How do close others to those with anxiety feel about treatment? Development and validation of the Treatment Concerns Questionnaire-Close Others.

Background: The close others (e.g., family members, romantic partners) of people with anxiety and related disorders are typically involved in their treatment decisions. However, we know little about close others' attitudes towards and concerns about their loved one starting cognitive-behavioural therapy (CBT).
Methods: Study one surveyed close others of those with anxiety and related disorders (n = 33) about their concerns about their loved one starting CBT. Thematic coding was completed, and items were developed to reflect these themes, comprising a measure of treatment concerns in close others. Study two involved the administration of the novel measure to a larger sample (n = 287) to evaluate its structure, reliability, and validity.
Results: Close others endorsed having treatment concerns of moderate intensity. The final 17-item measure, the Treatment Concerns Questionnaire-Close Others (TCQ-C), has a robust four-factor structure, with internally consistent subscales including "Adverse Reactions", "Personal/Family Consequences", "Lack of Commitment", and "Ineffectiveness". The measure shows moderate correlations with treatment expectations (convergent validity) and small correlations with respondent distress (discriminant validity). Conclusions: The value of this measure for clinicians and future directions for research are discussed. Supplementary Information: The online version contains supplementary material available at 10.1007/s10608-022-10318-9.

Introduction

It is estimated that up to one quarter of the North American population experiences an anxiety or related disorder in their lifetime (Somers et al., 2006). Anxiety can be effectively treated through psychological interventions, such as cognitive-behavioural therapy (Otte, 2011). However, only a small minority of people with anxiety seek treatment (Roness et al., 2005), and many delay treatment seeking for many years (e.g., one study found an average delay of 9–23 years; Wang et al., 2005). Given the high economic burden of anxiety disorders in North America (Kessler & Greenberg, 2002; Koerner et al., 2004), it is essential to understand barriers to treatment seeking.

Those with anxiety and related disorders balance both approach and avoidance motivations in their decision to seek treatment (Kushner & Sher, 1989). For example, the distress and impairment from anxiety may be a motivation to seek treatment, while the stigma and cost of treatment may motivate treatment delay. In addition to personal factors, one’s social context , including the opinions and attitudes of family members and close others, also affect this decisional balance. Social context has been identified as an important factor in health outcomes (Repetti et al., 2002) and in medical decision making (Pescosolido, 1992). Family members are often involved in health care decisions, including decisions about anxiety and related disorders (e.g. Geffken et al., 2006; Thompson-Hollands et al., 2014). People with mental health difficulties commonly consult family members before health care professionals when considering treatment options (Maria Bermúdez et al., 2010; Henshaw et al., 2013) and one of the most frequent reasons for entering treatment is family influence/support or impairments in personal relationships as a result of mental health difficulties (Orford et al., 2006; Polcin & Beattie, 2007; Tsogia et al., 2001). People are sensitive to the anticipated reactions and attitudes of close others (e.g., family, significant others) when considering therapy (Leaf, Bruce, & Tischler, 1986; Earnshaw et al., 2013).

Although there are multiple treatment options for individuals with anxiety and related disorders, cognitive-behavioural therapy (CBT) is an evidence-based psychological treatment (for a recent meta-analysis, see Cuijpers et al., 2016) that is considered the “gold standard” for anxiety treatment (David et al., 2018). The impact of the family/interpersonal context on decisions about a psychological treatment like CBT, specifically, is an important and interesting area of study. As noted above, loved ones’ support could be an important reason to seek treatment, and positive attitudes in loved ones may encourage someone to enter treatment earlier than they might otherwise. In contrast, just as perceptions of social stigma are related to reduced help-seeking efforts (d= -0.27; Clement et al., 2015), the anticipation that close others may respond poorly to a decision to seek treatment could foster apprehension about therapy (Corrigan et al., 2014). As Corrigan and colleagues noted, cultural factors or family members’ own negative experiences with therapy may cause them to discourage their loved one from seeking therapy. Additionally, even if treatment is initiated, the attitudes of close others have the power to influence treatment engagement and success. For example, youth with behavioural challenges with parents who perceive mental illness as unchangeable or see treatment as inconvenient are less likely to report therapeutic change and are more likely to drop-out (Morrissey-Kane & Prinz, 1999).

Although research in youth samples suggests that close others’ therapy attitudes can influence treatment engagement, response, and drop out, there has been little research on this topic, especially for people with anxiety and related disorders. Treatment acceptability research has surveyed parents of youth with ADHD (Berger et al., 2008) and family members of those with schizophrenia (Irani et al., 2004), finding that family members are supportive of pharmacological interventions for mental health treatment. Tarnowski and colleagues (1992) found that parents prefer therapy over medication for the treatment of childhood depression, and similar findings have been reported for childhood anxiety (Brown, Deacon, Abramowitz, Dammann, & Whiteside, 2007; Chavira et al., 2003). However, these studies focus on childhood disorders and many are hypothetical (e.g., “if I had a child with social anxiety…”). As such, we know little about the attitudes towards therapy held by people who are close to an adult with an anxiety or related disorder.

Despite research showing that family members perceive therapy positively, there is preliminary evidence that close others may also hold fears about therapy (e.g., that they may be blamed for the development of the disorder; Renshaw et al., 2005) and the logistics around therapy (e.g., practical obstacles like time and money; Kazdin et al., 1997). Additionally, close others may share concerns that clients themselves hold about therapy, such as worries about treatment failing or resulting in worsened symptoms (see the Treatment Ambivalence Questionnaire; Rowa et al., 2014). However, to date there has been no systematic study of treatment concerns in the close others of adults with anxiety and related disorders, nor is there a validated measure of loved ones’ concerns. Development of such a measure would allow us to explore the impact of treatment ambivalence of close others on their loved ones’ pursuit of and success in therapy. Additionally, given the recent clinical trend towards systemic interventions, including the involvement of important others in anxiety treatments (Chambless, 2012; Thompson-Hollands et al., 2014; and for a review see Carr, 2014), such a measure would be of use to clinicians when considering such involvement.

The purposes of the current studies are to develop and validate a measure of treatment ambivalence for close others of those with anxiety and related disorders, drawing from the methods used by Rowa et al., (2014). Rowa and colleagues surveyed those with anxiety and related disorders with an open-ended form eliciting treatment fears and concerns. They coded these responses to understand themes and used these themes to develop items for the Treatment Ambivalence Questionnaire, which they administered to a large sample and explored its factor structure, reliability, and validity. The current series of studies aimed to conduct a similar procedure to explore the treatment concerns of close others to those with anxiety. The first study sought to determine the nature and range of the concerns close others have about cognitive-behavioural therapy for their loved one with anxiety. Because attitudes are culturally influenced (Kushner & Sher, 1989, 1991) and there are different beliefs about family roles, mental illness, and therapy in different cultures, it is important to note that this sample is drawn from a Canadian population and may not reflect treatment concerns globally. Using the inductive method for item generation (Boateng et al., 2018), the qualitative data collected in study 1 was used to develop the Treatment Concerns Questionnaire-Close Others (TCQ-C). The second study examined the psychometric properties of the TCQ-C, including factor structure, internal consistency, and convergent and divergent validity.

Study 1: Measure Development

Many people will experience an anxiety or related disorder in their lifetime, but a minority will seek treatment. Close others can influence not only the course of a mental health difficulty, but also the person’s treatment-seeking efforts. Close others may be supportive of entering treatment, but they may also have concerns about treatment that can influence their loved ones’ decisions with respect totreatment seeking and drop out. Despite research suggesting that family members can influence therapy success, to date no research has directly surveyed loved ones of those with anxiety and related disorders about the concerns they have about their loved ones starting therapy. The purposes of the current study were to: (i) obtain qualitative data on the concerns people have about their loved one receiving CBT; (ii) identify the general themes following the principles of grounded theory, in which the conceptualization of an idea is conducted in a bottom-up manner (from data to theory; Chun Tie et al., 2019); and, (iii) use this data to develop a quantitative measure of close other treatment ambivalence.

Methods

Procedure

Participants were loved ones of people seeking treatment at an outpatient adult anxiety clinic in Ontario, Canada. We defined “loved one” broadly and included anyone who considers themselves to be a “close other” to the person with anxiety (e.g., family members, romantic partners). The study was advertised through posters in the clinic, at family education sessions, and posters given to clients to pass onto loved ones if they wished. People who were over 18 years of age and whose loved ones had not yet begun treatment or were in the first month of treatment were invited to participate in the online survey. Those who participated were able to enter a draw to win a gift card. The survey began with a brief overview of CBT to ensure that all participants had a basic understanding of this form of therapy. This overview was presented as 10 slides that, in lay terms, (a) introduced the idea that CBT is an evidence-based treatment, (b) provided an example of how thoughts and behaviours impact each other and levels of anxiety, (c) provided basic information about how CBT might address these maintaining factors in anxiety, and (d) provided resources for additional information. The survey also included demographic questions, a questionnaire about treatment expectations, and open-ended forms on which participants could report expected benefits of and concerns about treatment. Participants were asked to rate each the intensity of each concern from 1 (“minor concern”) to 10 (“major concern”). For the purposes of this study, we report only on respondents’ concerns and how those were used to create the measure.

Participants

Participants (that is, those reporting on their concerns about their loved one starting treatment) (n = 33) ranged from 18 to 65 years of age and were 43.15 years of age on average (SD = 12.03). About half (52.6%) of the participants were female (2.6% trans or non-binary). The majority (73.7%) of participants were White (5.3% East Asian, 2.6% Aboriginal, 2.6% Latin American, 2.6% South Asian). The education background of the sample was varied, with 24.2% having completed university or college, 21.1% having completed some university or college, 15.8% having completed a graduate level degree, and 10.5% having completed high school.

Most participants were a parent (39.5%) or a romantic partner (34.2%) of a client with anxiety; on average, participants knew the client for 18.77 years (SD = 10.34). Many (60.5%) were currently living with them, and were in frequent direct (i.e., face-to-face; µ = 35.07 hours/week, SD = 23.07, range = 0–80) and indirect (i.e., text, call, etc.; µ = 8.17 hours/week, SD = 20.51, range = 0-112) contact with the client.

Clients were 31.91 years of age on average (SD = 9.98, range = 18–54), 34.2% female (2.6% trans or non-binary), and 50% White (5.3% Asian, 44.7% information was not in clinical file). Diagnostic information (obtained through formal clinical assessments) was collected from the clinic with respondent and client permission. Diagnoses varied, with 15.8% of the sufferers having a primary diagnosis of OCD, 7.9% SAD, 7.9% GAD, 7.9% panic disorder, 5.3% depression, 2.6% PTSD; the remainder of the diagnoses were not retrievable from the client files (e.g., assessment was not yet complete or client had not consented to their data being used for research). Note that many of the clients had additional diagnoses, and all were seeking treatment for an anxiety or related (e.g., obsessive-compulsive or traumatic-stress) concern. Participants reported that 50% of those with anxiety had never received CBT, and for 87%, it was their first time seeking any kind of therapy for their anxiety. Participants reported that the clients had been experiencing anxiety for 16.37 years on average (SD = 10.40).

Results

Of the 33 participants, 23 (69.7%) reported having one or more treatment concerns. The modal number of concerns reported was 2 (range = 0–6). The average intensity of the concerns was 6.67 (SD = 1.95) on a 10-point scale, with no concern being rated lower than 3. In total, there were 55 unique concerns. The list of concerns was initially reviewed to gain an understanding of the range of responses. Then, stage 1 coding was completed, whereby each response was given a label that reflected the themes of that concern (“lower-order” or “conceptual coding”; Charmaz 2006). In accordance with grounded theory, lower-order codes were collapsed into categories or “higher-order” codes, which allow an understanding of larger themes in the data (Charmaz, 2006; Chun Tie et al., 2019). After reviewing concerns and codes, all three authors agreed that the qualitative data had reached saturation, as data was beginning to repeat, and no additional insights were identified. This number of participants is within the typical range for reaching saturation on qualitative data collection (Hennink, Kaiser, & Marconi, 2017; Marshall et al., 2013). See the discussion for a comment on coder reflexivity.

Category coding resulted in seven main themes (see Table 1). All authors reviewed the coding, and no conflicts arose. In the interest of creating a quantitative measure based on this qualitative information, conceptual codes were used to create items that captured the main concerns from our dataset (both broad themes as well as individual concerns). This resulted in a pool of 24 items. Then, all authors reviewed the items, eliminating or combining those that were redundant, and ensuring all remaining items showed face and content validity. The remaining measure consisted of 19 items (2–4 from each thematic category).

Table 1

Qualitative Themes: Close Others’ Concerns about Treatment

Theme	Frequency of Endorsement: n1 (%)	Example Response2
Treatment will not work	16 (24.24%)	“Therapy … won’t work. We have tried many things that [haven’t]”
Treatment failure will result in client hopelessness, depression, self-criticism	6 (9.09%)	“… He will feel even lower than he does now if it doesn’t work because he sees no better life for himself”
Treatment will result in an unwanted change in family relationship(s)	8 (12.12%)	“… The treatment will put stress on our relationship”
Treatment will lead to unwanted changes in client’s personality	2 (3.03%)	“He changes the person he is…”
Client will take on others’ anxieties	5 (7.57%)	“She will pick up other problems or symptoms from peers”
Client will not fully engage with therapy	11 (16.67%)	“… My loved one will not practice elements introduced in the treatment.”
Treatment will lead to increased anxiety, stress, depression, and/or self-criticism	18 (27.27%)	“Treatment will cause him stress by having to face his anxiety.”
1Although there were 55 unique concerns, some concerns contained content that spanned several categorical codes, resulting in them having more than one categorical code applied. There were 66 categorical codes in total. 2 Respondents provided permission for the use of anonymous quotations.

Discussion

Many participants reported treatment concerns and rated these concerns as important, with moderate to high intensity ratings. Seven main themes arose, with categories reflecting concerns about therapy not working, treatment failure leading the client to feel worse, therapy resulting in increased stress or new fears, therapy changing the loved one and the family relationships, and the loved one not fully committing to treatment. Participants reported that their loved ones had been experiencing anxiety for more than 15 years on average, which is consistent with previous research showing that treatment seeking is often delayed by a decade or more (Wang et al., 2005). Future research may explore whether the attitudes of close others are related to treatment seeking efforts and delays.

Interestingly, approximately one-third of participants did not report any concerns about treatment. This is consistent with previous research showing that family members are typically supportive of their loved one entering therapy (e.g., Chavira et al., 2003; Tarnowski et al., 1992). However, this study sampled close others of people who were already seeking treatment; it is likely that there would be greater ambivalence, on average, among close others of those who have not yet sought treatment (or do not intend to do so). Our sampling technique may have missed out on those who are not seeking treatment because close others’ treatment concerns have interfered with their doing so; that is, our survey may have not captured the most serious concerns about treatment. Since we sampled from an adult anxiety clinic and the majority of clients were seeking therapy for the first time, it is also possible that some of the clients may be seeking treatment now that they can consent on their own, after previously avoiding treatment due to loved one’s treatment concerns. Additionally, the majority of respondents in this sample were White; given research showing that culture can be an important variable in help-seeking attitudes (e.g., Mojaverian et al., 2013; Sun et al., 2016), it is important not to assume that these themes are reflective of those in other racial or ethnic groups. This may be particularly true for cultures in which beliefs about family, mental health, and/or therapy differ significantly from traditional Western beliefs. Future research may explore cross-cultural differences in the treatment attitudes of close others. In recognition of this limitation for the current study, the TCQ-C includes an open-ended question where respondents can report any additional concerns they have about treatment (see Appendix).

Several participants shared concerns about clients taking on others’ anxieties. The outpatient clinic from which loved ones were recruited offers both individual and group therapy services, and loved ones may have been referring to their loved one hearing others’ concerns in a group therapy format. Research is needed to explore how prevalent this concern is among patients and whether this feared “contagion” effect actually occurs in group therapy formats. However, inclusion of these concerns in the measure makes it more widely applicable; it may generate useful information about loved ones’ ambivalence regardless of the therapy format the sufferer is considering/entering.

Given that qualitative coding was involved in this study, reflection upon the potential role of the researchers in the research process is warranted. As a statement of reflexivity, the primary coder (first author) was a graduate student with a clinical training background and the two additional coders (second and third authors) were practicing psychologists and academic professors. All coders were White females, living in Canada, who use CBT as part of their clinical practice. These backgrounds likely played into the research design, collection, and interpretation. The purposes of the study are aligned with researcher beliefs that therapy is a valuable and worthwhile venture for many individuals, and the drive to understand and overcome barriers to treatment success. The work is also influenced by an assumption that families tend to want to support their loved ones, even if they find symptoms challenging. The authors’ prior experiences with clients and their families contributed to the assumption that family members can be ambivalent and have concerns about treatment. This assumption was incorporated into the open-ended form that was used to collect respondent treatment concerns; however, participants were also asked about perceived benefits of the treatment in an attempt to minimize demand characteristics. Prior experience with clients and the research literature on clients’ concerns about therapy may have been a source of influence on the coding of data and generation of themes. We attempted to minimize the effect of this prior knowledge by collecting qualitative information from close others rather than creating items of our own. We were agnostic about the concerns and themes that would arise. We attempted to create items that reflected both general trends in the data as well as individuals’ concerns in order to ensure that multiple perspectives were incorporated into the items that were created.

Loved ones reported significant worries about treatment. This warrants the development of a quantitative measure of close others’ concerns. Such a measure could fill an important gap in our understanding of the underutilization of anxiety treatment and could be used by researchers to study the relationship between ambivalence in close others, client ambivalence, and treatment success. Additionally, when working with clients with anxiety and related disorders, clinicians could use such a measure to understand the treatment attitudes of close others, address their concerns, and/or determine their level of involvement in therapy.

Study 2: Measure Validation

The purpose of the current study was to evaluate the psychometric properties of the TCQ-C through examining its factor structure, internal reliability, and construct validity. As such, we performed an exploratory factor analysis and examined the internal reliability of each subscale through calculating Cronbach’s alpha and Revelle’s omega coefficients. In addition, we examined construct validity through evaluating the measure’s relationship to a theoretically related construct (convergent validity with treatment expectations) and a theoretically unrelated construct (discriminant validity) that may have an indirect effect on responses (i.e., distress).

To explore convergent validity, a pre-existing measure of treatment expectations was selected. Theoretically, one’s treatment expectations should be related to one’s treatment attitudes, such that the more negative one’s attitudes about treatment, the more one feels treatment will not work. For example, Gonzalez, Tinsley, and Kreuder (2002) found that students’ attitudes and expectations both saw a positive shift after a psychoeducational intervention. Given that treatment expectations and treatment ambivalence are not redundant concepts, we expected them to have a significant but moderate correlation to each other.

Discriminant validity was examined through evaluating the relationship between the TCQ-C and a pre-existing measure of respondent distress. As in the psychometric assessment of the Treatment Ambivalence Questionnaire (Rowa et al., 2014), it was important to determine that one’s beliefs about treatment were not solely a representation of their current levels of distress. Based on the findings from that study, small correlations were expected between these constructs.

People who had a loved one with significant symptoms of an anxiety or related disorder were recruited from a number of sources: an outpatient anxiety clinic, a student population, a research database, posters in the community, and Amazon Mechanical Turk (“MTurk”). All participants confirmed that they had a loved one with significant, problematic, and/or excessive anxiety that has been present for six months or more. People who were over 18 years of age and who could read and understand English were eligible to participate; there were no age restrictions specified for the person with anxiety that participants reported on. Remuneration varied based on the recruitment method: students (n = 71) received course credit, MTurk participants (n = 225) received a small monetary compensation, and all others (n = 61) were entered into a draw to win a gift card. After consenting to participate, participants reviewed a brief description of CBT (the same that was used in Study 1) and completed demographic questions and other study measures (see Measures). Additionally, participants completed measures about their relationship with their loved one’s anxiety (e.g., criticism, accommodation) and about their own distress tolerance and fear of compassion; for the purposes of this study, we report only on those measures that are relevant to the psychometric properties of the TCQ-C.

Data Preparation

Given that the TCQ-C item means were similar across recruitment methods (multivariate GLM was significant, F(38,522) = 2.386, p < .001, Wilk’s Λ = 0.725, but none of the post-hoc group differences were significant when correcting for multiple comparisons) and the items did not show significant skew or kurtosis, the combination of data into one sample was deemed appropriate (Guilford, 1952) and even advantageous, as it provides increased heterogeneity and generalizability (Gaskin et al., 2017), and data from these different sources have been shown to be comparable in quality (Kees et al., 2017). Once combined, data quality was analyzed. Bot and duplicate responding detection resulted in the deletion of 4 participants’ data, failed attention checks resulted in the deletion of 18 participants’ data (a smaller proportion than is often reported in MTurk samples; e.g., Kaufmann, Schulze, & Veit, 2011 as cited by Bentley 2021), and two participants were removed from analyses due to reporting on themselves, rather than a loved one with anxiety. An additional 46 participants were removed from analyses due to having completed less than 50% of the items in the study (a similar proportion to previously reported community and MTurk samples; e.g., Zhang & Gearhart 2020).

The remaining participants’ (n = 287) data showed acceptable skew and kurtosis on all measures (Kline, 1998). Outliers were examined through Z-score analysis and visual inspection of box plots; values that were 3 or more standard deviations from the mean and that were discontinuous from the distribution (n = 9) were replaced with the respondent’s subscale mean value, provided that the respondent had completed 80% or more of the items on that subscale.

See Table 2 for demographic information about the respondents and their loved ones with anxiety. The majority of respondents were living in the United States of America (n = 169, 58.9%) or Canada (n = 107, 37.3%) when they participated, with < 5 participants each living in Barbados, China, Egypt, or India. The majority of their loved ones were also living in the United States of America (n = 171, 59.6%) or Canada (n = 104, 36.2%), with < 6 each currently living in China, Egypt, France, Germany, India, or South Korea.

Most respondents were a partner/spouse (n = 79, 27.5%), parent (n = 67, 23.3%), sibling (n = 61, 21.3%) or adult child (n = 45, 15.7%) of the person with anxiety. Other relationships (ns < 9) included grandparent, aunt/uncle, cousin, grandchild, close friend, and niece. The age of the sufferers was normally distributed, and the majority were adults: 9.9% were 6–17 years old, 33.4% were 18–29 years old, 26.4% were 30–45 years old, 22.2% were 46–64 years old, and 7% were 65–90 years old (1% did not report age). Most respondents (n = 257, 89.5%) had lived with the sufferer at some point, and many (n = 164, 57.1%) were living with them at the time of participation. On average, respondents knew the sufferer for 21.2 years (SD = 12.9). Respondents were in frequent direct (i.e., face-to-face; µ = 30.5 hours/week, SD = 33.0, range = 0-120) and indirect (i.e., text, call, etc.; µ = 8.9 hours/week, SD = 16.0, range = 0-112) contact with the sufferer.

Respondents reported that their loved one had been experiencing anxiety for 15.8 years on average (SD = 14.7). Most respondents reported that their loved one’s anxiety was impairing to a small (n = 112, 39%), moderate (n = 110, 38.3%), or great (n = 42, 14.6%) degree. Sixty-two percent of respondents (n = 180) reported that their loved one had received a diagnosis for their anxiety and 55.1% (n = 158) reported that their loved one had attended therapy for their anxiety. Twenty-two percent of respondents reported that the sufferer was in therapy at the time of the study.

Materials

Treatment Concerns Questionnaire-Close Others (TCQ-C; see Appendix). Treatment concerns were measured using the TCQ-C, the development of which was described in Study 1. The TCQ-C lists 19 potential concerns based on those reported by close others of people with anxiety. For each item, respondents rate how much they agree or disagree that they are concerned on a 7-point scale from “strongly disagree” to “strongly agree”.

Stanford Expectations for Treatment Scale, Family Modification (SETS-F). Treatment expectations were measured using a modified version of the SETS (Younger et al., 2012), which is a brief, 6-item scale, 3 items assessing for positive treatment expectations and 3 items for negative expectations. The measure shows acceptable internal consistency and predictive validity in health care settings (Younger et al., 2012). For the purposes of this study, two items were slightly modified to become specific to a loved ones’ anxiety treatment (e.g., original item “my condition will be completely resolved after this treatment” was modified to “my loved one’s anxiety will be completely resolved after this treatment”). In this sample, the SETS-F showed acceptable internal consistency (positive expectations subscale: Cronbach’s alpha (“α”) = 0.75; Revelle’s omega (ω) = 0.76 ; negative expectations subscale: α = 0.79; ω = 0.81).

Depression, Anxiety, Stress Scales (DASS-21; Lovibond & Lovibond 1995). Respondent distress was measured through the DASS-21, a 21-item self-report scale that assesses for depression, anxiety, and stress. Each item is rated based on respondent’s distress in the past week on a 4-point scale from “did not apply to me at all” to “applied to me very much, or most of the time”. This widely used scale shows good reliability and validity (Antony et al., 1998), and showed excellent internal reliability in this sample (α = 0.96; ω = 0.97). For the purposes of the current study, the DASS-21 was used as an indicator of respondent distress.

Exploratory factor analysis

Two-hundred and eighty-one (n = 281) participants completed the TCQ-C. Approximately 3% of those participants had 1 item missing from their response; missing items were treated with mean imputation. Items were within the recommended guidelines for skew and kurtosis (Watkins, 2018). Analyses were performed using SPSS version 24. Bartlett’s test of sphericity (approximate Chi-Square(136) = 2653.168, p < .001), and Kaiser-Myer-Olkin measure of sampling adequacy (0.89) confirmed that the correlation matrix was factorable. Maximum likelihood estimation was used to identify a latent factor structure. Oblimin rotation was employed, as we assumed that factors would be intercorrelated rather than orthogonal. Eigenvalues, scree plot, theoretical convergence, and percentage of variance accounted for by each factor were used to determine number of factors. Four factors showed eigenvalues over 1, accounted for a significant amount of variance, and represented the most parsimonious solution. A four-factor solution was robust across rotation methods and was also supported by the results of parallel analysis (conducted using RStudio, package “psych”, function “fa.parallel”): after 500 iterations, four eigenvalues from the observed correlations exceeded the eigenvalues from the simulated data correlations.

Item 1 (“treatment will change my loved one in unexpected ways”) showed discrepant loadings in the pattern matrix and the structure matrix, indicating that it is influenced by more than one factor; it was removed from analyses. Item 9 (“through treatment, my loved one may hear about others’ problems, and will think their own concerns are not valid”) loaded saliently on two factors (values within 0.10 of each other), so it was removed from analyses. The final four-factor solution accounted for 68.8% of the variance.

The first factor (“Adverse Reactions”, AR) was saliently loaded by six items, accounted for 41.89% of the variance, and showed good internal reliability (α = 0.86, ω = 0.90). The second factor (“Personal/Family Consequences”, PC) contained four items, accounted for 12.53% of the variance, and showed good internal reliability (α = 0.80, ω = 0.951). The third factor (“Lack of Commitment”, LC) was saliently loaded by four items, accounted for 7.62% of the variance, and good internal reliability (α = 0.89, ω = 0.90). The fourth and final factor (“Ineffectiveness”, IE) was saliently loaded by three items and accounted for 6.80% of the variance, with good internal reliability (α = 0.82, ω = 0.82). See Table 3 for descriptive statistics for each subscale and Table 4 for pattern coefficients and communalities of all items. When the 17 items are summed to make a total score, this total scale has excellent internal reliability (α = 0.91, ω = 0.94).

Table 3

Subscale Descriptive Statistics

	Descriptive Statistics
	Mean	SD	Skew	Kurtosis
AR	22.13	8.42	− 0.022	− 0.657
PC	9.21	4.79	1.04	0.722
LC	16.36	6.60	− 0.254	− 0.959
IE	10.98	4.54	0.095	− 0.878

TCQ-C Subscales: AR = Adverse Reactions, PC = Personal/Family Consequences, LC = Lack of Commitment, IE = Ineffectiveness; SD = standard deviation

Table 4

Factor Statistics

		Factor Loadings
Item #	Item	Communalities	F1	F2	F3	F4
3	My loved one will feel like a ‘failure’ if they do not ‘succeed’ in treatment.	0.41	0.53	0.00	− 0.08	0.11
10	Treatment will be overwhelming for my loved one.	0.60	.75	− 0.05	− 0.14	− 0.07
12	My loved one will feel hopeless or depressed if treatment does not help.	0.41	0.61	− 0.06	− 0.03	0.08
14	Treatment will create too much pressure for my loved one.	0.75	0.81	0.00	− 0.02	0.08
15	Through treatment, my loved one may learn about symptoms or fears others have, and will take on these fears or symptoms as their own.	0.45	0.55	0.21	− 0.01	− 0.02
17	Treatment will lead to worsened symptoms.	0.61	0.62	0.21	0.03	0.10
5	Treatment will negatively change the way my loved one sees me.	0.71	0.01	0.79	− 0.06	0.11
8	I will lose my relationship with my loved one because of this treatment.	0.68	− 0.10	0.82	− 0.10	0.09
11	If my loved one’s anxiety improves, other issues in our relationship/family will become more prominent.	0.34	0.09	0.56	0.00	− 0.14
13	Treatment will change my loved one in a negative way.	0.57	0.32	0.51	0.16	0.18
7	My loved one will not use the coping tools that they learn in treatment.	0.68	− 0.06	− 0.03	− 0.73	0.23
16	My loved one will not be fully committed to improving.	0.70	− 0.01	0.08	− 0.83	− 0.04
18	My loved one will not complete all the required components of the treatment.	0.67	0.04	0.03	− 0.78	0.01
19	If treatment is difficult, my loved one will become discouraged and give up.	0.74	0.30	− 0.03	− .70	− 0.04
2	Treatment will not work.	0.71	0.19	− 0.11	0.01	0.76
4	Treatment will be a waste of time and/or money.	0.54	− 0.03	0.20	− 0.09	0.64
6	Treatment will not be potent or comprehensive enough to help my loved one.	0.60	0.10	− 0.02	− 0.14	0.64

Extraction communalities are reported. Factor loadings are represented by pattern coefficients. F1 = Adverse Reactions (AR); F2 = Personal/Family Consequences (PC); F3 = Lack of Commitment (LC); F4 = Ineffectiveness (IE). See Appendix for final measure.

Respondents also had the option to complete an open-ended form with any additional concerns they had. Ninety-five respondents (33%) listed concerns here, many of which were more specific examples of items from the scale (e.g., reasons why the treatment may be distressing for their loved one in particular). Using the coding system from study 1, coding revealed that 22 described concerns fell under the Adverse Reactions category (e.g., group therapy being difficult for someone with social anxiety), 6 additional concerns were about personal/family consequences (e.g., support network not being encouraging of therapy), 34 concerns centered around the sufferer’s lack of commitment (e.g., they will not see therapy as applicable to them, will not co-operate, will stop therapy too early), and 12 concerns fell under the category of treatment being ineffective (e.g., therapist being unskilled, therapy not being tailored to their needs). Lastly, close others also endorsed concerns about treatment being too costly (n = 11), time-consuming (n = 6), or stigmatizing (n = 4).

Convergent and discriminant validity

We examined convergent validity through performing correlations between the TCQ-C subscales and the SETS, a measure of treatment expectations. As predicted, the TCQ-C subscales showed negative correlations with the positive expectations subscale of the SETS, and positive correlations with the negative expectations subscale of the SETS (see Table 5). Although all but one of these correlations were significant, beliefs about the ineffectiveness of treatment (TCQ-C-IE) were more strongly negatively related to positive expectations for treatment, and beliefs about adverse reactions and personal consequences of treatment were more strongly positively related to negative expectations for treatment.

Table 5

Correlations: TCQ-C Subscales, General Distress, Treatment Expectations

	TCQ-C-AR	TCQ-C-PC	TCQ-C-LC	TCQ-C-IE	DASS-S	DASS-A	DASS-D	SETS-POS	SETS-NEG
TCQ-C-AR	-
TCQ-C-PC	0.528**	-
TCQ-C-LC	0.560**	0.270**	-
TCQ-C-IE	0.574**	0.353**	0.539**	-
DASS-S	0.286**	0.083	0.282**	0.191*	-
DASS-A	0.264**	0.157*	0.244**	0.111	0.799**	-
DASS-D	0.291**	0.151*	0.335**	0.192*	0.745**	0.699**	-
SETS-POS	− 0.193*	− 0.011	− 0.323**	− 0.538**	− 0.137*	− 0.074	− 0.161*	-
SETS-NEG	0.539**	0.486**	0.242**	0.307**	0.037	0.113	0.065	− 0.046	-

TCQ-C = Treatment Concerns Questionnaire–Close Others; AR = Adverse Reactions, PC = Personal/Family Consequences, LC = Lack of Commitment, IE = Ineffectiveness; DASS = Depression, Anxiety, Stress Scales; DASS-S = Stress Scale, DASS-A = Anxiety Scale, DASS-D = Depression Scale; SETS-POS = Stanford Expectations for Treatment Scale, Positive Expectations Subscale, SETS-NEG = Stanford Expectations for Treatment Scale, Negative Expectations Subscale. *p < .05, **p < .001

Discriminant validity was explored through examining the correlations between the TCQ-C subscales and the respondent DASS scale scores. The correlations between these measures were small to moderate (rs ranging from 0.08 to 0.34; see Table 5), indicating that this measure is not just a reflection of one’s personal distress level.

The current study examined the psychometric properties of the TCQ-C. Exploratory factor analysis showed that the most parsimonious solution was a four-factor solution, yielding 4 subscales with good internal reliability and with small to moderate intercorrelations. These subscales reflect fears about treatment being ineffective, sufferers not fully committing to treatment, adverse reactions to treatment (e.g., increased symptoms) and personal consequences of treatment (e.g., changes in family dynamics). These four factors correspond well to the main themes that were derived in Study 1, with changes in personality and family relationships being grouped together to make the ‘personal/family consequences’ subscale and concerns about symptoms increasing, hopelessness, and taking on others’ symptoms being grouped together to make the ‘adverse reactions’ subscale, the subscale that accounted for the most variance. Future research could continue the validation of this factor structure through confirmatory factor analysis.

Many of these factors overlap with concerns clients themselves have about treatment as measured by the Treatment Ambivalence Questionnaire (TAQ; Rowa et al., 2014), a measure of client ambivalence for people with anxiety and related disorders. The TAQ contains subscales that reflect concerns about personal consequences (such as change in personality or relationships) and adverse reactions (such as an increase in anxiety or distress). Indeed, these concerns may reflect some truth: therapy can be distressing in the short term (especially exposure-based treatments; Thornton, 2017), treatment failure can lead to self-blame (Berk & Parker, 2009), CBT for anxiety does not help everyone (Loerinc et al., 2015), and therapy sometimes leads to negative outcomes for families (Szapocznik & Prado, 2007). The TCQ-C can help assess when close others hold exaggerated or problematic treatment fears.

The TAQ and TCQ-C can be used together to understand where clients and/or close others hold obvious misunderstandings about therapy. Depending on what is endorsed on these measures, psychoeducation could include messages that CBT can be stressful in the short-term, but it often leads to a decrease in anxiety symptoms in the long-term; treatment can involve the development of coping strategies to help the client manage the additional stress that therapy can bring; and that therapists do not encourage clients to sever close, non-toxic relationships. Family mental health awareness efforts are associated with better outcomes for clients (e.g., reduced symptoms and relapse) and families (e.g., reduced family distress and friction) (Lucksted, McFarlane, Downing, & Dixon, 2012). In their review, these authors emphasized the need for family education efforts to be tailored; the TCQ-C could serve as an efficient and effective tool for doing so. Researchers, too, could use the TAQ and TCQ-C conjointly to study whether there is an association between specific concerns that close others endorse (on the TCQ-C) and those that anxiety sufferers endorse (on the TAQ). This may shed light on the origin and maintenance of beliefs/attitudes that impede treatment seeking behaviours.

Close others are worried about their loved ones’ (lack of) commitment to treatment. Of the four factors, this factor showed the highest endorsement (highest average item score), whereas the personal consequences factor showed the lowest endorsement. At the same time, close others are optimistic about treatment, showing higher average positive expectations scores (mean = 3.90, SD = 1.23) than negative expectations scores (mean = 3.07, SD = 1.40) on the Stanford Expectations for Treatment Scale (t(280) = 9.94, p < .001; not reported above). If close others themselves are invested in the success of treatment, it is reasonable that they may be concerned about whether their loved one is equally invested. Alternatively, endorsement of these items may reflect disappointment in the failure of previous treatments; after all, CBT for anxiety has clinically significant effects for about half of those who try it (Loerinc et al., 2015), and outside observers may be in search of an explanation for why it hasn’t worked for their loved one. Attributing the negative outcomes in clients’ lives to their own actions or characteristics is associated with criticism and hostility in relatives (Barrowclough & Hooley, 2003). Future research could explore whether the TCQ-C “Lack of Commitment” subscale is particularly related to levels of criticism observed in the family or reported by the person with anxiety. Should close others endorse these items, clinicians may want to explore the origins of these concerns, and gently assess for criticism and client ambivalence about treatment.

The construct validity of the TCQ-C was also explored. Evidence for convergent validity was seen in the TCQ-C subscales’ correlations with a measure of treatment expectations. As expected, the strength of these correlations indicate that they are related, but not redundant concepts. One exception was that positive expectations did not significantly correlate with worries about the personal consequences of treatment. It appears that these concepts are unrelated, such that close others’ beliefs that treatment will be effective are not related to how concerned they are about treatment impacting family dynamics. In contrast, less worry about a loved one’s commitment or about the ineffectiveness of treatment is associated with more positive expectations for the outcomes of therapy. Given that hope and positive expectations are an important component of treatment success (for a recent meta-analysis, see Constantino, Vîslă, Coyne, & Boswell, 2018), the ineffectiveness subscale of the TCQ-C could be used to assess and address beliefs about treatment ineffectiveness in the early stages of treatment.

Evidence for discriminant validity lies in the small correlations seen between respondent distress and treatment ambivalence. Concerns about ineffectiveness of treatment and personal consequences of treatment seem to be especially unrelated to respondent distress. In contrast, concerns about adverse reactions to treatment and worries about lack of commitment showed a small relationship to depression scores (rs = 0.291 and 0.335, respectively), perhaps indicating that when one is more depressed, there is a general pessimism about their loved one’s ability to handle treatment.

A strength of this study is the diverse sampling techniques that were used, resulting in respondents from numerous countries, of different ages, and with a variety of relationships to the individual with anxiety. A limitation of this sample is that the majority of respondents and people with anxiety were White, limiting its generalizability to close others from other cultures and ethnicities. Traditions around family involvement in mental health (Chadda & Deb, 2013), stigma about mental illness (Papadopoulos et al., 2013), and accessibility of treatment (Snowden & Yamada, 2005), vary across cultures; future work would benefit from conducting similar research in non-White samples. In addition, some respondents (17%) reported having no direct contact with the person with anxiety; however, it should be noted that this study was conducted from August 2020-November 2021, when many individuals were refraining from in-person contact with those outside their household due to the COVID-19 pandemic. Lastly, this sample represents people who believe their loved one has problematic, excessive, and/or significant anxiety; however, a formal assessment was not conducted to determine diagnostic information. Future research may aim to replicate the factor structure of this measure in samples of those from different cultures and with a known clinical diagnosis. This measure may also be used to explore whether the type and strength of concerns differs as a function of the relationship to the client (e.g., parent versus romantic partner).

General discussion

This series of studies represents the first attempt to systematically investigate the concerns that close others have about their loved ones starting treatment for anxiety. Although previous research has discussed the importance of family attitudes (e.g., Morrissey-Kane & Prinz 1999), a measure of treatment concerns in close others had not previously been developed. In accordance with grounded theory principles (Chun Tie et al., 2019), we elicited concerns by close others and used this qualitative information to develop a quantitative measure of their treatment ambivalence. The final measure, with 17 items total, had strong internal reliability, a stable four-factor structure, and evidence of construct validity.

Across studies, the majority of respondents were parents or romantic partners of sufferers, indicating that these may be the close others that are most involved and/or invested in a clients’ mental health. Respondents in both studies had known the sufferer for many years (study 1 mean = 18.8 years, study 2 mean = 21.2 years) and the majority were living with the sufferer at the time of the study, which may be indicative of the close relationships between the respondents and the sufferers. Across studies, sufferers were mostly adults, with mean ages ranging from 32 to 36 years. The concerns investigated in these studies therefore apply mainly to adult sufferers with anxiety; future research may develop a similar measure for the close others of children with anxiety who are entering treatment.

While the samples in the two studies were similar in many ways, they differed in that the first study sampled loved ones of those who were treatment-seeking, while the second study sampled close others of anyone with excessive anxiety (with varying treatment status). Treatment concerns may be greater in a sample of those who are not awaiting treatment; on the other hand, treatment ambivalence may increase as a treatment start date approaches and the prospect of treatment becomes more “real” to close others. As Kushner & Sher (1991) write, different treatment fears may be relevant at different times throughout the help-seeking process. Some of the study two sufferers may be in different stages of therapy, while others may have no therapy experience at all. Close others of those with no therapy experience may have increased fears about lack of commitment. For example, on the open-ended item, some loved ones reported worries about the sufferer not being open to therapy, not seeing it as applicable to them, or “not taking it seriously”. For close others of people who are in therapy (22% of the study two sample), concerns about personal consequences may be more relevant, and concerns about ineffectiveness and adverse consequences may vary depending on the close other’s perception of the sufferer’s progress. Future research may seek to compare the responses of close others whose loved ones are in different stages of treatment seeking.

Whether or not close others are involved in treatment, their attitudes could impact the client’s motivation and treatment outcomes. Thus, assessing for close others’ attitudes could be a useful part of the intake process or as a mid-therapy assessment should the expected progress not be observed. This information can then be used to discuss with the client the impact of their close others’ attitudes or to consider offering close others a psychoeducation session before proceeding with treatment. Therapists can also use the TCQ-C as a tool to assess and address attitudes in preparation for joint therapy sessions with close others or when considering transitioning from individual to family therapy. Future research may explore whether close others’ concerns about treatment are related to rates of treatment-seeking in those with anxiety and whether TCQ-C scores predict treatment adherence, drop-out, or outcomes once the client has entered treatment.

Below is the link to the electronic supplementary material.

Supplementary Material 1

Table 2: Demographic Information
	Participant (Close Other)	Person with Anxiety
Age (years) Mean Standard Deviation	35.8 13.2	36.9 17.7
Gender, n (%) Male Female Gender non-binary, gender non-conforming Declined to answer	94 (32.8%) 188 (65.5%) 3 (1.0%) 1 (0.3%)	114 (39.7%) 164 (57.1%) 7 (2.4%) 2 (0.7%)
Ethnicity, n (%) White East Asian South Asian Black/African American Other	200 (69.7%) 21 (7.3%) 18 (6.3%) 14 (4.9%) 34 (11.8%)	195 (67.9%) 16 (5.6%) 20 (7.0%) 15 (5.2%) 41 (14.3%)
Highest level of education, n (%) High school or less Some or completed university/college degree Some or completed graduate degree Other or declined to answer	39 (13.6%) 203 (70.7%) 41 (14.3%) 4 (1.4%)	89 (31.0%) 144 (50.2%) 45 (15.7%) 9 (3.1%)