When can get my assignment back?

Methods for Collection of Participant-aided Sociograms for the Study of Social, Sexual and Substance-using Networks Among Young Men Who Have Sex with Men Abstract In this study, we adapted and tested a participant-aided sociogram approach for the study of the social, sexual, and substance use networks of young men who have sex with men (YMSM); a population of increasing and disproportionate risk of HIV infection. We used a combination of two interviewer-administered procedures: completion of a pre- numbered list form to enumerate alters and to capture alter attributes; and a participant-aided sociogram to capture respondent report of interactions between alters on an erasable whiteboard. We followed the collection of alter interactions via the sociogram with a traditional matrix-based tie elicitation approach for a sub-sample of respondents for comparison purposes. Digital photographs of each network drawn on the whiteboard serve as the raw data for entry into a database in which group interactions are stored. Visual feedback of the network was created at the point of data entry, using NetDraw network visualization software for comparison to the network structure elicited via the sociogram. In a sample of 175 YMSM, we found this approach to be feasible and reliable, with high rates of participation among those eligible for the study and substantial agreement between the participant-aided sociogram in comparison to a traditional matrix-based approach. We believe that key strengths of this approach are the engagement and maintenance of participant attention and reduction of participant burden for alter tie elicitation. A key weakness is the challenge of entry of interview-based list form and sociogram data into the database. Our experience suggests that this approach to data collection is feasible and particularly appropriate for an adolescent and young adult population.

This builds on and advances visualization-based approaches to social net\ work data collection. Keywords: Sociogram, MSM, youth, egocentric network Authors Lisa M. Kuhns , Ann & Robert H. Lurie Children’s Hospital of Chicago, Division of Adolescent Medicine, Chicago, IL. USA and Northwestern University, Feinberg School of Medicine, Department of Pediatrics, Chicago, IL USA.

M. Birkett , Northwestern University, Feinberg School of Medicine, Department of Medical Social Sciences, Chicago, IL USA. DOI: http://dx.doi.org/10.17266/35.1.1 Lisa M. Kuhns Ann & Robert H. Lurie Children’s Hospital of Chicago & Northwestern University Chicago, IL USA M. Birkett Northwestern University Chicago, IL USA B. Mustanski Northwestern University Chicago, IL USA S.Q. Muth Quintus-ential Solutions Colorado Springs, CO USA C. Latkin Johns Hopkins Bloomberg School of Public Health Baltimore, MD USA I. Ortiz-Estes Ann & Robert H. Lurie Children’s Hospital of Chicago Chicago, IL USA R. Garofalo Ann & Robert H. Lurie Children’s Hospital of Chicago & Northwestern University Chicago, IL USA S.Q. Muth , Quintus-ential Solutions, Colorado Springs, CO USA. C. Latkin, Johns Hopkins Bloomberg School of Public Health, Department of Health, Behavior and Society, Baltimore, MD USA.

I. Ortiz-Estes , Ann & Robert H. Lurie Children’s Hospital of Chicago, Chicago, IL USA. R. Garofalo , Ann & Robert H. Lurie Children’s Hospital of Chicago & Northwestern University, Chicago, IL USA. Acknowledgements We thank Sarah Brewster and Katie Andrews for their assistance with data collection and management and Crew 450 study participants for their time and effort. The project described herein was supported by grants from the National Institute on Drug Abuse: R01DA025548 and R01DA025548-S (PIs: R. Garofalo, B. Mustanski) The content is solely the responsibility of the authors and does not necessarily represent the official views of the National Institute on Drug Abuse or the National Institutes of Health.

Correspondence concerning this work should be addressed to Lisa M. Kuhns, Ann & Robert H. Lurie Children’s Hospital of Chicago, Chicago, 225 E. Chicago Ave. #161, Chicago, IL 60611.

1. Introduction In its third decade, the HIV epidemic continues to disproportionately affect men who have sex with men (MSM), but has shifted to increasingly affect young MSM (YMSM). In the United States (US), approximately 49,000 new HIV infections occur each year, with a third of incident cases occurring in youth below the age of 30 (Centers for Disease Control and Prevention, 2013a). Between 2008-2010, male-to-male sexual contact accounted for over half of estimated new infections annually; young Black MSM aged 13–24 accounting for more new infections than any other age group or race of MSM (Centers for Disease Control and Prevention, 2013b). Despite ongoing educational efforts, YMSM continue to engage in high risk sexual behaviors placing them at-risk for HIV (Mustanski, Newcomb, Du Bois, Garcia, & Grov, 2011).

Thus, over the past two decades a shift has occurred in the study of the epidemiology of sexually transmitted infections from a focus primarily on individual risk factors to a focus which includes characteristics of social and sexual networks (Aral, 1999; Wohlfeiler & Potterat, 2005). It has been proposed that social networks form part of a web of health causation, with social- structural conditions affecting the formation of social networks, which in turn transfer their influence to health through several basic pathways: social support, social influence, access to resources, social involvement, and person-to-person contact/contagion (Berkman & Glass, 2000; Smith & Christakis, 2008). These psychosocial and behavioral processes then impact health through more proximate mechanisms, including psychological stress responses and health behaviors (Berkman & Glass, 2000). In addition, sexual network ties confer risk for STI/HIV infection through member characteristics and behaviors (Aral, 1999). Among adolescents and young adults, partner characteristics such as age discordance, previous incarceration, STI diagnosis in the past year, other partners in the past year, and problems with alcohol and drugs have been found to influence individual STI/ HIV risk (Mustanski, Newcomb, & Clerkin, 2011; Newcomb & Mustanski, 2013; Staras, Cook, & Clark, 2009). Mathematical modeling studies have found that the structural characteristics of sexual networks, including patterns of mixing (Anderson, Gupta, & Ng, 1990), concurrency (Morris & Kretzschmar, 1997), and degree (Christley et al., 2005) impact disease spread, and that position within the network influences infection/ transmission (Christley et al., 2005). While networks in which transmission takes place have a common network structure, the actual level of transmission may be best determined by studying factors specific to a population group (Rothenberg & Muth, 2007).

Given the epidemic of HIV among YMSM, social network approaches to the study of HIV risk are an important area of focus (Clatts, Goldsamt, Neaigus, & Welle, 2003), although methods for network data collection developed and tested among YMSM are limited. YMSM constitute an unbounded, hidden, and stigmatized population which present challenges for the collection of network data. While methods of network- based recruitment, such as respondent-driven sampling (RDS; Heckathorn, 1997, 2002) have been developed C June | Issue 1| Volume 35 | insna.org Methods for Participant-Aided Sociograms for stigmatized and hidden populations and have been implemented with some success to sample MSM (see for example: Iguchi et al., 2009; Ramirez-Valles, Heckathorn, Vazquez, Diaz, & Campbell, 2005; Reisner et al., 2010; Rhodes et al., 2012; Schneider, Michaels, & Bouris, 2012), RDS was developed specifically for sampling hidden populations, not for the elucidation of the immediate network environment per se (see Dennis et al., 2013 as an exception). Given the unbounded nature of the population, the few studies of YMSM networks that have been completed have used egocentric network approaches (Clerkin, Newcomb, & Mustanski, 2011; Kapadia et al., 2013). This type of network study is called an “egocentric” study because all network information is derived from respondent, or “ego” perceptions (Marsden, 1990), rather than from firsthand reports from all individuals in the network. In this paper, we describe the development and testing of a novel participant-aided sociogram approach for a study of the personal networks of YMSM, assess the feasibility of data collection using this approach, and describe personal network characteristics of the target population. We define social networks as a set of nodes (e.g., persons) linked by a set of social relationships (e.g., friendships) of a specified type (Laumann, Galaskiewicz, & Marsden, 1978). A sociogram is a diagram used to represent the relationships between individuals in a group (Moreno, 1953). The sociogram creates a “picture” in which persons are represented as points and relationships are represented by lines between these points.

The sociogram, or network diagram, has become a mainstay of social network visualization (Wasserman & Faust, 1994); however, as a key tool of social network analysis, the network diagram has primarily been created after data collection has been completed. This is due in part to the focus in much of network analysis on whole networks, i.e., mapping networks of connections among members of entire populations (Hogan, Carrasco, & Wellman, 2007).

More recently, methods have developed for bringing the sociogram into the data collection process for collection of personal egocentric network data, i.e., in which the respondent (or “ego”) provides their direct contacts (or “alters”) \ as well as relationships between those alters (Hogan et al., 2007; Kennedy, Tucker, Green, Golinelli, & Ewing, 2012; Tucker et al., 2012). This process has advantages, including a high level of participant engagement in the network elicitation process, the immediate visual feedback, and verification of network structure by the respondent and the acceleration of the process of identifying ties between alters (Hogan et al., 2007).

Traditional approaches to elicit network connections require that the ego identify connections between each pair of alters, which is both labor-intensive for interviewers and tedious for respondents. Hogan and colleagues (2007) for example, developed a method to structure the network data as it is being collected in the form of real-time visualization. In their study of communication media and its impact on personal networks in Canada, a detailed network interview was completed with adults in which network members (i.e., somewhat to very close ties) were generated, names transcribed onto “Post-it” notes, and then arranged on a large sheet of paper. Alter ties were elicited by drawing circles around alters rather than completing all possible pairs. They found that this low-technology and interactive method improved interview quality, and was more time and cost-efficient than traditional matrix-based approaches that seek to capture all alter pairs (Hogan et al., 2007). Participant-aided visualization techniques have also been used among adolescent and young adult populations, specifically homeless youth at risk of HIV infection, in which network characteristics and structure are particularly salient (Rice, Barman-Achikari, Milburn, & Monro, 2012). Among homeless youth seeking services at shelters, drop-in centers, and street venues, Kennedy and colleagues (2012) completed network interviews in which names of social network contacts (i.e., “individual they knew, who knew them, and with whom they had contact during the past year or so”) were generated, alter characteristics were elicited, and then a traditional pair-based approach was used to elicit alter ties. Sexual partners were identified from among network members and questions about sexual risk were elicited about each partner. Visualizations of personal networks were then produced immediately using visualization software, and participants were asked to describe distinct clusters of alters (components) as well as alters with no connections (isolates) based on these visualizations. Composite indicators of network structure developed from these techniques were then included in multi-level models to analyze their impact on sexual risk (Kennedy et al., 2012).

Tucker and colleagues used a similar name generator in a study of homeless YMSM more specifically, i.e., “name people that you know and who know you and that you had contact with in the past 3 months” to generate a list of network members and elicited the perceived risk behaviors of alters (including sexual risk behavior and alcohol and illicit drug use), however no alter-to-alter ties were elicited, therefore analyses of network structure were limited (Tucker et al., 2012).

Thus, while participant-aided visualization techniques have been developed for health-related C insna.org | Volume 35 | Issue 1 | June Methods for Participant-Aided Sociograms research and utilized with young populations to measure network structure, some evidence suggests that they type of visualization method used is important. For example, McCarthy and colleagues (2007), among a small sample of adults, compared a freestyle network drawing technique such as that used in the Hogan study (2007) to a visualization approach based on a matrix-based elicitation of ties. Participants were instructed to first name forty-five alters with the instructions: “you know them and they know you, by sight or by name. You have had some contact with them in the past two years – by phone face-to-face, email, mail – and you could contact them again if necessary.” Respondents were instructed to create a representation of their network by drawing “social circles” and labeling them to describe each group.

Respondents then returned no later than a week later to complete a matrix of pairwise ties, which was mapped using visualization software. Study investigators found that the freestyle drawing often produced fewer and/or more homogenous ties than the visualization based on pairwise ties (McCarty, Molina, Aguilar, & Rota, 2007).

The investigators concluded that cognitive information on social network ties is not stored randomly, but arranged in patterns that aid recall and which correspond to social structure. Because this “shorthand” method of visualization may result in a consolidation of ties and thus a potential loss of data, these results call for further comparison of the two approaches in other populations and within larger samples.

2. Methods 2.1 Participants In the study described herein, we measured network structure in the context of an ongoing longitudinal study of a syndemic of health issues facing YMSM in Chicago: the nation’s third largest city and the epicenter of the HIV/AIDS epidemic in the Midwest. The purpose of the parent study is to characterize the prevalence, course, and predictors of a syndemic of health problems among YMSM. This syndemic includes substance abuse, experiences of violence, sexual risk taking, and internalizing mental health problems, which increases risk for HIV/STIs. We used a modified form of RDS to enroll YMSM between ages 16 and 20 in the parent study (Kuhns et al., 2014).

Participants for the network sub-study were recruited from the parent study at either the 12- or 24-month follow-up visits (i.e., at the 24-month visit if the 12-month visit had already been completed during the period of sub-study) from June 2011-October 2012. We chose these time points for strategic reasons: 1) to allow for sufficient prior interaction/trust-building between the respondent and the research team to facilitate collection of sensitive network-based information and 2) to coincide with HIV and STI testing, which is completed at baseline and 12-month intervals thereafter. The target sample size for the study (N=175) was determined based on an a priori power analysis linked to analysis goals (i.e., for multilevel analyses, see Birkett, Kuhns, Latkin, Muth, & Mustanski, in press; Mustanski, Birkett, Kuhns, Latkin, & Muth, 2014).

2.2 Procedures Because studying whole networks of YMSM would be impractical given limited resources (Latkin, Forman, Knowlton, & Sherman, 2003; Potterat, Woodhouse, Muth, & et al., 2004) we chose an egocentric or personal network data collection approach, gathering secondhand information about the immediate network environment from the respondents’ viewpoint. Similar to prior participant-aided visualization approaches described above, our network data collection procedures included three processes to elicit network-based information. This included asking the respondent to enumerate all persons with whom they have a certain type of connection, to describe characteristics of those individuals, and to describe social, substance use and sexual connections between these individuals. We used a combination of two interviewer-administered procedures to collect these data:

1) completion of a pre-numbered list form (i.e., in paper- and-pencil format) to enumerate alters and to capture alter attributes; and 2) a participant-aided sociogram to capture respondent report of interactions between alters (sexual, substance-using, and social networks).

2.3 Measures Name Generators . The enumeration of alters using name generators has been used extensively in network-based approaches (Wasserman & Faust, 1994) and has been found to be reliable for reports of individuals with whom the ego has repeated and salient interactions (Freeman, Romney, & Freeman, 1987). Our name generators and alter characteristic elicitation questions and procedures were based on prior studies of populations at risk of HIV infection (Auerswald, Muth, Brown, Padian, & Ellen, 2006; Latkin & Knowlton, 2005; Potterat, Rothenberg, & Muth, 1999; Potterat et al., 2004; Rothenberg, Baldwin, Trotter, & Muth, 2001). Egos were asked to elicit up to 40 alters on a pre-numbered list form, and provide sufficient identifying characteristics for tracking purposes (e.g., full C June | Issue 1| Volume 35 | insna.org Methods for Participant-Aided Sociograms names or failing that, nicknames or initials). We used a set of name generators to elicit social network members who provide emotional and/or instrumental support, including support for sexual minority-related issues: a. Name the people you are closest to, that is, people you see or talk to regularly and share your personal thoughts and feelings with. b. Can you think of other people who would give time and energy to help you? c. Can you think of other people who you could count on to lend or give you $25 or something of equal or greater value? d. Can you think of other people who you could turn to for help or advice about gay-related issues or problems (for example, if you were being harassed)? e. Can you think of other people you spend time with on a regular basis yet are not very close to you? We also asked whether or not the respondent had used substances with or had sex with any additional individuals not listed, or if they knew of other individuals who had used substances or had sex with two or more network members in order to identify additional substance-using and sexual partners not in the respondents’ social network (i.e., to elicit more complete substance-using and sexual networks) and to identify individuals in networks that might be in central positions in terms of sex and substance use behavior, but only weakly tied to the ego.

Name Interpreter. We then used a structured name interpreter to elicit demographic and behavioral characteristics of alters. Because signaling to respondents that large amounts of data would be collected on each alter might dampen enthusiasm for providing a complete list of names, we separated the name generator (on page 1 of the list form) from the questions on alter characteristics and relationships (pages 2-4 of the list form). This also facilitated a higher degree of confidentiality; we decoupled alter names from characteristics and securely stored them separately. Alter characteristics and relationships between the ego and alters collected in the name interpreter included: frequency of communication within the last 6 months (0=none to 5=daily), strength of the relationship (1=very close, 2=somewhat close, 3=not at all close), type of relationship (e.g., family member, friend, co-worker), estimated age (in years), race, gender, sexual orientation, and residential location (i.e., nearest cross-streets). We pilot tested use of the list form in multiple modalities and ultimately determined that collecting characteristics alter-wise, rather than question- wise, was more efficient.

In addition to demographic items, we included an additional set of questions regarding sexual and substance-using behavior between the ego and each alter, including: whether or not drugs or alcohol were ever used, frequency of drug or alcohol use in the last 6 months (0=never, 1=1-2 times, 2=once/month or less, 3=2-3 times/month, 4=1-2 times/week, 5=3-5 times/ week, 6=every day/almost every day), substances used in the last 6 months (list of 17 substances including alcohol), sexual contact at any time in the past, date of first/last sex, frequency by type of sex in the last 6 months (i.e., oral, vaginal, anal; 0=never, 1=1-2 times, 2=once/month or less, 3=2-3 times/month, 4=1-2 times/week, 5=3-5 times/ week, 6=every day/almost every day) and frequency of condom use in the last 6 months (1=always, 2=more than half the time, 3=about half the time, 4=less than half the time, 5=never).

Sociogram . After the name generator and name interpreters were completed, each respondent completed a sociogram to identify social, sexual, and substance- using connections between alters. Whereas Hogan and colleagues used a paper sociogram form with “post-it” notes to indicate alters, we chose a slightly more durable Figure 1: Sample social network drawing (left) and as converted to Netdraw figure (right) *Participant and ego ID and alter initials were changed to protect participant confidentiality. Colors and arrows in participant drawing were used only to distinguish connections between alters (i.e., to facilitate data entry, not to depict different types of connections or directionality).

C insna.org | Volume 35 | Issue 1 | June Methods for Participant-Aided Sociograms approach using erasable, writable magnets, and an erasable whiteboard (see Figures 1-3). We made this modification for our young target population because we anticipated more movement of alters around the whiteboard until the respondent settled on positions, and thus a need for more durable materials (i.e., we anticipated post-it notes losing their stickiness with repeated movement). Alter first names and last initial were transferred to numbered “write-on” magnets (i.e., pre-numbered with the corresponding alter identification number on the list form) and placed these on a magnetic whiteboard. Following Hogan’s method, we instructed respondents to arrange alters such that those who know each other were placed together, to draw a circle around any group of three or more alters who know each other well (to identify cliques) and to draw a line between any two alters (outside of a clique) who know each other well (to identify dyads). A digital photograph of this social network sociogram was taken (and verified using the view function), the white board was erased, and this process was completed for each of the other networks of interest – the substance-using and sexual networks. In these cases, respondents were directed to draw circles around respondents and lines between respondents who were known to the respondent to have used substances together or had sexual relations.

Database . Data from the name generator, name interpreter, and sociograms were then entered into a database programmed using Visual Basic in Microsoft Access. Groups of interactions depicted on the sociogram (both cliques and dyads) were directly entered into this database, which was programmed to automatically convert the groups to dyadic lists. The database was formatted such that the data entry tabs matched the name generator and name interpreter forms to facilitate data entry. Once names from the list form were initially entered, the remaining database forms were pre-populated with initials to facilitate data entry. Finally, instantaneous visual feedback of the network was provided at the point of data entry, with custom Visual Basic programming creating and launching NetDraw VNA files (Borgatti, 2002) directly from the database, facilitating verification. Assessment of reliability. Given prior evidence suggesting more comprehensive data elicited via matrix-based visualization versus the freestyle drawing approach cited above, we sought to test the reliability of a freestyle method of tie elicitation in comparison to the matrix-based elicitation. The freestyle drawing approach reflects, in effect, a “shortcut” to tie elicitation:

a respondent might forget or otherwise fail to describe ties between individuals when not specifically asked Figure 2: Sample sexual network drawing (left) and as converted to Netdraw figure (right)* *Participant and ego ID and alter initials were changed to protect participant confidentiality. Colors in participant drawing were used only to distinguish connections between alters (i.e., to facilitate data entry, not to depict different types of connections). Figure 3: Sample substance using network drawing (left) and as converted to Netdraw figure (right)* *Participant and ego ID and alter initials were changed to protect participant confidentiality. Colors in participant drawing were used only to distinguish connections between alters (i.e., to facilitate data entry, not to depict different types of connections or directionality).

C June | Issue 1| Volume 35 | insna.org Methods for Participant-Aided Sociograms about them. From a classic psychometric theory approach to measurement, measures are reliable to the extent that they are repeatable and free from error when persons, instruments, and conditions vary (Nunnally & Bernstein, 1994). In this case, we were interested in the consistency of alter-to-alter tie information when elicited using two different instruments – the participant-aided sociogram method and a traditional matrix-based approach. Thus, in a subsample of 52 participants, we assessed the reliability of reported social ties through the sociogram method in comparison to a matrix-based approach. Following initial data collection, we instructed participants to take a 5-10 minute break; we then elicited ties (between alters) using pairwise comparison from a random sample of 5 alters taken from the respondent’s list form. We asked respondents to indicate whether each pair of alters was socially connected, i.e., if they knew each other well, if they had a sexual relationship and if they had used substances together. We compared agreement between the two approaches using a coefficient of agreement, Cohen’s kappa (Cohen, 1960), calculated for each respondent with a mean value calculated across all respondents for each type of network.

3. Results Study enrollment was completed between June of 2011 and October of 2012. The study protocol was approved by the Institutional Review Boards at each site and respondents were provided with $25 for participation. A total of 204 parent study participants were approached for participation, and 179 (88%) agreed to participate, however, two participants did not show up for scheduled appointments and two were enrolled, but subsequently withdrew. The final sample size was 175 (86%). The reasons cited for non-participation included both logistical issues, e.g., no longer living in the area and scheduling conflicts, but also included lack of interest.

The two participants who withdrew cited privacy concerns as their reasons for withdrawal and asked to have any network data destroyed. The demographics of enrollees are parallel to those of the parent study in terms of age, race, and sexual orientation. Participants had a mean age of 20.1 (SD=1.4; range= 17.1 to 23.0) and were majority African American (53.7%) and gay-identified (82.9% identified as completely/mostly gay). Interviews lasted approximately 55 minutes on average (SD=21.5 minutes; range=19-129 minutes).

3.1 Name Generator and Alter Distribution The name generator elicited a total of 2,579 core social network alters within 175 personal networks, with an average of 14.7 alters (SD=8.4; range=3-40) reported per participant (76.1% of all alters named; see Table 1). Asking participants to list any additional network members with whom they had sex or used substances (but had not yet named) resulted in generation of 689 (20.3%) more names in the network. When asked about network members not directly tied to the ego, but who had sex or used substances with at least two network alters resulted in the generation of 122 (3.6%) more names (see Figure 4 for a visualization of these ego networks with tie type differentiated by node size). Only one respondent listed the maximum of 40 alters. The initial name generator, “Name the people you are closest to, that is, people you see or talk to regularly and share your personal thoughts and feelings with,” elicited the greatest number of responses, 1,008 alters (39.1%), followed by the second question (i.e., Can you think of other people who would give time and energy to help you?”), which generated 645 alters (25.0%). The name generator specific to gay-related support (i.e., Can you think of other people who you could turn to for help or advice about gay-related issues or problems, for example, if you were being harassed?) resulted in 262 additional alters (10.2%). Respondents identified social network alters largely by both first and last name (93.1%), with the next highest percentage being first name only (5.0%). All participants reported at least a first, last, or nickname for social network alters.

Respondents’ social network alters included primarily friends (60.0%), followed by family members (23.8%) and those identified as “acquaintance/associate” (4.7%). The strength of relationships was reported to be ‘very close’ to ‘somewhat close’ on average (i.e., mean=1.7; SD=0.7; range 1=very close, 3=not at all close). Respondents communicated with members of their networks (in the last 6 months) an average of weekly to a couple of times a monthly (i.e., mean=3.74; SD=1.3; range 0=none, 5=daily).

3.2 Alter Characteristics In terms of age, social network alters were largely in the 19-22 year old age group (46.3%), followed by 23-29 (17.9%); very few alters were 18 or younger (14.4%; see Table 2). By race, alters were primarily African American (45.0%), followed by White (23.8%) and Latino (22.1%).

Racial homophily across the core networks was 83.0% among African American respondents, 73.0% among Whites and 67.9% among Latinos. Respondents named alters of both sexes (50.7% male) and in terms of sexual orientation, gay (36.8%) or heterosexual (49.7%) alters.

Residential location data was provided for the vast C insna.org | Volume 35 | Issue 1 | June Methods for Participant-Aided Sociograms Figure 4 : Ego = dark gray color, node size 4, Social core = node size 3, Sex/substances = node size 2, \ Sex/substances alters only = node size 1, NB: colors used to differentiate ego networks.

C June | Issue 1| Volume 35 | insna.org Methods for Participant-Aided Sociograms Table 1 : Name Generator, Alter Distribution, and Relationship Characteristics Among 175 Ego-Networks of Young Men Who Have Sex with Men Entire Sample N % Core Network 2579 76.1% Substance-use or Sex only ties 689 20.3% Substance-use or Sex only ties to ≥2 alters (not ego) 122 3.6% Core Network Only N % Name generator:

Closest to/talk to/share feelings 1008 39.1% Time/energy to help 645 25.0% Lend or give you $25 339 13.1% Help for gay-related issues 262 10.2% Spend time with, not as close 325 12.6% Specificity of Name Information:

First and last names 2401 93.1% First name and first initial of last name 22 0.9% First name only 129 5.0% Nickname only 27 1.0% No name-based information 0 0.0% Type of relationship:

Immediate family/relative 614 23.8% Friend 1547 60.0% School personnel 37 1.4% Minister/church official 3 0.1% Community organization staff 18 0.7% Co-worker 59 2.3% Acquaintance/associate 121 4.7% Other 180 7.0% Strength of relationship (mean) 17 (1) Very close 1249 48.5% (2) Somewhat close 967 37.5% (3) Not at all close 361 14.0% Frequency of Communication, last 6 months (mean) 3.7 (0) Not at all 36 1.4% (1) Once or twice 170 6.6% (2) Three to six times 214 8.3% (3) At least a couple of times a month 507 19.7% (4) Weekly 825 32.0% (5) Daily 826 32.0% 3.4 Reliability of Sociogram vs. Matrix Elicitation of Alters In our reliability analysis, we found that the coefficient of agreement varied for social network ties in comparison to sex and substance use ties, but all estimates were substantial or nearly perfect according to common criteria (Landis & Koch, 1977), with overlapping confidence intervals for the social and substance use network reliability, but significantly different reliability for the sexual networks. The overall kappa for all ties was 0.74 (95% CI: 0.69, 0.80), whereas for social network ties it was 0.73 (95% CI: 0.66, 0.80), for substance use ties it was 0.68 (95% CI: 0.58, 0.77), and for sex ties it was 0.95 (0.86, 1.0). Reliability was not related to size of any of the networks (i.e., social, substance use or sexual) or overall network size (r=.039, p=.74). Of note, there are two ways our participant-aided sociogram approach could produce results different from the gold standard matrix-based approach: by producing false negatives or false positives. Our method was slightly more likely to produce false negatives than false positives, i.e., among 1,560 observations (i.e., in the sub-sample in which comparison between methods was analyzed); a failure to report a tie (i.e., in comparison to matrix-based elicitation) was observed in 46 cases and in 42 cases our method resulted in a reported link not reported in the matrix approach. Most false positives were reported among social network ties, whereas most false negatives were reported among substance-use ties.

majority of alters (98.9%). A total of 63.4% of alters were reported to live in Chicago and provided data sufficient to determine the community area within the city. Alters living in the city were approximately evenly split in the predominant areas with 35% on the north side, 26.2% on the south side and 35.5% on the west side of the city with very few residing in the city center (3.3%).

3.3 Sex and Substance-Using Behavior with Alters Respondents reporting using substances on average approximately 1-2 times in the last 6 months with 1,914 alters (56.5% of all ties, including sex/drug only and weak ties). The most frequently used substances included alcohol (41.2%), followed by marijuana (22.5%). In terms of sexual behavior, respondents reported having sex with 837 alters (24.7% of all ties, including sex/drug only and weak ties); most frequently male partners (92%). Only 56.0% reported always using a condom during anal sex with male partners in the prior 6 months.

C insna.org | Volume 35 | Issue 1 | June Methods for Participant-Aided Sociograms Table 2. Core Alter Characteristics Among 175 Ego-networks of Young Men Who Have Sex with Men N % Age (Mean; Range) 25.64 4 - 84 ≤15 28 1.1% 16-18 344 13.3% 19-22 1194 46.3% 23-29 462 17.9% 30-39 216 8.4% ≥40 310 12.0% Not reported or Unknown 25 1.0% Race African American 1161 45.0% Latino 569 22.1% White 614 23.8% Other 229 8.9% Not reported or Unknown 6 0.2% Sex Male 1307 50.7% Female 1222 47.4% Transgender 48 1.9% Not reported or Unknown 2 0.1% Sexual Orientation Bisexual 271 10.5% Gay 950 36.8% Queer 35 1.4% Heterosexual 1283 49.7% Not reported or Unknown 40 1.6% Residential Location Chicago 1634 63.4% North Side 572 35.0% Center 54 3.3% South Side 428 26.2% West Side 580 35.5% Illinois (including Chicago) 2306 89.4% Out-of-State 245 9.5% Not Reported or Unknown 28 1.1% 4. Discussion In this study, we sought to adapt and test a participant- aided sociogram approach for the study of the social, sexual, and substance use networks of YMSM; to assess the feasibility of data collection using this approach; and to describe personal network characteristics of the target population. In terms of feasibility, we found that potential participants were largely interested in participation (86% agreed to participate and completed interviews) and able to report first and last names of alters (93.1%) as well as sensitive behavior within networks, including substance use and sexual behavior. Only two participants who enrolled in the study later withdrew due to privacy concerns. It should be noted, however, that study participants were already enrolled in the larger parent study and were familiar with the study staff, which likely helped to overcome concerns about providing sensitive information about other individuals in their network.

We found the data collected via the sociogram approach to be of substantial reliability in comparison to a traditional matrix based approach, with an overall coefficient of agreement of .74 for all networks (Landis & Koch, 1977). Reporting of substance use ties in networks was similarly reliable at kappa =.68, while sexual tie reporting was nearly perfect at kappa=.95. The high reliability of sexual ties versus other types of ties may be due to the clear connection reflected in sexual relations versus the more “fuzzy” connection reflected in social ties. The nearly perfect reliability of sexual ties also suggests participants have good memory for reporting sex between alters in their network, and as such they may have a relatively high degree of certainty about these sexual interactions in which they were not directly involved. Similarly to McCarty and colleagues (2007), we found that the freestyle method of eliciting ties in our participant-aided sociogram method resulted in “false negatives,” that is, the failure to report a tie in comparison to the matrix-based elicitation; however this was fairly rare (i.e., 46 of 1,560 observations or 3%).

We also found that the sociogram approach resulted in “false positives,” that is when a tie was reported via the sociogram, but not in the matrix-based approach (i.e., 42 of 1,560 observations, or 2.7%). The false positives were more common for social ties whereas the false negatives were more common for substance use ties. In this case, the pattern may be less the result of social structural patterns (i.e., as McCarty and colleagues concluded with regard to their sample) and more the result of the relative stigma associated with substance use versus social support, resulting of under-reporting of substance use ties and over-reporting of social ties in a freestyle approach. The overall excellent reliability across types of ties suggests that the more efficient and less burdensome sociogram approach for measuring network structure is a viable alternative to asking participants to report on all ties between alters, especially in light of our recent advances in automating the data conversions to visualization software.

In terms of network size and distribution of alters, social networks were comprised of approximately 15 alters on average, although up to 40 could be listed (only C June | Issue 1| Volume 35 | insna.org Methods for Participant-Aided Sociograms one participant listed all 40), suggesting that limiting social network member names generated to 20, as was done in prior network studies of homeless youth (Kennedy et al., 2012; Tucker et al., 2012) may be a viable option and would further economize the interview process.

Approximately 10% of alter names were generated by using a gay-specific name generator, which suggests that a tailored approach to name generators may be important for this population, in particular. Social networks included primarily friends and to a lesser extent family members to whom respondents reported to be quite close emotionally (i.e., very close to somewhat close on average), but with whom they communicated with moderate frequency (i.e., weekly to a couple of times per month on average).

In terms of age and race/ethnicity, alters mirrored egos to a large extent. The majority of social network alters were ages 19-29, with very few alters ages 18 and under. This is an important finding and suggests that many school-age YMSM, i.e., those under age 19 in particular, may be isolated from networks of young adult MSM. This also suggests that studies seeking to recruit adolescent MSM should not rely on young adults for network-based recruitment as there appears to be little bridging across these two developmental groups. Racial homophily was high among all youth with the greatest degree of homophily among African American youth. In terms of gender, approximately half of alters were male and just under half were heterosexual, although a large percentage were reported to be gay (36.8%), as might be expected in a study of YMSM. Given the location of gay ghettos and gay-friendly neighborhoods on Chicago’s north side, the fact that alters came from all areas of the city suggests diverse geographic distribution of networks, beyond the most obvious neighborhoods, despite the high level of racial homophily among African America youth in particular.

Substance use between ego and alters was relatively common with egos reporting use of substances with over half of alters, although those substances were largely restricted to alcohol and marijuana and frequency of use was relatively low (i.e., 1-2 times per month on average). Sexual risk behavior between egos and alters was also common, with over half of respondents indicating that they did not use condoms with anal sex partners in the prior 6 months (see Birkett et al., in press and Mustanski et al., 2014 for further description of network factors related to sexual risk in the sample). It is important to note that close to an additional 4 substance- using and sexual ties were identified outside of the social network, highlighting the lack of complete overlap in these sets of ties.

While relative size of personal networks and their composition was not particularly surprising, we are not aware of similar studies that have attempted to estimate the size and characteristics of these networks among YMSM. An additional strength of this study and an extension of prior work is the efficient collection of information on the characteristics and structure of multiple networks (social, substance use and sexual) via this participant-aided visualization approach. While such analyses are beyond the scope of this paper, these data will allow for future analyses of the role of overlapping components and ties on health outcomes of interest.

It is our hope that this study will serve as a reference and resource upon which to build future personal network studies. Our experience demonstrates not only the feasibility of this approach for YMSM, but also the ability to collect large amounts of network data on a population at high risk of HIV infection. While we did not measure satisfaction with the research process, data collection staff reported spontaneous feedback from participants indicating a high level of interest and engagement in the data collection process, particularly with the personal network elicitation on the whiteboard.

In addition, staff found this approach straightforward and easy to implement and enjoyed the dynamic interaction with participants. Our experience with this population is consistent with the developmental literature which suggests that the ability to focus, particularly for long periods of time and the ability for complex thinking is still developing (Hunter & Sparrow, 2012), thus methods to make network data collection efficient, while maintaining validity and reliability, are quite important.

In terms of limitations, all network data was collected in interviewer-administered format, therefore social desirability bias may have been a factor in reporting, particularly reporting of sensitive behaviors. In addition, information on network alters was collected via report by the ego and thus is subject to potential errors associated with proxy reporting. While we sought to maintain the low technology approach of Hogan and colleagues (2007) to increase the generalizability of this method to low resource environments, we recognize that the use of whiteboards may not be easily portable. Therefore, this method, while quite appropriate for remote locations, may not be generalizable when portability is called for in field-based work. Data entry and management was perhaps the most challenging aspect, as this approach relied on paper and whiteboard capture initially, with subsequent computer data entry and verification. In future research on participant-aided visualization approaches, investigators should consider assessing low technology approaches such as those described here in comparison to data capture and visual feedback via electronic C insna.org | Volume 35 | Issue 1 | June Methods for Participant-Aided Sociograms devices and software such as EgoNet (McCarty, 2011) and VennMaker (Kronenwett & Duwaerts, 2014), which automate electronic data collection and visualization, for feasibility and satisfaction as well as reliability and validity, particularly in low resource settings where HIV- related research is often conducted. References Anderson, R.M., Gupta, S., & Ng, W. (1990). The significance of sexual partner contact networks for the transmission dynamics of HIV Journal of Acquired Immunodeficiency Syndrome , 3(4), 417–429. Aral, S. (1999). Sexual network patterns as determinants of STD rates: Paradigm shift in the behavioral epidemiology of STDs dade visible. Sexually Transmitted Diseases , 26(5), 262-264. Auerswald, C.L., Muth, S.Q., Brown, B., Padian, N., & Ellen, J.M. (2006). Does partner selection contribute to sex differences in sexually transmitted infection rates among African American adolescents in San Francisco. Sexually Transmitted Diseases , 33(8), 480-484. Berkman, LF, & Glass, T. (2000). Social integration, social networks, social support, and health.

In L. Berkman & I. Kawachi (Eds.), Social Epidemiology . New York: Oxford University Press. Birkett, M., Kuhns, L. M., Latkin, C., Muth, S.Q., & Mustanski, B. (in press). The sexual networks of racially diverse young men who have sex with men. Archives of Sexual Behavior . Borgatti, S.P. (2002). NetDraw software for network visualization. Lexington, KY: Analytic Technologies. Centers for Disease Control and Prevention. (2013a). HIV surveillance report 2011 (Vol. 23). Atlanta, GA: Centers for Disease Control and Prevention. Centers for Disease Control and Prevention. (2013b). HIV among Black/African American gay, bisexual and other men who have sex with men Fact Sheet . Christley, R.M., Pinchbeck, G.L., Bowers, R.G., Clancy, D., French, N.P., Bennett, R., & Turner, J. (2005).

Infection in social networks: Using network analysis to identify high-risk individuals.

American Journal of Epidemiology , 162(10), 1024–1031. Clatts, Michael C., Goldsamt, Lloyd, Neaigus, Alan, & Welle, Dorinda L. (2003). The social course of drug injection and sexual activity among YMSM and other high-risk youth: an agenda for future research. Journal of Urban Health , 80(4 Suppl 3), iii26-39. Clerkin, E. M., Newcomb, M. E., & Mustanski, B. (2011). Unpacking the racial disparity in HIV rates: the effect of race on risky sexual behavior among Black young men who have sex with men (YMSM). Journal of Behavioral Medicine , 34(4), 237-243. doi: 10.1007/s10865-010-9306- 4 Cohen, J. (1960). A coefficient of agreement for nominal scales. Educational and Psychological Measurement , 20(1), 37–46. Dennis, A. M., Murillo, W., Hernandez, F. D. M., Guardado, M. E., Nieto, A. I., Lorenzana de Rivera, I., . . . Paz-Bailey, G. (2013). Social network-based recruitment successfully reveals HIV-1 transmission networks among high-risk individuals in El Salvador. Journal of Acquired Immune Deficiency Syndromes , 63(1), 135-141. Freeman, L.C., Romney, A.K., & Freeman, S.C. (1987). Cognitive structure and informant accuracy.

American Anthropologist , 89(2), 310-325. Heckathorn, D.D. (1997). Respondent-driven sampling: A new approach to the study of hidden populations.

Social Problems , 44(2), 174-199. Heckathorn, D.D. (2002). Respondent-driven sampling II: Deriving valid population estimates from chain-referral samples of hidden populations.

Social Problems , 49(1), 11-34. Hogan, B., Carrasco, J.A., & Wellman, B. (2007). Visualizing Personal Networks: Working with Participant-aided Sociograms. Field Methods , 19(2), 116-144. Hunter, Scott J., & Sparrow, Elizabeth P. (2012). Executive function and disfunction: Identification, Assessment and Treatment : Cambridge University Press. Iguchi, M. Y., Ober, A. J., Berry, S. H., Fain, T., Heckathorn, D. D., Gorbach, P. M., . . . Zule, W. A. (2009).

Simultaneous recruitment of drug users and men who have sex with men in the United States and Russia using respondent-driven sampling:

Sampling methods and implications. Journal of Urban Health, 86, S5-S31. doi: 10.1007/s11524- 009-9365-4 Kapadia, F., Siconolfi, D. E., Barton, S., Olivieri, B., Lombardo, L., & Halkitis, P. N. (2013). Social support network characteristics and sexual risk taking among a racially/ethnically diverse sample of young, urban men who have sex with men.

AIDS Behav , 17(5), 1819-1828. doi: 10.1007/ C June | Issue 1| Volume 35 | insna.org Methods for Participant-Aided Sociograms s10461-013-0468-2 Kennedy, DP, Tucker, JS, Green, HD, Golinelli, D, & Ewing, B. (2012). Unprotected sex of homeless youth: Results form a multilevel analysis of individual, social network, and relationship factors. AIDS Behav , 16(7), 2015-2032. Kronenwett, M., & Duwaerts, T-S. (2014). VennMaker (Version 15) [Visual network mapping software]:

Kronenwett & Adolphs - Tools and Services for Social Network Analysis. Retrieved from www.

vennmaker.com Kuhns, L. M., Kwon, S., Ryan, D. T., Garofalo, R., Phillips, G., 2nd, & Mustanski, B. S. (2014).

Evaluation of respondent-driven sampling in a study of urban young men who have sex with men. J Urban Health , 92(1), 151-167. Landis, J. R., & Koch, G. G. (1977). The measurement of observer agreement for categorical data.

Biometrics , 33(1), 159-174. Latkin, C.A., & Knowlton, A. R. (2005). Micro-social structural approaches to HIV prevention: a social ecological perspective. AIDS Care , 17(Supplement 1), S102-S113. Latkin, CA, Forman, V, Knowlton, A, & Sherman, S. (2003). Norms, social networks, and HIV-related risk behaviors among urban disadvantaged drug users. Social Science & Medicine , 56(3), 456- 476. Laumann, E.O., Galaskiewicz, J., & Marsden, P.V. (1978). Community structure as Interorganizational linkages. Annual Review of Sociology , 4, 455- 484. Marsden, PV. (1990). Network data and measurement. Annual Review of Sociology , 16, 435-463. McCarty, C. (2011). EgoNet (Version GPLv2) [Visual network mapping software]: University of Florida. Software. Retrieved from www.

sourceforce.net McCarty, C., Molina, J.L., Aguilar, C., & Rota, L. (2007). A comparison of social network mapping and personal network visualization. Field Methods, 19(2), 145-162. Moreno, JL. (1953). Who shall survive? : Foundations of sociometry, group psychotherapy, and sociodrama (2nd ed.). Beacon, NY: Beacon House. Morris, M., & Kretzschmar, M. (1997). Concurrent partnerships and the spread of HIV. AIDS , 11(5), 641-648. Mustanski, B., Birkett, M., Kuhns, L. M., Latkin, C., & Muth, S.Q. (2014). The role of geographic and network factors in racial disparities in HIV among young men who have sex with men: An egocentric network study. AIDS and Behavior . Advance online publication. Mustanski, B., Newcomb, M. E., & Clerkin, E. M. (2011). Relationship characteristics and sexual risk-taking in young men who have sex with men. Health Psychology , 30(5), 597-605. Mustanski, B.S., Newcomb, M. E., Du Bois, S.N., Garcia, S.C., & Grov, C. (2011). HIV in young men who have sex with men: A review of epidemiology, risk and protective factors, and interventions.

Journal of Sex Research , 48(2), 218-253. Newcomb, M. E., & Mustanski, B. (2013). Racial differences in same-race partnering and the effects of sexual partnership characteristics on HIV risk in MSM: A prospective sexual diary study. Journal of Acquired Immune Deficiency Syndromes , 62(3), 329-333. Nunnally, JC, & Bernstein, IH. (1994). Psychometric theory (3rd ed.). New York: McGraw-Hill, Inc. Potterat, JJ, Rothenberg, RB, & Muth, SQ. (1999). Network structural dynamics and infectious disease propagation. International Journal of Std & Aids , 10, 182-185. Potterat, JJ, Woodhouse, DE, Muth, SQ, & et al. (2004). Network dynamism: History and lessons of the Colorado Springs study. In M. Morris (Ed.), Network epidemiology: A handbook for survey design and data collection (pp. 87-114). New York: Oxford University Press Inc. Ramirez-Valles, J., Heckathorn, D. D., Vazquez, R., Diaz, R. M., & Campbell, R. T. (2005). From networks to populations: The development and application of respondent-driven sampling among IDUs and Latino gay men. Aids and Behavior , 9(4), 387- 402. doi: 10.1007/s10461-005-9012-3 Reisner, S. L., Mimiaga, M. J., Bland, S., Skeer, M., Cranston, K., Isenberg, D., . . . Mayer, K. H.

(2010). Problematic alcohol use and HIV risk among Black men who have sex with men in Massachusetts. Aids Care , 22(5), 577-587. Rhodes, S. D., McCoy, T. P., Hergenrather, K. C., Vissman, A. T., Wolfson, M., Alonzo, J., . . . Eng, E. (2012). Prevalence estimates of health risk behaviors of immigrant Latino men who have sex with men. Journal of Rural Health , 28(1), 73-83. Rice, E., Barman-Achikari, A., Milburn, N. G., & Monro, W. (2012). Position-specific HIV risk in a large network of homeless youths. American Journal of Public Health , 102(1), 141-147. Rothenberg, R. , & Muth, S.Q. . (2007). Large-network C insna.org | Volume 35 | Issue 1 | June Methods for Participant-Aided Sociograms concepts and small-network characteristics:

Fixed and variable factors. Sexually Transmitted Diseases , 34(8), 604-612. Rothenberg, R., Baldwin, B., Trotter, R., & Muth, S.Q. (2001). The risk environment for HIV transmission:results from the Atlanta and Flagstaff network studies. Journal of Urban Health , 78(3), 419-432. Schneider, J., Michaels, S., & Bouris, A. (2012). Family network proportion and HIV risk among Black men who have sex with men. Journal of Acquired Immune Deficiency Syndromes , 61(5), 627-635. Smith, Kirsten P., & Christakis, Nicholas A. (2008). Social networks and health. Annual Review of Sociology , 34, 405-429. Staras, SA, Cook, RL, & Clark, DB. (2009). Sexual partner characteristics and sexually transmitted diseases among adolescents and young adults.

Sexually Transmitted Diseases , 36(4), 232-237. Tucker, JS, Hu, Jianhui, Golinelli, D, Kennedy, DP, Green, HD, & Wenzel, SL. (2012). Social network and individual correlates of sexual risk behavior among homeless MSM youth. J Adolesc Health , 51(4), 386-392. Wasserman, S, & Faust, K. (1994). Social network analysis: Methods and applications . New York: Cambridge University Press. Wohlfeiler, D., & Potterat, JJ. (2005). Using gay men’s sexual networks to reduce sexually transmitted disease (STD)/human immunodeficiency virus (HIV) transmission. Sexually Transmitted Diseases , suppl 32(10), s48-s52.

C June | Issue 1| Volume 35 | insna.org Methods for Participant-Aided Sociograms Copyright ofConnections (02261766)isthe property ofInternational NetworkforSocial Network Analysis (INSNA) anditscontent maynotbecopied oremailed tomultiple sitesor posted toalistserv without thecopyright holder'sexpresswrittenpermission. However,users may print, download, oremail articles forindividual use.