Cohort profile: the Korean Youth Health Behavior Panel Survey (KYHPS)

Study population and design

The KYHPS used a multistage cluster sampling design to obtain a nationally representative sample of students enrolled in elementary schools in Korea. The target population comprised sixth-grade students attending elementary schools in Korea in 2019, excluding those in schools located on Jeju Island. Jeju Island was excluded from the sampling frame primarily for efficiency and operational feasibility because of its small population share (approximately 1.45%), which was considered to have limited influence on national estimates. Including a geographically isolated and very small stratum would have substantially increased costs while providing only minimal gains in statistical precision, a concern that is particularly relevant in longitudinal panel surveys. Sixth-grade students were selected for both developmental and methodological reasons. This stage immediately precedes the transition to middle school, allowing the cohort to capture changes in health behaviors across a major educational transition. Including students in lower grades would have required a substantially longer follow-up period before the initiation of key health-risk behaviors could be observed, thereby reducing study efficiency. Furthermore, because the survey was self-administered, sixth-grade students were considered developmentally more capable of providing accurate and reliable responses. The sampling frame was constructed from the official nationwide school administrative database provided by the Korean Educational Development Institute on the basis of records as of April 2018. This database includes all registered elementary schools nationwide and therefore provides near-complete coverage of the eligible target population. The sampling process consisted of 3 stages: stratification, allocation, and selection. During stratification, 25 strata were constructed according to province and region size (metropolitan, small- and medium-sized cities, and rural areas). A compromise allocation method was used to secure an adequate number of students not only at the national level but also within each province and region size category. Specifically, the number of schools in each province was allocated proportionally to the square root of the number of schools, whereas within each province, the number of schools by region size was allocated in proportion to the number of schools. At least 2 schools were assigned to each stratum. A stratified multistage probability sampling method was then applied, with schools as the primary sampling units and classes as the secondary sampling units. Within each stratum, the school list was ordered by province, district, and school size, measured by the number of sixth-grade classes. Schools were subsequently selected by systematic probability sampling with probability proportional to size. Because the unit of selection within schools was the class rather than the individual student, the number of classes was considered a more appropriate measure of size for achieving stable self-weighting. A total of 260 schools were selected from a sampling frame of 3,741 eligible elementary schools using this stratified multistage probability sampling design. Within each sampled school, 2 or 3 classes were randomly selected, and all students in the selected classes were included in the survey sample [8].

Enrollment and follow-up

Baseline characteristics

At baseline, the weighted sample consisted of 51.5% boys and 48.5% girls. By residential area, the largest proportion of students lived in Gyeonggi-do (28.1%), followed by Seoul (15.9%), whereas smaller proportions lived in Sejong (0.9%) and Ulsan (2.9%). Most students rated their subjective academic performance as middle (40.2%) or upper-middle (38.9%), and only 0.9% rated it as low. Similarly, most students perceived their household economic status as middle (56.4%) or upper-middle (26.7%), whereas 0.6% reported it as low (Table 1). In addition to demographic characteristics, several key health behavior indicators were included to provide a more comprehensive baseline profile of the panel participants, including lifetime tobacco product use, lifetime alcohol use (sip- and glass-based), skipping breakfast on at least 5 days per week, fast-food consumption at least 3 times per week, and engaging in at least 60 minutes of physical activity on at least 5 days per week, as shown in Table 1.

Table 1.

Baseline demographic characteristics of participants in the Korean Youth Health Behavior Panel Survey (Wave 1)

Retention

Among the 5,051 sixth-grade students from 260 elementary schools who participated in the baseline survey, discontinuous participation, defined as missing 1 or 2 survey waves but returning in later waves and being counted under cross-sectional weights, was first observed in Wave 3. When these participants were included, panel retention rates were 93.1% (n=4,702 students) in Wave 2, 87.8% (n=4,438) in Wave 3, 87.3% (n=4,409) in Wave 4, 84.0% (n=4,243) in Wave 5, and 82.0% (n=4,141) in Wave 6. When only continuous participants, defined as those who completed every survey wave up to the given year and were used for longitudinal weight estimation, were considered, panel retention rates were slightly lower: 86.4% (n=4,365) in Wave 3, 82.6% (n=4,172) in Wave 4, 78.9% (n=3,986) in Wave 5, and 76.5% (n=3,864) in Wave 6 (Figure 1) [8].

Figure 1.

Panel retention rates of the Korean Youth Health Behavior Panel Survey from Wave 1 (baseline, 2019) to Wave 6 (2024). Values are presented as % (number). 1Continuous participants: participated in all waves up to the given year. 2Discontinuous participants: missed at least 1 previous wave but participated in the given wave.

To assess potential selection bias due to attrition, baseline characteristics were compared between participants who did and did not participate in Wave 6. The participation rate was significantly higher among boys than among girls (83.3 vs. 80.5%, p=0.009) and significantly lower among students from households reporting the highest economic status (77.8%, p=0.002). The relatively lower participation rate in the highest-economic-status group may indicate that survey incentives were less likely to motivate continued participation in this group. However, no significant differences were observed with respect to baseline academic performance, lifetime tobacco product use, lifetime alcohol use, skipping breakfast on at least 5 days per week, fast-food consumption at least 3 times per week, or engaging in 60 or more minutes of physical activity on at least 5 days per week (Table 2).

Table 2.

Comparison of baseline characteristics between participants and non-participants in Wave 6 of the Korean Youth Health Behavior Panel Survey

To minimize attrition, data collection was conducted by a nationwide professional survey agency with extensive experience in longitudinal panel studies. The same trained interviewers followed participants across waves to foster rapport and trust. Survey timing and locations were arranged flexibly to accommodate participants’ schedules. Modest incentives were provided to encourage continued participation. In addition, participants were contacted on major occasions, such as national holidays or birthdays, to maintain engagement with the panel.

For analyses examining change over time, longitudinal weights restricted to respondents who participated in all relevant waves should be applied. Cross-sectional weights should be used when analyzing outcomes at a specific time point.

Ethics statement

The study protocol was approved by the Institutional Review Board of Daegu Catholic University Medical Center (CR-19-107).

Survey items

Adolescence is the life stage during which exposure to the social determinants of health is most dynamic and diverse [9]. During this period, influences beyond the family environment become increasingly important, with peers, the media, and education playing major roles. At the same time, community-level and structural-level determinants continue to shape adolescent health trajectories [3]. In line with this perspective, the KYHPS survey instrument was developed on the basis of the ecological model of health behaviors. Initially, the KYHPS focused on 4 key domains of adolescent health behaviors: tobacco use, which was the primary focus, alcohol consumption, diet, and physical activity. Within each domain, both key behavioral variables and their predisposing factors were assessed. Consistent with the ecological model, questionnaire items were organized according to levels of influence, including individual, family, peer, school, and community factors.

The questionnaire was developed on the basis of results from a preliminary study conducted from 2017 to 2018 to establish the KYHPS. This preliminary study was informed by a nationwide literature review, including the KYRBS and international surveys targeting adolescents, and the questionnaire items were developed through comprehension and reliability testing [10]. For items introduced in later waves, outcome-related variables were primarily adopted from existing national surveys to ensure comparability and established validity. For other newly added items, particularly those assessing additional determinants or contextual factors, comprehension testing and pilot testing were conducted before implementation.

Table 3 provides an overview of the major survey domains and representative questionnaire items administered across survey waves. The tobacco use domain, a major focus of the KYHPS, included use of various tobacco products as well as items not covered in the KYRBS, such as experience with flavored tobacco. In Waves 1–4, an introductory screening question assessed lifetime use of any tobacco product regardless of product type. Beginning in Wave 5, this item was removed, and respondents were asked directly about their experience with individual tobacco products. The domain also included questions on perceptions, attitudes, and norms related to tobacco use. In Wave 5, additional items were introduced to assess perceptions of the relative harmfulness of e-cigarettes and heated tobacco products, as well as exposure to advertising and promotional activities and attitudes toward these products. The alcohol use domain included items on sipping-based drinking, which have not been included in previous national surveys in Korea, to evaluate the significance of early sipping experiences in shaping adolescent drinking behaviors [11] within the Korean context. Subsequently, items assessing behavioral changes related to the coronavirus disease 2019 (COVID-19) pandemic were added in Waves 3–4 (2021–2022). In Wave 3 (2021), the Generalized Anxiety Disorder 7-item scale and the Family Affluence Scale were incorporated. Wave 5 (2023) included additional items on subjective health status and physician-diagnosed conditions, including asthma and allergic rhinitis (Table 3).

Table 3.

Overview of student survey items across waves in the Korean Youth Health Behavior Panel Survey (Wave 1–6)

The parent questionnaire, which consisted of 33 items in Wave 1, addressed family relationships with the student as well as family-level factors that may influence the development of the student’s health behaviors. The school questionnaire, administered only in Wave 1, was designed to assess the level of interest in student health and related activities at the elementary school level. Detailed questionnaire items, including those administered to parents and school representatives, are available in the official KYHPS user guide and statistical reports [8,12].

Survey method

From Waves 1 to 5, students completed the survey using tablet-assisted self-interviews (TASI) after receiving instructions from trained interviewers who visited their households. This protocol was maintained during the COVID-19 pandemic, with interviewers continuing to conduct individual household visits. The parent survey was administered using the same method as the student survey through Wave 3; beginning in Wave 4, telephone interviews were also permitted. School representatives completed an online self-administered questionnaire.

In Wave 1, after sample schools had been selected and preparatory procedures completed, the survey was conducted from September 2019 to early January 2020. In Waves 2 and 3 (2020–2021), the survey period was delayed by approximately 2 months from the originally scheduled start in May because of the COVID-19 pandemic and was instead conducted between July and December. Beginning in Wave 4 (2022), the survey schedule returned to normal, with data collection taking place from May to September. The Wave 5 survey (2023) was conducted from May to November, and the Wave 6 survey (2024) from April to October, with most students participating during the first semester.

Tobacco product use

Annual incidence rates of lifetime tobacco product use by product type, including conventional cigarettes, e-cigarettes, and heated tobacco products, were calculated for each grade. Among participants who were continuously followed through Wave 6, the annual incidence of tobacco product initiation increased with grade progression: 0.29% in the first year of middle school, 1.34% in the second year of middle school, 2.38% in the third year of middle school, 3.29% upon entry into high school, and 3.22% in the second year of high school. In addition, longitudinal follow-up data showed that among adolescents who reported experience with 2 or more tobacco products, the first product used was most commonly conventional cigarettes (60.7%), followed by e-cigarettes (36.3%), heated tobacco products (1.8%), and unknown products (1.2%) [8].

Notably, in Wave 6 (11th grade), the prevalence of current e-cigarette use among female students exceeded that of conventional cigarette use for the first time [8]. This pattern has not yet been observed in the KYRBS [7], underscoring the need for continued surveillance. In Wave 6, 72.6% of current e-cigarette users were dual or multiple users who also used conventional cigarettes or heated tobacco products. This proportion was higher than that observed among current cigarette smokers (64.9%). In addition, 50.2% of e-cigarette users reported having attempted to quit, compared with 75.1% of cigarette smokers. Furthermore, 77.3% of adolescents with lifetime tobacco product use reported initiating use with a flavored product [8,13].

Alcohol use

Annual incidence rates of alcohol initiation also increased across grade levels. An analysis of data from Waves 1–5 identified several significant predisposing factors for initiation, including male sex, prior experience with sipping alcohol, intention to drink in the future, low perceived risk associated with alcohol use, having family members who drink, exposure to drinking scenes in the media, and witnessing alcohol consumption in public spaces [14]. The adjusted odds ratio for alcohol initiation among participants with prior sipping experience was 2.83 (95% confidence interval, 2.39 to 3.36) [13].