Skip to main content
NCBI home page
NIHPA Author Manuscripts logoLink to NIHPA Author Manuscripts
. Author manuscript; available in PMC: 2025 Sep 9.
Published in final edited form as: Psychol Addict Behav. 2025 Sep 1;39(8):733–742. doi: 10.1037/adb0001092

Validity of a Brief Egocentric Social Network Assessment in Adults with Alcohol Use Disorder: Direct Comparison of 5-Alter and 20-Alter Versions

Kyla L Belisario 1,2, Allan D Clifton 3, Amanda Doggett 1,2, Alba González-Roz 4, Samuel F Acuff 5,6, John F Kelly 5,6, James MacKillop 1,2,*
PMCID: PMC12416765  NIHMSID: NIHMS2099809  PMID: 40892553

Abstract

Objective:

Social network analysis offers a high-resolution framework for understanding social influences on alcohol use, but full-length assessments confer significant burden, giving rise to brief measures. However, few studies have empirically compared brief- and full-length assessments. To address this, the present study examined the internal and external validity of both brief and full egocentric social network assessments, and their ability to capture weak social ties.

Method:

In 405 adults (57.5% female) with alcohol use disorder, a full egocentric social network assessment estimated drinking behaviour in the ego’s 20 important alters, as well as their perceived closeness and frequency of interaction to their network, and the presence of supportive ties (mutual help organization members or treatment providers). The assessment yielded 4 social network drinking characteristics: percent drinking endorsement; percent heavy drinking endorsement; drinking frequency; and heavy drinking frequency. Measures from the full 20-alter assessment were compared to measures from the first 5 alters.

Results:

Associations between brief and full network measures were of large magnitude (rs=.53-.73, p<.0001). Internal psychometric properties of the social network drinking characteristics were robust and similar in both assessments and, in terms of external validity, 13/16 (81.3%) associations of network drinking with the ego’s drinking severity were equivalent across both assessments. However, the brief assessment had less representation of mutual help organization members and treatment providers (ps<.01), resulting in a higher percent of alters endorsing drinking (p<.05). No other significant differences were present amongst other network drinking characteristics.

Conclusions:

These findings provide support for brief egocentric social network assessments, but also reveal limitations in characterizing potentially important weak social ties, namely the presence of mutual help organization members and treatment providers. Brief or full-length versions may be variably appropriate depending on the research and clinical aims.

Keywords: alcohol use disorder, recovery, social network drinking density, weak social ties, psychometric validation

INTRODUCTION

Alcohol use disorder (AUD) is a complex disease with biological, psychological, and social determinants interacting with one another (MacKillop et al., 2022; MacKillop & Ray, 2017; Morris et al., 2023; Sliedrecht et al., 2019). Among the social factors is the influence of one’s social network, whereby those who drink alcohol tend to seek out others who drink alcohol (referred to as selection dynamics) and reciprocally, alcohol use in one’s social group influences their level of drinking (referred to as influence dynamics) (Knox et al., 2019; MacKillop & Ray, 2017; Strickland & Acuff, 2023; van den Ende et al., 2024). These dynamics have been studied using social network analysis (SNA), robustly establishing the links between both the presence of drinking and risky drinking in a network on an individual’s own drinking behaviour (Rosenquist, 2010; Stout et al., 2012; van den Ende et al., 2024). Much of this research has been conducted in samples of emerging adults in university settings with potent social environmental contributors (Bartel et al., 2020; Meisel et al., 2015; Russell et al., 2021), with fewer studies examining the influence of social network drinking in adults in general (Knox et al., 2019).

Similar to etiology, the social impacts on those attempting recovery from AUD work via mechanisms of selection and influence dynamics (Strickland & Acuff, 2023). Specifically, individuals making a recovery attempt often select a network of others who are recovering by joining mutual help organizations (MHO), such as Alcoholics Anonymous, and reciprocally are positively influenced by the presence of those in recovery or who are abstinent (Kelly et al., 2010). Both mechanisms can emerge through ‘weak social ties’, defined as relationships that form a less dense network which can provide the ego with connections outside1 of their own network (Granovetter, 1973), increasing their social capital (Burt, 2000; Panebianco et al., 2016). Relating to recovery specifically, and contrary to what the term may suggest, these “weak ties” actually represent recovery specific strengths as they are a way in which individuals can gain separation from a network of individuals who are not beneficial to recovery (Burt, 2000), and can provide a source of alternative reinforcement (Murphy et al., 2021). In fact, diversity in one’s social network was found to be important in early recovery (Roxburgh et al., 2024), and an absence of weak social ties has been related to a greater likelihood of relapse (Panebianco et al., 2016). Indeed, abstinence in one’s network with up to 3 degrees of separation has been found to have a significant impact on a person’s abstinence status (Rosenquist, 2010), highlighting the importance of weak social ties.

To understand the role of social networks in AUD both in terms of risk and protection, social network drinking characteristics is typically captured by one of several egocentric assessments, which vary in the number of individuals (referred to as alters) who make up the individual’s network. Egocentric instruments capture the social network of a single individual as opposed to sociocentric instruments which capture the whole social network of a larger organization such as a university dormitory or classroom (Burgette et al., 2021; Valente et al., 2004). In egocentric networks, egos are at the center of their social network, and alters are individuals who have ties with the ego but not necessarily anyone else within the network (Burgette et al., 2021). One of the briefest egocentric assessments is the Brief Alcohol Social Density Assessment (BASDA; Fortune et al., 2013), which assesses the drinking behaviours of the 4 closest alters, and has been validated in both young adults (MacKillop et al., 2013) as well as general community adults (Levitt et al., 2020). On the other hand, the most comprehensive social network instruments ask participants about their 20 or more closest alters (Meisel et al., 2015; Russell et al., 2024), with its ability to discriminate between those with AUD demonstrated in a sample of community adults (Levitt et al., 2024). However, longer (20+ alter) social network instruments have yet to be validated in a sample of adults with AUD. Another instrument is the Important People (IP) instrument which asks about the 10 closest alters (Hallgren et al., 2013). A brief IP consisting of 5 alters was validated as a sufficient approximation of the longer IP instrument, although this was only validated in emerging adults, and as such did not include network characteristics important to recovery such as those who are MHO members or treatment providers (Hallgren & Barnett, 2016).

Given the clinical and research utility of social network drinking characteristics and the relative paucity of research comparing brief with full-length assessments particularly amongst individuals with AUD, this study sought to evaluate similarities and differences when using a brief (5-alter) versus full (20-alter) questionnaire. Specifically, the aims of the study were to: i) characterize the similarities and differences between the brief and full social network drinking measures and percent supportive ties (treatment providers and MHO members); ii) evaluate the internal reliability, discriminative validity, and precision of the brief and full measures to distinguish between varying levels of social network drinking density (using information function curves); and iii) compare the external validity of the measures in relation to the ego’s drinking quantity and severity. It is expected that both the brief and full social network assessments will produce similar estimates of network drinking density, but that there will be differences in the percent of supportive ties (treatment providers and MHO members) due to the nature of the assessment whereby participants typically name their social network in order of perceived closeness, and thus will likely be less close to beneficial weak social ties within their network.

METHODS

Participants and Procedures

This study uses data from a longitudinal observational cohort study examining mechanisms of behaviour change in AUD+ individuals making a substantive recovery attempt. Participants were recruited across two sites in North America: Hamilton, Ontario (McMaster University/St. Joseph’s Healthcare Hamilton), and Boston, Massachusetts (Harvard Medical School/Massachusetts General Hospital) starting in 2019 at local inpatient and outpatient treatment centres. However, following the declaration of the COVID-19, participants were also recruited via social media and local bus advertisements. Eligibility criteria were: i) meet the Diagnostic and Statistical Manual of Mental Disorders, Fifth Edition (DSM-5) diagnosis of AUD; ii) identify alcohol as their primary substance used in the instance of dual use; iii) engage in high-risk drinking in the 30 days prior to their recovery attempt, defined as an average of at least >7/14 standard drinks per week for women/men, respectively; iv) have made a significant recovery attempt within the past 90 days or within 14 days after enrollment into the study; and v) be between 21-65 years of age. A significant recovery attempt was defined as currently or planning to participate in inpatient/outpatient treatment centres (prior to the COVID-19 pandemic), or, after the onset of the COVID-19 pandemic, was expanded to include if they have recently or plan to make a serious attempt to either abstain from drinking or to drink without problems. This study uses data collected at baseline through in-person interviews with a research assistant (RA; virtual interviews after the onset of the COVID-19 pandemic), reflecting drinking prior to the recovery attempt. Informed consent was provided by all participants, and ethics approval was obtained from the Hamilton Integrated Research Ethics Board (Protocol #3825) and the Partners Human Research Committee at Massachusetts General Hospital (Protocol #2017P002345) for the Hamilton and Massachusetts studies, respectively. Given the aims of study were to evaluate the social network drinking measures, strict quality criteria were applied; of the total sample (N = 501), participants were only included in this study if they had 100% complete drinking data for all 20 alters and demonstrated adequate attention/effort during the social network assessment (i.e., those with > 90% identical closeness ratings for the 20 alters were excluded for invariant responding). Out of the n = 96 removed, 36 (37.5%) were missing either data for all 20 alters, or had some missingness in the first 5 alters, and n = 2 (2.08%) were removed for not passing quality control (i.e. demonstrated invariant responding). The remaining participants (60.4%) were excluded for missingness on a measure for one or more alter that was not one of the first 5 alters. The requirement of complete drinking data for all 20 alters was driven by the rationale that weak social ties would be captured by a lengthier social network instrument, but not by shorter instruments typically used in the field.

The final sample size was n = 405 (81%). Participant characteristics are in Table 1; the sample can be characterized as predominantly non-Hispanic White (18.3% racial-ethnic minority), middle-aged (mean ± SD age = 40.6 ± 11.1) adults with severe AUD (mean ± SD symptom count = 8.3 ± 2.4), and with slightly more female2 participants (57.5%). Final sample characteristics compared to the full sample are provided in Supplemental Table S1, and characteristics disaggregated by study location are in Supplemental Table S2.

Table 1:

Sample characteristics of 405 recovery-oriented adults with Alcohol Use Disorder (AUD) including demographic, drinking, and social network characteristics.

Characteristic Overall Sample
(N = 405)
Study Location N (%)
Canada 243 (60.00%)
United States 162 (40.00%)
Mean ± SD Age 40.64 ± 11.10
Sex assigned at Birth N (%)
Female 233 (57.53%)
Male 172 (42.47%)
Race/Ethnicity* N (%)
Racial-Ethnic Minority 74 (18.32%)
Non-Hispanic White 330 (81.68%)
Sexual Orientation* N (%)
Sexual Minority 95 (23.51%)
Heterosexual 309 (76.49%)
Subjective Income+ N (%)
Not Enough 53 (13.15%)
Cut Back 123 (30.52%)
No Cut Back 97 (24.07%)
Enough for Extras 130 (32.26%)
Education+ N (%)
Less than Bachelors 231 (57.32%)
Bachelors or Higher 172 (42.68%)
AA Attendance+ N (%)
AA attendance 87 (21.59%)
No AA attendance 316 (78.41%)
Drinking Characteristics Mean ± SE
AUD Symptoms 8.31 ± 0.12
Percent Drinking Days 65.39 ± 1.47
Average Drinks per Drinking Day 10.05 ± 0.34
Percent HDD+ 53.30 ± 1.63
Social Network Characteristics N (%)
% with 1+ Alter who attends an MHO 201 (49.63%)
% with 1+ Alter who is a treatment provider 131 (32.35%)
Open in a new tab
*

Missing is n = 1 observation

+

Missing are n = 2 observations.

SD = Standard Deviation; SE = Standard Error; AA = Alcoholics Anonymous; AUD = Alcohol Use Disorder; HDD = Heavy Drinking Days; MHO = Mutual Help Organization.

Measures

Drinking-Related Measures:

Drinking was measured by the Timeline Follow-Back (TLFB; Sobell & Sobell, 1992), which is a calendar-based method of collecting daily consumption of standard alcoholic beverages for the past 90 days. The TLFB was used to capture percent drinking days, percent heavy drinking days (HDD; 4+ drinks for women, 5+ for men, aligning with the binge drinking definition from the National Institute of Alcohol Abuse and Alcoholism (NIAAA, 2020)), and average drinks per drinking day. Alcohol use disorder severity was measured by the Diagnostic Assessment and Research Tool (DART; McCabe et al., 2017); a DSM-5 AUD symptom count which has been validated in AUD populations (Garber et al., 2024).

Social Network Drinking Density:

Detailed compositional characteristics of the participant’s 20-closest personal connections within the past 3 months3 were captured using an egocentric social network questionnaire. The following instructions were given to participants by a trained RA in an interview as follows: “In this questionnaire I will ask you about the people who have been important to you and with whom you have had frequent contact during the past 3 months. Please think about people who are at least 12 years of age. This contact could have been in person, by phone, Skype, email, or text. The people who you name may be family members, friends, people from work or treatment, self-help groups, or anyone you see as having a significant impact on your life.”. To aid participants who demonstrated difficulty in naming 20 individuals, participants were prompted by RAs to name individuals in order of importance, although this wasn’t an explicit requirement. Subsequent network-related questions were completed by the participant using online survey capturing software (REDCap; Harris et al., 2019). To measure network drinking, there are 4 characteristics calculated from two frequency-related drinking patterns of alters: i) the perceived4 frequency the alter has consumed alcohol in the past month (“In the past month, how often does [alter name] drink alcohol?”), which calculates a) percent of alters who consume alcohol, and b) the average alcohol consumption frequency of alters; and ii) the perceived frequency the alter has had a HD day in the past month (“In the past month, how often does [alter name] have 5 (male) / 4 (female) drinks?”), which calculates a) percent of alters endorsing any heavy drinking (HD), and b) the average HD frequency of alters. Frequency categories were: Never (1), Monthly or less (2), 2-4 times a month (3), 2 times a week (4), and 4 or more times a week (5). The questionnaire also captures whether an alter is an active member of a recovery-based MHO, and whether they are a treatment provider such as a doctor, social worker, addictions attendant, or other type of health care provider. In total, the relevant number of items administered was 7 per alter (140 in total). A full breakdown of the questions used in the study are provided in Supplemental Table S3. The brief network characteristics were calculated from the first 5 alters of the full (20-alter) egocentric social network questionnaire. Additionally, the bottom 15-alters were calculated from the latter 15 alters of the full social network questionnaire to robustly assess the differences between the first 5 from the remaining 15 alters, removing any influence of the first 5 alters in the full 20-alter measures.

Data Analysis

Descriptive statistics were used to characterize the social network characteristics of the sample, with paired t-tests used to assess whether the brief (5-alter) and full (20-alter) measures were significantly different from one another. To assess internal reliability, both Cronbach’s α and McDonald’s ω was calculated. Internal discriminative validity was assessed through concordance of parallel (brief and full) network drinking density measures using Pearson correlations. The precision at which the brief and full measures could accurately assess varying levels of social network drinking density was assessed using item response theory (IRT). The strength of IRT is its capacity to evaluate the ability of several psychometric measures to measure a single underlying (i.e. latent) trait such as social network drinking density through discrimination (α) and difficulty (b) parameters which are modelled through information function curves. Discrimination is the ability of an item to differentiate between participants with a range of network drinking density. Higher discrimination values can be interpreted as the measure’s ability to adequately distinguish between those with varying levels of network drinking, with values between 1 and 2 considered good functioning of the measure (Bichi & Talib, 2018). Difficulty can be thought of as item severity (Cappelleri et al., 2014), and refers to whether an measure can distinguish between those with low or high levels of network drinking density. Specifically, positive values can be interpreted as the measure being able to precisely differentiate between those with higher network drinking, and that the measure is well-suited for use in samples with high drinking density (Cappelleri et al., 2014; Nguyen et al., 2014). Discrimination and difficulty parameters of the brief and full network measures were estimated using the lavaan package (Rosseel, 2012) using the structural equation modelling (SEM) approach (Zopluoglu, 2020). To assess external validity, Pearson correlations with alcohol-related outcomes of the ego were examined, and 95% Confidence Intervals (CIs) were calculated for the correlation coefficients using Fisher Z Transformations. Overlapping correlation coefficient CIs were considered supportive of equivalent associations between the ego’s alcohol outcomes for both the brief and full network measures. In addition, the Hittner et al., (2003) method for dependent, overlapping coefficients was used to determine whether correlation coefficients were statistically different from one another (Diedenhofen & Musch, 2015). All analyses were conducted in R (R Core Team, 2024).

Transparency and Openness

This study reports on how sample size and data exclusions were determined, and all measures reported in the study follow the Journal Article Reporting Standards (JARS; Kazak, 2018). Research materials are available upon request to the corresponding author, pending ethics review board approval. This study was not preregistered.

RESULTS

Differences between Brief and Full Social Network Characteristics

There was a higher percent of alters who are MHO members (8.68%; p < 0.001) or treatment providers (4.01%; p < 0.001) in the full (20-alter) measure compared to the brief (5-alter) measure (6.43% and 1.58%, respectively). Specifically, in the full social network, half (49.6%) of participants reporting one or more alters who attend an MHO, and a third (32.4%) reporting one or more alters who are treatment providers. Yet, in the brief social network, only 22.0% of participants report one or more alters who attend an MHO and only 4.4% report one or more alters who are treatment providers. This means that MHO members in the full (20-alter) network were not captured by the brief questionnaire more than half of the time (56%) and treatment providers in the full network are not captured by the brief questionnaire 86% of the time. Overall, one or more supportive tie (treatment providers or MHO members) was in the full network for 59.8% of participants, but this reduced to just 24.7% participants with one or more supportive tie in their brief social network.

Average social network drinking characteristics using the 5-alter, 20-alter, and latter 15-alter measures are in Figure 1. Paired t-tests in Table 2 reveal that both reported frequency of interaction and closeness with the alter were significantly different between the brief and full measures, as were the brief measures compared to the latter 15-alter measures (all ps < 0.001). These findings were expected, due to the nature of the instrument which suggests participants list alters in descending order of most significant relationship. Alters endorsing any alcohol use was lower in the full measure compared to the brief measure (p = 0.008), which is expected given the higher percent of MHO members in the latter 15-alters. However, there was no significant difference in network drinking frequency, HD frequency, and percent HD between brief and full measures (ps > 0.05), suggesting the brief assessment similarly captures social network drinking severity. The findings were consistent between the brief and latter 15-alter measures, suggesting that the first 5 alters are reflective of the latter 15 alters in one’s social network despite the presence of weaker social ties in the latter 15 alters.

Figure 1: Average 5-alter, 20-alter, and latter 15-alter social network drinking characteristics.

Figure 1:

Open in a new tab

Mean percent in network endorsing alcohol consumption and heavy drinking (HD), and mean frequency of alcohol consumption and HD ranging from 1 (Never) to 5 (4+ times a week). Average across the brief 5-alter (top left); Average across bottom 15-alters (bottom left), and all 20 individual alter averages represented (right, in order reported as denoted by the y-axis).

Table 2:

Paired t-tests comparing mean brief (5-alter) versus the full (20-alter) and the bottom 15-alter network measures.

Mean ± SE Brief versus FULL Brief versus Remaining
Alters
5-Alter 20-Alter d; p-value latter 15-alters d; p-value
Social Network Relationships:
Frequency of interaction with Alters* 3.72 ± 0.03 3.16 ± 0.02 1.04; < 0.001 2.97 ± 0.03 1.04; < 0.001
Closeness with Alters+ 3.20 ± 0.03 2.64 ± 0.03 1.27; < 0.001 2.45 ± 0.03 1.28; < 0.001
Social Network Drinking:
Alcohol Frequency of Alters† 2.81 ± 0.04 2.76 ± 0.04 0.07; 0.146 2.75 ± 0.04 0.07; 0.146
HD Frequency of Alters† 2.20 ± 0.04 2.25 ± 0.04 0.08; 0.104 2.26 ± 0.04 0.08; 0.104
% Alters Consuming Alcohol 70.96 ± 1.26 68.48 ± 1.10 0.13; 0.008 67.65 ± 1.10 0.13; 0.008
% Alters Endorsing HD 54.32 ± 1.46 54.86 ±1.21 0.02; 0.615 55.05 ± 1.33 0.02; 0.615
Social Network Support:
% Alters Attending MHO Meetings 6.43 ± 0.73 8.68 ± 0.71 0.18; < 0.001 9.42 ± 0.80 0.18; < 0.001
% Alters who are Treatment Providers 1.58 ± 0.42 4.01 ± 0.40 0.30; < 0.001 4.82 ± 0.48 0.30; < 0.001
Open in a new tab
*

Mean frequency of direct interaction with each alter through any means (in person, via telephone, text messaging, etc.) on a scale from 1 to 5: Not in the last year (1), once a month (2), once a week (3), multiple times a week (4), daily (5).

+

Mean perceived closeness to the alter on a scale from 1 to 4: Not close (1), slightly close (2), moderately close (3), very close (4).

†

Mean frequency of drinking/heavy drinking (HD) on a scale from 1 to 5: Never (1), Monthly or less (2), 2-4 times a month (3), 2 times a week (4), and 4 or more times a week (5).

SE = Standard Error; HD = Heavy Drinking; MHO = Mutual Help Organization

Internal Reliability and Discriminative Validity of Social Network Characteristics

Robust internal reliability was demonstrated in both the brief (α = 0.91; ω = 0.91) and full (α = 0.93; ω = 0.93) social network drinking measures. There were large magnitude positive correlations between the brief and full network measures (rmean = 0.71; rrange = 0.69 – 0.73, all ps < 0.001), and large magnitude positive correlations between brief and full network measures of the percent who are MHO members (r = 0.64; p < 0.001) and treatment providers (r = 0.53; p < 0.001). A heatmap of correlation coefficients is in Figure 2.

Figure 2: Pearson correlation coefficients of the brief 5-alter versus full 20-alter measures.

Figure 2:

Open in a new tab

Values in bold denote statistical significance at p < .05. The same social network characteristics in the 5-alter and 20-alter assessments form the diagonal (which are outlined). MHO = mutual help organization. See the online article for the color version of this figure.

Both the brief (5-alter) and full (20-alter) measures displayed similar abilities to distinguish between individuals with a wide range of network drinking density, as reported in the information function curves in Figure 3 and Supplemental Table S4. Specifically, both the 5- and 20-alter average HD frequency had the highest discrimination values (α5 = 1.39; α20 = 1.67), reflected by the steepest slope, whilst the 5- and 20-alter percent alcohol use had the lowest discrimination values (α5 = 1.05; α20 = 1.07), reflected by the shallowest slopes. The 5- and 20-alter average HD frequency measures displayed a similar positive difficulty estimate (b5 = 0.44; b20 = 0.40), reflective of an ability to differentiate with greater precision those with more severe levels of network drinking, whereas 5- and 20-alter percent alcohol use had the lowest level of difficulty (b5 = −0.82; b20 = −0.69), meaning the measure is better at distinguishing between those with low alcohol use in their network.

Figure 3: Item function curves of the brief 5-alter and full 20-alter social network drinking density measures.

Figure 3:

Open in a new tab

The vertical dotted lines depict the difficulty parameter, which corresponds to the social network drinking density (θ) value of the curve at a 0.5 probability (horizontal dotted line). The slope of the curve at this point corresponds to the discrimination parameter.

External Validity of Social Network Characteristics

The association of both the brief (5-alter) and full (20-alter) social network measures with the ego’s alcohol use quantity and severity are in Figure 4. For the associations with the ego’s percent HDD and AUD symptoms, the brief and full network measures were not significantly different from one another (all ps > 0.05). However, the associations of the ego’s percent drinking days with network drinking frequency were significantly different (p = 0.038), with the 20-alter correlation (r20 = 0.16, p = 0.001) nearly twice the size of the 5-alter correlation (r5 = 0.09, p = 0.077). Similarly, the association of the ego’s drinks per drinking day with percent network drinking were also significantly different (p = 0.004), with the 20-alter correlation being almost twice the size of the 5-alter correlation (r20 = −0.26, r5 = −0.14, ps < 0.005). The association of the ego’s drinks per drinking day with percent network HD were significantly different from one another (p = 0.025), although both correlations are themselves non-significant (r5 = 0.01, r20 = −0.08, both p > 0.05). All other 5-alter and 20-alter correlations with the ego’s alcohol-related measures were not significantly different (ps > 0.05).

Figure 4: Pearson correlation coefficients and Fisher Z 95% confidence intervals between the ego's drinking quantity and severity characteristics and the brief 5-alter (left) and full 20-alter (right) social network alcohol use characteristics.

Figure 4:

Open in a new tab

Bold denotes significant correlation coefficient of the social network measure (5-alter or 20-alter) with the ego characteristic), whilst an outline denotes statistically different dependent overlapping correlation coefficients by type of assessment (5-alter versus 20-alter). For example, the correlation coefficients of the 5-alter and 20-alter percent network drinking with the ego’s drinks per drinking day suggest that both network measures are significantly correlated with the ego characteristic, but that the magnitude of these correlations (5- and 20-alter) are significantly different from one another.

AUD = Alcohol Use Disorder; HDD = Heavy Drinking Days

DISCUSSION

The current study compared brief (5-alter) and full-length (20-alter) egocentric social network measures to evaluate both information loss and validity in the substantially shorter assessment. The measures of the brief assessment exhibited many similarities to the full assessment, including large associations of the brief network drinking measures with their respective full network drinking measures, indicating strong overlap. Psychometric performance of the brief and full network characteristics as assessed by information function curves were similar, demonstrating the ability of the brief alter to precisely measure varying levels of social network drinking density. Collectively, these results demonstrated robust internal validity. External validation revealed similar (i.e. non-significantly different) associations of the brief and full network characteristics with the ego’s alcohol use quantity and severity in many instances (81.3% of associations), with the exception of three significant differences where the brief measure produced associations which were halved.

In terms of differences, most were as expected, and reflected the ranking of alters in order of closeness. As such, closeness of alters was higher in the brief measure, and was unable to capture treatment providers most of the time (86%) and MHO members more than half of the time (56%) as these alters are more likely to be peripheral and belong to the bottom 15 alters. In this regard, the brief assessment over-estimated percent social network alcohol use, and was unable to capture individuals who may be important for the person making a recovery attempt by serving as weak social ties. That is, it under-estimated social ties who might help the individual to make connections to different networks outside of their own. This is of particular importance, as only 21.6% of the sample endorsed attending AA prior to their recovery, but half (49.6%) had one or more alters in their social network who attended AA or other similar MHO. The presence of these supportive ties in those who are not currently attending a MHO could be leveraged by clinicians to foster supportive relationships for the individual undergoing recovery, highlighting this potential loss of information in the brief network measures. Contrary to percent network drinking, HD frequency was not over-estimated in the brief assessment. Instead, brief HD frequency better discriminated between those with high drinking density in their social network compared to the full HD frequency measure (as seen in the item function curves), and, between the brief and full HD frequency measure there was 100% overlap with the ego’s drinking outcomes. This suggests that the brief HD frequency may provide higher resolution of social network drinking density than measures related to social network drinking percent or frequency alone, which has been demonstrated in other research (Russell et al., 2023).

Of interest, although several drinking network measures were associated with alcohol consumption and AUD severity, non-significant associations or significant negative associations were typically observed, which differs from previous studies which found positive associations (Knox et al., 2019). This may be because most previous work has used university students (Bartel et al., 2020, 2020; Meisel et al., 2015) and non-clinical adults (Levitt et al., 2020; Rosenquist, 2010), whereas the current sample comprises of adults with severe AUD initiating a recovery attempt. Resolution of an AUD can take an individual a mean of approximately 5 separate attempts (Kelly et al., 2019), meaning that participants may have had prior recovery attempts whereby they introduced supportive ties (i.e. MHO members or treatment providers) or those who abstain from drinking into their network. However, this is necessarily conjecture and warrants exploration in future studies which can account for the number of prior recovery attempts of the ego.

On balance, these findings suggest that the brief assessment is able to characterize core compositional features of the network but lacks resolution for more nuanced important features outside of the closest individuals. In future applications, the acceptability of these trade-offs will need to be considered on a study-by-study basis. For investigations in which the principal focus is network density of drinking or there are concerns about assessment burden, the brief measure would be expected to be viable. Indeed, these results largely converge with recent findings that a brief social network assessment exhibited slightly lower, but nonetheless similar diagnostic discrimination between adults with AUD and matched controls (Levitt et al., 2024). On the other hand, for studies that seek to capture the role of MHO members, treatment providers, or the role of weak social ties in general, these results suggest the brief assessment may potentially fall short and yield misleading estimates. This is necessarily an empirical question, and future research should examine whether brief egocentric social network assessments can adequately capture the change in social network composition of those making an active recovery attempt from AUD.

These findings need to be considered in the context of both its strengths and limitations. The study is the first to examine differences between brief and full-length SNA measures in a clinical AUD+ sample and did so in a relatively large sample and with high-resolution measures. A further strength is leveraging the same questionnaire, allowing for a direct comparison of the utility of 5-alter versus 20-alter social network characteristics using the same ranking of alters. However, a limitation of this approach is that truly independent responses to a brief measure alone were not available combined with an absence of explicit instructions to list alters in a given order (by importance, closeness, or other related ranking metric). Although it is not clear why participants would substantively change their responding in the absence of explicit instructions, it is fundamentally an empirical question. It is plausible that being faced with the task of naming 20 individuals, participants could have utilized various methods for recalling alters, such as starting with certain social circles such as family members before moving onto friends. Although closeness was empirically tested whereby participants perceived to be closer to the first 5 alters than the latter 15, the listing of alters in order of closeness cannot be assumed. Another limitation of the instrument used in this study is the time-period for which the social network measures cover. Specifically, the social network data collected covered the past 3 months prior to the baseline assessment data collected covered the past 3 months since the baseline assessment and did not necessarily overlap with the 90 days prior to the participants’ recovery attempt (which their drinking-related data span). Thus, it is possible that some alters could be relationships newly formed in a treatment program that they would not have had prior to their recovery journey. This may further explain the inverse (null/ negative) associations found between some of the social network drinking variables and the ego’s drinking. With regards to other limitations, a notable one is that the measures derived from the questionnaire assumes that all ties are equal in their support (Granovetter, 1973), and do not discriminate the perceived support of alters unlike other instruments such as the IP Drugs and Alcohol (IPDA; Zywiak et al., 2009). Despite alters being deemed as important to the ego, the supportive nature of these relationships for the ego’s sobriety are not measured and thus cannot be assumed. As such, the measurements calculated from the social network questionnaire do not allow for any weighting or filtering of alter by perceived support (or lack of support). Negative (Saunders et al., 2016) or non-supportive relationships (Anderson et al., 2021; Stout et al., 2012), and similarly perceived negative relationship of treatment providers can all negatively impact the individual (Meier et al., 2006; Nordfjaern et al., 2010). As such, negative relationships, particularly amongst those who are abstinent, may counteract any assumed positive impact on the individual with AUD. A further important limitation to consider is the potential for results to not be generalizable to other populations undergoing a recovery attempt, due to the requirement of non-missing drinking data for all 20-alters. Specifically, those excluded from the final sample due to missing/incomplete drinking data had on average a lower AUD symptom score, and were on average older than those included. It is possible that those with less severe AUD symptoms may be less likely to have any supportive ties in their full social network, and as such may be empirically different from the sample included in analyses which saw a high percent of supportive ties. As such, more research should be conducted within less severe AUD groups or those undergoing their first recovery attempt. Additionally, the final sample was predominantly non-Hispanic White, and as such results may not be generalizable to other racial or ethnic identities. This is important as racial-ethnic minorities face barriers to accessing MHO meetings (Zemore et al., 2023) and have a smaller and more closely knit social network (Flores et al., 2020), meaning credible differences in social network variables may be present.

CONCLUSION

In sum, compared to a full-length assessment, drinking measures derived from a brief social network assessment were found to be generally psychometrically robust and to generate substantively similar relations with alcohol use, albeit with less resolution for weak social ties to individuals in recovery-related settings, either informal (MHO) or formal (treatment providers). As such, these findings provide support for the use of a briefer social network instrument to understand social factors in AUD when those types of weak social ties are lower priority or when measurement efficiency is of high priority.

Supplementary Material

Text and graphics

Public Health Significance Statement:

  • This study validated the use of a brief 5-alter social network assessment derived from a full-length 20-alter assessment to capture network drinking characteristics in a sample of adults with alcohol use disorder.

  • Limitations were also identified in the brief assessment’s ability to capture potentially important weak social ties in the network, such as mutual help organization members or treatment providers, which may substantively influence recovery.

Acknowledgements:

The authors are grateful for the participation of study participants, as well as the ongoing support of research staff.

Funding:

This work is supported by the Peter Boris Chair in Addictions Research (JM), a Tier 1 Canada Research Chair in Translational Addiction Research (CRC-2020-00170; JM), and NIH grants (R01 AA025849, JM and JK; K24 AA022136, JK).

Footnotes

Conflicts of Interest: JM is a senior scientist and principal in BEAM Diagnostics, Inc., and has served as a consultant to Clairvoyant Therapeutics, Inc. There are no other conflicts of interest to declare.

1

An individual in the network which provides the ego with connections outside of their own social network is coined a bridge, and can provide what is called brokerage between two otherwise unconnected subgroups (Burgette et al., 2021).

2

The congruence between sex assigned at birth and cis-gender is high in this sample (99%), and happened to align with the sex binary. The decision to focus on sex assigned at birth is not to minimize gender identity nor the existence of gender identity outside of the binary.

3

The past 3-month time period refers to the past 90 days prior to the baseline assessment and does not necessarily overlap with the 90 days prior to the participants’ recovery attempt.

4

As alter characteristics are captured by the ego completing the questionnaire, alcohol use by the alters is perceived by the ego. For example, percent alcohol endorsement represents the percent of alters that the ego believes to consume alcohol, and not necessarily the percent of alters who endorse alcohol consumption.

REFERENCES

  1. Anderson M, Devlin AM, Pickering L, McCann M, & Wight D (2021). ‘It’s not 9 to 5 recovery’: The role of a recovery community in producing social bonds that support recovery. Drugs: Education, Prevention and Policy, 28(5), 475–485. 10.1080/09687637.2021.1933911 [DOI] [PMC free article] [PubMed] [Google Scholar]
  2. Bartel SJ, Sherry SB, Smith MM, Glowacka M, Speth TA, & Stewart SH (2020). Social Influences on Binge Drinking in Emerging Adults: Which Social Network Members Matter Most? Substance Abuse, 41(4), 480–484. 10.1080/08897077.2019.1709604 [DOI] [PubMed] [Google Scholar]
  3. Bichi AA, & Talib R (2018). Item Response Theory: An Introduction to Latent Trait Models to Test and Item Development. International Journal of Evaluation and Research in Education (IJERE), 7(2), 142. 10.11591/ijere.v7i2.12900 [DOI] [Google Scholar]
  4. Burgette JM, Rankine J, Culyba AJ, Chu K-H, & Carley KM (2021). Best Practices for Modeling Egocentric Social Network Data and Health Outcomes. HERD: Health Environments Research & Design Journal, 14(4), 18–34. 10.1177/19375867211013772 [DOI] [PMC free article] [PubMed] [Google Scholar]
  5. Burt RS (2000). The Network Structure Of Social Capital. Research in Organizational Behavior, 22, 345–423. 10.1016/S0191-3085(00)22009-1 [DOI] [Google Scholar]
  6. Cappelleri JC, Jason Lundy J, & Hays RD (2014). Overview of Classical Test Theory and Item Response Theory for the Quantitative Assessment of Items in Developing Patient-Reported Outcomes Measures. Clinical Therapeutics, 36(5), 648–662. 10.1016/j.clinthera.2014.04.006 [DOI] [PMC free article] [PubMed] [Google Scholar]
  7. Diedenhofen B, & Musch J (2015). cocor: A Comprehensive Solution for the Statistical Comparison of Correlations. PLoS ONE, 10(4), e0121945. 10.1371/journal.pone.0121945 [DOI] [PMC free article] [PubMed] [Google Scholar]
  8. Flores M, Ruiz JM, Goans C, Butler EA, Uchino BN, Hirai M, Tinajero R, & Smith TW (2020). Racial-ethnic differences in social networks and perceived support: Measurement considerations and implications for disparities research. Cultural Diversity & Ethnic Minority Psychology, 26(2), 189–199. 10.1037/cdp0000283 [DOI] [PubMed] [Google Scholar]
  9. Fortune EE, MacKillop J, Miller JD, Campbell WK, Clifton AD, & Goodie AS (2013). Social density of gambling and its association with gambling problems: An initial investigation. Journal of Gambling Studies, 29(2), 329–342. [DOI] [PubMed] [Google Scholar]
  10. Garber ML, Belisario K, Levitt EE, McCabe RE, Kelly J, & MacKillop J (2024). Psychometric validation of the Diagnostic Assessment Research Tool: Alcohol use disorder module. Alcohol and Alcoholism, 60(1), agae088. 10.1093/alcalc/agae088 [DOI] [PMC free article] [PubMed] [Google Scholar]
  11. Granovetter MS (1973). The Strength of Weak Ties. American Journal of Sociology, 78(6), 1360–1380. [Google Scholar]
  12. Hallgren KA, & Barnett NP (2016). Briefer assessment of social network drinking: A test of the Important People Instrument–5 (IP-5). Psychology of Addictive Behaviors, 30(8), 955–964. 10.1037/adb0000218 [DOI] [PMC free article] [PubMed] [Google Scholar]
  13. Hallgren KA, Ladd BO, & Greenfield BL (2013). Psychometric properties of the Important People Instrument with college student drinkers. Psychology of Addictive Behaviors, 27(3), 819–825. 10.1037/a0032346 [DOI] [PMC free article] [PubMed] [Google Scholar]
  14. Harris PA, Taylor R, Minor BL, Elliott V, Fernandez M, O’Neal L, McLeod L, Delacqua G, Delacqua F, Kirby J, & others. (2019). The REDCap consortium: Building an international community of software platform partners. Journal of Biomedical Informatics, 95, 103208. [DOI] [PMC free article] [PubMed] [Google Scholar]
  15. Hittner JB, May K, & Silver NC (2003). A Monte Carlo evaluation of tests for comparing dependent correlations. The Journal of General Psychology, 130(2), 149–168. [DOI] [PubMed] [Google Scholar]
  16. Kazak AE (2018). Editorial: Journal article reporting standards. American Psychologist, 73(1), 1–2. 10.1037/amp0000263 [DOI] [PubMed] [Google Scholar]
  17. Kelly JF, Greene MC, Bergman BG, White WL, & Hoeppner BB (2019). How Many Recovery Attempts Does it Take to Successfully Resolve an Alcohol or Drug Problem? Estimates and Correlates From a National Study of Recovering U.S. Adults. Alcoholism: Clinical and Experimental Research, 43(7), 1533–1544. 10.1111/acer.14067 [DOI] [PMC free article] [PubMed] [Google Scholar]
  18. Kelly JF, Stout RL, Magill M, & Tonigan JS (2010). The role of Alcoholics Anonymous in mobilizing adaptive social network changes: A prospective lagged mediational analysis. Drug and Alcohol Dependence, S0376871610003236. 10.1016/j.drugalcdep.2010.09.009 [DOI] [Google Scholar]
  19. Knox J, Schneider J, Greene E, Nicholson J, Hasin D, & Sandfort T (2019). Using social network analysis to examine alcohol use among adults: A systematic review. PLOS ONE, 14(8), e0221360. 10.1371/journal.pone.0221360 [DOI] [PMC free article] [PubMed] [Google Scholar]
  20. Levitt EE, Oshri A, Clifton A, Stout R, Costello MJ, VanDellen M, Kelly JF, & MacKillop J (2020). Psychometric Validation of the Brief Alcohol Social Density Assessment in an Adult Community Sample. Alcoholism: Clinical and Experimental Research, 44(1), 212–218. 10.1111/acer.14227 [DOI] [PMC free article] [PubMed] [Google Scholar]
  21. Levitt EE, Singh D, Clifton A, Stout R, Sweet L, Kelly JF, & MacKillop J (2024). Diagnostic discrimination of social network indicators in alcohol use disorder: Initial examination using high-resolution and brief assessments. Psychology of Addictive Behaviors, 38(6), 656–667. 10.1037/adb0001006 [DOI] [PMC free article] [PubMed] [Google Scholar]
  22. MacKillop J, Acker JD, Bollinger J, Clifton A, Miller JD, Campbell WK, & Goodie AS (2013). Brief Report: The Brief Alcohol Social Density Assessment (BASDA): Convergent, Criterion-Related, and Incremental Validity. Journal of Studies on Alcohol and Drugs, 74(5), 810–815. 10.15288/jsad.2013.74.810 [DOI] [PMC free article] [PubMed] [Google Scholar]
  23. MacKillop J, Agabio R, Feldstein Ewing SW, Heilig M, Kelly JF, Leggio L, Lingford-Hughes A, Palmer AA, Parry CD, Ray L, & Rehm J (2022). Hazardous drinking and alcohol use disorders. Nature Reviews Disease Primers, 8(1), 80. 10.1038/s41572-022-00406-1 [DOI] [Google Scholar]
  24. MacKillop J, & Ray LA (2017). The Etiology of Addiction. In MacKillop J, Kenna GA, Leggio L, & Ray LA (Eds.), Integrating Psychological and Pharmacological Treatments for Addictive Disorders (1st ed., pp. 32–53). Routledge. 10.4324/9781315683331-2 [DOI] [Google Scholar]
  25. McCabe R, Milosevic I, Rowa K, Shnaider P, Pawluk E, & Antony M (2017). Diagnostic Assessment Research Tool. Psychological Assessment. [Google Scholar]
  26. Meier PS, Donmall MC, McElduff P, Barrowclough C, & Heller RF (2006). The role of the early therapeutic alliance in predicting drug treatment dropout. Drug and Alcohol Dependence, 83(1), 57–64. [DOI] [PubMed] [Google Scholar]
  27. Meisel MK, Clifton AD, MacKillop J, & Goodie AS (2015). A social network analysis approach to alcohol use and co-occurring addictive behavior in young adults. Addictive Behaviors, 51, 72–79. 10.1016/j.addbeh.2015.07.009 [DOI] [PubMed] [Google Scholar]
  28. Morris J, Boness CL, Burton R, Dar-Nimrod I, & Moss A (2023). Balancing the bio in a biopsychosocial model of hazardous drinking and alcohol use disorders. Qeios. 10.32388/I1120F [DOI] [Google Scholar]
  29. Murphy JG, Dennhardt AA, & Gex KS (2021). Individual Behavioral Interventions to Incentivize Sobriety and Enrich the Natural Environment with Appealing Alternatives to Drinking. In Tucker JA & Witkiewitz K (Eds.), Dynamic Pathways to Recovery from Alcohol Use Disorder (1st ed., pp. 179–199). Cambridge University Press. 10.1017/9781108976213.015 [DOI] [Google Scholar]
  30. Nguyen TH, Han H-R, Kim MT, & Chan KS (2014). An Introduction to Item Response Theory for Patient-Reported Outcome Measurement. The Patient - Patient-Centered Outcomes Research, 7(1), 23–35. 10.1007/s40271-013-0041-0 [DOI] [PMC free article] [PubMed] [Google Scholar]
  31. NIAAA. (2020). National Institute on Alcohol Abuse and Alcoholism: Binge Drinking. [Google Scholar]
  32. Nordfjaern T, Rundmo T, & Hole R (2010). Treatment and recovery as perceived by patients with substance addiction. Journal of Psychiatric and Mental Health Nursing, 17(1), 46–64. [DOI] [PubMed] [Google Scholar]
  33. Panebianco D, Gallupe O, Carrington PJ, & Colozzi I (2016). Personal support networks, social capital, and risk of relapse among individuals treated for substance use issues. International Journal of Drug Policy, 27, 146–153. 10.1016/j.drugpo.2015.09.009 [DOI] [PubMed] [Google Scholar]
  34. R Core Team. (2024). R: A Language and Environment for Statistical Computing. R Foundation for Statistical Computing. https://www.R-project.org/ [Google Scholar]
  35. Rosenquist JN (2010). The Spread of Alcohol Consumption Behavior in a Large Social Network. Annals of Internal Medicine, 152(7), 426. 10.7326/0003-4819-152-7-201004060-00007 [DOI] [PMC free article] [PubMed] [Google Scholar]
  36. Rosseel Y. (2012). lavaan: An R Package for Structural Equation Modeling. Journal of Statistical Software, 48(2), 1–36. 10.18637/jss.v048.i02 [DOI] [Google Scholar]
  37. Roxburgh AD, Best D, Lubman DI, & Manning V (2024). Composition of social networks to build recovery capital differ across early and stable stages of recovery. Addiction Research & Theory, 32(3), 186–193. 10.1080/16066359.2023.2238594 [DOI] [Google Scholar]
  38. Russell AM, Barry AE, & Patterson MS (2023). Strategies for measuring peer alcohol use within egocentric network analysis. Journal of American College Health, 71(7), 2176–2183. [DOI] [PubMed] [Google Scholar]
  39. Russell AM, Monds L, Hing N, Kroll J, Russell AM, & Thorne HB (2024). Social associations and alcohol consumption in an Australian community sample: An egocentric social network analysis. Psychology of Addictive Behaviors, 38(2), 211. [DOI] [PubMed] [Google Scholar]
  40. Russell AM, Patterson MS, & Barry AE (2021). College Students’ Perceptions of Peer Alcohol Use: A Social Network Analytic Approach. Substance Use & Misuse, 56(1), 46–53. 10.1080/10826084.2020.1833929 [DOI] [PubMed] [Google Scholar]
  41. Saunders EC, McLeman BM, McGovern MP, Xie H, Lambert-Harris C, & Meier A (2016). The influence of family and social problems on treatment outcomes of persons with co-occurring substance use disorders and PTSD. Journal of Substance Use, 21(3), 237–243. [DOI] [PMC free article] [PubMed] [Google Scholar]
  42. Sliedrecht W, De Waart R, Witkiewitz K, & Roozen HG (2019). Alcohol use disorder relapse factors: A systematic review. Psychiatry Research, 278, 97–115. 10.1016/j.psychres.2019.05.038 [DOI] [PubMed] [Google Scholar]
  43. Sobell LC, & Sobell MB (1992). Timeline follow-back: A technique for assessing self-reported alcohol consumption. In Measuring alcohol consumption: Psychosocial and biochemical methods (pp. 41–72). Springer. [Google Scholar]
  44. Stout RL, Kelly JF, Magill M, & Pagano ME (2012). Association Between Social Influences and Drinking Outcomes Across Three Years. Journal of Studies on Alcohol and Drugs, 73(3), 489–497. 10.15288/jsad.2012.73.489 [DOI] [PMC free article] [PubMed] [Google Scholar]
  45. Strickland JC, & Acuff SF (2023). Role of social context in addiction etiology and recovery. Pharmacology Biochemistry and Behavior, 229, 173603. 10.1016/j.pbb.2023.173603 [DOI] [PMC free article] [PubMed] [Google Scholar]
  46. Valente TW, Gallaher P, & Mouttapa M (2004). Using Social Networks to Understand and Prevent Substance Use: A Transdisciplinary Perspective. Substance Use & Misuse, 39(10–12), 1685–1712. 10.1081/JA-200033210 [DOI] [PubMed] [Google Scholar]
  47. van den Ende MWJ, van der Maas HLJ, Epskamp S, & Lees MH (2024). Alcohol consumption as a socially contagious phenomenon in the Framingham Heart Study social network. Scientific Reports, 14, 4499. 10.1038/s41598-024-54155-0 [DOI] [PMC free article] [PubMed] [Google Scholar]
  48. Zemore SE, Mericle AA, Martinez P, Bergman BG, Karriker-Jaffe KJ, Patterson D, & Timko C (2023). Disparities in Alcoholics Anonymous Participation from 2000-2020 among U.S. Residents with an Alcohol Use Disorder in the National Alcohol Survey. Journal of Studies on Alcohol and Drugs, jsad.23-00086. 10.15288/jsad.23-00086 [DOI] [Google Scholar]
  49. Zopluoglu C. (2020). A finite mixture item response theory model for continuous measurement outcomes. Educational and Psychological Measurement, 80(2), 346–364. [DOI] [PMC free article] [PubMed] [Google Scholar]
  50. Zywiak WH, Neighbors CJ, Martin RA, Johnson JE, Eaton CA, & Rohsenow DJ (2009). The Important People Drug and Alcohol interview: Psychometric properties, predictive validity, and implications for treatment. Journal of Substance Abuse Treatment, 36(3), 321–330. 10.1016/j.jsat.2008.08.001 [DOI] [PMC free article] [PubMed] [Google Scholar]

Associated Data

This section collects any data citations, data availability statements, or supplementary materials included in this article.

Supplementary Materials

Text and graphics
NIHMS2099809-supplement-Text_and_graphics.docx (33.7KB, docx)

RESOURCES