Queen Essays

Annual Research Review: Adolescent mental health in the digital age: facts, fears, and future directions

Candice L. Odgers,1 and Michaeline R. Jensen2

1Department of Psychological Science, University of California, Irvine, Irvine, CA, USA; 2University of North Carolina at Greensboro, Greensboro, NC, USA

Adolescents are spending an increasing amount of their time online and connected to each other via digital technologies. Mobile device ownership and social media usage have reached unprecedented levels, and concerns have been raised that this constant connectivity is harming adolescents’ mental health. This review synthesized data from three sources: (a) narrative reviews and meta-analyses conducted between 2014 and 2019, (b) large-scale preregistered cohort studies and (c) intensive longitudinal and ecological momentary assessment studies, to summarize what is known about linkages between digital technology usage and adolescent mental health, with a specific focus on depression and anxiety. The review highlights that most research to date has been correlational, focused on adults versus adolescents, and has generated a mix of often conflicting small positive, negative and null associations. The most recent and rigorous large-scale preregistered studies report small associations between the amount of daily digital technology usage and adolescents’ well-being that do not offer a way of distinguishing cause from effect and, as estimated, are unlikely to be of clinical or practical significance. Implications for improving future research and for supporting adolescents’ mental health in the digital age are discussed. Keywords: Mental health; adolescence; depression; Internet usage; social media.

Introduction Adolescents have been early and enthusiastic adop- ters of digital technologies. Nearly all adolescents (95%) in the United States have at least one mobile device of their own, and 89% own a smartphone (Rideout & Robb, 2018). Similarly, a 2014 study of young people between the ages of 9 and 16 living across seven European countries reported that 80% of youth owned either a mobile or smartphone (Mascheroni & �Olafsson, 2014). Worldwide, rates of Internet and mobile phone access vary dramatically across high versus low-income countries; however, overall, one in three users of the Internet worldwide are under the age of 18 (Keeley & Little, 2017) and across both advanced and emerging economies younger (under the age of 35) versus older people (Taylor & Silver, 2018) are more likely to have access to the Internet, smartphones and social media.

Access to mobile devices begins early. Among our sample of young adolescents attending public schools in a large Southeastern state, close to half (48%) of 11-year-olds reported owning a mobile phone with a steep increase in ownership to 85% of adolescents by age 14 (Odgers, 2018). Young people are also spending an increasing amount of time online, with recent estimates in the United States placing older adolescents (aged 13–18) online view- ing of screen media for nonschool purposes at 6.67 hr per day, with their younger peers (aged 8– 12) spending, on average, 4.6 hr on screen media each day (Rideout, 2015).

Adolescents’ constant connectivity has led to con- cerns about how digital technologies may be influ- encing multiple aspects of adolescents’ lives, ranging from their levels of physical activity and their ability to interact with others in ‘real life’ to a more recent focus on whether too much time online is contributing to mental health problems among young people. Dis- cussions about the potential negative effects of smart- phones and social media are taking place alongside growing concerns regarding adolescents’ mental health. Recent increases in rates of depression, anx- iety and suicide, especially among girls (Mojtabai, Olfson, & Han, 2016) who are the heaviest users of new media, have led some to claim that smartphones and social media may be driving increases in suicidal behaviors, depression, and loneliness (Rosenstein & Sheehan, 2018; Twenge, Joiner, Rogers, & Martin, 2018). Alternative explanations for these increases have been provided and skepticism voiced regarding the claim that digital technology usage has led to increases in adolescent depression and related men- tal health problems (Daly, 2018; Livingstone, 2018); however, much of the conversation about contempo- rary adolescents’ mental health implicates digital technology usage as contributing to the worsening of mental health symptoms and well-being.

This paper reviews existing research regarding the association between digital technology use and men- tal health, with a specific emphasis on the potential influences of digital technology usage on adolescents’ experiences of depression and anxiety. The review integrates three main sets of information including recent: (a) meta-analyses summarizing the associa- tions between digital technology usage and mental health among youth, (b) findings from large-scale public access surveys and preregistered studies, and

Read the Commentary on this article at doi: 10.1111/jcpp.

13219

Conflict of interest statement: No conflicts declared.

© 2020 Association for Child and Adolescent Mental Health Published by John Wiley & Sons Ltd, 9600 Garsington Road, Oxford OX4 2DQ, UK and 350 Main St, Malden, MA 02148, USA

Journal of Child Psychology and Psychiatry 61:3 (2020), pp 336–348 doi:10.1111/jcpp.13190

PFI_12mmX178mm.pdf + eps format

(c) studies that have leveraged daily assessments of digital technology usage to understand both within- and between-person associations between adoles- cents’ digital technology usage and mental health. These three sources of information are triangulated to address the question of whether there are robust and practically significant associations between digital technology usage and adolescent mental health and, if so, for whom and under what circumstances digital technology usagemay amplify or reduce risk. Given a) the rapidly evolving nature of digital technologies usage among adolescents and b) the fact that a number of reviews and meta-analyses have recently been completed on this topic, a formal meta-analysis is not included. Instead, a synthesis of the main findings from recent reviews is provided alongside a review of key findings from large-scale datasets and daily and momentary studies. Finally, a set of future directions for research, policy and interventions are proposed, alongside a description of the steps that researchers, clinicians and policymakers will need to take to effectively support adolescents’ mental health in the digital age.

What do we currently know about the association between adolescent depression, mental health problems and digital technology usage? In the United States, there have been rapid and unprecedented increases in adolescent depressive symptoms (Keyes, Gary, O’Malley, Hamilton, & Schulenberg, 2019) and suicidal behavior (Burstein, Agostino, & Greenfield, 2019; Naghavi, 2019). Deaths by suicide have increased among every age group, but have been especially drastic among girls, where there has been a tripling of the suicide rate among 10- to 14-year-old girls from 1999 through 2017 (Hedegaard, Curtin, & Warner, 2018). It is important to note that the United States is an outlier with respect to these trends as rates of suicide worldwide continue to fall (Naghavi, 2019); nonethe- less, secular increases in emotional problems among young people have been observed, with increases in self-reported symptoms of anxiety and depression documented in countries such as Greece, Germany, Sweden, Iceland, Norway, China, and New Zealand from the 1980s onwards (Collishaw, 2015).

These increases have sounded alarms among par- ents, care providers and educators given the burden of disease and potentially devastating and deadly consequences for youth and their families. When plotted alongside increases in social media usage across this same time period, a powerful narrative has emerged that social media is driving changes in depressive symptoms and suicidal behaviors. Of course, the fact that two trend lines increase together does not mean that one phenomenon causes the other. Nonetheless, social media and digital technol- ogy usage has quickly emerged as a leading

candidate to explain the sudden jump in depression and related problems among girls.

Historically, adolescents who spent more time online were also more likely to report symptoms of depression and anxiety. But, these data come from a time when only a minority of young people were online, engaging in very different activities than what is seen today (in chat rooms talking with strangers versus online connecting with peers (George, Rus- sell, Piontak, & Odgers, 2018). Today, the majority of adolescents are online, typically connecting with offline friends and family (Reich, Subrahmanyam, & Espinoza, 2012). Moreover, as suggested by a recent synthesis of 37 studies, online communica- tion between young people is typically being used to support the ‘traditional’ tasks of offline friendships through arranging meet-ups, developing intimacy, and shows of affection (Yau & Reich, 2017).

Small associations still exist, as adolescents who report more depressive symptoms also tend to report spending more time online. However, as detailed below, a review of meta-analytic work and narrative reviews, recent large-scale public access and pre- registered studies, and daily and momentary assess- ments of digital technology usage and mental health, show that that associations between time online and internalizing symptoms are often (a) mixed between positive, negative, and null findings, (b) when pre- sent, are likely too small to translate into practically or clinically meaningful effects (explaining less than 0.5% of the variance in symptoms with poor adjust- ment for relevant confounding factors and estimates that are virtually always derived from correlation designs), and (c) are typically not distinguishable in terms of likely cause and effect. In addition, a recent systematic narrative review of 28 studies of online help-seeking behaviors indicated that many young people suffering from mental health problems are spending their time online searching for means of alleviating and better understanding their symptoms (Pretorius, Chambers, & Coyle, 2019).

Evidence Base 1. Meta-analytic studies and reviews

Six recent reviews summarizing the associations between digital technologies and adolescents’ mental health completed between the years of 2014 to 2019 are described below. The reviews were selected due to the fact that they targeted or included adolescent populations and included a focus specifically on the associations between amount of digital technology usage and mental health (see Table 1). The main results from each review are described briefly below, followed by a synthesis of findings and limitations across this work. Details on the individual studies included in the reviews are also provided in Table S1.

Three of the six reviews focused exclusively on adolescent or child populations. In one of the earliest and largest reviews, Best and colleagues (2014) conducted a systematic narrative review of 43

© 2020 Association for Child and Adolescent Mental Health

doi:10.1111/jcpp.13190 Adolescent mental health in the digital age 337

studies conducted between 2003 and 2013 focused on the association between online communication/ social media and well-being. Notably, their review included studies with wide ranging methodologies (e.g., other reviews, qualitative studies) and opera- tionalizations of digital technology use (e.g., technol- ogy-related problems and technology addiction alongside quantity of many different types of tech- nology use). Across studies, they observed contra- dictory evidence of mixed, null, and positive associations and emphasized the lack of robust causal research regarding the impact of social media on mental well-being among young people. With these limitations in mind, the authors then specu- lated on potential positive and negative impacts of social media for adolescents. Potential benefits of social media engagement that were identified included: increases in self-esteem, perceived social support and social capital, safe identity experimen- tation, and increased opportunities for self-disclo- sure. Specific potential harms of social media for well-being that were identified included: increased social isolation, depression, and cyberbullying.

In a 2017 systematic review, McCrae, Gettings and Pursell (2017) conducted a more focused review examining the association between social media use and depressive symptoms among children and adolescents (aged 5–18). Only 11 studies met eligi- bility for inclusion in the quantitative meta-analysis (focused on social networking sites and usage, restricted to English language publication, and con- ducted in general vs. clinical samples) resulting in a total N for the analysis of 12,646. The authors documented a small, but statistically significant, association between social media usage and depres- sive symptoms (r = .13, 95% CI: �.05 to 0.20), but noted the small number of studies, heavy reliance on cross-sectional designs (for 6 of the 11 studies), and difficulty in interpreting the clinical significance of

the findings due to the wide variation observed in sample sizes, methods, and results. The most recent systematic review in 2019 restricted the range of adolescents between 13 and 18 years of age and, again, only identified a small number of studies (N = 13) that met criteria for inclusion (Keles, McCrae, & Grealish, 2019). Eligibility for inclusion was determined based on age (13–18), measurement of social media usage as the exposure, measurement of depression, anxiety, or psychological distress by a validated instrument, and publication in peer reviewed journal, available in English. Of the 13 studies, 12 studies were cross-sectional. Again, the authors observed a general pattern of associations between social media usage and mental health problems, but noted that methodological limitations, the reliance on cross-sectional designs, and failure to include relevant mediators and moderators of associations, limited conclusions that could be drawn about the nature of this association. Impor- tantly, they highlighted the lack of longitudinal and experimental research in this area and, as such, emphasized that the relationship between social media and depression should be characterized in correlational versus causal terms.

The remaining three reviews included a mix of adults and adolescents in the sampling frame. Conclusions were consistent with those summarized for the adolescent populations above in that cross- sectional research designs, retrospective reporting of symptoms and digital technology usage, and small and mixed patterns of associations were the norm and often limiting factors in drawing reliable conclu- sions in this area (Baker & Algorta, 2016; Seabrook et al., 2016). For example, in a 2016 review exam- ining the association between frequency or time spent on SNS and depression, eight reported small positive associations, while twice as many found nonsignificant associations (Seabrook, Kern, &

Table 1 Recent Reviews on Youth Digital Technology Use and Mental Health

Study Design

Sample AgeMean (range) Sample size Mental health measure Tech use measure

Best et al. (2014) Systematic narrative review

Adolescents 43 studies Mental health and well- being

Online communication and social media

Baker et al. (2016) Systematic review of quant studies

Adolescents and Adults

30 studies Depression SNS

Seabrook et al. (2018) Systematic review

Adolescents and Adults

70 studies Depression and anxiety emphasis; Overall well- being

SNS

Huang (2010) Meta-analysis Adolescents and Adults

67 samples (61 studies) (N = 19,652)

Self-esteem, life satisfaction, loneliness and depression

SNS

Keles et al. (2017) Systematic review

13-18 13 papers Depression, anxiety and distress

Social media

McCrae et al. (2014) Systematic review

5 to 18 11 studies (N = 12,646)

Depression Social media

SNS, Social Networking Site.

© 2020 Association for Child and Adolescent Mental Health

338 Candice L. Odgers and Michaeline R. Jensen J Child Psychol Psychiatr 2020; 61(3): 336–48

Rickard, 2016). The authors concluded that the inconsistency across studies and lack of common themes or reproducible findings when varying mea- sures of SNS use were employed suggested that any association between social media and depression is likely to be conditional on a number of moderating factors and sensitive to variations in usage patterns, pre-existing vulnerabilities, and context. More recently, Huang (2017) performed a meta-analysis across 67 independent samples (61 studies), which included a mix of adolescents and young adults (N = 19,652). They reported that the mean correla- tion between time spent on social networking sites (SNS) and psychological well-being (comprised of self-esteem, life satisfaction, loneliness, and depres- sion) was r = �0.07 (95% CIs = �.04 to �.09), with associations for loneliness and depression that ranged from r’s = �0.08 and �.11, respectively. Main effects were not moderated by sample age or gender.

Table S1 provides additional details of the studies included in the six reviews which met inclusion criteria (adolescent sample; empirical analysis; available in English; measure of extent of digital technology use or engagement [i.e., studies which include only measures of technology-related prob- lems or ‘technology addiction’ excluded]; measures relevant to mental health [e.g., depression, anxiety, psychological well-being, loneliness, self-esteem]). The studies are summarized with respect to: the study design (cross-sectional, longitudinal, experi- mental), year of data collection, sample country, age of participants, measures of mental health and digital technology usage, and whether the study suggested that engagement with digital technology is harmful, helpful, or neither/unclear. Four main findings emerge from a review of the adolescent- focused studies detailed in this table. First, the majority of studies conducted to date are derived from cross-sectional surveys. Of the 29 studies included in Table S1, only 4 (14%) are longitudinal and only two studies included an experimental or quasi-experimental design. As a result, the ability to make causal inferences is extremely limited and does not allow for conclusions regarding whether increased time online or engagement with social media use causes changes in young people’s mental health.

The inconsistencies in the evidence reviewed and correlational nature of research to date raises ques- tions regarding how such a strong causal narrative has emerged regarding social media usage, time online, and adolescents’ mental health. An often- cited study when promoting the beneficial effects of reducing screen and social media time among ado- lescents comes from a study of Danish adults who were randomly assigned to take a break from Face- book. In this study, those assigned to take a Face- book break reported greater life satisfaction and more positive emotions compared to the control

condition who continued their Facebook use as usual (Tromholt, 2016). Results also suggested stronger effects among those whose use was already potentially problematic (as evidenced by heavy use, passive use, and envy of others on Facebook). However, the validity of this study and generalizabil- ity to adolescents is limited due to the fact that participants were unpaid adult volunteers recruited via Facebook ads, 86% of whom were women with an average age of 48 years, and all of whom were not blind to their condition prior to reporting on whether their mental health had improved after giving up Facebook. In contrast, experimental studies with college students have demonstrated that virtual communication can have positive impacts, with randomization to instant messaging and virtual communication leading to reductions in distress (Dolev-Cohen and Barak, 2013) and replenishment of self-esteem and perceived relational value after social exclusion (Gross, 2009). Additional experi- mental work with adolescent populations is sorely required, especially those that ensure participants are blind to study conditions and measure mental health using multiple informants.

Second, many studies have relied soley on screen time as the index of engagement with digital tech- nologies. Screen time is typically measured as the number of minutes or hours youth spend on a device or engaged in a particular online activity each day. The reliance on screen time metrics is a problem given that all screen time is not equal with respect to potential risks and benefits. Spending time on devices and screens is now a required part of many adolescents’ educational experiences and means of communication throughout the day with family and friends. Mobile devices have also become a primary means of accessing multiple modes of entertainment that have always appealed to adolescents, including streaming videos and movies, music, and gaming. In addition, screen time measures are typically gath- ered via retrospective self-reports from youth, which introduces recall bias, and are assessed alongside self-reported measures of mental health, which introduces common method or rater bias (Podsakoff, MacKenzie, Lee, & Podsakoff, 2003) into the research design and analysis. Finally, reducing a complex and multi-dimensional set of experiences into a single index of retrospective self-reports of the amount of time that youth spend in front of screens does not correspond well with objective measures of time spent online (correlations between objectively mea- sured and retrospectively reported screen time are estimated to be ~r = .20 (Ellis, 2019)). Across the 29 studies reviewed in Table 1, only two included objective or informant-rated measures of screen time or social media usage, and the majority did not go beyond relying on time-based summaries (e.g., 2 hr per day online) to characterize usage.

Third, most studies to date have relied on relatively small, nonrepresentative samples, which limits the

© 2020 Association for Child and Adolescent Mental Health

doi:10.1111/jcpp.13190 Adolescent mental health in the digital age 339

ability to both generalize back to the larger popula- tion of adolescents and to conduct adequately pow- ered interaction tests to identify which subpopulations may be most at risk, although there are exceptions to this trend (e.g., the Monitoring the Future Study and Millennium Cohort Study described in the next section). The vast majority of studies have been drawn from high-income and high-resource settings. Rates of mobile phone access and usage vary widely across low- to high-income settings, and potential impacts on adolescent health and well-being are likely to vary as well. This type of selective sampling and recruitment limits the gener- alizability of research findings and has resulted in conclusions being drawn almost exclusively from WEIRD (Western, Educated, Industrialized, Rich and Democratic) societies, an approach that is likely to heavily skew conclusions about potential impacts on adolescent mental health to a minority of adoles- cents worldwide (Henrich, Heine, & Norenzayan, 2010). The paucity of data from these settings impedes our understanding of potential impacts of digital technologies in middle- and low-income set- tings, where the vast majority of youth in the world are currently coming of age (World Health Organiza- tion, 2019).

Fourth, while a significant amount of time has been spent discussing issues related to negative impacts of digital technologies on adolescents, most empirical research on the effects of digital technolo- gies on well-being has focused on young children or adults (as evidenced by the small number of studies that met inclusion for the quantitative analyses above). More specifically, the early adolescent period has been neglected in prior research, despite the fact it is likely to be one of the most relevant times for understanding linkages between mental health and social media, as young people are making the transition biologically and socially to adolescence and, simultaneously, entering social media plat- forms and more complex digital environments. None of the studies reviewed above tested, or were pow- ered to test, whether associations differed by devel- opmental stage. Instead, when adolescence was considered separately, adolescents were treated as a homogenous group. Progress has been made in other areas with respect to mapping new media use on trajectories of adolescent brain development during this period (Crone & Konijn, 2018); however, what is currently needed is a developmentally cali- brated evaluation of the fit between the affordances and constraints of digital technologies and the core developmental tasks, competencies, and vulnerabil- ities that characterize the adolescent period more generally, and the transition to adolescence more specifically (Dahl, Allen, Wilbrecht, & Suleiman, 2018). Practically, there has been a blurring of the discussion in legal, clinical, and policy contexts between protections and screen time rules that are required for young children versus the approaches

required to help support and scaffold adolescents as they learn to navigate complex digital ecologies more independently.

To summarize, there has been widespread specu- lation that increases in depression and anxiety are being driven by changes in the way that adolescents interact with each other through social media and time online. The claims are that adolescents are increasingly losing out on opportunities for face-to- face interaction (Turkle, 2017), are likely to be harassed and victimized frequently online (Hamm et al., 2015), and are under constant assault by idealized and carefully curated images that may lead to upward social comparisons, envy, and, in turn, lower well-being and increasing rates of depression (Appel, Gerlach, & Crusius, 2016). However, a review of the existing research demonstrates inconsistent and primarily small associations between the quan- tity of digital technology usage and mental health, with no way to discern cause from effect. Additional research that is longitudinal, expands beyond WEIRD societies, integrates multiple indices of dig- ital technology usage and well-being, embeds exper- imental or quasi-experimental design features, and includes a sufficient, and representative number of young people spanning the entire adolescent period (ages 10–24) is needed. At present, narrative reviews and meta-analytic work do not support causal claims, or even strong and consistent correlational patterns, linking adolescents’ digital technology usage with mental health problems.

Evidence Base 2. Large-scale and multiple-cohort studies

Similar to findings from systematic reviews and meta-analyses, the most recent and rigorous large- scale and preregistered studies have not found strong support for a robust linkage between adoles- cents’ technology use and well-being. Using specifi- cation curve analysis across three national data sources of adolescents (N > 350,0000), two based in the United States and one in the UK, Orben and Przybylski (2019) demonstrated that choices related to the specification of variables capturing digital technology use, adolescent well-being, and con- founders can generate a myriad of effect sizes, with the most likely association being exceedingly small and explaining a small portion of the variance in well-being. More specifically, across their 3,221,225,472 analyses, technology use accounted for less than 1% (0.4%) of the variation in well-being. Again, the remaining small cross-sectional associa- tion between digital technology usage and well-being provided no credible way to disentangle cause from effect. In a related 2017 preregistered study of over 120,000 English adolescents, the authors found no robust associations between mental well-being and moderate use of digital technology (which character- izes use by most adolescents), with a measureable

© 2020 Association for Child and Adolescent Mental Health

340 Candice L. Odgers and Michaeline R. Jensen J Child Psychol Psychiatr 2020; 61(3): 336–48

‘albeit small’ negative associations (less than 1% of the variation explained) for those with high levels of engagement (Przybylski & Weinstein, 2017).

In a recent re-analysis of the Monitoring the Future Study (notably the same study and data that was used to signal initial alarms regarding the connection between social media/digital technology usage and depression; (Twenge et al., 2018)), daily social media use was not found to be a moderately strong or consistent risk factor for adolescents’ depressive symptoms (Kreski et al., submitted). The study analyzed data from 8th and 10th grade students, across 2009 to 2017, to assess the rela- tionship between self-reported daily social media use and depressive symptoms. The most consistent associations observed, after adjusting for confound- ing and stratifying by depression propensity, indi- cated that girls (but not boys) who had the lowest

propensity for depression had slightly increased risk for depressive symptoms with daily social media use exposure. Interestingly, as daily social media use has increased among adolescents in the United States, the associations between social media use and depressive symptoms across 2009 to 2017 have decreased in magnitude. Thus, while social media usage and depression have been both increasing over the last decade in the United States, the linkage between the two is mostly nonexistent, and when associations are detected, evidence indicates that they have become weaker over time. Across these large-scale cohort studies, the authors conclude that, as currently measured, social media usage is unlikely to be a meaningful contributor to increased depressive symptoms among youth in the United States and United Kingdom.

Evidence Base 3. Daily diary and ecological momentary assessment studies

Studies that have followed adolescents intensively using diary studies or Ecological Momentary Assess- ment (EMA) are also converging on a similar set of findings as those reviewed above, with small associ- ations that vary in direction between positive, neg- ative and null. Diary and EMA research designs allow for ‘in the moment’ data capture as young people report on their lived and recent experience and, more generally, enhance recall and produce more reliable and complete data on daily experiences (Shiffman, Stone, & Hufford, 2008). More specially, these methods have been shown to reduce the recall bias that is inherent in retrospective self-reports of experiences (which as detailed above is quite poor for estimates of time spent using technology; Ellis, 2019) and facilitate more accurate assessments of time allocation andmental health symptoms over the course of the day. Obtaining high density observa- tions of both digital technology usage and mental health also allows for an examination of within- person linkages between these experiences over time

while holding all stable all factors that are fixed within the individual and/or across time.

In our most recent EMA study (Jensen, George, Russell, & Odgers, 2019), adolescents were tracked on their smartphones to test whether more time spent using digital technology was linked to worse mental health outcomes. The study surveyed a pop- ulation representative sample of over 2100 youth, aged 11–15, followed by a 14-day ecological momen- tary assessment (EMA) via mobile phones with a representative sub-sample of approximately 400 youth in 2016–2017. The EMA portion of the study yielded 13,017 total observations over 5,270 study days and results demonstrated that adolescents’ baseline technology usage did not predict later men- tal health symptoms. Reports of mental health symptoms were gathered from the adolescents three times a day, and they also reported on their daily technology usage each night. There was no evidence that adolescents’ reported mental health was worse on days when they reported spending more versus less time on technology. When associations were observed, they were small and in the opposite direc- tion that would be expected given recent concerns about digital technology damaging adolescents’ men- tal health. For instance, teens who reported sending more text messages over the study period reported feeling better (less depressed) than teens who were texted less frequently. These findings are consistent with our prior research with adolescents deemed at risk for substance use and externalizing problems, where more time spent online, texting, and a greater number of texts sent were associated with less same day anxiety, andmore texts sent were also associated with less same day depression, although small same day linkages with increased externalizing problems were also observed (George et al., 2018).

EMA studies among older populations have gen- erated mixed findings. For example, in a study of college students using experience sampling, no sig- nificant associations emerged between daily social networking site use and depression (Jelenchick, Eickhoff, & Moreno, 2013). In an EMA of adults, momentary supportive online interactions were associated with momentary positive effect, but were not related to momentary negative affect (Oh, Ozkaya, LaRose, 2014). In contrast, another experi- ence sampling study (Kross et al., 2013) showed that quantity of Facebook use was associated with worse affect at the next time point (a lagged effect), but not the inverse (affect did not relate to next time point Facebook use). This study concluded that this effect was not attributable to loneliness, nor was it mod- erated by other risk factors.

Finally, a related and recently reported preregis- tered study from the United Kingdom examined associations between adolescents’ digital technology usage and life satisfaction over time (Orben, Dienlin, & Przybylski, 2019) using repeated within-person assessments to disentangle between-person

© 2020 Association for Child and Adolescent Mental Health

doi:10.1111/jcpp.13190 Adolescent mental health in the digital age 341

associations from within-person effects. Data were drawn from a large UK Household Longitudinal study, Understanding Society, which included 12,672 10- to 15-year-olds. The authors applied specification curve analysis and reported that across models, results were inconsistent, tended by be conditional (more likely to be present among females) on gender, with results that varied widely depending on how the data were analyzed. Most reported associations were small (‘arguably trivial’ as charac- terized by the authors) and in cases where stringent statistical controls were used, associations did not differ significantly from zero in over half of the models that were fit to the data. The authors concluded that, across the population (between people) social media use was not a strong predictor of adolescents’ life satisfaction and, over time, asso- ciations were likely to be reciprocal, small at best, stronger for females and largely dependent on the analytic approach adopted when analyzing the data.

To summarize, a review of meta-analytic work, large-scale preregistered studies, and intensive daily and momentary assessments provides little evidence that engagement with digital media has substantial associations with adolescents’ mental health symp- toms at the population level. It is also worth noting that one of the primary studies that has been frequently cited as a source of panic related to a possible connection between social media and depression is the Monitoring the Future Study in the United States. This paper (Twenge et al., 2018) reported on a correlation that accounted for <1% of the variation in depressive symptoms; that is 99.666% of the variation in adolescent’s depressive symptoms was due to other factors, and the small correlation between digital technology usage and depression (0.4%) was cross-sectional and was esti- mated based on both self-reported depressive symp- toms and technology usage. Similar to the vast majority of other studies reviewed here, there was no way to sort out cause from effect in this study. While it is true that small effects can have clinically meaningful and important implications for public health, this requires that the effects are causally estimated and there is compelling evidence of directionality and impacts. To date, the study designs and analytic approaches in this field have not been sufficient to support causal claims nor do they warrant the widespread panic related to smart- phones, social media and adolescent mental health.

Over the last year, other research teams have analyzed these same data (Kreski et al., submitted; Orben & Przybylski, 2019) and reported similar small initial associations between social media use and depressive symptoms. However, there are two important differences in the recent reporting from these same data. First, there has been an acknowl- edgement that results are highly dependent on how the models are specified and that associations are greatly reduced once potential confounders and

alternative specifications are considered. Second, even when the other teams have reported on the same initial small associations (using the same data set), the translation of the results has been in stark contrast to the message conveyed by the initial reports. That is, the message communicated from the recent analyses based on these data has been that there is no evidence of practically meaningful linkages between social media and contemporary adolescents’ depressive symptoms. The fact that the same data and effect sizes are reported across studies, but that they are communicated in dramat- ically different ways to the public, practitioners, and importantly to adolescents themselves, raises a number of questions related to the responsible and reproducible reporting of findings with public health importance from large, public use databases. That is, the stark contrast in how the findings are com- municated highlight the need to exercise caution and ensure that policies, parenting practices and the allocation of public health resources are based on robust facts versus common fears regarding how digital technologies influence young people (Uhls, 2016).

Overcoming fears and forging future directions for adolescents in the digital age Given the lack of evidence for strong connections between the amount of time that adolescents spend on social media and related technologies and their mental health, the question becomes: why has digital technology so quickly and adamantly been identified as a cause of recent upticks in adolescent depression? Some have suggested that each gener- ation is able to easily find fault in the choices, time- use, and overall character of the next and that moral panic around new technologies is an expected and well established cycle that plays out as new tech- nologies are introduced (Uhls, 2016). Another possi- bility is that the instincts and parental/clinical intuitions among those connecting social media with depression and anxiety are correct and the scientific community has simply not caught up or kept pace with new technologies in ways that allow us to capture their true impact and measurable effects. While future research may identify clear or stronger linkages, at present the available evidence falls short of the standard of proof required to identify digital technology use as a putative environmental cause of adolescent mental health problems. The scientific and medical community would not accept two lines traveling together as sufficient evidence to determine the cause of childhood cancer—a disease which also takes thousands of young people’s lives each year— we should not accept this standard in linking ado- lescents’ increasing depression and suicide with increases in social media use. Understanding the factors driving increasing rates of depression and suicide among young people constitutes a critically

© 2020 Association for Child and Adolescent Mental Health

342 Candice L. Odgers and Michaeline R. Jensen J Child Psychol Psychiatr 2020; 61(3): 336–48

important health crisis. If social media and smart- phones play a casual role, even a small one, we need to be able to effectively respond. To ensure that the scientific community is able to keep pace with the rapid evolution of new digital technologies and their potential linkages to adolescent well-being, careful attention to the following four issues will be required:

1. Adolescents’ online risk often mirrors offline vulnerabilities. Future research is needed to understand why offline risk signals online prob- lems and to support young people who are struggling in both spheres.

Adolescents with a history of prior victimization are more likely to be bullied, victimized, and solicited online (Kowalski, Giumetti, Schroeder, & Lattanner, 2014). Similarly, adolescents struggling with offline mental health problems are more likely to seek out more negative online content and spend more time passively ‘lurking’ versus engaging with others in online spaces (Underwood & Ehrenreich, 2017). Offline resources also matter, as youth from low- income families tend to report more negative spil- lover of negative experiences on social media to offline conflict, fights, and trouble at school (Odgers, 2018), while youth from more supportive and well- resourced homes are more likely to receive more scaffolding from adults and have more positive experiences online (Mascheroni & �Olafsson, 2014). Consistent with a ‘rich-get-richer’ model regarding who benefits most from time online (Kraut et al., 2002), longitudinal research has shown that chil- dren with higher quality social relationships (e.g., better reported relationships with friends, care- givers, siblings, and teachers) were more likely to become more frequent users of online communica- tion as adolescents (email, chats, or messaging) and, in turn, have more cohesive offline and online friendships (Lee, 2009).

Moving forward, research that integrates mea- sures of underlying mental health risk using, for example, family history, childhood risk, genetic propensity, or related markers of future mental health are required to trace how pre-existing vulner- abilities for mental health problems influence pat- terns of online usage and engagement and test whether pre-existing mental health risks moderate impacts of digital technology usage on well-being. A leading explanation for linkages between depressive symptoms and online engagement is that adoles- cents at higher risk for depressive symptoms may selectively use social media more, or differently. For example, youth who report psychological distress around their online activities and describe their technology use as including distressing or problem- atic elements, are also more likely to report psycho- logical distress in their offline lives (Andreassen et al., 2016; Augner & Hacker, 2012; Morrison &

Gore, 2010). Rigorous tests of reverse causation are required given that digital technology’s more nega- tive sides often appear among subgroups of adoles- cents with existing offline vulnerabilities (George & Odgers, 2015). At present, the over reliance on cross- sectional and correlational data make it impossible to determine whether problematic technology usage leads to mental health problems, or whether those with existing vulnerabilities are simply more likely to use technology in unhealthy ways. When consider- ing youth with existing vulnerabilities for mental health problems, there is also a danger in assuming a one-size-fits all explanation for this very diverse subgroup of adolescents, and for the influence of digital technology over time and across contexts. In general, there is a need to move beyond estimating one parameter to describe associations between adolescents’ digital technology usage and mental health, and importantly, not to simply replicate this ecological fallacy error when thinking about the population of adolescents (estimated at 1 in 5) suffering from a mental health problems. Instead, the next generation of digital mental health research for youth needs to ask when, under what conditions, and for whom does engagement with digital technol- ogy create opportunities, amplify risk, or neither. Both theoretically and empirically driven approaches (e.g., specification curve analyses) are required to better understand this type of heterogeneity in linkages across time, development, contexts, and adolescents.

Scientifically, accounting for unmeasured con- founding is a critical step in being able to under- stand mechanisms and model the interplay between offline and online risk. Practically, understanding how online and offline contexts interact is required to develop effective strategies for parenting and policies in the digital age. If, for example, online problems are largely determined by offline vulnerabilities, then much of our existing knowledge of how to promote healthy development among young people should translate into what has been described by many as a foreign digital landscape. For example, adolescents who are more vulnerable to upward social compar- isons and especially sensitive to peer and social rejection in offline social settings may benefit from being more closely monitored and supported when engaging in online interactions. Similarly, promoting supportive parent–child relationships that encour- age child disclosure, versus the adoption of overly restrictive of coercive parental monitoring strategies, may be equally effective in learning about young people’s unmonitored activities in both offline and online contexts. Just as interventions to prevent bullying within school settings have proven effective for reducing cyberbullying (Williford et al., 2013), parenting, and support strategies developed for use in offline spaces may translate well into supporting adolescents formation of healthy online relation- ships, interactions, and experiences.

© 2020 Association for Child and Adolescent Mental Health

doi:10.1111/jcpp.13190 Adolescent mental health in the digital age 343

2. Screen time is no longer a useful construct, but it still dominates research and public discourse. Researchers, policymakers and parents need to move beyond a singular focus on screen time and change the conversation to more accurately reflect how adolescents interact with digital tech- nologies in their daily lives.

Most measures of digital technology usage relied on in the studies reviewed above are reduced to a single measure of time spent online, or more recently, to time spent on a particular platform or type of online behavior. However, the nature of online interactions is likely to be more relevant for under- standing any potential mental health effects than is a global measure of the number of minutes or hours a youth spends online. Associations between online technology usage and mental health vary depending on the type and features of online activities. For example, online social networking site use tends to be related to less internalizing, to the extent that it includes positive interactions, enhances social sup- port, and facilitates social connectedness, and tends to be associated with more internalizing in instances when it is excessive, reduces time spent in in-person interactions, and in which interactions are negative or involve social comparisons (Clark, Algoe, & Green, 2018; Seabrook et al., 2016). Indeed, more nuanced studies of online activities among adolescents sug- gest that it is not the frequency but the type of social media usage that is associated with their depressive symptoms (Nesi, Miller, & Prinstein, 2017). It is also the case the social networking sites and platforms are evolving rapidly, from profiles that were origi- nally static portraits of the owner to dynamic ‘toolk- its’ that allow for interconnected streams of influence, conversations, and a mix of corporate, private, and public representations and uses of information and data (Ellison & Vitak, 2015). Ado- lescents are also engaging with multiple social media platforms which can change rapidly over time, creating challenges for researchers trying to capture the complex nature of their interactions and experi- ences in the online world. One innovative approach for capturing adolescents’ online engagement, that is not dependent on platform, is the EARS (Effortless Assessment of Risk States) which captures multiple indices of a person’s social and affective behavior via their naturalistic use of a smartphone, including the integration of a custom keyboard that logs, with the adolescents’ permission, text that is entered across social media platforms and other applications (Lind, Byrne, Wicks, Smidt, & Allen, 2018). Additional investments in developing and testing these types of flexible tools for research and clinical use are required, including approaches that include code- sign and interactive testing with adolescents them- selves.

More generally, in order to effectively move beyond a reliance on screen time metrics, alternative and

less burdensome methods of assessing mental health via mobile technologies are required, includ- ing, for example, scraping social media data to identify mental health risk (De Choudhury, Gamon, Counts, & Horvitz, 2013), and passively, and with consent, passively extracting data on the environ- ment, movements and digital traces left by young people that may be most relevant to their mental health (Mohr, Zhang, & Schueller, 2017; Nelson & Allen, 2018).

3. Digital technologies provide new opportunities to support all, but especially vulnerable, adoles- cents

The fears around the potential negative impacts of new technologies on young people have consumed much of the attention of policymakers, parents, and the medical community. What has been discussed less frequently is how new technologies could be leveraged to foster social connection and engage adolescents in ways that support their mental health. An emerging body of research suggests that if provided under the right conditions, online sup- ports and information can provide valuable forms of both instrumental and social support. Young people report going online frequently to seek out health information (Kauer, Mangan, & Sanci, 2014) and, those with lower social and emotional well-being, are more likely to report going online to seek support and to feel better about themselves (Rideout & Fox, 2018). Social networking sites may be used by young people in the face of setbacks (Toma & Hancock, 2013) and many young people turn to social media for support and advice related to their mental health symptoms (Pretorius et al., 2019), with some research suggesting that adolescents with moderate to severe depressive symptoms may be more likely (29) than their peers to turn to social media for emotional support (Rideout & Fox, 2018).

Supportive peers and networks carry important protective effects for young people’s mental health, and there is increasing evidence that online commu- nication may be a critical way that peer-to-peer support and communication occurs among adoles- cents. As reviewed above, digital communication is often used to support adolescents’ peer relationships by creating opportunities for displays of affection, intimate disclosure, and offline activities (Yau & Reich, 2017). Many studies now report positive associations and substantial overlap between ado- lescents online and offline interactions and relation- ship quality. For example, adolescents with stronger offline networks often report more robust online networks and, although increased time online tends to displace offline time with parents, parent–child relationships do not appear to be negatively influ- enced by these changes (for a review see George & Odgers, 2015). Interestingly, early experimental studies showed that virtual communication may

© 2020 Association for Child and Adolescent Mental Health

344 Candice L. Odgers and Michaeline R. Jensen J Child Psychol Psychiatr 2020; 61(3): 336–48

help adolescents ‘bounce back’ following social rejection (Gross, 2009) and, as such, may serve as a tool for providing social support when youth are separated from parents or loved ones physically.

The promise of digital technologies is that clini- cians, parents and researchers can now connect with adolescents where they spend much of their time and reach young people who may otherwise never enter a clinic or research laboratory. Digital tools offer the promise of taking evidence-based interventions to scale, reducing disparities in access to effective treatments and supports, and removing barriers to treatment resources (Lind et al., 2018). Peer-to-peer training and supports (e.g., mental health first aid), online support and referral systems (e.g., seven Cups of Tea) and the translation of evidence-based therapies, such as cognitive behav- ioral therapy, into digital format and delivery sys- tems, has provided proof of principal that digital technologies can be used to connect to and support young people. However, measurable progress in the development of interventions that support youth in online spaces will required interdisciplinary teams that bring expertise is not only the adolescent mental health, but also include those with expertise in communications, computer science, educational and learning sciences, pediatrics, and cultural anthropology/youth culture.

Despite the promise of supporting youth via digital technologies, a number of challenges remain, including the foundational problem that digital platforms and tools have not been designed or tailored developmentally for adolescents (Odgers, 2019). Instead, most wellness and mental health apps have been targeted toward adults or made for adults to use with or for their young children. Digital technologies are likely to provide a number of affordances that could be used to maintain and strengthen offline relationships, but relatively few evidence-based intervention efforts currently exist. The challenge will be moving past the ‘screen time debates’ and toward a set of productive investments in making digital technologies work in ways that effectively support youth.

1. The rapid adoption of new digital technologies may amplify existing inequalities in adolescent mental health and well-being. Equitable and inclusive research, policies, and intervention efforts are required to reduce the ‘new’ digital divide.

Historically, the introduction of new technologies have tended to benefit those who are best positioned to take advantage of the affordances that they provide. There is emerging evidence of ‘rich-get- richer’ effects related to adolescents’ online opportu- nities and experiences. For example, in our popula- tion representative sample of US adolescents, youth growing up economically disadvantaged families

were equally likely to have access to mobile devices but were more likely than their more affluent peers to perceive negative spillover of online experiences to problems in their offline lives (e.g., fights, trouble at school) (Odgers, 2018). In studies across Europe, children from wealthier versus poorer homes are more likely to receive two or more forms of active mediation of Internet safety by their parents (Mascheroni & �Olafsson, 2014) and in the United States, adolescents (aged 13–18) from low-income families spend twice as much time passively con- suming media than their peers from high-income families (with incomes >100,000 per year), and on average, spend about three more hours per day on screens.

Traditionally, the ‘digital divide’ has referred to differential access to new technologies. That gap still exists, but in many countries, it is shrinking (OECD, 2016).Whatwemaybeseeingnowistheemergenceofa new kind of digital divide, where differences in online experiences are amplifying risks among already vul- nerable adolescents. Lower versus higher income youth are increasingly living in two separate physical worlds as neighborhood, school, and other forms of segregation increase in the United States and else- where (Putnam, 2016); the concern is that this segre- gation of access, opportunities, and experiences will replicate itself online. The introduction and broad reach of digital technologies offers the promise of reducing health and educational disparities, but the fear is that if adequate supports are not provided, or technologies are not tailored, inequalities will be further amplified. As young people come of age in an increasingly unequal and stratified world, it is essen- tial thatequitywithrespect toaccess,experiences,and opportunities in both online and offline spaces is afforded (George et al, in press).

Conclusions Digitally, there have been unprecedented and rapid changes in howadolescents spend their time, connect totheworld,andcommunicatewitheachother.Mobile device ownership and social media use have reached unprecedented levels among adolescents. Perhaps this is not surprising as digital devices, and the affordances that they provide, are especially strong attractors for young people given their heighted need for affiliation, social approval, andnovelty seeking. As adolescents spend an increasing amount of time interactingwithdigital technologies, there isanurgent need to both understand effects of this usage and leveragenew technologies inways that support versus harm theirmental health andwell-being.

Unfortunately, most of the attention given to adolescents’ digital technology usage and mental health has focused on negative effects and has been based on weak correlational data. Over the past decade the rapid uptake of social media has fueled fears that social media platforms are causing serious

© 2020 Association for Child and Adolescent Mental Health

doi:10.1111/jcpp.13190 Adolescent mental health in the digital age 345

mental health problems. These fears have been extended down to children and were initially pro- moted based on scant evidence in a statement issued by the American Academy of Pediatrics Council on Communications and Media warning of the dangers of ‘excessive Facebook’ use among children and adolescents (O’Keeffe & Clarke-Pearson, 2011) and have since been fueled by a number of public calls to action on to protect children and adolescents from social media (Rosenstein & Sheehan, 2018). Research since that time has been mostly correla- tional, tends to focus on adults versus adolescents and has generated a mix of small positive, negative, and null associations. Most recently, large-scale preregistered studies have reported a lack of sizable or practically meaningful associations between ado- lescents’ digital technology usage and well-being.

Digital technologies are here to stay, and have become pervasive in the lives and relationships of youngpeople.Practically, it is critical toknowwhether recent fears about adolescents’ digital technology usage are justified as professional organizations release guidelines for parents, educators, and institu- tions based on incomplete and often contradictory findings. Policies restricting adolescents’ access to newtechnologiesareadvocated,butmaybe ill advised ifnewtechnologiesarebeingusedasavaluablesource ofsocialsupportorarerequiredinorder tobuilddigital and interpersonal (digitallymediated) skills for econo- mies of the future.With respect tomental health,what ismostneeded is a focusonhowto reachyoungpeople when they are in crisis and when support is needed most.

A theme that has consistently emerged across this research area relates to the overlap between

offline and online risk. This finding challenges the assumption, and a common message to parents, that the digital landscape and its effects are too complex, fast moving, or nuanced to fully under- stand or for us to help young people effectively navigate. A more likely explanation is that many of the same principles that guide healthy development and inform effective parenting will apply when supporting youth in their online activities and experiences. If this is true, then the good news for parents and policy makers is that existing evidence- based interventions and strategies may look differ- ent but will still be effective in supporting youth in the digital age.

Supporting information Additional supporting information may be found online in the Supporting Information section at the end of the article:

Table S1. Individual study details.

Acknowledgements C.L.O. is supported by the Jacobs Foundation and the Canadian Institute for Advanced Research. The authors have declared that they have no competing or potential conflicts of interest.

Correspondence Candice L. Odgers, Department of Psychological Science, 4556 Social and Behavioral Sciences Gateway, University of California, Irvine, Irvine, CA 92697-7085, USA; Email: [email protected]

Key Points

� Adolescents are early and enthusiastic adopters of digital technologies and are increasingly spending their time connecting to the online world and to each other through their devices. This constant connectivity has led to concerns that time spent online may be negatively impacting adolescents’ mental health and well- being.

� We synthesized recent findings across meta-analytic studies and narrative reviews, large-scale and preregistered cohort studies, and intensive assessment studies tracking digital technology use and mental health across time.

� Most research to date has been correlational, cross-sectional, mixed in terms of the directionality, and have resulted in small associations which leave no way of separating cause from effect.

� We recommend that future research use experimental and quasi-experimental methods and focus on online experiences versus screen time as well as heterogeneity in effects across diverse populations of youth. Knowledge generated from this research should allow researchers and practitioners to leverage online tools to reduce offline disparities and support adolescents’ mental health as they come of age in an increasingly digital and connected world.

© 2020 Association for Child and Adolescent Mental Health

346 Candice L. Odgers and Michaeline R. Jensen J Child Psychol Psychiatr 2020; 61(3): 336–48

References Andreassen, C.S., Billieux, J., Griffiths, M.D., Kuss, D.J.,

Demetrovics, Z., Mazzoni, E., & Pallesen, S. (2016). The relationship between addictive use of social media and video games and symptoms of psychiatric disorders: A large-scale cross-sectional study. Psychology of Addictive Behaviors, 30, 252–262.

Appel, H., Gerlach, A.L., & Crusius, J. (2016). The interplay between Facebook use, social comparison, envy, and depression. Current Opinion in Psychology, 9, 44–49.

Augner, C., & Hacker, G.W. (2012). Associations between problematic mobile phone use and psychological parameters in young adults. International Journal of Public Health, 57, 437–441.

Baker, D.A., & Algorta, G.P. (2016). The relationship between online social networking and depression: A systematic review of quantitative studies. Cyberpsychology, Behavior, and Social Networking, 19, 638–648.

Best, P., Manktelow, R., & Taylor, B. (2014). Online commu- nication, social media and adolescent wellbeing: A system- atic narrative review. Children and Youth Services Review, 41, 27–36.

Burstein, B., Agostino, H., & Greenfield, B. (2019). Suicidal attempts and ideation among children and adolescents in US emergency departments, 2007–2015. JAMA Pediatrics, 173, 598–600.

Clark, J.L., Algoe, S.B., & Green, M.C. (2018). Social network sites and well-being: The role of social connection. Current Directions in Psychological Science, 27, 32–37.

Collishaw, S. (2015). Annual Research Review: Secular trends in child and adolescent mental health. Journal of Child Psychology and Psychiatry, 56, 370–393.

Crone, E.A., & Konijn, E.A. (2018). Media use and brain development during adolescence. Nature Communications, 9, 588.

De Choudhury, M., Gamon, M., Counts, S., & Horvitz, E. (2013). Predicting Depression via Social Media. Paper pre- sented at the ICWSM.

Dahl, R.E., Allen, N.B., Wilbrecht, L., & Suleiman, A.B. (2018). Importance of nvesting in adolescence from a developmental science perspective. Nature, 54, 441–450.

Daly, M. (2018). Social-media use may explain little of the recent rise in depressive symptoms among adolescent girls. Clinical Psychological Science, 6, 295–296.

Dolev-Cohen, M., & Barak, A. (2013). Adolescents’ use of Instant Messaging as a means of emotional relief. Computers in Human Behavior, 29, 58–63.

Ellis, D.A. (2019). Are smartphones really that bad? Improving the psychological measurement of technology-related behav- iors. Computers in Human Behavior, 97, 60–66.

Ellison, N.B., & Vitak, J. (2015). Social network site affor- dances and their relationship to social capital processes. In S. Sundar (Ed.), The handbook of the psychology of commu- nication technology (pp. 205–227). Hoboken, NJ: Wiley- Blackwell.

George, M.J., & Odgers, C.L. (2015). Seven fears and the science of how mobile technologies may be influencing adolescents in the digital age. Perspectives on Psychological Science, 10, 832–851.

George, M.J., Russell, M.A., Piontak, J.R., & Odgers, C.L. (2018). Concurrent and subsequent associations between daily digital technology use and high-risk adolescents’ mental health symptoms. Child Development, 89, 78–88.

Gross, E.F. (2009). Logging on, bouncing back: an experimen- tal investigation of online communication following social exclusion. Developmental Psychology, 45, 1787–1793.

Hamm, M.P., Newton, A.S., Chisholm, A., Shulhan, J., Milne, A., Sundar, P., . . . Hartling, L. (2015). Prevalence and effect of cyberbullying on children and young people: A scoping review of social media studies. JAMA Pediatrics, 169, 770– 777.

Hedegaard, H., Curtin, S.C., & Warner, M. (2018).Suicide mortality in the United States, 1999–2017: US Department of Health and Human Services, Centers for Disease Control and Prevention.

Henrich, J., Heine, S.J., & Norenzayan, A. (2010). Most people are not WEIRD. Nature, 466, 29.

Huang, C. (2017). Time spent on social network sites and psychological well-being: A meta-analysis. Cyberpsychology, Behavior, and Social Networking, 20, 346–354.

Jensen, M., George, M.J., Russell, M.A., & Odgers, C.L. (2019). Young adolescents’ digital technology use and mental health symptoms: Little evidence of longitudinal or daily linkages. Clinical Psychological Science, 7, 1416–1433.

Jelenchick, L.A., Eickhoff, J.C., & Moreno, M.A. (2013). “Facebook depression?” Social networking site use and depression in older adolescents. Journal of Adolescent Health, 52, 128–130.

Kauer, S.D., Mangan, C., & Sanci, L. (2014). Do online mental health services improve help-seeking for young people? A systematic review. Journal of MEDICAL INTERNET RESEARCH, 16, e66.

Keeley, B., & Little, C. (2017).The State of the Worlds Children 2017: Children in a Digital World: ERIC.

Keles, B., McCrae, N., & Grealish, A. (2019). A systematic review: the influence of social media on depression, anxiety and psychological distress in adolescents. International Journal of Adolescence and Youth, 1–15. https://doi.org/ 10.1080/02673843.2019.1590851

Keyes, K.M., Gary, D., O’Malley, P.M., Hamilton, A., & Schu- lenberg, J. (2019). Recent increases in depressive symptoms among US adolescents: Trends from 1991 to 2018. Social Psychiatry and Psychiatric Epidemiology, 54, 987–996.

Kowalski, R.M., Giumetti, G.W., Schroeder, A.N., & Lattanner, M.R. (2014). Bullying in the digital age: A critical review and meta-analysis of cyberbullying research among youth. Psy- chological Bulletin, 140, 1073–1137.

Kraut, R., Kiesler, S., Boneva, B., Cummings, J., Helgeson, V., & Crawford, A. (2002). Internet paradox revisited. Journal of Social Issues, 58, 49–74.

Kreski, N., Platt, J., Rutherford, C., Olfson, M., Odgers, C.L., Schulenberg, J., & Keyes, K.M. (Submitted). Social media use and depressive symptoms among US adolescent stu- dents from 2009 through 2017: variations by levels of predicted depressive symptom risk.

Kross, E., Verduyn, P., Demiralp, E., Park, J., Lee, D.S., Lin, N., . . . Ybarra, O. (2013). Facebook use predicts declines in subjective well-being in young adults. PloS one, 8, e69841.

Lee, S.J. (2009). Online communication and adolescent social ties: Who benefits more from Internet use? Journal of Computer-Mediated Communication, 14, 509–531.

Lind, M.N., Byrne, M.L., Wicks, G., Smidt, A.M., & Allen, N.B. (2018). The Effortless Assessment of Risk States (EARS) tool: An interpersonal approach to mobile sensing. JMIR Mental Health, 5, e10334.

Livingstone, S. (2018). iGen: why today’s super-connected kids are growing up less rebellious, more tolerant, less happy– and completely unprepared for adulthood. In: Taylor & Francis.

Mascheroni, G., & �Olafsson, K. (2014). Net children go mobile: Risks and opportunities, 2nd ed. Milano, Italy: Educatt.

McCrae, N., Gettings, S., & Purssell, E. (2017). Social media and depressive symptoms in childhood and adolescence: A systematic review. Adolescent Research Review, 2, 315– 330.

Mohr, D.C., Zhang, M., & Schueller, S.M. (2017). Personal sensing: understanding mental health using ubiquitous sensors and machine learning. Annual Review of Clinical Psychology, 13, 23–47.

Mojtabai, R., Olfson, M., & Han, B. (2016). National trends in the prevalence and treatment of depression in adolescents and young adults. Pediatrics, 138, e20161878.

© 2020 Association for Child and Adolescent Mental Health

doi:10.1111/jcpp.13190 Adolescent mental health in the digital age 347

Morrison, C.M., & Gore, H. (2010). The relationship between excessive Internet use and depression: A questionnaire- based study of 1,319 young people and adults. Psy- chopathology, 43, 121–126.

Naghavi, M. (2019). Global, regional, and national burden of suicide mortality 1990 to 2016: systematic analysis for the Global Burden of Disease Study 2016. BMJ, 364, l94.

Nelson, B.W., & Allen, N.B. (2018). Extending the passive- sensing toolbox: Using smart-home technology in psycho- logical science. Perspectives on Psychological Science, 13, 718–733.

Nesi, J., Miller, A.B., & Prinstein, M.J. (2017). Adolescents’ depressive symptoms and subsequent technology-based interpersonal behaviors: A multi-wave study. Journal of Applied Developmental Psychology, 51, 12–19.

Odgers, C.L. (2018). Smartphones are bad for some teens, not all. Nature, 554, 432–434.

Odgers, C.L. (2019). Why digital tools have not yet revolution- ized adolescent health research and what we can do. Journal of Research on Adolescence, 29, 675–681.

OECD (2016). Are there differences in how advantaged and disadvantaged students use the Internet? Retrieved from Paris: /content/workingpaper/5jlv8zq6hw43-en.

Oh, H.J., Ozkaya, E., & LaRose, R. (2014). How does online social networking enhance life satisfaction? The relation- ships among online supportive interaction, affect, perceived social support, sense of community, and life satisfaction. Computers in Human Behavior, 30, 69–78.

O’Keeffe, G.S., & Clarke-Pearson, K. (2011). The impact of social media on children, adolescents, and families. Pedi- atrics, 127, 800–804.

Orben, A., Dienlin, T., & Przybylski, A.K. (2019). Social media’s enduring effect on adolescent life satisfaction. Proceedings of the National Academy of Sciences of the United States of America, 116, 10226–10228.

Orben, A., & Przybylski, A.K. (2019). The association between adolescent well-being and digital technology use. Nature Human Behaviour, 3, 173–182.

Podsakoff, P.M., MacKenzie, S.B., Lee, J.-Y., & Podsakoff, N.P. (2003). Common method biases in behavioral research: A critical review of the literature and recommended remedies. Journal of applied psychology, 88, 879–903.

Pretorius, C., Chambers, D., & Coyle, D. (2019). Young people’s online help-seeking and mental health difficulties: Systematic narrative review. Journal of Medical Internet Research, 21, e13873.

Przybylski, A.K., & Weinstein, N. (2017). A large-scale test of the Goldilocks Hypothesis: Quantifying the relations between digital-screen use and the mental well-being of adolescents. Psychological Science, 28, 204–215.

Putnam, R.D. (2016). Our kids: The American dream in crisis. New York: Simon and Schuster.

Reich, S.M., Subrahmanyam, K., & Espinoza, G. (2012). Friending, IMing, and hanging out face-to-face: overlap in adolescents’ online and offline social networks. Developmen- tal Psychology, 48, 356–368.

Rideout, V. (2015). The common sense census: Media use by tweens and teens. Los Angeles: Common Sense Media.

Rideout, V., & Fox, S. (2018). Digital health practices, social media use, and mental well-being among teens and young adults in the US. Hope Lab Report.

Rideout, V., & Robb, M.B. (2018). Social media, social life: Teens reveal their experiences. San Francisco, CA: Common Sense Media.

Rosenstein, B., & Sheehan, A. (2018). Open letter from Jana partners and Calstrs to Apple Inc. Available from https://thinkdifferentlyaboutkids.com/index.php?acc=1.

Seabrook, E.M., Kern, M.L., & Rickard, N.S. (2016). Social networking sites, depression, and anxiety: A systematic review. JMIR Mental Health, 3, e50.

Shiffman, S., Stone, A.A., & Hufford, M.R. (2008). Ecological momentary assessment. Annual Review of Clinical Psychol- ogy, 4, 1–32.

Taylor, K., & Silver, L. (2018).Smartphone ownership is growing rapidly around the world, but not always equally. Pew Research Center. Available from: https://www.pewre search.org/global/2019/02/05/smartphone-ownership-is- growing-rapidly-around-the-world-but-not-always-equally/ [last accessed 9 September 2019].

Toma, C.L., & Hancock, J.T. (2013). Self-affirmation underlies Facebook use. Personality and Social Psychology Bulletin, 39, 321–331.

Tromholt, M. (2016). The Facebook experiment: Quitting Facebook leads to higher levels of well-being. Cyberpsychol- ogy, Behavior, and Social Networking, 19, 661–666.

Turkle, S. (2017). Alone together: Why we expect more from technology and less from each other. London: Hachette UK.

Twenge, J.M., Joiner, T.E., Rogers, M.L., & Martin, G.N. (2018). Increases in depressive symptoms, suicide-related outcomes, and suicide rates among US adolescents after 2010 and links to increased new media screen time. Clinical Psychological Science, 6, 3–17.

Uhls, Y. (2016). Media moms & digital dads: A fact-not-fear approach to parenting in the digital age. New York, NY: Routledge.

Underwood, M.K., & Ehrenreich, S.E. (2017). The power and the pain of adolescents’ digital communication: Cyber victimization and the perils of lurking. American Psycholo- gist, 72, 144–158.

Williford, A., Elledge, L.C., Boulton, A.J., DePaolis, K.J., Little, T.D., & Salmivalli, C. (2013). Effects of the KiVa antibullying program on cyberbullying and cybervictimization frequency among Finnish youth. Journal of Clinical Child & Adolescent Psychology, 42, 820–833.

World Health Organization (2019). Coming of age: Adolescent health. Retrieved from https://www.who.int/health-topics/ adolescents/coming-of-age-adolescent-health.

Yau, J.C., & Reich, S.M. (2017). Are the qualities of adoles- cents’ offline friendships present in digital interactions? Adolescent Research Review, 3, 339–355.

Accepted for publication: 27 November 2019 First published online: 17 January 2020

© 2020 Association for Child and Adolescent Mental Health

348 Candice L. Odgers and Michaeline R. Jensen J Child Psychol Psychiatr 2020; 61(3): 336–48

Addictive Behaviors 137 (2023) 107524

Available online 17 October 2022 0306-4603/© 2022 The Author(s). Published by Elsevier Ltd. This is an open access article under the CC BY-NC-ND license (http://creativecommons.org/licenses/by- nc-nd/4.0/).

Collateral consequences of the school-to-prison pipeline: Adolescent substance use and developmental risk

Seth J. Prins a,*, Ruth T. Shefner a, Sandhya Kajeepeta a, Mark L. Hatzenbuehler b, Charles C. Branas a, Lisa R. Metsch a, Stephen T. Russell c

a Columbia University, United States b Harvard University, United States c University of Texas, Austin, United States

A R T I C L E I N F O

Keywords: School discipline School-to-prison pipeline Substance use Mental health Public health

A B S T R A C T

Objective: The adolescent health consequences of the school-to-prison pipeline remain underexplored. We test whether initiating components of the school-to-prison pipeline—suspensions, expulsions, and school poli- cing—are associated with higher school-average levels of student substance use, depressed feelings, and devel- opmental risk in the following year. Method: We linked 2003–2014 data from the California Healthy Kids Survey and the Civil Rights Data Collection from over 4,800 schools and 4,950,000 students. With lagged multi-level models, we estimated relationships between the school prevalence of total discipline, out-of-school discipline, and police-involved discipline, and standardized school-average levels of 6 substance use measures and 8 measures of developmental risk, respectively. Results: The prevalence of school discipline predicted subsequent school-mean substance use and developmental risk. A one-unit higher prevalence of total discipline predicted higher school levels (in standard deviations) of binge drinking alcohol (0.14, 95% CI: 0.11, 0.17), drinking alcohol (0.15, 95% CI: 0.12, 0.18), smoking tobacco (0.09, 95% CI: 0.06, 0.12), using cannabis (0.16, 95% CI: 0.14, 0.19), using other drugs (0.17, 95% CI: 0.14, 0.21), and violence/harassment (0.16, 95% CI: 0.12, 0.2). Total discipline predicted lower levels of reported community support (− 0.07, 95% CI: − 0.1, − 0.05), feeling safe in school (-0.12, 95% CI: − 0.16, − 0.09), and school support (− 0.16, 95% CI: − 0.19, − 0.12). Associations were greater in magnitude for more severe out-of- school discipline. Findings were inconsistent for police-involved discipline. Conclusion: Exclusionary school discipline and school policing—core elements of the school-to-prison pipe- line—are previously unidentified population predictors of adolescent substance use and developmental risk.

1. Introduction

Despite broad recognition of the public health crisis caused by mass criminalization and mass incarceration in the United States (US, Cloud, Parsons, & Delany-Brumsey, 2014; Wildeman, 2011), less is known about the public health implications of an auxiliary trend: the school-to- prison pipeline. The school-to-prison pipeline describes a carceral turn in public education, in which schools criminalize and punish the behavior of some students—especially Black and Latinx stu- dents—rather than provide quality education and support for underly- ing social/emotional or developmental needs (Mallett, 2016). Initiating components of the school-to-prison pipeline include exclusionary

discipline (suspension or expulsion) and police referrals/arrests in response to misbehavior. The pathways from school to the adult criminal legal system are empirically established (Bacher-Hicks, Billings, & Deming, 2019; Hemez, Brent, & Mowen, 2020), as are the wide-ranging individual and community health consequences of exposure to the criminal legal system (Cloud, Bassett, Graves, Fullilove, & Brinkley- Rubinstein, 2020; Hatzenbuehler, Keyes, Hamilton, Uddin, & Galea, 2015; Kajeepeta, Rutherford, Keyes, El-Sayed, & Prins, 2020, Kajeepeta et al. (2021)), especially for substance use (e.g., Binswanger, Blatchford, Mueller, & Stern, 2013; Møller et al., 2010). However, the population health consequences of the school-to-prison pipeline remain underex- plored. In the present study, we establish the first empirical evidence

* Corresponding author at: Columbia University Mailman School of Public Health, 722 W 168th Street, Room 521, New York, NY, United States. E-mail address: [email protected] (S.J. Prins).

Contents lists available at ScienceDirect

Addictive Behaviors

journal homepage: www.elsevier.com/locate/addictbeh

https://doi.org/10.1016/j.addbeh.2022.107524 Received 29 March 2022; Received in revised form 5 October 2022; Accepted 10 October 2022

Addictive Behaviors 137 (2023) 107524

2

that the prevalence of exclusionary school discipline is a potential determinant of school levels of adolescent substance use and other developmental risk factors.

In a recent study of over 4,800 schools comprising 4.9 million stu- dents in California, we found that schools with students who had higher average levels of student substance use and depressed feelings, less school and community support, and lower feelings of school safety had a subsequently higher prevalence of exclusionary school discipline and school-based police contact (Authors, 2021). We theorized that exposure to high levels of suspension, expulsion, and police contact in schools likely creates or exacerbates material and psychosocial conditions conducive to adolescent substance use, mental health problems, and developmental risk. In the present study, we explore this hypothesis and establish the first empirical evidence that the prevalence of exclusionary school discipline is not only a consequence, but a potential determinant of, school levels of adolescent substance use and other developmental risk factors.

1.1. Public education in the carceral state

Economic austerity in the US has had profound effects on public education. Systematic disinvestment in social services and infrastruc- ture, or “organized abandonment” (Gilmore, 2008; Harvey, 2018), exacerbated dramatic racial and class inequities in funding for public education and healthy child development (Chingos & Blagg, 2017; Giroux, 2013; Morgan & Amerikaner, 2018; Nguyen, Smith, & Granja, 2020; Urban Institute, 2017). As with social policy more broadly (Scull & Wacquant, 2009), local, state, and federal government managed the consequences of austerity in education with investments in social con- trol and punishment (Kupchik & Monahan, 2006), justified by racial- ized, largely manufactured crime panics (Welch, Price, & Yankey, 2002). As a result, schools have internalized carceral logics (Hirschfield, 2008; Kupchik & Monahan, 2006; Simon, 2007), as they increasingly function to manage, through criminalization, populations rendered surplus by the neoliberal transformation of the state (cf. Feeley & Simon, 1992; Wacquant, 2009).

This carceral turn in education increases adolescent criminalization in several ways, as school responses to student misbehavior have become more extreme and punitive (Hirschfield, 2008). “Zero toler- ance” policies, or rules that require the use of exclusionary discipline regardless of the severity or context of misbehavior (American Psycho- logical Association Zero Tolerance Task Force, 2008; Skiba & Knesting, 2001), were mandated in 75–90% of schools before the end of the 20th century (Mallett, 2016). Out-of-school suspension rates have more than doubled since the 1970s (Losen & Skiba, 2010), and students are more than twice as likely to be arrested in the month they are removed from school compared to months when they are not removed (Monahan, VanDerhei, Bechtold, & Cauffman, 2014). School securitization, including metal detectors, video surveillance, and police presence, has also increased dramatically (Mallett, 2016). Approximately 67% of high schoolers, 45% of middle schoolers, and 19% of elementary school students attended a school with at least one police officer present in the building in 2013–2014 (Lindsay, Lee, & Lloyd, 2018), and school-based arrests have increased 300–500% since the 1990s (Mallett, 2016).

Race and class disparities throughout the school-to-prison pipeline are profound. Black students, poor students, and students with disabil- ities are more likely to be disciplined than non-Black, wealthy, and non- disabled students (Fabelo et al., 2011; Freeman & Steidl, 2016). Black students are more than three times as likely to be suspended or expelled than white students, controlling for socioeconomic status and misbe- havior (Okonofua, Walton, & Eberhardt, 2016; Wallace, Goodkind, Wallace, and Bachman (2008)), and these racialized disparities likely contribute to the overrepresentation of Black people in the criminal legal system (Barnes & Motz, 2018; Rocque & Paternoster, 2011). At the school level, discipline and arrest rates are higher in districts with higher proportions of Black students and higher levels of disadvantage

(Freeman & Steidl, 2016; Mendez, Knoff, & Ferron, 2002).

1.2. Hypothesized adolescent substance use and mental health consequences of exclusionary school discipline

Adolescence is a critical developmental period for substance use initiation and psychiatric symptom incidence, and schools are the place that adolescents spend the majority of their time outside their homes (Ali et al., 2019; Dawson, Goldstein, Chou, Ruan, & Grant, 2008; Grant, 1998; Kessler et al., 2005, 2007; King & Chassin, 2007; Solmi et al., 2022). By 12th grade, 40% of US adolescents have used an illegal drug in the past year (primarily cannabis), and roughly 17% of 12th graders reported binge drinking in the past two weeks (Johnston et al., 2021). However, in 2019, of the 1.1 million adolescents who needed substance use treatment, only 6% received it in a specialty facility, and fewer than 1 in 10 adolescents with a substance use disorder (SUD) reported any past-year treatment (Substance Abuse and Mental Health Services Administration. (2019), 2019).

Within this context, more than a third of US adolescents who do access any mental health treatment access it only at school; they are disproportionately Black and low-income (Ali et al., 2019). Schools are thus crucial intervention targets for substance use and mental health treatment and prevention, and health equity therein. But instead, more than 10 million students in the US attend schools with police but no counselor, nurse, psychologist, or social worker (Whitaker et al., 2019).

Theory and evidence suggest that exposure to criminalization is developmentally harmful and has adverse effects on adolescent sub- stance use and mental health. For example, exposure to police stops increases trauma and anxiety symptoms among young men (Geller, Fagan, Tyler, & Link, 2014). Exposure to the criminal legal system can increase subsequent behavioral and substance use problems among adolescents (Huizinga, Henry, & Liberman, 2008). Students who attended schools with more severe exclusionary discipline policies had higher levels of depressive symptoms than students in schools with less severe exclusionary discipline policies (Eyllon, Salhi, Griffith, & Lincoln, 2020). And students who were subsequently suspended or expelled had nearly 50% higher odds of subsequent drug use compared with students who were not suspended or expelled (Dong & Krohn, 2020).

Finally, relationships with key supportive adults, such as teachers, are documented protective factors against adolescent substance use (Suldo, Mihalas, Powell, & French, 2008). Removing adolescents from school environments through exclusionary discipline, therefore, has the potential to exacerbate substance use risk, increase feelings of alienation and disengagement, and decrease feelings of social cohesion and phys- ical safety in schools (Noltemeyer, Ward, & Mcloughlin, 2015).

1.3. Study hypotheses

Prior research provides some direct evidence for the individual- and institutional-level substance use and mental health consequences of exclusionary school discipline and the school-to-prison pipeline. In the present study, we add to the body of evidence about institutional re- lationships between adolescent criminalization and adolescent health. If school discipline and policing are conceptualized as institutional re- sponses to organized abandonment, attendant disinvestment in adoles- cent health and developmental needs, and a mechanism of structural racism and criminalization (Gilmore, 2008; Hirschfield, 2008; Kupchik & Monahan, 2006), then we would expect exclusionary school discipline and policing to reproduce or exacerbate conditions harmful to adoles- cent health in ways apparent at the school level, not just the individual level.

To test this institutional-level hypothesis, we needed a unique data structure that contained school-level rather than student-level discipline data, as well as information on school-aggregate levels of student sub- stance use, mental health, and developmental risk factors. We also needed school-district covariates. To ensure detectable variation in the

S.J. Prins et al.

Addictive Behaviors 137 (2023) 107524

3

associations between these variables, we also required a sufficiently large sample of schools. We created this data structure by linking mul- tiple previously unconnected distinct sources, described below. Our hypothesis is that a higher prevalence of exclusionary school discipline and school-based police contact will be associated with subsequently higher school-aggregate levels of substance use, depressed feelings, and individual, peer, family, school, and community risk factors.

2. Method

2.1. Data

We connected 11 years of repeated cross-sectional California state data from three sources: adolescent health and wellbeing data from the statewide California Healthy Kids Survey (CHKS), school discipline prevalence data from the Civil Rights Data Collection (CRDC), and de- mographic data on California school districts from the American Com- munity Survey (ACS).

For outcome measures, we used 11 consecutive years (2003–2005 through 2013–2014) of the CHKS, in which approximately 85% of public-school districts in California participate annually (surveys are typically conducted in 2-year cycles). The survey is the largest of its kind in the US, and asks students about their behavior, experiences, and at- titudes related to their school, health, and well-being. The CHKS is administered anonymously to all 5th-, 7th-, 9th-, and 11th-grade stu- dents (Austin et al., 2011, 2013; Furlong, Ritchey, & O’Brennan, 2009; Hanson & Kim, 2007), and typically has a response rate greater than 70% (Austin, Hanson, & Polik, 2016). The sampling strategy and psy- chometric properties of CHKS measures have been described in-depth elsewhere (Austin & Duerr, 2004; Hanson & Austin, 2003; Hanson & Kim, 2007).

Exposure data come from the Civil Rights Data Collection (CRDC), a national survey of public schools in the US, which collects data on ed- ucation and civil rights issues, including school discipline (Office for Civil Rights, 2018). Since 2011, the CRDC has surveyed designated school officials and official records from all public schools (N = 97,172) in the US (response rate = 98–100%)(Office for Civil Rights, 2016). Prior to 2011, the CRDC used a stratified random, representative sample of all US public schools.

Covariate data come from the American Community Survey (ACS) Education Tabulation, a custom tabulation of ACS data for the National Center for Education Statistics (NCES) (National Center for Education Statistics. (n.d.). American Community Survey – Education Tabulation (ACS-ED). Education Demographic and Geographic Estimates. Retrieved September 15 (2020)). The data files, which contain publicly available demographic data for US school districts, are based on ACS five-year estimates and are updated annually.

We linked NCES school identifiers in the CRDC with CHKS unique County-District-School (CDS) codes using a crosswalk developed by NCES. School district demographic data from the ACS were also linked using NCES school identifiers.

2.2. Measures

2.2.1. Adolescent substance use and developmental risk factors For each measure described below, we calculated the mean or pro-

portion of student responses within each school, since schools are the primary unit of analysis. We then standardized the measures (i.e., calculated Z-scores) across all schools by year. Table S1 presents item composition and scoring for each measure.

Substance use, depressed feelings. In the CHKS, students reported how many times, respectively, in the past 30 days they had at least one drink of alcohol, binge drank (defined as four drinks for girls and five drinks for boys per drinking occasion), used cannabis, smoked a cigarette, and used a variety of other drugs (smokeless tobacco, inhalants, cocaine, methamphetamines, or amphetamines, ecstasy, LSD, or other

psychedelics, any other illegal drug). Alpha coefficients range from 0.90 to 0.98 (Hanson & Austin, 2003). Students also reported how many times in the past 30 days they felt depressed.

Community, home, peer, and school social support, and student resilience. Students reported, on a scale ranging from 0 (not at all true) to 3 (very true), whether they had support in their environments at home, in school, and in their community; support from and relationships with their friends; and their resilience, including items on self-efficacy, self- awareness, empathy, and problem-solving. We took the school mean of student responses to the items from each domain to create school-level summary measures, respectively, for community (8 items), home (8 items), peer (5 items), and school (9 items) social support, as well as student resilience (12 items). See online supplement Table S1 for the specific items in each of these domains. Alpha coefficients for these items ranged from 0.79 to 0.96 (Hanson & Austin, 2003).

Violence/harassment and school safety. Students reported how much they agreed that they felt safe in their schools and neighborhoods, scored from 1 (strongly disagree) to 5 (strongly agree). Students were also asked 18 questions about the number of times in the past 12 months they experienced violence and harassment in school, scored from 0 (zero times) to 3 (four or more times) (Russell, Sinclair, Poteat, & Koenig, 2012). See online supplement Table S1 for each of these items.

2.2.2. Exclusionary school discipline and police contact We constructed three measures of school discipline based on item

availability and our hypotheses about the health consequences of school discipline specifically, and the criminalization of students more broadly: total school discipline, out-of-school discipline only, and police-involved discipline.

The CRDC began collecting detailed school discipline data in 2009. Schools reported expulsions, out-of-school suspensions, in-school sus- pensions (when a student is removed from classes and activities but remains in the school building) and police-involved discipline (school- based arrests and police referrals). We divided the sum of these disci- pline measures by total enrollment to create a school-level total disci- pline prevalence proportion, covering the years 2009–2014. Before 2009, the CRDC collected data only on out-of-school-discipline, i.e., expulsions and out-of-school suspensions. To take complete advantage of all waves of available data, we created an out-of-school discipline prevalence proportion by dividing the sum of out-of-school-suspensions and expulsions by total enrollment. This measure covers the years 2003–2014. Finally, given the direct role that police play in student criminalization, we were interested in whether school policing alone predicted adolescent health outcomes. We created a school-level prev- alence proportion of police-involved discipline (school-based arrests and police referrals) divided by total enrollment, covering the years 2009–2014.

2.2.3. Potential confounders Since racialized group membership and class are strongly associated

with school discipline and the school-to-prison pipeline (Mendez et al., 2002; Rocque & Paternoster, 2011), as well as systematic disinvestment in child health and development (García, 2015; Johnson-Staub, 2017), we hypothesized several school- and school-district-level variables would confound the relationships among school discipline/policing and adolescent health and development outcomes. These included the school percentage of Black students; school district median age and median income; as well as the percentages of school district residents that were unemployed, had a high school degree, and identified as Black. Bivariate models testing the association between these confounders and the school discipline exposure variables and substance use/developmental risk outcome variables, respectively, supported their inclusion as controls.

2.3. Analysis

We fit multi-level linear models regressing each standardized health

S.J. Prins et al.

Addictive Behaviors 137 (2023) 107524

4

and well-being factor on each one-year-lagged school discipline mea- sure. Schools are the level one unit of analysis; schools were measured repeatedly over time (level 2). One-year lags were chosen to establish temporality. Models included random intercepts for school and controlled for year. In a second set of adjusted models, we added one- year-lagged confounding variables described above to each model. Models can be written as:

ytj = β00 + β10Xtj + u0j + etj

where ytj is the school-level value of a health and well-being outcome (e.g., binge drinking) for school j at time t, β00 is the average school mean outcome, Xtj is a vector of independent variables (the school discipline exposure as well as the confounding variables described above) for school j at time t, β10 is the vector of level-one fixed-effect parameters, u0j

represents individual school deviations around the average school mean (allowing for each school to have its own intercept, and accounting for nonindependence of observations of the same school over time), and etj

is the school j residual at time t (Bell & Jones, 2014; Diez Roux, 2002). The variance components are assumed to be normally distributed and independent (u0j N

( 0, σ2

u ) , etj N(0, σ2

e ) (Bell & Jones, 2014; Diez Roux, 2002). For each of the 42 models presented below, we examined quantile-quantile and residual plots and found that model assumptions of homoscedasticity and normally distributed residuals were generally met, with some evidence of skewing due to the presence of outliers. Including the lagged outcome variable as an independent variable did not appreciable alter our findings.

Given that the outcome variables are standardized, model co- efficients for the school discipline independent variables can be inter- preted as the change in standard deviations of the outcome associated with a 1-unit increase in the prevalence of school discipline. All analyses were conducted in R version 3.6.

The study was approved by the Institutional Review Boards of Columbia University.

3. Results

Table 1 displays the grand average characteristics of schools in the sample. The analytic sample contained data from 4,840 schools, repre- senting mean responses from 4,950,633 students. Students sampled were 30% white, 7.4% Black, 6.3% American Indian/Alaska Native, 10.6% Asian, 3.7% Native Hawaiian/Pacific Islander, and 43% Latinx. The mean school prevalence of total school discipline, out-of-school discipline, and police-involved discipline were 32%, 19%, and 2% respectively (Table 1). Figures S1-S4 present the means or proportions of all measures over time.

Fig. 1 presents results from adjusted multi-level linear models regressing the six-substance use/depressed feelings measures on the three one-year-lagged school discipline measures (18 models total). Tables S3 and S4 present unadjusted and adjusted (respectively) co- efficients, 95% CIs, and model fit statistics for each relationship. Adjusted estimates ranged from 0.09 (95% CI: 0.03, 0.14) for out-of- school discipline predicting subsequent school-mean-level of depressed feelings, to 0.39 (95% CI: 0.30, 0.48) for police-involved discipline predicting subsequent school-level cannabis use. In other words, a one-unit higher prevalence of police-involved discipline was associated with a 0.39 standard deviation higher school-mean level of cannabis use in the subsequent year.

After adjusting for school and school district confounders, higher total school discipline predicted subsequently higher school-mean levels of binge drinking alcohol, drinking alcohol, smoking tobacco, using cannabis, and using other drugs. Out-of-school discipline predicted subsequently higher binge drinking, depressed feelings, drinking alcohol, smoking tobacco, using cannabis, and using other drugs. Police- involved discipline predicted subsequently higher school-mean levels of drinking alcohol, using cannabis, and using other drugs.

Fig. 2 presents results from adjusted multi-level linear models regressing the eight developmental risk measures on the three one-year- lagged school discipline measures (24 models total). Tables S5 and S6 present unadjusted and adjusted (respectively) coefficients, 95% Cis, and model fit statistics for each relationship. Estimates ranged from − 0.24 (95% CI: − 0.39, − 0.08) for out-of-school discipline and subse- quent school-mean level of reported school support, to 0.28 (95% CI: 0.24, 0.33) for out-of-school discipline and subsequent school-level violence/harassment. In other words, a one-unit increase in the preva- lence of out-of-school discipline was associated with a − 0.24 standard deviation lower school-mean level of reported school support in the subsequent year.

In adjusted models, higher prevalence of total school discipline predicted lower school-mean levels of reported community support, feeling safe in school, school support, and higher school-mean levels of violence/harassment. Out-of-school discipline predicted lower subse- quent school-mean levels of reported community support, feeling safe in school, school support, and higher school-mean levels of violence/ harassment. Higher prevalence of police-involved discipline predicted lower subsequent school-mean levels of school support.

4. Discussion

We created an unprecedented longitudinal dataset linking statewide school discipline records and a statewide survey of student health and development from California, the state with the largest number of kindergarten-12th grade students in the US. We found that the preva- lence of exclusionary school discipline (suspension and expulsion) and school-based police contact—initiating components of the school-to-

Table 1 Grand average characteristics of California schools in the sample.

Variable Mean SD Min Max

Schools (N = 4,840) Students (N = 4,950,633) Demographics

Age 13.95 1.69 10 18 School proportion female 0.49 0.11 0 1 School proportion white 0.30 0.22 0 1 School proportion Black 0.07 0.09 0 1 School proportion Latinx 0.43 0.25 0 1

Substance use and depression outcomes (school proportion) Binge drink alcohol 0.42 0.50 0 5 Depressed past year 0.29 0.10 0.01 1 Drink alcohol 0.61 0.59 0 5 Smoke tobacco 0.42 0.64 0 5 Use cannabis 0.55 0.69 0 5 Use other drug 0.37 0.45 0 5

Developmental risk outcomes (raw scale score) Community support 1.99 0.33 0 3 Feel safe in neighborhood 4.01 1.06 1 5 Feel safe in school 3.50 0.43 1 5 Home support 1.70 0.70 0 3 Peer support 1.73 0.67 0 3 School support 1.60 0.23 0 3 Student resilience 1.69 0.70 0 3 Violence/harassment in school 0.43 0.15 0 2.5

School discipline exposures (prevalence) Out-of-school discipline (2003–2014) 0.19 0.38 0 10.53 Total school discipline (2009–2014) 0.32 0.61 0 15.09 Police-Involved Discipline (2009–2014)

0.02 0.11 0 5.73

School district sociodemographics Proportion high school grad 0.81 0.12 0.14 1 Proportion unemployed 0.06 0.02 0 0.3 Median household income ($) 63,129 21,461 18,750 23,8917 Proportion Black 0.06 0.06 0 0.49 Median age 36.4 6.2 20 66.3

Note: SOURCES: U.S. Department of Education, Office for Civil Rights, Civil Rights Data Collection, 2003–2014; California Healthy Kids Survey, 2003–2014; American Community Survey Education Tabulation, 2003–2014.

S.J. Prins et al.

Addictive Behaviors 137 (2023) 107524

5

prison pipeline—were associated with higher school-average levels of student substance use, depressed feelings, social support, and develop- mental risk factors in the following year.

Associations between total discipline and some outcomes were smaller than out-of-school discipline and confidence intervals some- times included the null. This is likely because the total discipline mea- sure included in-school suspensions, which are more common and less severe than out-of-school suspensions and expulsions. Findings for police-involved discipline were less consistent and warrant cautious interpretation; while the magnitudes of associations between police- involved discipline and many outcomes were often large (consistent with the hypothesized severity of this form of discipline), confidence intervals were wide and often included the null, indicating that esti- mates were imprecise. This is likely because police-involved discipline was rare relative to suspensions and expulsions. Findings for out-of- school discipline were more consistent and precise relative to total discipline and police-involved discipline, likely due in part because there were more waves of data available for this form of discipline.

While this institutional-level analysis does not permit inferences about individual-level pathways from suspension, expulsion, or school policing to negative adolescent health and developmental outcomes, our findings are nonetheless consistent with the view that as an educational paradigm, school discipline is not developmentally appropriate or responsive (and may be harmful) to adolescent health and develop- mental needs. At minimum, our findings suggest that schools that engage in more exclusionary discipline have students who, on average, subsequently engage in more substance use and have less community and school support.

Exposure to high levels of exclusionary discipline and policing in schools likely produces and reproduces material and psychosocial con- ditions that increase the risk of adolescent substance use and mental health problems, and erode social supports and healthy development (American Psychological Association Zero Tolerance Task Force, 2008). Meanwhile, heavy investments in school securitization and policing divert resources from school and community supports and services that might address the root causes of student disciplinary and health prob- lems. For example 90% of students in public schools experience staffing ratios for counselor, nurse, psychologist, and social worker positions that fail to meet professional standards (Whitaker et al., 2019). Our findings support efforts to reduce schools’ reliance on exclusionary

discipline and school-based policing in response to misbehavior, and instead invest in public health programs and personnel, including pri- mary prevention and behavioral health services.

Our findings are subject to several limitations. First, the CRDC did not require data reporting on school discipline for years prior to the 2013–2014 school year. This missing data may be informative, if schools’ failure to report was related to high rates of discipline, or may be random, if some schools chose not to report due to unfamiliarity with the questions or procedures. Second, our analyses are limited by the information provided in the CRDC, which is self-reported by designated officials who may be motivated to underreport school discipline, which may make estimates conservative. Regarding school discipline, the CRDC does not include information on reasons for reported school discipline, specific disciplinary infractions, or severity of behaviors that resulted in disciplinary measures. Third, approximately 15% of schools in California are not included in the CHKS, which may contribute to either random or biased missingness. Further, CHKS is a school-based sample and therefore does not include adolescents who had already been suspended, expelled, or incarcerated. This underrepresentation may have resulted in more conservative estimates of substance use, depressed feelings, and risk and resilience factors reported in this study, as these outcomes are likely elevated among students who have expe- rienced these forms of school discipline. Fourth, data from the CHKS are self-reported. Finally, because our data was aggregated to the school level, discipline prevalence ratios may reflect multiple suspensions, ex- pulsions, or police contacts for the same student. However, we do not have any reason to believe that this would systematically bias other students’ responses to the CHKS. School-level aggregation also limits our ability to make conclusions about individual-level behavior or associa- tions; at the same time, aggregation to the school level is appropriate for our research question, which seeks to examine institutional factors related to the school-to-prison pipeline. These limitations suggest that the relatively modest magnitudes of associations we found are likely conservative estimates. Nonetheless, even small effects can have a substantial impact when scaled over a large population of adolescents.

The purpose of the present study was to empirically establish school- level adolescent substance use, depressed feelings, and developmental risk factors as consequences of exclusionary school discipline and policing, known initiating components of the school-to-prison pipeline. As noted at the outset, there are strong theoretical reasons to believe,

Fig. 1. Results of 18 adjusted multi-level models regressing 6 standardized measures of substance use and depressed feelings on 3 lagged measures of school discipline.

S.J. Prins et al.

Addictive Behaviors 137 (2023) 107524

6

and initial evidence to suggest, that the relationships identified here are bidirectional, mutually compounding phenomena (Authors, 2021). Future research should identify the mechanisms for these bidirectional pathways and determine whether they remain consistent at the indi- vidual level. Moreover, it is likely the substance use and developmental risk outcomes mediate and modify each other with respect to school discipline. For example, it is possible that high rates of school discipline are worse for substance use outcomes in communities with less social support. Given the number of associations we tested, the distinct path- ways through which these more complex relationships are likely to operate, and the limitations of our data, fully exploring these pathways was beyond the scope of the present paper, but should be tested in future research.

Further, existing empirical evidence and gray literature has docu- mented the extent to which the school-to-prison-pipeline is an institu- tional mechanism of structural racism, given the profound racial disparities in the students it targets (Freeman & Steidl, 2016; Wallace, Jr. et al., 2008). In future research, we plan to determine the extent to which the associations identified in the present study are also racialized, and to what degree both the racialized criminalization of substance use and the consequences of community and school disinvestment help explain disparities in the school-to-prison pipeline.

This study found evidence that exclusionary school discipline and policing in schools—core elements of the school-to-prison pipeline—are a previously unidentified population predictors of adolescent substance use and developmental risk. As the social and health sciences continue to conduct research on the collateral consequences of mass criminalization and incarceration, they must also recognize schools as institutions that are implicated in the development and maintenance of carceral systems of control and the social production of poor health.

CRediT authorship contribution statement

Seth J. Prins: Conceptualization, Methodology, Formal analysis, Investigation, Resources, Writing – original draft, Supervision, Funding acquisition. Ruth T. Shefner: Writing – original draft, Writing – review & editing. Sandhya Kajeepeta: Writing – original draft, Writing – re- view & editing. Mark L. Hatzenbuehler: Conceptualization, Method- ology, Writing – review & editing. Charles C. Branas: Conceptualization, Methodology, Writing – review & editing. Lisa R. Metsch: Conceptualization, Methodology, Writing – review & editing. Stephen T. Russell: Conceptualization, Methodology, Writing – review & editing, Data curation.

Fig. 2. Results of 24 adjusted multi-level models regressing 8 measures of developmental risk on 3 lagged measures of school discipline.

S.J. Prins et al.

Addictive Behaviors 137 (2023) 107524

7

Declaration of Competing Interest

The authors declare that they have no known competing financial interests or personal relationships that could have appeared to influence the work reported in this paper.

Data availability

The authors do not have permission to share data.

Acknowledgements

This work was supported by the National Institute on Drug Abuse, National Institutes of Health [grant numbers K01 DA045955 and T32- DA037801].

Appendix A. Supplementary data

Supplementary data to this article can be found online at https://doi. org/10.1016/j.addbeh.2022.107524.

References

Ali, M. M., West, K., Teich, J. L., Lynch, S., Mutter, R., & Dubenitz, J. (2019). Utilization of Mental Health Services in Educational Setting by Adolescents in the United States. Journal of School Health, 89(5), 393–401. https://doi.org/10.1111/josh.12753

American Psychological Association Zero Tolerance Task Force. (2008). Are zero tolerance policies effective in the schools?: An evidentiary review and recommendations. American Psychologist, 63(9), 852–862. https://doi.org/10.1037/ 0003-066X.63.9.852

Austin, G., Bates, S., & Duerr, M. (2013). Guidebook To the Part II : Survey Content—Core Module. WestEd.

Austin, G., Bates, S., Duerr, M., Austin, G., & Bates, S. (2011). Guidebook to the California Health Kids Survey: Part III: Data use and Dissemination. http://file//localhost(null).

Austin, G., & Duerr, M. (2004). Guidebook for the California Healthy Kids Survey Part I. Administration 2004–2005 Edition. WestEd.

Austin, G., Hanson, T., Polik, J., & et al. (2016). School Climate, Substance Use, and Student Well-being in California, 2013-2015. Results of the Fifteenth Biennial Statewide Student Survey, Grades 7, 9, and 11. WestEd Health and Human Development Program.

Bacher-Hicks, A., Billings, S. B., & Deming, D. J. (2019). The School to Prison Pipeline: Long-Run Impacts of School Suspensions on Adult Crime (Working Paper No. 26257; Working Paper Series). National Bureau of Economic Research. https://doi.org/ 10.3386/w26257.

Barnes, J. C., & Motz, R. T. (2018). Reducing racial inequalities in adulthood arrest by reducing inequalities in school discipline: Evidence from the school-to-prison pipeline. Developmental Psychology, 54(12), 2328–2340. https://doi.org/10.1037/ dev0000613

Bell, A., & Jones, K. (2014). Explaining Fixed Effects: Random Effects Modeling of Time- Series Cross-Sectional and Panel Data. Political Science Research and Methods, 3(01), 133–153.

Binswanger, I. A., Blatchford, P. J., Mueller, S. R., & Stern, M. F. (2013). Mortality After Prison Release: Opioid Overdose and Other Causes of Death, Risk Factors, and Time Trends From 1999 to 2009. Annals of Internal Medicine, 159(9), 592–600. https://doi. org/10.7326/0003-4819-159-9-201311050-00005

Chingos, M., & Blagg, K. (2017). Do Poor Kids Get Their Fair Share of School Funding? (p. (p. 21).) Urban Institute.

Cloud, D. H., Bassett, M. T., Graves, J., Fullilove, R. E., & Brinkley-Rubinstein, L. (2020). Documenting and Addressing the Health Impacts of Carceral Systems. American Journal of Public Health, 110(S1), S5–S. https://doi.org/10.2105/AJPH.2019.305475

Cloud, D. H., Parsons, J., & Delany-Brumsey, A. (2014). Addressing mass incarceration: A clarion call for public health. American Journal of Public Health, 104(3), 389–391. https://doi.org/10.2105/AJPH.2013.301741

Dawson, D. A., Goldstein, R. B., Chou, S. P., Ruan, W. J., & Grant, B. F. (2008). Age at first drink and the first incidence of adult-onset DSM-IV alcohol use disorders. Alcoholism, Clinical and Experimental Research, 32(12), 2149–2160. https://doi.org/ 10.1111/j.1530-0277.2008.00806.x

Diez Roux, A. V. (2002). A glossary for multilevel analysis. Journal of Epidemiology & Community Health, 56(8), 588–594. https://doi.org/10.1136/jech.56.8.588

Dong, B., & Krohn, M. D. (2020). Sent Home versus Being Arrested: The Relative Influence of School and Police Intervention on Drug Use. Justice Quarterly: JQ, 37(6), 985–1011. https://doi.org/10.1080/07418825.2018.1561924

Eyllon, M., Salhi, C., Griffith, J. L., & Lincoln, A. K. (2020). Exclusionary School Discipline Policies and Mental Health in a National Sample of Adolescents without Histories of Suspension or Expulsion. Youth & Society, 0044118X20959591. https:// doi.org/10.1177/0044118X20959591

Fabelo, T., Thompson, M., Plotkin, M., Carmichael, D., Marchbanks, M., III, & Booth, E. (2011). Breaking Schools’ Rules: A Statewide Study of How School Discipline Relates to Students’ Success and Juvenile Justice Involvement. Council of State Governments

Justice Center. https://www.ojp.gov/ncjrs/virtual-library/abstracts/breaking-s chools-rules-statewide-study-how-school-discipline-0.

Feeley, M., & Simon, J. (1992). The new penology: Notes on the emerging strategy of corrections and its implications. Criminology, 30, 449–474. https://doi.org/10.1111/ j.1745-9125.1992.tb01112.x

Freeman, K. J., & Steidl, C. R. (2016). Distribution, Composition and Exclusion: How School Segregation Impacts Racist Disciplinary Patterns. Race and Social Problems, 8 (2), 171–185. https://doi.org/10.1007/s12552-016-9174-9

Furlong, M. J., Ritchey, K. M., & O’Brennan, L. M. (2009). Developing norms for the California Resilience Youth Development Module: Internal assets and school resources subscales. California School Psychologist, 14, 35–46. https://doi.org/ 10.1007/BF03340949

García, E. (2015). Inequalities at the Starting Gate: Cognitive and Noncognitive Skills Gaps between 2010–2011 Kindergarten Classmates (p. (p. 93).). Economic Policy Institute.

Geller, A., Fagan, J., Tyler, T., & Link, B. G. (2014). Aggressive policing and the mental health of young urban men. American Journal of Public Health, 104(12), 2321–2327. https://doi.org/10.2105/AJPH.2014.302046

Gilmore, R. W. (2008). Forgotten Places and the Seeds of Grassroots Planning. In C. R. Hale (Ed.), Engaging Contradictions: Theory, Politics, and Methods of Activist Scholarship. University of California Press. https://doi.org/10.1525/ 9780520916173.

Giroux, H. (2013). Can emocratic education survive in a neoliberal society? In C. Reitz (Ed.), Crisis and Commonwealth: Marcuse, Marx, McLaren (pp. 137–152). Lexington Books.

Grant, B. F. (1998). Age at smoking onset and its association with alcohol consumption and DSM-IV alcohol abuse and dependence: Results from the national longitudinal alcohol epidemiologic survey. Journal of Substance Abuse, 10(1), 59–73. https://doi. org/10.1016/S0899-3289(99)80141-2

Hanson, T. L., & Austin, G. (2003). Student health risks, resilience, and academic performance in California: Year 2 Report. Longitudinal Analysis. http://file//localh ost(null.

Hanson, T. L., & Kim, J. O. (2007). Measuring resilience and youth development: The psychometric properties of the Healthy Kids Survey. In Issues & Answers Report. Institute of Education Sciences, National Center for Education Evaluation and Regional Assistance, U.S. Department of Education.

Harvey, D. (2018). The limits to capital. Verso. Hatzenbuehler, M. L., Keyes, K., Hamilton, A., Uddin, M., & Galea, S. (2015). The

collateral damage of mass incarceration: Risk of psychiatric morbidity among nonincarcerated residents of high-incarceration neighborhoods. American Journal of Public Health, 105(1), 138–143. https://doi.org/10.2105/AJPH.2014.302184

Hemez, P., Brent, J. J., & Mowen, T. J. (2020). Exploring the School-to-Prison Pipeline: How School Suspensions Influence Incarceration During Young Adulthood. Youth Violence and Juvenile Justice, 18(3), 235–255. https://doi.org/10.1177/ 1541204019880945

Hirschfield, P. J. (2008). Preparing for prison?: The criminalization of school discipline in the USA. Theoretical Criminology, 12(1), 79–101. https://doi.org/10.1177/ 1362480607085795

Huizinga, D., Henry, K. L., & Liberman, A. M. (2008). The effect of arrest and justice system sanctions on subsequent behavior: Findings from longitudinal and other studies. In The Long View of Crime: A Synthesis of Longitudinal Research (Issue September, pp. 220–254).

Johnson-Staub, C. (2017). Addressing Racial Inequities in Child Care and Early Education Policy (p. (p. 24).). Center for Law and Social Policy.

Johnston, L. D., Miech, R. A., O’Malley, P. M., Bachman, J. G., Schulenberg, J. E., & Patrick, M. E. (2021). Monitoring the Future national survey results on drug use 1975–2020: Overview, key findings on adolescent drug use. Institute for Social Research, University of Michigan. http://www.monitoringthefuture.org/pubs/ monographs/mtf-overview2020.pdf.

Kajeepeta, S., Mauro, P. M., Keyes, K. M., El-Sayed, A. M., Rutherford, C. G., & Prins, S. J. (2021). Association between county jail incarceration and cause-specific county mortality in the USA, 1987–2017: A retrospective, longitudinal study. The Lancet Public Health, 6(4), e240–e248. https://doi.org/10.1016/S2468-2667(20)30283-8

Kajeepeta, S., Rutherford, C. G., Keyes, K. M., El-Sayed, A. M., & Prins, S. J. (2020). County Jail Incarceration Rates and County Mortality Rates in the United States, 1987–2016. American Journal of Public Health, 110(S1), S109–S115. https://doi.org/ 10.2105/AJPH.2019.305413

Kessler, R. C., Amminger, G. P., Aguilar-Gaxiola, S., Alonso, J., Lee, S., & Ustun, T. B. (2007). Age of onset of mental disorders: A review of recent literature. Current Opinion in Psychiatry, 20(4), 359–364. https://doi.org/10.1097/ YCO.0b013e32816ebc8c

Kessler, R. C., Berglund, P., Demler, O., Jin, R., Merikangas, K. R., & Walters, E. E. (2005). Lifetime prevalence and age-of-onset distributions of DSM-IV disorders in the National Comorbidity Survey Replication. Archives of General Psychiatry, 62(6), 593–602. https://doi.org/10.1001/archpsyc.62.6.593

King, K. M., & Chassin, L. (2007). A prospective study of the effects of age of initiation of alcohol and drug use on young adult substance dependence. Journal of Studies on Alcohol and Drugs, 68(2), 256–265. https://doi.org/10.15288/jsad.2007.68.256.

Kupchik, A., & Monahan, T. (2006). The New American School: Preparation for Post- Industrial Discipline. British Journal of Sociology of Education, 27(5), 617–631.

Lindsay, C. A., Lee, V., & Lloyd, T. (2018). June 21). Urban Institute: The prevalence of police officers in US schools. https://www.urban.org/urban-wire/prevalence-police- officers-us-schools.

Losen, D. J., & Skiba, R. J. (2010). Suspended Education: Urban Middle Schools in Crisis. Southern Poverty Law Center.

S.J. Prins et al.

Addictive Behaviors 137 (2023) 107524

8

Mallett, C. A. (2016). The School-to-Prison Pipeline: A Critical Review of the Punitive Paradigm Shift. Child and Adolescent Social Work Journal, 33(1), 15–24. https://doi. org/10.1007/s10560-015-0397-1

Mendez, L. M. R., Knoff, H. M., & Ferron, J. M. (2002). School demographic variables and out-of-school suspension rates: A quantitative and qualitative analysis of a large, ethnically diverse school district. Psychology in the Schools, 39(3), 259–277. https:// doi.org/10.1002/pits.10020

Møller, L. F., Matic, S., van den Bergh, B. J., Moloney, K., Hayton, P., & Gatherer, A. (2010). Acute drug-related mortality of people recently released from prisons. Public Health, 124(11), 637–639. https://doi.org/10.1016/j.puhe.2010.08.012

Monahan, K. C., VanDerhei, S., Bechtold, J., & Cauffman, E. (2014). From the school yard to the squad car: School discipline, truancy, and arrest. In None (Ed.), Journal of Youth and Adolescence (Vol. 43, Issue 7, pp. 1110–1122). Springer.

Morgan, I., & Amerikaner, A. (2018). Funding Gaps: An Analysis of School Funding Equity Across the U.S. and within Each State. The Education Trust.

National Center for Education Statistics. (n.d.). American Community Survey – Education Tabulation (ACS-ED). Education Demographic and Geographic Estimates. Retrieved September 15, 2020, from https://nces.ed.gov/programs/edge/Demographic/ACS.

Nguyen, U. S., Smith, S., & Granja, M. R. (2020). Young Children in Deep Poverty: Racial/ Ethnic Disparities and Child Well-Being Compared to Other Income Groups (p. 22). National Center for Children in Poverty, Bank Street Graduate School of Education.

Noltemeyer, A. L., Ward, R. M., & Mcloughlin, C. (2015). Relationship Between School Suspension and Student Outcomes: A Meta-Analysis. School Psychology Review, 44(2), 224–240. https://doi.org/10.17105/spr-14-0008.1.

Office for Civil Rights. (2016). 2013-14 Civil Rights Data Collection: A First Look (p. 13). U. S. Department of Education.

Office for Civil Rights. (2018). 2015-16 Civil Rights Data Collection: School Climate and Safety. U.S. Department of Education, Office for Civil Rights.

Okonofua, J. A., Walton, G. M., & Eberhardt, J. L. (2016). A vicious cycle: A social- psychological account of extreme racial disparities in school discipline. Perspectives on Psychological Science, 11(3), 381–398.

Rocque, M., & Paternoster, R. (2011). Understanding the antecedents of the “school-to- jail” link: The relationship between race and school discipline. The Journal of Criminal Law and Criminology, 101(2), 633–665.

Russell, S. T., Sinclair, K. O., Poteat, V. P., & Koenig, B. W. (2012). Adolescent Health and Harassment Based on Discriminatory Bias. American Journal of Public Health, 102(3), 493–495. https://doi.org/10.2105/AJPH.2011.300430

Scull, A., & Wacquant, L. (2009). Punishing the Poor: The Neoliberal Government of Social Insecurity. Archives Européennes de Sociologie: European Journal of Sociology, 50 (3), 499–503. https://doi.org/10.1017/S0003975609990312

Simon, J. (2007). Governing Through Crime: How the War on Crime Transformed American Democracy and Created a Culture of Fear. Oxford University Press.

Skiba, R. J., & Knesting, K. (2001). Zero tolerance, zero evidence: An analysis of school disciplinary practice. New Directions for Youth Development, 92, 17–43. https://doi. org/10.1002/yd.23320019204

Solmi, M., Radua, J., Olivola, M., Croce, E., Soardo, L., Salazar de Pablo, G., et al. (2022). Age at onset of mental disorders worldwide: Large-scale meta-analysis of 192 epidemiological studies. Molecular Psychiatry, 27(1), 281–295. https://doi.org/ 10.1038/s41380-021-01161-7

Substance Abuse and Mental Health Services Administration. (2019). Key Substance Use and Mental Health Indicators in the United States: Results from the 2019 National Survey on Drug Use and Health (HHS Publication No. PEP20-07-01-001, NSDUH Series H-55; p. 114). Center for Behavioral Health Statistics and Quality, Substance Abuse and Mental Health Services Administration. https://www.samhsa.gov/data/sites/ default/files/reports/rpt29393/2019NSDUHFFRPDFWHTML/ 2019NSDUHFFR1PDFW090120.pdf.

Suldo, S. M., Mihalas, S., Powell, H., & French, R. (2008). Ecological predictors of substance use in middle school students. School Psychology Quarterly, 23(3), 373–388. https://doi.org/10.1037/1045-3830.23.3.373

Urban Institute. (2017). How Has Education Funding Changed Over Time? How Has Education Funding Changed Over Time? http://urbn.is/fundingtrends.

Wacquant, L. (2009). Punishing the poor: The neoliberal government of social insecurity. Duke University Press.

Wallace, J. M., Jr., Goodkind, S., Wallace, C. M., & Bachman, J. G. (2008). Racial, Ethnic, and Gender Differences in School Discipline among U.S. High School Students: 1991–2005. The Negro. Educational Review, 59(1–2), 47–62.

Welch, M., Price, E. A., & Yankey, N. (2002). Moral Panic Over Youth Violence: Wilding and the Manufacture of Menace in the Media. Youth & Society, 34(1), 3–30. https:// doi.org/10.1177/0044118X02034001001

Whitaker, A., Torres-Guillén, S., Morton, M., Jordan, H., Coyle, S., Mann, A., et al. (2019). Cops and No Counselors: How the Lack of School Mental Health Staff Is Harming Students. American Civil Liberties Union. https://www.aclu.org/report/cop s-and-no-counselors.

Wildeman, C. (2011). Invited commentary: (Mass) Imprisonment and (Inequities in) Health. Am J Epidemiol, 173(5), 488–491. https://doi.org/10.1093/aje/kwq420

S.J. Prins et al.