Personal support networks, social capital, and risk of relapse among individuals treated for substance use issues

G Model

DRUPOL-1642; No. of Pages 8 International Journal of Drug Policy xxx (2015) xxx–xxx

Contents lists available at ScienceDirect

International Journal of Drug Policy journal homepage: www.elsevier.com/locate/drugpo

Research paper

Personal support networks, social capital, and risk of relapse among individuals treated for substance use issues Daria Panebianco a, Owen Gallupe b,*, Peter J. Carrington b, Ivo Colozzi c a

National Addiction Centre, Addictions Department, Institute of Psychiatry, Psychology and Neuroscience, King’s College London, 4 Windsor Walk, Denmark Hill, London, England SE5 8BB, United Kingdom b Department of Sociology and Legal Studies, University of Waterloo, 200 University Avenue West, Waterloo, Ontario, Canada N2L 3G1 c Department of Sociology and Business Law, Alma Mater Studiorum, University of Bologna, Strada Maggiore 45, 40125 Bologna, Italy

A R T I C L E I N F O

A B S T R A C T

Article history: Received 17 July 2014 Received in revised form 8 September 2015 Accepted 19 September 2015

Background: The success of treatment for substance use issues varies with personal and social factors, including the composition and structure of the individual’s personal support network. This paper describes the personal support networks and social capital of a sample of Italian adults after long-term residential therapeutic treatment for substance use issues, and analyses network correlates of posttreatment substance use (relapse). Methods: Using a social network analysis approach, data were obtained from structured interviews (90– 120 min long) with 80 former clients of a large non-governmental therapeutic treatment agency in Italy providing voluntary residential treatments and rehabilitation services for substance use issues. Participants had concluded the program at least six months prior. Data were collected on sociodemographic variables, addiction history, current drug use status (drug-free or relapsed), and the composition and structure of personal support networks. Factors related to risk of relapse were assessed using bivariate and multivariate logistic regression models. Results: A main goal of this study was to identify differences between the support network profiles of drug free and relapsed participants. Drug free participants had larger, less dense, more heterogeneous and reciprocal support networks, and more brokerage social capital than relapsed participants. Additionally, a lower risk of relapse was associated with higher socio-economic status, being married/ cohabiting, and having network members with higher socio-economic status, who have greater occupational heterogeneity, and reciprocate support. Conclusions: Post-treatment relapse was found to be negatively associated with the socioeconomic status and occupational heterogeneity of ego’s support network, reciprocity in the ties between ego and network members, and a support network in which the members are relatively loosely connected with one another (i.e., ego possesses ‘‘brokerage social capital’’). These findings suggest the incorporation into therapeutic programming of interventions that address those aspects of clients’ personal support networks. ß 2015 Elsevier B.V. All rights reserved.

Keywords: Substance abuse Social network Social support Social capital Relapse

Introduction Individual responses to treatment for substance use issues are highly variable. Multiple relapse episodes are common among treated substance users and are a normal part of rehabilitation (e.g., Kirshenbaum, Olsen, & Bickel, 2009; Polivy & Herman, 2002; Witkiewitz & Marlatt, 2004). However, explanations of relapse have historically emphasized individual-level factors (e.g., Dawson, Goldstein, & Grant, 2007; Koob & Le Moal, 2008; Sinha,

* Corresponding author. Tel.: +1 519 888 4567x33361; fax: +1 519 746 7326. E-mail addresses: [email protected] (D. Panebianco), [email protected] (O. Gallupe), [email protected] (P.J. Carrington), [email protected] (I. Colozzi).

2007), although there has been greater attention to the social nature of addiction since the mid-1980s. Recognizing that an exclusive focus on individual attributes has limited explanatory power to help understand substance use issues, numerous theoretical perspectives (e.g., social learning – Bandura, 1977; Sutherland & Cressey, 1978; social capital – Bourdieu, 1986; Coleman, 1988; social support – Cohen & Wills, 1985) and evidence-based treatments (e.g., the Community Reinforcement Approach –Meyers & Miller, 2001; Community Reinforcement and Family Training – Meyers, Miller, Hill, & Tonigan, 1998; Network Therapy – Copello, Williamson, Orford, & Day, 2006; Galanter, 1993) incorporate members of the individual’s social environment. In this study, we explore differences in both personal characteristics and the social environments of substance abuse treatment

http://dx.doi.org/10.1016/j.drugpo.2015.09.009 0955-3959/ß 2015 Elsevier B.V. All rights reserved.

Please cite this article in press as: Panebianco, D., et al. Personal support networks, social capital, and risk of relapse among individuals treated for substance use issues. International Journal of Drug Policy (2015), http://dx.doi.org/10.1016/j.drugpo.2015.09.009

G Model

DRUPOL-1642; No. of Pages 8 2

D. Panebianco et al. / International Journal of Drug Policy xxx (2015) xxx–xxx

recipients who had relapsed versus those who remained abstinent. In doing so, we draw on social support and social capital concepts using network analytic techniques. Social network approaches to the study of addiction and substance use have provided important insights into the social pressures that shape the behaviors of users. Social network analysis focuses on the individual’s social environment, the attributes of network members (the ‘‘composition’’ of the network), and the content and structure of relationships. Relationships are seen as channels through which people receive information (Burt, 2005), and the network structural profile plays a role in determining the amount and form of social support available to members of the network (Chua, Madej, & Wellman, 2011). Prior research has focused on the network structure of drug users, types of alters in their support network, and social support mobilized. Some of these studies have found that more dense (i.e., closed) networks that are composed of a larger number of drugusing members facilitate access to substances and are therefore risk factors for relapse (e.g., Harocopos, Lloyd, Kobrak, Jost, & Clatts, 2009; Koram et al., 2011; Latkin, Mandell, Oziemkowska, & Celentano, 1995; Rhoades et al., 2011; Rice, Milburn, & Monro, 2011). However, having family members and alters with higher educational backgrounds within the network appears to provide support for abstinence and helps protect against drug use (e.g., El-Bassel, Cooper, & Chen, 1998; Tyler, 2008). While social support is generally regarded as a positive characteristic of non-drug-using social networks (e.g., Debnam, Holt, Clark, Roth, & Southward, 2012; Kahn, Hessling, & Russell, 2003), it is often more complex and multidimensional as it relates to individuals with addiction issues (see Brown & Riley, 2005; Ellis, Bernichon, Yu, Roberts, & Herrell, 2004). Social supports can facilitate drug use by providing substance purchasing advice, money to buy drugs, and/or an appropriate place to use. However, supports can also promote abstinence-related selfefficacy (Stevens, Jason, Ram, & Light, 2014) and may help former users to maintain their drug free status and facilitate a return to the community by providing reassurance and opportunities (e.g., El-Bassel et al., 1998). Complicating matters is the risk that possessing strongly supportive ties could inhibit or delay the development of independent coping strategies. Reciprocity (the exchange of affect and/or resources in support relationships) may be effective in offsetting the risk of dependence. Some scholars have shown that a major factor in recovery from alcoholism is the ability to both receive and provide support (Gordon & Zrull, 1991). Relationships characterized by supportive equity might enhance feelings of self-efficacy, self-esteem, and sense of balance within the relationship, as well as decrease feelings of dependence on care providers for help (Gleason, Iida, Bolger, & Shrout, 2003). Therefore, being part of a network within which individuals are considered by alters to have the ability to provide support as well as receive it is likely to protect against relapse among treated users. The social capital perspective helps to explain the role of social support networks in addiction recovery (including abstinence maintenance). Since the 1970s, the social capital framework has received extensive research attention in a variety of disciplines. The basic premise of social capital is that investment in social relations provides instrumental and emotional returns, such as access to information, enhanced personal credentials and recognition, reinforcement of identity, and social support (Coleman, 1988; Lin & Erickson, 2008).1 Thus, social support can be seen as one type of (potential) return on an individual’s social capital (Song, 1 Social capital has been operationalised in many different ways in substance use studies. E.g., social participation and trust (Lundborg, 2005); collective norms (Kirst, 2009); support from non-drug-using friends (Cheung, 2009); school and family supports (Dufur et al., 2007).

Son, & Lin, 2011). Coleman (1988, 1990) proposed the concept of network closure to explain the mechanism through which social capital operates. Closed networks are insulated against outside forces by limiting membership to select individuals. In these tightly knit networks, there tends to be high levels of trust and people are expected to help each other. As it relates to the substance use literature, network closure has been found to be important in a number of ways including recognising the need for, and actually receiving, treatment (Tucker, Wenzel, Golinelli, Zhou, & Green, 2011), frequency of substance use (Wenzel et al., 2009), and odds of a non-fatal overdose (Tobin, Hua, Costenbader, & Latkin, 2007). In the context of individuals exiting treatment for substance use issues, integrating the former user into relationships based on trust and mutual reliance with nondrug-using others might produce social capital that helps to support their abstinence (Cheung & Cheung, 2003; Cheung, 2009). Social network theory also suggests that open networks of weak ties provide a different form of social capital. Being loosely connected to individuals who operate mainly in different groups exposes a person to new information, ideas, influences, and the ‘‘knowledge of the world beyond his own friendship circle’’ (Granovetter, 1973, p. 1371). Burt (1992, 2005) elaborated on these social capital benefits by arguing that both closed and open networks can have advantages. According to Burt, the close reciprocal connections that come with network closure provide different types of benefits than the informational benefits of connecting to various individuals and groups that are otherwise unconnected: a phenomenon that Burt called brokerage because ego is in a position to act as a broker between alters. That is, in networks with ‘‘structural holes’’ (gaps) between groups, occupying a brokerage role that connects those groups increases the likelihood of a person accessing new and rewarding opportunities (e.g., facilitating access to jobs or non substance use leisure activities) that they would otherwise not be able to take advantage of (Burt, 2005). Research on the relevance of brokerage to substance use careers has reported conflicting results. Jonas, Young, Oser, Leukefeld, and Havens (2012) found that those with higher brokerage scores tended to have a lower likelihood of daily cannabis use but a higher likelihood of daily OxyContin1 use. Lorant and Nicaise (2015) found higher levels of brokerage to be related to less frequent binge drinking among Belgian university students. While previous research has taken a variety of approaches to investigate the role that social networks play in structuring patterns of substance use, they generally take a narrow approach and do not account for a wider array of important factors. We attempt to address this by examining the relationship between abstinence maintenance and social network indicators of social capital and support that are potentially important but have not been extensively researched (including network structure, heterogeneity, reciprocity, and constraint) as well as accounting for important individual and relationship level characteristics. Specifically, this article describes the personal support networks of 80 Italian adults who received long-term, voluntary residential treatment for substance use issues. Personal support networks are composed of a single focal individual (‘‘ego’’), his/her support relationships with other individuals (‘‘alters’’), and relationships among the alters. This type of network in which one person is the main focus and the network is considered to be those connected to the person of interest (while accounting for the pattern of connections among the others) is often referred to as an ego-centred network. The main purpose of this study is to characterise and compare the ego-centred support networks of drug-free and relapsed individuals and to investigate the network characteristics that contribute to the maintenance of abstinence. To the best of our knowledge, this is the first study designed to

Please cite this article in press as: Panebianco, D., et al. Personal support networks, social capital, and risk of relapse among individuals treated for substance use issues. International Journal of Drug Policy (2015), http://dx.doi.org/10.1016/j.drugpo.2015.09.009

G Model

DRUPOL-1642; No. of Pages 8 D. Panebianco et al. / International Journal of Drug Policy xxx (2015) xxx–xxx

address the role of support networks and social capital among a sample of post-treatment relapsed and drug-free Italian drugusers. Methods Recruitment of participants and procedures Respondents were ex-clients (though some had returned to treatment) of a large non-governmental therapeutic substance abuse treatment agency providing voluntary residential treatments and rehabilitation services. This agency consists of three residential treatment centres, two for males and one for females. The average amount of time spent in treatment for people attending this agency is two years. To be eligible for this study, individuals must have left the treatment agency at least six months before the interview. The six month follow-up period is long enough after treatment that participants are likely to have settled into an autonomous lifestyle while also allowing for any shortterm treatment effects to wear off (i.e., even those with a high likelihood of relapse might not be captured as having relapsed if the interview was done immediately after treatment). Additionally, a substantial amount of time must have passed for there to be any confidence that the abstinence reported by drug-free participants is not short lived. Agency staff provided the sampling frame and contact information. Due to a variety of factors (e.g., invalid telephone numbers, non-response, incarceration, living abroad, death, or declining to participate in the study), 80 participants were interviewed out of approximately 110 that were contacted. Interviews were conducted at the treatment agency, public cafes, respondents’ workplaces and, when necessary, via video conference. Interviews were conducted between April and June 2012 and lasted approximately 90–120 min each. All participants provided informed consent. No honoraria were provided. The questionnaire was piloted by interviewing nine individuals. Measures Dependent variable The dependent variable was an indicator of whether or not the participant had relapsed. It was coded 1 if they had used any illicit drugs in the past 12 months and 0 if they had not. Fifteen respondents reported having relapsed while 65 remained drug-free. Independent variables Participants were asked to provide socio-demographic information such as sex, age, marital status (married/cohabiting or not), and socio-economic status (SES). SES was created by summing measures of education and occupational status. Response categories for education were: 0 = ‘‘never went to school/elementary school’’; 1 = ‘‘secondary school’’; 2 = ‘‘low school diploma’’; 3 = ‘‘high school diploma’’; 4 = ‘‘graduated from college (bachelor degree)’’; and 5 = ‘‘professional education’’. Response categories for occupation were: 0 = ‘‘unemployed’’; 1 = ‘‘unskilled worker’’; 2 = ‘‘skilled worker’’; 3 = ‘‘low-level white collar’’; 4 = ‘‘high-level white collar’’; and 5 = ‘‘professional’’ (Bearman, Moody, & Stovel, 2004). Unemployed respondents were asked to describe their last job. Participants provided information on their addiction history and past treatments. They reported the number of years they felt they were addicted and the types of drugs they generally used. There were three main drug use patterns: heroin-only, cocaineonly, and heroin and cocaine together (only two participants reported using other substances).

3

The social network analysis examined characteristics of, and relationships among, the people in the respondent’s support network. The social network component of the survey included the ‘‘name generator’’ and ‘‘name interpreter’’, two commonly used approaches to collecting social network data. With the name generator, individuals are asked to list the people they associate with (often divided into different domains) (Marsden, 1987, 2011). We used it to produce a list of personal support network members having a significant relationship with the participant and to whom he/she would ask for support/help with problems. Participants were asked to provide full names and/or nicknames of everyone they considered to be members of their support network. Network members were classified along the following lines: family, community (colleagues/classmates, friends, neighbours), people tied to the treatment agency (e.g., staff, former clients, or anyone involved in agency activities), and third sector organizations (e.g., social cooperatives, social clubs, sports associations, religious movements, political parties). The name interpreter (Marsden, 2011) is an instrument that provides information on the characteristics of a person’s contacts and the nature of the relationship between ego and their alters. Respondents reported on the occupation of alters and the strength of the tie with support members in terms of emotional closeness, duration of the bond, and frequency of communication. Additionally, respondents assessed the reciprocity in their support relationships with their alters, and reported on the presence and strength of ties among their alters. From the information collected using the name generator and name interpreters, the following network variables were created using SPSS (IBM Corp, 2011). Network size is the number of alters in the support network. We modeled the relationship between overall support network size and relapse risk, though the size of the family, community, treatment agency, and third sector organizations subnetworks are presented in Table 1 for descriptive purposes. Network closure is operationalized by a weighted measure of network density, following Marsden (1987). Network density is the number of ties between network members divided by the number of possible ties. However, in this study, ties were weighted by the strength of the relationship (as reported by ego). The tie between each pair of alters was coded 1 if the individuals were perceived as close, 0.5 if the tie was superficial or distant, and 0 if there was no tie between alters. This permitted the calculation of weighted network density, which might also be called mean tie strength (where an absent tie has a strength of 0), which ranges from 0.0 where no alters are connected, to 1.0 where all pairs of alters are closely tied to each other. Occupational heterogeneity is related to the diversity of work positions held by network members. Following Marsden’s (1987, 2011) approach to heterogeneity, it was calculated by taking the standard deviation of alters’ occupational status (coded on the 0–5 scale described above). A higher standard deviation across alters’ occupational status indicates that ego is connected to people at both higher and lower reaches of occupational status. This type of diversity ‘‘implies integration into several spheres of society, which is deemed advantageous for instrumental actions like gathering information’’ (Marsden, 1987, p. 124). Heterogeneity and density measures are meaningful only for respondents who report more than one alter; these measures were therefore calculated only for the 75 ego networks meeting that condition. Respondents were asked to list support members who have or would solicit help from them. Based on this, reciprocity was measured as the percentage of alters that ego perceives would request (or have requested) support from them to solve problems. Since alters are individuals who offer support to ego, reports that ego offers support to alters indicate a reciprocal relationship.

Please cite this article in press as: Panebianco, D., et al. Personal support networks, social capital, and risk of relapse among individuals treated for substance use issues. International Journal of Drug Policy (2015), http://dx.doi.org/10.1016/j.drugpo.2015.09.009

G Model

DRUPOL-1642; No. of Pages 8 D. Panebianco et al. / International Journal of Drug Policy xxx (2015) xxx–xxx

4 Table 1 Descriptive statistics.

Drug free % (n = 65) Sociodemographic and substance use characteristics Sexa Male = 0 (%) 84.62 Female = 1 (%) 15.38 Marital statusa Not married/cohabiting = 0 (%) 60.00 Married/cohabiting = 1 (%) 40.00 Type of substance useb Heroin + cocaine (%) 69.84 Heroin-only (%) 20.63 Cocaine-only (%) 9.52

c

Age Socio-economic statusc Number of years addictedc Support network profile Network size Overallc Familyd Communityc Treatment agencyc Third sector organizationc Occupational heterogeneityd Reciprocityc Alter socio-economic statusc Emotional closenessc Densityc Effective sizec Constraintc

Relapsed % (n = 15)

Full sample % (n = 80)

p

93.33 6.67

86.25 13.75

.377

93.33 6.67

66.25 33.75

.014

86.67 0.00 13.33

73.08 16.67 10.26

.145

Mean

Mean

Mean

39.26 3.77 10.38

36.33 1.60 14.00

38.71 3.36 11.06

9.10 1.62 6.49

9.06 2.83 4.12 1.86 0.25 1.32 84.90 60.18 3.42 0.77 4.57 0.43

6.47 1.47 1.00 4.00 – 0.63 33.20 44.77 3.23 0.94 2.53 0.70

8.57 2.58 3.54 2.26 0.25 1.21 75.74 57.51 3.39 0.79 4.21 0.48

4.64 1.70 3.13 2.53 0.78 0.56 35.09 14.91 0.49 0.23 2.81 0.27

SD

Min 21 1 1

0 0 0 0 0 0 0 24 2 0.28 1 0.17

Max

p

63 7 29

22 7 12 10 5 2.63 100 88 4 1 13.31 1.39

.297 .000 .019

.082 .005 .000 .041 – .000 .000 .000 .676 .008 .003 .090

n = 75 for density and occupational heterogeneity variables. Measures not valid for those with network size  1 (n = 5). n = 79 for emotional closeness and constraint index variables. Measures not valid for those with network size = 0 (n = 1). a Chi-square test. b Fisher’s exact test. n = 78. Two participants reported using other substances. c Mann–Whitney U test. d t-test.

Emotional closeness with alters was measured via a four point scale: 1 = ‘‘very distant’’; 2 = ‘‘somewhat distant’’; 3 = ‘‘somewhat close’’; 4 = ‘‘very close’’. These were averaged across all alters.2 Alters’ socio-economic status was derived from the occupational prestige score of network members by first calculating the occupational prestige of each alter using the International SocioEconomic Index (Ganzeboom & Treiman, 2003). The highest alter prestige score is used as our measure of alter SES as it is expected that individuals with access to higher SES alters within their personal support network will have more valued resources at their disposal (Lin, 2001) that can be mobilised to help prevent relapse. Burt’s concept of ego’s brokerage can be operationalized by the effective size of ego’s network, or by ego’s constraint (Burt, 1992, 2000). Effective size indicates the extent to which an individual connects various groups.3 Constraint indicates ‘‘the extent to which all of a person’s network time and energy is concentrated in one contact’’ (Burt, 2005, p. 26; see also Burt, 1992, chapter 2; Burt, 2 Length of bond (number of years ego had known alter) and frequency of communication (how often ego talked to alter: 1 = ‘‘once a month or less’’; 2 = ‘‘weekly’’; 3 = ‘‘daily’’) were also examined as measures of bond strength. Preliminary analyses showed that they were not related to relapse. Further, they did not scale together well. 3 According to Borgatti (1997), effective size for undirected, non-valued data is calculated using the following: i P h P j 1 p piqmiq 6¼ i j . where z piq ¼ Piqz ; i 6¼ j. j ij and z miq ¼ max iqðz Þ ; j 6¼ k. k

ik

2000).4 Both measures were calculated using UCINET (Borgatti, Everett, & Freeman, 2002). The two measures are closely related conceptually and empirically (Spearman’s r = .856, p < .01), though the fact that constraint is about concentrating ‘‘network time and energy. . .in one contact’’ as opposed to bridging disconnected groups ensures that the concepts are distinct enough to warrant an investigation into whether their effects on relapse differ. Analytic methods We first compared bivariate relationships between relapse and socio-demographic and social network dimensions using chi-square tests, Mann–Whitney U tests, Fisher’s exact tests, and t-tests. Next, bivariate logistic regression models were estimated in Stata (StataCorp, 2011) predicting membership in the relapse group (0 = drug-free, 1 = relapsed). This was followed up with multivariate logistic regressions predicting relapse from the variables that were significant in the bivariate regressions. We hesitate to draw firm conclusions from the multivariate analysis given the relatively small sample size. The multivariate analysis is instead intended to give a sense of the 4 According to Burt (2000 – Appendix available at http://faculty.chicagobooth. edu/ronald.burt/research/files/NSSC_APP.pdf), constraint is ‘‘the extent to which i’s network is directly or indirectly invested in a relationship with contact j: cij = (pij + Sqpiqpqj)2, for q 6¼ i, j, where pij is the proportion of i’s network time and energy invested in contact j, pij = zij/Sqziq, and variable zij is the strength of relationship between contacts i and j. The total in parentheses is the proportion of i’s relations that are directly or indirectly invested in connection with contact j. The sum of squared proportions, Sicij is the network constraint index C.’’

Please cite this article in press as: Panebianco, D., et al. Personal support networks, social capital, and risk of relapse among individuals treated for substance use issues. International Journal of Drug Policy (2015), http://dx.doi.org/10.1016/j.drugpo.2015.09.009

G Model

DRUPOL-1642; No. of Pages 8 D. Panebianco et al. / International Journal of Drug Policy xxx (2015) xxx–xxx

5

their personal networks as being emotionally close (mean for DFs = 3.42 out of 4; mean for Rs = 3.23; p = .676). Regarding social capital, while closure (weighted tie density) was high for both groups, it was significantly higher for Rs (mean = 0.94) than for DFs (mean = 0.77). Additionally, DFs had greater effective size (mean = 4.57) than Rs (mean = 2.53). DFs were also less constrained (mean = 0.43) than Rs (mean = 0.70) (though at p = .09, this difference is only marginally significant). Insofar as high levels of effective size and low levels of constraint indicate brokerage opportunities, drug free respondents tended to have more brokerage social capital than relapsed respondents. Table 2 provides more information on the relationship between the various predictors and relapse. Of the socio-demographic variables, being married or cohabiting and SES were negatively related to relapse. Compared to those not in a cohabiting relationship, married/cohabiting respondents were found to be less likely to relapse (b = 1.87; SE = 0.82). Higher levels of SES were associated with a reduction in the likelihood of being in the relapse group (b = 3.44; SE = 0.46). Age and sex were not significantly associated with the risk of relapse. With regard to addiction history, participants who reported only using heroin were found to be less likely to relapse (b = 2.10; SE = 0.42) than those who used both heroin and cocaine. There was no difference in the likelihood of relapse among participants who only used cocaine as compared to those who used both heroin and cocaine. Finally, participants who reported having been addicted for longer periods were found to be more likely to relapse (b = 0.08; SE = 0.04). Four network variables were found to be significantly related to relapse. The occupational heterogeneity of network members was negatively associated with ego’s relapse: participants with support networks in which the occupational status of alters was more homogeneous (less heterogeneous) were more likely to report having relapsed (b = 2.21; SE = 0.62). We also see that higher levels of reciprocity in the support network were related to lower odds of relapse (b = 0.04; SE = 0.01). Alter socio-economic status was also significantly related to relapse risk. Respondents with lower socioeconomic status alters were more likely to relapse (b = 0.08; SE = 0.04). Finally, brokerage (as indicated by the constraint index) was associated with relapse. Participants who were more constrained in their network had a greater likelihood of relapse (b = 3.00; SE = 1.13). Network size, density, effective size, and emotional closeness were not significantly associated with relapse.

robustness of the bivariate results. Firth logistic regression models, which produce ‘‘finite parameter estimates by means of penalized maximum likelihood estimation’’ (Heinze & Schemper, 2002, p. 2409), were employed to correct for issues of perfect separation (Albert & Anderson, 1984; Firth, 1993). In an effort to stabilise the results, bootstrap estimation was employed for both the bivariate and multivariate models (2000 replications per model) (Efron & Tibshirani, 1993). Multicollinearity was not found to be an issue (highest variance factors were less than 2).

Results Sample characteristics: socio-demographics and substance use The sample consisted of 65 (81%) drug free participants (DFs) and 15 (19%) relapsed participants (Rs) with males accounting for the majority of both groups (85% of DFs, 93% of Rs) (see Table 1). The mean age was 39 years for the DFs and 36 years for the Rs. DFs were significantly more likely to be married (40% versus 7% for Rs). The mean SES score for DFs was 3.8 versus 1.6 for Rs which suggests that relapsed participants were significantly more socioeconomically disadvantaged when compared with drug-free participants. Participants were not casual drug users (mean number of years addicted = 11.1). Rs tended to have longer addiction careers (mean = 14 years) than DFs (mean = 10 years). Similarities were found across the DF and R groups in terms of the types of substances used; the majority of respondents used both heroin and cocaine (70% for DFs, 87% for Rs). Social network characteristics The support networks of DF participants tended to be larger than those of Rs overall and in terms of family and community domains (see Table 1). Rs maintained a greater number of ties to the treatment agency despite no longer being in treatment. Reciprocity among drug free participants and their alters was very high (mean = 85% of alters would reciprocate) when compared with relapsed participants (mean = 33%). The highest SES among alters was greater for DFs (mean = 60) than Rs (mean = 45). Additionally, DFs had greater occupational heterogeneity (mean = 1.32) than Rs (mean = 0.63). Both drug free and relapsed participants portrayed

Table 2 Bivariate logit models predicting relapse group membership (0 = drug-free, 1 = relapsed). b

Sex (female = 1) Marital status (married/cohabiting = 1) Type of substance use (ref = heroin + cocaine) Heroin-only Cocaine-only Age Socio-economic status Number of years addicted Network size Occupational heterogeneity Reciprocity Alter socio-economic status Emotional closeness Density Effective size Constraint * **

0.60 1.87* – 2.10** 0.24 0.03 3.44** 0.08* 0.13 2.21** 0.04** 0.08* 0.74 4.34 0.38 3.00**

SE

(0.85) (0.82) – (0.42) (0.89) (0.02) (0.46) (0.04) (0.10) (0.62) (0.01) (0.04) (0.70) (3.68) (0.37) (1.13)

OR

0.55 0.15 – 0.12 1.27 0.97 0.03 1.08 0.88 0.11 0.96 0.92 0.48 76.71 0.68 20.09

95% CI of OR

n

Lower

Upper

0.10 0.03

2.92 0.77

80 80

0.05 0.22 0.92 0.01 1.01 0.72 0.03 0.94 0.86 0.12 0.06 0.33 2.20

0.28 7.17 1.01 0.08 1.17 1.07 0.37 0.99 0.99 1.90 102744 1.40 183.09

78 80 80 80 80 75 79 75 79 75 79 79

p < 0.05. p < 0.01.

Please cite this article in press as: Panebianco, D., et al. Personal support networks, social capital, and risk of relapse among individuals treated for substance use issues. International Journal of Drug Policy (2015), http://dx.doi.org/10.1016/j.drugpo.2015.09.009

G Model

DRUPOL-1642; No. of Pages 8 D. Panebianco et al. / International Journal of Drug Policy xxx (2015) xxx–xxx

6

Despite the small sample size, we included the predictor variables that were significant in the bivariate regressions in three multivariate models (see Appendix). Given the low statistical power of these models, it is not surprising that significant relationships tend to be washed out when other predictors are included. However, the direction and relative magnitude of the results are similar for most variables to the results of the bivariate models. This offers tentative support for these dynamics but should be confirmed using larger samples. The exceptions are the measures of constraint, marital status, and cocaine use. The direction of the effect of constraint and cocaine use switches across multivariate models; the direction of marital status switches from the bivariate to the multivariate models. Discussion The present study examined the personal support network, social capital, and individual characteristics of 80 Italian adults following treatment for substance use issues. The primary focus was on differences between the support networks of participants who report having relapsed versus those who report being abstinent. Although previous research has analysed substance use and health problems using social network and social capital frameworks, little attention has been paid to post-treatment support networks accounting for network structure and composition, relational properties, and closure and brokerage social capital, all of which are theoretically relevant contributors to relapse risk. We found that respondents who remained drug-free (‘‘DFs’’) between the time they left the treatment agency and the survey had on average larger and less dense support networks, with whom they maintained more reciprocal relationships, than those who relapsed (‘‘Rs’’). In particular, the support networks of relapsed participants consisted predominantly of family members and people from the treatment community. Consequently, their ties were mostly pre-existing (i.e., family) or, as suggested by the lack of reciprocity, asymmetrical relationships with the people of the treatment agency. The give-and-take of DFs’ ties suggests that interactions were not exclusively need-oriented, but characterized by higher levels of relational equity, and suggests ego’s ability to provide some degree of support to others. Reciprocity may work to enhance self-efficacy and self-esteem which may play a role in lowering the risk of relapse. Additionally, DFs had access to alters with higher socioeconomic status (cf. El-Bassel et al., 1998) and greater occupational heterogeneity than Rs. This suggests that DFs have more extensive and higher reaching support networks than relapsed subjects, which DFs may be able to draw on to help them maintain abstinence. Finally, DFs had greater average levels of brokerage social capital (i.e., support networks with greater effective size and less constraint) than Rs, whose networks were characterized by higher levels of closure (density). Although the networks of both groups were quite dense, the greater effective size of the networks of DFs suggests that they might be able to access a more diverse set of information than Rs, a phenomenon that, according to Burt (2005), tends to improve outcomes (in this case, reducing the risk of relapse). This study has several limitations. A primary issue is the extent to which the current findings are generalizable to other relevant populations. There were difficulties in recruiting participants, specifically relapsed subjects who often refused to be interviewed or did not show up for their interviews. This means that the sample is not necessarily representative of all clients of the treatment agency used for the study. Further, the average length of treatment was two years, substantially longer than what is available in most other regions. But while generalizability is clearly an issue, there is

no reason to suspect that the social network dynamics relating to relapse would systematically differ according to the length of treatment. However, this remains an empirical question that will need to be addressed by studies using other post-treatment populations. An additional limitation is that the information on personal support networks came exclusively from respondents’ (i.e., egos’) self-reports, which means that the measured strength and reciprocity of ties between ego and alters, and the measured presence or absence and strength of ties between pairs of alters, depend on egos’ perceptions. Informant recall of network ties is known to suffer from varying degrees of inaccuracy depending upon the particular context and steps taken by researchers to address the problem (see Bernard, Killworth, & Sailer, 1981; Bernard, Killworth, Kronenfeld, & Sailer, 1984; Marsden, 2011 for reviews of the issue). In this study, researchers were trained to help respondents consider their networks in terms of support characteristics (e.g., by providing examples of social support), a procedure that should have improved accuracy and reduced interrespondent variability in interpretations. Concerning the accuracy of respondents’ assessments of reciprocity, it may be the case that it is ego’s perception of reciprocity that is important, rather than ‘‘actual’’ reciprocity. Further, relapsed participants represented a small proportion of the sample and some had re-entered residential treatment at the time of the interview. This is essentially a sample size issue but it also highlights the importance of replication in other contexts. It would be informative to extend the scope of the study to a larger number of relapsed individuals who are not in treatment where emergent relational dynamics and individual needs may operate differently. Therefore, in order to extend the generalizability of these results, future research should attempt to replicate the study using a larger sample within a wider social context and should focus on post-treatment risk and protective factors including maintenance and reinforcement of the positive elements of therapy within the broader social network. Finally, the risk of type I error (finding a result to be significant when it is not) is elevated due to the number of statistical tests that were conducted. However, we chose not to employ p-value corrections given the problems associated with them: (a) their use means that ‘‘the interpretation of a finding depends on the number of other tests performed’’ (Perneger, 1998, p. 1236) and (b) they tend to be overly conservative5 (Nakagawa, 2004; Perneger, 1998). Our approach was therefore to set p < .05 as the cut point for statistical significance for each test while reminding readers of the possibility of false positives. The results of this study have implications for treatment programming, although recommendations must be considered tentative until the results are replicated. In general, the results support research on the integration of personal support networks into therapeutic programming (e.g., Copello et al., 2006; Galanter, 1993; Litt, Kadden, Kabela-Cormier, & Petry, 2007; Litt, Kadden, Kabela-Cormier, & Petry, 2009; Meyers & Miller, 2001; Meyers et al., 1998). More specifically, our study suggests that upon leaving treatment for substance use issues, individuals are often minimally integrated in the broader community; therefore, in addition to individual treatment, network interventions that imply a more dynamic role of social workers, counsellors and other health professionals are likely to improve individual capacity to avoid relapse. Optimizing the support network in the posttreatment phase should involve interacting with a diverse set of others from a variety of social domains that all promote substance avoidance or minimisation. 5 With 34 statistical tests in this study, a standard Bonferroni correction would require a p-value less than .05/34 = .00147 to be considered significant.

Please cite this article in press as: Panebianco, D., et al. Personal support networks, social capital, and risk of relapse among individuals treated for substance use issues. International Journal of Drug Policy (2015), http://dx.doi.org/10.1016/j.drugpo.2015.09.009

G Model

DRUPOL-1642; No. of Pages 8 D. Panebianco et al. / International Journal of Drug Policy xxx (2015) xxx–xxx

7

Conflict of interest statement

Conclusion Overall, the study shows the importance of social support networks in relation to relapse risk among individuals who have recently completed treatment for substance use issues. In particular, possessing support networks consisting of loosely interconnected higher-status alters in diverse occupations, and having reciprocal ties with alters, were found to be related to a reduced risk of relapse. We recommend increased attention to therapies that incorporate the broader social network into relapse prevention strategies.

None. Financial disclosure Preparation of the paper was supported by a research grant from the Social Science and Humanities Research Council (Canada). No funding source had any involvement in any aspect of the research design, execution of the research, or preparation of the paper.

Appendix. Multivariate logit models predicting relapse group membership (0 = drug-free, 1 = relapsed) Model 1

Marital status (married/cohabiting = 1) Type of substance use (ref = heroin + cocaine) Heroin-only Cocaine-only Socio-economic status Number of years addicted Occupational heterogeneity Reciprocity Alter socio-economic status Constraint Intercept n

Model 2

b

SE

OR

0.28 – 1.38 3.23 3.05* 0.13

(1.28) – (1.24) (2.90) (1.33) (0.09)

1.33 – 0.25 25.38 0.05 1.14

4.17

2.19 78

b

0.34 0.02 0.05 0.96 2.55

Model 3 SE

(1.10) (0.02) (0.05) (2.59) (3.14) 74

OR

0.71 0.98 0.95 2.61

b

SE

OR

0.62

(1.93)

1.86

2.61 0.57 2.13 0.17 0.59 0.01 0.10 1.01 8.96

(2.53) (3.69) (1.14) (0.13) (2.81) (0.02) (0.09) (4.15) (5.86) 72

0.07 0.57 0.12 1.19 0.55 0.99 0.90 0.36

* p < 0.05. **p < 0.01.

References Albert, A., & Anderson, J. A. (1984). On the existence of maximum likelihood estimates in logistic regression models. Biometrika, 71, 1–10. Bandura, A. (1977). Social learning theory. Englewood Cliffs, NJ: Prentice Hall. Bearman, P. S., Moody, J., & Stovel, K. (2004). Chains of affection: The structure of adolescent romantic and sexual networks. American Journal of Sociology, 110(1), 44–99. Bernard, H. R., Killworth, P., & Sailer, L. (1981). Summary of research on informant accuracy in social network data, and on the reverse small world problem. Connections, 4, 11–25. Bernard, H. R., Killworth, P., Kronenfeld, D., & Sailer, L. (1984). The problem of informant accuracy: The validity of retrospective data. Annual Review of Anthropology, 13, 495–517. Bourdieu, P. (1986). The forms of capital. In J. G. Richardson (Ed.), Handbook of theory and research for the sociology of education (pp. 241–258). New York: Greenwood. Borgatti, S. P. (1997). Structural holes: Unpacking Burt’s redundancy measures. Connections, 20, 35–38. Borgatti, S. P., Everett, M. G., & Freeman, L. C. (2002). UCINET for Windows: Software for social network analysis. Harvard, MA: Analytic Technologies. Brown, V. L., & Riley, M. A. (2005). Social support, drug use, and employment among low-income women. The American Journal of Drug and Alcohol Abuse, 31, 203–223. Burt, R. S. (1992). Structural holes. The social structure of competition. Cambridge, MA: Harvard University Press. Burt, R. S. (2000). The network structure of social capital. In R. I. Sutton, & B. M. Staw (Eds.), Research in organizational behavior (Vol. 22, pp. 345–423). Greenwich, CT: JAI Press. Burt, R. S. (2005). Brokerage and closure. An introduction to social capital. New York: Oxford University Press. Cheung, Y. W. (2009). A brighter side: Protective and risk factors in the rehabilitation of chronic drug abusers in Hong Kong. Hong Kong: The Chinese University Press. Cheung, Y. W., & Cheung, N. W. T. (2003). Social capital and risk level of posttreatment drug use: Implications for harm reduction among male treated addicts in Hong Kong. Addiction Research and Theory, 11(3), 145–162. Chua, V., Madej, J., & Wellman, B. (2011). Personal communities: The world according to me. In J. Scott, & P. J. Carrington (Eds.), The SAGE handbook of social network analysis (Vol. 8, pp. 101–115). London: Sage Publications. Cohen, S., & Wills, T. A. (1985). Stress, social support, and the buffering hypothesis. Psychological Bulletin, 98(2), 310–357. Coleman, J. S. (1988). Social capital in the creation of human capital. American Journal of Sociology, 94, S95–S120. Coleman, J. S. (1990). Foundations of social theory. Cambridge: Harvard University Press. Copello, A., Williamson, E., Orford, J., & Day, E. (2006). Implementing and evaluating social behaviour and network therapy in drug treatment practice in the UK: A feasibility study. Addictive Behaviors, 31, 802–810.

Dawson, D. A., Goldstein, R. B., & Grant, B. F. (2007). Rates and correlates of relapse among individuals in remission from DSM-IV alcohol dependence: A 3-year follow-up. Alcoholism: Clinical and Experimental Research, 31, 2036–2045. http:// dx.doi.org/10.1111/j.1530-0277.2007.00536.x Debnam, K., Holt, C. L., Clark, E. M., Roth, D. L., & Southward, P. (2012). Relationship between religious social support and general social support with health behaviors in a national sample of African Americans. Journal of Behavioral Medicine, 35, 179–189. Dufur, M. J., Parcel, T. L., & McKune, B. A. (2007). Comparing the efficacy of social capital created in different contexts: The case of adolescent substance use. Paper presented at the annual meeting of the American Sociological Association. Efron, B., & Tibshirani, R. J. (1993). An Introduction to the Bootstrap. New York: Chapman & Hall. El-Bassel, N., Cooper, D., & Chen, D. R. (1998). Social support and social networks among women on methadone. Social Service Review, 72, 379–401. Ellis, B., Bernichon, T., Yu, P., Roberts, T., & Herrell, J. M. (2004). Effect of social support on substance abuse relapse in a residential treatment setting for women. Evaluation and Program Planning, 27, 213–221. Firth, D. (1993). Bias reduction of maximum likelihood estimates. Biometrika, 80, 27–38. Galanter, M. (1993). Network therapy for substance abuse: A clinical trial. Psychotherapy: Theory, Research, Practice, Training, 30(2), 251–258. Ganzeboom, H. B. G., & Treiman, D. J. (2003). Three internationally standardised measures for comparative research on occupational status. In J. H. P. Hoffmeyer-Zlotnik & C. Wolf (Eds.), Advances in cross-national comparison. A European working book for demographic and socio-economic variables (pp. 159–193). New York: Kluwer Academic Press. Gleason, M. E. J., Iida, M., Bolger, N., & Shrout, P. E. (2003). Daily supportive equity in close relationships. Personality and Social Psychology Bulletin, 29, 1036–1045. Gordon, A. J., & Zrull, M. (1991). Social networks and recovery: One year after inpatient treatment. Journal of Substance Abuse Treatment, 8, 143–152. Granovetter, M. (1973). The strength of weak ties. American Journal of Sociology, 78, 1360–1380. Harocopos, A., Lloyd, A. G., Kobrak, P., Jost, J. J., & Clatts, M. C. (2009). New injectors and the social context of injection initiation. International Journal of Drug Policy, 20, 317–323. Heinze, G., & Schemper, M. (2002). A solution to the problem of separation in logistic regression. Statistics in Medicine, 21, 2409–2419. IBM Corp (2011). IBM SPSS Statistics for Windows, Version 20. 0. Armonk, NY: IBM Corp. Jonas, A. B., Young, A. M., Oser, C. B., Leukefeld, C. G., & Havens, J. R. (2012). OxyContin as currency: OxyContin use and increased social capital among rural Appalachian drug users. Social Science and Medicine, 74, 1602–1609. Kahn, J. H., Hessling, R. M., & Russell, D. W. (2003). Social support, health, and wellbeing among the elderly: What is the role of negative affectivity? Personality and Individual Differences, 35, 5–17.

Please cite this article in press as: Panebianco, D., et al. Personal support networks, social capital, and risk of relapse among individuals treated for substance use issues. International Journal of Drug Policy (2015), http://dx.doi.org/10.1016/j.drugpo.2015.09.009

G Model

DRUPOL-1642; No. of Pages 8 8

D. Panebianco et al. / International Journal of Drug Policy xxx (2015) xxx–xxx

Kirshenbaum, A. P., Olsen, D. M., & Bickel, W. K. (2009). A quantitative review of the ubiquitous relapse curve. Journal of Substance Abuse Treatment, 36, 8–17. http:// dx.doi.org/10.1016/j.jsat.2008.04.001 Kirst, M. J. (2009). Social capital and beyond: A qualitative analysis of social contextual and structural influences on drug-use related health behaviors. Journal of Drug Issues, 39(3), 653–676. Koob, G. F., & Le Moal, M. (2008). Addiction and the brain antireward system. Annual Review of Psychology, 59, 29–53. http://dx.doi.org/10.1146/annurev.psych.59. 103006.093548 Koram, N., Liu, H., Li, J., Li, J., Luo, J., & Nield, J. (2011). Role of social network dimensions in the transition to injection drug use: Actions speak louder than words. AIDS Behavior, 15, 1579–1588. Latkin, C. A., Mandell, W., Oziemkowska, M., & Celentano, D. (1995). Using social network analysis to study patterns of drug use among urban drug users at high risk for HIV/AIDS. Drug and Alcohol Dependence, 38(1), 1–9. Lin, N. (2001). Social capital. A theory of social structure and action. Cambridge: Cambridge University Press. Lin, N., & Erickson, B. H. (2008). Theory, measurement, and the research enterprise on social capital. In N. Lin & B. H. Erickson (Eds.), Social capital: An international research program (pp. 1–24). Oxford: Oxford University Press. Litt, M. D., Kadden, R. M., Kabela-Cormier, E., & Petry, N. (2007). Changing network support for drinking: Initial findings from the network support project. Journal of Consulting and Clinical Psychology, 75, 542–555. Litt, M. D., Kadden, R. M., Kabela-Cormier, E., & Petry, N. (2009). Changing network support for drinking: Network support project 2-year follow-up. Journal of Consulting and Clinical Psychology, 77, 229–242. Lorant, V., & Nicaise, P. (2015). Binge drinking at University: a social network study in Belgium. Health Promotion International, 30(3), 675–683. http://dx.doi.org/10. 1093/heapro/dau007 Lundborg, P. (2005). Social capital and substance use among Swedish adolescents – An explorative study. Social Science and Medicine, 61, 1151–1158. Marsden, P. V. (1987). Core discussion networks of Americans. American Sociological Review, 52, 122–131. Marsden, P. V. (2011). Survey methods for network data. In J. Scott, & P. J. Carrington (Eds.), The SAGE handbook of social network analysis (Vol. 25, pp. 370–388). London: Sage Publications. Meyers, R. J., & Miller, W. R. (2001). A community reinforcement approach to addiction treatment. New York: Cambridge University Press. Meyers, R. J., Miller, W. R., Hill, D. E., & Tonigan, J. S. (1998). Community reinforcement and family training (CRAFT): Engaging unmotivated drug users in treatment.

Journal of Substance Abuse, 10, 291–308. http://dx.doi.org/10.1016/S08993289(99)00003-6 Nakagawa, S. (2004). A farewell to Bonferroni: The problems of low statistical power and publication bias. Behavioral Ecology, 15(6), 1044–1045. Perneger, T. V. (1998). What’s wrong with Bonferroni adjustments. British Medical Journal, 316(7139), 1236–1238. Polivy, J., & Herman, C. P. (2002). If at first you don’t succeed: False hopes of self-change. American Psychologist, 57, 677–689. http://dx.doi.org/10.1037/0003-066X.57. 9.677 Rhoades, H., Wenzel, S. L., Golinelli, D., Tucker, J. S., Kennedy, D. P., Green, H. D., et al. (2011). The social context of homeless men’s substance use. Drug and Alcohol Dependence, 118, 320–325. Rice, E., Milburn, N. G., & Monro, W. (2011). Social networking technology, social network composition, and reductions in substance use among homeless adolescents. Prevention Science, 12, 80–88. Sinha, R. (2007). The role of stress in addiction relapse. Current Psychiatry Reports, 9, 388–395. http://dx.doi.org/10.1007/s11920-007-0050-6 Song, L., Son, J., & Lin, N. (2011). Social support. In J. Scott, & P. J. Carrington (Eds.), The SAGE handbook of social network analysis (Vol. 9, pp. 116–128). London: Sage Publications. StataCorp (2011). Stata Statistical Software: Release 12. College Station, TX: StataCorp LP. Stevens, E., Jason, L. A., Ram, D., & Light, J. (2014). Investigating social support and network relationships in substance use disorder recovery. Substance Abuse. http:// dx.doi.org/10.1080/08897077.2014.965870 Sutherland, E. H., & Cressey, D. R. (1978). Criminology (9th ed.). Philadelphia: JB Lippincott. Tobin, K. E., Hua, W., Costenbader, E. C., & Latkin, C. A. (2007). The association between change in social network characteristics and non-fatal overdose: Results from the SHIELD study in Baltimore, MD, USA. Drug and Alcohol Dependence, 87(1), 63–68. Tucker, J. S., Wenzel, S. L., Golinelli, D., Zhou, A., & Green, H. D. (2011). Predictors of substance abuse treatment need and receipt among homeless women. Journal of Substance Abuse Treatment, 40(3), 287–294. Tyler, K. A. (2008). Social network characteristics and risky sexual and drug related behaviors among homeless young adults. Social Science Research, 37, 673–685. Wenzel, S. L., Green, H. D., Tucker, J. S., Golinelli, D., Kennedy, D. P., Ryan, G., et al. (2009). The social context of homeless women’s alcohol and drug use. Drug and Alcohol Dependence, 105(1–2), 16–23. Witkiewitz, K., & Marlatt, G. A. (2004). Relapse prevention for alcohol and drug problems. American Psychologist, 59, 224–235. http://dx.doi.org/10.1037/0003066X.59.4.224

Please cite this article in press as: Panebianco, D., et al. Personal support networks, social capital, and risk of relapse among individuals treated for substance use issues. International Journal of Drug Policy (2015), http://dx.doi.org/10.1016/j.drugpo.2015.09.009