Skip to main content
NCBI home page
NIHPA Author Manuscripts logoLink to NIHPA Author Manuscripts
. Author manuscript; available in PMC: 2011 Mar 1.
Published in final edited form as: Drug Alcohol Depend. 2009 Nov 24;107(2-3):134. doi: 10.1016/j.drugalcdep.2009.09.012

Friendship networks of inner-city adults: A latent class analysis and multi-level regression of supporter types and the association of supporter latent class membership with supporter and recipient drug use

Amy SB Bohnert 1,2, Danielle German 3, Amy R Knowlton 3, Carl A Latkin 3
PMCID: PMC2822006  NIHMSID: NIHMS155358  PMID: 19939586

Abstract

BACKGROUND

Social support is a multi-dimensional construct that is important to drug use cessation. The present study identified types of supportive friends among the social network members in a community-based sample and examined the relationship of supporter-type classes with supporter, recipient, and supporter-recipient relationship characteristics. We hypothesized that the most supportive network members and their support recipients would be less likely to be current heroin/cocaine users.

METHODS

Participants (n=1,453) were recruited from low-income neighborhoods with a high prevalence of drug use. Participants identified their friends via a network inventory, and all nominated friends were included in a latent class analysis and grouped based on their probability of providing seven types of support. These latent classes were included as the dependent variable in a multi-level regression of supporter drug use, recipient drug use, and other characteristics.

RESULTS

The best-fitting latent class model identified five support patterns: friends who provided little/no support, low/moderate support, high support, socialization support, and financial support. In bivariate models, friends in the high, low/moderate, and financial support were less likely to use heroin or cocaine and had less conflict with and were more trusted by the support recipient than friends in the low/no support class. Individuals with supporters in those same support classes compared to the low/no support class were less likely to use heroin or cocaine, or to be homeless or female. Multivariable models suggested similar trends.

CONCLUSIONS

Those with current heroin/cocaine use were less likely to provide or receive comprehensive support from friends.

Keywords: drug use, social networks, social support, latent class analysis, multilevel modeling

1. Introduction

1.1 Background

Social support is well established as being critical to substance use cessation (Granfield and Cloud, 2001; McMahon, 2001; Scherbaum and Specka, 2008; Sobell et al., 2000; VanDeMark, 2007; Warner, 2004). While social support is widely agreed to be multidimensional (Veiel and Bauman, 1992), standard measurement of social support in addictions literature is usually limited to summary measures or focus on only one type of support (House and Kahn, 1985). As a result, much remains to be understood about the supportive functioning of friendship ties, and a greater understanding could greatly improve the utility of intake assessments at drug treatment centers.

1.1.1 Social Support and Drug Use Outcomes

Social support is defined as the resources provided to an individual by his or her social network members. Cohen (2004) writes that the mechanism behind social support is stress buffering, as support “eliminates or reduces effects of stressful experiences by promoting less threatening interpretations of adverse events and effective coping strategies” (pg. 677). Particular types of support that have been emphasized in previous studies of substance abuse include instrumental, emotional (able to talk to about private matters), informational (advice and knowledge), and social (spending time together and having fun).

In addition to mortality (Berkman et al., 2004), mental health (Kawachi and Berkman, 2001), and physical health (Cohen, 2004), social support is associated with reduced drug use and better drug treatment outcomes. Social support has been found to be an important factor in drug cessation among participants in treatment studies (McMahon, 2001; Scherbaum and Specka, 2008; VanDeMark, 2007; Warner, 2004). Evidence from qualitative studies suggest that network members also can provide support during the physical and psychological struggles of quitting drugs outside of treatment (Granfield and Cloud, 2001; Sobell et al., 2000). While these studies have demonstrated a relationship between social network support and successful drug cessation, the source of support, such as from friends, romantic partners, kin, and other network ties, which may have different influences on drug use and cessation, is rarely distinguished. Furthermore, while numerous studies find that social support is associated with health and treatment outcomes, much less is known about the content of social support relationships.

1.1.2 The Measurement of Social Support

Traditional methods of measuring social support frequently over-simplify and fail to adequately assess a person’s support resources, or address how specific network members do or do not specialize in the type of support provided (Wellman and Gulia, 1997). Summary scores are limited in their ability to provide insight into the relationship of support and drug use. Further research is needed to understand the relative importance of types of supporters.

Some research has focused on the effects of particular types or dimensions of support on drug use. For example, VanDeMark (2007) found instrumental support to be protective for relapse. The authors suggested instrumental support provided individuals who had quit more opportunities to take on new social roles, such as an employee, student, or parent. In contrast, Hanson and colleagues (1990) found that emotional support, but not instrumental and informational support, was associated with cessation. Furthermore, the ability of supporters to provide emotional support may have particular importance (relative to instrumental and informational support) in buffering the effects of emotional stressors by reframing life events, which was proposed by Cohen (2004) as a mechanism through which support impacts health. A better understanding of whether supporters provide specific kinds of support or multiple types of support, and the relation between these types of supporters and drug use outcomes, is needed in order to improve intervention strategies.

There is a burgeoning interest in latent variable methods to better model the complex inter-relationships between support types (Agneessens et al., 2006; Buckman et al., 2007; Buckman et al., 2008; Liebler and Sandefur, 2002). However, studies using a latent variable approach to measure social support have often measured the clusters of provision or exchange of support on a cumulative, network-wide level, rather than attempting to identify patterns of support provision by individual network members. Thus, the ways in which types or degrees of support provided by specific individuals contributes to drug use outcomes are obscured.

1.1.3 Friendship network support

Research suggests that most individuals report receiving support from friends (Liebler and Sandefur, 2002). Friends may be particularly important providers of support. Fiori and colleagues (2006) found in a sample of older adults that having a network of all friends and no kin was protective for depression compared a network of all family and no friends. While kin also provide important sources of support, the present study focused on the support provided by friends which are more often ties of choice and may be easier to alter. Furthermore, establishing and maintaining friendship ties may entail different social skills and normative expectations of reciprocity compared to kin. Consequently, friend supporters merit separate analysis.

1.2 The Present Study

The present study used an inner-city Baltimore community sample of adults (“index participants”) with a high prevalence of drug use to explore the relationship between specific types of supportive friendships and drug use. We used a latent variable modeling approach to identify types of supporters among friend network members. The present study addresses several gaps in the literature. First, it extends a limited amount of prior research using latent class analysis to describe the types of supporters among friend network members. Second, the present study furthers this avenue of research by examining the relationship of latent class membership with characteristics of both the supporter and the support recipient, with particular emphasis on drug use. We hypothesized that those network members who provide greater support and index participants with network members who provide greater support would be less likely to be current drug (heroin and/or cocaine) users. The findings improve our understanding of the dynamics of social support and how this information could be utilized to augment treatment and relapse prevention.

2. Methods

2.1 Sampling and recruitment

The present study was a secondary analysis of data from the baseline assessment of a network-based HIV prevention intervention study (the Self-Help In Eliminating Life-threatening Diseases study; Latkin et al., 2003a; Latkin et al., 2003b). Study procedures were reviewed and approved by the Committee on Human Research at the Johns Hopkins School of Public Health. All participants provided informed consent and were paid $20 at the completion of the interview.

Individuals were recruited into the study using targeted outreach in from neighborhoods in Baltimore, Maryland with high rates of drug use. Interested individuals were given a toll-free number to call and a written description of the study. Participants were eligible for the intervention if they had at least weekly contact with drug users, were 18 years of age or older, were willing to conduct outreach education on HIV risk behavior reduction, were willing to bring in a network member for assessment, and were not recently enrolled in other HIV behavioral interventions. Additional respondents (n = 597) were recruited from among social network members. Participants for the present analysis reported at least one social network member identified as a friend, which included 1,453 of the 1,617 baseline participants with network data. The 1,453 participants listed 5,969 friend network members, resulting in a mean number of friend network members per index participant of 4.1

2.2 Assessment and measures

2.2.1 Index-level variables

All data were collected in face-to-face interviews. Index participants reported information regarding themselves and their social network members. Drug use was assessed three ways: any use of heroin and/or cocaine via any route of administration in the prior six months, use of heroin and/or cocaine via injection in the prior six months, and using crack cocaine exclusively in the prior six months. Education was queried as years of schooling completed and a binary variable was coded to represent having completed a high school degree or equivalent. Homelessness was defined as being homeless for any period in the prior six months. Age was split into categories of 32 or fewer years of age, 33 to 42 years of age, and 43 to 67 years of age, based on the sample distribution.

2.2.2 Network member/Supporter-level variables

Information on network members was assessed in the index interview via a social network inventory (Barrera and Gottlieb, 1981). Names of network members were generated by asking 18 questions, such as “If you needed to borrow $25 or something valuable, is there anybody you know who would lend or give you $25, or more, or something that was valuable?,” “Who are the people that you do drugs with?,” and “If you needed a place to stay can you think of anyone who would let you stay at their place?” Respondents then reported the type of role relation with each network member, such as parent, boy/girlfriend, sexual partner, friend, and professional associate. A network member was considered a friend and included in analyses if the index identified that person as a “friend” (n = 5,831), “someone at work” (n = 23), “neighbor” (n = 71), or “housemate” (n = 44) (Liebler and Sandefur, 2002).

2.2.2.1 Social Support

Questions regarding the provision of types of support were chosen to cover a range of support dimensions identified in prior research. The types of support assessed and the frequency of endorsement for friend network members are reported in Table 1.

Table 1.

Description of the support questions included in the latent class analysis of friend network supporter types, and friend network and index participant characteristics included in latent class regression.

Friend network-level questions to assess support types (n = 5,969)
Question asked (Question name) Prevalence (%)
Who could you trust with your money (“Money”) 21.1
Who do you talk to about private matters (“Talk”) 17.5
Who would give up their time to help you (“Time”) 21.4
Who would lend you more than $25 (“Lend”) 21.5
With whom do you have fun and relax (“Hang”) 37.2
From whom would you ask for advice about STDs (“Advice”) 25.9
Who would or did let you stay at their place (“Stay”) 43.7
Friend network characteristics (n = 5,969)
Characteristic Prevalence (%)
Sex/Gender: male 57.6
Current heroin or cocaine user 57.6
Conflict with Index 11.5
Sees Index daily 35.7
Known Index at least 5 years 57.9
Is someone Index supports 6.7
Index trusts with life 26.3
Index particpant characteristics (n = 1,453)
Characteristic Prevalence (%)
Sex/Gender: Male 61.4
Drug use (past six month)
 -any heroin or cocaine 72.6
 -injection drug use 71.3
 -crack use only 4.3
 -any heroin use 64.7
High school degree 54.3
Age
 -age 18 to 32 18.9
 -age 33 to 42 50.5
 -age 43 to 67 30.6
Homelessness (past six month) 25.4
Open in a new tab
2.2.2.2 Supporter characteristics

In addition to support provision, information collected about each network member from the index included the types of drugs use, gender, provision of financial or material support by the index to the network member (reciprocity), and conflict (queried as “who are you often on bad terms with?”). For frequency of contact, a binary variable was created that indicated if a network member was a person who the index reported seeing “daily.” For trust, index participants were asked how much they trust each network member, on a scale of 1 (not at all) to 10 (trust with life). A binary variable was created that indicated for each network member if the index rated that person a “10” for trust. Length of time knowing each network member was dichotomized as “less than five years” or “five years or longer.”

2.3 Statistical Analyses

We executed all analyses using MPlus version 5.0 (Muthén and Muthén, 1998–2007), including imputation for missing data. Data was missing on only 3 of the 1,453 (0.2%) index participants and 143 of the 5,969 (2.4%) of network members.

2.3.1 Latent class analysis

We used latent class analysis (LCA; McCutcheon, 1987; Vermunt, 2008) to empirically define and describe subgroups of the friend network members in terms of the types of support they provide to the index. This method is a well-accepted for measuring outcomes related to drug use (Buckman et al., 2007; Buckman et al., 2008; Ghandour et al., 2008; Lynskey et al., 2006; Monga et al., 2007) and measuring complex social factors (Agneessens et al., 2006; Buckman et al., 2008; Fiori et al., 2006). Integrating a person-centered approach (e.g., latent variable analysis) with more common variable-centered approaches (e.g., regression modeling) was chosen in order to yield a greater understanding of social support resources important for drug use outcomes (Muthén and Muthén, 2000). Specifically, by integrating a latent class analysis with traditional variable-centered regression methods the analysis examined the relationship of supporter characteristics with the type of support they provide, as well as recipient characteristics associated with having particular types of supporters.

All LCA models included the seven binary items listed in Table 1 under “Friend network-level questions to assess support types.” All models were clustered by index. We determined the number of classes that exist in the sample by performing the analysis iteratively and specifying an additional class each time, and comparing the models with varying numbers of classes on measures of fit (minimizing Akaike’s Information Citeria, Bayesian Information Criteria, and sample-size adjusted Bayesian Information Criteria) and the interpretability of the results.

2.3.2 Latent class regression

We examined the relationship of index- and network member-level variables with latent class membership using a bivariate and multivariable latent class regression (Bandeen-Roche et al., 1997). We regressed the latent variable of class membership on covariates in the same step in which the measurement model is estimated. The latent class regression was conducted as a multi-level analysis, because there were multiple supporters per recipient. All supporter characteristics were specified as on a “within” level, while all index characteristics were specified on a “between” level.

3. Results

3.1 The Sample

The present sample was largely drug-experienced; 73% had used heroin and/or cocaine users in the prior six months, 88% had a known lifetime history of heroin and/or cocaine use, and 57.6% of all friend network members were current heroin and/or cocaine users. The sample was 95% Black. Only 7% reported ever having been married, but 59% reported currently having a main partner. Participants listed between one and twenty-four network members in their assessment. Participants who reported at least one network member as a “friend” as compared to those with no friends in their network (n=164) did not differ by race, sex/gender, education, or current drug use but were more likely to have a main partner (74.9% vs. 58.8%, p < 0.001).

3.2 Latent Class Analysis

Model selection information for the latent class analysis of supporter types for models having two, three, four, five, and six classes are presented in Table 2. While the AIC and the sample-size adjusted BIC were lower for a six-class model than a five class model, the BIC was higher, and there were problems with the empirical identifiability of this complex model, indicated by instability of estimates in replications. Qualitatively, the only difference between the six-class model and the five-class model was that one of the classes in the five-class model (Class #1 in Figure 1) became two classes, with estimated prevalences of 4 and 5%, and which only differed substantially in the probability of providing advice. In the interest of model identifiability and parsimony, a five-class model was selected.

Table 2.

Model selection information for latent class models of varying number of classes among the friend network members of index participants in the SHIELD study in Baltimore, MD.

# Latent Classes Parameters AIC BIC Sample-Size Adj. BIC Identifiabilitya
2 15 36378 36478 36430 OK
3 23 35837 35991 35918 R
4 31 35572 35780 35681 R
5 39 35366 35627 35503 R
6 47 35326 35641 35491 N
Open in a new tab
a

OK = no problems in modeling detected, R = replication assured the best solution has been reached, N = parameter estimates not trustworthy after increasing replications.

Figure 1.

Figure 1

Open in a new tab

Latent class model of support types among network friends: Item response probabilities of the five-class model. See table 1 for a description of the support items.

Figure 1 reports the class prevalences (based on estimated posterior probabilities) and the conditional probability of a friend providing each of the seven types of support, given class membership. The results of the latent class analysis indicated that types of supportive ties with friends included those friendships in which the friend primary provided a specific domain/type of support and friendships in which the friend had a fairly consistent probability of providing support across multiple domains. The classes were interpreted as follows:

Class 1: Moderate Support, Instrumental/Financial

Supporters in this class provide the index a moderate amount of support overall and were more likely than friends in any other group (except “High Support”) to be someone the index can borrow money from and stay with, as well as be someone who the index would trust with their own money. (Estimated prevalence of 8.2%)

Class 2: Moderate Support, Socialization

Supporters in this class provide the index with a moderate amount of support overall, as the probability of friends in this group providing any of the seven types of support ranged between 28 and 70%. The types of support that these friends were most likely to provide were being someone with whom the index “hangs out”, being someone who spends time with the index, and as someone the index feels they can talk to about private matters, suggesting companionship defined these relationships. (9.7%)

Class 3: Low to Moderate Support, Mixed type

Supporters in this class had a very low probability of providing any types of support except being someone from whom the index could ask for advice about STDs, stay at their residence, or spend time. Given that these items represent informational, instrumental, and socialization support respectively, this group did not appear to represent a particular type of support (9.5%)

Class 4: High Support

These supporters had the greatest probability of providing any and all of the seven types of support assessed. (13.1%)

Class 5: Low/No Support

Friends in this class were unlikely to provide any of the types of support, though about a third of friends in this class would provide emergency housing. (59.5%)

3.3 Latent Class Regression

3.3.1 Bivariate analyses of network-member level factors

Friends who use drugs were less likely to be in any of the support-type classes than the Low/No Support class except the Socialization Support class (#2). Male friends, compared to female friends, were also more likely to be in the Instrumental/Financial Support class (class #1) than the Low/No Support class.

There were differences between classes in the quality of the relationship of the supporter with the index (as reported by the index). Friends with whom the index reported conflict were also less likely to be in any of the supporter classes than the Low/No Support class except the Socialization Support class (#2). Network members who the index sees daily were more likely to be in the Socialization Support class (#2) and the High Support class (#4) than the Low/No Support class. Network members whom the index has known for five or more years and friend network members who the index rated the highest level of trust for were more likely to be in any of the support classes other than the Socialization Support class (#2), compared to the Low/No Support class. Finally, network members for whom the index provides financial or material assistance were more likely to be in the High Support class (#4) than the No/Low Support class, suggesting close, reciprocal ties.

3.3.2 Bivariate analyses of index/recipient level factors

Index participants who used cocaine and/or heroin in the prior 6 months were less likely to have Instrumental/Financial support friends (#1), Low to Moderate support friends (#3), and High Support friends (#4) than Low/No support friends. Index participants with greater social status (based on gender, education, age, and homelessness) had more supportive ties: male index participants, compared to female index participants, were more likely to have class #1, #3, and #4 friends compared to No/Low support friends, index participants with at least a high school degree were more likely to report having friends in the High Support class (#4) compared to index participants without a degree, older index participants were more likely to report having class #1 and #4 friends, and index participants who were homeless were less likely to report friends in classes #1, #3, and #4, compared to the Low/No Support class.

3.3.3 Multivariable and multilevel regression analyses

Table 4 reports the estimates of a multivariable latent class regression model with variables selected based on their importance to latent class membership in bivariate analyses, evidence from bivariate analyses that they varied within class, and an interest in not having variables that were strongly collinear. The analysis was replicated with 500 random starts and 10 optimizations for each of the 500 sets of start values, as well as 20 iterations, to assure that global rather than local maxima were reached for this complex model (Muthén and Muthén, 1998–2007); this process was replicated five times and model estimates matched to the thousandth decimal place among the five replications.

Table 4.

Results of the mutlivariable multi-level regression of supporter- and recipient- level factors on class membership to supporter group among friend network members in the SHIELD study, using the best-fitting five-class model.a

Variable Class 1: Financial Class 2: Socialization Class 3: Low to Mod. Class 4: High
Supporter-Level Adj. Estimate S.E. Adj. Estimate S.E. Adj. Estimate S.E. Adj. Estimate S.E.
 Sex/Gender: Male 0.22 0.16 0.14 0.15 −0.03 0.13 −0.03 0.13
 Current heroin or cocaine user −2.10** 0.14 −0.49* 0.25 −2.10** 0.14 −2.10** 0.14
 Conflict with Index −0.79** 0.24 0.43* 0.17 −0.79** 0.24 −0.09 0.17
Index-Level
 Sex/Gender: Male 0.13 0.27 0.01 0.19 0.18 0.17 0.45** 0.14
 Any heroin or cocaine use (past six months) 0.18 0.18 −0.17 0.21 0.09 0.18 −0.04 0.16
 High school degree 0.32* 0.16 −0.06 0.18 0.09 0.15 0.29* 0.14
 Age (reference is less that 33)
  -age 33 to 42 0.28 0.26 −0.17 0.22 0.17 0.21 0.46* 0.20
  -age 43 to 67 0.62* 0.27 −0.25 0.25 0.39+ 0.23 0.58** 0.22
 Homelessness (past six month) −0.40** 0.14 0.03 0.19 −0.40** 0.14 −0.40** 0.14
Open in a new tab
a

Class 5 (Low/No Support) is the reference.

+

p < 0.10

*

0.01 < p < 0.05,

**

p < 0.01

-- not able to estimate the coefficient due to a lack of variability within class on this variable

In the multivariable model, friends who used drugs were less likely to be in any of the other support classes than the Low/No Support class after controlling for index/recipient drug use and other covariates. Those friends with whom the index has conflict were less likely to be in class #1, #2, and #3 compared to the Low/No Support class. Male index participants were more likely to report having friends in the High Support (#4) class than the Low/No Support class after controlling for friends’ gender and other covariates. Additionally, there was no relationship between index drug use and class membership after controlling for other characteristics. Index particpants who were older and not homeless were more likely to report having friends in the High Support (#4) and Financial Support (#1) classes than the Low/No Support class. Index homelessness was also negatively associated with having friends in the Low to Moderate Support (#3) class compared to the Low/No Support class in multivariable modeling. Those index participants with at least a high school degree were more likely to have friends in the High Support (#4) and Financial Support (#1) classes, controlling for other index and friend characteristics.

4. Discussion

The present study revealed a fair degree of variety in the types of supporters among friends of this inner-city sample with high rates of drug use. Furthermore, latent variable modeling indicated that there are identifiable patterns in the support provided. The classes of supportive friends described types of friends that were likely to specialize in providing particular kinds of support (e.g. financial, socialization) and other types of friends who had a consistent probability of providing all of the kinds of support measured (e.g., high support, low/no support). It is notable that, in terms of the material aid often thought of as instrumental support (House and Kahn, 1985), financial help better differentiated between strongly supportive and weakly supportive friendship ties than being able to provide a place to stay. Because the latent class with the greatest estimated prevalence among the friend network members described supporters who provided no or little support, the findings give credence to the common reports of participants that they often feel socially isolated despite reporting large social networks.

Both supporter and index drug use was associated with a greater likelihood of a supporter being in the low/no support class than any of the more supportive classes. Prior studies have not been able to determine a direct association between drug use and the specific types of supportive ties available within social networks. These findings offer confirmation that, despite their relative need for support, drug users actually have fewer social support resources of all types within their networks compared to non-drug users. Furthermore, these findings show that drug users are less likely to provide support of any type to their friends. Given the high proportion of drug users within networks and the challenges of sustaining relationships with non-drug users, this finding points to an additional source of distress among those with limited resources. Prior research has demonstrated a strong association between index and network member drug use (Bohnert et al., 2009), which heightens the implications of these findings and provides a likely explanation for the attenuated association between index-level drug use and support multivariate analysis.

Previous social support research has relied upon singular measures of support, e.g. having friends to talk with about private matters as an indication of emotional support. In this study, the No/Low Support and Socialization Support type supporters were similar in terms of recipient and index drug use, suggesting that simple measures of emotional support may be insufficient for distinguishing the supportive quality of a relationship. Nonetheless, the ability in this study to identify friends in the Socialization Support class is relevant to the implementation of network-oriented interventions that promote risk reduction between high risk friends to promote HIV risk reduction (Latkin et al., 2003a).

It is notable that friends in the High Support (#4) class differed from friends in the Low/No Support group in almost every way assessed, suggesting that friends who provide support across multiple dimensions are those who are considered to have the close relationship traditionally thought of as being part of a supportive tie. The strong distinction in index drug use and support-recipient relationships between those friends in the High Support and the Financial Support compared to the Low/No Support classes adds evidence for a multi-dimensional approach to the measurement and interpretation of social support provision and helps validate the use of these particular social support measures as part of a multi-dimensional assessment of support. Additionally, we found that membership in highly supportive classes was positively associated with relationships lasting more than five years and with a high level of trust. These may be important qualities to include in future operationalization of supportive relationships.

4.1 Limitations

The present study focused only on friendship ties and was not able to characterize individuals whose support mainly comes from family members and romantic partners. Because the study eligibility criteria required that participants were willing to engage in outreach by discussing personal matters with others, it is possible that this sample may be more socially-oriented and have larger social networks than a general population of urban-dwelling adults. Many of the participants in the present sample are long-term drug users with episodic patterns of use, and the findings may not be generalizable to individuals with a shorter drug use history. Additionally, we are unable to determine causal relationships and cannot conclude any directional association between drug use and network support types.

4.2 Clinical Implications and Future Directions

Prior research suggests an association between social relationships and substance use behaviors and treatment participation (Davey-Rothwell et al., 2008; Latkin et al., 2004; Buchanan and Latkin, 2008). Drug treatment programs often make recommendations regarding social relationships, such as avoiding drug using friends and establishing frequent interactions with supportive non-users. However, the present study is one of the first to examine how social relationships among drug users are structured and how this structure may be related to drug use. Adding select questions regarding social network members to assessment tools used in clinical settings, such as to the support and family/social sections of the Addiction Severity Index (McLellan et al., 1992), which currently focuses exclusively on instrumental support, could increase the ability of a clinician to quickly and efficiently identify specific friends who could serve as social resources for promoting drug cessation. Interventions that provide additional resources and support from non-network sources as well as strengthen existing friend relationships are essential for disadvantaged individuals with drug use problems. Because of the lack of effectiveness found in large studies based on creating support ties with strangers, strengthening the existing ties may be a more effective and lasting approach to intervention (Cohen, 2004). Specific types of supporters from among the friend network members may be identified based on the profiles defined empirically in the present study, and potentially important supporters could be engaged in treatment planning. Additionally, social skills training in and outside of treatment settings could help individual in treatment to foster specific types of supportive ties with current network members.

The findings also have public health implications. While it was not possible to understand directional relationships in this study, the present study provides some support for social disadvantage is a fundamental cause of disease (Link and Phelan, 1995), as individuals with greater disadvantage were less likely to have friends who provided high levels of support. This finding indicates that those in most need of supportive ties are often those who have the fewest.

We also found evidence that social support is multidimensional. Consequently, social programs that address social support needs may have greater impact with multidimensional approaches including avenues for enhancing financial, emotional, and informational resources. Furthermore, Using a single item to measure social support is likely insufficient for studying the association with drug use outcomes. Future research should identify which of types of supporters are most relevant to specific drug-related outcomes, including aspects of drug use reduction, cessation and treatment outcomes. Future research should also explore similar questions within other types of relationships such as kin and romantic partners.

Table 3.

Results of the bivariate multi-level regression of supporter- and recipient- level factors on class membership to supporter group among friend network members in the SHIELD study, using the best-fitting five-class model.a

Variable Class 1: Financial Class 2: Socialization Class 3: Low to Mod. Class 4: High
Supporter-Level Estimate S.E. Estimate S.E. Estimate S.E. Estimate S.E.
 Sex/Gender: Male 0.47** 0.17 −0.04 0.15 −0.24 0.14 −0.10 0.11
 Current heroin or cocaine user −2.07** 0.13 −0.43 0.30 −2.07** 0.13 −2.07** 0.13
 Conflict with Index −0.55** 0.14 0.17 0.17 −0.55** 0.14 −0.55** 0.14
 See Index daily -- -- 0.30* 0.14 -- -- 0.38** 0.11
 Known Index at least 5 years 0.54** 0.15 -- -- 0.47** 0.15 0.89** 0.12
 Is someone Index supports 0.15 0.32 0.44 0.23 −0.47 0.37 0.65** 0.18
 Index trusts with life 1.94** 0.25 -- -- 1.32** 0.23 2.35** 0.15
Index-Level
 Sex/Gender: Male 0.39* 0.19 −0.18 0.17 0.46* 0.22 0.36** 0.13
 Drug use (past six month)
  -any heroin or cocaine −0.39* 0.16 −0.12 0.18 −0.63** 0.15 −0.63** 0.13
  -injection drug use −0.10 0.15 −0.09 0.15 −0.39** 0.14 −0.32** 0.12
  -crack use only −0.75 0.43 0.53 0.28 −0.48 0.39 −0.75 0.47
  -any heroin use −0.21 0.16 −0.16 0.16 −0.54** 0.17 −0.48** 0.12
 High school degree 0.24 0.16 0.15 0.15 0.16 0.17 0.38** 0.17
 Age (reference is less that 33)
  -age 33 to 42 0.39 0.26 −0.21 0.19 0.14 0.25 0.25 0.19
  -age 43 to 67 0.63* 0.27 −0.14 0.20 0.44 0.24 0.42* 0.20
 Homelessness (past six month) −0.36** 0.11 0.07 0.16 −0.36** 0.11 −0.36** 0.11
Open in a new tab
a

Class 5 (Low/No Support) is the reference.

*

0.01 < p < 0.05,

**

p < 0.01

-- not able to estimate the coefficient due to a lack of variability within class on this variable

Footnotes

Publisher's Disclaimer: This is a PDF file of an unedited manuscript that has been accepted for publication. As a service to our customers we are providing this early version of the manuscript. The manuscript will undergo copyediting, typesetting, and review of the resulting proof before it is published in its final citable form. Please note that during the production process errors may be discovered which could affect the content, and all legal disclaimers that apply to the journal pertain.

References

  1. Agneessens F, Waege H, Lievens J. Diversity in social support by role relations: A typology. Social Networks. 2006;28:427–441. [Google Scholar]
  2. Bandeen-Roche K, Miglioretti D, Zeger SL, Rathouz P. Latent variable regression for multiple discreet outcomes. J Am Stat Assoc. 1997;92:1375–1386. [Google Scholar]
  3. Barrera M, Jr, Gottlieb BH. Social support in the adjustment of pregnant adolescents: Assessment issues. In: Gottlieb BH, editor. Social Networks and Social Support. Sage Publications; Beverly Hills, CA: 1981. pp. 69–96. [Google Scholar]
  4. Berkman LF, Melchior M, Chastang JF, Niedhammer I, Leclerc A, Goldberg M. Social integration and mortality: a prospective study of French employees of Electricity of France-Gas of France: the GAZEL Cohort. Am J Epidemiol. 2004;159 (2):167–174. doi: 10.1093/aje/kwh020. [DOI] [PubMed] [Google Scholar]
  5. Bohnert ASB, Bradshaw C, Latkin CA. A social network perspective on heroin and cocaine use among adults: Evidence of bidirectional influences. Addiction. 2009;104 (7):1210–1218. doi: 10.1111/j.1360-0443.2009.02615.x. [DOI] [PMC free article] [PubMed] [Google Scholar]
  6. Buchanan AS, Latkin CA. Drug use in the social network of heroin and cocaine users before and after drug cessation. Drug Alcohol Depend. 2008;96 (3):286–289. doi: 10.1016/j.drugalcdep.2008.03.008. [DOI] [PMC free article] [PubMed] [Google Scholar]
  7. Buckman JF, Bates ME, Cisler RA. Social networks and their influence on drinking behaviors: differences related to cognitive impairment in clients receiving alcoholism treatment. J Stud Alcohol Drugs. 2007;68 (5):738–747. doi: 10.15288/jsad.2007.68.738. [DOI] [PMC free article] [PubMed] [Google Scholar]
  8. Buckman JF, Bates ME, Morgenstern J. Social support and cognitive impairment in clients receiving treatment for alcohol- and drug-use disorders: a replication study. J Stud Alcohol Drugs. 2008;69 (5):738–747. doi: 10.15288/jsad.2008.69.738. [DOI] [PMC free article] [PubMed] [Google Scholar]
  9. Cohen S. Social relationships and health. Am Psychol. 2004;59 (8):676–684. doi: 10.1037/0003-066X.59.8.676. [DOI] [PubMed] [Google Scholar]
  10. Darke S. Self-report among injecting drug users: A review. Drug and Alcohol Depend. 1998;51 (3):267–268. doi: 10.1016/s0376-8716(98)00028-3. [DOI] [PubMed] [Google Scholar]
  11. Davey-Rothwell MA, Kuramoto SJ, Latkin CA. Social networks, norms, and 12-step group participation. Am J Drug Alcohol Abuse. 2008;34 (2):185–193. doi: 10.1080/00952990701877086. [DOI] [PubMed] [Google Scholar]
  12. Fiori KL, Antonucci TC, Cortina KS. Social network typologies and mental health among older adults. J Gerontol B Psychol Sci Soc Sci. 2006;61 (1):25–32. doi: 10.1093/geronb/61.1.p25. [DOI] [PubMed] [Google Scholar]
  13. Ghandour LA, Martins SS, Chilcoat HD. Understanding the patterns of distribution of opioid analgesic dependence symptoms using a latent empirical approach. Int J Methods Psychiatr Res. 2008;17 (2):89–103. doi: 10.1002/mpr.232. [DOI] [PMC free article] [PubMed] [Google Scholar]
  14. Granfield R, Cloud W. Social context and “natural recovery”: the role of social capital in the resolution of drug-associated problems. Subst Use Misuse. 2001;36 (11):1543–1570. doi: 10.1081/ja-100106963. [DOI] [PubMed] [Google Scholar]
  15. Hanson BS, Isacsson SO, Janzon L, Lindell SE. Social support and quitting smoking for good. Is there an association? Results from the population study, “Men born in 1914,” Malmo, Sweden. Addict Behav. 1990;15 (3):221–233. doi: 10.1016/0306-4603(90)90065-6. [DOI] [PubMed] [Google Scholar]
  16. Harrison L. The validity of self-reported drug use in survey research: an overview and critique of research methods. NIDA Res Monogr. 1997;167:17–36. [PubMed] [Google Scholar]
  17. Harrison LD, Martin SS, Enev T, Harrington D. Office of Applied Studies. Substance Abuse and Mental Health Services Administration; Rockville, MD: 2007. Comparing drug testing and self-report of drug use among youths and young adults in the general population. DHHS publication no. SMA 07-4249, methodology series M-7. [Google Scholar]
  18. House JS, Kahn RL. Measures and concepts of social support. In: Cohen S, Syme SL, editors. Social Support and Health. Academic Press; New York: 1985. [Google Scholar]
  19. Kawachi I, Berkman LF. Social ties and mental health. J Urban Health. 2001;78 (3):458–467. doi: 10.1093/jurban/78.3.458. [DOI] [PMC free article] [PubMed] [Google Scholar]
  20. Latkin CA, Hua W, Tobin K. Social network correlates of self-reported non-fatal overdose. Drug Alcohol Depend. 2004;73 (1):61–67. doi: 10.1016/j.drugalcdep.2003.09.005. [DOI] [PubMed] [Google Scholar]
  21. Latkin CA, Sherman S, Knowlton A. HIV prevention among drug users: outcome of a network-oriented peer outreach intervention. Health Psychol. 2003a;22 (4):332–339. doi: 10.1037/0278-6133.22.4.332. [DOI] [PubMed] [Google Scholar]
  22. Latkin CA, Forman V, Knowlton A, Sherman S. Norms, social networks, and HIV-related risk behaviors among urban disadvantaged drug users. Soc Sci Med. 2003b;56 (3):465–476. doi: 10.1016/s0277-9536(02)00047-3. [DOI] [PubMed] [Google Scholar]
  23. Liebler CA, Sandefur GD. Gender differences in the exchange of social support with friends, neighbors, and co-workers at midlife. Soc Sci Res. 2002;31 (3):364–391. [Google Scholar]
  24. Link BG, Phelan J. Social conditions as fundamental causes of disease. J Health Soc Behav. 1995:80–94. (Extra Issue) [PubMed] [Google Scholar]
  25. Lynskey MT, Agrawal A, Bucholz KK, Nelson EC, Madden PA, Todorov AA, Grant JD, Martin NG, Heath AC. Subtypes of illicit drug users: a latent class analysis of data from an Australian twin sample. Twin Res Hum Genet. 2006;9 (4):523–530. doi: 10.1375/183242706778024964. [DOI] [PubMed] [Google Scholar]
  26. McCutcheon AL. Latent Class Analysis. Sage; Thousand Oaks, CA: 1987. [Google Scholar]
  27. McLellan AT, Kushner H, Metzger D, Peters R, Smith I, Grissom G, Pettinati H, Argeriou M. The Fifth Edition of the Addiction Severity Index. J Subs Abuse Treat. 1992;9 (3):199–213. doi: 10.1016/0740-5472(92)90062-s. [DOI] [PubMed] [Google Scholar]
  28. McMahon RC. Personality, stress, and social support in cocaine relapse prediction. J Subst Abuse Treat. 2001;21 (2):77–87. doi: 10.1016/s0740-5472(01)00187-8. [DOI] [PubMed] [Google Scholar]
  29. Monga N, Rehm J, Fischer B, Brissette S, Bruneau J, El-Guebaly N, Noel L, Tyndall M, Wild C, Leri F, Fallu JS, Bahl S. Using latent class analysis (LCA) to analyze patterns of drug use in a population of illegal opioid users. Drug Alcohol Depend. 2007;88 (1):1–8. doi: 10.1016/j.drugalcdep.2006.08.029. [DOI] [PubMed] [Google Scholar]
  30. Muthén LK, Muthén BO. MPlus user’s guide. 5. Muthen & Muthen; Los Angeles: 1998–2007. [Google Scholar]
  31. Muthén B, Muthén LK. Integrating person-centered and variable-centered analyses: growth mixture modeling with latent trajectory classes. Alcohol Clin Exp Res. 2000;24 (6):882–891. [PubMed] [Google Scholar]
  32. Rosengren A, Orth-Gomer K, Wedel H, Wilhelmsen L. Stressful life events, social support, and mortality in men born in 1933. BMJ. 1993;307 (6912):1102–1105. doi: 10.1136/bmj.307.6912.1102. [DOI] [PMC free article] [PubMed] [Google Scholar]
  33. Scherbaum N, Specka M. Factors influencing the course of opiate addiction. Int J Methods Psychiatr Res. 2008;17 (S1):S39–S44. doi: 10.1002/mpr.244. [DOI] [PMC free article] [PubMed] [Google Scholar]
  34. Sobell LC, Ellingstad TP, Sobell MB. Natural recovery from alcohol and drug problems: methodological review of the research with suggestions for future directions. Addiction. 2000;95 (5):749–764. doi: 10.1046/j.1360-0443.2000.95574911.x. [DOI] [PubMed] [Google Scholar]
  35. VanDeMark NR. Policy on reintegration of women with histories of substance abuse: A mixed methods study of predictors of relapse and facilitators of recovery. Subst Abuse Treat Prev Policy. 2007;2:28. doi: 10.1186/1747-597X-2-28. [DOI] [PMC free article] [PubMed] [Google Scholar]
  36. Veiel HOF, Bauman U. The many meanings of social support. In: Veiel HOF, Bauman U, editors. The Meaning and Measurement of Social Support. Hemisphere Publishing Cooperation; Bristol, PA: 1992. pp. 1–9. [Google Scholar]
  37. Vermunt JK. Latent class and finite mixture models for multilevel data sets. Stat Methods Med Res. 2008;17 (1):33–51. doi: 10.1177/0962280207081238. [DOI] [PubMed] [Google Scholar]
  38. Warner LA, Alegria M, Canino G. Remission from drug dependence symptoms and drug use cessation among women drug users in puerto rico. Arch Gen Psychiatry. 2004;61 (10):1034–1041. doi: 10.1001/archpsyc.61.10.1034. [DOI] [PubMed] [Google Scholar]
  39. Wellman B, Gulia M. Where does social support come from? The social network basis of interpersonal resources for coping with stress. In: Maney A, editor. Social Conditions, Stress, Resources and Health. National Institute of Mental Health; Rockville, MD: 1997. [Google Scholar]

RESOURCES