analyses were conducted.

DISCUSSION

This study investigated the role of social support in HD relative to OCD as a clinical control group and healthy controls. We sought to replicate previous findings that individuals with HD experience reduced social support (Barton et al., 2021) through a more comprehensive evaluation of individual's support network and felt social support. These findings were replicated; those with HD and OCD both had smaller social networks than HC, but reduced support was specific to HD. This included significant differences in reported emotional support and aid. The HD group reported significantly higher loneliness and thwarted belongingness relative to both OCD and HC. Interestingly, despite indications from the literature base, this study found no significant difference in perceived criticism or trauma between groups indicating that other mechanisms warrant future research.

Findings from Barton et al. (2021) were replicated on the original MOSS-SSS and also observed using the NSSQ-R which is a more comprehensive evaluation. Whilst Barton et al. (2021) found that OCD and HD experienced reduced (but comparable) loneliness, the present study found HD to have significantly greater loneliness. It is possible that completing the NSSQ-R induced greater feelings of loneliness. The present study occurred during the COVID-19 pandemic, and it is possible that lockdowns had a particularly exacerbating effect on individuals with HD. For example, it may be hypothesised that being confined to a home that is excessively cluttered may increase feelings of difference and loneliness. Individuals with HD may also be reliant on support groups (which were often halted during the pandemic) for a sense of connection. Yet challenges arising from the pandemic such as this are likely to have equally impacted the OCD group. Moreover, the present results are consistent with previous findings from community samples that HD symptoms are correlated with reduced social support (Chen et al., 2022) and increased loneliness (Yap & Grisham, 2019). It builds upon studies showing associations between HD and poorer social functioning that used demographic variables as a proxy for social support (Archer et al., 2019).

Previous research has found differences in relationship status between HD and OCD and HC (Barton et al., 2021). In this study, HD did not differ significantly in terms of being married/living as a couple. However, in both the current study and Barton et al. (2021) HD report reduced social support compared to OCD and HC, so partnership appears unlikely to account for reduced social support in HD.

Previous studies suggest increased trauma within HD which was not found in this study. Studies typically report correlations between increased trauma and increased hoarding symptoms (Przeworski et al., 2014; Shaw et al., 2016) rather than between-group comparisons established by a SCID as in this study. Trauma may increase certain HD symptoms but may not be specific to HD. It is also possible our sample is not representative (e.g., individuals with a trauma history may be less likely to participate in studies with a telephone interview, or as the sample had higher partnership rates this may correspond to reduced adverse relationships). It could also be that the LEC did not capture HD-related trauma (e.g., sudden material deprivation) so future research should consider alternative measures.

Strengths and limitations

This study utilized the SCID to assign individuals to diagnostic groups and included a comparison of HD with comparably impaired clinical controls as well as HCs. It included a comprehensive measure of social support which includes consideration and rating of all listed individuals.

The recruitment method (samples of convenience from social media and personal contacts including other trainee research) may have impacted the representativeness of the groups. Whilst paper questionnaires were available to those without internet access, most of the recruitment was online (especially as support groups were no longer face-to-face due to COVID-19). The use of the SCID means validity of groupings was strong but the recruitment strategy likely impacted who was aware of the study and volunteered. However, this would be expected to be comparable across groups so is unlikely to account for group differences. The cross-sectional nature of the study means, of course, that causal inferences cannot be drawn. The sample predominantly consisted of well-educated Caucasian females limiting generalizability. A more diverse sample is required to increase representation. Exploration of how support and belonging may vary between protected characteristics (e.g., culture) may indicate important protective factors or barriers to feeling supported. The sample was also self-selected, with some clinical participants recruited via support groups. Individuals attending support groups may do so as they have smaller social networks and support. However, it is also possible that these individuals are more open to support. Future research should record whether individuals belong to support groups and experience professional contacts as supportive.

There were differences between groups which may have influenced the dependent variables. Follow-up correlational analyses indicated that age had little impact on key dependent variables. Whilst the OCD group was younger, this had a minor role on HD having comparatively reduced social support and increased loneliness and TB.

Education also had little impact on social support. Interestingly, education accounted for high rates of variance for loneliness and TB. This might strengthen the notion that these are separate constructs, as education may influence the experience of loneliness and thwarted belongingness (but not receiving support). The samples were highly educated, particularly the OCD and HC groups. Future research with similar level of education between groups (and greater spread of education throughout groups) would help to understand the role of education. Group differences concerning receipt of professional help (talking therapies and medication) are also of note, especially the higher rates of talking therapy in OCD compared to HD which may buffer reduced support and loneliness. Qualitative studies may be helpful in exploring thoughts and experiences around these observed differences.

Theoretical implications

This study highlights features apparent in both clinical groups (e.g., smaller network size) and those in HD only (e.g., low perceived support, high loneliness and thwarted belongingness). Salkovskis and Forrester (2002) differentiate between disorder-specific and disorder-relevant constructs; that is, variables which uniquely occur in a particular disorder (disorder specific) versus those occurring across disorders but which may be relevant to both. Characteristics specific to HD would occur more in HD compared to individuals with other disorders but may not be unique to HD. As well as being specific, they need to be relevant to the formation or maintenance of difficulties. To consider whether variables are specific and relevant, Salkovskis and Forrester (2002) promote a theory-driven approach. Therefore, given that differences in perceived support were found to be specific and fit with the compensatory model of HD (Yap & Grisham, 2019), experimental studies are required to assess whether this is a maintaining factor in HD.

Social support, loneliness and TB represented related but distinct constructs (Berkman et al., 2000; Van Orden et al., 2012). In their model, Berkman et al. (2000) describe how social networks provide opportunities for social support and social engagement which influence inner needs or states such as TB. The model also describes other inner states such as self-efficacy and self-esteem. In their qualitative study, Taylor et al. (2019) found that barriers to engaging with social support included negative feelings about the self, particularly shame and guilt around clutter. Whilst perceived criticism was not significantly different between groups in this study, stigma and shame in HD are likely to be important factors. Berkman et al. (2000) describe broader social-cultural influences on social networks, and more negative public perception has been shown towards HD than conditions such as OCD (Chasson et al., 2018); participants who knew someone with HD reported greater disdain and blame. It is of note that measures in this study focused on listing individuals who provide support but there may be other people, not listed, who are unsupportive or critical.

As reduced social support is as a reliable predictor of suicidal ideation, attempts and completion (Van Orden et al., 2010), individuals with HD may have a heightened risk given the significantly higher rates of PB and TB. High rates of suicidality were found in treatment-seeking individuals with HD who were less likely to be married or have children and more likely to live alone (Archer et al., 2019). Further, recent meta-analyses demonstrate a high prevalence of lifetime suicidal ideation and lifetime suicide attempts in OCD and HD (13.5% and 24.1%, respectively, for suicide attempts; Pellegrini et al., 2020, 2021). In a population sample, HD severity was significantly associated with PB after controlling for negative affect, but there was no association with TB (Raines et al., 2016). The authors hypothesised that this may be explained by the negating effect of increased object attachment in providing a sense of belonging. In contrast, this study highlights the significance of TB in clinical samples and suggests that object attachment does not prevent loneliness. We note that in the current study, we replicated the findings of Gordon et al. (2013) in that individuals with HD scored more highly on beliefs about hoarding measure on subscales measuring fears of material deprivation and attachment disturbance. These differences warrant further investigation in terms of their relationship with perceived support.

Research implications

The key question is whether feeling less social support is a cause or a consequence of HD. Qualitative studies could help to explore social identity within HD and to understand how individuals feel (or do not feel) connected with and supported by others. Barriers to receiving help could be explored including whether individuals are not offered, do not ask for, or accept support, and the extent to which help refers to or is restricted by clutter. Exploration of the support felt from different multi-agency professionals could have clinical implications for staff training and co-production.

As social interactions are two-way, qualitative interviews with supporters of individuals with HD could help explore social support, connectedness and other interpersonal factors such as blame and conflict. Exploring the role of stigma (public stigma, stigma within personal and professional networks and self-stigma) could be valuable.

There is also scope for experimental-based designs in furthering understanding of social support. This could include utilizing induction procedures involving, for example, priming memories of supportive and unsupportive interactions in the context of in vivo discarding, with ratings of ability to discard those items as the dependent variable.

Clinical implications

Whether or not the results indicate that feeling unsupported and lonely is a cause or a consequence of HD, professionals should consider enabling individuals to access opportunities for support and connection, including but not confined to HD-based support groups. It would be important to establish if improving support and belonging leads to less reliance on objects and changes to acquiring and/or discarding. The results also have implications for helping supporters to provide and maintain support. This may involve psychoeducation, facilitating discussions around support, strengthening connectedness and if applicable promoting reconciliation. This may be especially important where individuals do not recognize difficulties with hoarding. If these individuals lose support and connectedness, this may reduce well-being and motivation to change. As such, promoting understanding and support for individuals with HD as well as their social networks is required.

CONCLUSION

This study found that individuals with HD and OCD have reduced social networks and that reduced social support, loneliness and thwarted belonging were specific to HD relative to OCD and HC. Further research is required to explore the nature of felt support, whether it forms a maintenance factor and identification of potential mechanisms. The findings have important implications for advocating and promoting individual and systemic support for HD.

AUTHOR CONTRIBUTIONS

Victoria Edwards was involved in conceptualization, methodology, data curation, investigation, formal analysis, writing—original draft and writing—review and editing. Paul M. Salkovskis was involved in conceptualization, methodology, formal analysis, supervision and writing—review and editing. Victoria Bream was involved in conceptualization, methodology, supervision and writing—review and editing.

ACKNOWLEDGEMENTS

We would like to acknowledge and thank everyone who dedicated their time to supporting and participating in this research.

FUNDING INFORMATION

This research did not receive any specific funds or grants.

CONFLICT OF INTEREST STATEMENT

None of the authors have any competing interests to declare.

Open Research

DATA AVAILABILITY STATEMENT

Research data are not shared. Due to the sensitive nature of the questions asked in this study, participants were assured that their data would remain confidential and would not be shared.