Among members of the general population, the quality of family relationships plays a critical role in shaping an individual’s health, quality of life [1], and well-being across the lifespan [2]. Because of the substantial influence family relationships have on an individual’s health, it is imperative for treatment to consider and address the family relationships of persons with serious mental illness (SMI). A focus on the relationships of persons with SMI is consistent with a recovery-oriented approach, which maintains that recovery is supported through relationships and social networks [3]. Qualitative evidence suggests persons with SMI consider the creation of relationships to be both an indicator and a process of recovery [4], and quantitative studies demonstrate that for persons with SMI, supportive family relationships are correlated with improved functional outcomes [5], overall health outcomes [6], decreased hospitalization rates and lengths [6, 7] and lower mortality rates [8]. Collectively, empirical evidence suggests that positive familial relationships of persons with SMI are a vital component of the recovery process.

To date, much of the literature on families of persons with SMI has focused on negative family dynamics, including caregiving burden and distress [9, 10]. While research does indicate that these dynamics may occur in some families of persons with SMI, this does not preclude positive family dynamics from occurring as well. Indeed, available research indicates that positive family dynamics such as contributions made by persons with SMI [11], caregiving gratification [12], and closeness in personal relationships [13] tend to be the rule rather than the exception. By overlooking the positive dynamics that occur in families, such scholarship may contribute to researchers and human service professionals developing narrow and incomplete understandings of the family relationships of persons with SMI. Further, by largely ignoring the positive qualities of families of persons with SMI, research in this area likely contributes to the stigma of mental illness and is not aligned with the recovery movement nor the strengths-based perspective. One positive aspect of family relationships among persons with SMI that is under-researched and deserving of attention is relationship quality. Knowing rates of positive relationship quality among persons with SMI and examining the factors associated with such may provide researchers and mental health professionals with a fuller and more accurate understanding of the rich and varied family relationships of persons with SMI.

Employing a U.S. community-recruited sample of persons with SMI, the primary objective of the current analyses is to examine the extent to which proposed factors are associated with relationship quality with the adult relative with whom they spend the most time—hereafter referred to as reference relatives (RR). The secondary objective is to determine the extent to which the relationship between the use of routine limit-setting practices by RR toward participants and relationship quality is mediated by perceived emotional over-involvement of RR.

Background Literature

Approximately 5.2% of the United States population (an estimated 13.1 million people) lives with SMI [14]. Within this population, family relationships have historically been examined using three frameworks: (1) family caregiving burden, (2) expressed emotion (EE), and (3) family stress and coping models. Each of these frameworks has been criticized for their narrow scope that ignores the depth and strengths of families within their own contexts [15]. To provide a more holistic and accurate depiction of the family relationships of persons with SMI, it is necessary to describe their relationship quality with family members as well as factors that may impact relationship quality.

Unfortunately, there is a dearth of literature on relationship quality among persons with SMI. One of available studies found that compared to the general population, persons with SMI were less satisfied with their social relationships [16]. Another study found that the more perceived social support between a person with SMI and their romantic partner, the higher each partner rated their relationship satisfaction [17]. Finally, though it was not the explicit focus, Hooley and Teasdale [18] reported a connection between EE, perceived criticism, and marital satisfaction in persons living with depression.

While little is known about the factors that influence relationship quality among persons with SMI, more attention has been paid to this topic within the general population. Among the general population, there is sparse and/or contradicting evidence regarding whether demographics such as age, race, socioeconomic status, and partners’ employment status influence relationship quality [19,20,21,22]. As such, we proposed that demographic factors were not significantly associated with relationship quality. Research generally suggests that in multivariate analyses, family outcomes (e.g., family functioning and violence) are largely not associated with clinical variables (e.g., diagnosis and symptoms), but instead are correlated with family dynamics such as communication, problem-solving, and limit-setting [23, 24]. We proposed that relationship quality was not associated with clinical variables among persons with SMI (diagnosis, psychiatric hospitalization in the past year, illegal drug use, or regular use of psychiatric medications). Relationship interaction factors between participants and RR conceptualized as supportive (presented below) were hypothesized to be positively associated with relationship quality. Relationship interaction factors conceptualized as problematic were proposed to be negatively associated with relationship quality.

Supportive Relationship Interaction Factors

Co-residence and Frequency of in-person Contact

With intimate partners and married couples, co-residence and relationship satisfaction are correlated and appear to be mediated by commitment to the relationship [25, 26]. However, co-residence must be considered from a cultural context, as not all cultures accept co-residence prior to marriage in the case of romantic relationships. Therefore, how a culture views co-residence is influenced by the age and type of relationships among others, which in turn may influence relationship satisfaction [27]. Dainton and Aylor [28] note the associations of physical presence and routine interactions with relationship quality. Family members having higher levels of in-person contact with each other may afford them greater opportunities for relational maintenance behaviors, such as sharing everyday tasks, which are associated with increased relationship quality [29].

Financial Assistance and Assistance with Activities of Daily Living

Researchers tend to collapse various dimensions of providing and receiving help (e.g., emotional support, helping around house, and financial contributions) into a single, overarching concept of “support” [30]. Among married couples and within friendships, the provision of various forms of support towards each other is positively correlated with relationship quality [19, 30, 31].

Problematic Relationship Interaction Factors

Psychological Abuse and Physical Violence

Within the literature on intimate partners, psychological abuse [32] and physical violence [33,34,35] are generally associated with decreased relationship satisfaction; although, there is occasional conflicting evidence when people do not consider themselves to be in a physically or psychologically violent relationship or if people engage in reciprocal behaviors [36].

Limit-setting Practices and Emotional over-involvement

While family members typically acknowledge their relative with SMI’s autonomy and interact with them in a collaborative manner, for various reasons (e.g., to increase medication compliance and prevent problematic behaviors), many family members also attempt to modify the behavior of persons with SMI through encouragement, verbal pressure, and contingency contracts—termed limit-setting practices [37]. To date, no studies have directly explored the relationship between limit-setting practices and relationship quality between adults; however, limit-setting practices have repeatedly been associated with family conflict and violence [for a review see 38], with evidence indicating that limit-setting practices often precede assaults [39]. There is likely considerable variation in how limit-setting practices are enacted by family members and how they are perceived by persons with SMI. When limit-setting practices are perceived—even if not immediately—by persons with SMI as necessary, well-intended, and mutually agreed upon (collaboratively developing and later enacting a psychiatric advanced directive is a potential example), they may be less likely to contribute to conflict and poor relationship quality. Alternatively, when limit-setting practices are perceived by persons with SMI as unilaterally imposed, unnecessary, non-collaborative, coercive, or hostile, they may be strongly connected to negative family outcomes, including poor relationship quality. Within the general population, both hostility and coercion contribute to decreased relationship quality [40, 41].

EE includes the construct of emotional over-involvement of family members towards persons with SMI, with EE being associated with relapse among persons living with schizophrenia [42] and depression [18]. We hypothesize that limiting-setting practices used by family members and perceived emotional over-involvement will be associated with decreased relationship quality.

Conceptualization of Mediation by Perceived Emotional Over-Involvement

We conceptualize that limit-setting practices decrease relationship quality because they are often perceived by persons with SMI as being unnecessary and unwelcome, thereby being perceived as acts of emotional overinvolvement. As such, we propose that the relationship between limit-setting practices and relationship quality will be mediated by perceived emotional over-involvement of RR.

Methods

This article utilizes secondary analysis of data from a study pertaining to conflict between persons with mental illness and their relatives [43]. Five hundred twenty three adults diagnosed with a mental illness completed an online survey between December 2015 and April 2017. Participants provided information regarding themselves, their RR, and their interactions with one another. The investigators decided to conduct an online survey as surveys tend to result in less social desirability bias than interviews and allow for recruiting geographically and clinically diverse samples. Evidence suggests that the majority of persons with serious and non-serious mental illness use the internet [44, 45].

Sampling and Recruitment Procedures

Participants were recruited from a myriad of mental health organizations across the U.S. (described in more detail in the Results section), using online (i.e., website and social media postings, email listservs, and e-newsletters), and in-person (paper flier) advertisements. Advertisements described the study as seeking to understand the interactions persons with mental illness have with their family members. Advertisements contained a web address from which prospective participants could obtain more information about the survey and how to participate. Most people (85%) who began the survey completed it. There were no significant differences in the gender or race/ethnicity of completers versus non-completers. Eligibility criteria included residing in the U.S., being at least 18 years of age, and having been diagnosed with a mental illness, per self-report.

Measures

Common survey guidelines [46] were adhered to in creating the survey and by incorporating the input of two mental health professionals. The survey was pretested by three people meeting eligibility criteria for participation. In addition, an instructed response question was utilized to identify careless responders and the data from such participants was removed from the sample [47].

The dependent variable was measured with the question “In the past 6 months, about how often did having this family member in your life make you happy?” (Response options = not at all, less than once a month, once a month, once a week, and more than once a week).

All variables pertaining to the relationship interactions of participants and RR were measured regarding the prior 6 months. Level of caregiving with activities of daily living provided by RR towards participants was measured with the sum of seven questions adapted from the Family Experiences Interview Schedule [48], yielding a Cronbach’s alpha in the present study of 0.81. Total scores range from 0 to 28 with higher scores indicating a greater level of caregiving. Frequency of financial assistance provided by RR to participants was measured with the sum of two questions (e.g., “About how often did this family member personally pay for or give you money for basic living necessities such as food, transportation, or rent?”) modified from a previous study [24], yielding Cronbach’s alpha of 0.78 in this study. Total scores range from 0 to 8 with higher scores indicating more frequent financial assistance. Routine limit-setting practices used by RR toward participants was measured with the sum of 5 questions constituting the Routine Limit-Setting subscale of the Family Limit-Setting Scale [37]. These questions were modified to reflect that participants were being asked about limit-setting practices engaged in toward them (instead of by them). Questions were asked about the frequency in which the RR engaged in specific behaviors toward participants aiming to influence their behaviors (for example: “…how often did your RR suggest that you should change your behavior?”). Total scores ranged from 0 to 20 with higher scores indicating greater use of routine limit-setting practices. These questions have a Cronbach’s alpha of 0.87 in this study. Perceived emotional over-involvement of RR regarding participants was measured with the sum of two dichotomous questions (“Is this family member always nosing into your business?” And “Is this family member overprotective with you?”) adapted from the Level of EE Scale [49]. Scores ranged from 0 to 2 with higher scores indicating greater perceived emotional over-involvement. Adapted from a previous study [50], victimization and perpetration of psychological abuse between participants and RR were each measured with the sum of 4 questions assessing the frequency of acts of criticism, yelling, destruction of property, and threats of minor violence. Total scores range from 0 to 20 with higher scores indicating greater psychological abuse. These measures yielded Cronbach’s alphas above 0.70 in this study. Victimization and perpetration of physical violence between RRs and participants were measured with questions closely adapted from the MacArthur Community Violence Instrument [51]. In the present analyses, the occurrence of violence was defined as any act of threatening with a knife, gun, or other lethal object with weapon in hand or any act of battery regardless of whether resulting in injury or involving the use of a weapon; sexual violence was not measured. The following variables were measured dichotomously (yes/no): co-residence of RR and participants, RR officially managing the money of participants as their representative payee, and RR unofficially managing the money of participants. Remaining relationship interaction variables were measured with single straight-forward questions, with ordinal response options of “less than once a month”, “once a month”, “once a week”, and “more than once a week”. All variables pertaining to participants and RR (listed in Tables 1 and 2) were measured with single straight-forward questions.

Table 1 Characteristics of participants and reference relatives (RR)
Full size table
Table 2 Factors associated with high relationship quality between participants and reference relatives (N = 523)
Full size table

Analysis

Multivariate ordered logistic regression (for ordinal dependent variables) was performed; however, tests indicated the proportional odds assumption was violated for several independent variables. As such, dichotomous logistic regression [52] was employed. We dichotomized the dependent variable where considered to have clinical significance and yielding the greatest statistical power. High relationship quality was defined as a response of “more than once a week” to the question “In the past 6 months, about how often did having this family member in your life make you happy?”; low relationship quality was defined as a response of “once a week” or less. Unadjusted (bivariate) odds ratios (OR) for independent variables were first computed. All independent variables with statistically significant unadjusted OR (alpha level < 0.05) were then entered into a multivariate model, estimating adjusted OR. The secondary objective of examining the extent to which the relationship between routine limit-setting practices and relationship quality is mediated by perceived emotional over-involvement was examined by completing the four steps recommended by Baron and Kenny [53]. We used Sobel testing [54] to assess the statistical significance of the indirect effects of perceived emotional over-involvement on relationship quality, with established adjustments made for dichotomous dependent variables [55].

Results

Descriptive Statistics

Participants (N = 523) reported that they became aware of the study through the following sources: National Alliance on Mental Illness (n = 127, 24%), Depression and Bipolar Support Alliance (n = 118, 23%), consumer-run organizations (n = 68, 13%), Facebook groups pertaining to mental illness (n = 58, 11%), inpatient or outpatient mental health treatment (n = 37, 7%), the International Association of Peer Supporters (n = 16, 3%), state division of mental health listservs (n = 12, 2%), a chronic suicide support forum (n = 9, 2%), and We Search Together (n = 7, 1%). 14% (n = 71) provided nondescript responses (e.g., “online”) or did not respond. The mean ± SD age of participants was 43.1 ± 14.39 years (range 18–73). The vast majority were female (80%) and Non-Hispanic Caucasian (88%). One-third were employed full-time, with 21% employed part-time. The median category of education level was “some college”, with the median category of annual income being $10,000 - $19,999. Nearly three-fourths of participants reported their primary mental health disorder was either bipolar disorder (39%) or depression (34%). 84% (n = 442) were regularly taking mental health medications and 16% (n = 83) had been hospitalized for psychiatric reasons in the past year. Participants reported residing in 42 states in the U.S. (not including AL, DE, HI, MS, MT, ND, NV, and WV) and the District of Columbia. Nearly half of participants were related to RR as their romantic partner (46%), with 27% being their child, and the remaining 27% being other family members (parent, sibling, grandchild, etc.). Nearly three-fourths (70%, n = 367) of participants and RR lived together in the past 6 months.

The mean ± SD age of RR was 50.32 ± 16.89 years (range 18–87). Half (n = 260) were female, with the race/ethnicity of RR mirroring that of participants. Close to half were employed full-time, with 12% employed part-time. The median category of education level was “some college”, with the median category of annual income being $20,000 - $39,999. Nearly one-third of RR were reported to have a mental health disorder. More detailed information regarding the characteristics of participants and RR is provided in Table 1.

Participants’ responses to the question “In the past 6 months, About how often did having this family member in your life make you happy?” were “not at all” (6%, n = 30), “less than once a month” (7%, n = 35), “once a month” (9%, n = 45), “once a week” (18%, n = 92), and “more than once a week” (61%, n = 621).

Correlates of Relationship Quality

Unadjusted and adjusted odds ratios (OR) for high relationship quality are presented in Table 2. Adjusted OR indicate that high relationship quality was significantly more likely when participants and RR had more frequent in-person contact, when participants more frequently provided assistance to RR with activities of daily living, and when the level of caregiving provided by RR to participants was greater. Inversely, high relationship quality was significantly less likely when participants were related to RR as their child or other family member (as compared to their romantic partner), and when there was a greater level of perceived emotional over-involvement and psychological abuse by RR toward participants. In multivariate analyses, relationship quality was not significantly associated with any demographic or clinical characteristic of participants or RR.

Mediation

The adjusted logit beta for routine limit-setting with relationship quality after controlling for all variables in the final multivariate model (presented in Table 2) except perceived emotional over-involvement was − 0.069 (95% CI [-0.13, − 0.01], p = .034). The regression beta for routine limit-setting with perceived emotional over-involvement after controlling for all other independent variables in the multivariate model was 0.05 (95% CI [0.10, 0.22], p < .001). The adjusted logit beta for routine limit-setting with relationship quality after controlling for all variables in the final multivariate model (presented in Table 2) including perceived emotional over-involvement was − 0.044 (95% CI [-0.11, 0.22), p = .19). Results of the Sobel test, which tests whether the mediation effect is statistically significant, indicated that the relationship between routine limit-setting and relationship quality is significantly mediated by perceived emotional over-involvement (z = -2.92, SE = 0.009, p < .001).

Discussion

Most participants in this study reported high levels of relationship quality with their RR, with more than 60% responding that having their RR in their life made them happy more than once a week. As argued in the introduction, positive family qualities are clearly common in families that include a member with SMI. This study found that as compared to participants who reported their romantic partner was their closest relationship, participants who indicated their parents (p = .001) or other family members (p = .02) as their closest relationship were less likely to report high relationship quality within those relationships. Romantic relationships tend to be more voluntary than parental or other familial relationships. One possible explanation for the above findings is that when people are in romantic relationships with low relationship quality, they may choose to leave the relationship [56]. Conversely, it may be more difficult to sever relationships with family of origin, resulting in lower levels of relationship quality as compared to romantic partners.

Remaining factors found to be associated with relationship quality examined in this study may serve as targets for clinicians to support and facilitate recovery among persons with SMI. In efforts to enhance the relationship quality of persons with SMI, it is advisable for clinicians to assess clients’ perceived relationship quality with family members. Given the importance of social relationships in experiencing recovery, clinicians should assess relationship quality when working with persons with SMI. When clients express an interest in improving relationship quality, it may be helpful for clinicians to assess the presence of factors that contribute to and detract from relationship quality. Then, in collaboration with clients, clinical efforts may include enhancing the presence of interactions positively associated with relationship quality (conceptualized as supportive interactions) and decreasing the occurrence of interactions negatively associated with relationship quality (conceptualized as problematic interactions).

From the perspective of persons with SMI, this study found that one supportive interaction associated with relationship quality is increased rates of in-person contact with RR (p < .001). Based on research in the general population, relationship quality is likely bolstered not merely by spending time together but spending time together while engaged in positive interactions. For example, Hogan and colleagues [57] found that couples who spent a larger proportion of their time talking to each other without arguing reported greater relationship satisfaction than those who spent less time together. However, if in-person contact was spent arguing, relationship satisfaction was lowered when couples spent more time together [57]. As it is likely that positive, in-person contact with relatives may increase overall relationship quality in persons with SMI, clinicians should encourage clients to spend more time with their family members engaging in positive interactions.

In this study, participants providing assistance to RR in completing activities of daily living was correlated with high relationship quality (p = .001). This finding is comparable to studies in the general population that find the provision of support contributes to relationship quality [30]. As previously described, most persons with SMI provide a level of support to their family members [11]. It may behoove clinicians to recognize the capacity of persons with SMI to contribute to their families and households; furthermore, clinicians should support and encourage persons with SMI to invest in their familial relationships through helping with meal preparation, shopping, and other chores. Doing so may not only increase relationship quality but may also facilitate the process of recovery.

It was found that relationship quality was positively associated with receiving caregiving from the RR (p = .001). Within the general population, caregiving is considered to be an indicator of support [58] and receiving higher levels of support is related to higher relationship quality [30]. Clinicians may encourage clients to be open to receiving care from their relatives. Unfortunately, within the literature on persons with SMI, caregiving has largely been conceptualized as negative and burdensome on family members [59]. Framing caregiving as a problem to be coped with likely creates increased stigma for persons with SMI and, by extension, for their families and is not in alignment with a recovery-oriented approach. Clinicians should recognize the mutuality within the relationships of persons with SMI and encourage clients to feel comfortable receiving familial support, which is often an indicator of a strong relationship [17].

This study also found that experiencing psychological abuse (p < .001) and perceiving RR as emotionally over-involved (p = .001) were negatively associated with relationship quality. Among the broader literature on relationship violence, increased levels of psychological abuse are related to lower levels of relationship satisfaction [34]. Clinicians should monitor when clients are experiencing psychological abuse and perceive their relatives to be emotionally over-involved, as these experiences likely decrease relationship quality, as well as impede recovery and exacerbate symptomology. When clinicians are aware of the presence of these interactions, they may offer or refer out to family-based interventions such as psychoeducation [60] or family therapy [61] which have been found to reduce EE. The finding that the relationship between limit-setting practices used by RR and relationship quality is mediated by perceived emotional over-involvement (p < .001) suggests that when family members set limits with persons with SMI, persons with SMI often perceive family members to be emotionally over-involved, with this perception deteriorating relationship quality. There is considerable variation in the strategies family members can use to set limits (e.g., verbal encouragement, contingency contracts) as well as the immediate and broader relationship contexts in which limits can be set (e.g., regarding warmth, trust, and collaboration). It is likely that many of these varying elements impact the degree to which persons with SMI interpret family members as emotionally over-involved when they set limits. Among the general population, receiving unwanted advice (but not advice generally) from parents is predictive of negative mood among adult children [62]. Also, among the general population, research has found that receiving social support from a partner is beneficial only when recipients perceive the support as responsive—i.e., that their partner understands, validates, and cares for them [63]. We recommend that practitioners support family members and persons with SMI in agreeing when it is necessary to set limits. Moreover, when family members do set limits with persons with SMI, they should be supported in doing so in a manner that strives to collaborate with persons with SMI (e.g., agreeing in advance on the situations in which limits will be set and how limits will be set) and conveys care, warmth, and understanding (e.g., explaining with warmth that limits are being set out of concern for the wellbeing of persons with SMI). Future research should more comprehensively explore the many strategies family members use to set limits with persons with SMI and examine potential differences in the extent to which particular strategies—and the relational contexts in which they are employed—are related to relationship quality.

Finally, this study found that clinical characteristics of participants were not significantly related to relationship quality, consistent with related findings in the literature [23]. These results have implications for clinical practice as a clinical focus on a reduction of psychiatric symptoms may not meaningfully improve relationship quality and may or may not be in alignment with clients’ goals. Overall, this paper suggests that a meaningful way to increase relationship quality among persons with SMI may be by increasing supportive interactions between persons with SMI and their family members and decreasing problematic ones. Collectively, these findings indicate that family relationship quality of persons with SMI are influenced by factors similar to those for members of the general population and may benefit from similar treatments and targets of intervention, particularly if symptom management is not in alignment with clients’ goals. Clinicians who focus on clients’ overall levels of functional wellbeing, which includes their relationship quality with family members, are practicing in alignment with a recovery-oriented approach as they support clients in living enriched and meaningful lives.

Strengths and Limitations

A strength of the present article is that it used a large national sample of persons with SMI and that it fills a considerable gap in the literature. Limitations of the sample employed is that it had inadequate diversity regarding sex and race/ethnicity. Unfortunately, recruiting from mental health support and education organizations commonly produces samples that are disproportionately female and Non-Hispanic Caucasian [64, 65], likely reflecting the member composition of such organizations. The sample also had disproportionately high levels of educational attainment. Examining other proxies for impairment, however, participants appear similarly or more impaired than the U.S. population of persons with SMI. It is estimated that 54.5% of persons with SMI are employed [66] and 7% have received inpatient mental health treatment in the past year [67]. In the present sample, 53% of participants were employed and 16% had been hospitalized in the past year. Research on this topic going forward should involve a more diverse sample, ideally utilizing a longitudinal design enabling claims about the temporal ordering of relationship dynamics in connection with relationship quality.

Another limitation of the present study is that the dependent variable was measured with a single question. Unfortunately, there is considerable variation in how relationship quality is measured, particularly within different relationship types (i.e., romantic partners, family members, friendships). In the literature on the general population, increased happiness derived from a romantic relationship or friendship is known to be positively associated with relationship satisfaction [68,69,70,71]. Furthermore, happiness in relationship is occasionally used as a dimensional measure of relationship satisfaction [72]. As such, we consider the question we used to measure relationship quality to be a proxy of this construct. Future research conducted in this area should utilize a multidimensional measure of relationship quality.