US adolescents’ friendship networks and health risk behaviors: a systematic review of studies using social network analysis and Add Health data

Sexual behavior

Adolescents who engage in unprotected sexual behaviors have a considerably higher risk of experiencing an unintended pregnancy or contracting STIs, including HIV (Tapert et al., 2001), than those who do not engage in these behaviors. As of 2011, the YRBS reported that the percentage of adolescents (grades 9 through 12) responding positively to the question “ever had sexual intercourse” was 47.4% (Centers for Disease Control and Prevention, 2012d). Although this percentage is high, it represents a decline: in 1991, more than half (54.1%) of adolescents in grades 9 through 12 reported that they had engaged in sexual intercourse.

Due to its many negative health and psychosocial consequences (not the least of which are sexual abuse and statutory rape), having had sexual intercourse before the age of 13 is another problematic behavior among adolescents. Rates for this behavior have plateaued between 2001 and 2009 (Centers for Disease Control and Prevention, 2012d), and have dropped 4% (from 10.2% to 6%) compared to 1991.

Alcohol use

In the United States, alcohol use by adolescents is illegal (under age 21) and also remains a public health problem because it is associated with different risk behaviors, including tobacco use and unprotected sexual intercourse. Data from the YRBS documents that, in 2011, an estimated 70.8% of adolescents reported they “ever had at least one drink of alcohol on at least 1 day” during their lifetime (Centers for Disease Control and Prevention, 2012a). This statistic shows the percentage fell 11% compared to 1991.

By 2011, 38.7% of adolescents reported that they had “had at least one drink of alcohol on at least 1 day” during the 30 days before the survey (Centers for Disease Control and Prevention, 2012a). This percentage dropped from 50.8% in 1991, a 12% decrease.

Tobacco use

Smoking is related to morbidity and mortality, and is a leading cause of chronic diseases (e.g., cardiac disease and vascular disease). Although smoking under the age of 18 years is illegal in the US (US Food and Drug Administration, 2014), data from the YRBS in 2011 indicate that 44.7% of teens reported they “had tried cigarette smoking” (Centers for Disease Control and Prevention, 2012e). This rate has fallen by 25% since 1991 (70.1%). Another problematic smoking behavior among adolescents (i.e., “smoked cigarettes on at least 1 day”) when assessed in 2011, indicated 18.1% were smokers (Centers for Disease Control and Prevention, 2012e). This rate has fallen 9.4% since 1991 (from 27.5%). While during the period 1991–1997 the rates had gradually increased to 36.8%, the numbers have steadily decreased during between1999 and 2011.

Marijuana and cocaine use

For adolescents, marijuana and cocaine use can cause unexplained changes in personality or attitudes such as anxiety, poor social skills, interpersonal alienation, and poor impulse control. These drugs also can affect physical development (e.g., brain and nerve damage, respiratory problems, and blood pressure) (National Institute on Drug Abuse, 2014a; Brook, Balka & Whiteman, 1999; Shedler & Block, 1990; Volkow et al., 2014). Moreover, they can lead adolescents to other risky behaviors (e.g., sexual intercourse or drinking alcohol). Marijuana and cocaine use are illegal for adolescents in the US (National Institute on Drug Abuse, 2014b), yet in 2011, 39.9% of adolescents reported, in the YRBS, ever using marijuana “one or more times” (Centers for Disease Control and Prevention, 2012b). This rate has steadily increased by 8.6% since 1991. Between 1991 and 1999 the increase was even larger, from 31.3% to 47.2%.

During that same time-period, an estimated 6.8% of adolescents reported they “ever used any form of cocaine one or more times” (Centers for Disease Control and Prevention, 2012b). Moreover, from 1991 to 1999, the rates had slightly increased (5.9% to 9.5%).

Studies’ characteristics

Fifteen studies met our inclusion criteria. Most studies (n = 14) were conducted in the US, and one paper, in France. Most reviewed studies (n = 11) were published between 2009 and 2014, perhaps because social network analysis only recently became popular as an analytical tool. Even though network data were collected in Wave I of the Add Health survey (1994–1995), we found the earliest publication on social networks among the reviewed studies was published in 2001.

All reviewed studies appeared in journals with impact factors ranging from 1.638 to 4.266. Four were published in the Journal of Adolescent Health, and two studies were published in Addictive Behaviors. The other journals (the American Sociological Review, the American Journal of Public Health, Health Psychology, Developmental Psychology, Social Science & Medicine, Academic Pediatrics, the Journal of Adolescence, the Journal of Health Economics, and Perspectives on Sexual and Reproductive Health) published one report each.

Studies’ findings

1) Which adolescents’ risky health behaviors have been examined using SNA and the Add Health data?

The studies in this review utilized SNA to examine adolescents’ substance use—drinking, smoking, and marijuana use—and sexual behavior—specifically, sexual intercourse.

Eight studies examined adolescents’ alcohol consumption behaviors, focusing on different aspects of adolescents’ drinking (Ali & Dwyer, 2010; Clark & Lohéac, 2007; Fujimoto & Valente, 2012a; Fujimoto & Valente, 2012b; Fujimoto & Valente, 2013; Jaccard, Blanton & Dodge, 2005; Kreager & Haynie, 2011; Mundt, 2011). Among these reports, six studied adolescents’ drinking frequency as affected by best (or close) friends, peer group, affiliated members (e.g., sports and club activities), or direct (and indirect) friends (Ali & Dwyer, 2010; Clark & Lohéac, 2007; Fujimoto & Valente, 2012a; Fujimoto & Valente, 2012b; Fujimoto & Valente, 2013; Mundt, 2011). Moreover, two studies (Jaccard, Blanton & Dodge, 2005, and Kreager & Haynie, 2011) investigated adolescents’ level of drinking (specifically, bingeing) as influenced by friends.

In the six studies focused on drinking frequency, researchers used various questions from the Add Health questionnaires, including: “During the past 12 months, on how many days did you drink alcohol?” and “Think of all the times you have had a drink during the past 12 months, how many drinks did you usually have each time?”; “Over the past 12 months, on how many days have you gotten drunk or ‘very, very high’ on alcohol?”; and “During the past 12 months, how often did you get drunk?” In the two studies examining level of drinking (specifically, bingeing), researchers used the following questions: “Over the past twelve months, on how many days did you drink five or more drinks in a row?” Both the adolescents and their friends were asked this question.

Seven studies focused on cigarette use or smoking behaviors among adolescents (Alexander et al., 2001; Ali & Dwyer, 2009; Clark & Lohéac, 2007; Fujimoto & Valente, 2012a; Fujimoto & Valente, 2012b; Lakon, Hipp & Timberlake, 2010; Pollard et al., 2010). All seven examined adolescents’ frequency of smoking as influenced by various friendships, such as close or best friends, popular friends, mutual friends, or direct (and indirect) friends.

In these seven studies, authors used various questionnaire items, including: “During the past 30 days (past 12 months), on how many days did you smoke cigarettes?”; “During the past 30 days, on the days you smoked, how many cigarettes did you smoke each day?”; or “Of your 3 best friends, how many smoke at least 1 cigarette a day?”

Three studies investigated sexual behavior (intercourse) (Ali & Dwyer, 2011; Jaccard, Blanton & Dodge, 2005; Sieving et al., 2006). These studies examined the frequency of sexual intercourse as being influenced by close friends. Researchers used the questions: “Have you ever had sexual intercourse?”; “In what month and year did you have sexual intercourse most recently?”; or “If you had sexual intercourse, your friends would respect you more” from a section on “Motivations to Engage in Risky Behaviors.”

Two studies (Clark & Lohéac, 2007; Tucker et al., 2014) focused on adolescents’ marijuana use as influenced by peer groups. In this study, researchers used the questionnaire item: “During the past 30 days, how many times did you use marijuana?” In these studies, authors examined the frequency of marijuana use as influenced by adolescents’ friendships.

Even though the Add Health questionnaires have items addressing two different behaviors in tandem (e.g., sexual intercourse + drinking; and sexual intercourse + drugs), none of the reviewed studies examined more than one behavior at a time.

(2) What research findings have been identified in the literature relevant to friendship networks’ impact on adolescents’ risk behaviors?

Alcohol use

As mentioned previously, eight reviewed studies investigated the relationship between drinking alcohol and friendship networks (drinking frequency and amount of drinking). For instance, the study conducted by Fujimoto & Valente (2012a) examined how various friendship types influenced adolescents’ substance use, including drinking (frequency). Authors classified three types of friendships: mutual friendships, directional friendships, and intimate friendships (see Fig. 2 for diagrams of the three types of friendships examined by Fujimoto & Valente, 2012a). A mutual friendship was defined as reciprocated friends (knowing each other as friends). A directional friendship was defined as an unreciprocated nomination that originated either from an ego or from an alter (i.e., ego-nominating friend and alter-nominating friend). An intimate friendship was defined as closest or best friends who were being first nominated (Fig. 2). These three friendship types were based on friendship nominations that students were asked to make as they nominated five best male friends and five best female friends from the Add Health data.

Fujimoto & Valente (2012a) found mutual friends were more likely to influence their friends’ drinking behavior (frequency) than a directional friendship in the previous year (AOR = 2.07; p < 0.001). Moreover, in the directional friendships among unreciprocated alters, the authors found ego-nominating friends (Fig. 2) were slightly more influential in adolescents’ drinking behavior than alter-nominating friends (AOR = 2.02; p < 0.001). Paradoxically, for the intimate relationships (Fig. 2), the study indicated that non-best friends were more likely to influence adolescents’ past year drinking than best friends (AOR = 2.62; p < 0.001).

Fujimoto & Valente (2013) also examined the influence on adolescents’ drinking (and drinking frequency) of friends and affiliated members in sports and club activities. Adolescents were asked in which school organized clubs or sports they participated. Based on this information, authors divided activities into 12 types of sports and clubs, such as playing chess, studying French, and basketball.

Moreover, Fujimoto & Valente (2013) categorized friendships as (1) all nominated friends (adolescent nominated the alter as a friend, the equivalent to “directional friendships” in Fig. 2), and (2) only reciprocated friends (both adolescents mutually called each other friends) (Fig. 2). Fujimoto & Valente (2013) then created affiliation models based on nominated friends (i.e., General affiliation—the influence from all members’ friendships; Nominated-friends’ affiliation—the influence from adolescents who were nominated as friends; and Nonfriends’ affiliation—the influence from adolescents who were not nominated friends). Affiliation models based on reciprocated friends were also developed (i.e., General affiliation—the influence from all members’ friendships; Reciprocated-friends’ affiliation—the influence from adolescents who had at least one reciprocated friend; and Nonreciprocated-friends’ affiliation—the influence of a nonreciprocated friend).

In the nominated friends’ general affiliation model, sports members influenced adolescents’ drinking and frequency of drinking (AOR = 1.20; p < 0.05), and club members only affected adolescents’ drinking (AOR = 1.46; p < 0.01). This study additionally demonstrated that friends who drink were also more likely to affect adolescents’ drinking and drinking frequency (AOR = 1.55; p < 0.001). In the nominated-friends’ affiliation model, this study indicated club members significantly influenced adolescents’ drinking and drinking frequency (AOR = 1.15; p < 0.01) (Fujimoto & Valente, 2013). In the nonfriends’ affiliation model, club members who were not friends were more likely to affect drinking and drinking frequency of adolescents (AOR = 1.37; p < 0.01). In the reciprocated friends’ affiliation model, sports members who were mutual friends with adolescents significantly influenced drinking and frequent drinking (AOR = 1.16; p < 0.01). Moreover, club members who were mutual friends were more influential in adolescents’ drinking and frequency of drinking (AOR = 1.22; p < 0.001) than the results based on the nominated-friends affiliation model (p < 0.01). Additionally, in the nonreciprocated friends’ affiliation model, club members significantly influenced drinking and drinking frequency of adolescents (AOR = 1.25; p < 0.05) (Fujimoto & Valente, 2013).

Uniquely, this study showed that club members who have no friendship ties with others influenced other adolescents’ drinking behavior within the affiliation friendship networks. This finding can be explained by the fact that club members do not need to be intimate friends to be connected to each other, because they share many common interests and behaviors, even if they are not friends.

In another study, Jaccard, Blanton & Dodge (2005) evaluated how close friends influence adolescents’ binge drinking. In this study, close friends were defined as those who were nominated by adolescents. Authors found a statistical significance in the behavioral similarity (binge drinking) between adolescents and their close friends ((unstandardized regression coefficient) 0.15; p < 0.006). Additionally, the study demonstrated that when adolescents’ drinking behavior changed between Wave I (1995) and Wave II (1996) of data collection, their close friends’ binge drinking also changed accordingly during the same time period ((unstandardized regression coefficient) 0.12; p < 0.05).

The other five studies showed similar findings, indicating that friendships that matter, among adolescents, were more likely to exert influence upon adolescents’ drinking behavior. For instance, among these five studies, Clark & Lohéac (2007) found that “if participation in drinking alcohol by the male peer group in the same school year increases by 25%, the adolescent’s probability of drinking alcohol increases 4.5%” (p. 773). Likewise, the study by Ali & Dwyer (2010) showed that if the number of close friends who drink increased by 10%, other adolescents’ drinking would increase by 2%. Authors also found “a 10% increase in drinking among grade-level peers …associated with a 4% increase in individual drinking” (p. 340).

Tobacco use

Seven of the fifteen studies reported friendship influence on adolescents’ frequency of smoking. For instance, Ali & Dwyer (2009) categorized peer network as not only close friends who were nominated by the adolescents, but also those who were classmates and others from the same grade in school. A key finding from the study was that “having up to 25 percentage of close friends as smokers increases the probability of smoking by 5% …whereas being in a class containing up to 25% smokers increases the likelihood of smoking by 10%” (p. 406).

In another study, Fujimoto & Valente (2012b) investigated the influence of peer networks on adolescent’s substance use (smoking cigarettes and drinking alcohol), based on contagion mechanisms, in terms of cohesion and structural equivalence. Cohesion referred to relationships within a network, for which there are direct ties or exchange of influence. Structural equivalence referred to relationships among adolescents who occupy similar positions as others within friendship networks (see Fig. 3 for diagrams of cohesion and structural equivalence in a network). Authors defined peers as those who were nominated by friends. In their analysis, they utilized a network exposure model to assess both cohesion and structural equivalence measuring peers’ risk taking in terms of social distances (at four steps away from other adolescents—friends of friends of friends of friends). The results indicated “the odds ratios for cohesion exposures to smoking were statistically significant up to distance two (but not significant for distances greater than two) with the highest magnitude at distance one (OR = 1.50; p < 0.001), followed by distance two (OR = 1.40; p < 0.001)” (p. 1957).

These findings suggest that direct or indirect friends (a friend or the friend of a friend) were more likely to influence adolescents’ smoking behavior than friends at distance three or four (the friend(1)-of-a-friend(2)-of-a-friend(3), or the friend(1)-of-a-friend(2)-of-a-friend(3)-of-a-friend(4)). Moreover, the researchers found that for structural equivalence exposure to smoking, “…exposure effects were statistically significant for all distances with the highest in magnitude at distance one (OR = 1.99; p < 0.001), followed by distance two (OR = 1.83; p < 0.001), distance three (OR = 1.59; p < 0.001) and distance four (OR = 1.59; p < 0.001)” (Fujimoto & Valente, 2012b, p. 1957). These findings suggest that adolescents, who were one and two steps away in the network structure, were more likely to affect adolescents’ smoking behavior than adolescents at three or four steps away in terms of social distances.

The other five studies showed similar results, namely, that various close friendships, such as best, popular, and mutual friends, were more likely to influence adolescents’ smoking behavior than non-close friends. For instance, Alexander et al. (2001) indicated that if adolescents have best friends who are cigarette smokers, those adolescents’ probability of smoking increases two fold. Similarly, the study conducted by Pollard et al. (2010) demonstrated that “…a greater number of one’s best friends (who) smoked, and increases in the perceived number of best friends who smoked over a one-year period, were associated with greater odds of an adolescent being [a smoker]…” (p. 682).

Sexual intercourse

Three studies focused on how adolescents’ friendships influence each other’s sexual behavior (specifically, intercourse). For instance, Sieving et al. (2006) classified close friends as those who were being nominated. Close friends were based on friendship nominations by students who were asked to nominate best male and female friends. Researchers found that “…for every 1% increase in sexually experienced friends at Wave I (1995), the odds that young people initiated sex by Wave 2 (1996) increased by 1%” (p. 17).

In another study, Ali & Dwyer (2011) defined peer group as not only close friends who were nominated by adolescents, but also those who were classmates and others from the same grade in school. In this study, they found that if the number of close friends initiating sex increased by 10%, an adolescent’s probability of initiating sex would also increase by 5%.

The third study (Jaccard, Blanton & Dodge, 2005), showed similar findings, indicating that close friends were more likely to exert influence on adolescents to engage in sex. They found “…of target individuals whose closest friends engaged in sexual activity across the two waves (Waves I and II), 56% also engaged in sexual intercourse across the waves” (p. 141).

All reviewed studies examining sexual behaviors examined single behaviors; examined two or more behaviors in tandem, such as sexual activity with alcohol consumption, even though the Add Health data contains this information.

Marijuana use

Among the 15 studies reviewed, two examined the influence of friendship networks on adolescent’s marijuana use. Tucker et al. (2014) examined how different types of friendships (i.e., friend reciprocity, friend popularity, and popularity difference) affected adolescents’ marijuana use in two saturated schools samples (i.e., school 1 and 2) at Wave I and II. Friend reciprocity was defined as mutual friends (being nominated by each other as friends). Friend popularity was defined based on “the total number of friendship nominations received by a nominated friend at a given wave…” (p. 68). Popularity difference was defined as “the difference in number of friend nominations received…” (p. 68). Researchers found that two friendships (i.e., reciprocity and friend popularity) were more likely to influence adolescents’ marijuana use. Regarding friend reciprocity, authors found “in school 1, there was a significant positive interaction between friends’ influence on marijuana use and friend reciprocity…adolescents tended to adopt the (marijuana) use behaviors of their mutual friends…” (p. 72). For friend popularity, the researchers found that there was only statistical significance in school 2, indicating popular friends were more likely to influence adolescents’ marijuana use.

The second study, conducted by Clark & Lohéac (2007), examined peer group effects on adolescents’ alcohol, smoking and marijuana use. However, in this study, authors did not find that friendships influenced adolescents’ marijuana use.

(3) What is the methodological quality of this body of literature?

As previously mentioned, we assigned a methodological quality score (MQS—with a possible range of 1 to 9 points) to each reviewed study. Table 2 presents the distribution of reviewed studies in terms of the MQS criteria. The average MQS was 4.6 (SD = 1.24), with actual scores ranging from 2 to 7 points. Among the 15 studies, seven (46.7%) scored below 4.5. The main problems affecting studies scoring below average were lack of visuals for the networks and/or for the analyses.

Eleven reviewed studies (73.3%) focused on studying a single behavior (most commonly, smoking or drinking). Four studies (26.7%) analyzed two or more behaviors, such as alcohol and tobacco use, but each behavior was examined separately. None examined two or more risk behaviors simultaneously (e.g., sexual intercourse with drug or alcohol consumption).

Regarding using or adopting theoretical frameworks, while the majority (10 studies) employed a theoretical framework, five studies (33.3%) failed to do so. Among the 10 studies employing a theoretical framework, six (40%) provided some theoretical explanation or rationale and four studies (26.7%) presented a scientific or behavioral theory: Social Learning Theory and Social Comparison Theory. Eight studies (53.3%) did not test a hypothesis. Seven reports (46.7%) tested a proposed hypothesis such as “…influence from mutual friendships has stronger influence on adolescent drinking and smoking than non-mutual friendships” (Fujimoto & Valente, 2012a, p. 137) or “adolescents with higher proportions of sexually experienced close friends are more likely to initiate sexual intercourse than others” (Sieving et al., 2006, p. 14).

Only one reviewed study (6.7%) provided visual graphics for the networks examined, while four studies (26.7%) provided illustrations for how friendship influences egos and their alters. Fourteen studies (93.3%) employed and reported both descriptive and inferential statistics in their data analysis. One study (6.7%) reported only inferential statistics. More than half of the reviewed studies (66.7%) made recommendations for prevention or intervention programs, based on their network-related findings. Lakon, Hipp & Timberlake (2010), for instance, propose in their study: “These friendship pairs could be targeted for a school-based intervention, either to help both adolescents in a pair remain nonsmokers or so that they could help each other stop smoking” (p. 1226–1227).