Cross-classified Random Effects Models Research Articles

How well do workplace-based assessments support summative entrustment decisions? A multi-institutional generalisability study.

Assessment of the Core Entrustable Professional Activities for Entering Residency requires direct observation through workplace-based assessments (WBAs). Single-institution studies have demonstrated mixed findings regarding the reliability of WBAs developed to measure student progression towards entrustment. Factors such as faculty development, rater engagement and scale selection have been suggested to improve reliability. The purpose of this investigation was to conduct a multi-institutional generalisability study to determine the influence of specific factors on reliability of WBAs. The authors analysed WBA data obtained for clerkship-level students across seven institutions from 2018 to 2020. Institutions implemented a variety of strategies including selection of designated assessors, altered scales and different EPAs. Data were aggregated by these factors. Generalisability theory was then used to examine the internal structure validity evidence of the data. An unbalanced cross-classified random-effects model was used to decompose variance components. A phi coefficient of >0.7 was used as threshold for acceptable reliability. Data from 53 565 WBAs were analysed, and a total of 77 generalisability studies were performed. Most data came from EPAs 1 (n = 17 118, 32%) 2 (n = 10 237, 19.1%), and 6 (n = 6000, 18.5%). Low variance attributed to the learner (<10%) was found for most (59/77, 76%) analyses, resulting in a relatively large number of observations required for reasonable reliability (range = 3 to >560, median = 60). Factors such as DA, scale or EPA were not consistently associated with improved reliability. The results from this study describe relatively low reliability in the WBAs obtained across seven sites. Generalisability for these instruments may be less dependent on factors such as faculty development, rater engagement or scale selection. When used for formative feedback, data from these instruments may be useful. However, such instruments do not consistently provide reasonable reliability to justify their use in high-stakes summative entrustment decisions.

Exploring Disparities for Obesity in Korea Using Hierarchical Age-Period-Cohort Analysis With Cross-Classified Random Effect Models.

This research article investigates the age, period, and birth cohort effects on prevalence of obesity in the Korean population, with the goal of identifying key factors to inform effective public health strategies. We analyzed data from the Korea National Health and Nutrition Examination Survey, spanning 2007-2021, including 35,736 men and 46,756 women. Using the hierarchical age-period-cohort (APC) analysis with cross-classified random effects modeling, we applied multivariable mixed logistic regression to estimate the marginal prevalence of obesity across age, period, and birth cohort, while assessing the interaction between APC and lifestyle and socioeconomic factors. Our findings reveal an inverted U-shaped age effect on obesity, influenced by smoking history (P for interaction = 0.020) and physical activity (I for interaction < 0.001). The period effect was positive in 2020 and 2021, while negative in 2014 (P for period effect < 0.001). A declining trend in obesity prevalence was observed in birth cohorts from 1980s onward. Notably, disparities in obesity rates among recent birth cohorts have increased in relation to smoking history (P for interaction = 0.020), physical activity (P for interaction < 0.001), and residence (P for interaction = 0.005). Particularly, those born after 1960 were more likely to be obese if they were ex-smokers, physical inactive, or lived in rural areas. These findings highlight growing disparities in obesity within birth cohorts, underscoring the need for targeted health policies that promote smoking cessation and physical activity, especially in rural areas.

Approaches to Item-Level Data with Cross-Classified Structure: An Illustration with Student Evaluation of Teaching

Student evaluation of teaching (SET) questionnaires are ubiquitously applied in higher education institutions in North America for both formative and summative purposes. Data collected from SET questionnaires are usually item-level data with cross-classified structure, which are characterized by multivariate categorical outcomes (i.e., multiple Likert-type items in the questionnaires) and cross-classified structure (i.e., non-nested students and instructors). Recently, a new approach, namely the cross-classified IRT model, was proposed for appropriately handling SET data. To inform researchers in higher education, in this article, the cross-classified IRT model, along with three existing approaches applied in SET studies, including the cross-classified random effects model (CCREM), the multilevel item response theory (MLIRT) model, and a two-step integrated strategy, was reviewed. The strengths and weaknesses of each of the four approaches were also discussed. Additionally, the new and existing approaches were compared through an empirical data analysis and a preliminary simulation study. This article concluded by providing general suggestions to researchers for analyzing SET data and discussing limitations and future research directions.

Propensity Score Matching with Cross-Classified Data Structures: A Comparison of Methods

In the current study, we compare propensity score (PS) matching methods for data with a cross-classified structure, where each individual is clustered within more than one group, but the groups are not hierarchically organized. Through a Monte Carlo simulation study, we compared sequential cluster matching (SCM), preferential within cluster matching (PWCM), greedy matching (GM), and optimal full matching (OFM), using propensity scores from four different models. The results indicated that the four matching methods performed well when PSs were estimated with logistic regression containing both level-1 and level-2 covariates. When the level-2 covariates were omitted in the logistic regression PS model, matching methods resulted in biased treatment effect estimates. However, omission of level-2 covariates did not result in biased estimates when the PS model was a logistic cross-classified random effects model (CCREM). SCM and PWCM outperformed GM and OFM with a logistic CCREM that included level-1 and level-2 covariates.

The salience of social context, opioid antagonist use, and prior opioid exposure as determinants of fatal and non-fatal opioid overdoses.

This study examines the salience of social context for opioid overdoses in Boston from 2014 to 2019. Longitudinal negative binomial models with random effects indicated that higher levels of concentrated disadvantage, residential instability, and illicit drug activity increased annual block group counts of opioid overdoses. Logistic hierarchical and cross-classified random effects models indicated that the use of Narcan and greater exposure to drugs through previous opioid overdose and contextual lillicit drug crime activity reduced the odds of fatal opioid overdose relative to non-fatal opioid overdose. The findings suggest that the accurate tracking of both fatal and non-fatal overdoses, and a consideration of the broader social context, can facilitate effective public health resource allocation to reduce opioid overdoses.

Modern applications of cross-classified random effects models in social and behavioral research: Illustration with R package PLmixed.

Cross-classified random effects models (CCREMs) have been developed for appropriately analyzing data with a cross-classified structure. Despite its flexibility and the prevalence of cross-classified data in social and behavioral research, CCREMs have been under-utilized in applied research. In this article, we present CCREMs as a general and flexible modeling framework, and present a wide range of existing models designed for different purposes as special instances of CCREMs. We also introduce several less well-known applications of CCREMs. The flexibility of CCREMs allows these models to be easily extended to address substantive questions. We use the free R package PLmixed to illustrate the estimation of these models, and show how the general language of the CCREM framework can be translated into specific modeling contexts.

Assessing the effects of student perceptions of instructional quality: A cross-subject within-student design

This study examines associations between student-perceived instructional quality and a student’s motivation and achievement using unique data from Germany in which 9th-grade students reported on instructional quality in their English and German classes in the same academic year. Student outcomes, including self-efficacy, course grades, and standardized achievement were measured in both subjects. We utilized a cross-classified random effects model to decompose the variance in student reports of teacher support and monitoring, finding that 17% and 16%, respectively, of the variation in ratings was due to the student reporter with 26% and 27%, respectively, due to the teacher being rated. Students and teachers were further nested within classes of students who stayed together across the two subjects; however, almost no variance was attributable to class groupings. Student-reported teacher support was positively associated with achievement gains and self-efficacy at the within-student, between-student, and between-teacher levels whereas monitoring was inconsistently associated with outcomes. Standardized betas of perceived teacher support on test score gains were small (<0.10), small-to-medium on course grade gains (0.16–0.24), and medium to large on self-efficacy (0.33–0.52). Measuring the same students with different teachers allows us to disentangle dyadic student-teacher variance from student rater variance in student perceptions and outcomes and facilitates the use of powerful within-student analytical techniques to quantify the effects of educational environments on student learning and self-efficacy.

Trends in activities of daily living disability among Chinese older adults from 1998 to 2018: an age-period-cohort analysis.

This study aims to investigate the age, period, and cohort effects on trends in activities of daily living (ADL) disability among Chinese older adults; and to explore these three temporal effects on gender and residence disparities in disability. We utilized multiple cross-sectional waves of the Chinese Longitudinal Healthy Longevity Survey data (1998-2018), including 89,511 participants aged above 65years old. Our measurement of disability is the number of ADL items (dressing, bathing, indoor transferring, toileting, eating, and continence) participants can't perform independently. Hierarchical age-period-cohort cross-classified random effects models were conducted to investigate age, period and cohort trends in ADL disability. Results showed that ADL disability increased with age at an increasing rate. A V-shaped cohort trend and a fluctuated period trend were identified. Females and urban residents were associated with more ADL limitations. When age increased, the gender and residence gaps in disability further increased. The cohort-based gender and residence inequalities in ADL limitations converged with successive cohorts. The period-based residence gap in ADL limitations diverged throughout the 20-year period, while the corresponding period-based change in gender disparity was not significant. These findings suggested that age, period, and cohort had different and independent effects on ADL disability among Chinese older adults. The age effect on trends in ADL is stronger compared to period and cohort effects. The gender and residence disparities in disability increased with age and decreased with successive cohorts. These patterns might help inform healthcare planning and the priorities for medical resource allocation accordingly.

Age, Period, and Cohort Effects on Literacy Skills across Life Stages

Literacy skills are essential adult competencies for economic, social, political and cultural participation, which are linked to higher quality of life. Literacy skills are known to be lower for older age groups. However, relatively little is known about cohort and period effects, which provide clues to the sociohistorical impacts on literacy, in addition to the well-known age effects, over the life course. This study analyzed three nationally representative cross-sectional survey data of the U.S. adults at five time points, from the 1994 International Adult Literacy Survey (IALS), 2003 Adult Literacy and Life Skills Survey (ALL), and the 2012/2014/2017 Program for International Assessment of Adult Competencies (PIAAC). The total analytic sample was 17,450 adults age between 18 and 65 years old. The literacy measures were re-scaled (0-500 points) to be comparable across the surveys. An age-period-cohort hierarchical linear model (i.e., cross-classified random effects model) was constructed using the Bayes estimator. Individuals were cross-classified based on 14 five-year birth cohort and 5 periods (survey years) information. Results showed that literacy skills improved [95% credibility-interval (CI) for linear effect of age = (0.31, 1.07), but the rate of improvement declined over time, faster rates of decline in later life stages [95% CI for quadratic effect of age = (-0.17, -0.09)]. Additionally, the notable variability across the cohorts and periods [95% CI variances = (5.34, 52.52) and (2.30, 172.01), respectively] were identified. Possible explanations for the identified age, period and cohort effects on literacy and implications for adult competencies in later life are evaluated.

Trends in Cognitive Function Among Chinese Elderly From 1998 to 2018: An Age-Period-Cohort Analysis.

Objectives: To investigate the effects of age, period, and cohort (APC) on trends in cognitive function among the Chinese elderly, and to explore how gender gaps in cognitive function change with age, period, and cohort.Methods: This study used data from the Chinese Longitudinal Healthy Longevity Survey (CLHLS) from 1998 to 2018, and included 90,432 participants aged above 65 years old. The measurement of cognitive function was the score of the Mini-Mental State Examination (MMSE). Cross-classified random-effect models were used to investigate age, period, and cohort trends in cognitive function.Results: Mini-Mental State Examination scores decreased with age at an increasing rate. While the cohort effect was nearly stable, the period effect demonstrated a downward trend from 1998 to 2002 followed by a nearly flat line. Females were associated with lower MMSE scores than males. When age increased, the gender gaps in MMSE scores further increased. The period-based gender gaps in MMSE scores diverged throughout the 20 years, while the cohort-based gender disparities in MMSE scores converged with successive cohorts.Conclusions: Age, period, and cohort had different and independent effects on cognitive function among the Chinese elderly. The effect of age was stronger than that of period and cohort. Gender disparities in cognitive function increased with age and period, and decreased with successive cohorts.

Exploring the relations between child and word characteristics and preschoolers' word-learning

Preschool vocabulary interventions have reported modest increases in learning of target words, with wide variability among participants. To design interventions that are effective for all learners, more fine-grained information is needed. In the present study, English-only and dual language learner preschoolers (n = 128) were taught new words during shared book-reading. Using cross-classified random effects models, we investigated how both child and word characteristics related to vocabulary learning. Preschoolers were assessed on general receptive vocabulary (PPVT), executive function skills, and expressive knowledge of target words at pretest and posttest. Target words were rated on their perceptual accessibility, as captured by imageability, concreteness, and consistency of shape. Controlling for other predictors, target word pretest scores, PPVT, executive function skills, and words' imageability were all positive significant predictors of posttest vocabulary. Tailoring instructional strategies to better account for these factors may help to support the learning of a range of children and word types.

Age Patterns in Self-Reported Cognitive Impairment Among Older Latino Subgroups and Non-Latino Whites in the United States, 1997–2018: Implications for Public Health Policy

Background and ObjectivesU.S. Latinos are a heterogeneous population with unique characteristics related to individual-level socioeconomic and contextual factors based on nativity status and country of origin. Population aging and greater public awareness of dementia may contribute to an increasing prevalence of self-reported cognitive impairment. However, population-level trends in self-reported cognitive impairment among Latinos are unclear and it is unknown whether there are differences among Latino subgroups. Thus, this study aims to examine heterogeneity in self-reported cognitive impairment among older U.S. Latino subgroups.Research Design and MethodsWe used data from the 1997–2018 National Health Interview Survey to document age-specific patterns in self-reported cognitive impairment among U.S.-born Mexican, foreign-born Mexican, island-born Puerto Rican, foreign-born Cuban, and U.S.-born non-Latino Whites aged 60 and older. We estimated hierarchical age–period–cohort cross-classified random effects models (HAPC-CCREM) to isolate age patterns in self-reported cognitive impairment across disaggregated Latino subgroups and U.S.-born non-Latino Whites.ResultsThe overall prevalence of self-reported cognitive impairment increased from 6.0% in 1997 to 7.1% in 2018. This increase was evident among U.S.-born non-Latino Whites and U.S.-born and foreign-born Mexicans but not other Latino subgroups. Fully adjusted HAPC-CCREM estimates indicated that Latinos were more likely to self-report cognitive impairment than U.S-born non-Latino Whites (b = 0.371, p < .001). When disaggregated by Latino subgroup, the difference in the likelihood for self-reported cognitive impairment compared to U.S.-born non-Latino Whites was greatest for island-born Puerto Ricans (b = 0.598, p < .001) and smallest for foreign-born Cubans (b = 0.131, p > .05).Discussion and ImplicationsWe found evidence of considerable heterogeneity in the age patterns of self-reported cognitive impairment among U.S. Latino subgroups. We also detected large differences in the likelihood for self-reported cognitive impairment between U.S. Latino subgroups compared to U.S.-born non-Latino Whites. These results underscore the importance of differentiating between unique Latino subpopulations when studying population-level trends in cognitive function.

Place of study, field of study and labour-market region: What matters for wage differences among higher-education graduates?

This paper focuses on the structure and extent of wage differences among graduates of different higher-education institutions in Germany. We ask how large these differences are and how they relate to fields of study and regional labour markets. The results from our application of cross-classified random-effects models to a cohort of the DZHW Graduate Panel show that there is a considerable amount of wage variation depending on the graduates’ alma mater. However, this variation can be fully explained by structural characteristics: Selection based on individual characteristics is of only minor importance, while regional labour markets do matter. Most of all, however, the differences relate to fields of study.

The trainer matters: Cross-classified models of trainee reactions.

Despite the common belief that "training is only as effective as the trainer providing it" (Osborn, 2018, para. 1), training theory tends to underemphasize the trainer and instead focuses on training content and design as sources of training effectiveness. In this article, we examine whether the role of the trainer should be more central to training theory. We address this issue using a dataset of trainee reactions from more than 10,000 employees enrolled in professional development courses. We suggest that trainee reactions are more likely to be influenced by the trainer than by the content. Thus, trainee reactions should reflect more between-trainer variance than between-content variance. Across 2 studies in online and face-to-face contexts, cross-classified random-effects models provide general support for our hypotheses, with 1 notable exception: the trainer matters less for trainee reactions in online courses. Our findings suggest the trainer matters more than previously thought and, thus, training theory should incorporate the role of the trainer in training effectiveness. Based on our findings, we suggest that training researchers should (a) model the trainer as a source of variation in training evaluation metrics, (b) examine the effect of the trainer at multiple levels of analysis, and (c) explicitly model the role of the trainer in training theory and design. (PsycInfo Database Record (c) 2021 APA, all rights reserved).

Gender-specific temporal trends in overweight prevalence among Chinese adults: a hierarchical age-period-cohort analysis from 2008 to 2015

BackgroundAs a key health risk, the prevalence of overweight has been strikingly increasing worldwide. This study aimed to disentangle the net age, period, and cohort effects on overweight among Chinese adults by gender.MethodsData came from the Chinese General Social Survey from 2008 to 2015, which was a repeated cross-sectional survey (n = 55,726, aged 18 and older). χ2 or t tests were used to estimate the gender disparities in overweight and socioeconomic status (SES). A series of hierarchical age-period-cohort cross-classified random-effects models were performed using SAS version 9.4 to estimate the overall and gender-specific temporal trends of overweight, as well as the association between SES and overweight. Further, a series of line charts were used to present the age and cohort variations in overweight.ResultsAfter controlling for covariates, significant age and cohort effects were observed among adults in China (b = 0.0205, p < 0.001; b = 0.0122, p < 0.05; respectively). Specifically, inverted U-shaped age effects were identified for both genders, with a high probability of overweight occurring in middle age (b = –0.0012, p < 0.001). Overweight was more prevalent among men than women before 60 years old, and this trend reversed thereafter (b = –0.0253, p < 0.001). Moreover, men born during the war (before 1950) and reform cohorts (after the 1975s) demonstrated a substantial decline in overweight, while men born in 1950–1975 showed an increasing trend in overweight prevalence (b = 0.0378, p < 0.05). However, the cohort effect on women was not statistically significant. Additionally, a higher SES was related to an elevated probability of overweight.ConclusionGender-specific age and cohort effects on the prevalence of overweight were observed among Chinese adults. Both China and other developing countries need to pay attention to the coming obesity challenge and related health inequality. Full life-cycle overweight prevention interventions should focus on middle-aged adults, men born in the war and reform eras, and adults with a higher SES.

The application of meta-analytic (multi-level) models with multiple random effects: A systematic review.

In meta-analysis, study participants are nested within studies, leading to a multilevel data structure. The traditional random effects model can be considered as a model with a random study effect, but additional random effects can be added in order to account for dependent effects sizes within or across studies. The goal of this systematic review is three-fold. First, we will describe how multilevel models with multiple random effects (i.e., hierarchical three-, four-, five-level models and cross-classified random effects models) are applied in meta-analysis. Second, we will illustrate how in some specific three-level meta-analyses, a more sophisticated model could have been used to deal with additional dependencies in the data. Third and last, we will describe the distribution of the characteristics of multilevel meta-analyses (e.g., distribution of the number of outcomes across studies or which dependencies are typically modeled) so that future simulation studies can simulate more realistic conditions. Results showed that four- or five-level or cross-classified random effects models are not often used although they might account better for the meta-analytic data structure of the analyzed datasets. Also, we found that the simulation studies done on multilevel meta-analysis with multiple random factors could have used more realistic simulation factor conditions. The implications of these results are discussed, and further suggestions are given.

Social transition and health inequality in China: an age-period-cohort analysis

ObjectivesTo examine self-rated health (SRH) trajectories by age, period and cohort in the dynamic environment of China and to focus on sex, educational, family economic status (FES) and regional disparities in these temporal trajectories. Study designRepeated cross-sectional data from the World Value Survey from 1990 to 2012 (n = 7731) were used. MethodsAn age-period-cohort (APC) model with an intrinsic estimator method was applied, using SRH as a continuous variable to analyse SRH trajectories by age, period and cohort in different groups in China. Robust analyses were conducted using SRH as a binary variable and a hierarchical APC cross-classified random effects model. ResultsSRH was adversely associated with age, and a global trend of declining SRH was observed between 1990 and 2012; however, SRH generally increased across successive cohorts in China. SRH was better in males, individuals from the eastern region, and in those with higher levels of education and higher FES. Regional and FES disparities in SRH increased with age and across successive cohorts; however, educational disparities in SRH first decreased and then increased with age and across successive cohorts, and sex disparities in SRH decreased across successive cohorts, in general. ConclusionsA well-performing social environment is related to better population health. The cumulative disadvantage in health among individuals from underprivileged regions/families with less education should be the focus of social transition in China.

Tobacco Retail Density and Initiation of Alternative Tobacco Product Use Among Teens

PurposeThe rise of noncigarette, alternative tobacco product (ATP) use among adolescents may be due in part to an increase in retail availability of ATPs. We examined whether proximity and density of tobacco retailers near students’ homes are associated with a higher likelihood of initiating ATP use over time. MethodsUsing data from 728 adolescents (aged 13–19 years at baseline) residing in 191 different neighborhoods and attending 10 different California high schools, longitudinal multilevel and cross-classified random effect models evaluated individual-level, neighborhood-level, and school-level risk factors for ATP initiation after 1 year. Covariates were obtained from the American Community Survey and the California Department of Education. ResultsThe sample was predominantly female (63.5%) and was racially and ethnically diverse. Approximately one third of participants (32.5%) reported ever ATP use at baseline, with 106 (14.5%) initiating ATP use within 1 year. The mean number of tobacco retailers per square mile within a tract was 5.66 (standard deviation = 6.3), and the average distance from each participant's residence to the nearest tobacco retailer was .61 miles (standard deviation = .4). Living in neighborhoods with greater tobacco retailer density at baseline was associated with higher odds of ATP initiation (odds ratio = 1.22, 95% confidence interval = 1.07–2.12), controlling for individual and school factors. ConclusionsTobacco retailers clustered in students’ home neighborhood may be an environmental influence on adolescents’ ATP use. Policy efforts to reduce adolescent ATP use should aim to reduce the density of tobacco retailers and limit the proximity of tobacco retailers near adolescents’ homes and schools.

SOCIOCULTURAL VARIABILITY IN SELF-REPORTED COGNITIVE IMPAIRMENT AMONG OLDER LATINOS IN THE UNITED STATES

Cognitive impairment is a major public health concern in the United States. Research indicates cognitive impairment is higher for older U.S. Latinos than non-Latino whites, due in part to Latinos having longer life expectancy, lower educational attainment, and a higher prevalence of diabetes and cardiovascular disease. Prior studies on cognition have largely examined “Latinos” as a monolithic group. However, Latinos are heterogeneous in composition with unique socio-cultural characteristics based on nativity and country of origin. Accordingly, we used data from the 1997-2017 National Health Interview Survey (NHIS) to document age-specific trends in in self-reported cognitive impairment among US-born Mexican, foreign-born Mexican, island-born Puerto Rican, foreign-born Cuban, and non-Latino white adults aged 60 and older. Given the repeated cross-sectional nature of these data, we estimated hierarchical age period–cohort (HAPC) cross-classified random-effects model (CCREM) to isolate age trends in self-reported cognitive impairment across Latino subgroups and non-Latino whites. Results indicate significant heterogeneity among Latino subgroups, with island-born Puerto Ricans exhibiting the highest rates of cognitive impairment and foreign-born Cubans the lowest. Conversely, US-born and foreign-born Mexicans exhibited rates in between these two. All Latino subgroups statistically differed from non-Latino whites. Socio-demographic controls account for approximately 33%-45% of the disparity, but fully account for foreign-born Cubans and non-Latino whites differences. These findings indicate the importance of considering nativity and country of origin when assessing cognitive outcomes among older Latinos. Understanding minority and immigrant differences in cognitive impairment has implications for the development and implementation of culture-appropriate programs to promote healthy brain aging.

Effects of educational policy on unhappiness of middle and high school students in Korea, 2006-2016

Abstract Background Educational policies may have a substantial impact on health and well-being of students, who spend most of their daily lives at school. Policies on secondary education in Korea has changed in the late 2000s, which diversified college entrance exams (CEEs) and made the National Curriculum (NC) more autonomous to mitigate students’ burden of preparing for CEEs. This study aimed to estimate effects of educational policy changes on unhappiness, which is known to be associated with health status such as mortality, morbidity, and prognosis, of students in secondary education. Methods Using repeated cross-sectional self-reported data of 773,362 students in the Korea Youth Risk Behavior Web-based Survey during 2006-2016, cross-classified random effects models were fitted to estimate the fixed cohort effects of educational policy changes on unhappiness while considering individual factors and random effects of periods and grade cohorts. Results Students’ unhappiness increased when the NC began to become autonomous (odds ratio [OR] 1.21, 95% confidence interval [CI] 1.03-1.42), but it increased less after the autonomous NC was fully applied (OR 1.14, 95% CI 0.92-1.42). Two times of diversification of CEEs increased unhappiness (OR 1.21, 95% CI 1.06-1.38 for the first diversification and OR 1.33, 95% CI 1.07-1.67 for the second diversification). Unhappiness was high in students with high stress level (OR 7.59, 95% CI 7.43-7.75), low academic performance (OR 1.48, 95% CI 1.44-1.52), and low household economic status (OR 2.32, 95% CI 2.26-2.39). Conclusions Enhancement of autonomy of the NC and diversification of CEEs increased students’ unhappiness. Educational policy changes that diversified competencies required for college admission might result in an increase in academic burden in the context of Korea, where entrance into prestigious universities is crucial for success in life. Improvement of educational policy is needed to increase students’ happiness. Key messages Educational policies may have a substantial impact on health and well-being of students, who spend most of their daily lives at school. Education policy changes in Korea increased students’ unhappiness, which might be due to an increase in academic burden resulted from diversification of competencies required for college admission.