PEDIATRICS Vol. 106 No. 5 November 2000, pp. 1017-1021

School Disconnectedness: Identifying Adolescents at Risk

Andrea E. Bonny, MD^*, Maria T. Britto, MD, MPH^*, , Brenda K. Klostermann, PhD, Richard W. Hornung, DrPH, MS, and Gail B. Slap, MD, MS^*

From the ^* Children's Hospital Medical Center, Division of Adolescent Medicine, Cincinnati, Ohio; and  University of Cincinnati College of Medicine, Institute for Health Policy and Health Services Research.

	ABSTRACT

Top Abstract Methods Results Discussion References

Objective.  School connectedness, or the feeling of closeness to school personnel and the school environment, decreases the likelihood of health risk behaviors during adolescence. The objective of this study was to identify factors differentiating youth who do and do not feel connected to their schools in an effort to target school-based interventions to those at highest health risk.

Methods.  The study population consisted of all students attending the 7th through 12th grades of 8 public schools. The students were asked to complete a modified version of the in-school survey designed for the National Longitudinal Study of Adolescent Health (Add Health). The school connectedness score (SCS) was the summation of 5 survey items. Bivariate analyses were used to evaluate the association between SCS and 13 self-reported variables. Stepwise linear regression was conducted to identify the set of factors best predicting connectedness, and logistic regression analysis was performed to identify students with SCS >1 standard deviation below the mean.

Results.  Of the 3491 students receiving surveys, 1959 (56%) submitted usable surveys. The sample was 47% white and 38% black. Median age was 15. Median grade was 9th. The SCS was normally distributed with a mean of 15.7 and a possible range of 5 to 25. Of the 12 variables associated with connectedness, 7 (gender, race, extracurricular involvement, cigarette use, health status, school nurse visits, and school area) entered the linear regression model. All but gender were significant in the logistic model predicting students with SCS >1 standard deviation below the mean.

Conclusions.  In our sample, decreasing school connectedness was associated with 4 potentially modifiable factors: declining health status, increasing school nurse visits, cigarette use, and lack of extracurricular involvement. Black race, female gender, and urban schools were also associated with lower SCS. Further work is needed to better understand the link between these variables and school connectedness. If these associations are found in other populations, school health providers could use these markers to target youth in need of assistance.  Key words:  school connectedness, adolescent health.

School violence has focused national attention on the identification of youth who feel alienated, disconnected, or stigmatized.^1,2 Connectedness has also emerged as a new concept in the literature on adolescent risk behavior.^3-13 Connectedness to school and family demonstrate strong associations with safer behaviors and better health outcomes during adolescence.^11,14-16 School connectedness is defined as an adolescent's experience of caring at school and sense of closeness to school personnel and environment.¹⁴

Resnick et al¹⁵ found that high school students with high connectedness, compared with those with low connectedness scores, had significantly lower rates of emotional distress, suicidal ideation, suicidal behavior, violence, substance use, and early sexual initiation. School connectedness was more protective than any other factor, including family connectedness, against absenteeism, delinquency, polydrug use, unintentional injury, and pregnancy.

The association of school connectedness with health suggests a new strategy for the identification of and intervention with youth at highest risk for adverse outcomes. Public awareness and an emerging professional literature on school connectedness have given rise to primary and secondary prevention strategies designed, respectively, to promote connectedness for all students and to identify and help disconnected students. Hawkins et al⁷ reported that interventions, applied during the elementary school years, successfully improved school bonding at age 18 years, suggesting that school connectedness is malleable. Secondary prevention strategies have been more limited in scope and evaluation. Although school personnel and peers often believe that they can identify disenfranchised students, there is no evidence that their impressions are correct.

The objective of this study was to identify potentially modifiable factors that differentiate youth who do or do not feel connected to their schools. Identification of such factors is a first step toward developing school-based prevention strategies that are targeted correctly to those at highest behavioral and health risk.

	METHODS

Top Abstract Methods Results Discussion References

Sample

The target population consisted of all students in grades 7 through 12 of 8 public, urban or suburban schools (n = 4773). The 8 schools are part of a longitudinal, school-based intervention program. Schools were chosen for their high rates of adverse outcomes such as school failure, teen pregnancy, or child abuse. The median income in the zip codes where the majority of the students resided ranged from $6930 to $31 101 (1990 Census figures). Students in classes designated as either special behavioral or developmentally handicapped were excluded from the study (~5%). Letters to parents explaining the study were sent home with all eligible students. Parents could decline their child's participation by letter or phone call. Students could decline participation by speaking with school personnel or leaving survey items blank. Of the 4534 estimated eligible adolescents, 843 (19%) were absent from school on the study day, and teachers did not distribute surveys to an estimated 200 students (~4%). The proportion of students who were absent from school on the survey day was typical for the participating school systems. Of the remaining 3491 students, 2260 (65%) submitted surveys.

Ninety-eight surveys (4%) were eliminated because of >5 out-of-range answers. Eighty-eight surveys (4%) were eliminated because over one half of the items were left blank. One hundred fifteen surveys did not contain sufficient information to calculate the outcome measure of interest, the school connectedness score (SCS). The final sample for analysis consisted of 1959 subjects. The study protocol was approved by the institutional review boards of the participating institutions and the participating school districts.

Data Collection

Subjects were asked to complete a modified version of the in-school survey designed for the National Longitudinal Study of Adolescent Health (Add Health).^14,17 This comprehensive questionnaire elicits information pertaining to demographic characteristics, school performance and conduct, extracurricular involvement, personal worries and concerns, relationships with friends and school personnel, risk behaviors, health status, and use of health services. Questions pertaining to sexual activity were eliminated at the request of the participating school districts. Eleven questions, further exploring health services utilization, were added to the Add Health Survey following pretesting for clarity among ~30 adolescents. The final survey included 106 questions. The cross-sectional survey was administered in the spring of 1998 at the end of year 1 of a 3-year, longitudinal, school-based intervention program.

Data Analysis

The main outcome measure was SCS derived from 5 survey items as shown in Table 1.

Bivariate analyses were used to explore the association between SCS and each of the following self-reported variables: gender, race, grade, grade-age synchrony, maternal education, paternal education, adoption status, academic performance, extracurricular involvement, cigarette use, alcohol use, perceived health status, and school nurse visits. Students were categorized as young for grade if their age was younger than the minimum age for grade as defined by the school district. Students were categorized as old for grade if they were 2 or more years older than the minimum age required for grade. Academic performance was tabulated from the average self-reported performance in 4 major subjects at completion of the last grading period. Letter grades and values ranged from A = 1 to D or worse = 4. Cigarette and alcohol use were categorized into nonusers and users. Users reported monthly to daily use. Students who reported use only once or twice in the past year were categorized as nonusers. Initially, cigarette and alcohol use were categorized into 3 levels: none, moderate, and high. Because of small numbers in either the moderate or high use categories, these 2 categories were combined. School nurse visits were categorized as none, low (1-2 visits/school year), medium (3-10 visits/school year), and high (>10 visits/school year).

Student t tests were used to assess the difference in SCS between risk factors with 2 levels. Analysis of variance was used for variables with 3 or more potential values. All candidate variables were considered eligible for a stepwise linear regression model designed to identify the set of factors best explaining connectedness. Significant variables with >2 potential values were tested to assess a possible linear relationship with the SCS. When a linear relationship could not be confirmed, the variable was coded as a series of dummy variables.

Logistic regression analysis was performed to identify subjects with SCS >1 standard deviation below the mean. Initially, both linear and logistic models were stratified by school to adjust for possible school-specific effects. Because of problems with collinearity in the linear model, school type (urban or suburban) was substituted for individual schools. Data analyses were conducted using SAS Statistical Software, Version 6.12 (SAS Institute, Inc, Cary, NC).¹⁸

	RESULTS

Top Abstract Methods Results Discussion References

Of the 1959 subjects in the study sample, 882 were male (45%) and 1071 were female (55%). The sample was predominantly white (1137 subjects; 61%) and black (620 subjects; 33%). Median age was 15 years, and median grade was ninth.

The SCS was normally distributed across a range of 5 to 25 with a mean of 15.7 (standard deviation = 4.7; Fig 1). Cronbach  coefficient was .77. Internal consistency of the scale was comparable to that found on national data (Cronbach  = .75 on 5-item SCS scale).

View larger version (36K): [in this window] [in a new window]   Fig. 1.   Distribution of SCS.

The associations of SCS with single variables are summarized in Table 2. SCS was positively associated with male gender, white race, and level of parental education. It was not associated with grade level in school or adoption status. It was negatively associated with grade-age asynchrony. Students who were either young for grade or old for grade reported lower SCS than did students whose age was grade-appropriate.

SCS was positively associated with better academic performance, more extracurricular involvement, and better perceived health. Increasing school nurse visits, cigarette use, and alcohol use were associated with decreasing SCS. Students from suburban schools reported higher SCS than did their urban counterparts.

Table 3 shows the results of the linear and logistic regression analyses for SCS. Six variables entered both models: race, extracurricular involvement, cigarette use, perceived health status, school nurse visits, and school type. One additional variable, gender, entered the linear regression model. The 6 variables comprising the linear regression model accounted for 17% of the variance in SCS. The regression models were examined with the demographic variables (gender, race, grade, and parental education) used as both confounders and independent variables, and no differences in the models were seen.

Simple correlation between all the independent variables was examined, while evaluating the linear regression model for collinearity. Correlation between perceived health status and school nurse visits was weak (Kendall's  = .15). The correlation between race and school type was high (phi coefficient = .79). Despite a high correlation between these 2 variables, both remained independently associated with SCS when the sample was not stratified by gender.

	DISCUSSION

Top Abstract Methods Results Discussion References

Previous studies have demonstrated strong inverse associations between school connectedness and many adolescent risk behaviors.^14,15 These reports, however, do not describe the personal characteristics of students who do or do not feel connected to their schools. In this study of 7th through 12th grade students, significant associations between school connectedness and several modifiable factors were seen. Students with higher SCS reported better academic performance and more extracurricular involvement. They described better health status, fewer school nurse visits, and less use of cigarettes and alcohol. This association between SCS and health status is consistent with previous studies on predictors of perceived health status in adolescents. These studies demonstrate that the perception of physical health among youth is mediated by factors, such as school achievement, participation in sports, supportive social environments, and psychological well-being rather than minor or chronic health problems.^16,19 Furthermore, although some statistically significant differences in mean SCS were small, Resnick et al¹⁴ found that even smaller changes in SCS were clinically relevant and associated with better health outcomes.

A variety of sociodemographic factors were also found to be associated with school connectedness. Gender and race differences were seen. Boys reported feeling more connected than girls. White students reported feeling more connected than did black students. Students who reported lower parental education, age-grade asynchrony, or urban school type also reported feeling less connected to their schools. Although the urban high schools are larger than the suburban high schools, the urban middle schools are either similar in size or smaller than the suburban middle schools. School size, therefore, could not consistently explain the differences in school connectedness between students at urban or suburban schools. No information was known regarding class sizes at the participating schools. The suburban areas are more cohesive neighborhoods than are the urban counterparts, and in general, suburban students tend to attend schools closer to home. Complex differences between urban and suburban schools, including neighborhood structure and services available, may play a role in affecting school connectedness.

The nature of the association between school connectedness and the independent variables needs to be considered. This survey is a cross-sectional survey of self-reported data and does not contain direct measures of behaviors. Previous studies have found that self-reported data that are collected anonymously and outside the home are reliable.²⁰ The cross-sectional nature of data collection, however, precludes any inferences about the cause and effect relationships between school connectedness and the independent variables. It may be that the underlying factors that make a student feel more connected also make that student more likely to be involved in extracurricular activities or perform better in school. In this case, these variables would be associated, but not causally related. In contrast, perhaps interventions that enhance the number of extracurricular activities available in schools will have a positive effect on school connectedness. Future longitudinal studies should be aimed at understanding the nature of the association among these variables.

A recent longitudinal study, conducted in the elementary school years, demonstrated that interventions designed to increase school bonding had enduring effects in reducing violent behavior, heavy drinking, and sexual intercourse by age 18 years.⁷ Interventions included parental education, teacher training, and skills training for children. Intervention students demonstrated better school bonding and school commitment and less high-risk behaviors at age 18 years. Earlier reports from this study also showed positive effects on school bonding during grades 1 through 4.²¹ This study suggests that school connectedness is indeed malleable. It is still unclear, however, how late in a student's school career can interventions positively affect school connectedness and what interventions can best implement such change. Several factors found to be associated with school connectedness in our study are potentially modifiable: extracurricular involvement, cigarette use, perceived health status, and frequency of school nurse visits. It is worthwhile to consider whether modification of these factors in adolescence and at younger ages could result in positive changes in students' sense of connectedness to school.

Other limitations to this study must be considered. Characteristics of the population of students who did and did not complete the in-school survey need to be considered when evaluating the study results. The school systems included in our study, which were public, urban or suburban, may not be representative of other areas. Furthermore, students who did not participate in the survey may have affected the study results. The survey administered to the students consisted of 106 questions, and the questions comprising the SCS were late in the survey. No changes to the order of questions within the survey were made, so as not to affect the psychometric properties of the questionnaire. Students with lower motivation would have had more difficulty completing the entire questionnaire. Resnick et al¹⁵ reported a 75% completion rate on the survey, compared with our 65%. The demographic distribution of the students in the study by Resnick et al was much wider, including all sociodemographic levels. In light of the current study's focus on schools with a much higher prevalence of at-risk youth, the lower completion rate seems expected. No information was available on the survey nonrespondents. Students who were not in attendance at school or who did not complete the survey would seem more likely to be poorly connected to their schools.

The identification of students at highest risk for health compromising behaviors is an essential first step in the design of successful school-based intervention programs. Our exploratory data analyses, aimed at finding factors associated with school connectedness, suggest that several factors may help identify youth at highest risk: extracurricular uninvolvement, cigarette use, poor perceived health, and frequent school nurse visits. Because students with low SCS seem more likely to make frequent school nurse visits, school nurses may be an excellent resource for identifying disconnected youth. If similar associations are found between school connectedness and these variables in other populations, school health providers could use these variables as markers of adolescents in need of assistance. Future studies need to be designed to further define the relationships between school connectedness and the above variables and to identify any causal effects.

	ACKNOWLEDGMENTS

We thank the students, teachers, and administrators of the participating schools for their help with this study.

	FOOTNOTES

Received for publication May 17, 1999; accepted Jan 27, 2000.

Reprint requests to (A.B.) Children's Hospital Medical Center, Division of Adolescent Medicine, 3333 Burnet Ave, Cincinnati, OH. E-mail: bonna0{at}chmcc.org

	ABBREVIATIONS

SCS, school connectedness score; Add Health, National Longitudinal Study of Adolescent Health.

	REFERENCES

Top Abstract Methods Results Discussion References

1. Rees M. Can we buy safer schools? The New York Times. April 27, 1999:A30
2. Goode E. When violent fantasy emerges as reality. The New York Times. April 25, 1999:24
3. Allen J, Aber J, Leadbeater B Adolescent problem behaviors: the influence of attachment and autonomy. Psychiatr Clin North Am 1990; 13:455-467 [Medline]
4. Chase-Lansdale P, Wakschlag L, Brooks-Gunn J A psychological perspective on the development of caring in children and youth: the role of the family. J Adolesc 1995; 18:515-556 [CrossRef]
5. Hawkins J, Weis J The social development model: an integrated approach to delinquency prevention. J Primary Prev 1985; 6:73-97
6. Hawkins J, Catalano R, Miller J Risk and protective factors for alcohol and other drug problems in adolescence and early adulthood: implications for substance abuse prevention. Psychol Bull 1992; 112:64-105 [CrossRef][Medline]
7. Hawkins J, Catalano R, Kosterman R, Preventing adolescent health-risk behaviors by strengthening protection during childhood. Arch Pediatr Adolesc Med 1999; 153:226-234 [Abstract/Free Full Text]
8. Jessor R Risk behavior in adolescence: a psychosocial framework for action. J Adolesc Health 1991; 12:597-605 [CrossRef][Medline]
9. Luthar S, Ziegler E Vulnerability and competence: a review of research on resilience in children. Am J Orthopsychiatry 1991; 61:6-22 [Medline]
10. McBride C, Curry S, Cheadle A, School-level application of a social bonding model to adolescent risk-taking behavior. J Sch Health 1995; 65:63-68 [Medline]
11. Rutter M. Protective factors in children's response to stress and disadvantage. In: Rolf J, Kent MD, eds. Primary Prevention of Psychopathology, III: Social Competence in Children. Boston, MA: Hanover University Press of New England; 1978:49-74
12. Rutter M Resilience: some conceptual considerations. J Adolesc Health 1993; 14:626-631 [CrossRef][Medline]
13. Werner EE, Smith RS. Vulnerable But Invincible: A Longitudinal Study of Resilient Children and Youth. New York, NY: McGraw-Hill; 1982
14. Resnick MD, Bearman PS, Blum RW, Protecting adolescents from harm: findings from the National Longitudinal Study on Adolescent Health. JAMA 1997; 278:823-832 [Abstract]
15. Resnick MD, Harris L, Blum R The impact of caring and connectedness on adolescent health and well-being. J Paediatr Child Health 1993; 29:S3-S9
16. Mechanic D, Hansell S Adolescent competence, psychological well-being, and self-assessed physical health. J Health Soc Behav 1987; 28:364-374 [CrossRef][Medline]
17. Bearman P, Jones J, Udry J. The national longitudinal study of adolescent health: research design. 1997. Available at: http://www.cpc.unc.edu/projects/addhealth/design.html
18. Cody R, Smith J. Applied Statistics and the SAS Programming Language. 4th ed. Upper Saddle River, NJ: Prentice-Hall Inc; 1997
19. Vingilis E, Wade TJ, Adlaf E What factors predict student self-rated physical health. J Adolesc 1998; 21:83-97 [CrossRef][Medline]
20. Brener N, Collins J, Kann L, Reliability of the Youth Risk Behavior Survey questionnaire. Am J Epidemiol 1995; 141:575-580 [Abstract/Free Full Text]
21. Hawkins J, Von Cleve E, Catalano R Reducing early childhood aggression: results of a primary prevention program. J Am Acad Child Adolesc Psychiatry 1991; 30:208-217 [Medline]