from support to overload: patterns of positive and ... · jointly, in the family networks of adults...

Fr

Ma

b

c

d

a

AA

KFSSCM

i(K1fePpmnapsw

“tS

h00

Social Networks 47 (2016) 59–72

Contents lists available at ScienceDirect

Social Networks

jo ur nal homepage: www.elsev ier .com/ locate /socnet

rom support to overload: Patterns of positive and negative familyelationships of adults with mental illness over time

arlène Sapina,b,∗, Eric D. Widmerc,b, Katia Iglesiasd

Swiss Center of Expertise in Social Sciences (FORS), University of Lausanne, SwitzerlandSwiss National Centre of Competence in Research “LIVES – overcoming vulnerability: life course perspectives”, SwitzerlandDepartment of Sociology, University of Geneva, SwitzerlandCentre for the Understanding of Social Processes (MAPS), University of Neuchâtel, Switzerland

r t i c l e i n f o

rticle history:vailable online 19 May 2016

eywords:amily network structureocial capitalocial supportonflictental health

a b s t r a c t

Family relationships account for much of the support available to individuals with mental illness.Although some studies have acknowledged the importance of family support, and while others haveunderlined the harmful effects of negative relationships, research has seldom empirically considered thecomplex web of positive and negative relationships in family networks. This research hypothesised thatsocial capital has distinct consequences for psychological health depending on the presence or absence ofnegative family relationships. Through a five-wave follow-up of 60 individuals undergoing psychotherapyin a private practice, the study explored the structural features of positive and negative relationships, con-sidered jointly, in the family networks of adults with mental illness. Four patterns of relationships werefound: bonding social capital, bridging social capital, overload and ego-centred conflict. Compared to indi-
viduals within a bonding or bridging social capital pattern, those experiencing overload and ego-centredconflict patterns showed higher levels of psychological distress. These results highlight the importanceof considering the structural dimensions of positive and negative relationships together to understandthe lasting connection between family networks and the psychological health of individuals with mentalillness.
© 2016 The Authors. Published by Elsevier B.V. This is an open access article under the CC BY-NC-ND

Family members are crucial for mental health because fam-ly relationships are a primary resource of care and supportAgneessens et al., 2006; Fehr and Perlman, 1985; Furstenberg andaplan, 2004; Haines et al., 1996; Wellman and Worthley, 1989,990). At the same time, family networks may produce stress-ul and negative interactions among their members (Fingermant al., 2004; Newsom et al., 2005; Pagel et al., 1987; Unger andowell, 1980), which may contribute in their own way to poorsychological health. Research has stressed for a long time theultidimensionality of personal networks as well as the useful-

ess of using typology for capturing it (for instance, Pescosolidond Levy, 2002; Wellman and Potter, 1999). Accordingly, this study
roposed a typological approach of the structural dimensions ofupport and conflict within the family networks of individualsith severe psychiatric disorders. Overall, four types adequately
∗ Corresponding author at: Swiss National Centre of Competence in ResearchLIVES – overcoming vulnerability: life course perspectives” & Swiss Center of Exper-ise in Social Sciences (FORS) University of Lausanne, Geopolis, CH-1015 Lausanne,witzerland. Tel.: +41 21 692 38 42; fax: +41 21 692 37 35.

E-mail address: [email protected] (M. Sapin).

ttp://dx.doi.org/10.1016/j.socnet.2016.04.002378-8733/© 2016 The Authors. Published by Elsevier B.V. This is an open access article

/).

license (http://creativecommons.org/licenses/by-nc-nd/4.0/).

capture the diversity of relational patterns present in familynetworks. These were associated with the psychological health ofindividuals followed for a year and a half.

1. Supportive family relationships

Personal networks transcend narrow reciprocity or the sum ofreciprocity (Kadushin, 2002; Wellman and Worthley, 1990). Thesocial support provided by specific dyads is embedded in broadernetworks of relationships, within which structure matters (Faberand Wasserman, 2002; Walker et al., 1993). The flux of resourcesthat circulate in personal networks and family networks makeup much of the social capital that individuals mobilise to dealwith daily life, either to reduce life’s uncertainty (Kadushin, 2002;Wellman and Worthley, 1990) or to face hardships and life crises(Perry and Pescosolido, 2012).

With regards to a health-related crisis, research on personal
networks has highlighted the significant role that family mem-bers hold in the process of making the decision to seek help andenter into treatment (Carpentier, 2013; Carpentier and Ducharme,2003; Rogers et al., 1999; Wellman, 2000). It has also examined the
under the CC BY-NC-ND license (http://creativecommons.org/licenses/by-nc-nd/4.

dx.doi.org/10.1016/j.socnet.2016.04.002

http://www.sciencedirect.com/science/journal/03788733

http://www.elsevier.com/locate/socnet

http://crossmark.crossref.org/dialog/?doi=10.1016/j.socnet.2016.04.002&domain=pdf

http://creativecommons.org/licenses/by-nc-nd/4.0/










mailto:[email protected]

dx.doi.org/10.1016/j.socnet.2016.04.002











6 Netw

dpnp(ldamFinPipppeulwLnoDois

orm(and2ml1saooottfv1oteetp

srtws(p(s

0 M. Sapin et al. / Social

ifferent roles and functions that personal networks play in theathways through which individuals go during their illness jour-eys. Some studies have considered the extent to which healthroblems, as other disruptive life events, change personal networksCornwell, 2009; Perry and Pescosolido, 2012). Analyses of the evo-ution of the personal networks of individuals with severe mentalisorders have shown that network changes are closely related ton individual’s ability to adjust to the different stages of his or herental illness and treatment career (Perry and Pescosolido, 2012).

amily members hold a prominent position in such a network, play-ng a crucial role particularly in the first stages, when the supporteeds are the greatest (Carpentier et al., 1999; Karp, 2001; Perry andescosolido, 2012). Friends and relatives from the extended fam-ly might be activated during the most demanding and disturbingeriod following the mental illness onset, but generally, they do notrovide long-term assistance when formal treatment is in place androfessionals take over some of their support functions (Carpentiert al., 1999; Karp, 2001). Parent and adult children, in contrast,sually provide long-term assistance (Karp, 2001). Along these

ines, other studies have highlighted the burden that an individualith a mental illness might put on family caregivers (Karp, 2001;

oukissa, 1995; Muhlbauer, 2002; Rogers et al., 1999), as well as theegative consequences that professional intervention might haven caregivers and their family (Carpentier, 2013; Carpentier anducharme, 2003). In this line of research, the family relationshipsf individuals with severe mental illnesses seem to be importantn relation to their psychological health, as they themselves are toome extent caregivers for other family members.

Family members are key players for social support in researchn personal networks and health research. Other work similarlyeports that family relationships account for much of the infor-al support available to individuals with psychiatric disorders

Corrigan and Phelan, 2004; Evert et al., 2003; Tolsdorf, 1976). Liter-ture also emphasises that the social support provided by intimateetwork members positively affects psychological health, eitherirectly or by buffering stress (Berkman et al., 2000; Cohen et al.,000; Cohen and Wills, 1985; Kawachi and Berkman, 2001). Familyembers’ role as support providers for individuals with psycho-

ogical problems has been emphasised heavily (Fehr and Perlman,985; Hooley and Hiller, 1997; Pierce et al., 1996). Nevertheless,ocial support is a multifaceted and complex construct (Antonuccind Akiyama, 1995; Pierce et al., 1997a). In general, various typesf relationships ensure distinctly different types of resources, butverall, family members provide long-lasting multiplex aids moreften than other members of personal networks do. Research showshat close family members who provide major instrumental helpend to also offer emotional support, while family members orriends who are less close provide either emotional support or ser-ices, but generally not both (Agneessens et al., 2006; Haines et al.,996; Wellman and Worthley, 1989, 1990). In addition, not all typesf support have been considered as equally important in relationo psychological health; prominence has been given to perceivedmotional support (Lin et al., 1999; Thoits, 1995). In the context ofmotional support within family relationships, research stresseshe beneficial effects of family support and family cohesion forsychological adjustment (Moos et al., 1998).

Such approaches to family support were criticised for over-implifying the complexities of family configurations and theirole in adaptation (Coyne and DeLongis, 1986). In a configura-ional perspective (Widmer, 2010; Widmer and Jallinoja, 2008),hich defines families as personal networks, supportive relation-

hips with family members were conceptualised as social capital
Widmer, 2010) or the relational resources resulting from theossession of a durable network of acquaintances or recognitionBourdieu, 1985). In a structural perspective, two main types ofocial capital have been distinguished: bonding and bridging (Burt,
orks 47 (2016) 59–72

1995; Coleman, 1988; Szreter and Woolcock, 2004). Bonding socialcapital is characterised by network closure, that is, redundant tieswithin a group with a high density of relationships (Coleman,1988). In dense networks, most if not all individuals are intercon-nected, which strengthens expectations, claims, obligations, andtrust among members through the increased collective dimen-sion of normative control. Therefore, in dense personal networks,support has a collective nature, as individuals are likely to coordi-nate their efforts when helping another. Distinctly, others conceivesocial capital as bridging (Burt, 1995), as a function of structuralholes and brokerage opportunities: weaker connections amongsubgroups in a personal network create holes that provide focalpersons, such as brokers, with opportunities to mediate and controlthe flow of information among group members. Regarding motivesand fundamental human needs, bonding social capital addressessafety, fulfilling the need to be in a secure structure and offeringa sense of belonging. In contrast, bridging social capital serves thebasic human need for autonomy and self-reliance (Kadushin, 2002).In other terms, bridging social capital gives a feeling of mastery,competence, and control over the environment.

Social capital has been shown to have positive consequencesfor individuals in various contexts, promoting physical and psy-chological health (Kawachi and Berkman, 2001; Lin et al., 1999;Moren-Cross and Lin, 2006; Song et al., 2011). The distinctionbetween the bridging and bonding types of social capital has beenconsidered promising (Whitley and McKenzie, 2005). However,apart from a few exceptions, the structural features of social capitalhave seldom been used empirically in physical and mental healthresearch. In a study on older adults and their personal networks,it was notably found that poor physical and mental health wererelated to bonding social capital and that bridging social capitalmight be too demanding for individuals with mental health andcognitive impairments (Cornwell, 2009).

2. Supportive and upsetting family relationships

Family configurations often constitute a significant source ofstress, conflict, and ambivalence (Connidis and McMullin, 2002;Widmer, 2010; Sapin, 2013). In highly cohesive family contexts,interferences from and control by family members are frequentand possibly consequential (Johnson and Milardo, 1984; Widmer,2010). Indeed, empirical research has increasingly suggested thatpositive and negative contents are inherent in personal relation-ships (DeLongis et al., 2004; House et al., 1988; Rook, 1984).Research has emphasised the significant effects of social strainsand social negativity on mental health (Bertera, 2005; Pagel et al.,1987; Newsom et al., 2005; Schuster et al., 1990). The literaturehas called for considering all supportive and negative relation-ships in personal networks, stressing that their effects interact(DeLongis et al., 2004; Major et al., 1997; Rook, 1998; Schuster et al.,1990). On the one hand, highly supportive personal networks maymoderate the immediate, detrimental impact of problematic rela-tionships (Schuster et al., 1990); on the other hand, conflict maydecrease the positive impact of supportive relationships (DeLongiset al., 2004; Okun and Keith, 1998). In addition, research has foundthat the effects of the upsetting relationships might be heightenedin personal networks also characterised by supportive relation-ships (Pagel et al., 1987, 1987; Uchino et al., 2004). The scarcityof resources within personal and family networks leads to conflictsof interest and competition, even if it is latent, hidden or misunder-stood (Sprey, 1969, 1999; Sillars et al., 2004; Walker et al., 1993;
Widmer, 2010). Therefore, social capital should be considered inrelation to negative relationships within relational patterns thathave structural features of their own (Kearns and Leonard, 2004;Labianca et al., 1998).

M. Sapin et al. / Social Networks 47 (2016) 59–72 61

Table 1Basic demographic characteristics, main diagnosis, and mental health problem history of study participants at baseline (N = 60).

Variable Range Mean SD % of sample

Demographic characteristicsAge 25–64 42.2 10.1Female 73.8%Primary education 33.3%Secondary education 45.2%Tertiary education 21.4%

Main diagnosisAnxiety or mood disorders 21.7%Personality disorders 66.7%Psychotic disorders 11.7%

History of mental health problemsInstitutionalised in the past 51.7%Beginning of the treatment in another practice or institution 71.1%

0–38

5–60

salmlhdpWtpesbfmM

2

crocwnpas(titdsstc

ttahr

Onset of the first psychiatric episode (in year)

Age of onset of mental disorder 1In psychiatric treatment since less than two year

In addition, positive and negative relationships may have per-isting effects over time. Many relationship difficulties are ongoing,nd these persistent, negative social exchanges can be particu-arly injurious to psychological health. Negative social exchanges

ay represent stable or chronic patterns of relationships withonger-term consequences (Newsom et al., 2008). Previous workighlighted that supportive relationships are associated with loweristress levels in the short run, while negative relationships arearticularly detrimental in the long run (Newsom et al., 2003).hen examined jointly with stressful life experiences, such as rela-

ionship losses, disruptive events and functional impairment, thearticular nature of positive and negative relationship effects onxperienced distress varies according to the type and level of thetressors (August et al., 2007). Therefore, to assess the interrelationetween family networks and psychological health, the structuraleatures of relationships need to be analysed as whole patterns at

ore than one point in time (Bergman et al., 1991; Bergman andagnusson, 1997).

.1. Perspective and aims

Some evidence stresses that the effect of social capital on psy-hological health should be considered in relation to negativeelationships with relational patterns that have structural featuresf their own. As such, the development of a typology linking socialapital and conflict seems necessary to uncover the various ways inhich positive and negative relationships intermingle in personaletworks. Various typologies of personal networks have been pro-osed (for instance, Agneessens, 2013; Wellman and Potter, 1999),lthough none, to our knowledge, included conflict as a dimen-ion, despite its obvious importance for small group dynamicsWidmer, 2010; Simmel, 1999; Coser, 1956). In general, configura-ional approaches are concerned with how the pattern of multiplendependent variables is related to the dependent variables ratherhan with how individual independent variables are related to theependent variables (Delery and Doty, 1996). Numerous dimen-ions of personal networks impact mental health, and researchtresses that it might be the overall “feel and look” of such networkshat matters (Pescosolido and Levy, 2002), to which support andonflict greatly contribute.

This study explored supportive and straining relationships inhe family networks of individuals with mental illnesses. Rather
han considering positive relationships (social capital) and neg-tive relationships (conflicts) as alternative interactions, it wasypothesised that they merge into qualitatively distinct patterns ofelationships (H1). Second, it examined the interrelation between
10.5 9.732.6 9.4

30.0%

patterns of family relationships and psychological health. It waspostulated that family-based social capital has different conse-quences for individuals depending on the negative relationshipsin which it is embedded (H2). Overall, we expected that supportwhen it comes with conflict is associated with poorer psychologi-cal health compared with situations where it comes alone (H2a).In addition, the patterns of family relationships, depending onthe resources they provide and the strains they cause, might beexpected to have not only synchronic consequences on psycho-logical health but also a cumulative impact in the longer term,as individuals were embedded in enduring straining patterns. Itwas hypothesised that patterns of family relationships also havediachronic effects on the levels of distress experienced by individ-uals with mental illnesses (H2b).

3. Methods

3.1. Participants

Data were collected during an exploratory study on familynetworks and mental health (Widmer and Sapin, 2008). To gainaccess to individuals with mental illnesses, a purposive strategy(Bernard, 2012) was used to make contacts through a privatepsychotherapy practice in the French area of Switzerland. Afterthe ethics committee granted approval, all clients of the privatepractice were informed formally of the study. Among a hun-dred patients undergoing psychotherapy at the private practiceat the beginning of the study, sixty-one individuals agreed to beinterviewed. One case was excluded from the analysis becauseof missing precise information on mental health at baseline andat subsequent time points. As it is difficult to obtain access tothis population, the response rate of this study (60%) is usual(e.g. Link et al., 1989; Perry and Pescosolido, 2012). All respon-dents were outpatients and went through treatment over thefive-wave assessment. Characteristics of the sample are presentedin Table 1. On average, respondents were in their mid-40s. The aver-age age was 43, and 74% were women. They presented a varietyof serious mental disorders, including personality disorders, psy-chosis, and anxiety and mood disorders. When grouped accordingto the dominant diagnostic of the 10th revision of the Interna-tional Statistical Classification of Diseases (ICD-10) (World HealthOrganization [WHO], 1993), the sample consisted of 40 individ-
uals with personality and behaviour disorders (F60–F69); sevenindividuals with schizophrenia, schizotypal or delusional disorders(F20–F29); and 13 individuals with both or either mood (F30–F39)and neurotic, stress-related or somatoform disorders (F40–F48).

6 Netw

Ttpppspotaa

3

ot2iDblivwawfitobibtw

dwwtacsfappvrejfnLietwoa

3

fn


he individuals in this sample had a lower level of educationhan the average in Switzerland. The majority of them each had arevious history of mental health problems and was referred byublic health institutions of primary care. All patients underwentsychotherapy for periods of time ranging from several months toeveral years, most of them for more than two years (70%). The sam-le included individuals for whom the onset of the mental disordersccurred at different ages (with a range from 15 to 60). Amonghem, 71% had previous experiences of psychiatric treatment, andbout half had each been institutionalised in a psychiatric facilityt one point in his or her life.

.2. Procedure

Every 3 months over a year and a half, the respondents filledut the family network method (FNM) questionnaire, an egocen-ric network collection tool (Widmer, 1999; Widmer and La Farga,000; Widmer et al., 2005). At each interview wave, the partic-

pants also completed the Symptom Check-List-90-R (SCL-90-R,erogatis, 1983; Pellet, 1997), an inventory commonly used byoth clinicians and researchers to evaluate psychological prob-

ems and symptoms. Both questionnaires were filled in duringn-person interviews by trained health professionals from the pri-ate practice. Among the 60 outpatients, the entire five-wave studyas completed by 43 respondents. Four waves of interviews were

vailable for 48 respondents; 12 patients answered only one or twoaves (20% of the sample). Attrition was the highest between therst and second waves (9 individuals dropped out of the observa-ion), primarily due to individuals who quit therapy temporarilyr definitively; one quit because of healing, while the others quitecause of relational difficulties with the therapist. The analysis

ncluded all 252 interviews, because up to five waves were possi-le for each of the 60 individuals. Throughout all 252 interviews,wo observations on mental health and one on family networksere missing.

In order to control for the potential misleading effect of theropouts (20% of the individuals interviewed at waves 1 and 2),e compared individuals who were interviewed in at least fouraves of the follow-up with those who left the study before

he third wave on basic demographic variables, such as age, sexnd level of education, as well as on main diagnosis. No statisti-ally significant differences between the dropouts and those whotayed in the sample were found for any of the variables. The dif-erences in their mental health problem history measures werelso statistically non-significant, except that individuals who hadrevious experience of psychiatric treatment in another privateractice or institution were less likely to drop out of the obser-ation (�2(1) = 4.36, p < .05). The attrition might, to some extent, beelated to stigma associated with mental health. Individuals, at anarlier stage of their mental health trajectory, might be more sub-ect to the stigma associated with being labelled and might morerequently experience withdraw from social interactions that wereot with their closest friends and family members (Link et al., 1989;ink and Phelan, 2001; Perry and Pescosolido, 2012). However, thendividuals who dropped out of the observation, on average, did notxperience higher levels of distress at baseline (t = 1.19). In ordero control for the potential effect of attrition, all of the analysesere performed on the whole sample, as well as on the subsample

f individuals for whom at least four waves of observation werevailable.

.3. Measures

At each wave, the respondents were first invited to list all of theiramily members who played a significant role, either positive oregative, in their lives that year. The size and composition of their

orks 47 (2016) 59–72

family network were therefore time variant through the follow-up. The average size across the waves ranged from 9.3 (SD = 4.0; atwave 1) to 6.6 (SD = 2.8; at wave 5). The correlations between thesize scores across waves were medium to high (ranging from r = .32to r = .62). Based on this free definition of their family network,the respondents later reported all of the relationships of supportand conflict that existed among their family members, accordingto them. As in other cognitive network studies (Krackhardt, 1987),the respondents described both their relationships with their fam-ily members and the relationships among their family members.The question measuring support asked, “Who would provide X withemotional support when he or she is sad during routine or minortroubles?” The question for conflict asked, “Every family has its con-flicts and tensions. In your opinion, who often makes X upset orangry?” The respondents were asked these questions about everyfamily member whom they had identified as significant at the timeof the interview.

To investigate the patterns of relationships characterising fam-ily networks, two classical dimensions of social capital in personalnetworks were considered: cohesion and centrality (Wassermanand Faust, 1994). Indices were computed with UCINET 6 (Borgattiet al., 2002). The centrality of the respondents was computed byreferring to three measurements for both support and conflict rela-tionships. The in-degree centrality of the respondents indicated thenumber of individuals whom the respondents supported, whiletheir in-degree centrality for conflict counted the relatives whowere bothered by the respondent. The out-degree centralities ofthe respondents represented the individuals who supported therespondents or caused conflict for them. In- and out-degree cen-tralities refer to a local type of network centrality (Wasserman andFaust, 1994). To also have a less local measure of centrality andtherefore a better assessment of bridging social capital, we alsocomputed the respondents’ betweenness centrality, which cap-tures the proportion of connections in the family network in whichthe respondents occupy an intermediary position (Wasserman andFaust, 1994).

Cohesion in the family network was operationalised by den-sity, which is an indicator of group cohesion evidenced in positiverelationships such as friendship and interaction (Wasserman andFaust, 1994). It corresponds to the number of activated connectionsdivided by the number of potential connections. The densities ofboth emotional support and conflict relationships were calculated.The greater the density of supportive relationships, the greater thesocial cohesion is, while increased density of negative relationshipswould generally be the opposite, except if all negative ties impli-cate a few individuals. In this case, the group might bond overcommon, central problematic individuals (Everett and Borgartti,2014). In addition to density, at the local level, the number of tiesbetween individuals directly connected to the respondent was alsocomputed for support and conflict.

Psychological health was measured using the Symptom Check-List-90-R (SCL-90-R, Derogatis, 1983; Pellet, 1997). The answers tothe 90 items, rated on a 5-point Likert-scale (ranging from 0, “notat all,” to 4, “extremely”), can be combined in three indices thatprovide global measures of mental health: the Global Severity Index(GSI), which includes all symptoms of the nine specific sub-scales;the Positive Symptom Total (PST), which indicates the diversity ofsymptoms (considering only their presence or absence); and thePositive Symptom Distress Index (PSDI), which is the GSI dividedby the PST, a measure of individual distress independent of the typeof experienced psychiatric symptoms. As the sample was composedof individuals experiencing various psychiatric disorders, our anal-
yses focussed on the PSDI, which assessed levels of psychologicaldistress independently of the number or kind of symptoms experi-enced by individuals. The average PSDI scores across waves rangedfrom 1.97 (SD = .51; at wave 3) to 2.17 (SD = .54; at wave 1). The

Netw

c(

3

mcatmApciccR

wOssbtpapit

isrRaantratttcuS(wstdauwtoddateswcp

M. Sapin et al. / Social

orrelations between the PSDI scores across the waves were highranging from r = .61 to r = .75).

.4. Analytical strategy

Because we were concerned with how the whole patterns ofultiple intercorrelated independent variables are related to psy-

hological health as a dependent variable, we developed a clusternalysis that considered all network indices together to find pat-erns related to Hypothesis H1. Accordingly, we applied a validated

ultivariate statistical approach of clustering (Lebart et al., 1997).fter factor reduction (Tabachnick and Fidell, 1996) of all sup-ort and conflict centrality and cohesion indices, using principalomponent analysis with varimax rotation to extract the signif-cant factors, the factor scores were inputted into a hierarchicallustering analysis based on correlation distances and the Wardlustering algorithm (Lebart et al., 1997). This analysis was run in

(R Development Core Team, 2011).The Pearson correlations among all of the global and local net-

ork indices of cohesion and centrality are presented in Table 2.verall, numerous indices are intercorrelated, not only within the

upport or conflict relationships, but also between both dimen-ions. Note that many indices related to conflict are not negativelyut positively correlated with indices related with support. In addi-ion, the correlations existing between the various network indicesrohibit their use as distinct independent variables in regressionnalysis because of collinearity issues. These results strongly sup-ort a typological approach that considers all indices together. For

nformation purposes, the correlations between these indices andhe network size were added to Table 2.

The relationship between the patterns of positive and negativenterdependencies and levels of psychological distress (Hypothe-is H2) was then assessed using multilevel mixed-effects linearegression models, with participants as random effect (Bryk andaudenbush, 1992; Hox, 2002; Pinheiro and Bates, 2000; Snijdersnd Bosker, 2011). Such random-intercept regression modelsccount for the nested nature of our repeated measures (level 1,1 = 250) within respondents (level 2, n2 = 60) through time. Suchwo-level variance component models, with the respondent-levelandom intercepts, account for within-subject heterogeneity andre a good fit for unbalanced data containing missing observa-ions across waves, which was the case here. They also allowedhe cumulative effects of potential straining patterns of family rela-ionships through time to be explored at the individual level, whenonsidering the proportion of straining patterns across the follow-p (Bryk and Raudenbush, 1992). Analyses were performed usingTATA’s xtmixed command with maximum likelihood estimationStataCorp, 2011). Because the data on individual distress (PSDI)ere skewed to the right, some of the respondents reporting higher

cores, a log transformation of the individual distress was appliedo satisfy the random intercept model’s assumptions of normalistribution and the homoscedasticity of random effects (Kohlernd Kreuter, 2009; Snijders and Berkhof, 2004). A first series ofnivariate analyses, with all variables estimated as fixed effects,as carried out to independently assess the relationships among

he basic demographic variables, the main diagnosis, and the setf measures on mental health problem history at baseline withistress level, at the individual level. The independent effects onistress of the patterns of family relationships and network size, as

control variable, were also assessed at observation level. In ordero investigate Hypothesis H2b, we tested the independent fixedffect of the proportion of straining patterns of family relation-
hips across the follow-up, at the individual level, while estimatinghether being embedded in enduring straining patterns was asso-
iated with higher levels of distress. Finally, we also estimated theotential effect of the averaged network size through the follow-up.

orks 47 (2016) 59–72 63

Following this first set of analyses, we implemented a combina-tion of forward and backward stepwise strategies (Collett, 2003;Snijders and Bosker, 2011) to select the variables that proved to bestatistically significant in a multivariate model. Finally, we carriedout a fixed-effects panel regression analysis to further gauge therobustness of the association between the patterns of positive andnegative family interdependencies and the level of distress expe-rienced by our sample (Kohler and Kreuter, 2009). Fixed-effectsmodels control for time-invariant unobserved heterogeneity ofindividuals (Allison, 2009; Kohler and Kreuter, 2009) and partlysolve selection issues (Clarke et al., 2010). However, such fixed-effects models for panel data cannot be used to investigate specifictime-invariant individual characteristics, such as the cumulativeeffect of straining patterns through time. Multilevel mixed-effectsregressions will therefore be presented. All of the models were per-formed for the whole sample and for the subsample of individualsfor whom all observation points were available.

4. Results

This section first presents the clustering of positive and negativefamily relationships. Then, the interrelations between the patternsand individual distress are described, as well as the cumulativestraining effect of these patterns through time.

4.1. Clusters of family relationships

Following the principal component analysis, five factors witheigenvalues greater than one were retained, altogether explaining82.3% of the variance. The first rotated factor (34% of the variance)represents the in- and out-degree neighbourhood of the respon-dents in their support relationships. It captures the extent to whichthe respondent’s direct family environment features a high or a lowlevel of support relationships and whether the structural featuresof these ties present a bonding form. In other words, it refers to thenumber of supported family members by the respondent, as wellas the number of family members who support the respondent andthe number of supportive ties that exist among them. The secondfactor is similar but for conflict relationships (20.4% of the vari-ance); it represents the in-degree centralities of the respondentsin conflicting relationships, as well as the number of conflict tiesthat exist between their conflicting alters. The third factor (11% ofthe variance) embodies the betweenness centrality of the respon-dents in their conflict relationships. Similarly, the fourth factor (9%of the variance) captures the betweenness centrality of the respon-dents in their support relationships, and the fifth factor (7.8% of thevariance) measures the extent to which the conflict and supportnetworks are densely connected overall.

Then, we inputted the five factor scores into a hierarchical clus-ter analysis (Rapkin and Luke, 1993; Lebart et al., 1997; Everittet al., 2011). Several standard cluster cut-off criteria (average sil-houette width [ASW], point biserial correlation [PBC], and Huber’sGamma Somers’ D [HGSD]) clearly supported four clusters as thebest grouping, compared to adjacent cluster numbers (Hennig andLiao, 2013; Kaufman and Rousseeuw, 1990). The global ASW (.40for four clusters) and the dendrogram (see Fig. 1) provided strongevidence that the data were highly structured in these four pat-terns. Table 3 presents the mean of the network indices for eachof the four clusters. Each of them attained a satisfying ASW score(ranging from .25 to .57). The differences among the clusters werestatistically significant.

Cluster 1 (24.6% of the observations) exhibits a pattern of bond-ing social capital, as the density of support relationships was highand the betweenness centrality of the respondents in their supportrelationships was low. The number of ties between alters directly

64

M.

Sapin et

al. /

Social N

etworks

47 (2016)

59–72

Table 2Pearson correlations between support and conflict network indices used to define structural patterns of positive and negative relationships (N = 252).

1 2 3 4 5 6 7 8 9 10 11 12

Support 1. Density of supportrelationships

–

2. Betweenness centralityof respondent in supportnetwork

.21*** –

3. Number of alters towhom respondentprovided support

.09 .42*** –

4. Number of supportiveties between alters towhom respondentprovided support

.32*** .11. .72*** –

5. Number of alters whoprovided support torespondent

.18*** .51*** .53*** .46*** –

6. Number of supportiveties between alters whoprovided support torespondent

.40*** .13* .44*** .77*** .64*** –

Conflict 7. Density of conflictrelationships

.29*** .06 −.14* −.07 −.20** −.05 –

8. Betweenness centralityof respondent in conflictrelationships

.06 .04 .04 .08 .05 .12. .36*** –

9. Number of altersbothered by respondent

−.02 .12. .35*** .25*** .26*** .23*** .22*** .64*** –

10. Number of conflict tiesbetween alters bothered byrespondent

.04 .16* .34*** .28*** .18** .17** .36*** .34*** .64*** –

11. Number of alters whocaused conflict forrespondent

−.02 .13* .34*** .28*** .23*** .21*** .28*** .59*** .58*** .53*** –

12. Number of conflict tiesbetween alters who causedconflict for respondent

−.04 .12. .24*** .18** .06 .02 .29*** .27*** .33*** .63*** .70*** –

Size −.35*** .02 .62*** .40*** .43*** .24*** −.37*** .06 .40*** .26*** .36*** .22***

*** p < .001; ** p < .01; * p < .05; . p < .10.Note: Size was not included in structural pattern definition.

M.

Sapin et

al. /

Social N

etworks

47 (2016)

59–72

65

Table 3Clusters of structural interdependencies in family relationships according to the support and conflict network indices.

Network indices Bonding social capital(n = 62, 24.6%)

Bridging social capital(n = 76, 30.2%)

Overload(n = 70, 27.8%))

Ego-centred conflict(n = 44, 17.5%)

Total (N = 252) F

Mean (SD) Mean (SD) Mean (SD) Mean (SD) Mean (SD)

Support Density of supportrelationships

0.44 (0.23) 0.35 (0.17) 0.17 (0.10) 0.24 (0.12) 0.30 (0.19) 33.4***

Betweenness centrality ofrespondent in supportnetwork

5.32 (10.67) 40.78 (22.75) 14.45 (15.59) 16.56 (19.31) 20.51 (22.53) 51.3***

Number of alters to whomrespondent providedsupport

3.24 (2.97) 3.83 (2.41) 4.26 (3.04) 2.70 (2.09) 3.61 (2.73) 3.5*

Number of supportive tiesbetween alters to whomrespondent providedsupport

7.76 (12.32) 2.43 (3.15) 4.43 (6.81) 1.43 (2.91) 4.12 (7.71) 8.3***

Number of alters whoprovided support torespondents

2.47 (2.24) 3.47 (1.68) 1.96 (1.33) 2.86 (2.24) 2.70 (1.94) 8.6***

Number of supportive tiesbetween alters whoprovided support torespondent

5.16 (9.53) 1.87 (2.29) 1.10 (1.91) 1.34 (3.25) 2.37 (5.39) 8.2***

Conflict Density of conflictrelationships

0.23 (0.22) 0.16 (0.15) 0.16 (0.12) 0.21 (0.14) 0.18 (0.16) 3.1*

Betweenness centrality ofrespondent in conflictrelationships

3.22 (8.89) 2.26 (5.47) 7.49 (10.59) 26.32 (20.16) 8.15 (14.25) 47.9***

Number of alters botheredby respondent

1.29 (1.36) 1.07 (0.96) 2.16 (1.72) 3.52 (1.89) 1.85 (1.71) 30.0***

Number of conflict tiesbetween alters bothered byrespondent

0.65 (1.38) 0.30 (0.76) 1.86 (2.72) 1.05 (1.63) 0.95 (1.87) 10.1***

Number of alters whocaused conflict forrespondent

1.11 (1.10) 1.09 (0.98) 2.31 (1.89) 2.48 (1.15) 1.68 (1.48) 18.8***

Number of conflict tiesbetween alters who causedconflict for respondent

0.31 (0.90) 0.33 (0.85) 2.34 (3.71) 0.41 (0.87) 0.90 (2.27) 15.4***

Size 5.15 (3.02) 5.57 (2.56) 8.40 (4.02) 6.84 (2.75) 6.47 (3.42) 14.5***

*** p < .001; ** p < .01; * p < .05.

66 M. Sapin et al. / Social Netw

Fig. 1. Dendrogram of the hierarchical cluster analysis of structural patterns.

Fig. 2. Illustrations of bonding and b

orks 47 (2016) 59–72

connected to the respondents that provides them with collectivesupport was also large. The density of conflict relationships wasaverage, with as much as 23% of all pairs of network members con-nected with an upsetting relationship. However, the respondentsdid not occupy an intermediary position in conflicts. Fig. 2 illus-trates this pattern of family relationships. The arrows point towardssupport providers or family members who are sources of conflict.In Fig. 2a, the respondent has no intermediary position betweenfamily members who are part of a connected set of supportiverelationships. All of the family members are interconnected, andmost give and receive support from each other. A fair number ofthem are connected through conflict relationships. Cluster 2 (30.2%of the observations) features a bridging social capital pattern. It ischaracterised by a high betweenness centrality of respondents insupport relationships. Individuals in such a pattern supported a sig-nificant number of family members, who also provided them withsupport. There were, however, few supportive ties between alters.Regarding conflict, far fewer relationships were present. Fig. 2bdepicts such a pattern characterised by the central position of therespondent in support relationships. If one removes the respondentfrom his or her network, it splits into several components. Conflict
relationships are less dense.
Cluster 3 (27.8% of the observations) presents an overload sit-uation, because the respondent provides support to many familymembers in a context characterised by a large number of negative

ridging social capital patterns.

M. Sapin et al. / Social Networks 47 (2016) 59–72 67

d and

rtfoaosoc

odiopbotfelr

Fig. 3. Illustrations of overloa

elationships. The respondents embedded in this pattern of rela-ionships provided support to many alters and received supportrom only a minority of them. Therefore, there was an imbalancef support given and received by these respondents, which wasssociated with a large density of conflict ties. The first illustrationf Fig. 3a presents the overload pattern. The respondent providesupport to many family members; in return, she receives supportnly from her spouse and her daughters. There is a high density ofonflict relationships.

Cluster 4 (17.5% of the observations) describes a high centralityf respondents in family conflicts. It is also characterised by a lowensity of support relationships and a low centrality of respondents

n support relationships. The density of conflict is higher than in anyf the other patterns, and the respondents have a high intermediaryosition in negative relationships. A large number of family mem-ers irritate these respondents, and they upset far more membersf their family network. Overall, these respondents are at the cen-re of conflicts in their family network, in a relational context with
ew supportive relationships. Accordingly, this pattern was calledgo-centred conflict. Fig. 3b depicts such a pattern characterised by aow density of support and high centrality of respondents in conflictelationships.
ego-centred conflict patterns.

Overall, in support of Hypothesis H1, conflict and support rela-tionships did not exclude each other but were summed up bycluster analysis into qualitatively distinct patterns that mixed themin various ways. The first pattern clearly featured bonding socialcapital with mixed collective support (high density and low central-ity for support), with average density of conflict but low centralityof respondents in such conflicts. The second pattern featured bridg-ing social capital (low network density and high centrality ofrespondents for support, low density and low centrality in conflict).The third pattern featured an overload situation: the intermedi-ary position of respondents in supportive relationships (bridgingsocial capital) was related with a high degree of centrality in conflictrelationships and with a strong imbalance in the support receivedand given. The fourth pattern exhibited a high density of conflict inwhich the respondents occupied a central position, with a low den-sity of support and a low centrality of the respondents in supportiverelationships.

4.2. Individual distress

We assessed Hypothesis H2, about the relationship betweenpatterns of family relationships and their potential cumulative

68 M. Sapin et al. / Social Networks 47 (2016) 59–72

Table 4Multilevel mixed-effects linear regression models of structural patterns of family relationships on individual distress (SCL-90-R Positive Symptom Distress Index – log ofPSDI).

Univariate Multivariate

Coeff. (SE) Coeff. (SE)

Level 2Intercept .52*** .15Age (in years) .0004 .003 .002 .003

SexFemale Ref RefMale .04 .07 .05 .07

Education n.c.Primary education −.08 .08Secondary education RefTertiary education −.17. .09

Main diagnosisAnxiety or mood disorders −.01 .10 .06 .11Personality disorders Ref RefPsychotic disorders −.14. .08 −.08 .08

Institutionalised in the past n.c.Never RefPreviously .07 .06

Duration of the psychiatric treatment in the current private practice n.c.Less than one year .02 .10Between one or two years −.04 .08Between two and five years .11 .09Since more than five years Ref

Onset of the first psychiatric episode (in years) −.001 .004 n.c.

Beginning of the treatment in another practice or institution .03 .07 n.c.

Age of onset of mental disorder .003 .004 n.c.

Average network size through the follow-up .009 .012 n.c.

Cumulative effect of straining patterns through the follow-up .23** .08 .14 .09

Level 1Time

w1 (baseline) Ref Refw2 −.10*** .03 −.10** .03w3 −.11*** .03 −.10** .03w4 −.11*** .03 −.10** .03w5 −.08* .04 −.08** .03

ClustersBridging social capital pattern Ref RefOverload pattern .11** .03 .09* .04Ego-centred conflict pattern .13** .04 .11** .04Bonding social capital pattern .02 .03 .02 .03

Network size .007. .003 n.c.

Random effect�0 .21 .02�2 .15 .01

Deviance between empty model and multivariate model 37.01n1 248

*N

eaatrit(

itto

n2

** p < .001; ** p < .01; * p < .05; . p < .10.ote: Multilevel mixed-effects models on the logarithm of PSDI.

ffects in the longer term with individual distress level (PSDI), using series of multilevel mixed-effects regression models (Pinheirond Bates, 2000; Snijders and Bosker, 2011). Individuals’ charac-eristics were defined as level 2 (n2 = 60), and clusters of familyelationships as level 1 (n1 = 250). The intra-class correlation of thenitial null model on psychological distress (log of PSDI) showedhat the most important variance occurred at the individual levelICC = .67).

A first series of univariate analyses (see Table 4) assessed the
ndependent fixed effects of the patterns of family relationships,he basic background variables at baseline, the main diagnos-ic, and variables measuring mental health problem history. Inrder to estimate the effect of being embedded in enduring
59

straining patterns of family relationships, the number of times anindividual was embedded in either an overload or ego-centred con-flict pattern was computed and then divided by the number ofobservation points. The independent fixed effect of such a mea-sure of cumulation of straining patterns through time was assessedat the individual level. Finally, univariate analyses were also per-formed to check the independent fixed effect of network size, atthe observation level (the variable being time-variant), as well asaveraged at the contextual level of individuals.

The results of these series of univariate analyses showed thatthe basic background variables and the measures on mental healthproblem history were not significantly related to individual dis-tress levels. Similarly, the main diagnosis was not associated

Netw

wivaeilrlrpsotidnmssstivii

aucflioHttss

sratsuconaoaIgi

ftWi�itisS(


ith distress levels. The patterns of positive and negative familynterdependencies, in contrast, were significantly related to indi-idual distress (�2(3) = 16.13). Compared to the respondents in

bridging social capital pattern, individuals in the overload andgo-centred conflict patterns showed higher distress scores, whilendividuals in a bonding social capital pattern exhibited identicalevels of distress. Network size did not demonstrate a significantelationship at either the observation level or at the individualevel (respectively �2(1) = 1.47 and �2(1) = .57). To the contrary, theesult of univariate analysis of the cumulative effect of strainingatterns through time was statistically significant (�2(1) = 7.32),howing that individuals who were frequently embedded in anverload or an ego-centred conflict patterns of family relationshipshrough the follow-up experienced higher levels of distress. Wedentified the set of variables that best accounted for individualistress based on backward and forward stepwise selection tech-iques (Collett, 2003; Snijders and Bosker, 2011). The multivariateodel of Table 4 assessed the effects of patterns of family relation-

hips and the cumulative effects of being embedded in enduringtraining patterns through the follow-up on distress level, withex, age, and the main diagnosis as covariates. The difference inhe deviance of this model with the initial null model was signif-cant (�2(12) = 37.01), with a substantial increase of the explainedariance (R2 = 16.11%). Analyses conducted on the subsample ofndividuals who participated in the entire follow-up displayed fullydentical results.

The results of Table 4 confirmed Hypothesis H2a regarding thessociation between patterns of family relationships and individ-al distress level. Compared to the respondents in bridging socialapital patterns, individuals in the overload and ego-centred con-ict patterns showed higher scores of distress, while individuals

n a bonding social capital pattern exhibited comparable levelsf distress. The multivariate analyses did not, however, confirmypothesis H2b. The cumulative effect of being in straining pat-

erns through the follow-up became non-significant, showing thathe effect of the patterns was mainly synchronic. The analysis of theubsample of individuals who participated in the entire follow-uptudy displayed similar results.

The results of the multivariate multilevel mixed-effect regres-ion model in Table 4 also showed a significant effect of time. Theespondents reported higher levels of psychological distress (PSDI)t baseline compared to all subsequent waves of observations. Afterhis decrease in the intensity of distress between the first andecond waves, distress level remained stable. Another assessmentsing a multivariate, multilevel, mixed-effect regression model wasarried out on the data from Waves 2 to 5, while disregardingbservations of the first wave. Again, the results confirmed the sig-ificant relationships between the patterns of family relationshipsnd psychological distress. Interestingly, supplementary analysisf attrition (results not reported) showed that relational patternslso had an effect on the likelihood of dropping out of the study.ndividuals in an overload pattern at baseline had a significantlyreater chance of quitting the study before wave 3 than individualsn other patterns.

In addition, we performed a fixed-effect panel regression tourther check that these results were not biased by some omit-ed individual variables (Allison, 2009; Kohler and Kreuter, 2009).

hen assessed with network size, the results gave rise to sim-lar effects (fixed effect model �2(4) = 15.34; multilevel model2(4) = 16.14). Compared to individuals in a bridging social cap-

tal pattern, those in an overload pattern (b = .10*, SE = .04) andhose in an ego-centred conflict patterns (b = .12**, SE = .05) exhib-
ted higher levels of distress, while respondents in a bondingocial capital pattern reported comparable levels of distress (b = .03,E = .04). Network size again proved to have a non-significant effectb = .0007, SE = .0033).
orks 47 (2016) 59–72 69

In sum, supporting Hypothesis H2a, patterns of positive and neg-ative family relationships were significantly related to the levelsof distress experienced by individuals undergoing psychotherapy.Individuals embedded in the overload and ego-centred conflictpatterns of family relationships exhibited higher levels of distressthan individuals within bonding or bridging social capital patterns.Hypothesis H2b, on the cumulative effect of being embedded inenduring straining patterns through the follow-up on distress level,was not confirmed. The effects of the patterns on distress levelwere mainly synchronic, occurring at the time when informationon family networks was collected. Neither the basic demographicvariables nor the main diagnosis or mental health problems historymeasures were significantly associated with distress level.

5. Discussion

This study explored supportive and straining relationships inthe family networks of individuals with mental illness. In linewith previous literature underlying the multidimensionality of per-sonal and family networks (Wellman and Giulia, 1999; Wellmanand Potter, 1999; Widmer, 2010), and the fruitfulness of exploringhow different dimensions of personal networks might combine toimpact individuals’ lives and health (Pescosolido and Levy, 2002),we developed a typological approach to uncover the prevalent pat-terns of relationships in such networks. The results confirmed thatstructural dimensions of positive relationships (social capital) andnegative (conflict) relationships were not alternative interactionsin family networks, as they often appear together but mix in dif-ferent ways. The bonding social capital pattern featured a highlevel of interconnections in family networks in both support andconflict, whereas the overload pattern featured a high centralityof individuals with mental health problems, both in conflict andsupport relationships with family members. Alternatively, somepatterns showed different structural features in support and con-flict relationships. Individuals with mental illness embedded inthe ego-centred conflict pattern, as its name reveals, experiencedlarger numbers of conflict relationships and very few support-ive ties. Bridging social capital only depicted centrality in supportrelationships. Overall, conflict and support are loosely associateddimensions of family interdependencies that combine in diversebut identifiable ways, which need to be further research in largerrepresentative samples.

Patterns of positive and negative family relationships proved tobe related with psychological health during the course of mentalillness. The results were mainly synchronic, with current pat-terns correlating with current psychological adjustment. Distinctconfigurations of family ties can reach the same final state of psy-chological distress in bringing up strains. Similarly, other distinctconfigurations can provide social capital, which helps individ-uals with mental illness for coping with their illness trajectory(Carpentier, 2013). Individuals who are embedded in a patterncharacterised by bonding social capital, as with those in a bridgingsocial capital pattern, experienced less stressful interdependencies,allowing them to better adjust to mental illness. Bonding social cap-ital in a family provides a sense of being cared for, loved, valued,and accepted (Kadushin, 2002), although conflicts relationships arepresent. Despite such conflict, the feeling of security associatedwith bonding social capital decreases the perception of stressful-ness in challenging life situations (Pierce et al., 1997b) and mightbe a contributing factor to healing (Moos et al., 1998).

Individuals with bridging social capital featured an equal level
of distress to those with bonding social capital. With a weak pres-ence of conflict and a balance between the support received andgiven, the individuals with mental illness embedded in a bridgingsocial capital pattern did not experience higher levels of distress,

7 Netw

dpTfshoaw2

pamcotHpttmtItIarpsfpltbtaa

lotduthIa1Fps1

opamsabebpta


espite the cognitive resources needed to deal with their centralosition associated with bridging social capital (Cornwell, 2009).hey supported some family members, but also received supportrom them in a balanced way. Individuals with mental illness inuch configuration might experience a feeling of autonomy and ofaving some control of their environment (Kadushin, 2002). Previ-us work has underlined the benefits of balanced exchanges thatllow individuals with mental illness to keep their own autonomy,hich prove to be crucial in the recovery process (Davidson et al.,

009; DiMatteo and Hays, 1981; Rapp, 2006).The central position of individuals for support in the overload

attern was associated with a high level of conflict relationshipsnd an imbalance between support received and given. Therefore,ore support relationships did not necessarily translate into less

onflict in family networks (Sapin, 2013; Widmer, 2010). In theverload pattern, individuals had shares of bridging social capi-al because they connected otherwise unrelated family members.owever, despite the beneficial effects of associated with a centralosition, such as promoting individual autonomy and informa-ional resources (Burt, 1995, 2000), various reasons likely madehis situation too demanding for many individuals dealing with

ental illness. For example, they might lack the necessary cogni-ive resources to deal with this central position (Cornwell, 2009).n addition, the high centrality of individual with mental illness inheir support relationships was associated with family demands.ndividuals embedded in an overload pattern perceived themselvess giving family members far more emotional support than theyeceived. Their role as a support provider, thus, became demanding,otentially hindering their psychological health as they faced thetresses stemming from their mental disorders. Centrality in theamily network could then place great strain on individuals withoor psychological resources because dealing with various unre-

ated family members would be more costly and stressful for themhan for others, particularly when they are themselves supportedy few of them. This result is in line with earlier studies findinghat an overload of demands from family members (Veiel, 1993)nd family conflict (Moos et al., 1998) predicted recurrent episodesnd non-recovery.

The ego-centred conflict pattern was associated with a highevel of distress. In this pattern, negative relationships focussedn individuals with mental health problems, likely giving themhe feeling that their mental illness increased their family’s bur-en (Karp, 2001; Loukissa, 1995). This focus of family tensions mayndermine their self-view and others’ views of them, decreasingheir self-esteem and self-efficacy (Lakey et al., 1994), and mayave increased the perceived stressfulness of their life situations.

ndividuals with mental illness embedded in such family networksre likely to be the subject of interference (Johnson and Milardo,984) and the targets of criticism and hostility (Hooley, 2007).amily members’ expression of criticism towards a person indeedredicts recurrent episodes of and non-recovery from mood orchizophrenic disorders (Butzlaff and Hooley, 1998; Hooley et al.,986; Miklowitz, 2004).

This research is exploratory and based on a purposive sample ofutpatients from only one psychotherapy practice. The outpatientsresented various psychiatric disorders, and a majority of them had

long history of mental health problems. They were beyond theost acute distressful phases of their mental illness journey and, as

uggested by the results of their evolution through time and of thettrition. Some of them were still in a phase characterised by insta-ility and crises, while other moving towards stability (Carpentiert al., 1999; Muhlbauer, 2002). Following previous studies, it can
e hypothesised that the presence and evolution of the relationalatterns might be diverse over time, as some disorders give riseo more relational changes and different phases of a mental illnessre associated with specific needs and distinct formal treatment
orks 47 (2016) 59–72

arrangements (Perry and Pescosolido, 2012). This research focuseson the relational patterns in family networks of individuals withmental illness, in relation to their experienced distress, during alimited time period of their illness trajectory in which they weremoving towards stability, without being able to take into accountthe various sociological determinants of mental illness or the socialdimensions of the recovery process (Davidson et al., 2009; Pilgrimand McCranie, 2013; Rapp, 2006).

This study yields insights into how family support and conflictin family networks may be connected with the psychological healthof adults with mental illness. Evidence was provided that individ-uals with mental illness develop distinct relational patterns thatare associated with psychological health. Only a minority of indi-viduals experience social capital without family conflict. Variouspatterns mixing negative and positive family relationships exist,which are more than the sum of the support and conflict ties. Thewhole patterns were related with distinct levels of psychologicaldistress. In order to better understand the interrelation betweenfamily contexts and the psychological health of individuals withmental illness, the investigation of positive relationships in familynetworks should be therefore complemented by the inclusion ofnegative relationships. Beyond its scope in mental illness research,this study aims to identifying fundamental ways in which familyand personal networks vary in their structural organisation andhow such a structural organisation impact on individuals’ lives.It also stresses the importance of conflict as “building block” ofpersonal communities (Wellman and Potter, 1999).

Acknowledgements

We express our gratitude towards Dr Patricia Dumas and Mar-ianne Chevalier for the fruitful collaboration. We also thank Prof.André Berchtold for his statistical advice on this paper, as well as theanonymous reviewers who helped us to improve the manuscript.

This publication is part of the research works conducted at theSwiss National Centre of Competence in Research “LIVES – Over-coming Vulnerability: Life Course Perspectives”, which is financedby the Swiss National Science Foundation.

References

Agneessens, F., 2013, May. Is social capital more than the sum of its parts? Tax-onomies of ego-network position of employees as a predictor of individualperformance. In: Paper presented at the SUNBELT XXXIII Conference of theInternational Network for Social Network Analysis, Hamburg Deutschland, May21–26, 2013.

Agneessens, F., Waege, H., Lievens, J., 2006. Diversity in social support by role rela-tions: a typology. Social Networks 28, 427–441.

Allison, P.D., 2009. Fixed effects regression models. In: Quantitative Applications inthe Social Sciences. Sage Publications, Thousand Oaks, CA.

Antonucci, T.C., Akiyama, H., 1995. Convoys of social relations: family and friend-ships within a life span context. In: Blieszner, R., Bedford, V.H. (Eds.), Handbookof Aging and the Family. Greenwood Press, Westport, CT, pp. 355–371.

August, K.J., Rook, K.S., Newsom, J.T., 2007. The joint effects of life stress and negativesocial exchange on emotional distress. J. Gerontol. 62B (5), S304–S331.

Bergman, L.R., Magnusson, D., 1997. A person-oriented approach in research ondevelopmental psychopathology. Dev. Psychopathol. 9 (2), 291–319.

Bergman, L.R., Eklund, G., Magnusson, D., 1991. Studying individual development:problems and methods. In: Magnusson, D., Bergman, L.R., Rudinger, G., Törestad,B. (Eds.), Problems and Methods in Longitudinal Research. Cambridge UniversityPress, Cambridge, UK, pp. 1–27.

Berkman, L.F., Glass, T., Brisette, I., Seeman, T.E., 2000. From social integration tohealth: Durkheim in the new millennium. Soc. Sci. Med. 51, 843–857.

Bernard, H.R., 2012. Social Research Methods: Qualitative and QuantitativeApproaches, 2nd ed. Sage, Thousand Oaks, CA.

Bertera, E.M., 2005. Mental health in U.S. adults: the role of positive social supportand social negativity in personal relationships. J. Soc. Pers. Relat. 22 (1), 33–48.

Borgatti, S., Everett, M.G., Freeman, L.C., 2002. UCINET 6 for Windows [Software].Analytic Technologies, Lexington, KY.

Bourdieu, P., 1985. The form of capital. In: Richardson, J.C. (Ed.), Handbook of The-ory and Research for the Sociology of Education. Greenwood, New York, pp.241–258.

http://refhub.elsevier.com/S0378-8733(16)30138-1/sbref0655













































































































































































































































































Netw

B

B

B

B

C

C

C

C

C

C

C

CC

C

C

CC

D

D

D

D

D

E

E

E

F

F

F

F

H

H

H

H

H


ryk, A.S., Raudenbush, S.W., 1992. Hierarchical Linear Models: Applications andData Analysis Methods. Sage, Newbury Park, CA.

urt, R., 1995. Structural holes. In: The Social Structure of Competition. HarvardUniversity Press, Cambridge, MA.

urt, R.S., 2000. The network structure of social capital. In: Sutton, R.I., Staw, B.M.(Eds.), Research in Organisational Behavior, vol. 22. JAI Press, Greenwich, CT.

utzlaff, R.L., Hooley, J.M., 1998. Expressed emotion and psychiatric relapse: A meta-analysis. Arch. Gen. Psychiatry 55 (6), 547–552.

arpentier, N., 2013. Entry into a care trajectory: individualization process,networks, and emerging project. Sage Open, April–June 2013, 1–9, http://dx.doi.org/10.1177/2158244013494215.

arpentier, N., Lesage, A., White, D., 1999. Family influence on the first stages of thetrajectory of patients diagnosed with severe psychiatric disorders. Fam. Relat.48, 397–403.

arpentier, N., Ducharme, F., 2003. Care-giver network transformation: the need foran integrated perspective. Aging Soc. 23, 507–525.

larke, P., Crawford, C., Steele, F., Vignoles, A.F., 2010. The choice between fixedand random effects models: some considerations for educational research. In:Discussion Paper No. 5287. IZA.

ohen, S., Wills, T.A., 1985. Stress, social support, and the buffering hypothesis.Psychol. Bull. 98 (2), 310–357.

ohen, S., Underwood, L.G., Gottlieb, B.H., 2000. Social relationship and health. In:Cohen, S., Underwood, L.G., Gottlieb, B.H. (Eds.), Social Support Measurementand Intervention. A Guide for Health and Social Scientists. Oxford UniversityPress, Oxford, pp. 3–25.

oleman, J., 1988. Social capital and the creation of human capital. Am. J. Sociol. 94,95–121.

ollett, D., 2003. Modelling Binary Data. Chapman & Hall/CRC, Boca Raton.onnidis, I.A., McMullin, J.A., 2002. Sociological ambivalence and families ties: a

critical perspective. J. Marriage Fam. 64 (3), 558–567.ornwell, B., 2009. Good health and the bridging of structural holes. Soc. Netw. 31

(1), 92–103.orrigan, P.W., Phelan, S.M., 2004. Social support and recovery in people with serious

mental illness. Community Ment. Health J. 40 (6), 513–523.oser, L.A., 1956. The Functions of Social Conflict. The Free Press, New York.oyne, J.C., DeLongis, A., 1986. Going beyond social support: the role of social rela-

tionships in adaptation. J. Consult. Clin. Psychol. 54 (4), 454–460.avidson, L., Tondora, J., Staeheli Lawless, M., O’Connell, M.J., Rowe, M., 2009. A

practical guide to recovery-oriented practice. In: Tools for Transforming MentalHealth Care. Oxford University Press, New York.

elery, J.E., Doty, D.H., 1996. Modes of theorizing in strategic human resource man-agement: tests of universalistic, contingency, and configurational performancepredictions. Acad. Manage. J. 39 (4), 802–835.

eLongis, A., Capreol, M., Hotlzman, S., O’Brien, T., Campbell, J., 2004. Social sup-port and social strain among husbands and wives: a multilevel analysis. J. Fam.Psychol. 18 (3), 470–479.

erogatis, L.R., 1983. Administration, Scoring and Procedures, 2nd ed. Clinical Psy-chometric Research, Baltimore, MD.

iMatteo, M.R., Hays, R., 1981. Social support and serious illness. In: Gottlieb,B.H. (Ed.), Social Networks and Social Support. Sage, Beverly Hills, London, pp.117–148.

verett, M.G., Borgartti, S.P., 2014. Networks containing negative ties. Soc. Netw. 38,111–120.

veritt, B.S., Landau, S., Leese, M., Stahl, D., 2011. Cluster Analysis, 5th ed. John Wiley& Sons, Chichester, UK.

vert, H., Harvey, C., Trauer, T., Herrman, H., 2003. The relationship between socialnetworks and occupational and self-care functioning in people with psychosis.Soc. Psychiatry Psychiatr. Epidemiol. 38, 180–188.

aber, A.D., Wasserman, S., 2002. Social support and social networks: synthesisand review. In: Levy, J.A., Pescosolido, B.A. (Eds.), Social Networks and Health(Advances in Medical Sociology, vol. 8). Emerald Group Publishing Limited, Bin-gley, pp. 29–72.

ehr, B., Perlman, D., 1985. The family as a social network and support system. In:L’Abate, L. (Ed.), The Handbook of Family Psychology and Therapy. The DorseyPress, Homewood, IL, pp. 323–356.

ingerman, K.L., Hay, E.L., Birditt, K.S., 2004. The best of ties, the worst of ties:close, problematic, and ambivalent social relationships. J. Marriage Fam. 66 (3),792–808.

urstenberg, F.F., Kaplan, S.H., 2004. Social capital and the family. In: Scott, J., Treas,J., Richards, M. (Eds.), The Blackwell Companion to the Sociology of Families.Blackwell Publishing, Malden, MA, pp. 218–232.

aines, V.A., Hurlbert, J.S., Beggs, J.T., 1996. Exploring the determinants of supportprovision: provider characteristics, personal networks, community contexts,and support following life events. J. Health Soc. Behav. 37 (3), 252–264.

ennig, C., Liao, T., 2013. How to find an appropriate clustering for mixed-type vari-ables with application to socio-economic stratification. J. R. Stat. Soc. 62 (3),309–369.

ooley, J.M., 2007. Expressed emotion and relapse of psychopathology. Annu. Rev.Clin. Psychol. 3, 329–352.

ooley, J.M., Hiller, J.B., 1997. Family relationships and major mental disorder: riskfactors and preventive strategies. In: Duck, S. (Ed.), Handbook of Personal Rela-
tionships. Theory, Research, Interventions. John Wiley & Sons, Chichester, UK,pp. 621–628.
ooley, J.M., Orley, J., Teasdale, J.D., 1986. Levels of expressed emotion and relapsein depressed patients. Br. J. Psychiatry 148, 642–647.

orks 47 (2016) 59–72 71

House, J.S., Umberson, D., Landis, K.R., 1988. Structures and processes of social sup-port. Annu. Rev. Sociol. 14, 293–318.

Hox, J., 2002. Multilevel Analysis: Techniques and Application. Lawrence ErlbaumAssociates, Mahwah, NJ.

Johnson, M.P., Milardo, R.M., 1984. Network interference in pair relationships: asocial psychological recasting of Slater’s theory of social regression. J. MarriageFam. 46 (4), 893–899.

Kadushin, C., 2002. The motivational foundation of social networks. Soc. Netw. 24(1), 77–91.

Karp, D., 2001. The Burden of Sympathy: How Families Cope with Mental Illness.Oxford University Press, New York.

Kaufman, L., Rousseeuw, P.J., 1990. Finding Groups in Data. An Introduction to ClusterAnalysis. Wiley, New York.

Kawachi, I., Berkman, L.F., 2001. Social ties and mental health. J. Urban Health: Bull.N.Y. Acad. Med. 78 (3), 458–467.

Kearns, J.N., Leonard, K.E., 2004. Social networks, structural interdependence, andmarital quality over the transition to marriage: a prospective analysis. J. Fam.Psychol. 18 (2), 383–395.

Kohler, U., Kreuter, F., 2009. Data Analysis using Stata, 2nd ed. Stata CorporationPress.

Krackhardt, D., 1987. Cognitive social structures. Soc. Netw. 9, 109–134.Labianca, G., Brass, D.J., Gray, B., 1998. Social networks and perceptions of intergroup

conflicts: the role of negative relationships and third parties. Acad. Manage. J.41 (1), 55–67.

Lakey, B., Tardiff, T.A., Drew, J.B., 1994. Negative social interactions: assessment andrelations to social support, cognition, and psychological distress. J. Soc. Clin.Psychol. 13 (1), 42–62.

Lebart, L., Morineau, A., Piron, M., 1997. Statistique exploratoire multidimension-nelle. Dunod, Paris.

Lin, N., Ye, X., Ensel, W.M., 1999. Social support and depressed mood: a structuralanalysis. J. Health Soc. Behav. 40, 344–359.

Link, B.G., Cullen, F.T., Struening, E., Shrout, P.E., Dohrenwend, B.P., 1989. A modifiedlabeling theory approach to mental disorders: an empirical assessment. Am.Sociol. Rev. 54 (3), 400–423.

Link, B.G., Phelan, J., 2001. Conceptualizing stigma. Annu. Rev. Sociol. 27, 363–385.Loukissa, D.A., 1995. Family burden in chronic mental illness: a review of research

study. J. Adv. Nurs. 21, 248–255.Major, B., Zubek, J.M., Cooper, M.L., Cozzarelli, C., Richards, C., 1997. Mixed

messages: implications of social conflict and social support within close rela-tionships for adjustment to a stressful life event. J. Pers. Soc. Psychol. 72 (6),1349–1363.

Miklowitz, D.J., 2004. The role of family systems in severe and recurrent psychi-atric disorders: a developmental psychopathology view. Dev. Psychopathol. 16,667–688.

Moos, R.H., Cronkite, R.C., Moos, B., 1998. Family and extrafamily resourcesand the 10-year course of treated depression. J. Abnorm. Psychol. 107 (3),450–460.

Moren-Cross, J.L., Lin, N., 2006. Social network and health. In: Binstock, R.H., George,L.K., Cutler, S.J., Hendricks, J., Schulz, J.H. (Eds.), Handbook of Aging and the SocialSciences. , 6th ed. Elsevier Academic Press, San Diego, CA, pp. 111–127.

Muhlbauer, S.A., 2002. Navigating the storm of mental illness: phases in the family’sjourney. Qual. Health Res. 12, 1076–1092.

Newsom, J.T., Nishishiba, M., Morgan, D.L., Rook, K.S., 2003. The relative importanceof three domains of positive and negative social exchanges: a longitudinal modelwith comparable measures. Psychol. Aging 18 (4), 746–754.

Newsom, J.T., Mahan, T.L., Rook, K.S., Krause, N., 2008. Stable negative socialexchanges and health. Health Psychol. 27 (1), 78–86.

Newsom, J.T., Rook, K.S., Nishishiba, M., Sorkin, D.H., Mahan, T.L., 2005. Under-standing the relative importance of positive and negative exchanges: examiningspecific domains and appraisals. J. Gerontol. Ser. B 60 (6), P304–P312.

Okun, M.A., Keith, V.M., 1998. Effects of positive and negative social exchanges withvarious sources on depressive symptoms in younger and older adults. J. Gerontol.53B, 4–20.

Pagel, M.D., Erdly, W.W., Becker, J., 1987. Social networks: we get by with (and inspite of) a little help from our friends. J. Pers. Soc. Psychol. 53 (4), 793–804.

Pellet, J., 1997. La Symptom Check-List SCL-90R. In: Guelfy, J.D. (Ed.), L’évaluationclinique standardisée en psychiatrie, 1. Editions médicales Pierre Fabre,Boulogne, pp. 77–85.

Perry, B.L., Pescosolido, B.A., 2012. Social network dynamics and biographical dis-ruption: the case of “first-timers”. Am. J. Sociol. 118 (1), 134–175.

Pescosolido, B.A., Levy, J.A., 2002. The role of social networks in health, illness, diseaseand healing. In: Levy, J.A., Pescosolido, B.A. (Eds.), Social Networks and Health(Advances in Medical Sociology, vol. 8). Emerald Group Publishing Limited, Bin-gley, pp. 3–25.

Pierce, G.R., Lakey, B., Sarason, I.G., Sarason, B.R., 1997a. Sourcebook of Social Supportand Personality. Plenum Press, New York.

Pierce, G.R., Sarason, B.R., Sarason, I.G., Henderson, C.A., 1996. Conceptualizing andassessing social support in the context of the family. In: Pierce, G.R., Sarason,B.R., Sarason, I.G. (Eds.), Handbook of Social Support and the Family. PlenumPress, New York, pp. 3–23.

Pierce, T., Baldwin, M.W., Lydon, J.E., 1997b. A relational schema approach to social
support. In: Pierce, G.R., Lakey, B., Sarason, I.G., Sarason, B.R. (Eds.), Sourcebookof Social Support and Personality. Plenum, New York, pp. 19–47.
Pilgrim, D., McCranie, A., 2013. Recovery and Mental Health: A Critical SociologicalAccount. Palgrave Macmillan, New York.
































































dx.doi.org/10.1177/2158244013494215

dx.doi.org/10.1177/2158244013494215

dx.doi.org/10.1177/2158244013494215

dx.doi.org/10.1177/2158244013494215

dx.doi.org/10.1177/2158244013494215

dx.doi.org/10.1177/2158244013494215

dx.doi.org/10.1177/2158244013494215
















































































































































































































































































































































































































































































































































































































































































































































































































































































































































































































































































































































































































































































7 Netw

P

R

R

R

R

R

R

S

S

S

S

S

S

S

S

SS

S

T

T


inheiro, J.C., Bates, D.M., 2000. Mixed Effects Models in S and S-Plus. Springer, NewYork.

Core Team, 2011. R: A Language and Environment for Statistical Computing.R Foundation for Statistical Computing, Vienna, Austria, ISBN 3-900051-07-0,http://www.R-project.org/.

app, C., 2006. The Strengths Model: Case Management with People with PsychiatricDisabilities. Oxford University Press, New York.

apkin, B.D., Luke, D.A., 1993. Cluster analysis in community research: epistemologyand practice. Am. J. Community Psychol. 21 (2), 247–277.

ogers, A., Hassell, K., Nicolaas, G., 1999. Demanding Patients? Analysing the Use ofPrimary Care. Open University Press, Philadelphia.

ook, K., 1984. The negative side of social interaction: impact on psychological well-being. J. Pers. Soc. Psychol. 46 (5), 1097–1108.

ook, K.S., 1998. Investigating the positive and negative sides of personal relation-ships: through a lens darkly? In: Spitzberg, B.H., Cupach, W.R. (Eds.), The DarkSide of Close Relationships. Erlbaum, Mahwah, NJ, pp. 369–393.

apin, M., Unpublished doctoral thesis 2013. Configurations de proches et adaptationdans le parcours de vie d’individus en situation de vulnérabilité. University ofLausanne, Faculty of social and political sciences, pp. 316.

chuster, T.L., Kessler, R.C., Aseltine, R.H., 1990. Supportive interactions, negativeinteractions, and depressive mood. Am. J. Community Psychol. 18 (3), 423–438.

illars, A., Canary, D.J., Tafoya, M., 2004. Communication, conflict, and the quality offamily relationships. In: Vangelisti, A.L. (Ed.), Handbook of Family Communica-tion. Erlbaum, Mahwah, NJ, pp. 413–446.

immel, G., 1999. Sociologie. Etudes sur les formes de la socialisation. Presse Uni-versitaires de France, Paris.

nijders, T.A.B., Bosker, R.J., 2011. Multilevel Analysis: An Introduction to Basic andAdvanced Multilevel Modeling, 2nd ed. Sage, London.

nijders, T., Berkhof, J., 2004. Diagnostic checks for multilevel models. In: de Leeuw,J., Kreft, I. (Eds.), Handbook of Quantitative Multilevel Analysis. Sage, ThousandOaks, CA, pp. 141–175.

ong, L., Son, J., Lin, N., 2011. Social support. In: Scott, J., Carrington, P. (Eds.), SageHandbook of Social Network Analysis. Sage, London, pp. 116–128.

prey, J., 1999. Family dynamics: an essay on conflict and power. In: Sussman, M.B.,Steinmetz, S.K. (Eds.), Handbook of Marriage and the Family. , 2nd ed. PlenumPress, New York, pp. 667–685.

prey, J., 1969. The family as a system in conflict. J. Marriage Fam. 31 (4), 699–706.tataCorp, 2011. Stata Statistical Software: Release 12. StataCorp LP, College Station,

TX.zreter, S., Woolcock, M., 2004. Health by association? Social capital, social theory,

and the political economy of public health. Int. J. Epidemiol. 33 (4), 650–667.abachnick, B., Fidell, L., 1996. Using Multivariate Statistics. Harper-Collins, New

York.hoits, P.A., 1995. Coping, social support processes: where are we? What next? J.

Health Soc. Behav. 35, 53–79.

orks 47 (2016) 59–72

Tolsdorf, C.C., 1976. Social networks, support and coping: an exploratory study. Fam.Process 15, 407–417.

Uchino, B.N., Holt-Lunstad, J., Smith, T.W., Bloor, L., 2004. Heterogeneity in socialnetworks: a comparison of different models linking relationships to psycholog-ical outcomes. J. Soc. Clin. Psychol. 23 (2), 123–139.

Unger, D.G., Powell, D.R., 1980. Supporting families under stress: the role of socialnetworks. Fam. Relat. 29 (4), 566–574.

Veiel, H.O., 1993. Detrimental effects of kin support networks on the course ofdepression. J. Abnorm. Psychol. 102 (3), 419–429.

Walker, M.E., Wasserman, S., Wellman, B., 1993. Statistical models for social supportnetworks. Sociol. Method Res. 22, 71–98.

Wasserman, S., Faust, K., 1994. Social Network Analysis: Methods and Applications.Cambridge University Press, Cambridge, UK.

Wellman, B., 2000. Partner in illness: who helps when you are sick. In: Kelner, M.,Wellman, B., Saks, M., Pescosolido, B. (Eds.), Complementary and AlternativeMedicine: Challenge and Changes. Overseas Publishers Association, Amsterdam,pp. 143–161.

Wellman, B., Giulia, M., 1999. The network basis of social support: a network is morethan the sum of its ties. In: Wellman, B. (Ed.), Networks in the Global Village.Westview Press, Boulder [etc.], pp. 83–118.

Wellman, B., Potter, S., 1999. The elements of personal communities. In: Wellman, B.(Ed.), Networks in the Global Village. Westview Press, Boulder [etc.], pp. 49–118.

Wellman, B., Worthley, S., 1989. Brothers’ keepers: situating kinship relations inbroader networks of social support. Sociol. Perspect. 32 (3), 273–306.

Wellman, B., Worthley, S., 1990. Different strokes from different folks: communityties and social support. Am. J. Sociol. 96 (3), 558–588.

Whitley, R., McKenzie, M., 2005. Social capital and psychiatry: review of the litera-ture. Harv. Rev. Psychiatry 13 (2), 71–84.

World Health Organization, 1993. The ICD-10 Classification of Mental and BehavioralDisorders. Diagnostic Criteria for Research. World Health Organization, Geneva.

Widmer, E.D., 1999. Family contexts as cognitive networks: a structural approachof family relationships. Pers. Relat. 6, 487–503.

Widmer, E.D., 2010. Family Configurations. A Structural Approach to Family Diver-sity. Ashgate Publishing, London.

Widmer, E.D., Chevalier, M., Dumas, M., 2005. Le “Family Network Method”(FNM): Un outil d’investigation des configurations familiales à disposition desthérapeutes. Thérapie Familiale 26 (4), 427–445.

Widmer, E.D., Jallinoja, R. (Eds.), 2008. Beyond the Nuclear Family. Families in aConfigurational Perspective. Peter Lang, Bern.

Widmer, E.D., La Farga, L.-A., 2000. Family networks: a sociometric method to study
relationships in family. Field Methods 12 (2), 108–128.
Widmer, E.D., Sapin, M., 2008. Families on the move. Insights about how familyconfigurations of individuals change. In: Widmer, E.D., Jallinoja, R. (Eds.), Beyondthe Nuclear Family. Families in a Configurational Perspective. Peter Lang, Bern,pp. 279–302.











http://www.r-project.org/








































































































































































































































































































































































































































































































































































































































































































































from support to overload: patterns of positive and ... · jointly, in the family networks of adults...

Documents