Substance Use, Recovery Capital, and Social Networks of Adolescents Recently Initiating Inpatient Treatment

Abstract

Background:

A variety of personal and environmental resources (recovery capital: RC) support recovery for adolescents. Social identity theory suggests one’s substance use behavior is strongly influenced by social network behavior. This study examined how the level of pretreatment substance use was associated with RC and social network characteristics among youth entering inpatient treatment. Understanding these relationships may help inform initial treatment engagement and post-treatment transitions.

Method:

Participants (N=108; Mage=17.1; 61% male; 70% White) completed standardized measures on admission to inpatient treatment: demographics, past 90-day substance use, recovery capital, and social network composition/behavior. Regression analyses were used to examine relationships between AOD use before treatment and (1) recovery capital and (2) level of identification with network groups.

Results:

On average, in the 90 days prior to entry into treatment adolescents consumed alcohol on 20.3 days, used cannabis on 46.5 days, and used other substances (e.g., cocaine) on 17.3 days in the 90 days. The mean RC score was 36.8 (range 25-50). Being male and having more treatment days before baseline were associated with higher RC. Pre-treatment substance use was not consistently associated with RC levels. 70% of participants included at least one family or friend group who used AODs. Participants with a recovery group identified more strongly with that group than with substance-using groups (M=7.28 compared to M=6.63 for groups using substances heavily; scale 0-10).

Conclusions:

Adolescents appear to have flexible social identities and potentially risky post-treatment relationships, highlighting the need to engage them in recovery-supportive networks following treatment.

Graphical Abstract

1. Introduction

Alcohol and other drug (AOD) use disorders among adolescents result in many serious immediate and distal outcomes, including early mortality (Feldstein Ewing et al., 2014; Patrick et al., 2016; Schulenberg and Maggs, 2001; Schulte and Hser, 2013; Squeglia et al., 2014; Wilson et al., 2015). Adverse outcomes may be mitigated, however, if the adolescent engages in treatment and initiates recovery, i.e., the “process of change through which individuals improve their health and wellness, live a self-directed life, and strive to reach their full potential” (Substance Abuse and Mental Health Services Administration (SAMHSA), 2012). Recovery capital is the total set of resources a person possesses for sustaining recovery. It is comprised of four major domains: human, financial, social, and community resources (Granfield and Cloud, 1999; Hennessy et al., 2018). In particular, the domain of social recovery capital (e.g., sober friends, family, and peer groups) is vital to reducing substance use, especially for adolescents (Anderson et al., 2007; Brown et al., 1989; Jurinsky et al., 2022; Ramo et al., 2012). Adolescents are in the developmental stage of forming new social identities in a variety of social contexts. Their social environments and social group memberships begin to re-shape their social identity as they seek to be accepted by peer groups by behaving in ways aligned with perceived group norms (e.g., in the case of substance use, perceived or observed use by group members; Field et al., 2024). Thus, recovery becomes more or less difficult depending on the degree to which social influences promote or discourage substance use within the larger context of available recovery capital (Anderson et al., 2007; Jurinsky et al., 2022; Stone et al., 2012).

However, few studies have examined associations between levels of substance use, recovery capital, and social influences in the early stages of treatment among adolescents. This is an important gap when considering approaches to best engage adolescents in treatment and plan for their transition back to potentially risky environments after discharge. Adolescents who strongly identify with groups using substances may be harder to engage in treatment and they may be at high relapse risk after treatment ends. This study examines the relationships among levels of substance use, recovery capital, and social network characteristics in a sample of adolescents who recently entered inpatient treatment for substance use disorders (SUD).

1.1. Adolescent substance use, recovery, and recovery capital

In 2022 in the US, over 2.2 million youth reported needing SUD treatment (Substance Abuse and Mental Health Services Administration, 2023). The process of recovery is difficult for adolescents and often involves multiple returns to use and subsequent rounds of treatment (Buckheit et al., 2018). Recovery Capital theory suggests that resources at different ecological levels (individual, interpersonal, community) are necessary for recovery (Granfield and Cloud, 2001, 1999). Developed among adults and adapted for adolescents, the Recovery Capital for Adolescents Model (RCAM) is comprised of human, financial, social, and community recovery capital domains that support recovery (Hennessy et al., 2018). RCAM posits that the more positive resources an adolescent has for recovery, the higher the likelihood that they will recover (Best and Hennessy, 2021). Conversely, lacking these factors, or having direct barriers in these domains, makes recovery more difficult. Recovery capital and barriers can change over time and vary across settings, especially social environments (Granfield and Cloud, 1999; Hennessy, 2017). Social recovery capital, a key component of RCAM, refers to social resources (e.g., sober friends, supportive family, pro-social engagement and the absence of using friends and networks) (Hennessy et al., 2018). It contributes to overall recovery capital by fostering coping skills, abstinence self-efficacy, and motivation, both through modeling these characteristics and direct encouragement and reinforcement (Bandura, 1986).

1.2. Social Influence and Recovery Capital in Adolescent Substance Use

Social influence plays a powerful role in shaping adolescent behavior, including substance use, as adolescents are in the process of developing their social identities (Cheng and Lo, 2015; Gibbons et al., 2012; Godley et al., 2005). Social identity, defined by group memberships, guides one’s behavior based on the perceived norms and values, including high risk behaviors such as the use of those groups, including high risk behaviors such as the use of substances. For example, evidence demonstrates that adolescents with family members or close peers who use substances are at higher risk of returning to substance use after treatment (Anderson et al., 2008, 2007; Eddie and Kelly, 2017; Gonzales et al., 2012; Jurinsky et al., 2022; Savolainen et al., 2018).

The Social Identity Model of Recovery highlights how this process could evolve. When using substances, and early in recovery, adolescents often remain tied to substance-using groups. These groups continue to exert strong influence and result in maintaining a “using” identity leading to continued substance use (Best et al., 2016). Over time, exposure to, and engagement with, sober and recovery-oriented groups can shift an adolescent’s social identity toward sobriety (Mackenzie et al., 2016; Mawson et al., 2016). Stable recovery typically reflects a dominant connection to recovery-supportive and non-using groups (Best et al., 2016, 2014; Hennessy et al., 2025, 2024). To support this shift, it is crucial to understand the structure and influence of adolescents’ social networks, including promoters of, and barriers to, developing recovery capital, and more specifically, social recovery capital through sober or recovery-supportive groups.

Recovering adolescents are in a transitional phase of their identity development. Although they may have begun to distance themselves from the “using” group’s social identity, they also may not align with a “recovery” identity. Barriers (e.g., social influence, personal challenges) after leaving treatment can lead to recovery setbacks. For example, if youth strongly identify with groups using substances, it may make it harder for them to engage with sober, recovery-supportive groups and with the treatment approach when entering treatment. Thus, strong ties to influential groups using substances may delay behavior change. The RCAM suggests that the way an adolescent addresses these challenges is determined by recovery capital resources. Social learning theory suggests these resources will be in part determined by countervailing social network influences (Bandura, 1986), through both recovery-supportive networks and relationships as well as bridges to community assets and resources. The objectives of this study were to characterize the nature of the social networks of recovering adolescents within the first month of entering treatment and the association between pretreatment levels of substance use, recovery capital, and social network structure. We then examined how these relationships can inform our expectations for youth entering treatment.

1.3. Study Aims

We had three specific hypotheses related to days of substance use prior to entering treatment and recovery capital: Adolescents with (1) higher alcohol use, (2) higher cannabis use, or (3) higher other substance use would report lower levels of baseline recovery capital. We had four hypotheses related to social network structure based on social identity theory: (1) adolescents with greater alcohol use prior to treatment (i.e., days of use prior to treatment) would report (1a) lower identification with non-alcohol using groups and (1b) higher identification with groups using alcohol; and (2) adolescents with greater other substance use prior to treatment (i.e., days of use prior to treatment) would report (2a) lower identification with non-substance using groups and (2b) higher identification with groups using substances.

Methods

The study was approved by the Massachusetts General Hospital (MGH) IRB and pre-registered on the OSF framework (June 20, 2023: https://doi.org/10.17605/OSF.IO/VPBCJ). Starting in July 2023, we recruited adolescents aged 12–19 entering a 12-step–based inpatient treatment center for substance use disorders (SUD) in the Midwest United States. The program served youth with any type of substance use and incorporated a complementary set of evidence-based approaches, including 12-step facilitation and psychosocial interventions (e.g., group therapy, individual counseling). The typical length of inpatient treatment at this program is 14–30 days.

A standardized survey battery was administered at intake to inpatient treatment and for one year posttreatment (3-, 6-, 12-mo). The battery consisted of clinical research interviews, self-administered questionnaires, and a social identity mapping procedure. The current study is based on cross-sectional baseline data collected to date, focusing on adolescents with completed baseline assessments (N = 108).

2.1. Participant Enrollment

Eligible participants were adolescents (12-19 years old) entering inpatient treatment (see above). Potential participants were identified at treatment intake and approached by a research assistant within 1-3 weeks of entry for possible participation in this study. The majority (69%) had their baseline assessment within 14 days of treatment intake (only 7% had it within the first week). A parent consented for adolescent minors to provide assent. Age was significantly associated with the time between treatment entry and their baseline assessment (r = −0.35, p<.05), as it took longer to enroll younger participants given the additional task of getting parent consent. Once enrolled, a baseline study visit was scheduled within 1-2 days. Participants were paid $40 for completing the baseline. Of those screened, 83% consented and completed a baseline assessment.

2.2. Data collection

Data were collected in person by a trained BA-level research assistant (RA). The study visit took up to 2 hours but was broken up into several smaller blocks of time to accommodate the treatment center’s daily activities. During the study visit the RA conducted a structured interview and the participant completed two self-administered survey blocks via REDCap (Harris et al., 2009). The RA also facilitated the online social identity mapping (Bentley et al., 2020).

2.3. Measures

Alcohol, cannabis, and other substance use in the 90 days prior to treatment entry, and days of school attendance, were assessed by the RA using a Timeline Follow-Back (TLFB) approach (Sobell et al., 1985; Sobell and Sobell, 1992). The TLFB has good test-retest reliability, and high validity for recording verifiable events (Hjorthøj et al., 2012; Sobell et al., 1985; Tonigan et al., 1997). Symptoms of SUD were recorded by the RA using the Mini-International Neuropsychiatric Interview (M.I.N.I.) KID, English Version 7.0.2 (Sheehan et al., 2010). The MINI KID is a short, structured diagnostic interview, developed by psychiatrists and clinicians. It was used in this study to assess alcohol and other substance use disorder symptoms and their severity (mild, moderate, severe) according to DSM-5 criteria. Mental health treatment history, age of first AOD use, and general health rating (fair-excellent) were recorded using the Adolescent Diagnostic Interview (Winters and Henly, 1993)

Self-administered surveys included (1) demographics (using question wording from the NIH “All of Us” research program, December 2018, Version 8) and (2) Assessment of Recovery Capital (Groshkova et al., 2013). The Assessment of Recovery Capital measures personal recovery capital (e.g., confidence and motivation for recovery) and social recovery capital (e.g., recovery support from friends, involvement in positive social activities) based on how the person feels on that day. Items are dichotomously scored (0 = no, 1 = yes), resulting in a score ranging from 0-25 within each domain, and a score of 0–50 overall. Higher scores indicate higher recovery capital. We adapted some questions with wording appropriate for adolescents (See Supplement Table S1). The internal consistency in this sample was moderate to good (alpha=0.85).

Social network information was derived from a social identity mapping task. Social identity mapping is a data collection process in which the participant creates a map of their various social groups, including each group’s substance use behaviors. The resulting map can help to identify additional aspects of social recovery capital, and barriers to building it, by identifying the influences associated with groups important to the participant (Beckwith et al., 2018; Cruwys et al., 2016; Jurinsky et al., 2022; Mawson et al., 2016). The electronic, online version was used in this study for ease of data collection (Bentley et al., 2020), using the standardized facilitator guide recently developed with adolescents (Hennessy et al., 2024). In this process, an RA uses a standardized guide to pose a series of questions to the participant. The questions prompt the participant to create a map of their social groups, with themselves at the center and other social groups represented by circles (see Supplement, Figure S1 for examples from a male and female participant). The map captures the following information: (1) number of social groups identified by the respondent and (2) number of group members (0-10+) per group, (3) group importance (1-5, from least important to most important), (4) time spent with the group (0-10; with 10 as the most), (5) level of identification with group (0-10: 10 as the highest), (6) level of conflict between self and each group, and between groups with other groups (0-3; no, low, medium, high conflict) represented by differently weighted lines drawn between groups, (7) alcohol use of group members (rated 0-10 for the entire group; 10 as the highest), and (8) substance use (substances aside from alcohol and nicotine/tobacco) of groups color-coded by the degree of participant-rated perceived severity (heavy=red, casual=yellow, none=blue, and in recovery=green). These data can be used to calculate several variables representing group-level behaviors and social identity (e.g., ratio of substance-using groups to overall number of groups). Based on the descriptive labels provided by participants for groups, we categorized them into the following group types: Family, Friends, Treatment (e.g., staff, “Therapist”), Treatment Peers (e.g., “Treatment peers”, Recovery Support Service Peers (e.g., “Friends from AA”)), Pets, Work, School (e.g., classmates, teachers), Religious, and Sports team. Groups could be categorized in multiple types (e.g., “Volleyball friends” was considered both a “Friend” and “Sports team” group).

2.4. Participant Characteristics

Of the 108 participants recruited, 61% were male and the average age was 17.1 years (SD = 1.5). The majority were Non-Hispanic White (70%). Participants’ mean recovery capital (RC) on a scale of 0-50 at study baseline was 36.8 (SD=7.1; sample range: 20-49), indicating fairly high RC. On average, in the 90 days before treatment, participants had consumed alcohol for 20.3 days (SD = 21.8), used cannabis for 46.5 days (SD = 28.1), and used another substance for 17.3 days (SD = 25.3). The top “other” substance that youth indicated were problems for them included opiates (22%) and stimulants (10%). Scores on the Mini-International Neuropsychiatric Interview for Children and Adolescents (Sheehan, 2010) indicated that 98% of participants met criteria for a non-alcohol SUD and 70% met criteria for an alcohol use disorder (AUD), for a total of 66% with both (only two participants presented with just an AUD). Demographic characteristics of the sample are shown in Table 1.

Table 1.

Demographic Characteristics and Intake Substance Use of Sample (N=108).

Characteristic	Total Sample (N = 108)	Males (n = 66)	Females (n = 42)	Welch’s T-Test for Sex Differences
Age	17.14 (1.54)	16.95 (1.53)	17.43 (1.53)	tdf(89.44)= 1.57; p = 0.1207
Age, first received AOD treatment/counseling	15.64 (1.89)	15.53 (1.89)	15.79 (1.91)	tdf(51.38)= 0.52; p = 0.6076
Days between intake and baseline	13.36 (6.42)	14.23 (7.32)	12.00 (4.41)	tdf(107.65)= −1.97; p = 0.0512
Assessment of Recovery Capital Score (0-50)	36.80 (7.14)	38.53 (6.76)	34.07 (6.93)	tdf(87.71)= −3.29; p = 0.0014
Personal Recovery Capital (0-25)	18.34 (4.02)	19.50 (3.61)	16.52 (4.00)	tdf(82.50)= −3.91; p = 0.0002
Social Recovery Capital (0-25)	18.45 (3.82)	19.03 (3.83)	17.55 (3.66)	tdf(92.53)= −2.01; p = 0.0468
Days of alcohol use in prior 90 days	20.33 (21.83)	19.64 (21.39)	21.43 (22.70)	tdf(85.41)= 0.41; p = 0.6836
Days of cannabis use in prior 90 days	46.54 (28.06)	45.89 (27.31)	47.55 (29.49)	tdf(84.25)= 0.29; p = 0.7708
Days of other substance use in prior 90 days	17.44 (25.30)	14.55 (21.81)	22.00 (29.68)	tdf(70.15)= 1.40; p = 0.1647
Total instances, AOD treatment/counseling	1.41 (2.00)	1.38 (2.14)	1.45 (1.78)	tdf(96.09)= 0.17; p = 0.8663
General health rating (Fair-Excellent: 1-4)	2.83 (0.72)	2.92 (0.73)	2.68 (0.69)	tdf(90.91)= −1.72; p = 0.0880
AUD Status
Mild AUD	0.13 (0.34)	0.15 (0.36)	0.10 (0.31)
Moderate AUD	0.14 (0.35)	0.19 (0.40)	0.07 (0.26)
Severe AUD	0.72 (0.45)	0.66 (0.48)	0.83 (0.38)
SUD Status
Mild SUD	0.04 (0.20)	0.05 (0.22)	0.03 (0.16)
Moderate SUD	0.05 (0.22)	0.05 (0.22)	0.05 (0.22)
Severe SUD	0.91 (0.29)	0.90 (0.30)	0.93 (0.27)
Gender
Man	63 (58.3%)	63 (95.5%)
Woman	40 (37.0%)		40 (95.2%)
Non-binary	4 (3.7%)	2 (3.0%)	2 (4.8%)
Transgender	1 (0.9%)	1 (1.5%)
Race and Ethnicity
Non-Hispanic White	76 (70.4%)	49 (74.2%)	27 (64.3%)
American Indian or Alaska Native	3 (2.8%)	1 (1.5%)	2 (4.8%)
Asian	3 (2.8%)	2 (3.0%)	1 (2.4%)
Black, African American, or African	2 (1.9%)	1 (1.5%)	1 (2.4%)
Hispanic, Latino, or Spanish	3 (2.8%)	2 (3.0%)	1 (2.4%)
Middle Eastern or North African	2 (1.9%)	1 (1.5%)	1 (2.4%)
Native Hawaiian or other Pacific Islander	1 (0.9%)	1 (2.4%)
Multiple race/ethnicities selected	18 (16.7%)	10 (15.2%)	8 (19.0%)
American Indian, Blacka	1 (6%)
American Indian, Hispanica	1 (6%)
American Indian, Whitea	4 (22%)
American Indian, Black, Whitea	1 (6%)
White, Asiana	2 (11%)
White, Blacka	3 (16.7%)
White, Hispanica	5 (28%)
White, Middle Easterna	1 (6%)
Received counseling/treatment for anxiety	87 (87.9%)	49 (83.1%)	38 (95.0%)
Received counseling/treatment for depression	93 (93.9%)	53 (89.8%)	40 (100.0%)
Currently attending school	66 (61.1%)	46 (69.7%)	20 (47.6%)

Note.

^aThe percent represents the proportion of each of these racial/ethnic identities out of the multiracial/ethnic category. Given the small cell sizes, the sex of the participants within each multiracial/ethnic category is not further reported.

2.5. Data Analysis

All data analyses were conducted in Stata SE (V. 17). We first examined the descriptive characteristics of all measures and network characteristics. Given the unequal distribution of sex, we calculated Welch’s t-test for sex differences on main demographic and social network variables. Given that the sample was primarily White, and due to the modest sample size, participant race was dichotomized to Non-Hispanic white versus non-white to ensure adequate cell counts and stable parameter estimation in the regression analyses. We next conducted seven sets of regressions, predicting the following dependent variables: (1) Baseline Recovery Capital; Identification with (2) alcohol-using groups, (3) non-alcohol-using groups, (4) heavy substance use groups, (5) casual substance use groups, (6) no-substance-use groups, and (7) recovery groups. For each dependent variable models were run with each of three independent variables (pre-treatment alcohol use; pre-treatment cannabis use; pre-treatment other substance use) and covariates (age, sex (Male = 1), White race, and days from treatment entry to baseline assessment). Statistical significance was initially defined a priori as p < .05 (two-tailed). To account for multiple comparisons, Bonferroni corrections were applied within outcome domains. For recovery capital analyses (3 tests), the adjusted significance threshold was α = .017, and for social identity analyses (18 tests), α = .0028. Although unadjusted p-values are reported, statistical significance was determined using these adjusted thresholds.

2. Results

3.1. Nature of social networks of adolescents in inpatient treatment

Table 2 presents key participant social network characteristics displayed overall, and then separately by sex and MINI KID indications of AUD/SUD (Supplement Table S2 includes information for additional social network variables). Tables 3 and Supplement Table S3 present information on categories of group types (e.g., family, friend) and characteristics associated with them by these categories.

Table 2.

Network characteristics around AOD use (by sex and AUD/SUD status)

	Whole Sample (N =108) M (SD)	Males (n =66) M (SD)	Females (n=42) M (SD)	Welch’s t-test for Sex Differences	Sample with AUD (n=76) M (SD)	Sample with SUD (n=102) M (SD)
	4.88	4.67	5.21	tdf(60.56) = 1.24;	4.97	4.90
Total number of groups on SIM	(2.02)	(1.54)	(2.59)	p = 0.2206	(2.10)	(2.06)
	6.65	6.73	6.52	tdf(88.22) = −0.58;	6.60	6.67
Mean number of people in groups on SIM	(1.81)	(1.80)	(1.83)	p = 0.5601	(1.71)	(1.83)
	1.04	0.86	1.31	tdf(82.99) = 2.53;	1.07	1.05
Total groups with heavy SU (#)	(0.90)	(0.84)	(0.92)	p = 0.0132	(0.88)	(0.91)
	1.70	1.70	1.71	tdf(88.23) = 0.07;	1.78	1.69
Total groups with casual SU (#)	(1.25)	(1.25)	(1.27)	p = 0.9449	(1.16)	(1.27)
	1.22	1.26	1.17	tdf(87.48) = −0.42;	1.21	1.24
Total groups without SU (#)	(1.08)	(1.07)	(1.10)	p = 0.6738	(1.00)	(1.09)
	1.06	1.00	1.14	tdf(82.19) = 0.77;	1.01	1.08
Total recovery groups (#)	(0.92)	(0.88)	(0.98)	p = 0.4433	(0.90)	(0.90)
	4.07	3.96	4.25	tdf(75.00) = 0.90;	4.31	4.05
Average AU level of group (0-10)	(1.52)	(1.38)	(1.72)	p = 0.3706	(1.44)	(1.55)

Note. Avg. = Average. AU = alcohol using. SU = substance using.

Table 3.

Network characteristics of each group type (N = 108)

	Total Number of Groups	Mean Number of Each Group Type M (SD)	Mean Severity of Alcohol Use by Group 0-10 M (SD)	Average Identification with Groups 0-10 M (SD)	Average Importance of Groups 1-5 M (SD)
Family Groups (n = 104)	135	1.30 (0.71)	4.03 (2.33)	6.46 (2.53)	3.41 (0.86)
Friend Groups (n = 102)	185	1.81 (1.18)	5.02 (2.3)	7.28 (1.96)	2.83 (0.96)
Treatment Groups (n = 34)	38	1.12 (0.33)	1.81 (2.53)	5.56 (2.27)	2.34 (1.04)
Treatment Peer Groups (n = 50)	53	1.06 (0.24)	2.8 (3.96)	7.08 (2.35)	2.34 (1.22)
RSS Peer Groups (n = 23)	24	1.04 (0.21)	1.06 (1.51)	7.13 (2.88)	2.65 (1.03)
Work Groups (n = 34)	37	1.09 (0.29)	4.84 (2.77)	4.69 (2.54)	1.47 (1.19)
School Groups (n = 25)	26	1.04 (0.20)	4.33 (3.18)	4.28 (2.75)	1.26 (0.97)
Religious Groups (n = 4)	4	1 (0)	3.55 (3.11)	5.47 (2.79)	1.93 (1.18)
Sports Team Groups (n = 15)	17	1.13 (0.35)	0 (NA)	7 (2.83)	2.75 (0.96)

3.1.1. Number and type of groups

Participants listed 4.9 social groups on average (SD=2.0; range 2-17), and a mean of 6.7 people per group (SD=1.8). They reported on average about 1 high, 2 moderate and 2-3 low conflict lines between groups in their networks. Participants were most likely to indicate at least one family or friend group, and often depicted more than one (up to 6 family and 8 friend groups). 46% had a treatment peer group and 21% had a recovery support service peer group.

3.1.2. Substance use behaviors of groups

In this sample, 70% of participants identified at least one heavy substance using (SU) group, with 97% having either a heavy or casual SU group. Of these, 12% of participants indicated having at least one family group who heavily used substances and 57% indicated having a least one friend group who heavily used substances. Most participants also had at least one non-SU and/or recovery group in their overall social network (96%); 12% were family groups and 22% were friend groups. Males had fewer heavy SU groups compared to females (M=0.86 versus 1.3, respectively; p<0.05). On a scale of 0-10, from no drinking to heavy drinking, participants reported an average alcohol use (AU) rating across groups of 4.07 (SD=1.52). The majority of friend and family groups used alcohol (83-87%), with friends rated only slightly higher than family on AU levels (5.02 versus 4.03, respectively).

3.1.3. Group identification, importance, and time spent

Overall, participants rated the strongest identification with recovery groups (M=7.28, SD = 2.23), followed by heavy, casual, and non-SU groups, but had similar levels of identification with both AU and non-AU groups (M=6.5 and M=6.7, respectively). Participants rated the highest level of importance for non-SU groups (M=2.97, SD = 0.96), followed by recovery, casual, and heavy SU groups. They also rated similar levels of importance for AU and non-AU groups (M=2.7 and M=2.6, respectively). Participants reported less time with AU than non-AU groups. The most time was spent with recovery groups, followed by heavy, non-, and casual SU groups. Family is rated the most important (M=3.41) group, followed by friends (M=2.83), but friends (M=7.28), treatment peers (M=7.08), and religious (7.0) are rated as the groups they most identify with, followed by family (M=6.46).

3.2. AOD use severity before Treatment and Recovery Capital at Baseline

We expected that those with higher AOD use in the 90-days prior to treatment entry would report lower recovery capital. Prior to adjustment, only days of other substance use, and not alcohol or cannabis use, was associated with lower RC. Yet, this association did not meet the Bonferroni-adjusted significance threshold (α = .017) and should be interpreted cautiously (Table 4). After correction for multiple comparisons, RC remained significantly associated with sex and days between treatment and baseline. Males had higher RC compared to females. More days between treatment intake and baseline was associated with higher RC.

Table 4.

Models Predicting Baseline Recovery Capital from Past 90-Day Alcohol, Cannabis, and Other Substance use

Models Predicting Baseline Recovery Capital (n = 108)
	Model 1						Model 2						Model 3
	b	SE	t	p	95% CI		b	S E	t	p	95% CI		b	SE	t	p	95% CI
AU prior to TX (90-day)	−0.03	0.03	−0.85	0.398	−0.08	0.03
Cannabis Use prior to TX (90-day)							−0.02	0.02	−0.72	0.473	0.06	0.03
Other Substance Use prior to TX (90-day)													−0.05	0.025	−2.07	0.041	−0.10	−0.002
Age	0.08	0.45	0.17	0.863	−0.81	0.97	0.01	0.45	0.02	0.981	0.88	0.91	0.20	0.45	0.45	0.655	−0.68	1.08
Sex (Male = 1)	3.70	1.33	2.77	0.007	1.06	6.35	3.72	1.34	2.78	0.006	1.07	6.37	3.37	1.32	2.54	0.012	0.74	5.99
White race	2.12	1.81	1.17	0.245	−1.48	5.72	2.03	1.81	1.12	0.265	−1.56	5.63	2.08	1.78	1.17	0.245	−1.45	5.61
Days (TX entry to baseline)	0.31	0.11	2.85	0.005	0.09	0.52	0.30	0.11	2.70	0.008	0.08	0.51	0.33	0.11	3.10	0.002	0.12	0.54
Constant	27.81	8.36	3.33	0.001	11.23	44.39	29.43	8.68	3.39	0.001	12.22	46.64	26.08	8.26	3.16	0.002	9.70	42.46
	R2 = 0.18 (Adj. R2 = 0.14) F(5, 102) = 4.49; Prob > F = 0.0010						R2 = 0.18 (Adj. R2 = 0.14) F(5, 102) = 4.44; Prob > F = 0.0011						R2 = 0.21 (Adj. R2 = 0.17) F(5, 102) = 5.35; Prob > F = 0.0002

Note. Exact p-values are reported. Bolded coefficients indicate statistical significance based on Bonferroni correction for multiple comparisons (α = .017).

3.3. AOD use severity before Treatment and the Identification with Alcohol Using v. Non-Alcohol Using Groups and Heavy, Casual, Recovery and Non-Substance Using Groups

We examined the relationship between history of AOD use at baseline and levels of identification with groups based on the AOD use status of those groups (Tables 5 and 6). There were few significant relationships across all models prior to Bonferroni adjustment, none remained after this adjustment, and the models held little explanatory value (Tables 5 and 6).

Table 5.

Models Predicting Participant Identification with Groups Who Used/Did not Use Alcohol Based on Severity of 90-Day Substance Use Prior to Treatment (N = 108)

	Level of Identification with Groups Using Alcohol
	Model 1						Model 2						Model 3
	b	SE	t	p	95% CI		b	SE	t	p	95% CI		b	S E	t	p	95% CI
AU prior to TX (90-day)	0.03	0.05	0.55	0.581	−0.07	0.12
Cannabis Use prior to TX (90-day)							0.08	0.037	2.23	0.028	0.009	0.16
Other Substance Use prior to TX (90-day)													0.06	0.04	1.34	0.183	−0.03	0.14
Age	0.02	0.73	0.02	0.981	−1.44	1.47	0.23	0.719	0.33	0.746	−1.19	1.66	−0.11	0.74	−0.15	0.878	−1.57	1.35
Sex (Male = 1)	−1.96	2.18	−0.90	0.372	−6.28	2.37	−1.93	2.132	−0.9	0.368	−6.16	2.3	−1.60	2.18	−0.73	0.466	−5.93	2.73
White race	−5.27	2.96	−1.78	0.078	−11.1	0.6	−5.12	2.892	−1.8	0.080	−10.9	0.61	−5.23	2.94	−1.78	0.078	−11.05	0.60
Days (TX entry to baseline)	0.07	0.18	0.40	0.687	−0.28	0.42	0.14	0.175	0.82	0.417	−0.2	0.49	0.05	0.17	0.28	0.784	−0.30	0.39
Constant	26.33	13.66	1.93	0.057	−0.75	53.4	18.2	13.85	1.31	0.192	−9.27	45.7	28.2	13.6	2.07	0.041	1.16	55.21
	R2 = 0.04 (Adj. R2 = −0.004) F(5, 102) = 0.91; Prob > F = 0.4768						R2 = 0.08 (Adj. R2 = 0.04) F(5, 102) = 1.89; Prob > F = 0.1033						R2 = 0.086 (Adj. R2 = 0.01) F(5, 102) = 1.22; Prob > F = 0.3040
	Level of Identification with Groups Not Using Alcohol
	Model 1						Model 2						Model 3
	b	SE	t	p	95% CI		b	SE	t	p	95% CI		b	S E	t	p	95% CI
AU prior to TX (90-day)	−0.03	0.03	−1	0.323	−0.09	0.03
Cannabis Use prior to TX (90-day)							0.02	0.02	0.62	0.534	−0.03	0.06
Other Substance Use prior to TX (90-day)													−0.03	0.03	−0.96	0.340	−0.08	0.03
Age	0.01	0.48	0.02	0.984	−0.94	0.96	0.01	0.48	0.02	0.986	−0.95	0.96	0.05	0.48	0.10	0.919	−0.91	1.01
Sex (Male = 1)	−0.34	1.42	−0.2	0.811	−3.16	2.48	−0.30	1.43	−0.21	0.832	−3.13	2.53	−0.49	1.44	−0.34	0.732	−3.34	2.35
White race	2.02	1.93	1.05	0.298	−1.81	5.85	1.94	1.94	1.00	0.319	−1.90	5.78	1.94	1.93	1.01	0.316	−1.89	5.77
Days (TX entry to baseline)	0.13	0.11	1.15	0.253	−0.10	0.36	0.15	0.12	1.30	0.196	−0.08	0.38	0.15	0.11	1.27	0.206	−0.08	0.37
Constant	2.48	8.91	0.28	0.782	−15.20	20.15	0.91	9.26	0.10	0.922	−17.47	19.29	1.58	8.96	0.18	0.860	−16.18	19.35
	R2 = 0.02 (Adj. R2 = −0.03) F(5, 102) = 0.43; Prob > F = 0.8249						R2 = 0.02 (Adj. R2 = −0.03) F(5, 102) = 0.36; Prob > F = 0.8773						R2 = 0.01 (Adj. R2 = −0.03) F(5, 102) = 0.30; Prob > F = 0.9144

Note. Exact p-values are reported. Bolded coefficients indicate statistical significance based on Bonferroni correction for multiple comparisons (α = .0028).

Table 6.

Models Predicting Participant Identification with Groups’ Degree of Substance Use Based on Severity of 90-Day Substance Use Prior to Treatment (N = 108)

	Level of Identification with Groups Heavily Using Substances
	Model 1						Model 2						Model 3
	b	SE	t	p	95% CI		b	SE	t	p	95% CI		b	SE	t	p	95% CI
AU prior to TX (90-day)	−0.002	0.02	−0.11	0.912	−0.03	0.03
Cannabis Use prior to TX (90-day)							0.03	0.01	2.06	0.04	0.001	0.05
Other Substance Use prior to TX (90-day)													0.03	0.01	1.92	0.058	−0.001	0.05
Age	−0.09	0.25	−0.36	0.717	−0.58	0.40	−0.03	0.24	−0.14	0.89	−0.51	0.45	−0.17	0.24	−0.69	0.493	−0.65	0.32
Sex (Male = 1)	−1.12	0.73	−1.53	0.129	−2.58	0.33	−1.10	0.72	−1.54	0.13	−2.53	0.32	−0.94	0.73	−1.29	0.200	−2.38	0.50
White race	−0.03	1.00	−0.03	0.979	−2.00	1.95	−0.01	0.98	−0.01	0.99	−1.95	1.92	−0.05	0.98	−0.05	0.960	−1.99	1.89
Days (TX entry to baseline)	−0.01	0. 06	−0.19	0.851	−0.13	0.11	0.01	0.06	0.22	0.82	−0.10	0.13	−0.02	0.06	−0.33	0.742	−0.13	0.10
Constant	7.03	4.59	1.53	0.129	−2.08	16.14	4.49	4.67	0.96	0.34	−4.77	13.74	7.89	4.54	1.74	0.085	−1.10	16.89
	R2 = 0.02 (Adj. R2 = −0.02) F(5, 102) = 0.50; Prob > F = 0.7733						R2 = 0.06 (Adj. R2 = 0.02) F(5, 102) = 1.37; Prob > F = 0.2419						R2 = 0.06 (Adj. R2 = 0.01) F(5, 102) = 1.26; Prob > F = 0.892
	Level of Identification with Groups Casually Using Substances
	Model 1						Model 2						Model 3
	b	SE	t	p	95% CI		b	SE	t	p	95% CI		b	SE	t	p	95% CI
AU prior to TX (90-day)	0.01	0.01	0.97	0.335	−0.01	0.04
Cannabis Use prior to TX (90-day)							0.001	0.01	0.11	0.92	−0.02	0.02
Other Substance Use prior to TX (90-day)													−0.01	0.01	−0.99	0.326	−0.04	0.01
Age	0.09	0.21	0.46	0.650	−0.32	0.51	0.11	0.21	0.54	0.59	−0.30	0.53	0.14	0.21	0.69	0.495	−0.27	0.56
Sex (Male = 1)	0.08	0. 62	0.13	0.900	−1.14	1.30	0.07	0.62	0.11	0.92	−1.16	1.29	−0.01	0.62	−0.02	0.981	−1.25	1.22
White race	−0.41	0.84	−0.49	0.627	−2.07	1.25	−0.37	0.84	−0.44	0.66	−2.03	1.30	−0.36	0.84	−0.43	0.668	−2.02	1.30
Days (TX entry to baseline)	−0.02	0.05	−0.41	0.680	−0.12	0.08	−0.02	0.05	−0.43	0.67	−0.12	0.08	−0.02	0.05	−0.39	0.697	−0.12	0.08
Constant	4.09	3.86	1.06	0.292	−3.57	11.74	3.99	4.02	0.99	0.32	−3.98	11.96	3.72	3.88	0.96	0.339	−3.97	11.4
	R2 = 0.02 (Adj. R2 = −0.03) F(5, 102) = 0.36; Prob > F = 0.8755						R2 = 0.01 (Adj. R2 = −0.04) F(5, 102) = 0.17; Prob > F = 0.9725						R2 = 0.02 (Adj. R2 = −0.03) F(5, 102) = 0.37; Prob > F = 0.8707
	Level of Identification with Groups Not Using Substances
	Model 1						Model 2						Model 3
	b	SE	t	p	95% CI		b	SE	t	p	95% CI		b	SE	t	p	95% CI
AU prior to TX (90-day)	0.0009	0.015	0.06	0.956	−0.03	0.031
Cannabis Use prior to TX (90-day)							−0	0.01	−0.33	0.74	−0.03	0.02
Other Substance Use prior to TX (90-day)													−0	0.01	−0.38	0.701	−0.03	0.02
Age	0.0753	0.233	0.32	0.747	−0.39	0.538	0.07	0.23	0.29	0.77	−0.4	0.53	0.09	0.24	0.39	0.700	−0.38	0.56
Sex (Male = 1)	1.56	0.693	2.25	0.027	0.184	2.935	1.56	0.69	2.25	0.03	0.182	2.93	1.52	0.7	2.18	0.032	0.138	2.91
White race	0.732	0.942	0.78	0.439	−1.14	2.6	0.73	0.94	0.78	0.44	−1.13	2.6	0.74	0.94	0.79	0.434	−1.13	2.6
Days (TX entry to baseline)	−0.026	0.056	−0.47	0.641	−0.14	0.085	−0.03	0.06	−0.53	0.6	−0.14	0.08	−0	0.06	−0.44	0.660	−0.14	0.09
Constant	1.7178	4.343	0.4	0.693	−6.9	10.33	2.11	4.5	0.47	0.64	−6.81	11	1.55	4.36	0.36	0.723	−7.1	10.2
	R2 = 0.06 (Adj. R2 = 0.01) F(5, 102) = 1.21; Prob > F = 0.3098						R2 = 0.06 (Adj. R2 = 0.01) F(5, 102) = 1.23; Prob > F = 0.2994						R2 = 0.06 (Adj. R2 = 0.01) F(5, 102) = 1.24; Prob > F = 0.2957
	Level of Identification with Groups in Recovery
	Model 1						Model 2						Model 3
	b	SE	t	p	95% CI		b	SE	t	p	95% CI		b	SE	t	p	95% CI
AU prior to TX (90-day)	−0.02	0.02	−0.89	0.377	−0.05	0.02
Cannabis Use prior to TX (90-day)							0.01	0.01	0.81	0.42	−0.02	0.04
Other Substance Use prior to TX (90-day)													0.01	0.02	0.78	0.44	−0.02	0.04
Age	−0.04	0.26	−0.14	0.886	−0.55	0.48	−0.03	0.26	−0.12	0.91	−0.55	0.49	−0.1	0.26	−0.34	0.74	−0.61	0.43
Sex (Male = 1)	−0.57	0.77	−0.74	0.462	−2.11	0.96	−0.55	0.77	−0.71	0.48	−2.09	0.99	−0.5	0.78	−0.61	0.54	−2.03	1.07
White race	0.43	1.05	0.41	0.684	−1.66	2.52	0.39	1.05	0.37	0.71	−1.69	2.48	0.38	1.05	0.36	0.72	−1.71	2.46
Days (TX entry to baseline)	0.11	0.06	1.74	0.085	−0.02	0.23	0.12	0.06	1.91	0.06	0.00	0.25	0.11	0.06	1.72	0.09	−0.02	0.23
Constant	4.61	4.85	0.95	0.344	−5.01	14.23	3.51	5.03	0.7	0.49	−6.47	13.49	4.97	4.88	1.02	0.31	−4.70	14.6
	R2 = 0.05 (Adj. R2 = −0.001) F(5, 102) = 0.97; Prob > F = 0.4376						R2 = 0.04 (Adj. R2 = −0.003) F(5, 102) = 0.97; Prob > F = 0.4376						R2 = 0.04 (Adj. R2 = −0.003) F(5, 102) = 0.94; Prob > F = 0.4601

Note. Exact p-values are reported. Bolded coefficients indicate statistical significance based on Bonferroni correction for multiple comparisons (α = .0028).

4. Discussion

As might be expected from the Social Identity Model of Recovery (Best et al., 2016), the social network results from this sample of adolescents who had recently entered inpatient treatment suggest a complex picture of social networks during recovery. Overall, adolescents reported diverse social networks that included both groups using substances and recovery-supportive groups. Nearly all participants identified at least one group using substances, particularly among friends, but encouragingly, the majority also reported at least one recovery or abstinent group. Consistent with prior work on the importance of recovery-supportive relationships for youth (Eddie and Kelly, 2017; Jurinsky et al., 2022), adolescents reported their strongest sense of identification with recovery groups, and rated non-substance-using groups as the most important. These findings suggest that, despite high levels of exposure to substance-using groups outside treatment, many youth entering treatment reported strong identification with a treatment or recovery-oriented group, potentially indicating early engagement with recovery identities. It is noteworthy that over a third of the participants had developed recovery-oriented social networks at baseline, although this could have been due to previous treatment episodes or biased reporting due to social desirability.

The degree of family substance use reported here was also striking. Most family groups used alcohol, and 43% were using other substances casually, with another 10% using heavily. Substance use among family members is a known recovery barrier and, if left unchanged, will be a challenge for youth to overcome upon treatment exit (Anderson et al., 2008, 2007; Hennessy et al., 2018). Although we did not capture the specific type of non-alcohol substance being used, these patterns may reflect broader shifts in substance use norms in some U.S. policy contexts. (e.g., evolving recreational cannabis landscape (Cerdá et al., 2020)). Regardless of substance type, these findings suggest that further work is needed to support young people whose family members use substances, particularly if they are returning to that home environment.

In examining recovery capital, the level of pre-treatment substance use was not consistently linked to lower RC across substances. Although days of “other” substance use (e.g., opioids, stimulants) showed an association with lower recovery capital at the unadjusted level, this did not meet the Bonferroni-adjusted significance threshold. Thus, while this pattern suggests that this type of substance use may represent a more significant risk to developing recovery resources or to greater loss of recovery capital, perhaps due to the use of these other substances causing greater social or functional impairment, relationship loss/conflict, self-stigma (e.g., leading to treatment avoidance, or lack of disclosure, worsening consequences and isolating individuals (Ayhan et al., 2024; Birtel et al., 2017; Blyth et al., 2024; Granfield and Cloud, 2001; Jurinsky et al., 2022; Neale et al., 2014), these results should be cautiously interpreted.

Patterns also emerged when examining social networks by sex. Males reported fewer heavy substance-using groups than females and were more likely to identify with non-substanceusing groups. This suggests in line with other research that social context may play a differential role in shaping recovery resources for males versus females, with potential implications for tailoring treatment and recovery supports (Field et al., 2024). Being male was also consistently associated with higher baseline recovery capital scores, a finding that aligns with theory and adult research suggesting that females may have more barriers to initially building recovery capital (Abreu Minero et al., 2022; Granfield and Cloud, 2001, 1999). It may take longer for females to be identified for treatment, leading to more negative consequences accruing, and lower recovery capital at treatment entry. Also, participants had higher recovery capital scores the longer they were in treatment before their study baseline, providing preliminary support for what seems obvious but has not been empirically examined: inpatient treatment helps build recovery capital for adolescents even after only a short time.

Contrary to expectations, substance use prior to treatment was not strongly associated with group identification. This could be due to several reasons. First, some participants might not have been using substances with others or to feel part of specific groups but rather to cope with negative feelings (e.g., anxiety, depression; (Armeli et al., 2021; Creswell, 2021; Creswell et al., 2014). Indeed, given the high prevalence of previous mental health treatment receipt for adolescents in this sample, that is one plausible explanation for the lack of group identification and substance use behaviors. Second, due to greater perceived distance from substance-using groups during treatment, it is likely that the salience of group identification was also low. The follow-up data will be important to examine whether group identification and group relationships shift after treatment.

4.1. Limitations

As this study presents baseline data from an observational sample of primarily White adolescents entering a single treatment center, its results must be interpreted with caution. Adolescents who were doing better overall were potentially more motivated to enter this treatment or to participate. Indeed, this sample had fairly high recovery capital and low variability. In the future, recruiting a more racially and socioeconomically diverse sample, across different types of treatment may provide additional nuance. Given that the consent process for minors involved reaching parents external to treatment, these youth may have had more involved/motivated family members than youth who did not enroll. Examining these relationships cross-sectionally does not capture the dynamic nature of relationships and does not allow for causality to be established between pre-treatment substance use, social network group composition, and recovery capital. Finally, although respondents supplied indications of the degree of conflict between groups, the nature of those conflicts was not captured. Future research should examine conflict in more detail for adolescent recovery processes.

4.2. Conclusions

Together, these results underscore the complexity of adolescent recovery environments and the potential flexibility of their social identity in specific environments. Adolescents entering treatment simultaneously report ties to both high-risk and recovery-supportive groups and may return to spending time with high-risk groups, especially when substance use is present within their homes. Future research should further examine the intersection of substance use, social groups, and both interpersonal characteristics and reasons for substance use (Meisel et al., 2023) to better understand treatment and recovery trajectories among adolescents. Understanding the interplay of these factors could better help to target key mechanisms of change maintenance for adolescents leaving treatment.

Highlights.

• We examined recovery capital (RC), substance use, and networks in recovering youth

• Alcohol or cannabis use before treatment was not linked to lower RC

• Higher other substance use before treatment was linked to lower RC

• Males and those with longer treatment days before the baseline survey had higher RC

• 46% identified strongly with their treatment peer group