Drug and Alcohol Dependence,
Elsevier Scientific Publishers
275
30 (1992) 275 - 287
Ireland Ltd.
Elementary schoolchildren’s use of alcohol, cigarettes and marijuana and classmates’ attribution of socialization Patricia Children’s Health Promotion,
J. Bush and Ronald J. Iannotti
Department of Community and Family Medicine, 415 Kober Cogan Bldg., Georgetown University School of Medicine, Washington, DC 20007 (USA)
(Accepted February 23, 1992)
In each of 2 years beginning in fourth and fifth grades, urban elementary public schoolchildren completed surveys about abusable substance use and health promoting behaviors and completed an instrument that permitted each child to have a socialization score attributed by classmates. A factor weighted 12-item scale was developed from 15 items in three domains (personal, interpersonal and school). The scale was positively correlated over the 2 years and positively correlated with a healthful activities scale in both years. Conditional multiple logistic regression, matching on school classroom, indicated that socialization was negatively associated with use of alcohol without parental permission and cigarettes in both years (grades 4 - 5, grades 5 - 6) and with use of marijuana in year 1. Socialization measured in year 1 was negatively associated with cigarette use in year 2 and with onset of use from year 1 to year 2. Shyness, a non-socialization scale item, was negatively associated with use of cigarettes in both years and with use of alcohol without permission and use of marijuana in year 2. Being ‘good at sports’ was an attribute positively associated with alcohol use without permission and cigarette use in year 2. Results suggest that elementary schoolchildren can ascribe social characteristics to their classmates that are associated with and predict health related behaviors. Key words: alcohol; cigarettes;
marijuana; schoolchildren;
epidemiology
Introduction The relationships between use and onset of use of alcohol, cigarettes and marijuana by elementary schoolchildren and a measure of social adaptation, i.e., socialization, attributed by their classmates, were examined in this study. Although other personal characteristics, e.g., aggressiveness (Kellam et al., 1980) and risk taking (Bush and Iannotti, 1985), have been associated with the use of abusable substances by preadolescents, social influences in three domains (family, school, peer) at the environmental and interpersonal levels have been viewed as more important (Hawkins, in press). Correspondence to: Patrica J. Bush, Children’s Health Promotion, Department of Community and Family Medicine, 415 Kober Cogan Bldg., Georgetown University School of Medicine, Washington, D.C. 20007, USA. 0376X716/92/$05.00 0 1992 Elsevier Scientific Publishers Printed and Published in Ireland
Consistent positive correlations have been observed between preadolescent children’s use and expectations to use abusable substances and use by family members. For example, Bush and Iannotti (1985) found that the number of family users of alcohol, marijuana and cigarettes, was correlated with the children’s use and use intentions. In earlier research, Shute et al. (1981) showed that half of a sample of preschool children exposed to smoking at home indicated they intended to smoke in the future compared to only 11% who were not exposed. For adolescents, the influence of parental drug use has been found to be less than the influence of use by older siblings’ and peers’ (Brook et al., 1983, 1989; Needle et al., 1986). After reviewing the literature, Glynn (1981) concluded that the influence of parents never completely disappears, but that it recedes in favor of peers as children move into adolescence. Ireland Ltd.
276
Lack of academic success and low academic aspirations consistently have been associated with use of abusable substances (Jessor and Jessor, 1977; Kandel, 1978; Paulson and Coombs, 1990; Robins, 1980, Smith and Fogg, 1978), although the relationships may be spurious. Substance use and academic failure may both result from prior problems such as peer rejection. Peers also have been found to influence the substance use of preadolescents (Bush and Iannotti, 1985; Iannotti et al., 1986; Iannotti and Bush, 1992). This influence may be differentiated according to whether it represents: (1) peer behavior, (2) peer beliefs and attitudes, (3) perceived peer behaviors, or (4) perceived peer beliefs and attitudes. Relative to the influence of peers, Iannotti and Bush (1992) found that the perceived use of friends explained more of the variance in the alcohol, cigarette and marijuana use of fourth and fifth grade students than either the actual use of three identified best friends or the actual use of the students’ classmates. However, the actual use of the classmates explained more of the variance than the actual use of the friends. In view of the consistent finding that peer use is among the strongest correlates of substance use among adolescents (Barnes and Welte, 1986; Elliott et al., 1985; Hansen et al., 1987; Kandel, 1978, 1986; Needle et al., 1986; Newcomb and Bentler, 1986) and has some predictive value in preadolescents (Gillmore, 1990; Iannotti and Bush, 1992), the mechanism for how the influence of peers is transmitted is of interest. The question remains, for example, whether the availability of abusable substances as an influence on a child’s use is mediated by psychological and other characteristics of the child that affect relationships with peers. In a panel of children followed from preschool into adolescence, Block and his colleagues (1988; Shedler and Block, 1990) found that psychological differences among frequent drug users, experimenters and abstainers could be traced from early childhood to age 18. Also, Kellam and his associates (1980) found that psychological measured characteristics, e.g., aggressiveness
at ages 6 and 7 predicted drug use a decade later whereas children rated as ‘shy’ were least likely to become drug users. Lerner and Vicary (1984) followed 5-year olds into adulthood and found that a ‘difficult’ childhood, characterized by frequent negative moods and withdrawal, predicted regular alcohol, tobacco and marijuana use in adulthood. These studies argue strongly for early assessment and tracking of children prospectively so that targeted interventions can be developed. However, early psychological assessment for large cohorts of children using standard methods is impractical. The instruments are long and require skilled personnel, or reliance must be placed on subjective reports of parents and teachers. A psychometric instrument was developed by Smith (1967) that permitted eighth and ninth grade students to rate their classmates on 18 socialization characteristics. At follow-up ll- 12 years later, the ratings were found associated with use of cigarettes, alcohol, marijuana, cocaine and ‘hard drugs’; the direction of the relationship consistently indicated that poor socialization preceded use (Smith and Fogg, 1978; Smith, 1986). The results obtained by Smith (1978; 1986), Kellam et al., (1980) and Iannotti and Bush (1992) have relatively high validity because they predicted substance use by characteristics attributed by respondents other than the subjects themselves. In most research inquiring into characteristics associated with use of abusable substances, the characteristics as well as the use are reported by the subjects. In the research discussed here, following Smith (1967; 1978; 1986), children reported their own behaviors and rated their classmates on social attributes and the relationships between these and use were investigated cross-sectionally and longitudinally in preadolescents. This approach was consistent with that of Kellam et al. (1980) who advocated use of ‘natural raters’ to assess the social adaptational status of children, suggesting peers as well as parents and teachers as appropriate raters. Of additional interest was the relationship of healthful behaviors to the use of abusable
substances and socialization. It has been suggested that healthful behaviors, e.g., exercise, protect against abusable substance use (Perry and Jessor, 1985). If healthful behaviors are negatively related to use of abusable substances, then socialization may be a predisposing factor for both of these. Our first hypothesis is that schoolchildren who are rated highest by their classmates on socialization are least likely to use abusable substances; our second hypothesis is that schoolchildren who are rated highest by their classmates on socialization will be most likely to exhibit healthful behaviors. Methods Populations
and data collection
In 1988- 1989, a classroom survey inquiring into abusable substance use was administered to public elementary school students in grades 4 - 5 in Washington, DC. A subset of these students also was given a sociometric questionnaire. In the second year (grades 5 - 6) both instruments were administered to all of the students. To test the hypotheses, analyses were limited to students in the 296 classes of 10 students or more who completed the survey and were ranked by classmates on the sociometric questionnaire in year 2 (N = 5995); of these students, 1904 in 153 classes of 10 or more students also have survey data and sociometric rankings in year 1. In year 1, another 3857 students had survey data only or were in smaller classes. In the first year, subjects were attending classes in 111 schools (94.1% of all DC elementary schools); 76.5% of students were in fourth grade and 23.5% were in fifth grade; less than 1% refused to complete the survey, but about 12% of the students were absent from class at the time of the survey or had transferred. Of the cohort, 86.1% self-identified themselves as black, 3.8% as white, 3.9% as Hispanic, 1.0% as Asian and 5.1% as ‘other.’ The median and modal age was 10 years (range 8- 12), with 51.0% female. In the second year, 109 (92.4%) schools participated; 74.5% of the students were in fifth
grade and 25.5% were in sixth grade; 88.9% self-identified themselves as black, 3.4% as white, 4.1% as Hispanic 0.8% as Asian and 2.7% as ‘other’; 50.6% were female. There were no significant gender differences between subjects completing both a drug survey and a sociometric questionnaire in year 1 and subjects completing them in year 2, (O.R. 0.937; 95% C.I. 0.843, 1.040). For none of the substance-use-variables were there significant differences between subjects completing the drug survey in both years and subjects completing the drug surveys in year 1 only. However, boys represented 51.8% of subjects completing drug surveys in year 1 only, and 48.3% of those completing drug surveys in both years (O.R. 1.157; 95% C.I. 1.038, 1.289). The rates of reported ‘crack’ use were too low to be used in any analyses. Survey Following assurances of confidentiality, students completed the fixed choice question survey in their classrooms. Each question with corresponding choices was projected on a slide screen and read to the class by a member of the research staff. Children marked their responses on special answer sheets created to correspond to these questions without containing any indication as to the content of questions or answers. Items were read as often as necessary until all children had an opportunity to mark their answer sheets. The answer sheets, which did not contain identifiers, were placed by the children in envelopes bearing identifying codes, sealed and dropped through a slot in a box for transport to the research site. These annual identifying codes are linked for longitudinal tracking through limited access computer files that contain the school system identification code and the student’s name. Behaviors assessed were use of cigarettes, alcohol (beer, wine, liquor, wine cooler), marijuana and cocaine or crack and five healthful behaviors, including eating properly, exercising and getting enough sleep. Healthful behaviors were assessed in absolute terms, i.e., ‘Which things are usually true for you?’ The healthful
278
activity scale consisted of five items (year 1: mean 2.37, SD. 1.05; year 2: mean 2.39, SD. 1.08). Students were also asked about their school grades (1 - 5, ‘F’ to ‘A’) they most often got in school. Risk taking was measured by a drawing of a stick figure on a wall that varied in height from ground level to the same height as the figure. The students were told to indicate where of five heights they would jump off if they were the figure on the wall. Students were asked if they had been offered cigarettes, alcohol (beer, wine, liquor, wine cooler) and marijuana by other kids and responses were summed to form a O-3 point scale. To increase validity of responses, questions relating to substance use were posed in a way which assumed the child had used the substance, e.g., ‘How old were you when you first used . ..?’ with the option ‘Never have’ following the choices. Similarly phrased questions also inquired into the frequency of use of each abusable substance, e.g., ‘How many times have you . ..?’ with choices followed by the option ‘Never have.’ A child was not considered to have used a substance unless a positive response was recorded for at least two out of three questions inquiring into its use. In year 1, 0.5% of children with three or more inconsistent responses overall were dropped from all analyses; in year 2, 0.02% were dropped for this reason. Dichotomous variables were created to indicate whether a substance had ever been used. Another alcohol use variable indicated whether the child had ever had alcohol without the child’s mother knowing about it. Additional variables measured heavier use: for alcohol whether and the number of times the child had had more than a sip of alcohol; for cigarettes, whether and the number of times the child had had more than just a puff. Because gender and race have been associated with differences in abusable substance use (Barnes and Welte, 1986; Newcomb et al., 1987), it is important to control for them in analyses. Although a school level of socioeconomic status (SES) was obtained (percentage of children eligible for the federally funded school lunch program), use of conditional logistic analyses, which
took account of the students’ school and classroom, made inclusion of the SES variable unnecessary. Sociometric
questionnaire
Unlike the drug survey, the sociometric questionnaire was self-adminstered (see Table I). The sociometric questionnaire permitted all children in the class to be ranked by their classmates on social competency indicators. Three categories of variables intended to tap personal, interpersonal and school competency and shown to relate to early substance use were identified in the literature. The technique, a much simplified version of that used by Smith (1967) with older students, consisted of having each student record by number (from a numbered list of classmates), the two students who in his/her opinion best represented the opposite ends of a scale, e.g., most likes school and least likes school (see Table I). Children were permitted to select themselves. This process resulted in each child having a score for each item that reflected the percentage of times he/she was nominated by classmates for that item. Suppose, for example, there were 30 children in a class; they nominated 60 children as most liking school and 60 children as least liking school and a particular child was nominated twice (3.3% of the responses) as most liking school and 6 times (10% of the responses) as least liking school. This child’s score for likes school would be - 6.7 (3.3% - 10%). If a child was not nominated at all for an item, the score was zero. At each administration, four versions of the instrument varied the order of the same items to reduce order effects and increase confidentiality. A pilot version contained 30 items and asked students to record four classmates who represented the most and the four who represented the least of each item. The pilot version was administered to 94 low to low-middle socioeconomic status fourth grade students in four classes. The pilot study indicated that the instrument was too long for most students to complete. Therefore, the number of classmates each student was asked to pick for each item was
279
Table I.
Sociometric instrument
(version 1 of 4)* used for peer ratings.
Put down the number of the two students in your class who you think are the MOST and who are the LEAST of each thing. The first one is an example. LEAST
MOST 12 -
3 -
Lucky Liked by teachers Likes school Takes chances Picks on others Fast at doing things Popular Can’t sit still in school Good at sports Picked on by others School work is neat Quits Can’t be trusted Shy Needs help to do things Friendly Three best friends
5 xx
16 xx
*A fourth of the items are rotated from bottom to top to yield four versions used in each classroom.
reduced to four and items were eliminated using the following criteria: (1) items not understood as reported by field staff; (2) items failing to differentiate among the students and (3) multiple items differentiating similarly. Finally, the list was balanced to achieve 5 personal, 5 interpersonal and 5 school items. Using the sociometric data from the larger second year sample, factor analyses (Table II) were used to derive factor scores and to reduce the data. Twelve of the 15 items in the principle components factor had loadings exceeding 0.35 in year 2 and achieved satisfactory internal reliability for children the age of those in this study in both years (Cronbach’s alpha 0.81 in year 2 and 0.70 in year 1). *Three of the items, ‘shy’, ‘good at sports’ and ‘picked on by others’, were dropped from the socialization scale because of low factor loadings and to increase internal reliability. The correlation between the socialization factors using the same 12 items in both years was
0.55 (P 5 0.0001). The range of the socialization factor was -5.41 to +6.01 (mean 0.051, SD. 0.986) in year 1 and -8.04 to + 5.37 (mean 0.000, S.D. 1.00) in year 2. As another indicator to the validity of the socialization scale, it was examined for its relationship to the students’ estimates of their usual school grades. The socialization scale was associated with the subjects’ positively estimates of their usual school grades in year 2 (r = 0.34, P 5 O.OOOl), but not in year 1 (r = 0.01, P 2 0.05). However, socialization in year 1 predicted year 2 usual school grade estimates (r = 0.28, P I 0.0001). There was no reverse relationship, i.e., usual school grades in year 1 did not predict socialization in year 2 (T = 0.01, P s- 0.05). As a traditional measure of socialization, each student had the opportunity to be nominated as a best friend by classmates on the sociometric questionnaire (Table I). Each student nominated three classmates as best friends. Popularity
280 Table II.
Socialization scale item factor loadings in year 1 (Grades 4 - 5, N = 1904) and year 2 (Grades 5 - 6, N = 5995). Year 1
Year 2
Items with positive loadings Fast at doing things Popular School work is neat Friendly Can be trusted Liked by teachers Likes school
0.53 0.46 0.59 0.58 0.61 0.73 0.64
0.48 0.46 0.60 0.64 0.66 0.80 0.71
Items Needs Takes Picks Can’t Quits
0.41 0.22 0.38 0.28 0.13
0.55 0.40 0.54 0.52 0.37
3.00 0.70 0.054 - 5.41, 6.01
3.94 0.81 0.00 - 8.04,5.57
with negative loadings help to do things chances on others sit still in school
Eigenvalue: Cronbach’s alpha: * Mean: Range:
*All negative items reversed; items dropped because of low factor loadings and to increase alpha coefficient included ‘shy’, ‘good at sports’, and ‘picked on by others’.
represented the percentage of times of the total number of nominations made by the class that a child was named as a best friend. Popularity and socialization were correlated in both years (year 1: r = 0.30, P 5 0.0001; year 2: r = 0.29; P I 0.0001). analyses Conditional logistic regression models were used to estimate the degree of association between the socialization scale and use and onset of use of alcohol, cigarettes and marijuana. These models took account of the school classroom of the subjects and provided statistical control of gender and race, covariates with potentially confounding effects. Conditional multiple logistic regression permitted assessment of models that also included the three nonsocialization scale items, i.e., ‘shy’, ‘good at sports’, and ‘picked on by others’ and permitted exploration of interaction effects.
Statistical
Results
Table III shows the percentage of subjects reporting use of alcohol, alcohol without parental permission, cigarettes and marijuana. Boys were more likely to report use in both years, and, except for marijuana, rates of use were higher in year 2 (grade 5- 6) than in year 1 (grade 4 - 5). The conditional logistic regressions of the use variables on the socialization scale (holding the student’s gender and race constant and matching on the school classroom) are shown in Table IV. This table includes the results of crosssectional analyses (both socialization and substance use measured in the same years) and longitudinal analyses (socialization measured in year 1 and substance use measured in year 2 and socialization measured in year 1 and onset of substance use measured from year 1 to year 2). The results indicate that in both years, children attributed higher socialization scores by their
281 Table III. Percentage of urban children reporting use of alcohol, cigarettes, and year 2 (grades 5 - 6, N = 5995). Substance
Alcohol Ever had sip or more Drank without parental permission Cigarettes Ever had puff or more Had more than a puff Marijuana Ever had puff or more
and marijuana in year 1 grades 4 - 5, N = 5761)
Year 2
Year 1 Boys
Girls
Boys
Girls
58.1% 18.6
47.0% 11.4
59.5% 20.4
52.3% 15.3
21.4 15.3
16.0 11.6
25.1 18.9
20.5 15.1
2.8
1.4
2.2
1.0
classmates were less likely to report ever having used alcohol without parental permission, ever having had a puff of a cigarette, ever having had more than a puff of a cigarette and ever having had a puff or more of marijuana. Also, socialization measured in year 1 was negatively associated with ever having smoked cigarettes (puff or more and more than a puff) measured in year 2 and with cigarette smoking onset (puff or more and more than a puff) from year 1 to year 2. Relatively low socialization scores in year 1 increase the relative risk of onset of cigarette smoking by 24% and of onset of smoking more than a puff by 30%. The odds ratios are consistently stronger for smoking more than a puff than for smoking only a puff or more. The conditional multiple logistic regressions examining the effects of adding the three nonsocialization scale items, shy, good at sports and picked on by others, to the models containing the socialization scale are shown in Table V. ‘Picked on by others’ did not contribute significantly to any of the models. ‘Shy’ contributed negatively to all four of the crosssectional year 2 models and to cigarette smoking (puff or more) in year 1, but provided no significant contribution to any of the longitudinal models. Good at sports contributed positively to the models for alcohol use without permission and with cigarette use (puff or more and more than a puff) in year 2 but not in year 1 cross-
sectional analyses. A weak positive association was found between good at sports measured in year 1 and cigarette use (puff or more) in year 2. Neither shy nor good at sports predicted onset of use in any model. Examination of the multiple conditional regression results for year 1 cross-sectional analyses indicated that girls were less likely than boys to have used alcohol with parental knowledge and to have smoked marijuana; black students were less likely to have smoked marijuana. Year 2 cross-sectional analyses indicated that girls and black students were less likely to have smoked marijuana and that black students were less likely to have smoked more than a puff of a cigarette. Being female also protected the student in longitudinal analyses (year 1 socialization measures, year 2 use) from alcohol use without parental permission and from marijuana use. Being black protected against marijuana use and onset in both longitudinal models. Those models that showed gender or race effects were examined for interactions with other independent variables, but none were found. Because the relationship between shyness and substance use was consistent with that found by Kellam et al. (1980), further analysis was performed to increase understanding of the role of shyness. Kellam et al. rejected the idea that shyness protected because it limited opportunities to associate with users; their analyses
(N = 1854) 0.812 0.689 - 0.955 0.012
Use onset from year 1 to year 2 Socialization-Year 1 (N = 1843) O.R. 1.008 95% C.I. 0.883- 1.151 0.901 P-value
(N = 1859) 0.757 0.632-0.906 0.002
(N = 1865) 0.736 0.632 - 0.857 < 0.001
(N = 5972) 0.683 0.633 - 0.737 < 0.001
(N = 1859) 1.497 0.965 - 2.324 0.072
(N = 1869) 1.371 0.861- 2.183 0.184
(N = 5992) 0.769 0.630 - 0.939 0.010
(N = 1894) 0.514 0.331-0.799 0.003
Marijuana (puff + )
*Matching variable = school classroom; gender and race held constant; ‘use’ variables coded 0 - 1 no use, use; socialization coded low to high; Ns vary within models because of missing values. “O.R. = Odds Ratio; C.I. = Confidence Interval.
(N = 1864) 0.759 0.665 - 0.868 < 0.001
(N = 5967) 0.711 0.664 - 0.760 puff)
of alcohol, cigarette, and marijuana use and onset on the attributed socialization
(puff +)
(N = 1894) 0.719 0.626 - 0.825
Elsevier Scientific Publishers
275
30 (1992) 275 - 287
Ireland Ltd.
Elementary schoolchildren’s use of alcohol, cigarettes and marijuana and classmates’ attribution of socialization Patricia Children’s Health Promotion,
J. Bush and Ronald J. Iannotti
Department of Community and Family Medicine, 415 Kober Cogan Bldg., Georgetown University School of Medicine, Washington, DC 20007 (USA)
(Accepted February 23, 1992)
In each of 2 years beginning in fourth and fifth grades, urban elementary public schoolchildren completed surveys about abusable substance use and health promoting behaviors and completed an instrument that permitted each child to have a socialization score attributed by classmates. A factor weighted 12-item scale was developed from 15 items in three domains (personal, interpersonal and school). The scale was positively correlated over the 2 years and positively correlated with a healthful activities scale in both years. Conditional multiple logistic regression, matching on school classroom, indicated that socialization was negatively associated with use of alcohol without parental permission and cigarettes in both years (grades 4 - 5, grades 5 - 6) and with use of marijuana in year 1. Socialization measured in year 1 was negatively associated with cigarette use in year 2 and with onset of use from year 1 to year 2. Shyness, a non-socialization scale item, was negatively associated with use of cigarettes in both years and with use of alcohol without permission and use of marijuana in year 2. Being ‘good at sports’ was an attribute positively associated with alcohol use without permission and cigarette use in year 2. Results suggest that elementary schoolchildren can ascribe social characteristics to their classmates that are associated with and predict health related behaviors. Key words: alcohol; cigarettes;
marijuana; schoolchildren;
epidemiology
Introduction The relationships between use and onset of use of alcohol, cigarettes and marijuana by elementary schoolchildren and a measure of social adaptation, i.e., socialization, attributed by their classmates, were examined in this study. Although other personal characteristics, e.g., aggressiveness (Kellam et al., 1980) and risk taking (Bush and Iannotti, 1985), have been associated with the use of abusable substances by preadolescents, social influences in three domains (family, school, peer) at the environmental and interpersonal levels have been viewed as more important (Hawkins, in press). Correspondence to: Patrica J. Bush, Children’s Health Promotion, Department of Community and Family Medicine, 415 Kober Cogan Bldg., Georgetown University School of Medicine, Washington, D.C. 20007, USA. 0376X716/92/$05.00 0 1992 Elsevier Scientific Publishers Printed and Published in Ireland
Consistent positive correlations have been observed between preadolescent children’s use and expectations to use abusable substances and use by family members. For example, Bush and Iannotti (1985) found that the number of family users of alcohol, marijuana and cigarettes, was correlated with the children’s use and use intentions. In earlier research, Shute et al. (1981) showed that half of a sample of preschool children exposed to smoking at home indicated they intended to smoke in the future compared to only 11% who were not exposed. For adolescents, the influence of parental drug use has been found to be less than the influence of use by older siblings’ and peers’ (Brook et al., 1983, 1989; Needle et al., 1986). After reviewing the literature, Glynn (1981) concluded that the influence of parents never completely disappears, but that it recedes in favor of peers as children move into adolescence. Ireland Ltd.
276
Lack of academic success and low academic aspirations consistently have been associated with use of abusable substances (Jessor and Jessor, 1977; Kandel, 1978; Paulson and Coombs, 1990; Robins, 1980, Smith and Fogg, 1978), although the relationships may be spurious. Substance use and academic failure may both result from prior problems such as peer rejection. Peers also have been found to influence the substance use of preadolescents (Bush and Iannotti, 1985; Iannotti et al., 1986; Iannotti and Bush, 1992). This influence may be differentiated according to whether it represents: (1) peer behavior, (2) peer beliefs and attitudes, (3) perceived peer behaviors, or (4) perceived peer beliefs and attitudes. Relative to the influence of peers, Iannotti and Bush (1992) found that the perceived use of friends explained more of the variance in the alcohol, cigarette and marijuana use of fourth and fifth grade students than either the actual use of three identified best friends or the actual use of the students’ classmates. However, the actual use of the classmates explained more of the variance than the actual use of the friends. In view of the consistent finding that peer use is among the strongest correlates of substance use among adolescents (Barnes and Welte, 1986; Elliott et al., 1985; Hansen et al., 1987; Kandel, 1978, 1986; Needle et al., 1986; Newcomb and Bentler, 1986) and has some predictive value in preadolescents (Gillmore, 1990; Iannotti and Bush, 1992), the mechanism for how the influence of peers is transmitted is of interest. The question remains, for example, whether the availability of abusable substances as an influence on a child’s use is mediated by psychological and other characteristics of the child that affect relationships with peers. In a panel of children followed from preschool into adolescence, Block and his colleagues (1988; Shedler and Block, 1990) found that psychological differences among frequent drug users, experimenters and abstainers could be traced from early childhood to age 18. Also, Kellam and his associates (1980) found that psychological measured characteristics, e.g., aggressiveness
at ages 6 and 7 predicted drug use a decade later whereas children rated as ‘shy’ were least likely to become drug users. Lerner and Vicary (1984) followed 5-year olds into adulthood and found that a ‘difficult’ childhood, characterized by frequent negative moods and withdrawal, predicted regular alcohol, tobacco and marijuana use in adulthood. These studies argue strongly for early assessment and tracking of children prospectively so that targeted interventions can be developed. However, early psychological assessment for large cohorts of children using standard methods is impractical. The instruments are long and require skilled personnel, or reliance must be placed on subjective reports of parents and teachers. A psychometric instrument was developed by Smith (1967) that permitted eighth and ninth grade students to rate their classmates on 18 socialization characteristics. At follow-up ll- 12 years later, the ratings were found associated with use of cigarettes, alcohol, marijuana, cocaine and ‘hard drugs’; the direction of the relationship consistently indicated that poor socialization preceded use (Smith and Fogg, 1978; Smith, 1986). The results obtained by Smith (1978; 1986), Kellam et al., (1980) and Iannotti and Bush (1992) have relatively high validity because they predicted substance use by characteristics attributed by respondents other than the subjects themselves. In most research inquiring into characteristics associated with use of abusable substances, the characteristics as well as the use are reported by the subjects. In the research discussed here, following Smith (1967; 1978; 1986), children reported their own behaviors and rated their classmates on social attributes and the relationships between these and use were investigated cross-sectionally and longitudinally in preadolescents. This approach was consistent with that of Kellam et al. (1980) who advocated use of ‘natural raters’ to assess the social adaptational status of children, suggesting peers as well as parents and teachers as appropriate raters. Of additional interest was the relationship of healthful behaviors to the use of abusable
substances and socialization. It has been suggested that healthful behaviors, e.g., exercise, protect against abusable substance use (Perry and Jessor, 1985). If healthful behaviors are negatively related to use of abusable substances, then socialization may be a predisposing factor for both of these. Our first hypothesis is that schoolchildren who are rated highest by their classmates on socialization are least likely to use abusable substances; our second hypothesis is that schoolchildren who are rated highest by their classmates on socialization will be most likely to exhibit healthful behaviors. Methods Populations
and data collection
In 1988- 1989, a classroom survey inquiring into abusable substance use was administered to public elementary school students in grades 4 - 5 in Washington, DC. A subset of these students also was given a sociometric questionnaire. In the second year (grades 5 - 6) both instruments were administered to all of the students. To test the hypotheses, analyses were limited to students in the 296 classes of 10 students or more who completed the survey and were ranked by classmates on the sociometric questionnaire in year 2 (N = 5995); of these students, 1904 in 153 classes of 10 or more students also have survey data and sociometric rankings in year 1. In year 1, another 3857 students had survey data only or were in smaller classes. In the first year, subjects were attending classes in 111 schools (94.1% of all DC elementary schools); 76.5% of students were in fourth grade and 23.5% were in fifth grade; less than 1% refused to complete the survey, but about 12% of the students were absent from class at the time of the survey or had transferred. Of the cohort, 86.1% self-identified themselves as black, 3.8% as white, 3.9% as Hispanic, 1.0% as Asian and 5.1% as ‘other.’ The median and modal age was 10 years (range 8- 12), with 51.0% female. In the second year, 109 (92.4%) schools participated; 74.5% of the students were in fifth
grade and 25.5% were in sixth grade; 88.9% self-identified themselves as black, 3.4% as white, 4.1% as Hispanic 0.8% as Asian and 2.7% as ‘other’; 50.6% were female. There were no significant gender differences between subjects completing both a drug survey and a sociometric questionnaire in year 1 and subjects completing them in year 2, (O.R. 0.937; 95% C.I. 0.843, 1.040). For none of the substance-use-variables were there significant differences between subjects completing the drug survey in both years and subjects completing the drug surveys in year 1 only. However, boys represented 51.8% of subjects completing drug surveys in year 1 only, and 48.3% of those completing drug surveys in both years (O.R. 1.157; 95% C.I. 1.038, 1.289). The rates of reported ‘crack’ use were too low to be used in any analyses. Survey Following assurances of confidentiality, students completed the fixed choice question survey in their classrooms. Each question with corresponding choices was projected on a slide screen and read to the class by a member of the research staff. Children marked their responses on special answer sheets created to correspond to these questions without containing any indication as to the content of questions or answers. Items were read as often as necessary until all children had an opportunity to mark their answer sheets. The answer sheets, which did not contain identifiers, were placed by the children in envelopes bearing identifying codes, sealed and dropped through a slot in a box for transport to the research site. These annual identifying codes are linked for longitudinal tracking through limited access computer files that contain the school system identification code and the student’s name. Behaviors assessed were use of cigarettes, alcohol (beer, wine, liquor, wine cooler), marijuana and cocaine or crack and five healthful behaviors, including eating properly, exercising and getting enough sleep. Healthful behaviors were assessed in absolute terms, i.e., ‘Which things are usually true for you?’ The healthful
278
activity scale consisted of five items (year 1: mean 2.37, SD. 1.05; year 2: mean 2.39, SD. 1.08). Students were also asked about their school grades (1 - 5, ‘F’ to ‘A’) they most often got in school. Risk taking was measured by a drawing of a stick figure on a wall that varied in height from ground level to the same height as the figure. The students were told to indicate where of five heights they would jump off if they were the figure on the wall. Students were asked if they had been offered cigarettes, alcohol (beer, wine, liquor, wine cooler) and marijuana by other kids and responses were summed to form a O-3 point scale. To increase validity of responses, questions relating to substance use were posed in a way which assumed the child had used the substance, e.g., ‘How old were you when you first used . ..?’ with the option ‘Never have’ following the choices. Similarly phrased questions also inquired into the frequency of use of each abusable substance, e.g., ‘How many times have you . ..?’ with choices followed by the option ‘Never have.’ A child was not considered to have used a substance unless a positive response was recorded for at least two out of three questions inquiring into its use. In year 1, 0.5% of children with three or more inconsistent responses overall were dropped from all analyses; in year 2, 0.02% were dropped for this reason. Dichotomous variables were created to indicate whether a substance had ever been used. Another alcohol use variable indicated whether the child had ever had alcohol without the child’s mother knowing about it. Additional variables measured heavier use: for alcohol whether and the number of times the child had had more than a sip of alcohol; for cigarettes, whether and the number of times the child had had more than just a puff. Because gender and race have been associated with differences in abusable substance use (Barnes and Welte, 1986; Newcomb et al., 1987), it is important to control for them in analyses. Although a school level of socioeconomic status (SES) was obtained (percentage of children eligible for the federally funded school lunch program), use of conditional logistic analyses, which
took account of the students’ school and classroom, made inclusion of the SES variable unnecessary. Sociometric
questionnaire
Unlike the drug survey, the sociometric questionnaire was self-adminstered (see Table I). The sociometric questionnaire permitted all children in the class to be ranked by their classmates on social competency indicators. Three categories of variables intended to tap personal, interpersonal and school competency and shown to relate to early substance use were identified in the literature. The technique, a much simplified version of that used by Smith (1967) with older students, consisted of having each student record by number (from a numbered list of classmates), the two students who in his/her opinion best represented the opposite ends of a scale, e.g., most likes school and least likes school (see Table I). Children were permitted to select themselves. This process resulted in each child having a score for each item that reflected the percentage of times he/she was nominated by classmates for that item. Suppose, for example, there were 30 children in a class; they nominated 60 children as most liking school and 60 children as least liking school and a particular child was nominated twice (3.3% of the responses) as most liking school and 6 times (10% of the responses) as least liking school. This child’s score for likes school would be - 6.7 (3.3% - 10%). If a child was not nominated at all for an item, the score was zero. At each administration, four versions of the instrument varied the order of the same items to reduce order effects and increase confidentiality. A pilot version contained 30 items and asked students to record four classmates who represented the most and the four who represented the least of each item. The pilot version was administered to 94 low to low-middle socioeconomic status fourth grade students in four classes. The pilot study indicated that the instrument was too long for most students to complete. Therefore, the number of classmates each student was asked to pick for each item was
279
Table I.
Sociometric instrument
(version 1 of 4)* used for peer ratings.
Put down the number of the two students in your class who you think are the MOST and who are the LEAST of each thing. The first one is an example. LEAST
MOST 12 -
3 -
Lucky Liked by teachers Likes school Takes chances Picks on others Fast at doing things Popular Can’t sit still in school Good at sports Picked on by others School work is neat Quits Can’t be trusted Shy Needs help to do things Friendly Three best friends
5 xx
16 xx
*A fourth of the items are rotated from bottom to top to yield four versions used in each classroom.
reduced to four and items were eliminated using the following criteria: (1) items not understood as reported by field staff; (2) items failing to differentiate among the students and (3) multiple items differentiating similarly. Finally, the list was balanced to achieve 5 personal, 5 interpersonal and 5 school items. Using the sociometric data from the larger second year sample, factor analyses (Table II) were used to derive factor scores and to reduce the data. Twelve of the 15 items in the principle components factor had loadings exceeding 0.35 in year 2 and achieved satisfactory internal reliability for children the age of those in this study in both years (Cronbach’s alpha 0.81 in year 2 and 0.70 in year 1). *Three of the items, ‘shy’, ‘good at sports’ and ‘picked on by others’, were dropped from the socialization scale because of low factor loadings and to increase internal reliability. The correlation between the socialization factors using the same 12 items in both years was
0.55 (P 5 0.0001). The range of the socialization factor was -5.41 to +6.01 (mean 0.051, SD. 0.986) in year 1 and -8.04 to + 5.37 (mean 0.000, S.D. 1.00) in year 2. As another indicator to the validity of the socialization scale, it was examined for its relationship to the students’ estimates of their usual school grades. The socialization scale was associated with the subjects’ positively estimates of their usual school grades in year 2 (r = 0.34, P 5 O.OOOl), but not in year 1 (r = 0.01, P 2 0.05). However, socialization in year 1 predicted year 2 usual school grade estimates (r = 0.28, P I 0.0001). There was no reverse relationship, i.e., usual school grades in year 1 did not predict socialization in year 2 (T = 0.01, P s- 0.05). As a traditional measure of socialization, each student had the opportunity to be nominated as a best friend by classmates on the sociometric questionnaire (Table I). Each student nominated three classmates as best friends. Popularity
280 Table II.
Socialization scale item factor loadings in year 1 (Grades 4 - 5, N = 1904) and year 2 (Grades 5 - 6, N = 5995). Year 1
Year 2
Items with positive loadings Fast at doing things Popular School work is neat Friendly Can be trusted Liked by teachers Likes school
0.53 0.46 0.59 0.58 0.61 0.73 0.64
0.48 0.46 0.60 0.64 0.66 0.80 0.71
Items Needs Takes Picks Can’t Quits
0.41 0.22 0.38 0.28 0.13
0.55 0.40 0.54 0.52 0.37
3.00 0.70 0.054 - 5.41, 6.01
3.94 0.81 0.00 - 8.04,5.57
with negative loadings help to do things chances on others sit still in school
Eigenvalue: Cronbach’s alpha: * Mean: Range:
*All negative items reversed; items dropped because of low factor loadings and to increase alpha coefficient included ‘shy’, ‘good at sports’, and ‘picked on by others’.
represented the percentage of times of the total number of nominations made by the class that a child was named as a best friend. Popularity and socialization were correlated in both years (year 1: r = 0.30, P 5 0.0001; year 2: r = 0.29; P I 0.0001). analyses Conditional logistic regression models were used to estimate the degree of association between the socialization scale and use and onset of use of alcohol, cigarettes and marijuana. These models took account of the school classroom of the subjects and provided statistical control of gender and race, covariates with potentially confounding effects. Conditional multiple logistic regression permitted assessment of models that also included the three nonsocialization scale items, i.e., ‘shy’, ‘good at sports’, and ‘picked on by others’ and permitted exploration of interaction effects.
Statistical
Results
Table III shows the percentage of subjects reporting use of alcohol, alcohol without parental permission, cigarettes and marijuana. Boys were more likely to report use in both years, and, except for marijuana, rates of use were higher in year 2 (grade 5- 6) than in year 1 (grade 4 - 5). The conditional logistic regressions of the use variables on the socialization scale (holding the student’s gender and race constant and matching on the school classroom) are shown in Table IV. This table includes the results of crosssectional analyses (both socialization and substance use measured in the same years) and longitudinal analyses (socialization measured in year 1 and substance use measured in year 2 and socialization measured in year 1 and onset of substance use measured from year 1 to year 2). The results indicate that in both years, children attributed higher socialization scores by their
281 Table III. Percentage of urban children reporting use of alcohol, cigarettes, and year 2 (grades 5 - 6, N = 5995). Substance
Alcohol Ever had sip or more Drank without parental permission Cigarettes Ever had puff or more Had more than a puff Marijuana Ever had puff or more
and marijuana in year 1 grades 4 - 5, N = 5761)
Year 2
Year 1 Boys
Girls
Boys
Girls
58.1% 18.6
47.0% 11.4
59.5% 20.4
52.3% 15.3
21.4 15.3
16.0 11.6
25.1 18.9
20.5 15.1
2.8
1.4
2.2
1.0
classmates were less likely to report ever having used alcohol without parental permission, ever having had a puff of a cigarette, ever having had more than a puff of a cigarette and ever having had a puff or more of marijuana. Also, socialization measured in year 1 was negatively associated with ever having smoked cigarettes (puff or more and more than a puff) measured in year 2 and with cigarette smoking onset (puff or more and more than a puff) from year 1 to year 2. Relatively low socialization scores in year 1 increase the relative risk of onset of cigarette smoking by 24% and of onset of smoking more than a puff by 30%. The odds ratios are consistently stronger for smoking more than a puff than for smoking only a puff or more. The conditional multiple logistic regressions examining the effects of adding the three nonsocialization scale items, shy, good at sports and picked on by others, to the models containing the socialization scale are shown in Table V. ‘Picked on by others’ did not contribute significantly to any of the models. ‘Shy’ contributed negatively to all four of the crosssectional year 2 models and to cigarette smoking (puff or more) in year 1, but provided no significant contribution to any of the longitudinal models. Good at sports contributed positively to the models for alcohol use without permission and with cigarette use (puff or more and more than a puff) in year 2 but not in year 1 cross-
sectional analyses. A weak positive association was found between good at sports measured in year 1 and cigarette use (puff or more) in year 2. Neither shy nor good at sports predicted onset of use in any model. Examination of the multiple conditional regression results for year 1 cross-sectional analyses indicated that girls were less likely than boys to have used alcohol with parental knowledge and to have smoked marijuana; black students were less likely to have smoked marijuana. Year 2 cross-sectional analyses indicated that girls and black students were less likely to have smoked marijuana and that black students were less likely to have smoked more than a puff of a cigarette. Being female also protected the student in longitudinal analyses (year 1 socialization measures, year 2 use) from alcohol use without parental permission and from marijuana use. Being black protected against marijuana use and onset in both longitudinal models. Those models that showed gender or race effects were examined for interactions with other independent variables, but none were found. Because the relationship between shyness and substance use was consistent with that found by Kellam et al. (1980), further analysis was performed to increase understanding of the role of shyness. Kellam et al. rejected the idea that shyness protected because it limited opportunities to associate with users; their analyses
(N = 1854) 0.812 0.689 - 0.955 0.012
Use onset from year 1 to year 2 Socialization-Year 1 (N = 1843) O.R. 1.008 95% C.I. 0.883- 1.151 0.901 P-value
(N = 1859) 0.757 0.632-0.906 0.002
(N = 1865) 0.736 0.632 - 0.857 < 0.001
(N = 5972) 0.683 0.633 - 0.737 < 0.001
(N = 1859) 1.497 0.965 - 2.324 0.072
(N = 1869) 1.371 0.861- 2.183 0.184
(N = 5992) 0.769 0.630 - 0.939 0.010
(N = 1894) 0.514 0.331-0.799 0.003
Marijuana (puff + )
*Matching variable = school classroom; gender and race held constant; ‘use’ variables coded 0 - 1 no use, use; socialization coded low to high; Ns vary within models because of missing values. “O.R. = Odds Ratio; C.I. = Confidence Interval.
(N = 1864) 0.759 0.665 - 0.868 < 0.001
(N = 5967) 0.711 0.664 - 0.760 puff)
of alcohol, cigarette, and marijuana use and onset on the attributed socialization
(puff +)
(N = 1894) 0.719 0.626 - 0.825
