420 likes | 438 Views
This study explores how peer network characteristics impact adolescent substance use, focusing on prominence, integration, and proximity to users. Data collected from 6th to 10th graders in North Carolina. Analysis involved network measures, recent substance use, and demographics.
E N D
Social Network Characteristics and Substance Use: Findings from a Longitudinal Study of Adolescent Health Risk Behaviors
Susan T. Ennett,1 Karl E. Bauman,1Andrea Hussong,2 Robert Faris,3Vangie A. Foshee,1 Patrick Curran,2 Rob DuRant4 1Department of Health Behavior and Health Education 2Department of Psychology 3Department of Sociology The University of North Carolina at Chapel Hill 4Departments of Pediatrics and Public Health Sciences Wake Forest University
Hypothesis 1 Adolescents with a) more peer network prominence, b) stronger peer network integration, and c) further distanceto substance users will be less likely to have used substances recently than adolescents with less peer network prominence, weaker peer network integration, and closer distance to substance users Substances: cigarettes, alcohol, or marijuana
Hypothesis 2 Adolescents with a) more peer network prominence, b) stronger peer network integration, and c) further distance to substance users will increase recent substance use less with time than adolescents with less prominence, weaker integration, and closer proximity to substance users Substances: cigarettes, alcohol, or marijuana
Data The Public School Systems in 3 Central North Carolina Counties (Vance, Person, Moore) Wave 1: 6th, 7th, and 8th Graders, Spring 2002 Surveyed again every 6 months until Spring 2004 when in grades 8 - 10 (total of 5 waves) Present analyses: 1st 3 waves
Data (cont.) Eligible N (Wave 1) = 5906* Wave 1: 88.4% (N = 5220) completed 1 hour questionnaires, most in classrooms _______________________________ *Ineligible: exceptional children in self-contained classrooms and students without English proficiency
Data (cont.) Sample in present analyses: Age 11 through 15 and followed through Wave 3 (84% of eligible, N=4961, Observations=12,135) Dropped 259 Wave 1 respondents: Alternative school students (N=65) Outside age range (N=24) Missing information (N=170)
Data for Network Measures Write first names of up to 5 closest friends, starting with best friend Assign unique number to each friend from school directory Not in directory: assign “0000”
Network Boundaries School and Grade (N=26)
Recent cigarette use in a 6th grade network (blue=nonuser, yellow=user,gray=missing info)
Recent alcohol use in a 6th grade network (blue=nonuser, yellow=user,gray=missing info)
Recent marijuana use in a 6th grade network (blue=nonuser, yellow=user,gray=missing info)
Prominence in Network 1. Normed indegree: No. of nominations received by ego divided by no. of possible nominations (UCINET) 2. Betweeness centrality: Extent to which ego links unlinked pairs of friends (UCINET)
Network Integration 3. Social position: group member, bridge, or isolate (SAS IML, James Moody) 4. Density of network neighborhood: No. of friendship ties among alters divided by total number of possible ties (UCINET) 5. Number of out-of-network (grade) nominations (UCINET)
Proximity to User 6. Number of users in adolescent’s neighborhood, excluding ego (UCINET) 7.Number of links to nearest substance user (UCINET)
Also measured: Recent (within 3 months) cigarette, alcohol, and marijuana use. Binary. Demographics: grade in school, sex, race/ethnicity (white vs. other), family structure (2 parent vs. other)
Sample Characteristics Grade in school: % 6.5 35.6 7.5 33.3 8.5 31.1 Sex: 50.8% female Race/ethnicity: 52.5% white 2 parent families: 67.8%
Analysis Strategy Cohort sequential design 12,135 observations (4961 respondents, 3 waves) Age rather than wave of data collection to indicate time
Analysis Strategy (cont.) 3-level (time, school, individual) Hierarchical Generalized Linear Models Nested repeated measures within adolescents and adolescents within schools Separate for each network variable and substance Always control for demographic variable when significantly related to substance use
Slopes Intercepts Age
Hypothesis 1 Adolescents with a) more peer network prominence, b) stronger peer network integration, and c) further distanceto substance users will be less likely to have used substances recently than adolescents with less peer network prominence, weaker peer network integration, and closer distance to substance users Substances: cigarettes, alcohol, or marijuana
Tests of Hypothesis 1 Intercept (age 11) association between network variable and age 11 substance use
Hypothesis 1 Findings Social network variables by substance use:Beta(SE) Cigarettes Alcohol Marijuana Prominence Normed 3.94 9.03**0 6.07 indegree (1.27) (3.43) (5.16) Betweeness .01 .02*0 .03 centrality (.02) (.01) (.02) *p<.05, **p<.01, 0opposite to hypothesized direction
Hypothesis 1 Findings (cont.) Cigarettes Alcohol Marijuana Integration Social Position Isol. vs. Gp. 1.99* 2.57** 1.44 (1.01) (.92) (1.34) Brid. vs. Gp. .40 -.12 .65 (.34) (.30) (.46) Density -1.28 -1.77 -2.37 (1.06) (.95) (1.53) Out-of-network .29* .30* -.14 nominations (.14) (.13) (.19) *p<.05, **p<.01
Hypothesis 1 Findings (cont.) Cigarettes Alcohol Marijuana Proximity No. of friend .93*** .29 1.29*** users (.17) (.15) (.23) Distance to -.63* -.23 -1.19*** nearest user (.27) (.20) (.32) ***p<.001
Hypothesis 2 Adolescents with a) more peer network prominence, b) stronger peer network integration, and c) further distance to substance users will increase recent substance use less with time than adolescents with less prominence, weaker integration, and closer proximity to substance users Substances: cigarettes, alcohol, or marijuana
Tests of Hypothesis 2 Hypothesis 2: Network variables predict different slopes of substance use Social network variable by substance use by time
Hypothesis 2 Findings 24 interactions tested, 2 significantly different slopes: Position (isolate vs. group member) and alcohol use Beta = - 1.25, SE=.40, p<.001 (opposite to hypothesized direction) Number of marijuana using friends and marijuana use Beta = -0.18, p<.05)
Conclusions Integration (social position and out-of-network nominations) and proximity (number of users and distance to users) are implicated in substance use. Prominence (normed indegree and betweeness centrality) are less implicated in substance use. Intercept and slope findings can yield substantially different impressions
Conclusions (cont.) Three statistically significant findings opposite to hypothesized direction Findings vary by substance: No hypothesis accepted for all substances No corrections for multiple statistical tests Findings could change as the cohort ages Findings could differ for other network measures, later-measured network measures, and for other substances
Network variables(Prominence) Variable Mean SD Min Max 1.Normed indegree .06 .04 0 .29 2.Betweeness 9.96 11.95 0 140.07 centrality
Network variables(Integration) Variable 3.Social position: 53% Gp, 44% Brid, 3% Isol (N=4961) Mean SD Min Max 4. Density .18 .17 0 1.00 5. Out-of-network nominations .80 1.16 0 5
Network variables(Proximity to user) Variable Mean SD Min Max 6.No. users Cigarette .75 1.03 0 7 Alcohol .83 1.02 0 7 Marijuana .67 1.02 0 8 7. Distance Cigarette 1.71 .82 1 5 Alcohol 1.63 .80 1 5 Marijuana 1.84 .91 1 6
Recent substance use by age (% ) Age Cigarette Alcohol Marijuana N (Obs) 11 6 7 2 808 12 11 11 5 3306 13 17 18 11 3947 14 23 27 18 3116 15 29 31 23 958
Correlations between Network Variables Prominence Normed Indegree x Betweeness Centrality: .51 Integration Social Position x Density: .36 Social Position x No. Out-of network Nominations: .13 Density x No. out of network Nominations -.32
Correlations between Network Variables (cont.) Proximity No. Users x Distance: Cigarettes: -.64 Alcohol: -.64 Marijuana: -.61
Analysis Strategy (cont.) Age No. of Observations 11 808 12 3306 13 3947 14 3116 15 958 Total 12135