Advertisement

Prevalence and Patterns of Polysubstance Use in a Nationally Representative Sample of 10th Graders in the United States

  • Kevin P. Conway
    Correspondence
    Address correspondence to: Kevin P. Conway, Ph.D., Division of Epidemiology, Services, and Prevention Research, National Institute on Drug Abuse, 6001 Executive Boulevard, Suite 5185, Bethesda, MD 20892-9589.
    Affiliations
    Division of Epidemiology, Services, and Prevention Research, National Institute on Drug Abuse, National Institutes of Health, Department of Health and Human Services, Bethesda, Maryland
    Search for articles by this author
  • Genevieve C. Vullo
    Affiliations
    Division of Epidemiology, Services, and Prevention Research, National Institute on Drug Abuse, National Institutes of Health, Department of Health and Human Services, Bethesda, Maryland

    Kelly Government Solutions, Bethesda, Maryland
    Search for articles by this author
  • Brandon Nichter
    Affiliations
    Division of Epidemiology, Services, and Prevention Research, Eunice Kennedy Shriver National Institute of Child Health and Human Development, National Institutes of Health, Department of Health and Human Services, Bethesda, Maryland
    Search for articles by this author
  • Jing Wang
    Affiliations
    Division of Epidemiology, Services, and Prevention Research, Eunice Kennedy Shriver National Institute of Child Health and Human Development, National Institutes of Health, Department of Health and Human Services, Bethesda, Maryland
    Search for articles by this author
  • Wilson M. Compton
    Affiliations
    Division of Epidemiology, Services, and Prevention Research, National Institute on Drug Abuse, National Institutes of Health, Department of Health and Human Services, Bethesda, Maryland
    Search for articles by this author
  • Ronald J. Iannotti
    Affiliations
    Division of Epidemiology, Services, and Prevention Research, Eunice Kennedy Shriver National Institute of Child Health and Human Development, National Institutes of Health, Department of Health and Human Services, Bethesda, Maryland
    Search for articles by this author
  • Bruce Simons-Morton
    Affiliations
    Division of Epidemiology, Services, and Prevention Research, Eunice Kennedy Shriver National Institute of Child Health and Human Development, National Institutes of Health, Department of Health and Human Services, Bethesda, Maryland
    Search for articles by this author

      Abstract

      Purpose

      The current study examines the prevalence and demographic correlates of self-reported substance use and identifies subgroups of polysubstance users among a cohort of United States 10th-grade students.

      Methods

      A nationally representative school-based cohort of United States 10th-grade students completed the NEXT Generation Health Study baseline survey in spring 2010 (N = 2,524).

      Results

      Past-year use of marijuana was most common among illicit drugs (26%), followed by misuse of medication (9%) and use of other illicit drugs (8%). During the past month, alcohol use was reported by more than one third (35%), binge drinking by 27%, and cigarette smoking by 19%. Results further show that substance use varied somewhat by demographic characteristics. Results from the latent class analysis of polysubstance use indicated a four-class solution as the best-fitting model; class 1 (59%) included the nonuser group; class 2 (23%) comprised the predominant alcohol user group; class 3 (11%) formed the predominant marijuana user group; and class 4 (8%) was characterized as the predominant polysubstance user group. Somatic and depressive symptoms varied significantly by class membership, with predominant polysubstance users reporting elevated levels of somatic and depressive symptoms.

      Conclusions

      The findings from this national study of 10th-grade students indicate high rates of substance and polysubstance use. The high level of depressive and somatic symptoms among polysubstance users indicates the need for mental health screening and referral.

      Keywords

      Implications and Contribution
      These findings underscore the ubiquitous yet heterogeneous nature of adolescent substance use and reveal that the multiple classes of substance users differ in terms of use patterns, somatic reports, and depressive symptomatology. The differences among classes implicate polysubstance use as a broad indicator of severity deserving research and clinical attention.
      National epidemiological studies estimate that most adolescents will engage in some form of substance use by the time they graduate from high school. The 2011 Monitoring the Future Survey (MTF) found that 26.4% have tried an illicit drug by eighth grade, 40.8% by 10th grade, and 51.8% by 12th grade [
      • Johnston L.D.
      • O'Malley P.M.
      • Bachman J.G.
      • Schulenberg J.E.
      Monitoring the Future national results on adolescent drug use: overview of key findings, 2011.
      ]. The findings of MTF further show that adolescent users tend to report using a variety of specific substances at an early age. Among 10th graders, for example, rates of past-year use were 49.8% for alcohol, 28.8% for marijuana/hashish, 6.6% for amphetamines, and a range from 2.6% to 5.9% for misuse of medications (e.g., oxycodone HCl, cough medicine) [
      • Johnston L.D.
      • O'Malley P.M.
      • Bachman J.G.
      • Schulenberg J.E.
      Monitoring the Future national results on adolescent drug use: overview of key findings, 2011.
      ].
      Polysubstance use, defined as using multiple substances within a specified period of time, remains an understudied characteristic of youthful substance use [
      • Mitchell C.M.
      • Plunkett M.
      The latent structure of substance use among American Indian adolescents: An example using categorical variables.
      ,
      • Connell C.
      • Gilreath T.
      • Aklin W.
      • Brex R.
      Social-ecological influences on patterns of substance use among non-metropolitan high school students.
      ,
      • Whitesell N.R.
      • Beals J.
      • Mitchell C.M.
      • et al.
      Latent class analysis of substance use: Comparison of two American Indian reservation populations and a national sample.
      ,
      • Dierker L.C.
      • Vesel F.
      • Sledjeski E.M.
      • et al.
      Testing the dual pathway hypothesis to substance use in adolescence and young adulthood.
      ,
      • Connell C.M.
      • Gilreath T.D.
      • Hansen N.B.
      A multiprocess latent class analysis of the co-occurrence of substance use and sexual risk behavior among adolescents.
      ]. Previous findings on polysubstance use come largely from studies applying latent class analyses (LCAs) to identify multiple classes or groups of adolescents, defined by their patterns of substance use. Among the school-based surveys of adolescents in middle or high school, between 15% and 39% of adolescents could be classified as polysubstance users. Precise estimates varied by sample, reporting period, and definition of polysubstance use [
      • Mitchell C.M.
      • Plunkett M.
      The latent structure of substance use among American Indian adolescents: An example using categorical variables.
      ,
      • Connell C.
      • Gilreath T.
      • Aklin W.
      • Brex R.
      Social-ecological influences on patterns of substance use among non-metropolitan high school students.
      ,
      • Whitesell N.R.
      • Beals J.
      • Mitchell C.M.
      • et al.
      Latent class analysis of substance use: Comparison of two American Indian reservation populations and a national sample.
      ,
      • Dierker L.C.
      • Vesel F.
      • Sledjeski E.M.
      • et al.
      Testing the dual pathway hypothesis to substance use in adolescence and young adulthood.
      ,
      • Connell C.M.
      • Gilreath T.D.
      • Hansen N.B.
      A multiprocess latent class analysis of the co-occurrence of substance use and sexual risk behavior among adolescents.
      ]. Results from the 2005 Youth Risk Behavior Surveillance (YRBS) [
      • Connell C.M.
      • Gilreath T.D.
      • Hansen N.B.
      A multiprocess latent class analysis of the co-occurrence of substance use and sexual risk behavior among adolescents.
      ] depict four classes based on past-month use among grades 9–12, as follows: alcohol experimenters (38%), nonusers/abstainers (27%), occasional polysubstance users (23%), and frequent polysubstance users (13%). The 2001–2002 Add Health data [
      • Dierker L.C.
      • Vesel F.
      • Sledjeski E.M.
      • et al.
      Testing the dual pathway hypothesis to substance use in adolescence and young adulthood.
      ] identified five classes of students in grades 7–12 based on use during the past month or year: low use (55%), alcohol only (15%), alcohol or marijuana (8%), cigarettes (8%), and three-substance use (14%) [
      • Dierker L.C.
      • Vesel F.
      • Sledjeski E.M.
      • et al.
      Testing the dual pathway hypothesis to substance use in adolescence and young adulthood.
      ]. Given that medication is now the most commonly abused substance after alcohol, marijuana, and tobacco [
      • Johnston L.D.
      • O'Malley P.M.
      • Bachman J.G.
      • Schulenberg J.E.
      Monitoring the Future national results on adolescent drug use: overview of key findings, 2011.
      ,

      Substance Abuse and Mental Health Services Administration, Results from the 2010 National Survey on Drug Use and Health: Summary of National Findings, NSDUH Series H-41, HHS Publication No. (SMA) 11-4658. Rockville, MD: Substance Abuse and Mental Health Services Administration, 2011.

      ], it seems both timely and useful that these substances be examined separately to better characterize contemporary patterns of use.
      In addition to examining prevalence of classes defined by polysubstance use, several studies have linked adolescent polysubstance use with deleterious outcomes, including substance dependence [
      • Whitesell N.R.
      • Beals J.
      • Mitchell C.M.
      • et al.
      Latent class analysis of substance use: Comparison of two American Indian reservation populations and a national sample.
      ], smoking in adulthood [
      • Dierker L.C.
      • Vesel F.
      • Sledjeski E.M.
      • et al.
      Testing the dual pathway hypothesis to substance use in adolescence and young adulthood.
      ], and risky sexual behavior [
      • Connell C.M.
      • Gilreath T.D.
      • Hansen N.B.
      A multiprocess latent class analysis of the co-occurrence of substance use and sexual risk behavior among adolescents.
      ]. Furthermore, many studies have reported that substance abuse commonly co-occurs with mental health problems in epidemiologic samples of adults [
      • Kessler R.C.
      • Crum R.M.
      • Warner L.A.
      • et al.
      Lifetime co-occurrence of DSM-III-R alcohol abuse and dependence with other psychiatric disorders in the National Comorbidity Survey.
      ,
      • Conway K.P.
      • Compton W.
      • Stinson F.S.
      • Grant B.F.
      Lifetime comorbidity of DSM-IV mood and anxiety disorders and specific drug use disorders: Results from the National Epidemiologic Survey on Alcohol and Related Conditions.
      ,
      • Regier D.A.
      • Farmer M.E.
      • Rae D.S.
      • et al.
      Comorbidity of mental disorders with alcohol and other drug abuse: Results from the Epidemiologic Catchment Area (ECA) Study.
      ,
      • Compton W.M.
      • Conway K.P.
      • Stinson F.S.
      • et al.
      Prevalence, correlates, and comorbidity of DSM-IV antisocial personality syndromes and alcohol and specific drug use disorders in the United States: Results from the national epidemiologic survey on alcohol and related conditions.
      ,
      • Merikangas K.R.
      • Mehta R.L.
      • Molnar B.E.
      • et al.
      Comorbidity of substance use disorders with mood and anxiety disorders: Results of the International Consortium in Psychiatric Epidemiology.
      ] and adolescents [
      • Costello E.J.
      • Mustillo S.
      • Erkanli A.
      • et al.
      Prevalence and development of psychiatric disorders in childhood and adolescence.
      ,
      • Merikangas K.R.
      • He J.P.
      • Burstein M.
      • et al.
      Lifetime prevalence of mental disorders in U.S. adolescents: Results from the National Comorbidity Survey Replication–Adolescent Supplement (NCS-A).
      ,
      • Marmorstein N.R.
      • Iacono W.G.
      • Malone S.M.
      Longitudinal associations between depression and substance dependence from adolescence through early adulthood.
      ], with psychopathology onset often predating substance abuse [
      • Costello E.J.
      • Mustillo S.
      • Erkanli A.
      • et al.
      Prevalence and development of psychiatric disorders in childhood and adolescence.
      ,
      • Merikangas K.R.
      • He J.P.
      • Burstein M.
      • et al.
      Lifetime prevalence of mental disorders in U.S. adolescents: Results from the National Comorbidity Survey Replication–Adolescent Supplement (NCS-A).
      ,
      • Burke Jr., J.D.
      • Burke K.C.
      • Rae D.S.
      Increased rates of drug abuse and dependence after onset of mood or anxiety disorders in adolescence.
      ]. However, past research has been limited in two key ways. First, most studies focus on a specific substance (e.g., cannabis) or on substance use at an aggregate level (e.g., any substance use), rather than a more detailed characterization based on profiles of polysubstance use. Second, most studies of comorbidity report disorder-level associations between specific substance use disorders and specific psychiatric disorders [
      • Conway K.P.
      • Compton W.
      • Stinson F.S.
      • Grant B.F.
      Lifetime comorbidity of DSM-IV mood and anxiety disorders and specific drug use disorders: Results from the National Epidemiologic Survey on Alcohol and Related Conditions.
      ,
      • Regier D.A.
      • Farmer M.E.
      • Rae D.S.
      • et al.
      Comorbidity of mental disorders with alcohol and other drug abuse: Results from the Epidemiologic Catchment Area (ECA) Study.
      ,
      • Compton W.M.
      • Conway K.P.
      • Stinson F.S.
      • et al.
      Prevalence, correlates, and comorbidity of DSM-IV antisocial personality syndromes and alcohol and specific drug use disorders in the United States: Results from the national epidemiologic survey on alcohol and related conditions.
      ] or between any substance use disorder and specific psychiatric disorders [
      • Kessler R.C.
      • Crum R.M.
      • Warner L.A.
      • et al.
      Lifetime co-occurrence of DSM-III-R alcohol abuse and dependence with other psychiatric disorders in the National Comorbidity Survey.
      ,
      • Merikangas K.R.
      • Mehta R.L.
      • Molnar B.E.
      • et al.
      Comorbidity of substance use disorders with mood and anxiety disorders: Results of the International Consortium in Psychiatric Epidemiology.
      ]. Although polysubstance use has been associated with depressive symptoms in college students [
      • McCabe S.E.
      • Cranford J.A.
      • Morales M.
      • Young A.
      Simultaneous and concurrent polydrug use of alcohol and prescription drugs: Prevalence, correlates, and consequences.
      ], to our knowledge only two studies examined polysubstance use and depressive symptoms among adolescents [
      • Connell C.
      • Gilreath T.
      • Aklin W.
      • Brex R.
      Social-ecological influences on patterns of substance use among non-metropolitan high school students.
      ,
      • Dierker L.C.
      • Vesel F.
      • Sledjeski E.M.
      • et al.
      Testing the dual pathway hypothesis to substance use in adolescence and young adulthood.
      ]. Additional research is needed not only on the association between adolescent polysubstance use and depressive symptomatology in particular, but also on somatic reports such as headaches, stomachaches, and other musculoskeletal pains, which are common expressions of underlying depression in adolescents [
      • Bernstein G.A.
      • Massie E.D.
      • Thuras P.D.
      • et al.
      Somatic symptoms in anxious-depressed school refusers.
      ,
      • Egger H.L.
      • Costello E.J.
      • Erkanli A.
      • Angold A.
      Somatic complaints and psychopathology in children and adolescents: Stomach aches, musculoskeletal pains, and headaches.
      ].
      The purposes of the current study are threefold: (1) to examine the prevalence and demographic correlates of substance use; (2) to identify classes and subgroups according to substance use patterns; and (3) to examine associations between classes and subgroups and depressive and somatic symptoms.

      Methods

      The NEXT Generation Health Study is a longitudinal study of a cohort of 10th-grade United States (U.S.) students beginning in spring 2010. School-based assessments included a survey of health behaviors with a focus on four areas: obesity and obesity-related behaviors (e.g., physical activity, sedentary behavior, diet, and sleep), substance use, dating violence, and driving.

      Sample and procedure

      We recruited a nationally representative cohort of U.S. students in grade 10 using a multistage stratified design. Primary-sampling units consisted of school districts or groups of school districts stratified across the nine U.S. Census divisions. Within this sampling framework, we randomly selected and formally recruited 137 schools; 80 (58.4%) agreed to participate. We randomly selected 10th-grade classes within each recruited school and recruited 3,796 students to participate; we obtained youth assent and parental consent from 2,619 students (69.0%). Of those consented, 2,524 completed the Wave 1 (baseline) survey, for an overall completion rate of 66.5%. We oversampled African-American students to provide better prevalence estimates and permit comparisons across subpopulations; given the prevalence of Hispanic youth in this sample, the cohort already included an adequate sample of Hispanic youth to meet these criteria. In Wave 1, trained research assistants in the 10th-grade classrooms administered confidential self-report surveys. The Institutional Review Board of the Eunice Kennedy Shriver National Institute of Child Health and Human Development reviewed and approved the study protocol.

      Measures

      Demographics

      Students provided information about race/ethnicity and family composition. Parents of youth provided parental education via a paper survey completed when they gave informed consent for their child's participation. The item, obtained from the Add Health study, indicates the highest grade completed at the time of the interview.

      Substance use

      Prevalence in the past month was reported separately for cigarette smoking, alcohol use, and binge drinking. On a 7-point scale (1 = never; 2 = once or twice; 3 = three to five times; 4 = six to nine times; 5 = 10 to 19 times; 6 = 20–39 times; and 7 = ≥40 times), respondents reported on how many occasions in the past 30 days they had smoked cigarettes and on how many occasions they drank alcohol. To assess binge drinking, on a 6-point scale (1 = none; 2 = 1; 3 = 2; 4 = 3–5; 5 = 6–9; 6 = ≥10 times) respondents reported how many times in the past month they had four or more (females) or five or more (males) drinks in a row on an occasion. The alcohol questions came from the European School Survey Project on Alcohol and Other Drugs and the National Longitudinal Study of Adolescent Health study (Add Health), with the measure of binge drinking conforming to the NIAAA standard [

      National Institute on Alcohol Abuse and Alcoholism. NIAAA Newsletter (NIH publication number 04-5346); 01.01.2004.

      ]. Prevalence in the past year was reported separately for marijuana/hashish use, medication use to get high, and other illicit drug use. On a 7-point scale (1 = never; 2 = once or twice; 3 = 3–5 times; 4 = six to nine times; 5 = 10 to 19 times; 6 = 20–39 times; 7 = ≥40 times), respondents reported how often in the past 12 months they had used marijuana, ecstasy, amphetamines (methamphetamine, ice, glass, and speed), opiates (heroin, morphine, and smack), medication to get high, cocaine, glue or solvents, lysergic acid diethylamide, anabolic steroids, and other drugs (open-ended). For reporting prevalence of substance use and conducting LCAs, we created six dichotomous variables (never vs. ever used during past month or past year) on cigarette smoking, alcohol use, binge drinking, marijuana use, medication misuse, and other illicit drug use.

      Depressive and somatic symptomatology

      Depressive symptomatology was assessed by a self-reported screener for Diagnostic and Statistical Manual of Mental Disorders–3rd Edition, Revised depression [

      Dahlberg L, Toal S, Swahn M, Behrens C. Measuring violence-related attitudes, behaviors, and influences among youths: a compendium of assessment tools. 2. 01.01.2005. Atlanta (GA): Centers for Disease Control and Prevention, National Center for Injury Prevention and Control.

      ]. On a 5-point scale (1 = never; 2 = seldom; 3 = sometimes; 4 = often; and 5 = always), respondents reported how often they endorsed the following six items over the past 30 days: feeling very sad, feeling grouchy/irritable or in a bad mood, feeling hopeless about the future, feeling like not eating or eating more than usual, sleeping a lot more or a lot less than usual, and experiencing difficulty concentrating on schoolwork. The internal reliability of this scale was high (Cronbach α = .82). In addition, somatic symptoms were assessed through self-report [
      • Hetland J.
      • Torscheim T.
      • Aura L.
      Subjective health complaints in adolescence: Dimensional structure and variation across gender and age.
      ]. On a 5-point scale (1 = rarely or never; 2 = about every month; 3 = about every week; 4 = more than once a week; and 5 = about every day), respondents reported how often they endorsed the following four items over the past 6 months: having a headache, having a stomachache, having a backache, and feeling dizzy. The internal reliability of this scale was moderate (Cronbach α = .70).

      Analytical approach

      We conducted a series of LCAs in two steps. The first step was to choose the optimal number of classes by specifying separate LCA models with various numbers of classes. We tested the models with and without demographic covariates, which included gender, race/ethnicity, and parent education. We first conducted the most parsimonious one-class model, and then estimated successive models with two to five classes. For LCA with covariates, with class membership as outcome and demographic variables as predictors, the model is analogous to a logistic or multinomial logistic regression. We compared models on a series of statistical fit indices, as well as conceptual considerations: (1) We compared information criteria, including the Akaike Information Criterion [
      • Akaike H.
      On the entropy maximization principle.
      ], Bayesian Information Criterion (BIC) [
      • Raftery A.
      Bayesian model selection in social research.
      ], and sample size adjusted BIC (ABIC); (2) Pearson and likelihood ratio chi-square statistics were reported, with nonsignificant p values indicating good fit; (3) the selected model should have adequate classification quality, as shown by relative entropy and average classification probabilities (ACPs) with entropy (ACP values near 1 indicate high certainty and reliability in the classification); and (4) we considered the practical interpretability of the classes in comparing models with similarly adequate fit statistics. More weight was given to BIC and ABIC in choosing the number of classes, because a recent simulation study has shown that BIC performs better than other information criteria and likelihood ratio tests in identifying the appropriate number of latent classes [
      • Nylund K.L.
      • Asparouhov T.
      • Muthén B.O.
      Deciding on the number of classes in latent class analysis and growth mixture modeling: A Monte Carlo simulation study.
      ]. After we chose the appropriate number of classes, we compared depressive and somatic symptoms across the extracted latent classes. We used the posterior probability-based multiple imputations to conduct paired mean comparisons.
      We used the statistical software package Mplus 6.1 [

      Muthén LK, Muthén BO. Mplus user’s guide. 5th ed; 1998.

      ] for model-fitting. To accommodate the complex sampling structure of the data, we examined LCA models with stratification, cluster, and sampling weights. Furthermore, Mplus enabled us to make use of all available data, including cases with some missing responses, through estimation by Full Information Maximum Likelihood [
      • Schafer J.L.
      • Graham J.W.
      Missing data: Our view of the state of the art.
      ]. We compared depressive and somatic symptoms across latent class membership with the AUXILIARY option in Mplus.

      Results

      Sample

      The demographic characteristics of the sample appear in Table 1. There were slightly more females than males: 54.4% female and 45.6% male. Race/ethnicity, parent education, family structure, and region of residence in the U.S. are also reported.
      Table 1Sample characteristics of NEXT Generation Health Study (N = 2,524)
      CategoriesnWeighted %SE
      Gender (N = 2,519)
       Male1,13245.61.7
       Female1,38754.41.7
      Race/ethnicity (N = 2,510)
       White1,09257.95.5
       African-American48517.53.7
       Hispanic80119.63.3
       Other
      “Other” includes Asians, Hawaiians/Pacific Islanders, and American Indians/Alaskan Natives.
      1324.91.1
      Asian791.7.5
      Hawaiian/Pacific Islander13.4.2
      American Indian/Alaskan Native402.81.0
      Parent education (highest of both) (N = 2,361)
       Less than high school diploma3348.42.0
       High school/graduate equivalency diploma60225.12.1
       Some college or technical school86539.71.7
       Bachelor's degree32114.41.6
       Graduate degree23912.42.0
      Family structure (N = 2,524)
       Both biological1,32052.12.2
       Stepparents41919.21.3
       Single parent–mother45716.81.6
       Single parent–father592.7.6
       Others2699.21.1
      Region (N = 2,524)
       South1,09839.33.4
       Northeast36817.04.1
       West68224.72.8
       Midwest37619.03.3
      SE = standard error.
      a “Other” includes Asians, Hawaiians/Pacific Islanders, and American Indians/Alaskan Natives.
      Table 2 lists the prevalence of substance use, by sex, race, parent education, family structure, and geographic region. Of the 2,524 respondents, 29% reported using any illicit drugs in the past year (data not shown). During the past month, 35% reported alcohol use, 27% binge drinking, and 19% cigarette smoking. Past-year use of marijuana was 26%, misuse of medication was 9%, and other illicit drugs (e.g., ecstasy, amphetamine, lysergic acid diethylamide, opiates, cocaine, anabolic steroids, glue/solvents) was 8%.
      Table 2Prevalence of illicit drug use (past 12 months) and nicotine or alcohol use and binge drinking (past month) in U.S. adolescents in the NEXT Generation Health Study (N = 2,524)
      CategoryPast yearPast 30 days
      Marijuana (N = 2,485)Medication to get high (N = 2,482)Other illicit drug (N = 2,482)Smoked cigarettes (N = 2,498)Drank alcohol (N = 2,497)Binge drinking (on one occasion) (N = 2,477 )
      % (SE)% (SE)% (SE)% (SE)% (SE)% (SE)
      Total (N = 2,524)26.3 (2.2)8.7 (1.3)8.3 (.97)18.9 (2.6)35.3 (2.3)27.2 (1.9)
      Gender
       Male (n = 1,132)30.0 (3.5)a10.5 (2.2)6.9 (1.3)20.0 (3.0)34.2 (2.7)28.8 (2.3)
       Female (n =1,386)23.1 (1.8)b7.2 (1.3)9.5 (1.3)17.9 (2.7)36.1 (2.9)25.9 (2.4)
      Race/ethnicity
       White (n =1,092)25.9 (2.7)8.7 (1.2)6.7 (1.1)a21.3 (2.6)36.4 (3.0)29.4 (2.4)
       African-American (n = 485)20.5 (3.2)6.7 (3.7)8.2 (2.7)a,b9.7 (3.2)31.8 (5.6)18.6 (4.4)
       Hispanic (n = 801)32.1 (4.6)10.1 (3.0)11.7 (1.6)b18.8 (6.2)34.5 (4.3)27.7 (4.2)
       Other (n = 132)28.1 (7.6)10.7 (6.6)14.4 (5.7)b23.5 (8.3)39.0 (10.2)30.1 (7.5)
      Parent education
       Less than high school diploma (n = 334)26.1 (4.3)a,b4.0 (1.9)11.0 (2.9)16.4 (7.0)a,b33.9 (5.4)a,b26.1 (4.9)a,b
       High school or graduate equivalency diploma (n = 602)30.5 (3.2)a9.3 (2.1)7.7 (2.0)25.3 (3.9)a34.6 (2.4)a,b30.4 (3.6)a
       Some college or technical school (n = 865)27.1 (3.8)a,b8.1 (1.8)10.3 (2.5)18.2 (3.2)b35.2 (3.1)a,b27.4 (2.9)a,b
       Bachelor's degree (n = 321)25.6 (7.3)a,b13.1 (4.2)5.1 (1.5)16.8 (3.7)a,b41.3 (6.3)a30.0 (6.0)a
       Graduate degree (n = 239)15.0 (3.8)b8.1 (3.3)3.6 (1.7)14.4 (3.2)b31.4 (4.7)b17.6 (3.4)b
      Family structure
       Both biological (n = 1,320)22.2 (2.1)a6.9 (1.5)5.6 (1.0)a15.4 (1.7)a33.4 (2.6)23.9 (2.5)a
       Stepparents (n = 419)32.7 (2.7)b12.3 (2.3)12.8 (2.4)b,c20.3 (2.5)b37.6 (3.2)34.4 (3.7)b
       Single parent–mother (n = 457)25.9 (4.1)ab9.0 (2.5)6.4 (1.7)a20.0 (4.3)a,b36.6 (4.7)29.1 (4.3)a,b
       Single parent–father (n = 59)23.0 (9.6)a,b8.8 (6.9)11.5 (7.0)a–c28.6 (5.3)b31.0 (2.9)20.2 (7.0)a,b
       Others (n = 269)37.3 (7.4)b11.2 (3.9)16.5 (5.0)c30.6 (9.2)b40.6 (5.7)29.3 (5.9)a,b
      Region
       South (n = 1,098)22.9 (3.8)6.8 (1.5)8.5 (1.3)18.6 (3.8)a,b33.7 (3.0)25.1 (3.1)
       Northeast (n = 368)23.4 (3.1)6.6 (1.9)5.0 (2.3)11.7 (1.9)a37.1 (5.7)25.8 (5.1)
       West (n= 682)34.8 (5.0)12.0 (3.5)11.7 (2.2)20.4 (7.4)a,b36.5 (6.9)29.6 (4.7)
       Midwest (n = 376)24.8 (4.5)10.4 (3.5)6.5 (2.0)23.7 (4.2)b35.6 (1.5)29.8 (1.7)
      SE = standard error.
      Prevalence was a weighted percentage controlling for the complex survey design. For each demographic characteristic, we conducted separate logistic regression models for each of the six substance use variables. We conducted pairwise contrast analyses when we detected an overall difference. “Other” for race/ethnicity includes Asians, Hawaiians/Pacific Islanders, and American Indians/Alaskan Natives. We conducted analyses with SAS, version 9.2.
      Within each model, two categories with different superscripts (denoted a, b, c) were significantly different in using the particular substance, according to Wald chi-square test statistics at α = .05. Categories with significant differences are bolded. For example, there was a significant racial/ethnic difference with regard to prevalence of other illicit drug. Further contrast analyses showed a significant difference between whites and Hispanics, and between whites and other racial/ethnic groups.
      Other findings indicate that substance use varied somewhat by demographic characteristics. Marijuana use was more prevalent among males than females. Compared with whites, Hispanics and those in the “Other” race/ethnicity category (including Asians, Hawaiians/Pacific Islanders, and American Indians/Alaska Natives) were more likely to use other illicit drugs. There were several differences by parent education category. For example, the prevalence of marijuana use, cigarette smoking, and binge drinking was lower among children of parents with graduate degrees compared with high school diploma or graduate equivalency diploma. By family structure, significantly lower prevalence was found among children with biological parents than children whose family structure falls into other categories, including stepparents (marijuana use, other illicit drug use, cigarette smoking, binge drinking), single father (cigarette smoking), and other family structure (marijuana use, other illicit drug use, and cigarette smoking). In addition, children from single-mother families reported lower use of other illicit drugs compared with stepparents and other family structure. Last, prevalence of smoking cigarettes was lower among those living in the Northeast than the Midwest.

      Latent classes

      Table 3 lists results of fit statistics for LCA models with one to five classes, and for the LCA with covariates models with two to five classes. Pearson and likelihood chi-square statistics test departure of the model from the observed data, and were significant for models with one to three classes, but not for models with four or five classes. In terms of the information criteria, the four-class solution without covariates had the minimum values on BIC and ABIC, and the four-class solution with covariates had the minimum values on BIC. In addition, the four-class solution also had acceptably high values on entropy (without covariates, .832; with covariates, .806) and ACPs (without covariates, range .869–.921; with covariates, range .846–.915). Thus, by model fit statistics, we chose the four-class solution as the best-fitting model. An examination on class interpretability for models with two to five classes further indicates the four-class model as optimal in terms of practical implication and distinctiveness of each latent class.
      Table 3Model fit statistics for models with one to five latent classes
      Without covariates, N = 2,524 tests of model fit12345
      Information criteria
      Adjusted BIC is BIC adjusted for sample size. A model with smaller AIC, BIC, and adjusted BIC value is considered to have a better fit.
       AIC14,343.711,533.111,315.211,147.011,135.5
       BIC14,378.711,608.911,431.911,304.511,333.8
       Adjusted BIC14,359.711,567.611,368.311,218.711,225.8
      Chi-square test
      Pearson and likelihood chi-square statistics test departure of the model from the observed data.
       Pearson χ2582.7221.2145.642.327.4
       df5550433629
      p.000.000.000.217.550
       Likelihood ratio G2255.0177.3101.541.531.5
       df5550433629
      p.000.000.000.243.342
      Entropy
      The entropy statistics range from 0 to 1, with values closer to 1 indicating a better classification quality.
      .854.831.832.849
      Range of ACPs.958–.964.859–.957.869–.921.787–.954
      With covariates, N = 2,350 tests of model fit2345
      Information criteria
      Adjusted BIC is BIC adjusted for sample size. A model with smaller AIC, BIC, and adjusted BIC value is considered to have a better fit.
       AIC10,783.710,568.210,365.610,304.0
       BIC10,893.110,752.610,624.910,638.2
       Adjusted BIC10,832.810,650.910,481.910,453.9
      χ2 test
      Pearson and likelihood chi-square statistics test departure of the model from the observed data.
       Pearson χ2211.6121.944.140.8
       df50433629
      p.000.000.167.072
       Likelihood ratio G2169.9113.743.141.3
       df50433629
      p.000.000.195.065
      Entropy
      The entropy statistics range from 0 to 1, with values closer to 1 indicating a better classification quality.
      .849.828.806.819
      Range of ACPs.944–.969.823–.976.846–.915.695–.920
      ACP = average latent class probabilities for most likely latent class membership; AIC = Akaike's Information Criteria; BIC = Bayesian Information Criteria.
      a Adjusted BIC is BIC adjusted for sample size. A model with smaller AIC, BIC, and adjusted BIC value is considered to have a better fit.
      b Pearson and likelihood chi-square statistics test departure of the model from the observed data.
      c The entropy statistics range from 0 to 1, with values closer to 1 indicating a better classification quality.
      For the four-class solution of LCA with covariates, Figure 1 reports the latent class prevalence (i.e., the proportion of the sample composing each class) and item probability (i.e., the likelihood of reporting each behavior within a particular latent class). Adolescents in class 1 (59.3%) formed the nonuser group, with very low probabilities of using any of the substances, such as medication to get high (.000), illicit drugs (.006), cigarettes (.007), marijuana (.019), alcohol (.071), and binge drinking (.023). Adolescents in class 2 (22.6%) comprised the predominant alcohol user group, with high probability of drinking (.982) and moderately high probability of binge drinking (.735), moderate probabilities of using marijuana (.418) and smoking cigarettes (.413), and low probabilities of misusing medication (.074) and using other illicit drugs (.066). Adolescents in class 3 (10.5%) formed the predominant marijuana user group, with much higher probability of using marijuana (.578) than using other substances, such as cigarettes (.270) or medication to get high (.139) or other illicit drugs (.156), and virtually no use of alcohol (.000) or binge drinking (.076). Adolescents in class 4 (7.6%) formed the predominant polysubstance user group, who had moderate to very high probabilities of using each of the substances, including marijuana (1.000), alcohol (.979), binge drinking (.906), cigarettes (.711), medication to get high (.658), and other illicit drugs (.541).
      Figure thumbnail gr1
      Figure 1Item probability for each latent class in the four-class model and associations with depressive and somatic symptomatology. SE = standard error.

      Latent class analysis with covariates

      Covariates included gender (male as reference), race/ethnicity (white as reference), and parent education (three categories: low, median, and high, with high as reference). We excluded family structure and region because of possible multi co-linearity (with race/ethnicity and parent education) and the negative impact on statistical power of including many covariates. We set the nonuser class as the reference group. Table 4 reports the results. Females were less likely than males to be in the predominant marijuana user compared with the nonuser class (odds ratio [OR] = .34; confidence interval [CI] .18–.65). Compared with children with high parent education (bachelor's or graduate degrees), those with low parent education (high school degree or less) were more likely to be in the predominant alcohol user class (OR = 1.57; CI 105–2.33) or the predominant marijuana user class (OR = 3.32; CI 1.06–10.37).
      Table 4Results of LCA with covariates: sociodemographic differences (n = 2,350)
      Odds Ratio95% Confidence Interval
      Class 2. Predominant alcohol user
       Female versus male.86.56–1.31
       Race/ethnicity (vs. white)
      African-American.65.29–1.48
      Hispanic.80.32–1.97
      Other.82.31–2.15
       Parent education (vs. high school)
      Low (≤ high school)1.571.052.33
      Median (some college or technical school)1.15.86–1.54
      Class 3. Predominant marijuana user
       Female versus male.34.18.65
       Race/ethnicity (vs. white)
      African-American.97.23–4.07
      Hispanic1.26.76–2.08
      Other1.55.54–4.49
       Parent education (vs. high school)
      Low (≤ high school)3.321.0610.37
      Median (some college or technical school)2.28.93–5.61
      Class 4. Predominant polysubstance user
       Female versus male.95.51–1.75
       Race/ethnicity (vs. white)
      African-American.47.07–3.16
      Hispanic1.52.58–3.99
      Other1.75.27–11.57
       Parent education (vs. high school)
      Low (≤ high school).85.30–2.39
      Median (some college or technical school)1.02.32–3.25
      Significant odds ratios are in bold print. The latent class of nonuser was set as the reference group for the multinomial logistic regression model. For gender, race/ethnicity, and parent education, the reference groups were male, white, and high-parent education (bachelor's and graduate degree). Because of missing information on covariates, we excluded 174 participants for this analysis.

      Associations

      The four classes differed with respect to depressive and somatic symptomatology, which we included as distal outcomes with the AUXILIARY option in Mplus, shown in Figure 1. We did not adjust for covariates in this analysis. Nonusers reported significantly lower levels of depressive and somatic symptoms than any other class: predominant polysubstance users (p < .01), predominant alcohol users (p < .01), and predominant marijuana users (p < .05). Predominant polysubstance users reported significantly (p < .01) higher levels of depressive and somatic symptoms than those in all other classes. Predominant alcohol users and predominant marijuana users did not significantly differ in their reports of depressive or somatic symptoms.

      Discussion

      The primary study findings are that (1) there were four classes of drug users: nonusers, predominant drinkers, marijuana users, and polysubstance users; and (2) the predominant polysubstance users scored higher on depressive and somatic symptomatology. The overall prevalence of substance use in this sample is consistent with other national studies, such as the Monitoring the Future Study, and highlights its ubiquitousness among American youth. Past-year rates of marijuana use among 10th graders, for example, were similar between the NEXT study (26%) and the 2011 MTF study (29%) [
      • Johnston L.D.
      • O'Malley P.M.
      • Bachman J.G.
      • Schulenberg J.E.
      Monitoring the Future national results on adolescent drug use: overview of key findings, 2011.
      ]. Past-year rates of medication misuse among 10th graders were slightly higher in the NEXT study (8.7%) compared with the 2011 MTF study (range, 2.6%–5.9%). This difference may stem from a definitional difference of medication misuse: NEXT asks about use of “medication to get high,” whereas MTF asks about use of specific medications (e.g., oxycodone HCl, hydrocodone/paracetamol, methylphenidate) “not under a doctor's orders.”
      Regarding noteworthy demographic factors associated with substance use, children who live with both biological parents were less likely than children who live with stepparents to use marijuana, medication to get high, other illicit drugs, or cigarettes, and to binge drink. This finding is consistent with the well-supported notion that lack of parental monitoring is a risk factor for substance use [

      Substance Abuse and Mental Health Services Administration, Results from the 2010 National Survey on Drug Use and Health: Summary of National Findings, NSDUH Series H-41, HHS Publication No. (SMA) 11-4658. Rockville, MD: Substance Abuse and Mental Health Services Administration, 2011.

      ,
      • Hawkins J.D.
      • Catalano R.F.
      • Miller J.Y.
      Risk and protective factors for alcohol and other drug problems in adolescence and early adulthood: Implications for substance abuse prevention.
      ].
      The multiple classes defined by polysubstance use reported here compare with other studies [
      • Dierker L.C.
      • Vesel F.
      • Sledjeski E.M.
      • et al.
      Testing the dual pathway hypothesis to substance use in adolescence and young adulthood.
      ,
      • Connell C.M.
      • Gilreath T.D.
      • Hansen N.B.
      A multiprocess latent class analysis of the co-occurrence of substance use and sexual risk behavior among adolescents.
      ] and underscore the considerable heterogeneity that exists among young substance users. In our sample, as in Add Health, abstainers comprised the largest and majority class of substance users [
      • Dierker L.C.
      • Vesel F.
      • Sledjeski E.M.
      • et al.
      Testing the dual pathway hypothesis to substance use in adolescence and young adulthood.
      ]. By comparison, in the 2009 YRBS, with a wider age range in the sample, alcohol users formed the largest class [
      • Connell C.M.
      • Gilreath T.D.
      • Hansen N.B.
      A multiprocess latent class analysis of the co-occurrence of substance use and sexual risk behavior among adolescents.
      ]. All three studies found that polysubstance users formed a smaller class of youth: NEXT, with 7.6% of the sample characterized as predominant polysubstance users; YRBS, with 13% frequent polysubstance users and 23% occasional polysubstance users; and Add Health, with 14% three-substance users. Both the NEXT and YRBS identified that one polydrug class stood apart from the others in the seriousness of their substance use patterns, symptomatology, and risk behaviors [
      • Connell C.M.
      • Gilreath T.D.
      • Hansen N.B.
      A multiprocess latent class analysis of the co-occurrence of substance use and sexual risk behavior among adolescents.
      ].
      In addition to these findings, several results are particularly important and novel. First, the analyses found a strong representation of medication misuse in the predominant polysubstance user class (66%). Medication misuse is associated with serious adverse consequences including addiction and unintentional overdose [
      • Compton W.M.
      • Volkow N.D.
      Abuse of prescription drugs and the risk of addiction.
      ,
      • Hall A.J.
      • Logan J.E.
      • Toblin R.L.
      • et al.
      Patterns of abuse among unintentional pharmaceutical overdose fatalities.
      ]. Second, the predominant polysubstance users reported binge drinking at rates equal to, or even higher than, predominant alcohol users. Binge drinking during adolescence is a hazardous practice associated with significant short-term and long-term negative consequences [
      • Chassin L.
      • Pitts S.C.
      • Prost J.
      Binge drinking trajectories from adolescence to emerging adulthood in a high-risk sample: Predictors and substance abuse outcomes.
      ,
      • Donath C.
      • Grassel E.
      • Baier D.
      • et al.
      Predictors of binge drinking in adolescents: Ultimate and distal factors—a representative study.
      ]. Whereas our study measured the use of various substances over a period of time and not simultaneous polysubstance use (e.g., misusing medications while drinking alcohol), other research reported high rates and negative consequences of simultaneous use in adolescents and college students [
      • McCabe S.E.
      • Cranford J.A.
      • Morales M.
      • Young A.
      Simultaneous and concurrent polydrug use of alcohol and prescription drugs: Prevalence, correlates, and consequences.
      ,
      • McCabe S.E.
      • West B.T.
      • Teter C.J.
      • Boyd C.J.
      Co-ingestion of prescription opioids and other drugs among high school seniors: Results from a national study.
      ]. Third, the predominant polysubstance users reported elevated levels of depressive symptoms and more somatic problems such as headaches, stomachaches, and backaches, which often underlie depression [
      • Bernstein G.A.
      • Massie E.D.
      • Thuras P.D.
      • et al.
      Somatic symptoms in anxious-depressed school refusers.
      ,
      • Egger H.L.
      • Costello E.J.
      • Erkanli A.
      • Angold A.
      Somatic complaints and psychopathology in children and adolescents: Stomach aches, musculoskeletal pains, and headaches.
      ]. The association of polysubstance use with mental health symptoms is concerning, because comorbid depressive psychopathology often predicts substance use, worsened clinical course, and medical illnesses among substance abusers [
      • Rounsaville B.J.
      • Dolinsky Z.S.
      • Babor T.F.
      • Meyer R.E.
      Psychopathology as a predictor of treatment outcome in alcoholics.
      ,
      • Compton III, W.M.
      • Cottler L.B.
      • Jacobs J.L.
      • et al.
      The role of psychiatric disorders in predicting drug dependence treatment outcomes.
      ,
      • Harter M.C.
      • Conway K.P.
      • Merikangas K.R.
      Associations between anxiety disorders and physical illness.
      ]. Our findings strongly suggest that polysubstance users in the 10th grade may already be experiencing significant impairments and may be at high risk for deleterious outcomes, an interpretation warranting confirmation by longitudinal analysis. Finally, polysubstance use may be a broad indicator of severity prevalent in over 7% of 10th graders. There is a need to reinforce the call for better substance use and mental health screening as well as referral to and availability of care, as recommended by the American Academy of Pediatrics Association and the Society for Adolescent Medicine [
      • American Academy of Pediatrics
      Coding for pediatrics.
      ,
      • American Academy of Pediatrics
      Policy statement—alcohol use by youth and adolescents: A pediatric concern.
      ,
      • Kapphahn C.J.
      • Morreale M.C.
      • Rickert V.I.
      • Walker L.R.
      Financing mental health services for adolescents: A position paper of the Society for Adolescent Medicine.
      ].
      The results reported here should be considered in light of several limitations and strengths. Study limitations include the somewhat attenuated response rate, limited sample size relative to other national epidemiological studies, and cross-sectional design. The strengths include the nationally representative sampling frame, oversampling of African-American and Hispanic youth to allow for robust comparisons across certain race/ethnic groups, assessment of depressive and somatic symptoms, and the examination of misuse of medications separately from other illicit drugs.

      Acknowledgments

      This project (contract number HHSN267200800009C) was supported in part by the intramural research program of the Eunice Kennedy Shriver National Institute of Child Health and Human Development (NICHD), and the National Heart, Lung and Blood Institute (NHLBI), the National Institute on Alcohol Abuse and Alcoholism (NIAAA), and Maternal and Child Health Bureau (MCHB) of the Health Resources and Services Administration (HRSA), with supplemental support from the National Institute on Drug Abuse (NIDA).
      The views and opinions expressed in this article are those of the authors and do not necessarily represent the views of the National Institute on Drug Abuse, the Eunice Kennedy Shriver National Institute of Child Health and Human Development, the National Institutes of Health, or any other governmental agency.

      References

        • Johnston L.D.
        • O'Malley P.M.
        • Bachman J.G.
        • Schulenberg J.E.
        Monitoring the Future national results on adolescent drug use: overview of key findings, 2011.
        Institute for Social Research, University of Michigan, Ann Arbor (MI)2012
        • Mitchell C.M.
        • Plunkett M.
        The latent structure of substance use among American Indian adolescents: An example using categorical variables.
        Am J Commun Psychol. 2000; 28: 105-125
        • Connell C.
        • Gilreath T.
        • Aklin W.
        • Brex R.
        Social-ecological influences on patterns of substance use among non-metropolitan high school students.
        Am J Commun Psychol. 2010; 45: 36-48
        • Whitesell N.R.
        • Beals J.
        • Mitchell C.M.
        • et al.
        Latent class analysis of substance use: Comparison of two American Indian reservation populations and a national sample.
        J Stud Alcohol. 2006; 67: 32-43
        • Dierker L.C.
        • Vesel F.
        • Sledjeski E.M.
        • et al.
        Testing the dual pathway hypothesis to substance use in adolescence and young adulthood.
        Drug Alcohol Depend. 2007; 87: 83-93
        • Connell C.M.
        • Gilreath T.D.
        • Hansen N.B.
        A multiprocess latent class analysis of the co-occurrence of substance use and sexual risk behavior among adolescents.
        J Stud Alcohol Drugs. 2009; 70: 943-951
      1. Substance Abuse and Mental Health Services Administration, Results from the 2010 National Survey on Drug Use and Health: Summary of National Findings, NSDUH Series H-41, HHS Publication No. (SMA) 11-4658. Rockville, MD: Substance Abuse and Mental Health Services Administration, 2011.

        • Kessler R.C.
        • Crum R.M.
        • Warner L.A.
        • et al.
        Lifetime co-occurrence of DSM-III-R alcohol abuse and dependence with other psychiatric disorders in the National Comorbidity Survey.
        Arch Gen Psychiatry. 1997; 54: 313-321
        • Conway K.P.
        • Compton W.
        • Stinson F.S.
        • Grant B.F.
        Lifetime comorbidity of DSM-IV mood and anxiety disorders and specific drug use disorders: Results from the National Epidemiologic Survey on Alcohol and Related Conditions.
        J Clin Psychiatry. 2006; 67: 247-257
        • Regier D.A.
        • Farmer M.E.
        • Rae D.S.
        • et al.
        Comorbidity of mental disorders with alcohol and other drug abuse: Results from the Epidemiologic Catchment Area (ECA) Study.
        JAMA. 1990; 264: 2511-2518
        • Compton W.M.
        • Conway K.P.
        • Stinson F.S.
        • et al.
        Prevalence, correlates, and comorbidity of DSM-IV antisocial personality syndromes and alcohol and specific drug use disorders in the United States: Results from the national epidemiologic survey on alcohol and related conditions.
        J Clin Psychiatry. 2005; 66: 677-685
        • Merikangas K.R.
        • Mehta R.L.
        • Molnar B.E.
        • et al.
        Comorbidity of substance use disorders with mood and anxiety disorders: Results of the International Consortium in Psychiatric Epidemiology.
        Addict Behav. 1998; 23: 893-907
        • Costello E.J.
        • Mustillo S.
        • Erkanli A.
        • et al.
        Prevalence and development of psychiatric disorders in childhood and adolescence.
        Arch Gen Psychiatry. 2003; 60: 837-844
        • Merikangas K.R.
        • He J.P.
        • Burstein M.
        • et al.
        Lifetime prevalence of mental disorders in U.S. adolescents: Results from the National Comorbidity Survey Replication–Adolescent Supplement (NCS-A).
        J Am Acad Child Adolesc Psychiatry. 2010; 49: 980-989
        • Marmorstein N.R.
        • Iacono W.G.
        • Malone S.M.
        Longitudinal associations between depression and substance dependence from adolescence through early adulthood.
        Drug Alcohol Depend. 2010; 107: 154-160
        • Burke Jr., J.D.
        • Burke K.C.
        • Rae D.S.
        Increased rates of drug abuse and dependence after onset of mood or anxiety disorders in adolescence.
        Hosp Commun Psychiatry. 1994; 45: 451-455
        • McCabe S.E.
        • Cranford J.A.
        • Morales M.
        • Young A.
        Simultaneous and concurrent polydrug use of alcohol and prescription drugs: Prevalence, correlates, and consequences.
        J Stud Alcohol. 2006; 67: 529-537
        • Bernstein G.A.
        • Massie E.D.
        • Thuras P.D.
        • et al.
        Somatic symptoms in anxious-depressed school refusers.
        J Am Acad Child Adolesc Psychiatry. 1997; 36: 661-668
        • Egger H.L.
        • Costello E.J.
        • Erkanli A.
        • Angold A.
        Somatic complaints and psychopathology in children and adolescents: Stomach aches, musculoskeletal pains, and headaches.
        J Am Acad Child Adolesc Psychiatry. 1999; 38: 852-860
      2. National Institute on Alcohol Abuse and Alcoholism. NIAAA Newsletter (NIH publication number 04-5346); 01.01.2004.

      3. Dahlberg L, Toal S, Swahn M, Behrens C. Measuring violence-related attitudes, behaviors, and influences among youths: a compendium of assessment tools. 2. 01.01.2005. Atlanta (GA): Centers for Disease Control and Prevention, National Center for Injury Prevention and Control.

        • Hetland J.
        • Torscheim T.
        • Aura L.
        Subjective health complaints in adolescence: Dimensional structure and variation across gender and age.
        Scand J Public Health. 2002; 30: 223-230
        • Akaike H.
        On the entropy maximization principle.
        in: Krishnaiah P. Applications of statistics. North-Holland, Amsterdam1977 (p. 27–41)
        • Raftery A.
        Bayesian model selection in social research.
        Sociol Method. 1995; 25: 111-163
        • Nylund K.L.
        • Asparouhov T.
        • Muthén B.O.
        Deciding on the number of classes in latent class analysis and growth mixture modeling: A Monte Carlo simulation study.
        Structural Equation Modeling. 2007; 14: 535-569
      4. Muthén LK, Muthén BO. Mplus user’s guide. 5th ed; 1998.

        • Schafer J.L.
        • Graham J.W.
        Missing data: Our view of the state of the art.
        Psychol Method. 2002; 7: 147-177
        • Hawkins J.D.
        • Catalano R.F.
        • Miller J.Y.
        Risk and protective factors for alcohol and other drug problems in adolescence and early adulthood: Implications for substance abuse prevention.
        Psychol Bull. 1992; 112: 64-105
        • Compton W.M.
        • Volkow N.D.
        Abuse of prescription drugs and the risk of addiction.
        Drug Alcohol Depend. 2006; 83: S4-S7
        • Hall A.J.
        • Logan J.E.
        • Toblin R.L.
        • et al.
        Patterns of abuse among unintentional pharmaceutical overdose fatalities.
        JAMA. 2008; 300: 2613-2620
        • Chassin L.
        • Pitts S.C.
        • Prost J.
        Binge drinking trajectories from adolescence to emerging adulthood in a high-risk sample: Predictors and substance abuse outcomes.
        J Consult Clin Psychol. 2002; 70: 67-78
        • Donath C.
        • Grassel E.
        • Baier D.
        • et al.
        Predictors of binge drinking in adolescents: Ultimate and distal factors—a representative study.
        BMC Public Health. 2012; 12: 263
        • McCabe S.E.
        • West B.T.
        • Teter C.J.
        • Boyd C.J.
        Co-ingestion of prescription opioids and other drugs among high school seniors: Results from a national study.
        Drug Alcohol Depend. 2012; 126: 65-70
        • Rounsaville B.J.
        • Dolinsky Z.S.
        • Babor T.F.
        • Meyer R.E.
        Psychopathology as a predictor of treatment outcome in alcoholics.
        Arch Gen Psychiatry. 1987; 44: 505-513
        • Compton III, W.M.
        • Cottler L.B.
        • Jacobs J.L.
        • et al.
        The role of psychiatric disorders in predicting drug dependence treatment outcomes.
        Am J Psychiatry. 2003; 160: 890-895
        • Harter M.C.
        • Conway K.P.
        • Merikangas K.R.
        Associations between anxiety disorders and physical illness.
        Eur Arch Psychiatry Clin Neurosci. 2003; 253: 313-320
        • American Academy of Pediatrics
        Coding for pediatrics.
        16th ed. American Academy of Pediatrics, Elk Grove Village (IL)2011
        • American Academy of Pediatrics
        Policy statement—alcohol use by youth and adolescents: A pediatric concern.
        Pediatrics. 2010; 5: 1078-1087
        • Kapphahn C.J.
        • Morreale M.C.
        • Rickert V.I.
        • Walker L.R.
        Financing mental health services for adolescents: A position paper of the Society for Adolescent Medicine.
        J Adolesc Health. 2006; 39: 456-458