Perceptions of School Climate Shape Adolescent Health Behavior: A Longitudinal Multi-School Study

Abstract

Background:

Adolescent behaviors and academic outcomes are thought to be shaped by school climate. We sought to identify longitudinal associations between school climate measures and downstream health and academic outcomes.

Methods:

Data from a longitudinal survey of public high school students in Los Angeles were analyzed. Eleventh grade health and academic outcomes (dependent variables, e.g., substance use, delinquency, risky sex, bullying, standardized exams, college matriculation), were modeled as a function of 10^th grade school climate measures (independent variables: institutional environment, student-teacher relationships, disciplinary style), controlling for baseline outcome measures and student/parental covariates.

Results:

The 1114 student respondents (87.8% retention), were 46% male, 90% Latinx, 87% born in the USA, and 40% native English speakers. Greater school order and teacher respect for students were associated with lower odds of multiple high risk behaviors including 30-day alcohol use (OR 0.81 95%CI [0.72, 0.92] and OR 0.73 [0.62, 0.85]) and 30-day cannabis use (OR 0.74 [0.59, 0.91] and OR 0.76 [0.63, 0.92]). Neglectful disciplinary style was associated with multiple poor health and academic outcomes while permissive disciplinary style was associated with favorable academic outcomes.

Implications for School Health Policy, Practice, and Equity:

School health practitioners may prospectively leverage school environment, teacher-student relationships, and disciplinary style to promote health and learning.

Conclusions:

Our findings identify specific modifiable aspects of the school environment with critical implications for life course health.

Adolescence is a critical period of development marked by the formation of self-concept and identity, independence from parental guidance, and growth in cognitive and socioemotional skills such as empathy, resilience, and creativity.¹ However, some adolescents also begin to engage in risky behaviors, such as use of tobacco, cannabis, alcohol and other substances. ^2–4 These behaviors are significant, as they can negatively influence this important developmental period and contribute to a vicious cycle whereby risky behaviors interfere with school engagement and academic performance and vice versa. This negative feedback loop is suggested by Richard Jessor’s Theory of Problem Behavior,^5,6 which proposes that school climate, including the social environment of peers, contributes to adverse adolescent behaviors and outcomes including school disengagement, risky behaviors, and academic failure. These adolescent behaviors in turn influence the school climate, as when groups of students normalize delinquent behaviors, undermining academic engagement more broadly. This vicious cycle in adolescence can have significant downstream effects in adulthood, potentially affecting educational and socioeconomic opportunities as well as overall health outcomes.^7,8

While Jessor’s theory suggests reciprocal effects between a negative school climate and adolescent risky behaviors, it may also suggest that a positive school climate could create a virtuous cycle of improved academic success, greater school engagement, academically and prosocially supportive peers, and better academic and behavioral outcomes among teens.^9–11 This is supported by prior literature which has shown that positive school climate is linked to better academic performance,¹² student wellbeing, and school engagement, and lower rates of problem behaviors such as disruptive, antisocial, violent, bullying, or delinquent behavior.^13–15 Although there is no standardized measure of school climate, there are several domains which have been used to characterize school climate and show predictive potential,¹⁶ among them: the institutional environment (e.g., school chaos/order, safety),^17,18 student-teacher relationships (e.g., teacher respect for students, teacher support of college),^19–21 and disciplinary styles.²² However, prior studies have primarily only examined a limited set of school climate variables and adolescent risky behaviors and most have been limited to cross-sectional designs.^12,23,24 As a result, it is still unknown which aspects of school climate might be targeted to improve specific academic or health outcomes.

The present study sought to identify and compare associations between school climate measures across multiple domains and multiple downstream health and academic outcomes longitudinally. We examined data from an on-going longitudinal natural experiment (the RISE-UP - Reducing Health Inequalities through Social and Educational Change Follow Up - Study), which has followed a cohort of mostly Latinx students starting at the beginning of 9^th grade through age 23 currently. Racial/ethnic and socioeconomic disparities exist in youth outcomes and comparative studies suggest that ethnicity may influence perception of school climate.^16,25–27 Furthermore, within the Latinx experience there exists a diversity of national and indigenous heritages, degrees of acculturation and integration, and immigration stories. While ethnicity was not the focus of this analysis, we believe study of this predominantly Latinx population makes a valuable contribution to the extant literature. Additionally, RISE-UP examined a range of school climate characteristics, including school safety and order, student-teacher relationships and support, and disciplinary style. We examined a variety of different adolescent behaviors (alcohol and cannabis use, alcohol and cannabis misuse, delinquent behaviors, violence, high-risk sexual behaviors, bullying), and academic outcomes (truancy, changing schools, grades, standardized test scores, and matriculation into a 4-year college), so as to permit a more holistic analysis that incorporates health and behavior with cognitive development.

METHODS

Participants

This is a secondary analysis of data from the RISE-UP (Reducing Health Inequalities through Social and Educational Change Follow Up) study, a longitudinal natural experiment designed to assess the effects of high-performing high schools on health behaviors among low-income, minority adolescents in Los Angeles. Five high-performing charter high schools were selected based on: (1) enrollment of predominantly economically disadvantaged students (i.e., qualifying for free or reduced lunch), (2) academic performance in the top tertile of public schools in Los Angeles County based on 2012 Academic Performance Index derived from standardized test scores, and (3) use of an admissions lottery. Eighth grade students who were applying for 9^th grade admission into high school were randomly sampled from the admissions lottery list of “winners” and “losers” during two consecutive years in the spring before entry into high school (spring 2013 and 2014). To be eligible, students had to speak English or Spanish fluently and reside in Los Angeles County. Of 1509 eligible students, 1270 were enrolled and consented to participate in the study (16% refusal rate). Further details of the original study are published elsewhere.¹¹ The institutional review board of the RAND Corporation and the University of California Los Angeles approved this study. Written parental consent and student assent were obtained from all participants.

Procedure

Participants completed a baseline, face-to-face, computer-assisted survey from March of 8^th grade through November of 9^th (baseline 9^th grade survey). Similar follow-up surveys were conducted in the spring semester of 10^th grade and 11^th grade (2015 to 2017). Interviews were conducted in the patient’s primary language with the aid of bilingual research assistants and in a sufficiently private location of the participant’s choice. A computer-assisted self-interview was used to minimize social desirability bias for potentially sensitive topics related to substance use and sexual and delinquent behaviors.^28–30 A total of 1159 students completed the survey in 10^th grade and 1114 students completed the survey in 11^th grade for an 87.8% retention rate through 11^th grade.

Instrumentation

Adolescent behaviors and academic outcomes.

At each survey, students reported their frequency of alcohol and cannabis use in the last 30 days,³¹ dichotomized (no use vs any use). Students also completed an alcohol misuse scale³² and a cannabis misuse scale (alpha = 0.85), which assessed high risk substance use behaviors (use of alcohol/cannabis on school property, using by oneself, binge use) and its negative consequences (blacking out, missing school, regret, getting into trouble at school, getting into trouble at home, and poor concentration). Scale items were dichotomized (endorsed vs not) and summed to produce a total score with higher scores representing greater misuse characteristics. Students reported on delinquent behaviors that are associated with negative life outcomes using the delinquent behavior index from the National Longitudinal Study of Adolescent to Adult Health³³ and included: painting graffiti, damaging someone else’s property, shoplifting or stealing, running away from home, driving a car without the owner’s permission, burglary, armed robbery, selling illicit drugs, participation in a gang in the last year, and having ever participated in a gang fight. The score was dichotomized (zero vs one or more behaviors). Students were asked if they carried a weapon such as a real gun or knife in the last 30 days and if they had been in a physical fight in the last 12 months. These questions were combined into one dichotomous variable of any of the two behaviors (none vs one or both behaviors). Students responded to several questions about high-risk sex behaviors including not using contraception, ever becoming pregnant, and having multiple sexual partners (none vs one or more high risk sex behaviors). Students also answered two questions about bullying at school in the last 12 months, which were dichotomized: 1) whether someone had bullied or picked on them (none vs. any) and 2) whether they themselves had bullied or picked on someone (none vs. any).

We also collected information on several academic outcomes. For truancy, students reported if they had cut or skipped classes in the last 12 months, dichotomized (never vs one or more times). Students also responded whether they transferred to another school for any reason in the last academic year, dichotomized (never vs one or more times). We obtained student grade point average (GPA) from official school transcript records (1091 out of 1270 students from baseline sample). We used self-reported GPA when we could not obtain school transcripts (163 of 179 with missing transcripts). We obtained standardized test scores for each student for 8^th grade and 11^th grade from the California Department of Education. Math and English proficiency were determined by the California Standardized Testing and Reporting Program (8^th grade) and the California Assessment of Student Performance and Progress (11^th grade). We compared those who failed to meet 11^th grade standard versus those who were proficient or above. We obtained data on college matriculation into a 4-year college from the National Student Clearinghouse, a nonprofit organization providing enrollment and degree-verification services to colleges and universities. These data were obtained on 10/30/2019 corresponding to about 1.5–2.5 years after the end of 12^th grade.

Perception of school climate.

In the 10^th and 11^th grade surveys, students were asked about several aspects of their school environment. These school climate measures are not comprehensive but chosen to represent a diversity of school climate domains. School order refers to the amount of confusion and chaos in the classroom³⁴ and was assessed using a scale based on the Confusion, Hubbub, and Order Scale developed by Matheny and colleagues.³⁵ We analyzed the measure as school order, the inverse of school chaos, so that higher scores indicated a more positive school climate (Cronbach alpha = 0.68, 9 items, range 1–4). School safety was assessed using the Chicago Consortium on School Research Student Perceptions of Safety Scale, a 4-item measure of self-reported safety in and around school (Cronbach alpha = 0.63 range, 0–3, with higher scores indicating greater perceived school safety).³⁶

Using a modified questionnaire from the annual survey of Chicago public schools³⁷ students reported perceptions of teacher-student relationships on a four point scale from strongly disagree to strongly agree. We simply summated the responses to three questions (“my teachers always try to be fair”, “when my teachers tell me to do something, I know he/she has a good reason”, “my teachers treat me with respect”) into a single variable representing perceived teacher respect for students (Cronbach alpha = 0.72, higher scores indicating greater perceived teacher respect for students). We combined three additional questions (“teachers at this school help students plan for college outside of class time,” “teachers expect most students in this school to go to college,” and “teachers in this school feel that it is a part of their job to prepare students to succeed in college”) into a second variable representing teacher support for college (Cronbach alpha = 0.72, higher scores indicating greater perceived support for college).

School disciplinary style was assessed according to student ratings of school support and structure as previously described.²² These two rating scales were categorized into tertiles, and then combined to create a single perceived school disciplinary style variable with five categories: authoritative (highest tertile for both support and structure), authoritarian (lowest tertile for support, highest tertile for structure), permissive (highest tertile for support, lowest tertile for structure), neglectful (lowest tertile in support and structure), and average (remainder of the sample).

Covariates.

At the beginning of 9^th grade, students reported information on their personal demographics: gender (male vs not), Latinx ethnicity (Latinx vs not), birthplace (USA born vs not), native language (English vs not), as well as parental characteristics: birthplace (one or more parents born in USA vs not), employment (one or more parents employed full-time vs not), level of education (one or more parents graduated from high school vs not) and their parent’s parenting style (normal, authoritative, authoritarian, indulgent, neglectful).^38,39 We use this measure of parenting style as a gross approximation for potential parental behavioral confounders, understanding that parenting effects may be moderated by child-, environment-, and culture-specific factors which make validation of this scale in diverse populations challenging.⁴⁰

Data Analysis

We conducted linear and logistic regression analyses to examine the relationship between each school climate variable and each adolescent health, behavioral, and academic outcome separately. For these analyses, the continuous school climate variables (school order, school safety, teacher respect and teacher support for college) were standardized so that a 1-point change in each scale equaled one standard deviation. All models were adjusted for student gender, Latinx ethnicity, USA birthplace, native English language, parental birthplace, parental employment, parental education and parenting style. In each model, we also controlled for the outcome measures at baseline (end of 8^th grade/beginning of 9^th grade). For models examining GPA and standardized test scores, we controlled for these outcomes from middle school. All models used generalized estimating equations with a random effect for school to adjust for clustering of outcomes at the school-level. The analyses were restricted to the sample of respondents who completed baseline, 10^th grade and 11^th grade surveys. Among this analytic sample, values were missing for 2.3% or less of the sample for any single measure. 5.1% of the sample were missing data for the 8^th grade standardized test scores, 11.7% were missing 11^th grade standardized test scores, 2.3% were missing transcript and self-reported GPA from middle school, and 0.3% were missing transcript and self-reported GPA from high school. Missing values were multiply imputed using 100 replicates so as to maximize the use of available data across a large number of variables. Sensitivity analyses using listwise deletion produced similar results. STATA 14.0 (College Station, TX) was used for all analyses.

RESULTS

The original RISE-Up sample was comprised of 1270 students at baseline (9^th grade), 91% (1159) of whom completed the 10^th grade survey. This study was limited to the 1114 students (88%) who completed the baseline through 11^th grade surveys. Table 1 summarizes student and parental demographic characteristics. Just under half of the sample were males (46%), 90% were Latinx, 87% were born in the USA, and 40% were native English speakers. One-quarter of students reported having at least one parent born in the USA, 89% had one or more parent working full-time, and 52% had one or more parents graduate from high school. Compared to those in the analytic sample, subjects who were lost to follow up before the 11^th grade survey were more likely to be male (55% vs 46%, p=0.04), white (18% v 10% p=0.002), native English speaker (49% vs. 40%, p=0.02), and have at least 1 parent born in the USA (36% vs 25%, p=0.003). Those who were lost to follow up were less likely to have at least 1 parent working full-time (81% vs 88%, p=0.02). There were no differences between the analytic sample and those lost to follow-up in parental education, birth in the USA, and parenting style.

TABLE 1.

Student and parental characteristics reported by participants in the RISE-Up Study, Los Angeles, CA (N = 1114)

	Percent of students
Student demographics
Male	46.3%
Latinx	90.3%
USA Born	87.3%
Native English speaker	39.7%
Parental characteristics
Born in the USA, one or more parents	25.1%
Full-time employed, one or more parents	88.0%
Graduated high school, one or more parents
No	43.5%
Yes	51.6%
Unsure	4.8%
Parenting style *
Normal	50.0%
Authoritative	20.2%
Authoritarian	9.4%
Indulgent	8.9%
Neglectful	11.5%

^*Parenting style scale developed by Lamborn et al ^38,39.

Prevalence of outcomes.

A minority of the sample reported engaging in risky behaviors (Table 2). At 11^th grade, 15% reported using alcohol and 11% reported using cannabis in the last 30 days. One-fifth (22%) reported engaging in one or more alcohol misuse behaviors in the past year, such as drinking alcohol at school, getting into trouble because of alcohol, or missing school because of alcohol use (mean score 1.02, SD 2.72). Sixteen percent reported engaging in similar cannabis misuse behaviors (mean score 0.67, SD 2.22). One-fifth (22%) of the sample also reported engaging in one or more delinquent behaviors in the last year such as stealing, graffiti, selling drugs or being in a gang. One in eight students (13%) reported either carrying a weapon in the last 30 days or being in a physical fight in the last 12 months. Approximately 9% of students reported engaging in high-risk sex. Nearly one in five (19%) reported being the victim of bullying and 15% reported bullying others in the last 12 months.

Table 2.

Health, behavioral, and academic outcomes of participants in the RISE-Up study, Los Angeles, CA (N = 1114)

Health and behavior outcomes
30-day alcohol use, %	15.3%
30-day cannabis use, %	11.4%
Alcohol misuse, mean (sd)	1.02 (2.72)
Cannabis misuse, mean (sd)	0.67 (2.22)
Delinquent behaviors, one or more, %	21.6%
Violence, %	12.8%
High risk sex, %	8.6%
Victim of bullying, %	18.8%
Bullied others, %	15.2%
Academic outcomes
Truancy, %	21.8%
Changed schools since 9th grade, %	23.0%
GPA, mean (sd)	2.83 (0.68)
Proficient in Math,*%	34.7%
Proficient in English,*%	70.6%
Matriculated in a 4-year College, %	43.1%

^*Proficient or above 11^th grade standard on California Assessment of Student Performance and Progress, a state-wide standardized exam.

Among respondents, 22% of students reported being truant. From the start of 9^th grade to the time of 11^th grade survey, 23% of students reported changing schools at least once. Mean high school GPA was 2.83 (SD 0.68), 35% and 71% of students were proficient in Math and English on 11^th grade standardized tests respectively, and 43% matriculated at a 4-year college after high school.

Associations between school climate and adolescent health and behaviors.

After controlling for baseline (9^th grade) variables including student covariates, parental covariates, and the relevant outcome measure, perceived elements of school climate at 10^th grade were associated with health and behavioral outcomes reported in 11^th grade (Table 3). Greater perceived school order was associated with lower odds of 30-day alcohol and cannabis use (OR 0.81 95%CI [0.72, 0.92] and OR 0.74 [0.59, 0.91], respectively), lower scores on both alcohol and cannabis misuse scales (β −0.22 [−0.44, 0] and β −0.23 [−0.39, −0.07], respectively), and lower odds of reporting a delinquent behavior (OR 0.79 [0.68, 0.93]), high risk sex (OR 0.65 [0.54, 0.78]), being a victim of bullying (OR 0.8 [0.71, 0.9]), and bullying others (0.79 [0.68, 0.91)). Greater perceived school safety was associated with lower odds of being a victim of bullying (OR 0.79 [0.70, 0.89]) as well as bullying others (OR 0.76 [0.67, 0.87]), but none of the other risky behaviors.

Table 3.

Associations Between School Climate (10^th grade) and Health Outcomes (11^th grade), Los Angeles, CA (N=1114).

	30-day Alcohol Use [OR, 95%CI]	30-day Cannabis Use [OR, 95%CI]	Alcohol Misuse [β, 95%CI]	Cannabis Misuse [β, 95%CI]	Delinquent Behaviors [OR, 95%CI]	Violence [OR, 95%CI]	High Risk Sex [OR, 95%CI]	Victim of Bullying [OR, 95%CI]	Bullied Others [OR, 95%CI]
School environment
Order*	0.81 [0.72, 0.92]	0.74 [0.59, 0.91]	−0.22 [−0.44, 0]	−0.23 [−0.39, −0.07]	0.79 [0.68, 0.93]	0.83 [0.68, 1.01]	0.65 [0.54, 0.78]	0.80 [0.71, 0.90]	0.79 [0.68, 0.91]
Safety**	0.92 [0.77, 1.10]	0.91 [0.74, 1.12]	−0.17 [−0.39, 0.05]	−0.14 [−0.30, 0.02]	0.92 [0.8, 1.06]	0.95 [0.79, 1.14]	0.98 [0.81, 1.20]	0.79 [0.70, 0.89]	0.76 [0.67, 0.87]
Teacher relationship
Teacher respect for students	0.73 [0.62, 0.85]	0.76 [0.63, 0.92]	−0.34 [−0.54, −0.15]	−0.23 [−0.35, −0.12]	0.75 [0.64, 0.88]	0.85 [0.73, 0.99]	0.80 [0.66, 0.97]	0.88 [0.73, 1.05]	0.82 [0.69, 0.97]
Teacher support for college	0.82 [0.70, 0.97]	0.90 [0.75, 1.08]	−0.24 [−0.43, −0.05]	−0.22 [−0.43, −0.02]	0.90 [0.78, 1.04]	0.88 [0.74, 1.06]	0.86 [0.68, 1.09]	0.97 [0.82, 1.14]	0.95 [0.81, 1.10]
Disciplinary Style ***
Average	reference	reference	reference	reference	reference	reference	reference	reference	reference
Authoritative	0.70 [0.44, 1.11]	0.65 [0.32, 1.33]	−0.33 [−0.66, 0.01]	−0.13 [−0.47, 0.21]	0.62 [0.39, 1.00]	0.99 [0.60, 1.63]	0.62 [0.36, 1.07]	0.62 [0.42, 0.93]	0.86 [0.57, 1.30]
Authoritarian	1.45 [0.80, 2.62]	3.26 [1.63, 6.52]	0.42 [−0.38, 1.22]	0.5 [−0.45, 1.45]	1.36 [0.71, 2.63]	1.22 [0.58, 2.55]	0.44 [0.11, 1.73]	0.84 [0.44, 1.60]	0.63 [0.23, 1.72]
Permissive	0.30 [0.08, 1.05]	1.15 [0.34, 3.87]	0.02 [−1.32, 1.36]	0.11 [−0.51, 0.73]	0.48 [0.11, 2.12]	1.34 [0.40, 4.46]	0.63 [0.13, 3.00]	1.32 [0.62, 2.83]	0.56 [0.18, 1.76]
Neglectful	1.30 [0.90, 1.88]	2.71 [1.79, 4.12]	0.43 [0.01, 0.85]	0.47 [0.16, 0.77]	1.99 [1.34, 2.97]	1.36 [0.83, 2.23]	1.75 [1.09, 2.81]	0.77 [0.56, 1.05]	1.02 [0.74, 1.42]

Abbreviations: OR – Odds Ratio, β – regression coefficient. Bold font indicates estimate significant to p<0.05. All models adjusted for: student gender, Latinx ethnicity, USA birthplace, native English language, parental birthplace, parental employment, parental education, parenting style, outcome measure at 9^th grade.

^*Wong et al.’s measure of school order with higher scores indicates a more positive school climate ³⁴.

^**Chicago Consortium on School Research Student Perceptions of Safety Scale, higher scores indicating greater perceived safety ³⁶.

^***School Disciplinary Style Score ²².

In regard to student-teacher relationships, teacher respect for students was protective for 30-day alcohol use (OR 0.73 [0.62, 0.85], 30-day cannabis use (OR 0.76 [0.63, 0.92]), alcohol misuse (β −0.34 [−0.54, −0.15]), cannabis misuse (β −0.23 [−0.35, −0.12]), delinquent behaviors (OR 0.75 [0.64, 0.88]), violence (OR 0.85 [0.73, 0.99]), high risk sex (OR 0.80 [0.66, 0.97]), and bullying others (OR 0.82 [0.69, 0.97]). Perceived teacher support for college was protective for 30-day alcohol use (OR 0.82 [0.70, 0.97]) and alcohol and cannabis misuse (β −0.24 [−0.43, −0.05] and β −0.22 [−0.43, −0.02], respectively).

Associations with school disciplinary style used the average category as reference. The authoritative style (high support and structure) was protective against reporting delinquent behaviors (OR 0.62 [0.39, 1.00]) and being a victim of bullying (OR 0.62 [0.42, 0.93]). In contrast, the authoritarian style (low support, high structure) was a strong risk factor for 30-day cannabis use (OR 3.26 [1.63, 6.52]). The permissive disciplinary style was not significantly associated with any health or behavioral outcomes but was positively associated with some academic outcomes (see below). Finally, the neglectful style was a risk factor for 30-day cannabis use (OR 2.71 [1.79, 4.12]), alcohol misuse (β 0.43 [0.01, 0.85]), cannabis misuse (β 0.47 [0.16, 0.77]), delinquent behaviors (OR 1.99 [1.34, 2.97]), and high risk sex (OR 1.75 [1.09, 2.81]).

Association between school climate and adolescent academic outcomes.

Perceived school climate variables were also associated with academic outcomes (Table 4). Greater school order was associated with lower odds of truancy (OR 0.72 [0.61, 0.86]) and greater odds of matriculating in a 4-year college (1.14 [1.00, 1.31]). Greater perceived school safety was associated with lower odds of truancy (OR 0.83 [0.72, 0.96]). Teacher respect for students was associated with lower odds of truancy (OR 0.84 [0.72, 0.99]), as well as slightly higher GPA (β 0.03 [0, 0.06]) and greater odds of Math and English proficiency (OR 1.16 [1.02, 1.31], OR 1.19 [1.04, 1.37], respectively). Perceived school disciplinary style was also associated with academic outcomes. Authoritative disciplinary style was associated with lower truancy (OR 0.67 [0.46, 0.97]). Permissive style was associated with lower odds of truancy (OR 0.43 [0.23, 0.82]), higher GPA (β 0.24 [0.11, 0.36]), higher odds of English proficiency (OR 2.86 [1.12, 7.33]), and greater odds of matriculation in a 4-year college (OR 3.31 [1.38, 7.96]). Neglectful disciplinary style was associated with increased truancy (OR 1.75 [1.23, 2.49]), increased odds of changing schools (OR 1.29 [1.04, 1.61]) and lower GPA (β −0.08 [−0.15, −0.01]).

Table 4.

Associations Between School Climate (10^th grade) and Academic Outcomes (11^th grade), Los Angeles, CA (N=1114).

	Truancy [OR, 95%CI]	School change [OR, 95%CI]	GPA [β, 95%CI]	Math proficient [OR, 95%CI]	English proficient [OR, 95%CI]	4-year College matriculation [OR, 95%CI]
School environment
Order*	0.72 [0.61, 0.86]	1.02 [0.93, 1.13]	0.02 [−0.01, 0.06]	1.05 [0.92, 1.20]	1.06 [0.93, 1.20]	1.14 [1.00, 1.31]
Safety**	0.83 [0.72, 0.96]	0.96 [0.87, 1.07]	0.01 [−0.01, 0.04]	1.10 [0.96, 1.27]	1.04 [0.93, 1.15]	1.04 [0.95, 1.14]
Teacher relationship
Teacher respect for students	0.84 [0.72, 0.99]	0.98 [0.90, 1.06]	0.03 [0, 0.06]	1.16 [1.02, 1.31]	1.19 [1.04, 1.37]	1.06 [0.93, 1.21]
Teacher support for college	0.88 [0.75, 1.02]	0.94 [0.85, 1.03]	0 [−0.04, 0.04]	1.00 [0.84, 1.18]	1.13 [0.96, 1.32]	1.04 [0.93, 1.15]
Disciplinary Style ***
Average	Reference	reference	reference	Reference	reference	reference
Authoritative	0.67 [0.46, 0.97]	1.07 [0.78, 1.46]	0.02 [−0.05, 0.10]	0.95 [0.65, 1.40]	1.37 [0.82, 2.30]	1.19 [0.79, 1.80]
Authoritarian	0.82 [0.44, 1.54]	1.17 [0.82, 1.68]	−0.04 [−0.15, 0.08]	0.58 [0.29, 1.15]	0.95 [0.53, 1.70]	0.87 [0.45, 1.68]
Permissive	0.43 [0.23, 0.82]	0.80 [0.50, 1.29]	0.24 [0.11, 0.36]	1.00 [0.47, 2.14]	2.86 [1.12, 7.33]	3.31 [1.38, 7.96]
Neglectful	1.75 [1.23, 2.49]	1.29 [1.04, 1.61]	−0.08 [−0.15, −0.01]	0.76 [0.53, 1.10]	1.04 [0.74, 1.47]	1.03 [0.79, 1.35]

Abbreviations: OR – Odds Ratio, β – regression coefficient. Bold font indicates estimate significant to p<0.05. All models adjusted for: student gender, Latinx ethnicity, USA birthplace, native English language, parental birthplace, parental employment, parental education, parenting style, outcome measure at 9^th grade.

^*Wong et al.’s measure of school order with higher scores indicates a more positive school climate ³⁴.

^**Chicago Consortium on School Research Student Perceptions of Safety Scale, higher scores indicating greater perceived safety ³⁶.

^***School Disciplinary Style Score ²².

DISCUSSION

In the process of displacing millions of adolescents from school settings across the nation and the world, SARS-CoV-2 has reminded parents and policymakers alike of the irreplaceable role schools have in adolescent growth and health. It has also reinvigorated interest in the importance of the social climate that each school cultivates. These findings add longitudinal evidence that student-reported metrics of school climate – including an orderly environment, teacher-student relationships, and disciplinary style – are important upstream predictors of both health and academic outcomes in subsequent years. Departing from the current literature that tends to isolate one or two school climate variables and outcomes, this analysis took a comprehensive approach in analyzing the longitudinal relationship between multiple school climate variables and an array of both health and academic outcomes. This permits a more holistic analysis that better captures the effect of a multifaceted school climate not just on cognitive development but also on health and behavioral development.

The findings support some widely held beliefs about school climate and challenge others. While perceptions of school order and teacher respect for students were protective for nearly all risky behaviors, perceptions of safety were surprisingly only associated with less bullying. This supports some researchers’ assertion that, except in the case of bullying, school safety is only inconsistently protective for many outcomes.⁴¹ Similarly, perceptions of order and teacher respect for students was beneficial for a number of academic outcomes. Yet surprisingly, teacher support for college was not linked to any of our objective academic outcomes. Other authors have attributed such divergences to variations in measurement,¹³ however, as yet unidentified modifiers such as cultural norms cannot be excluded. Aligning well with the literature,^14,22,24 neglectful disciplinary style (low structure, low support) was associated with several serious and concerning behaviors (i.e., cannabis use and misuse, alcohol misuse, delinquent behaviors, and high-risk sex) as well as poor academic outcomes (truancy, changing schools, and GPA). In contrast, permissive disciplinary style (low structure, high support) was strongly predictive of English proficiency and college matriculation, providing additional evidence for the critical role of teacher support.^14,42–45 These findings are notably different than a previous study in which perceptions of school safety, respect, and authoritative disciplinary style (high structure and high support) were each consistently predictive of beneficial social-emotional health outcomes including depression, stress, self-efficacy, grit, and hopelessness,⁴⁶ which may hint at important differences between the determinants of these outcomes.

There are several potential mechanisms for these findings interpreted through the lens of Jessor’s theory of problem behavior, which framed problem behavior as the developmental result of three systems: personality, perceived environment, and behavior.⁵ School order may be protective for most problem behaviors by permitting students greater personal control and elevating expectations for academic achievement. Teacher respect for students validates prior literature on the importance of social support and quality of positive social relationships, but the more limited significance of teacher support for college suggests that this school climate measure may either be overly specific or otherwise does not reflect an important contextual factor for student behaviors and decision-making. Interestingly, the data revealed a far less robust association between an authoritative disciplinary style (high structure and support) and outcomes than anticipated. This could be due to several reasons including: 1) study design bias (e.g., insufficient lag between exposure and outcomes, sample size, imprecise measurement of disciplinary style) or 2) true lack of an association indicating that high risk behaviors are less subject to authoritative school disciplinary style than mental health outcomes. That estimates were qualitatively concordant with a protective effect but not statistically significant, suggests the former as a more likely explanation. Authoritarian disciplinary style (low support, high structure) was associated with increased cannabis use and the neglectful disciplinary style (low support and low structure) was associated with increased risky behavior and worsened academic outcomes. Meanwhile the permissive disciplinary style (high support, low structure) had the opposite effect for academic outcomes. Together these findings add additional evidence that teachers can be a key source of social support for adolescent development and that relationships with teachers influence both academic outcomes and engagement in risky behaviors.

Limitations

The study results should be considered within the following limitations. Measures of school climate were based upon student report. Thus it is possible that the student’s behaviors influenced their perceptions of school climate rather than vice versa. We tried to mitigate this reverse causal relationship by examining the relationship of school climate perceptions measured at 10^th grade with outcomes measured at 11^th grade, controlling for the outcome at baseline. Nevertheless, the field of school climate is moving toward multidimensional measures which include student self-report and studies have confirmed high internal validity of many student-reported variables when examined at the school level.²¹ Behavioral outcomes were also obtained via student self-report. Some outcomes were dichotomized to aid interpretability and because some outcomes are relatively rare, however, dichotomizing can introduce measurement bias, reduce statistical power, and underestimate variability between groups. The parental disciplinary style used as a covariate grossly approximated potential confounders such as parental monitoring, however, this measure may introduce measurement bias as these questions may not align well with commonly held conceptions of parenting in Latinx communities.^47,48 This study was observational in nature so we cannot draw conclusions about causality. The generalizability of this study is further hampered by the great heterogeneity of school climate definitions in the literature.^49,50 The baseline differences between the analytic sample and those lost to follow-up were relatively minor but could indicate the possibility of attrition bias. Lastly, the student population was mostly Latinx. While the sample was roughly similar to the population of students in the Los Angeles Unified School District, the results may not be generalizable to other student populations. Future studies may delve deeper into the Latinx experience and how school climate may better promote adolescent health and development.

Implications for School Health Policy, Practice, and Equity

Our findings support ongoing reform initiatives to measure student perspectives and intervene on school climate with the expectation of downstream benefits to student health and academic achievement, with reaffirmation in a predominantly Latinx student population. Such interventions have become more common including social-emotional learning interventions, school-wide positive behavioral interventions and supports, bullying prevention, community development programs, and interventions to improve teacher working conditions.⁵¹ These interventions are understandably cross-disciplinary. Therefore our findings also reaffirm school health practitioners’ use of the CDC’s Whole School, Whole Community, Whole Child framework and direct engagement with youth toward cross-disciplinary systems change.^52,53

CONCLUSIONS

As the isolation of SARS-CoV-2 has reminded us, the settings in which youth study and create relationships with peers and adults outside the home shape their long-term health and educational trajectories with consequences that reverberate throughout the life course. Adolescents are at the threshold of transitioning to adulthood from a developmental and social perspective and hence the potential consequences of risky health behaviors such as substance use or poor academic performance can be highly impactful on their long-term health trajectories. School climate is a measurable and changeable construct^17,50 and could be a valuable public health target, particularly relevant during adolescence when youth are especially sensitive to social influences.⁵⁴ To the extent that school climate drives health, such an approach may broaden the scope of schools from centers for academic learning to platforms for human thriving, transforming vicious cycles to virtuous ones.