The Association Between School Bonding and ...

2 downloads 0 Views 293KB Size Report
Feb 11, 2015 - Keywords: Smoking, adolescents, school bonding, school attachment, school commitment. INTRODUCTION ... large representative sample of secondary schools students in Chile. METHODS ...... [22] Libbey HP. Measuring ...
Substance Abuse

ISSN: 0889-7077 (Print) 1547-0164 (Online) Journal homepage: http://www.tandfonline.com/loi/wsub20

The Association Between School Bonding and Smoking Amongst Chilean Adolescents Jorge Gaete PhD, Alan Montgomery PhD & Ricardo Araya PhD To cite this article: Jorge Gaete PhD, Alan Montgomery PhD & Ricardo Araya PhD (2015): The Association Between School Bonding and Smoking Amongst Chilean Adolescents, Substance Abuse, DOI: 10.1080/08897077.2014.991862 To link to this article: http://dx.doi.org/10.1080/08897077.2014.991862

Accepted author version posted online: 11 Feb 2015. Published online: 11 Feb 2015. Submit your article to this journal

Article views: 11

View related articles

View Crossmark data

Full Terms & Conditions of access and use can be found at http://www.tandfonline.com/action/journalInformation?journalCode=wsub20 Download by: [University of London]

Date: 09 November 2015, At: 14:18

SUBSTANCE ABUSE, 0: 1–9, 2015 Copyright Ó Taylor and Francis Group, LLC ISSN: 0889-7077 print / 1547-0164 online DOI: 10.1080/08897077.2014.991862

The Association Between School Bonding and Smoking Amongst Chilean Adolescents

Downloaded by [University of London] at 14:18 09 November 2015

Jorge Gaete, PhD,1 Alan Montgomery, PhD,2,3 and Ricardo Araya, PhD4 ABSTRACT. Background: The objective of the research was to study the association between school bonding dimensions (school commitment and school attachment) and current adolescent smoking in Chile, controlling for confounding variables using the fifth Chilean School Population National Substance Use Survey, 2003 (CHSS-2003) data set. Methods: The CHSS-2003 is a stratified cross-sectional survey that gathers information about personal, familial, peer, and school factors and cigarette use using a self-reported questionnaire. Complete data from 21,956 adolescent students for all the variables of interest were used in the analyses. An exploratory factor analysis (EFA) was performed in order to explore the construct validity of the questionnaire and create the main exposure and potential confounding variables. Multivariable logistic regression analyses were undertaken to study the association between school bonding and smoking. Results: The construct validity of the school attachment and school commitment scales was mainly supported by the EFA. Multivariable analyses showed strong evidence that, after adjusting for factors from different domains, school commitment (student’s good grades and school attendance) appears to have a clear inverse association with current smoking (odds ratio [OR] D 0.46, 95% confidence interval [CI]: 0.38–0.56). On the other hand, school attachment (their feelings towards their school and their teachers) was not associated with adolescent smoking (OR D 1.16, 95% CI: 0.88–1.53). Conclusions: School commitment was strongly associated with current smoking. It is important to further study this variable with the aim of ascertaining whether or not interventions that improve school commitment may prevent or reduce smoking amongst adolescent students.

Keywords: Smoking, adolescents, school bonding, school attachment, school commitment INTRODUCTION Smoking is a major epidemic throughout the world with a wide range of detrimental consequences upon the health of the population. Smoking seems to start at an early age,1 and efforts to prevent smoking early in life could avert some of these untoward consequences. The prevalence of smoking among adolescents has remained stable or declined in most Western countries.1,2 However, this declining trend is not seen in many developing countries,

1

School of Psychology, Universidad de los Andes, Santiago, Chile Nottingham Clinical Trials Unit, University of Nottingham, Nottingham, United Kingdom 3 School of Social and Community Medicine, University of Bristol, Bristol, United Kingdom 4 Department of Epidemiology and Population Health, London School of Hygiene and Tropical Medicine, London, United Kingdom Correspondence should be addressed to Jorge Gaete, PhD, Associate Professor, School of Psychology, Universidad de los Andes, Monse~nor  Alvaro del Portillo 12455, Las Condes, Santiago, Chile. E-mail: [email protected] 2

particularly among girls.3 For instance, a World Health Organization (WHO) multisite study revealed that the highest rate of current smoking (39.6%) among adolescents was found in a region of Chile.4 Several surveys carried out in Chile over the last decade3 have consistently shown that more than a third of the young population (8th to 12th graders) currently smoke.3 Other forms of consuming tobacco are virtually inexistent in Chile. Several studies and reviews have investigated risk and protective individual, family, peer, and environmental factors in adolescent smoking.5–8 Relatively fewer studies have focused on the association between school-related factors and students’ smoking.6 The school is where adolescents spend a considerable proportion of their time and learn and develop skills that may be used when they become adults. Additionally, the school provides a normative influence and an opportunity for social bonding different from that obtained from the family. School bonding is one of these school-related factors, which has been investigated in greater depth in relation to educational outcomes such as academic performance, school discipline or substance misuse.9 However, there has been a relative scarcity of studies focusing on a possible association between school bonding and smoking behavior.10

Downloaded by [University of London] at 14:18 09 November 2015

2

SUBSTANCE ABUSE

School bonding is a multidimensional construct that refers to the relationship between students and their school. Maddox and Prinz’s review identified 4 dimensions or constructs within this concept: (i) attachment to school (liking or belonging to the school); (ii) attachment to personnel (such as feeling that the teachers support students); (iii) school commitment (beliefs and behaviors in relation to academic performance); and (iv) school involvement (participation in extracurricular activities supported by the school).11 Other authors include sense of belonging as another indicator of the relationship between students and school.12 Therefore, there are several measurements to assess this multidimensional school bonding concept.13,14 Some of these measures have focused on feelings towards schools (school attachment),15,16 others on the importance of academic work for their lives (school commitment),17–19 whilst others on sense of belonging (school belonging).20 However, some experts think that school attachment and school commitment should be considered the key elements of school bonding.21–23 Inconsistent results have been found when studying the association between school attachment16,24,25 or a more general construct of school bonding26,27 and smoking among adolescents. Nonetheless, the protective effect of school commitment against smoking appears more consistent in previous reports.28–30 Additionally, there are no studies exploring explicitly the differential effect of school attachment and school commitment on smoking controlling for each other and for other potential confounding variables. The aims of this study were (1) to explore the questionnaire to ascertain which items could be include in each one of the main domains (i.e., school attachment or school commitment); (2) to assess the association of school bonding, defined as school attachment and school commitment, with current smoking, taking into account other school-related variables as well as variables from personal, school, family, peer, and drug use perceptions domains in a large representative sample of secondary schools students in Chile.

METHODS

Ethical Issues Passive consent was used after ethical approval by the Ministry of Education. Parents and children were informed of their right to withdraw from the study.

Subjects The CHSS-2003 consisted of a self-administered questionnaire given to all students consenting to enter the study between October and November 2003. All students in the randomly selected classes were eligible to participate in the study. Complete data from 21,956 students were included in the analyses presented in this paper.

Instruments The self-reported questionnaire consisted of 198 items and gathered information about 3 main areas: sociodemographic features (10 items), history of substance use (e.g., cigarette use; 54 items), and students’ perceptions of personal and social life (134 items). The questionnaire was built using items from different sources such as the “Monitoring The Future Questionnaire”2 and items developed by a group of Chilean experts responsible for the CHSS-2003. The questionnaire design is in line with recommendations by the United Nations Office on Drugs and Crime for conducting school surveys,31 and it has been validated in a wide range of settings.32 Items referring to students’ perceptions about their personal and social life had not been previously validated. These items assess different constructs, such as school bonding,11 parental monitoring,33 and school climate.34 In the first stage, we assessed the multidimensionality and looked for findings to support the construct validity of this part of the questionnaire through performing an exploratory factor analysis (EFA).35 Subsequently we created scales representing each one of these dimensions and, finally, tested for their internal consistency (reliability). Some additional items known as potential risk factors for smoking were selected from a set of items that did not load on any of the main factors.

Data Source Data come from the Fifth Chilean National Survey on Substance Use in Schools (CHSS-2003).3 This survey used a stratified nationally representative sample of students attending 8th to 12th grades in state, subsidized, and private urban schools in Chile, representing 86 municipalities with >30,000 inhabitants each.

Study Design The sampling frame was the 2002 Chilean School Registry, which indexes all schools in Chile and is updated yearly by the Chilean Ministry of Education. The strata were defined by school type (state, subsided, and private) and grade level (8th, 9th, 10th, 11th, and 12th grades). This method yielded a maximum of 15 strata per municipality. The primary sampling units were classes (40 students per class on average). These classes were randomly selected with a probability proportional to the number of students in each school. The secondary sampling unit corresponds to the students in each of the selected classes. Approximately half of all students in each selected class were randomly chosen for inclusion.

Exploratory Factor Analysis Selection of items A systematic approach for the creation of the main exposure variables and potential confounders was followed. Several aspects were considered to decide if these variables would be included in the EFA: (1) items should measure the intensity or quantity of a construct; (2) items needed to have a theorized relationship with the latent construct intended to be measured35; (3) items should have a clear response format; and (4) item redundancy should be avoided. Items with problems in any of the aspects stated above were excluded from the EFA, and those with “don’t know” response options were recoded as missing data, in view of the difficulty to ascertain what the student was trying to answer when selecting this alternative. A sensitivity EFA with promax rotation was performed using 5 possible EFA scenarios depending on the amount of missing data. These scenarios ranged from all 107 items included to situations with only 95 items included. The set of

GAETE, MONTGOMERY, AND ARAYA

items chosen was based on the largest possible sample size retained and the structure of the common factors obtained through the EFA.

Downloaded by [University of London] at 14:18 09 November 2015

Selection of the items within factors The criterion used for selection of factors was having an eigenvalue over 1 and theoretical plausibility. Items loading 0.32 on a factor were considered part of that factor. For items with factor loadings higher than 0.32 in more than 1 factor (cross-loading), the item was ascribed to the factor where it had the highest loading factor. The items selected in each factor were used to create weighted scales: each item in the factor was weighted according to the loading from the promax rotated factor matrix35,36 and added to generate a total composite score for the scale. Then, in order to facilitate the interpretation and comparison between scales, a standardization method35 was followed to obtain scores ranging from 0 to 1 in each scale. A higher score in the scale reflected a more positive view of the construct. The internal consistency of the generated scales was assessed by calculating the Cronbach’s alpha for each scale; however, this information was considered informative rather than decisive when determining which items to include in each factor.

Outcome variable Current smoking was a binary outcome defined as a student who had smoked 1 or more cigarettes on at least 1 day in the last 30 days previous to the survey.

Exposure and confounding variables The main exposure variables relating to school bonding were identified from the EFA using Maddox and Printz’s definitions of school bonding dimensions.11 Potential confounders were extracted from the EFA and from other parts of the questionnaire. These variables represented different domains: personal, family, school, peers, and drug use perceptions.

Data Analyses Associations between school bonding dimensions and current smoking were tested using univariable and multivariable logistic regression models. All multivariable models included age and sex as confounders. In addition, univariable models were used to identify other potential confounding variables: if any variable was associated with both smoking and any of the school bonding variables at P  .2, then it was entered in a subsequent model as a potential confounder. If adding this confounding variable to the model changed the strength of the association between any of the school bonding variables (school attachment, school commitment, and time doing homework) and current smoking by §10% of the original odds ratio (OR), this variable was kept for inclusion in the process of building the final model. Confounding variables were grouped according to the domain that they represent (personal, school, family, peers, and drug use perceptions). A model for each domain was built. Each of these models included all main exposure variables, age and sex, and the confounders belonging to the same domain. In the process of building the final model, all the main exposures relating to school bonding and all confounders selected

3

through the procedure described were entered in the multivariable regression model. Finally, only those confounders with an independent association at P  .05 with the outcome were kept in the final model. Effect modification by sex and age was also evaluated using appropriate interaction terms in the regression models. To make the results easier to interpret, age was recoded creating quintiles. All analyses were performed using STATA 9.2.37 The analyses used the “svy” suite of commands in order to take account of the stratified sample design.

RESULTS Participants and Descriptive data A total of 23,359 students had complete data in order to be included in the EFA. A small proportion of the questionnaires (n D 1403) were excluded from the regression models because of incomplete data. Therefore, 21,956 students’ questionnaires were entered into the full analyses of this paper. Table 1 presents characteristics of students included in all of the analyses. The sample excluded from the rest of the analyses (n D 1403) did show a few differences with the rest of the sample: there were more boys, the proportion of students living with the biological parents was smaller, and so was the proportion of students with parents working full-time, and whose parents were living together. The proportion of current smokers was also smaller in this group. Of those included in all analyses, over half were female (51.2%) and the mean (SD) age was 15.6 (1.6), with a range of 12 to 21. Most of the students came from families where parents were married and lived together (64.5%). The majority of the students reported that their father (76.8%) or mother (50.2%) was working (full- or part-time) (Table 1). Current smoking was reported by 49.9% of the students. Girls smoked in a higher proportion than boys (Girls, 51.2% vs. Boys, 48.8%; P < .001). Mean age of initiation of cigarette use was 12.8 (2.0) (Table 1), without differences between girls and boys (P D .056).

Identification and Creation of Exposure Variables From EFA Variables excluded from the EFA Out of 134 items, 27 were excluded because they were nominal items (n D 3), did not have a theorized relationship with potential latent constructs (n D 9), and had question design problems (n D 14), and 1 item was regarded as redundant.

Sensitivity EFA Out of the remaining 107 items, 18 items included a “don’t know” response option. After applying the 5 scenarios described in the Methods, the number of subjects included in the EFA increased from 4996 (all 107 items) to 23,359 (all items except those with “don’t know” options). The number of factors was either 25, for the data with all 107 items, or 23, for the other 4 scenarios. The factorial structure was very similar in all solutions.

4

SUBSTANCE ABUSE TABLE 1 Sociodemographic and Smoking Behaviour Variables Sample not used in regression models (n D 1403)

Variable Age in years Age of first consumption of cigarettes Number of days of cigarette use in the last month Grade (School year) 8th grade 9th grade 10th grade 11th grade 12th grade

Mean

SD

15.4 12.7 8.5

1.7 2.2 11.6

15.6 12.8 7.9

1.6 2.1 11.4

13.7 14.7 15.8 16.8 17.8

0.7 0.8 0.8 0.8 0.9

13.6 14.7 15.7 16.7 17.7

0.7 0.7 0.7 0.7 0.7

Downloaded by [University of London] at 14:18 09 November 2015

n Sex Male Female People who lives with the student Biological father and mother Father and stepmother Mother and stepfather Biological father only Biological mother only None of the above Father’s main activity Working full-time Working part-time Looking for work or unemployed Temporarily unavailable for work Pensioner Another activity I do not know or it is not applicable Mother’s main activity Working full-time Working part-time Looking for work or unemployed Temporarily unavailable for work Pensioner Housewife (if she does not work) Another activity I do not know or it is not applicable Parents’ marital status Married and live together Not married but live together Separated or divorced Never been married Widow/er Another situation Lifetime cigarette use Current smoker (had smoked at least 1 day in last 30 days)

Thus, it was decided to use the data that excluded the items with “don’t know” options. Nineteen factors with 2 items were considered in the creation of the exposure variable scales. Three 1item factors were retained as exposure variables using the original categorical variable status. One factor was left without items after applying the criterion for cross-loading. The features of the scales can be seen in Table 2.

Sample used in all analyses (n D 21,956)

%

Mean

n

SD

%

736 667

51.8 48.5

10,703 11,253

48.8 51.2

758 47 70 153 179 181

54.4 3.2 4.8 10.4 13.7 12.5

15,236 215 1369 401 3721 870

69.0 1.0 6.3 1.8 17.3 3.9

676 230 60 22 53 87 225

49.0 15.4 4.7 1.7 3.3 6.2 15.9

12,887 3898 945 326 648 1085 2001

58.6 18.0 4.4 1.5 2.7 5.0 9.0

198 109 26 14 9 217 40 611

15.5 7.8 1.7 0.9 0.7 14.6 2.4 44.9

7008 3871 753 281 217 8729 1097 0

32.2 17.9 3.5 1.3 1.0 39.0 5.0 0

570 71 196 85 139 134 1040 517

43.2 5.3 13.4 6.4 9.0 9.0 71.7 35.2

14,156 1588 3202 1453 574 983 17,412 11,279

64.1 7.4 14.8 6.7 2.6 4.5 77.9 49.9

Main exposure variables The description of the school bonding variables is given in Table 3. School attachment and school commitment were clear factors identified from EFA. School attachment explored the students’ feeling toward their schools, such as liking and belonging to the schools, and relationship with their teachers. School

GAETE, MONTGOMERY, AND ARAYA

5

TABLE 2 Features of the Scales With 2 Items Extracted From the Exploratory Factor Analysis (EFA)

Downloaded by [University of London] at 14:18 09 November 2015

Name of the scale Main school bonding exposures School commitment School attachment Personal Negative feelings Positive feelings Sport interest Religiosity Belong to an alternative group Belong to social groups School School stream School climate School drug environment School aggressiveness Being bullied Family Parental monitoring Peers Time with friends Drug use perception Opinion about drug use Acceptance of marijuana use Drug accessibility Drug offer

Number of items

Mean

SD

Median

Percentile 25th

Percentile 75th

Cronbach’s alpha

3 5

0.71 0.60

0.25 0.21

0.82 0.60

0.44 0.44

0.88 0.76

.4665 .6867

5 5 2 2 2 9

0.51 0.68 0.32 0.27 0.34 0.06

0.32 0.30 0.30 0.35 0.36 0.12

0.57 0.76 0.24 0.11 0.40 0

0.19 0.49 0.08 0 0 0

0.81 1 0.56 0.44 0.40 0.10

.6688 .6229 .4986 .7271 .3660 .6089

2 2 2 6 2

0.68 0.51 0.37 0.91 0.90

0.25 0.26 0.32 0.15 0.18

0.67 0.48 0.47 0.98 1

0.50 0.31 0 0.89 0.89

0.84 0.67 0.50 1 1

.7150 .4875 .6880 .7561 .5261

10

0.76

0.17

0.79

0.66

0.89

.6744

2

0.44

0.29

0.43

0.20

0.63

.5052

7 4 3 4

0.56 0.59 0.49 0.88

0.32 0.26 0.31 0.21

0.57 0.59 0.33 1

0.31 0.41 0.33 0.85

0.86 0.77 0.68 1

.8983 .7207 .8438 .8187

commitment explored self-reported academic performance and level of absenteeism. We included a single item enquiring about “time spent doing homework.” This is another element of “school commitment,” which did not emerge from the EFA. Nonetheless, we decided to include it because of its relevance in giving an alternative insight on school commitment.

Association Between School Bonding and Smoking Results are presented in Table 4. The unadjusted (Model 0) and adjusted by age and sex (Model 1) models showed clear inverse associations between each school bonding variable and current smoking. Adjusting for age and sex practically did not alter the ORs of each school bonding variable on current smoking. In all models adjusting for confounders (Models 2–6), there was attenuation of associations but most associations between school bonding variables a current smoking remained significant at P  .01. In the fully adjusted final model (Final model in Table 4), “school commitment” showed an inverse association with current smoking (OR D 0.46, 95% confidence interval [CI]: 0.38–0.56; P < .001). There was no evidence of association between either school attachment (P D .262) or time doing homework (P D .089) and current smoking. There was evidence of an interaction between age and school attachment on the primary outcome (P < .001). It appears that school attachment is associated with current smoking only among the oldest students. There was a suggestion of an interaction between age and school commitment (P D .071), but no evidence of any interaction between age and time doing homework (P D .548) on current smoking. See Table 5.

There was no evidence of any interaction between sex and either school attachment (P D .188), school commitment (P D .325), or time doing homework (P D .986) on smoking.

DISCUSSION This is a large study of adolescent students in a middle-income country involving a wide range of potential risk and protective factors for smoking. Firstly, no previous studies had explored the psychometric properties of the CHSS-2003 questionnaire. We found that it has a multidimensional structure. Regarding the main exposures of this study, school attachment and school commitment, the EFA identified that items loaded well into these latent factors. For the purpose of this research, 2 scales were considered as main exposure variables in the regression analyses: school attachment and school commitment. School attachment had a high Cronbach’s alpha, demonstrating good internal reliability. On the contrary, school commitment had a Cronbach’s alpha of .47, a value just below what is often regarded as acceptable. Some authors have argued that researchers cannot base their decisions for using or not a scale only under this criterion,38 but the content of the items (face validity) and the factor loadings in the case of using an EFA to assess the item structure need also to be considered.36 Having said that, one explanation for the low reliability of the school commitment subscale may be related to the few number of items (3 items) included. The number of variables included in a scale might affect the reliability, as suggested by some authors.39 The same explanation could be given for other subscales built out from this questionnaire that showed similarly low reliability scores. However, due to the fact that this study is a secondary analysis of a

6

SUBSTANCE ABUSE TABLE 3 Description of School Bonding Variables

Main exposure

Item

School attachment Q. 25

Q. 26 Q. 35 Q. 36 Q. 37

Downloaded by [University of London] at 14:18 09 November 2015

School commitment Q. 31

Q. 32 Q. 34

Time doing homework

Description

Mean (SD)

Cronbach’s alpha

5-item scale. Standardized from 0 to 1. How happy are you usually about going to school? A: 1 D Not at all; 2 D Slightly happy; 3 D Moderately happy; 4 D Fairly happy; 5 D Very happy. Generally speaking, do you feel part of your school? A: 1 D No; 2 D More or less; 3 D Yes. Are there any teachers in your school with whom you could discuss a personal problem and receive advice? A: 1 D No; 2 D Yes. How would do you describe the relationship between you and your teachers? A: 1 D Poor; 2 D Good; 3 D Very good; 4 D Excellent. Would you say most of your teachers are concerned about their students’ welfare? A: 1 D A little concerned; 2 D Somewhat concerned; 3 D Fairly concerned; 4 D Very much concerned. 3-item scale. Standardized from 0 to 1. What is your overall end of year grade point average (GPA) at end of year? A: Less than 4.5 D 1; Between 4.4 and 4.9 D 2; Between 5.0 and 5.4 D 3; Between 5.5 and 5.9 D 4; Between 6.0 and 6.4 D 5; Between 6.5 and 7.0 D 6. During the last year, have you often been absent, for any reason? A: 1 D Yes; 2 D No. How many times during the last school year, have you not attended lessons without prior permission (bunking off)? A: 1 D Many times; 2 D Several times; 3 D Occasionally; 4 D Never. 1 item. Categorical variable. In a normal day, after school, how many hours do you spend doing homework? A: 0 D 0 hours; 1 D Less than half an hour; 2 D Between half and 1 hour; 3 D Between 1 and 2 hours; 4 D Between 3 and 4 hours; 5 D More than 3 hours.

0.60 (0.21)

.6867

0.71 (0.25)

.4665

data set already collected, the inclusion of new items was not possible. Secondly, main results showed strong evidence that, after adjusting for factors from different domains, school commitment (student’s good grades and school attendance) appears to have a clear inverse association with current smoking. Although there was a suggestion that more time doing homework was associated

NA

NA

with less smoking, the evidence for this was less compelling. Finally, school attachment (their feelings towards their school and their teachers) was not associated with smoking. These findings help to see the complexity of the effect of school bonding and its dimensions. The school bonding dimensions used in this study included Maddox and Prinz’s 3 constructs: attachment to school and

TABLE 4 Unadjusted and Adjusted Logistic Regression Models of the Effect of the School Bonding Variables on Current Smoking School bonding variable

Model 0

Model 1

Model 2

Model 3

Model 4

Model 5

Model 6

Final model

School attachment School commitment Time doing homework 0 hours Less than half an hour Between half and 1 hour Between 1 and 2 hours Between 2 and 3 hours More than 3 hours

0.26 0.14

0.48 0.25

0.61 0.27

0.68 0.31

0.75 0.31

0.46 0.30

0.94 0.36

1.22 (0.90–1.66) 0.37 (0.30–0.45)

1.00 0.71 0.48 0.41 0.32 0.29

1.00 0.83 0.62 0.53 0.41 0.38

1.00 0.86 0.67 0.59 0.47 0.44

1.00 0.89 0.70 0.62 0.49 0.46

1.00 0.90 0.71 0.63 0.50 0.48

1.00 0.83 0.65 0.56 0.45 0.41

1.00 0.96 0.83 0.76 0.63 0.59

1.00 1.00 (0.90–1.12) 0.94 (0.85–1.05) 0.94 (0.84–1.06) 0.84 (0.68–1.02) 0.80 (0.61–1.05)

Note. Model 0: Unadjusted. Model 1: This model included sex and age, and all 3 school bonding variables. Models 2–6 include age, sex, plus adjustment for number of items relating–the following domains: personal (Model 2), school (Model 3), family (Model 4), peers (Model 5), and drug use perception (Model 6). Final model: This model included sex and age, all 3 school bonding variables and the following confounders: negative feelings, religiosity, sport interest, and going home after school, school climate, school aggressiveness, school misbehaviour, drug lessons, past family member drug use, current family member drug use, time with friends, relationship with friends, opinion about drug use, acceptance of marijuana use, drug accessibility, drug offer, and marijuana accessibility.

GAETE, MONTGOMERY, AND ARAYA

7

TABLE 5 Interactions Between Age and School Attachment and School Commitment on Current Smoking Age group quintile (age)

Downloaded by [University of London] at 14:18 09 November 2015

1 (12–14) 2 (15) 3 (16) 4 (17) 5 (18–21)

n

School attachment on current smoking

School commitment on current smoking

5285 4177 4551 4791 3152

0.91 [0.70–1.19] 0.84 [0.53–1.34] 1.45 [0.96–2.18] 1.62 [0.95–2.75] 1.92 [1.32–2.80]

0.40 [0.28–0.58] 0.40 [0.24–0.68] 0.44 [0.28–0.68] 0.41 [0.30–0.56] 0.72 [0.49–1.05]

attachment to staff (both included as a single scale of school attachment after the EFA), and school commitment (which involved 2 variables in this study: school commitment and time doing homework). Thus, conclusions based on this study can only be reached for these dimensions and not for others included in the school bonding construct proposed by other authors, such as school involvement.11 Nonetheless, other authors accept that school attachment and school commitment are the main dimensions of how students bond to school.22 The comparison of these results with those from other studies is difficult because of the wide variation in definitions used for different aspects of school bonding. For example, there are only 2 studies24,40 that included similar, but not identical, concepts as those used in this research to describe what we have called “school attachment” and “school commitment.” In the cross-sectional analysis of West et al.,40 the results showed that when “school involvement” (similar to “school attachment”) and “school engagement” (similar to “school commitment” but without the academic performance component) were added to the multivariable model, only “school engagement” remained associated with smoking. Bryant et al.24 explored the association between 1 “school bonding” variable (similar to “school attachment”) and 3 “school commitment” variables (school interest, status of academic success at school, and school effort) and smoking. However, the “school commitment” variable used in Bryant et al.’s study did not incorporate the academic performance and school absenteeism aspects included in our research, but did include other aspects as mentioned previously and as suggested by the Maddox and Prinz’s review. In their multivariable analysis, neither “school bonding” nor any of the “school commitment” variables were associated with smoking. Five cross-sectional studies assessed the association between school bonding variables and smoking, controlled by “academic performance” in their analyses.41–45 In all cases, school bonding was associated with smoking even after adjusting for academic performance. Two of these studies showed that only academic performance remained associated with smoking in the multivariable analysis that included also school attachment.24,46 In another study, neither of these variables were associated with smoking.47 Finally, Johnson and Hoffmann30 showed that a higher school commitment (lower “school misconduct”) was a protective factor for smoking behavior whereas higher academic performance was a risk factor for smoking. As far as school absenteeism (truancy) is concerned, fewer studies have controlled the association between school bonding variables and smoking by this variable. Only 3 cross-sectional studies included a general concept of school bonding26,41 or school commitment49 and truancy in the same multivariable analysis. Two of them showed that the association between school bonding,

truancy, and smoking were all significant in the fully adjusted multivariable models,41,48 whereas Resnick et al. showed that only truancy remained associated in their fully adjusted models.26 None of the studies reviewed reported interactions between school attachment and age. Some authors have suggested that the expression of school bonding dimensions change during adolescence as a result of the degree of influence that schools have on students’ lives.13 Our finding that 18-year-old or older students with better school attachment had a higher likelihood of smoking supports the idea that school bonding dimensions are not static but may change with age.13 One possible explanation is that for students at older age, schools may be more tolerant to smoking behavior given the fact that students are approaching adulthood; this tolerance may help students to have more positive feelings towards their schools. However, this explanation should be tested in future studies.

Strengths and Limitations This is a large study and the first to specifically assess in a systematic way the effect of possible confounders from several domains (school, family, peer, and personal) on the association between school bonding dimensions and smoking behavior. Most previous studies have assessed the impact of confounders stemming from a single domain using different approaches. Also, most previous studies do not approach confounders in a systematic way, so it is difficult to interpret the relative impact of each one of these possible confounders. However, there are some limitations. This is a cross-sectional study, so no causal associations can be inferred, but the unique population, the lack of studies from low- and middle-income countries, and the large sample size make it worth considering. As in all such large surveys, it is impossible to cover all potentially interesting variables. Some possible confounders were not included, especially those related to role modeling and domainspecific normative values on adolescents’ behavior. There were no data relating to parents’, siblings’, or friends’ smoking behavior or about their attitudes towards cigarette or drug use. These variables have been considered as important in some social theories of substance use among adolescents.5 In view of this, we were unable to test these models. Given that there was a strong suggestion that data were missing at random, it was not considered necessary to replace missing data through regression models or imputations. In any case, it is worth emphasizing that whatever the method chosen to replace missing values, this decision is not exempted of other problems and limitations. School commitment could have an important role in preventive programs. Bond et al. carried out a study to improve school commitment and to assess its impact on several outcomes, ranging

Downloaded by [University of London] at 14:18 09 November 2015

8

SUBSTANCE ABUSE

from depression to school relationships.49 This intervention, Gatehouse Project, was aimed at individuals and the school as whole. The key elements were “the establishment and support of a school-based adolescent health team, the identification of relevant risk and protective factors in each school’s social and learning environment from student surveys, and, through the use of these data, the identification and implementation of effective strategies that address the school environment issues.”50 Most of the activities were based on supporting academic achievement. It showed a consistent 3% to 5% risk difference between the intervention and the control group for any drinking, and between regular smoking and friends’ alcohol and tobacco use.49 Overall, we found good evidence of the psychometric properties of the questionnaire used in the CHSS-2003, even though we consider that this questionnaire could be improved by including more items regarding the constructs studied here, such as school commitment. Mainly, we found that school commitment has a strong effect on reducing the likelihood of smoking among adolescents. Therefore, interventions supporting academic achievements and school commitment may help to reduce tobacco use as well as achieving other important milestones. Future research should include more comprehensive studies of school bonding and its multidimensional composition, and its effect on using other substances of abuse, and ideally use a longitudinal design that allows a better understanding of causality.

ACKNOWLEDGMENTS The authors would like to acknowledge to CONACE (Consejo Nacional para el Control de Estupefacientes) for providing access to the dataset of the Fifth Chilean National Survey on Substance Use in Schools.

[3]

[4]

[5]

[6]

[7]

[8]

[9]

[10]

[11]

[12]

[13]

[14]

FUNDING This research was part of Jorge Gaete’s PhD degree at the University of Bristol. The PhD study was funded by Alban Scholarship (European Union), Beca Presidente de la Republica de Chile (Chile), and Universidad de los Andes Scholarship (Chile). The authors declare that they have no conflicts of interest.

[15]

[16]

[17]

AUTHOR CONTRIBUTIONS

[18]

Jorge Gaete conceived the research idea and the design, performed the data analyses, interpretation, and writing. Ricardo Araya contributed to the research conception and design, and reviewed the analyses and the writing. Alan Montgomery reviewed the analyses and the writing.

[19]

[20]

REFERENCES [21] [1] Simons-Morton BG. Prospective analysis of peer and parent influences on smoking initiation among early adolescents. Prev Sci. 2002;3:275–283. [2] Johnston LD, O’Malley PM, Bachman JG, Schulenberg JE. Monitoring the Future National Results on Adolescent Drug Use: Overview

[22]

of Key Findings, 1999. Bethesda, MD: National Institute on Drug Abuse; 2000. CONACE (Consejo Nacional para el Control de Estupefacientes). Quinto Estudio Nacional de Drogas en la Poblacion Escolar de Chile, 2003. Santiago, Chile: Ministerio de Interior, Chile; 2003. The Global Youth Tobacco Survay Collaborative Group. Tobacco use among youth: a cross country comparison. Tob Control. 2002;11:252–270. Petraitis J, Flay BR, Miller TQ. Reviewing theories of adolescent substance use: organizing pieces in the puzzle. Psychol Bull. 1995;117:67–86. Tyas SL, Pederson LL. Psychosocial factors related to adolescent smoking: a critical review of the literature. Tob Control. 1998;7:409–420. Aveyard P, Markham WA, Cheng KK. A methodological and substantive review of the evidence that schools cause pupils to smoke. Soc Sci Med. 2004;58:2253–2265. Sellstrom E, Bremberg S. Is there a “school effect” on pupil outcomes? A review of multilevel studies. J Epidemiol Community Health. 2006;60:149–155. Catalano RF, Haqqerty KP, Oesterle S, Fleming CB, Hawkins JD. The importance of bonding to school for healthy development: findings from the Social Development Research Group. J Sch Health. 2004;74:252–261. Dornbusch SM, Erickson KG, Laird J, Wong CA. The relation of family and school attachment to adolescent deviance in diverse groups and communities. J Adolesc Res. 2001;16:396–422. Maddox SJ, Prinz RJ. School bonding in children and adolescents: conceptualization, assessment, and associated variables. Clin Child Fam Psychol Rev. 2003;6:31–49. Goodenow C. The psychological sense of school membership among adolescents: scale development and educational correlates. Psychol Sch. 1993;30:79–90. Oelsner J, Lippold MA, Greenberg MT. Factors influencing the development of school bonding among middle school students. J Early Adolesc. 2011;31:463–487. O’Farrell S, Morrison G. A factor analysis exploring school bonding and related constructs among upper elementary students. Contemp School Psychol. 2003;8:53–72. Dickens DD, Dieterich SE, Henry KL, Beauvais F. School bonding as a moderator of the effect of peer influences on alcohol use among American Indian adolescents. J Stud Alcohol Drugs. 2012;73:597–603. Chen C-Y, Wu C-C, Chang H-Y, Yen L-L. The effects of social structure and social capital on changes in smoking status from 8th to 9th grade: results of the Child and Adolescent Behaviors in Longterm Evolution (CABLE) study. Prev Med. 2014;62:148–154. Cavendish W, Nielsen AL, Montague M. Parent attachment, school commitment, and problem behavior trajectories of diverse adolescents. J Adolesc. 2012;35:1629–1639. Sheryl AH, Stephanie MP, Herrenkohl TI, Toumbourou JW, Catalano RF. Student and school factors associated with school suspension: a multilevel analysis of students in Victoria, Australia and Washington State, United States. Child Youth Serv Rev. 2014;36:187–194. Henry K, Knight K, Thornberry T. School disengagement as a predictor of dropout, delinquency, and problem substance use during adolescence and early adulthood. J Youth Adolesc. 2012;41:156–166. Millings A, Buck R, Montgomery A, Spears M, Stallard P. School connectedness, peer attachment, and self-esteem as predictors of adolescent depression. J Adolesc. 2012;35:1061–1067. Hawkins JD, Guo J, Hill KG, Battin-Pearson S, Abbott RD. Longterm effects of the seattle social development intervention on school bonding trajectories. Appl Dev Sci. 2001;5:225–236. Libbey HP. Measuring student relationship to school: attachment, bonding, connectedness, and engagement. J Sch Health. 2004;74:274–283.

Downloaded by [University of London] at 14:18 09 November 2015

GAETE, MONTGOMERY, AND ARAYA [23] Maurizi LK, Grogan-Kaylor A, Granillo MT, Delva J. The role of social relationships in the association between adolescents’ depressive symptoms and academic achievement. Child Youth Serv Rev. 2013;35:618–625. [24] Bryant AL, Schulenberg JE, O’Malley PM, Bachman JG, Johnston LD. How academic achievement, attitudes, and behaviors relate to the course of substance use during adolescence: a 6year, multiwave national longitudinal study. J Res Adolesc. 2003;13:361–397. [25] Ennett ST, Flewelling RL, Lindrooth RC, Norton EC. School and neighborhood characteristics associated with school rates of alcohol, cigarette, and marijuana use. J Health Soc Behav. 1997;38:55–71. [26] Resnick MD, Bearman PS, Blum RW, et al. Protecting adolescents from harm. Findings from the National Longitudinal Study on Adolescent Health. JAMA. 1997;278:823–832. [27] Leatherdale ST, Cameron R, Brown KS, McDonald PW. Senior student smoking at school, student characteristics, and smoking onset among junior students: a multilevel analysis. Prev Med. 2005;40:853–859. [28] Moore L, Roberts C, Tudor-Smith C. School smoking policies and smoking prevalence among adolescents: multilevel analysis of crosssectional data from Wales. Tob Control. 2001;10:117–123. [29] LeMaster PL, Connell CM, Mitchell CM, Manson SM. Tobacco use among American Indian adolescents: protective and risk factors. J Adolesc Health. 2002;30:426–432. [30] Johnson RA, Hoffmann JP. Adolescent cigarette smoking in U.S. racial/ethnic subgroups: findings from the National Education Longitudinal Study. J Health Soc Behav. 2000;41:392–407. [31] United Nations Office on Drugs and Crime. GAP Toolkit Module 3: Conducting School Surveys on Drug Abuse. New York: United Nations; 2003. [32] Johnston LD, Driessen F, Kokkevi A. Surveying Student Drug Misuse: A Six-Country Pilot Study. Strasbourg, France: Council of Europe; 1994. [33] Hayes L, Hudson A, Matthews J. Parental monitoring: a process model of parent-adolescent interaction. Behav Change. 2003;20:13–24. [34] Freiberg HJ. School Climate: Measuring and Sustaining Healthy Learning Environments. London, Philadelphia: Falmer Press; 1999. [35] Pedhazur EJ, Pedhazur Schmelkin L. Construct validation. In: Pedhazur EJ, Pedhazur Schmelkin L, eds. Measurement, Design, and Analysis: An Integrated Approach. London: Lawrence Erlbaum Associates; 1991:52–80. [36] de Vaus D. Building scales. In: Surveys in Social Research. Vol. 5. London: Routledge; 2002:180–199.

9

[37] StataCorp. STATA [computer program]. Version 9.2. College Station, TX: StataCorp; 2007. [38] Schmitt N. Uses and abuses of coefficient alpha. Psychol Assess. 1996;8:350–353. [39] Velicer WF, Fava JL. Affects of variable and subject sampling on factor pattern recovery. Psychol Methods. 1998;3:231–251. [40] West P, Sweeting H, Leyland A. School effects on pupils’ health behaviours: evidence in support of the health promoting school. Res Papers Educ. 2004;19:261–291. [41] Maes L, Lievens J. Can the school make a difference? A multilevel analysis of adolescent risk and health behaviour. Soc Sci Med. 2003;56:517–529. [42] Abdelrahman AI, Rodriguez G, Ryan JA, French JF, Weinbaum D. The epidemiology of substance use among middle school students: the impact of school, familial, community and individual risk factors. J Child Adolesc Subst Abuse. 1998;8:55–75. [43] Lewinsohn PM, Brown RA, Seeley JR, Ramsey SE. Psychosocial correlates of cigarette smoking abstinence, experimentation, persistence and frequency during adolescence. Nicotine Tob Res. 2000;2:121–131. [44] Pinilla J, Gonzalez B, Barber P, Santana Y. Smoking in young adolescents: an approach with multilevel discrete choice models. J Epidemiol Community Health. 2002;56:227–232. [45] Zhu BP, Liu M, Wang SQ, et al. Cigarette smoking among junior high school students in Beijing, China, 1988. Int J Epidemiol. 1992;21:854–861. [46] Bryant AL, Schulenberg J, Bachman JG, O’Malley PM, Johnston LD. Understanding the links among school misbehavior, academic achievement, and cigarette use: a national panel study of adolescents. Prev Med. 2000;1:71–87. [47] Bryant AL, Zimmerman MA. Examining the effects of academic beliefs and behaviors on changes in substance use among urban adolescents. J Educ Psychol. 2002;94:621–637. [48] Hagquist C. Variations in adolescents’ smoking and alcohol behaviour between Swedish schools: an ecological analysis. Drugs Educ Prev Policy. 1997;4:139–150. [49] Bond L, Patton G, Glover S, et al. The Gatehouse Project: can a multilevel school intervention affect emotional wellbeing and health risk behaviours? J Epidemiol Community Health. 2004;58:997–1003. [50] Bond L, Glover S, Godfrey C, Butler H, Patton GC. Building capacity for system-level change in schools: lessons from the Gatehouse project. Health Educ Behav. 2001;28:368–383.