View in EDS HTML Full Text PDF Full Text

A Comparison of Repeat Cross-Sectional and Longitudinal Results from the COMPASS Study: Design Considerations for Analysing Surveillance Data over Time

Saved in:
Bibliographic Details
Title: A Comparison of Repeat Cross-Sectional and Longitudinal Results from the COMPASS Study: Design Considerations for Analysing Surveillance Data over Time
Language: English
Authors: Butler, Alexandra E., Battista, Kate, Leatherdale, Scott T., Meyer, Samantha B., Elliott, Susan J., Majowicz, Shannon E.
Source: International Journal of Social Research Methodology. 2022 25(5):597-609.
Availability: Routledge. Available from: Taylor & Francis, Ltd. 530 Walnut Street Suite 850, Philadelphia, PA 19106. Tel: 800-354-1420; Tel: 215-625-8900; Fax: 215-207-0050; Web site: http://www.tandf.co.uk/journals
Peer Reviewed: Y
Page Count: 13
Publication Date: 2022
Document Type: Journal Articles
Reports - Research
Descriptors: Research Design, Longitudinal Studies, Data Analysis, Case Studies, Social Science Research, Public Health, Public Policy, Foreign Countries, Youth, Eating Habits
Geographic Terms: Canada
DOI: 10.1080/13645579.2021.1922804
ISSN: 1364-5579
1464-5300
Abstract: Selection of appropriate study design and analytical methods is critical for producing robust research, as the design and analytical approach used can ultimately shape results and their interpretation. The objective of this research was to examine how findings from a large repeat cross-sectional data system compare to those from a sample of longitudinal data nested within the same data system, using milk and milk alternative consumption as a case study. Overall, the repeat cross-sectional and longitudinal findings were consistent, and this article highlights the strengths and limitations of both methodologies under examination. Application to social research and public health and policy are discussed.
Abstractor: As Provided
Entry Date: 2023
Accession Number: EJ1367264
Database: ERIC
Full text is not displayed to guests.
FullText Links:
  – Type: pdflink
    Url: https://content.ebscohost.com/cds/retrieve?content=AQICAHj0k_4E0hTGH8RJwT4gCJyBsGNe_WN95AvKlDbXJGqwxwEE_agdTK11EhJEdN7HUSRoAAAA4zCB4AYJKoZIhvcNAQcGoIHSMIHPAgEAMIHJBgkqhkiG9w0BBwEwHgYJYIZIAWUDBAEuMBEEDMLTiZdq-6lUOVMuKQIBEICBm3rsjQc1cX31bEiJu1_wYQAdMlXWCi0tpoDrQe8sNGvAG5wdkvZ30T4_fE-F1yO7n-e8r4Mab7THmAf9y6eiroE8SPv4zXTsrzIjcZn_LPsQLcp4UJIi_GZuqLW9iRp3yBrd1wqfm2bAWVNUc9CUhyD0GA-kwPTjxn9sUS10ICtSb2dVkHoycoP-2njHY6qFcLyrPXuJBONlSdGb
Text:
  Availability: 1
  Value: <anid>AN0159232770;9eb01sep.22;2022Sep23.01:22;v2.2.500</anid> <title id="AN0159232770-1">A comparison of repeat cross-sectional and longitudinal results from the COMPASS study: design considerations for analysing surveillance data over time </title> <p>Selection of appropriate study design and analytical methods is critical for producing robust research, as the design and analytical approach used can ultimately shape results and their interpretation. The objective of this research was to examine how findings from a large repeat cross-sectional data system compare to those from a sample of longitudinal data nested within the same data system, using milk and milk alternative consumption as a case study. Overall, the repeat cross-sectional and longitudinal findings were consistent, and this article highlights the strengths and limitations of both methodologies under examination. Application to social research and public health and policy are discussed.</p> <p>Keywords: Cross-sectional design; longitudinal design; cohort study; milk and milk alternatives; youth</p> <hd id="AN0159232770-2">Introduction</hd> <p>Evidence-informed research is the foundation for informing effective public health policies, and greatly impacts the health status of the population (Brownson et al., [<reflink idref="bib4" id="ref1">4</reflink>]; Tudisca et al., [<reflink idref="bib39" id="ref2">39</reflink>]). Selection of appropriate study design and analytical methods is critical for producing robust research, as the design and analytical approach used can ultimately shape results and their interpretation (Louis et al., [<reflink idref="bib23" id="ref3">23</reflink>]). Much of public health research relies on observational (i.e. non-experimental) or quasi-experimental research designs, in which researchers have no control over participant behaviours regarding the intervention or predictive factors of interest. Such designs can involve examining behaviour at a particular point in time or examine behavioural trends and patterns over time. The collection of multi-time point observational data typically falls into two overarching study designs: repeat cross-sectional and longitudinal. Both approaches have unique modeling and methods implications that can lead to alternative results that can threaten validity (Louis et al., [<reflink idref="bib23" id="ref4">23</reflink>]). Herein, we draw on empirical data to demonstrate the importance of considering the implications of study design<emph>a priori</emph>, and to discuss the strengths and limitations between cross-sectional and longitudinal research.</p> <p>Cross-sectional data are collected at a single time point, while repeat cross-sectional data are collected at two or more time points, though not necessarily on the same sample of participants (although in some study contexts they may be the same individuals participating at more than one time point). Surveillance studies typically use a cross-sectional design, as they are often simpler and more cost-effective (Berger, [<reflink idref="bib2" id="ref5">2</reflink>]; Setia, [<reflink idref="bib35" id="ref6">35</reflink>]). On the other hand, some surveillance (or evaluation) studies collect longitudinal data where data are collected on the same participants over multiple time points. While repeat cross-sectional data can be used to examine changes in sample characteristics over time (e.g. population prevalence, averages), it cannot be used to examine changes in individual outcomes over time due to the differences between the samples at the different time points. Longitudinal data can be used to examine change in individual outcomes, and as such temporality of effects within individuals can be established to elucidate cause and effect with the appropriate statistical modeling (Caruana et al., [<reflink idref="bib7" id="ref7">7</reflink>]).</p> <p>In public health research, longitudinal studies are often used to evaluate relationships between disease development and risk factors, intervention outcomes, and individual outcome trajectories (Caruana et al., [<reflink idref="bib7" id="ref8">7</reflink>]). Longitudinal data can allow researchers to control for time in their analysis and provide the capacity to examine the variability between and within participants over time. However, longitudinal data are subject to loss of follow-up (attrition), which can jeopardize the internal validity of the study and ultimately the quality of the research findings (Dettori, [<reflink idref="bib12" id="ref9">12</reflink>]). Additionally, longitudinal studies typically have smaller sample sizes, are more financially demanding (Caruana et al., [<reflink idref="bib7" id="ref10">7</reflink>]), and are less representative without sample replenishment compared to repeat cross-sectional studies.</p> <p>Typically, cross-sectional and longitudinal results are consistent when using data from the same sample and study instrument (Costa et al., [<reflink idref="bib9" id="ref11">9</reflink>]; Darling, [<reflink idref="bib10" id="ref12">10</reflink>]; Terracciano et al., [<reflink idref="bib38" id="ref13">38</reflink>]). However, if research produces disparate results between cross-sectional and longitudinal data (McCormick et al., [<reflink idref="bib25" id="ref14">25</reflink>]; Mueller et al., [<reflink idref="bib26" id="ref15">26</reflink>]), the differences may be due to varaitions between the samples (cohort effects), or circumstances during which the research was conducted (period effects) (Cook & Ware, [<reflink idref="bib8" id="ref16">8</reflink>]). Understanding the differences between longitudinal and cross-sectional data is critical when studying populations during transitional periods (McCormick et al., [<reflink idref="bib25" id="ref17">25</reflink>]) and previous research has demonstrated examples of varying results between cross-sectional and longitudinal data (Mullis et al., [<reflink idref="bib27" id="ref18">27</reflink>]; Ribeiro et al., [<reflink idref="bib32" id="ref19">32</reflink>]; Rönnlund et al., [<reflink idref="bib33" id="ref20">33</reflink>]; Zuckermann et al., [<reflink idref="bib43" id="ref21">43</reflink>]). For example, research examining changes in adolescent self-esteem did not find self-esteem scores to be significantly different across high school grades when examined cross-sectionally, whereas longitudinal findings in the same sample demonstrated clear developmental trends in self-esteem over time (Mullis et al., [<reflink idref="bib27" id="ref22">27</reflink>]); in this case, the within-person variability for emerging self-esteem demonstrated that younger adolescents are more sensitive to changes, and self-esteem was found to positively increase over the high school years (Mullis et al., [<reflink idref="bib27" id="ref23">27</reflink>]). A recent study examining adolescent cannabis use found that while no differences in cannabis use were observed across the longitudinal cohorts, repeat cross-sectional analysis of the same sample demonstrated an increase in odds of ever trying cannabis over time (Zuckermann et al., [<reflink idref="bib43" id="ref24">43</reflink>]). Adolescence is a stage characterized by complex changes and variation in the timing of individual mental and physical development (Giedd et al., [<reflink idref="bib14" id="ref25">14</reflink>]; McCormick et al., [<reflink idref="bib25" id="ref26">25</reflink>]; Palmert & Boepple, [<reflink idref="bib28" id="ref27">28</reflink>]; Parent et al., [<reflink idref="bib29" id="ref28">29</reflink>]; Spear, [<reflink idref="bib36" id="ref29">36</reflink>]) that can vary by rate and direction. Therefore, cross-sectional studies may be poorly suited to examine and delineate behavioural changes during this time period, and may mislead or contradict longitudinal findings (Kraemer et al., [<reflink idref="bib17" id="ref30">17</reflink>]; Maxwell & Cole, [<reflink idref="bib24" id="ref31">24</reflink>]; Mullis et al., [<reflink idref="bib27" id="ref32">27</reflink>]). Given the differences in development and behaviour during adolescence, it is important that research questions that seek to examine changes, address within-person variance.</p> <p>The objective of this research was to examine how findings from a large repeat cross-sectional data system compare to those from a sample of longitudinal data nested within the same data system. This analysis was part of a larger project that aimed to examine milk and milk alternative (MMA) consumption among Canadian youth (Butler, Battista, Leatherdale et al., [<reflink idref="bib5" id="ref33">5</reflink>]; Butler, Battista, Leatherdale et al., [<reflink idref="bib6" id="ref34">6</reflink>]). Specifically, we analysed data from the COMPASS study on MMA consumption among Canadian youth. COMPASS data are ideal to answer the research questions of interest given the study's approach of having large repeat cross-sectional samples where some participants are also followed longitudinally. MMA consumption is an appropriate focus outcome for this methodological assessment because we expected observable time trends in consumption patterns (Health Canada, [<reflink idref="bib15" id="ref35">15</reflink>]) and overall declining MMA consumption in the last decade, (Pierre, [<reflink idref="bib30" id="ref36">30</reflink>]) and because there were minimal missing data (allowing for a strong comparison between repeat cross-sectional and longitudinal results).</p> <hd id="AN0159232770-3">Methods</hd> <p>COMPASS is a prospective cohort study (2012–2021) collecting longitudinal hierarchical data at the student- and school-levels in Canada, and was designed to evaluate if policy and program changes impact youth health behaviour outcomes over time. Each year of data collection in COMPASS harvest a large repeat cross-sectional sample whereby linked longitudinal data are nested within these samples over time (through grade 9–12). Using an active-information passive-consent protocol, eligible students participated in COMPASS during class time. This paper uses the six most recent waves (i.e. school years; 2013–14 to 2018–19) of student-level data from Canadian secondary schools participating in COPMASS (<ulink href="http://www.compass.uwaterloo.ca">www.compass.uwaterloo.ca</ulink>) (Leatherdale et al., [<reflink idref="bib19" id="ref37">19</reflink>]). Using a unique, self-generated code (Qian et al., [<reflink idref="bib31" id="ref38">31</reflink>]), students were anonymously linked in grade 9 through grade 12, resulting in three complete 4-year cohorts. Primary reasons for unlinked data are absenteeism in any given data collection or incomplete responses to questions used to generate unique identifier codes. Full details on the data linkage process are available online (Battista et al., [<reflink idref="bib1" id="ref39">1</reflink>]; Bredin & Leatherdale, [<reflink idref="bib3" id="ref40">3</reflink>]; Qian et al., [<reflink idref="bib31" id="ref41">31</reflink>]). COMPASS received ethics approval from the University of Waterloo Human Research Ethics Committee (ORE #: 30118) and all participating school boards.</p> <hd id="AN0159232770-4">Sample</hd> <p>We analysed data collected from a convenience sample of grades 9–12 students (aged: 14–18), attending 50 secondary schools in Ontario (N = 45), and Alberta (N = 5), Canada, that participated in all six waves of COMPASS data collection. Student data were examined as both repeat cross-sectional data (wave 1 [2013–14] n = 26,586; wave 2 [2014–15] n = 25,460; wave 3 [2015–16] n = 25,450; wave 4 [2016–17] n = 24,860; wave 5 [2017–18] n = 24,481; and wave 6 [2018–19] n = 23,781), and as three longitudinal cohort samples (cohort 1 [2013–14 to 2016–17] n = 2,405; cohort 2 [2014–15 to 2017–18] n = 2,066; and cohort 3 [2015–16 to 2018–19] n = 2,066). The longitudinal cohort sample consisted of a nested subset of students who were in grade 9 at baseline and for whose responses were successfully linked across 4 years.</p> <hd id="AN0159232770-5">Measures</hd> <p>Student-level data were collected through paper-based, self-administered, anonymous questionnaires that were completed during class time. The questionnaire was designed to measure a variety of health behaviours and outcomes, including dietary behaviours. The average response rate for the student questionnaire was approximately 80%, with the primary reason for non-response being absenteeism at the time of the survey. The parental refusal rate was less than 0.25%.</p> <hd id="AN0159232770-6">Outcome variable</hd> <p>MMA consumption was measured by asking students 'Yesterday, from the time you woke up until the time you went to bed, how many servings of milk and alternatives did you have? One "food guide" serving of milk or milk alternatives includes milk, fortified soy beverage, reconstituted powder milk, canned milk, yogurt or kefir (another type of cultured milk products), and cheese'. To assist students with quantifying the number of servings consumed, images of serving size examples excerpted from the 2011 food guide were provided within the question. Daily MMA servings were measured as continuous, and response options included: none, 1 serving, 2 servings, 3 servings, 4 servings, 5 servings and 6 or more servings. Moderately high test–retest reliability (ICC 0.69) and fair validity (ICC 0.60) have been observed for the self-reported MMA consumption measure in the student COMPASS questionnaire (Leatherdale & Laxer, [<reflink idref="bib21" id="ref42">21</reflink>]).</p> <p>Students also reported demographic data (grade, gender, ethnicity) and health behaviour data. Health behaviour indicators examined in this study included: weight goals, physical activity, daily breakfast consumption, participation in a school breakfast program, bringing lunch from home and purchasing snacks from vending machines. These variables were selected based on their significant association with student MMA consumption in our previous analyses (Butler, Battista, Leatherdale et al., [<reflink idref="bib5" id="ref43">5</reflink>]; Butler, Battista, Leatherdale et al., [<reflink idref="bib6" id="ref44">6</reflink>]) and operational definitions for these correlates are described in detail elsewhere (Butler, Battista, Leatherdale et al., [<reflink idref="bib5" id="ref45">5</reflink>]). Restaurants, convenience stores, and MMA type were not examined in this analysis due to our previous findings (Butler, Battista, Leatherdale et al., [<reflink idref="bib6" id="ref46">6</reflink>]) that suggested there was limited variability in student MMA consumption between schools. However, MMA consumption patterns across school-level factors that were observed to possibly influence consumption (breakfast program participation and staff nutrition training) were assessed.</p> <p>For the repeat cross-sectional sample, we examined whether behaviours were reported within each calendar-year of the study. For the longitudinal cohort, the patterns of these behaviours were examined over time as students progressed through grades 9–12, where students were grouped as always/mostly, never, or intermittently having the behaviour over grade. For example, meeting the physical activity guidelines, students who were classified as 'always' if they reported the behaviour for all years of participation in COMPASS, 'mostly' if they reported the behaviour at least 3 times during participation, 'intermittently' if they reported the behaviour in some years, and 'never' if they had never reported the behaviour for any years. This grouping provides the most straightforward approach to understanding the impact of changing behaviours over time.</p> <hd id="AN0159232770-7">Analyses</hd> <p>Students with missing data on variables of interest were removed from all analyses, resulting in a complete case repeat cross-sectional sample of 23,837 students in wave 1, 22,766 students in wave 2, 22,696 students in wave 3, 22,041 students in wave 4, 21,516 students in wave 5, and 20,879 students in wave 6, and a complete-case longitudinal cohort sample of 4,755 students (cohort 1: 1,733, cohort 2: 1,489, cohort 3: 1,533). Sample descriptives were examined at each time point for the repeat cross-sectional sample, as well as at baseline for each cohort in the longitudinal sample. The number of MMA servings over time was graphed for each demographic and behaviour subgroup.</p> <p>Regression models were run on both the repeat cross-sectional and longitudinal samples in order to examine trends in MMA consumption. Initial null models were run to examine the level of interclass correlation at the school and student levels as appropriate. Predictive models were run stepwise by first incorporating the main effects to examine the overall trends in MMA consumption, and then adding time interaction terms to examine the influence of demographics and health behaviours on these trends. Using a stepwise approach, we present the full model with the main effects as well as a model with significant interactions from an overall type 3 test (UCLA: Statistical Consulting Group, [<reflink idref="bib40" id="ref47">40</reflink>]).</p> <p>In the repeat cross-sectional sample, differences in MMA consumption trends by demographics and health behaviours were examined using ordinal mixed effects regression models with a random school intercept term to account for the school clustering effect. Calendar-year was used as a categorical predictor variable to examine (potentially non-linear) overall trends in MMA consumption while controlling for sample demographics and health behaviours. Interaction terms with calendar-year were then included to examine whether this trend varied for certain demographic or behavioural subgroups.</p> <p>In the longitudinal sample, differences in MMA consumption trends over time as students progressed through high school were examined by cohort as well as by demographics and health behaviours. The longitudinal data were modeled using an ordinal-mixed effects regression model with a random student intercept term to account for within-student correlation of responses (i.e. that a given student's responses over time will be related). Time was used as a categorical predictor to examine trends in MMA consumption over the course of high school. Time, in this case, corresponds to the data collection baseline/follow-up point and is analogous to students progressing through high school grades from grade 9 to grade 12. Cohort was also used as a categorical predictor to examine trends in MMA consumption for students starting high school in different calendar-years, and a cohort–time interaction was included to determine whether the time differences in MMA consumption varied by cohort. Demographic and health behaviour were controlled for as main effects, with time interactions later included to examine whether the time trend throughout high school varied for certain subgroups. All models were run using the GLIMMIX procedure in SAS 9.4 (SAS Institute, [<reflink idref="bib34" id="ref48">34</reflink>]).</p> <p>Lastly, post-hoc analyses were conducted to explore if changes in MMA consumption were evident among sub-populations or environmental contexts as per our overall grant objective of investigating factors related to MMA consumption. A latent class analysis (LCA) was conducted to attempt to identify subgroups of students with a different trajectory than the overall downward trend of MMA consumption over time. In addition, difference-in-difference models were conducted to examine whether school-level changes had a significant impact on MMA consumption changes in the most recent 2 years of study.</p> <hd id="AN0159232770-8">Results</hd> <p></p> <hd id="AN0159232770-9">Sample descriptives</hd> <p>Baseline sample descriptives for student demographics, health behaviour and MMA consumption, in both the repeat cross-sectional and the 3 longitudinal cohort ssamples are presented in Supplementary Table 1. We observed a consistent decline in mean MMA consumption over time for the repeat cross-sectional sample and the longitudinal samples. Within the repeat cross-sectional model, the majority of the sample self-identified as non-racialized and attended school in Ontario, and male and female genders were represented equally. The longitudinal sample had more females and students who self-identified as non-racialized when compared to the repeat cross-sectional sample.</p> <hd id="AN0159232770-10">Within-school and within-student correlation in MMA consumption</hd> <p>Consistent with our previous findings (Butler, Battista, Leatherdale et al., [<reflink idref="bib6" id="ref49">6</reflink>]), the ICC for the repeat cross-sectional null model was low (1.2%), suggesting that there was little variability in student MMA consumption between schools. The longitudinal model yielded an ICC of 50.1%, meaning that half of the variability in MMA consumption was due to variability between students, while half was due to within-student variability in consumption over time for a given student.</p> <hd id="AN0159232770-11">Changes in MMA consumption over time</hd> <p>A decline in MMA consumption over time was observed within both the repeat cross-sectional (Table 1) and longitudinal (Table 2) analyses across all student characteristics and health behaviours examined.</p> <p>Table 1. Ordinal random intercept regression results for repeat cross-sectional model of students from the 50 Canadian secondary schools who participated in the COMPASS study by main effects and significant interactions.</p> <p> <ephtml> <table><thead><tr><td /><td /><td>Model 1a Main effects</td><td>Model 1b <sup>Ψ</sup> Significant interactions</td></tr></thead><tbody><tr><td>Year</td><td>2013–14 (ref)</td><td><bold>OR (95% CI)</bold></td><td><bold>OR (95% CI)</bold></td></tr><tr><td /><td>2014–15</td><td>0.90 (0.88–0.93)**</td><td>0.84 (0.79–0.90)**</td></tr><tr><td /><td>2015–16</td><td>0.78 (0.76–0.81)**</td><td>0.70 (0.66–0.75)**</td></tr><tr><td /><td>2016–17</td><td>0.70 (0.67–0.72)**</td><td>0.64 (0.60–0.69)**</td></tr><tr><td /><td>2017–18</td><td>0.62 (0.60–0.64)**</td><td>0.55 (0.52–0.59)**</td></tr><tr><td /><td>2018–19</td><td>0.52 (0.50–0.54)**</td><td>0.45 (0.42–0.48)**</td></tr><tr><td>Grade</td><td>9 (ref)</td><td /><td /></tr><tr><td /><td>10</td><td>0.97 (0.94–0.99)*</td><td>0.92 (0.86–0.98)**</td></tr><tr><td /><td>11</td><td>0.92 (0.90–0.95)**</td><td>0.80 (0.75–0.85)**</td></tr><tr><td /><td>12</td><td>0.88 (0.85–0.90)**</td><td>0.76 (0.72–0.82)</td></tr><tr><td>Province</td><td>Alberta (ref)</td><td /><td /></tr><tr><td /><td>Ontario</td><td>1.01 (0.87–1.18)</td><td>1.02 (0.88–1.19)</td></tr><tr><td>Gender</td><td>Female (ref)</td><td /><td /></tr><tr><td /><td>Male</td><td>2.02 (1.98–2.06)**</td><td>2.02 (1.98–2.06)**</td></tr><tr><td>Ethnicity</td><td>Non-racialized (ref)</td><td /><td /></tr><tr><td /><td>Racialized</td><td>0.69 (0.67–0.71)**</td><td>0.69 (0.67–0.71)**</td></tr><tr><td>Weight Goals</td><td>No weight goal (ref)</td><td /><td /></tr><tr><td /><td>Lose weight</td><td>0.85 (0.83–0.88)**</td><td>0.85 (0.83–0.88)**</td></tr><tr><td /><td>Gain weight</td><td>1.32 (1.28–1.37)**</td><td>1.33 (1.28–1.37)**</td></tr><tr><td /><td>Stay the same weight</td><td>1.07 (1.04–1.10)**</td><td>1.07 (1.04–1.10)**</td></tr><tr><td>Meet Physical Activity guidelines</td><td>No (ref)</td><td /><td /></tr><tr><td /><td>Yes</td><td>1.44 (1.41–1.46)**</td><td>1.44 (1.14–1.47)**</td></tr><tr><td>Eat Breakfast every day</td><td>No (ref)</td><td /><td /></tr><tr><td /><td>Yes</td><td>1.43 (1.40–1.46)**</td><td>1.43 (1.40–1.46)**</td></tr><tr><td>Participate in Breakfast program</td><td>No (ref)</td><td /><td /></tr><tr><td /><td>Yes</td><td>1.22 (1.19–1.26)**</td><td>1.22 (1.18–1.26)**</td></tr><tr><td>Bring lunch from home</td><td>0–3 days (ref)</td><td /><td /></tr><tr><td /><td>4–5 days</td><td>1.22 (1.19–1.24)**</td><td>1.22 (1.19–1.24)**</td></tr><tr><td>Buy snacks from vending machine</td><td>0 days (ref)</td><td /><td /></tr><tr><td /><td>1–5 days</td><td>1.30 (1.26–1.33)**</td><td>1.30 (1.26–1.33)**</td></tr><tr><td><bold><italic>Significant interaction terms within repeat cross-sectional sample</italic></bold></td></tr><tr><td>Year * Grade 10</td><td>2013–14</td><td /><td>–</td></tr><tr><td /><td>2014–15</td><td>–</td><td>1.06 (0.97–1.16)</td></tr><tr><td /><td>2015–16</td><td>–</td><td>1.04 (0.95–1.14)</td></tr><tr><td /><td>2016–17</td><td>–</td><td>1.05 (0.96–1.15)</td></tr><tr><td /><td>2017–18</td><td>–</td><td>1.09 (0.99–1.19)</td></tr><tr><td /><td>2018–19</td><td>–</td><td>1.12 (1.02–1.22)*</td></tr><tr><td><bold>Year * Grade 11</bold></td><td>2013–14</td><td>–</td><td>–</td></tr><tr><td /><td>2014–15</td><td>–</td><td>1.14 (1.04–1.25)**</td></tr><tr><td /><td>2015–16</td><td>–</td><td>1.26 (1.15–1.37)**</td></tr><tr><td /><td>2016–17</td><td>–</td><td>1.12 (1.02–1.22)*</td></tr><tr><td /><td>2017–18</td><td>–</td><td>1.20 (1.10–1.32)**</td></tr><tr><td /><td>2018–19</td><td>–</td><td>1.26 (1.14–1.38)**</td></tr><tr><td><bold>Year * Grade 12</bold></td><td>2013–14</td><td>–</td><td>–</td></tr><tr><td /><td>2014–15</td><td>–</td><td>1.09 (1.00–1.20)</td></tr><tr><td /><td>2015–16</td><td>–</td><td>1.18 (1.07–1.29)**</td></tr><tr><td /><td>2016–17</td><td>–</td><td>1.18 (1.07–1.29)**</td></tr><tr><td /><td>2017–18</td><td>–</td><td>1.21 (1.10–1.33)**</td></tr><tr><td /><td>2018–19</td><td>–</td><td>1.27 (1.15–1.39)**</td></tr></tbody></table> </ephtml> </p> <p>1 * Significance at p < 0.05, ** Significance at p < 0.01,</p> <p>2 <sups>Ψ</sups>All significant time interactions were based on overall type 3 effect.</p> <p>Table 2. Linked longitudinal sample of the students from the 50 Canadian secondary schools who participated in the COMPASS study (2013–2019) demonstrating the odds of greater MMA consumption by main effects (linked by cohort and time) and significant interactions.</p> <p> <ephtml> <table><thead><tr><td>Longitudinal model</td><td /><td>Model 2a Main effects (with time and cohort)</td><td>Model 2b <sup>Ψ</sup> Significant interactions</td></tr></thead><tbody><tr><td /><td /><td><bold>OR (95% CI)</bold></td><td><bold>OR (95% CI)</bold></td></tr><tr><td>Cohort</td><td>Cohort 1 (ref)</td><td /><td /></tr><tr><td /><td>Cohort 2</td><td>0.82 (0.69–0.97)*</td><td>0.82 (0.69–0.97)*</td></tr><tr><td /><td>Cohort 3</td><td>0.62 (0.52–0.73)**</td><td>0.62 (0.52–0.73)**</td></tr><tr><td>Time</td><td>Grade 9 (Year 1) (ref)</td><td /><td /></tr><tr><td /><td>Grade 10 (Year 2)</td><td>0.68 (0.60–0.77)**</td><td>0.61 (0.53–0.70)**</td></tr><tr><td /><td>Grade 11 (Year 3)</td><td>0.50 (0.50–0.65)**</td><td>0.50 (0.43–0.57)**</td></tr><tr><td /><td>Grade 12 (Year 4)</td><td>0.42 (0.37–0.47)**</td><td>0.38 (0.33–0.44)**</td></tr><tr><td>Time*Cohort 2</td><td>Grade 9 (Year 1) (ref)</td><td /><td /></tr><tr><td /><td>Grade 10 (Year 2)</td><td>1.11 (0.92–1.33)</td><td>1.11 (0.93–1.33)</td></tr><tr><td /><td>Grade 11 (Year 3)</td><td>0.92 (0.76–1.10)</td><td>0.92 (0.77–1.10)</td></tr><tr><td /><td>Grade 12 (Year 4)</td><td>0.82 (0.69–0.99)*</td><td>0.82 (0.69–0.99)*</td></tr><tr><td>Time*Cohort 3</td><td>Grade 9 (Year 1) (ref)</td><td /><td /></tr><tr><td /><td>Grade 10 (Year 2)</td><td>1.19 (0.99–1.42)</td><td>1.19 (0.99–1.42)</td></tr><tr><td /><td>Grade 11 (Year 3)</td><td>1.10 (0.92–1.31)</td><td>1.09 (0.91–1.31)</td></tr><tr><td /><td>Grade 12 (Year 4)</td><td>1.09 (0.91–1.31)</td><td>1.09 (0.91–1.31)</td></tr><tr><td>Province</td><td>Alberta (ref)</td><td /><td /></tr><tr><td /><td>Ontario</td><td>0.94 (0.70–1.26)</td><td>0.94 (0.71–1.26)</td></tr><tr><td>Gender</td><td>Female (ref)</td><td /><td /></tr><tr><td /><td>Male</td><td>2.99 (2.26–3.36)**</td><td>2.46 (2.12–2.85)**</td></tr><tr><td>Ethnicity</td><td>Non-racialized (ref)</td><td /><td /></tr><tr><td /><td>Racialized</td><td>0.49 (0.42–0.56)**</td><td>0.49 (0.42–0.56)**</td></tr><tr><td>Weight goals</td><td>Changing Goals Over Time (ref)</td><td /><td /></tr><tr><td /><td>Lose Weight Always</td><td>0.73 (0.64–0.83)**</td><td>0.73 (0.64–0.83)**</td></tr><tr><td /><td>Gain Weight Always</td><td>1.44 (1.14–1.82)**</td><td>1.44 (1.13–1.82)**</td></tr><tr><td>Meet physical activity guidelines</td><td>Never (ref)</td><td /><td /></tr><tr><td /><td>Always</td><td>2.66 (2.22–3.19)**</td><td>2.66 (2.22–3.20)**</td></tr><tr><td /><td>Intermittent</td><td>1.49 (1.30–1.70)**</td><td>1.49 (1.30–1.70)**</td></tr><tr><td>Eat breakfast every day</td><td>Never (ref)</td><td /><td /></tr><tr><td /><td>Always</td><td>2.11 (1.81–2.47)**</td><td>2.11 (1.81–2.47)**</td></tr><tr><td /><td>Intermittent</td><td>1.51 (1.33–1.72)**</td><td>1.51 (1.33–1.72)**</td></tr><tr><td>Participate in breakfast program</td><td>Never (ref)</td><td /><td /></tr><tr><td /><td>Always/Mostly</td><td>1.60 (1.27–2.01)**</td><td>1.60 (1.27–2.01)**</td></tr><tr><td /><td>Intermittent</td><td>1.15 (1.00–1.31)*</td><td>1.15 (1.00–1.31)*</td></tr><tr><td>Bring lunch from home</td><td>Never (ref)</td><td /><td /></tr><tr><td /><td>Always</td><td>1.32 (1.12–1.56)**</td><td>1.33 (1.13–1.56)**</td></tr><tr><td /><td>Intermittent</td><td>1.24 (1.07–1.44)**</td><td>1.25 (1.08–1.44)**</td></tr><tr><td>Buy snacks from vending machine</td><td>Never (ref)</td><td /><td /></tr><tr><td /><td>Always/Mostly</td><td>1.25 (0.97–1.60)</td><td>1.25 (0.98–1.60)</td></tr><tr><td /><td>Intermittent</td><td>1.14 (1.01–1.30)*</td><td>1.15 (1.01–1.30)**</td></tr><tr><td><bold><italic>Significant interaction terms within longitudinal sample</italic></bold></td><td /><td /></tr><tr><td>Time*Male</td><td>Grade 9 (Year 1) (ref)</td><td /><td /></tr><tr><td /><td>Grade 10 (Year 2)</td><td /><td>1.30 (1.12–1.51)**</td></tr><tr><td /><td>Grade 11 (Year 3)</td><td /><td>1.35 (1.16–1.57)**</td></tr><tr><td /><td>Grade 12 (Year 4)</td><td /><td>1.24 (1.07–1.45)**</td></tr></tbody></table> </ephtml> </p> <ulist> <item>3 * Significance at p < 0.05, ** Significance at p < 0.01,</item> <item>4 <sups>Ψ</sups>All significant time interactions were based on overall type 3 effect.</item> </ulist> <hd id="AN0159232770-12">Repeat cross-sectional results</hd> <p>As shown in the main effects model (Model 1a) in Table 1, there was a pronounced downward trend in the odds of greater MMA consumption over time, after controlling for demographics and health behaviours. The odds of consuming greater MMA servings in 2018–19 were 0.45 (95%CI: 0.38–0.54), meaning that students were less than half as likely to consume a greater number of MMA servings in 2018–19 than they were in 2013–14, regardless of the other factors in the model (e.g. gender, grade). Results remained consistent even after including significant interaction terms (Model 1b). Significant interactions were observed between time and grade suggesting that, while students in higher grades have lower overall odds of consuming more MMA servings, their decrease in MMA consumption over time was less of a downwards slope than what was seen for students in grade 9. MMA consumption for each demographic and behaviour subgroup is illustrated in Supplementary Figure 1 and showed consistent downward trends, with visible differences in slope by grade.</p> <hd id="AN0159232770-13">Longitudinal results</hd> <p>As shown in the main effects model (Model 2a) in Table 2, there was a downward trend in MMA consumption over time as students progressed through high school. The odds of consuming greater MMA servings were also lower among the later calendar-year cohorts. The cohort–time interactions in Model 2a show few differences in MMA consumption trends with cohort, except for some evidence that cohort 2 had a slightly steeper decrease in consumption at time 4 than cohort 1. The model including significant interactions terms (Model 1b) showed consistent results, with gender being the only variable to have a significant time interaction. This interaction shows that in addition to males having higher overall odds of consuming a greater number of MMA servings than females (OR: 2.46, 95% CI: 2.12–2.85), they also have less of a downward slope over time (T2 OR: 1.30, 95%CI: 1.12–1.51; T3 OR: 1.35, 95%CI: 1.16–1.57; T4 OR: 1.24, 95%CI: 1.07–1.45). Time interactions with all other health behaviours were insignificant, suggesting no difference in MMA consumption trends among these different subgroups. Downward trends in the number of MMA servings across all subgroups are illustrated in Supplementary Figure 2.</p> <hd id="AN0159232770-14">Predictors of MMA consumption</hd> <p>In both the repeat cross-sectional sample (Model 1b) and longitudinal sample (Model 2b), male students had higher overall odds of greater MMA consumption and students who identified as racialized had lower overall odds of greater MMA consumption, while no significant differences were seen by the province. The magnitudes of the effect were more pronounced for the longitudinal sample (Table 2 model 2a) compared to the repeat cross-sectional sample (Table 1 model 1a), though the overall trends were similar. Additionally, the longitudinal model showed a significant time interaction with gender that was not observed in the repeat cross-sectional model.</p> <hd id="AN0159232770-15">Post-hoc subgroup analyses</hd> <p>Results from the LCA (Supplementary file B) suggested that, with the exception of one very small group (n = 64), there were no classes of students that maintained or increased their MMA consumption over time. No demographic or health behaviour variables measured in the main analysis were found to be predictive of membership in this increasing consumption class of 64 students, and therefore it is unclear whether the small group increasing their MMA consumption is practically meaningful. Results from the difference-in-difference models (Supplementary file B) showed that changes to breakfast programs or staff nutrition training at the school level during the study period did not influence student MMA consumption in the most recent 2 years of the study.</p> <hd id="AN0159232770-16">Discussion</hd> <p>Using a nationally unique surveillance data system where longitudinal samples are nested within large repeat cross-sectional samples of youth (Leatherdale et al., [<reflink idref="bib19" id="ref50">19</reflink>]), the present study highlights similarities and differences between the repeat cross-sectional and longitudinal findings when examining Canadian youth MMA consumption over time. The repeat cross-sectional findings provide insight into what MMA consumption among youth looks like at the population level, and how it changes within these samples over time. The longitudinal findings indicate how MMA consumption declines among specific groups within this sample (e.g. males) as well as within individuals over time. Overall, the repeat cross-sectional and longitudinal findings were consistent: both showed a downward trend in MMA consumption among Canadian youth over our 6-year study period. Repeat cross-sectional findings indicate that overall MMA consumption declines over time in this population, whereas longitudinal finding demonstrated an overall decline in MMA consumption across both time and cohorts. While our findings suggest that there are differences between MMA consumption over time and between cohorts, the absence of an interaction effect indicated no significant cohort effect (which refers to trends differences between groups over time (Wei et al., [<reflink idref="bib41" id="ref51">41</reflink>])) was observed in this study. Consistent with our previous cross-sectional research (Butler, Battista, Leatherdale et al., [<reflink idref="bib5" id="ref52">5</reflink>]; Butler, Battista, Leatherdale et al., [<reflink idref="bib6" id="ref53">6</reflink>]), the majority of the demographics and health behaviours examined in both the repeat cross-sectional and longitudinal analyses demonstrated significant overall associations with MMA consumption; however, there was no evidence that any behaviours altered consumption trajectories over time.</p> <p>Longitudinal models follow the same subjects over time and let us understand how an individual's MMA consumption has changed over time. These models allow the within-person variability and between-person variability to be separated. Based on the ICC (50.1%) from the longitudinal model, half of the variability in MMA consumption was due to between-student variability, while half was due to variability in MMA consumption over time for a given student (within-student variability). High within-student variability would indicate that the individual student response varies over time, and initial MMA consumption (e.g. in grade 9) is not correlated with MMA consumption at later time points (e.g. in grade 12). On the other hand, low within-student variability would indicate that MMA consumption was fairly consistent for a given student over time (i.e. if a student reported low MMA consumption at one time point, they are likely to also report low consumption at other time points). The findings of this research appear to be confounded by time, as there are moderate within-student variability and similar results between cross-sectional and longitudinal models.</p> <p>The repeat cross-sectional and longitudinal modeling approaches both provide insight into trends over time in terms of calendar-year progression and grade progression throughout high school; however, these are captured in two different, and not directly comparable, ways lending to different interpretations. Within the repeat cross-sectional model, population-level trends are measured using a snapshot of separate samples of students in each calendar-year, where each snapshot includes students in all grades of high school. Grade differences in MMA consumption over time were examined using a year-grade interaction effect to determine whether the overall calendar-year trends differed by grade. In contrast, the longitudinal model only includes a subset of students who were in grade 9 in their baseline year of participation. While students are followed over time each year, they are also progressing through grades 9 to 12; as such, the calendar-year and grade effects on MMA consumption are confounded. These effects are partially elucidated by following staggered cohorts with different calendar-year start dates. The cohort main effect highlights calendar-year trends in overall MMA consumption levels, while the cohort–time interaction highlights differences in the grade progression effect by calendar-year. These two modeling approaches to examining time and grade trends address different research questions: the repeat cross-sectional design allows us to examine overall population trends in MMA consumption, while the longitudinal design provides insight into within-person developmental trends.</p> <p>In order to advance health programing and evidence-informed policy, the best available data and selection of both appropriate methodology (Zaccai, [<reflink idref="bib42" id="ref54">42</reflink>]) and study designs (Leatherdale, [<reflink idref="bib18" id="ref55">18</reflink>]) are required. While our repeat cross-sectional and longitudinal findings produced mostly consistent results, each method provided distinct information to inform best practice and policy development. Our repeat cross-sectional findings demonstrate population-level trends, which are used to understand aggregate change among a population over time, the prevalence of the outcome and exposures of interest, and compare how subgroups (e.g. males and females) differ. In general, repeat cross-sectional results are important for public health planning and surveillance, and remain a useful and cost-effective approach to examine the health and health behaviours of the population at a single time point (Setia, [<reflink idref="bib35" id="ref56">35</reflink>]). In contrast, our longitudinal findings demonstrate individual-level changes that occur over time. Longitudinal research is well suited to examine individual trajectories across time and detect developmental transitions (i.e. adolescent populations where the timing of individual development is varied(Giedd et al., [<reflink idref="bib14" id="ref57">14</reflink>]; McCormick et al., [<reflink idref="bib25" id="ref58">25</reflink>]; Palmert & Boepple, [<reflink idref="bib28" id="ref59">28</reflink>]; Parent et al., [<reflink idref="bib29" id="ref60">29</reflink>]; Spear, [<reflink idref="bib36" id="ref61">36</reflink>])). Such results have great utility for public health professionals and can be used to provide evidence for how individual behaviours (or wellbeing) change over time as a result of an intervention, program, or policy reform.</p> <p>The behaviour under examination in this study (MMA consumption) may have been a driving factor for the similar findings within both analytic scenarios. It is possible that examining adolescent health behaviours that are typically confounded by time and demonstrate stronger within-person variability (e.g. mental health (Kessler et al., [<reflink idref="bib16" id="ref62">16</reflink>]) and substance use (Leatherdale & Burkhalter, [<reflink idref="bib20" id="ref63">20</reflink>]; Leatherdale & Rynard, [<reflink idref="bib22" id="ref64">22</reflink>]), or body mass index (Leatherdale & Rynard, [<reflink idref="bib22" id="ref65">22</reflink>])) may produce more pronounced differences across methodologies as observed in other research (Mullis et al., [<reflink idref="bib27" id="ref66">27</reflink>]; Rönnlund et al., [<reflink idref="bib33" id="ref67">33</reflink>]; Zuckermann et al., [<reflink idref="bib43" id="ref68">43</reflink>]). It is also possible that disparate results may have been observed if data on MMA consumption was collected within a younger population (Garriguet, [<reflink idref="bib13" id="ref69">13</reflink>]). Regardless of the challenges associated with collecting longitudinal data, continued funding for research that can provide evidence at both the population- and individual-levels is needed as these data offer the capacity to evaluate population level trends, as well as individual trajectories and sub-population variability.</p> <p>Our findings have several strengths and limitations. A major strength of this research is the availability of 6 years of repeat cross-sectional and linked longitudinal data derived from a large sample of youth. These robust data allow for comparisons of methodological approaches and exploration of different behaviour trends among subpopulations over time. Another benefit of this type of approach, is the ability to triangulate findings. Triangulating findings using multiple methodologies to study the same research objective provide a deeper understanding of the true phenomenon (Denzin, [<reflink idref="bib11" id="ref70">11</reflink>]) and can reduce biases that may be present when examining a topic using solely repeat cross-sectional or longitudinal methods. In a repeat cross-sectional model, it is possible that the changes in MMA are due to the sample consisting of different subjects with varying MMA consumption patterns each year. To account for this, this study only included schools that had participated in all years of the data examined for this study. Additionally, we controlled for sociodemographic factors in the repeat cross-sectional models to account for potential sample differences. However, COMPASS uses a purposive sampling method and therefore this sample is not representative of all secondary schools in Canada. Self-reported MMA consumption over time is subject to recall bias; however, this dietary measure is consistent with other school-based research surveillance platforms that require students to report on past 24-hour food and drink consumption (Storey & Mccargar, [<reflink idref="bib37" id="ref71">37</reflink>]).</p> <hd id="AN0159232770-17">Conclusion</hd> <p>Overall, the repeat cross-sectional and longitudinal findings were consistent, and this study provides a deeper examination of the utility and purpose of both methodologies. This research was used to triangulate findings using multiple methodologies to study the same research objective, and provides a deeper understanding of the true phenomenon of interest (Denzin, [<reflink idref="bib11" id="ref72">11</reflink>]). While results from both the repeat cross-sectional and longitudinal methods draw similar conclusions, both types of data and analytical methods remain necessary for public health research and evaluation. These findings highlight the importance of selecting the research design most appropriate for the research question(s) at hand, as well as the strengths and limitations between cross-sectional and longitudinal research, and their application to policy, practice, and social research methods.</p> <hd id="AN0159232770-18">Disclosure statement</hd> <p>This research was supported by Dairy Farmers of Canada (DFC). As per the research agreement, aside from providing financial support, DFC have no decision-making role in the design and conduct of the studies, data collection, and analysis or interpretation of the data. Researchers maintain independence in conducting their studies, own their data, and report the outcomes regardless of the results. The decision to publish the findings rests solely with the researchers, as does the choice of journal(s) to which the papers are submitted (although DFC encourages publication in Canadian journals or those with a Canadian readership). As per the conditions of the grant, the researchers provided DFC with a copy of this manuscript before its submission; DFC did not provide any comments. SEM has served as a paid Expert on behalf of the Attorney General of Canada in legal proceedings, providing evidence on the public health risks and benefits of unpasteurized milk.</p> <hd id="AN0159232770-19">Supplementary material</hd> <p>Supplemental data for this article can be accessed https://doi.org/10.1080/13645579.2021.1922804.</p> <ref id="AN0159232770-20"> <title> References </title> <blist> <bibl id="bib1" idref="ref39" type="bt">1</bibl> <bibtext> Battista, K., Qian, W., Bredin, C., Leatherdale, S.T. Student Data Linkage over Multiple Years. COMPASS Technical Report Series. (2019); 6(3): Waterloo, Ontario: University of Waterloo. [Internet]. Available from: https://uwaterloo.ca/compass-system/student-data-linkage-over-multiple-years</bibtext> </blist> <blist> <bibl id="bib2" idref="ref5" type="bt">2</bibl> <bibtext> Berger, MPF. (1986). A comparison of efficiencies of longitudinal, mixed longitudinal, and cross-sectional designs. Journal of Educational Statistics, 11 (3), 171 – 181. https://doi.org/10.3102/10769986011003171</bibtext> </blist> <blist> <bibl id="bib3" idref="ref40" type="bt">3</bibl> <bibtext> Bredin, C., & Leatherdale, S. T. Methods for linking COMPASS student-level data over time. COMPASS Technical Report Series. 2013;1(2). Waterloo, Ontario: University of Waterloo. [Internet]. Available from: https://uwaterloo.ca/compass-system/publications/methods-linking-compass-student-level-data-over-time</bibtext> </blist> <blist> <bibl id="bib4" idref="ref1" type="bt">4</bibl> <bibtext> Brownson, R. C., Chriqui, J. F., & Stamatakis, K. A. (2009). Understanding evidence-based public health policy. American Journal of Public Health, 99 (9), 1576 – 1583. https://doi.org/10.2105/AJPH.2008.156224</bibtext> </blist> <blist> <bibl id="bib5" idref="ref33" type="bt">5</bibl> <bibtext> Butler, A. E, Battista, K., Leatherdale, S.T., Elliott, S., Meyers, S., & Majowicz, S. E. Correlates of milk and milk alternative consumption among Canadian secondary school students, circa 2017/2018. COMPASS Technical Report Series. (2020);7(1): Waterloo, Ontario: University of Waterloo. [ Internet ]. Available from : https://uwaterloo.ca/compass-system/publications/correlates-milk-and-milk-alternative-consumption-among</bibtext> </blist> <blist> <bibl id="bib6" idref="ref34" type="bt">6</bibl> <bibtext> Butler, A. E, Battista, K., Leatherdale, S. T., Meyer, S., Elliott, S. J., & Majowicz, S. E. (2020). Environmental factors of youth milk and milk alternative consumption. American Journal of Health Behavior, 44 (5), 666 – 680. https://doi.org/10.5993/AJHB.44.5.10</bibtext> </blist> <blist> <bibl id="bib7" idref="ref7" type="bt">7</bibl> <bibtext> Caruana, E. J., Roman, M., Hernández-Sánchez, J., & Solli, P. (2015). Longitudinal studies. Journal of Thoracic Disease, 7 (11), E537 – 40. https://doi.org/10.3978/j.2072-1439.2015.10.63</bibtext> </blist> <blist> <bibl id="bib8" idref="ref16" type="bt">8</bibl> <bibtext> Cook, N. R., & Ware, J. H. (1983). Design and analysis methods for longitudinal research. Annual Review of Public Health, 4 (1), 1 – 23. https://doi.org/10.1146/annurev.pu.04.050183.000245</bibtext> </blist> <blist> <bibl id="bib9" idref="ref11" type="bt">9</bibl> <bibtext> Costa, P. T., McCrae, R. R., & Löckenhoff, C. E. (2019). Personality across the life span. Annual Review of Psychology, 70 (1), 423 – 448. https://doi.org/10.1146/annurev-psych-010418-103244</bibtext> </blist> <blist> <bibtext> Darling, N. (2005). Participation in extracurricular activities and adolescent adjustment: Cross-sectional and longitudinal findings. Journal of Youth and Adolescence, 34 (5), 493 – 505. https://doi.org/10.1007/s10964-005-7266-8</bibtext> </blist> <blist> <bibtext> Denzin, N. K. (2012). Triangulation 2.0*. Journal of Mixed Methods Research, 6 (2), 80 – 88. https://doi.org/10.1177/1558689812437186</bibtext> </blist> <blist> <bibtext> Dettori, J. (2011). Loss to follow-up. Evidence-Based Spine-Care Journal, 2 (1), 7 – 10. https://doi.org/10.1055/s-0030-1267080</bibtext> </blist> <blist> <bibtext> Garriguet D. (2008). Beverage consumption of children and teens. Health reports, 19(4), 17–22. Available from: https://www150.statcan.gc.ca/n1/en/pub/82-003-x/2008004/article/6500820-eng.pdf?st=52yhAMtS</bibtext> </blist> <blist> <bibtext> Giedd, J. N., Blumenthal, J., Jeffries, N. O., Castellanos, F. X., Liu, H., Zijdenbos, A., Paus, T., Evans, A. C., & Rapoport, J. L. (1999). Brain development during childhood and adolescence: A longitudinal MRI study [2]. Nature Neuroscience, 2 (10), 861 – 863. https://doi.org/10.1038/13158</bibtext> </blist> <blist> <bibtext> Health Canada. Canada's dietary guidelines [Internet]. 2019. Available from : https://food-guide.canada.ca/static/assets/pdf/CDG-EN-2018.pdf</bibtext> </blist> <blist> <bibtext> 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. https://doi.org/10.1001/archpsyc.62.6.593</bibtext> </blist> <blist> <bibtext> Kraemer, H. C., Yesavage, J. A., Taylor, J. L., & Kupfer, D. (2000). How can we learn about developmental processes from cross-sectional studies, or can we? American Journal of Psychiatry, 157 (2), 163 – 171. https://doi.org/10.1176/appi.ajp.157.2.163</bibtext> </blist> <blist> <bibtext> Leatherdale, S. T. (2019). Natural experiment methodology for research: A review of how different methods can support real-world research. International Journal of Social Research Methodology, 22 (1), 19 – 35. https://doi.org/10.1080/13645579.2018.1488449</bibtext> </blist> <blist> <bibtext> Leatherdale, S. T., Brown, K. S., Carson, V., Childs, R. A., Dubin, J. A., Elliott, S. J., Faulkner, G., Hammond, D., Manske, S., Sabiston, C. M., Laxer, R. E., Bredin, C., & Thompson-Haile, A. (2014). The COMPASS study: A longitudinal hierarchical research platform for evaluating natural experiments related to changes in school-level programs, policies and built environment resources. BMC Public Health, 14 (1): 331. https://doi.org/10.1186/1471-2458-14-331</bibtext> </blist> <blist> <bibtext> Leatherdale, S. T., & Burkhalter, R. (2012). The substance use profile of Canadian youth: Exploring the prevalence of alcohol, drug and tobacco use by gender and grade. Addictive Behaviors, 37 (3), 318 – 322. https://doi.org/10.1016/j.addbeh.2011.10.007</bibtext> </blist> <blist> <bibtext> Leatherdale, S. T., & Laxer, R. E. (2013). Reliability and validity of the weight status and dietary intake measures in the COMPASS questionnaire: Are the self-reported measures of body mass index (BMI) and Canada's food guide servings robust? International Journal of Behavioral Nutrition and Physical Activity, 10 (1), 42. https://doi.org/10.1186/1479-5868-10-42</bibtext> </blist> <blist> <bibtext> Leatherdale, S. T., & Rynard, V. (2013). A cross-sectional examination of modifiable risk factors for chronic disease among a nationally representative sample of youth: Are Canadian students graduating high school with a failing grade for health? BMC Public Health, 13 (1):569. https://doi.org/10.1186/1471-2458-13-569</bibtext> </blist> <blist> <bibtext> Louis, T. A., Robins, J., Dockery, D. W., Spiro, A., & Ware, J. H. (1986). Explaining discrepancies between longitudinal and cross-sectional models. Journal of Chronic Diseases, 39 (10), 831 – 839. https://doi.org/10.1016/0021-9681(86)90085-8</bibtext> </blist> <blist> <bibtext> Maxwell, S. E., & Cole, D. A. (2007). Bias in cross-sectional analyses of longitudinal mediation. Psychological Methods, 12 (1), 23 – 44. https://doi.org/10.1037/1082-989X.12.1.23</bibtext> </blist> <blist> <bibtext> McCormick, E. M., Qu, Y., & Telzer, E. H. (2017). Activation in context: Differential conclusions drawn from cross-sectional and longitudinal analyses of adolescents' cognitive control-related neural activity. Frontiers in Human Neuroscience, 11: 141. https://doi.org/10.3389/fnhum.2017.00141</bibtext> </blist> <blist> <bibtext> Mueller, S., Wagner, J., Drewelies, J., Duezel, S., Eibich, P., Specht, J., Demuth, I., Steinhagen-Thiessen, E., Wagner, G. G., & Gerstorf, D. (2016). Personality development in old age relates to physical health and cognitive performance: Evidence from the Berlin aging study II. Journal of Research in Personality, 65, 94 – 108. https://doi.org/10.1016/j.jrp.2016.08.007</bibtext> </blist> <blist> <bibtext> Mullis, A. K., Mullis, R. L., & Normandin, D. (1992). Cross-sectional and longitudinal comparisons of adolescent self-esteem. Adolescence, 27 (105), 51 – 61. https://search.proquest.com/docview/195933001?pq-origsite=gscholar&fromopenview=true</bibtext> </blist> <blist> <bibtext> Palmert, M. R., & Boepple, P. A. (2001). Commentary: Variation in the timing of puberty: Clinical spectrum and genetic investigation. Journal of Clinical Endocrinology and Metabolism, 86 (6), 2364 – 2368. https://doi.org/10.1210/jcem.86.6.7603</bibtext> </blist> <blist> <bibtext> Parent, A. S., Teilmann, G., Juul, A., Skakkebaek, N. E., Toppari, J., & Bourguignon, J. P. (2003). The timing of normal puberty and the age limits of sexual precocity: Variations around the world, secular trends, and changes after migration. Endocrine Reviews, 24 (5), 668 – 693. https://doi.org/10.1210/er.2002-0019</bibtext> </blist> <blist> <bibtext> Pierre, M., St. Changes in Canadians' preferences for milk and dairy products. 2017. Minister of Industry. [Internet]. Available from : https://www150.statcan.gc.ca/n1/pub/21-004-x/2017001/article/14786-eng.htm</bibtext> </blist> <blist> <bibtext> Qian, W., Battista, K., Bredin, C., Brown, K., & Leatherdale, S.T. Assessing longitudinal data linkage results in the COMPASS study. COMPASS Technical Report Series. 2015; 3(4). Waterloo, Ontario: University of Waterloo. [Internet]. Available from : https://uwaterloo.ca/compass-system/publications/assessing-longitudinal-data-linkage-results-compass-study</bibtext> </blist> <blist> <bibtext> Ribeiro, J. D., Pease, J. L., Gutierrez, P. M., Silva, C., Bernert, R. A., Rudd, M. D., & Joiner, T. E. (2012). Sleep problems outperform depression and hopelessness as cross-sectional and longitudinal predictors of suicidal ideation and behavior in young adults in the military. Journal of Affective Disorders, 136 (3), 743 – 750. https://doi.org/10.1016/j.jad.2011.09.049</bibtext> </blist> <blist> <bibtext> Rönnlund, M., Nyberg, L., Bäckman, L., & Nilsson, L. G. (2005). Stability, growth, and decline in adult life span development of declarative memory: Cross-sectional and longitudinal data from a population-based study. Psychology and Aging, 20 (1), 3 – 18. https://doi.org/10.1037/0882-7974.20.1.3</bibtext> </blist> <blist> <bibtext> SAS Institute. (2012). SAS version 9.4. SAS Institute Inc.</bibtext> </blist> <blist> <bibtext> Setia, M. S. (2016). Methodology series module 3: Cross-sectional studies. Indian Journal of Dermatology, 61(3), 261–264. https://doi.org/10.4103/0019-5154.182410</bibtext> </blist> <blist> <bibtext> Spear, L. P. (2000). The adolescent brain and age-related behavioral manifestations. Neuroscience and Biobehavioral Reviews, 24(4),417–463. doi https://doi.org/10.1016/s0149-7634(00)00014-2</bibtext> </blist> <blist> <bibtext> Storey, K. E., & Mccargar, L. J. (2012). Reliability and validity of Web-SPAN, a web-based method for assessing weight status, diet and physical activity in youth. Journal of Human Nutrition and Dietetics, 25 (1), 59 – 68. https://doi.org/10.1111/j.1365-277X.2011.01181.x</bibtext> </blist> <blist> <bibtext> Terracciano, A., McCrae, R. R., Brant, L. J., & Costa, P. T. (2005). Hierarchical linear modeling analyses of the NEO-PI-R scales in the Baltimore longitudinal study of aging. Psychology and Aging, 20 (3), 493 – 506. https://doi.org/10.1037/0882-7974.20.3.493</bibtext> </blist> <blist> <bibtext> Tudisca, V., Valente, A., Castellani, T., Stahl, T., Sandu, P., Dulf, D., Spitters, H., Van de Goor, I., Radl-Karimi, C., Syed, M.A. and Loncarevic, N. (2018). Development of measurable indicators to enhance public health evidence-informed policy-making. Health Research Policy System, 16 (1), 1 – 13.</bibtext> </blist> <blist> <bibtext> UCLA: Statistical Consulting Group. Analyzing and visualizing interactions in SAS [Internet]. 2020 [ cited 2020 Aug 25 ]. Available from : https://stats.idre.ucla.edu/sas/seminars/analyzing-and-visualizing-interactions/</bibtext> </blist> <blist> <bibtext> Wei, X., Zang, Y., Jia, X., He, X., Zou, S., Wang, H., Shen, M., & Zang, J. (2017). Age, period and cohort effects and the predictors of physical activity and sedentary behaviour among Chinese children, from 2004 to 2011. BMC Public Health, 17 (1):353. https://doi.org/10.1186/s12889-017-4215-x</bibtext> </blist> <blist> <bibtext> Zaccai, J. H. (2004). How to assess epidemiological studies. Postgraduate Medical Journal, 80 (941), 140 – 147. https://doi.org/10.1136/pgmj.2003.012633</bibtext> </blist> <blist> <bibtext> Zuckermann, A. M. E., Battista, K., Belanger, R., Haddad, S., Butler, A., Costello, M. J. E., Leatherdale, S. T. (2021). Trends in youth cannabis use across cannabis legalization: Data from the COMPASS prospective cohort study. Preventive Medicine Reports, 22, 101351. https://doi.org/10.1016/j.pmedr.2021.101351</bibtext> </blist> </ref> <aug> <p>By Alexandra E. Butler; Kate Battista; Scott T. Leatherdale; Samantha B. Meyer; Susan J. Elliott and Shannon E. Majowicz</p> <p>Reported by Author; Author; Author; Author; Author; Author</p> <p></p> <p>Alexandra E. Butler is a PhD student in the School of Public Health and Health Systems at the University of Waterloo. She is currently a Project Manager for the COMPASS system where she is involved in the robust generation of evidence advancing Canadian youth health and wellbeing. Her research is focused on primary prevention and population health evaluation across various behavioural and environmental domains.</p> <p>Kate Battista is a PhD candidate in the School of Public Health and Health Systems at the University of Waterloo. She holds a Master of Mathematics degree in biostatistics from the University of Waterloo. She is employed as the Program Manager on the COMPASS project, where she oversees dataset management and data quality for a hierarchical prospective cohort study following 75,000+ Canadian youth. Her primary research interest is in using machine learning methods to evaluate policy and program changes within large datasets.</p> <p>Professor Scott T. Leatherdale is a behavioural scientist whose research focuses on advancing a systems science approach to primary prevention activities, evaluating complex population-level health interventions across multiple risk factor domains focused on youth, and creating research infrastructure to facilitate large population-based learning systems in chronic disease prevention.</p> <p>Associate Professor Samantha B. Meyer is an applied social scientist whose research focuses on understanding social and structural factors that influence health behaviour. Ultimately her research is used to inform health and risk communication strategies to improve individual and population health.</p> <p>Professor Susan J. Elliott is a medical geographer with expertise in global environment and health specializing in food, water and vulnerable populations.</p> <p>Associate Professor Shannon E. Majowicz is an infectious disease epidemiologist whose research focuses on food-related diseases in Canadian and international contexts, including translating findings to enhance population and public health.</p> </aug> <nolink nlid="nl1" bibid="bib39" firstref="ref2"></nolink> <nolink nlid="nl2" bibid="bib23" firstref="ref3"></nolink> <nolink nlid="nl3" bibid="bib35" firstref="ref6"></nolink> <nolink nlid="nl4" bibid="bib12" firstref="ref9"></nolink> <nolink nlid="nl5" bibid="bib10" firstref="ref12"></nolink> <nolink nlid="nl6" bibid="bib38" firstref="ref13"></nolink> <nolink nlid="nl7" bibid="bib25" firstref="ref14"></nolink> <nolink nlid="nl8" bibid="bib26" firstref="ref15"></nolink> <nolink nlid="nl9" bibid="bib27" firstref="ref18"></nolink> <nolink nlid="nl10" bibid="bib32" firstref="ref19"></nolink> <nolink nlid="nl11" bibid="bib33" firstref="ref20"></nolink> <nolink nlid="nl12" bibid="bib43" firstref="ref21"></nolink> <nolink nlid="nl13" bibid="bib14" firstref="ref25"></nolink> <nolink nlid="nl14" bibid="bib28" firstref="ref27"></nolink> <nolink nlid="nl15" bibid="bib29" firstref="ref28"></nolink> <nolink nlid="nl16" bibid="bib36" firstref="ref29"></nolink> <nolink nlid="nl17" bibid="bib17" firstref="ref30"></nolink> <nolink nlid="nl18" bibid="bib24" firstref="ref31"></nolink> <nolink nlid="nl19" bibid="bib15" firstref="ref35"></nolink> <nolink nlid="nl20" bibid="bib30" firstref="ref36"></nolink> <nolink nlid="nl21" bibid="bib19" firstref="ref37"></nolink> <nolink nlid="nl22" bibid="bib31" firstref="ref38"></nolink> <nolink nlid="nl23" bibid="bib21" firstref="ref42"></nolink> <nolink nlid="nl24" bibid="bib40" firstref="ref47"></nolink> <nolink nlid="nl25" bibid="bib34" firstref="ref48"></nolink> <nolink nlid="nl26" bibid="bib41" firstref="ref51"></nolink> <nolink nlid="nl27" bibid="bib42" firstref="ref54"></nolink> <nolink nlid="nl28" bibid="bib18" firstref="ref55"></nolink> <nolink nlid="nl29" bibid="bib16" firstref="ref62"></nolink> <nolink nlid="nl30" bibid="bib20" firstref="ref63"></nolink> <nolink nlid="nl31" bibid="bib22" firstref="ref64"></nolink> <nolink nlid="nl32" bibid="bib13" firstref="ref69"></nolink> <nolink nlid="nl33" bibid="bib11" firstref="ref70"></nolink> <nolink nlid="nl34" bibid="bib37" firstref="ref71"></nolink>
Header DbId: eric
DbLabel: ERIC
An: EJ1367264
AccessLevel: 3
PubType: Academic Journal
PubTypeId: academicJournal
PreciseRelevancyScore: 0
IllustrationInfo
Items – Name: Title
  Label: Title
  Group: Ti
  Data: A Comparison of Repeat Cross-Sectional and Longitudinal Results from the COMPASS Study: Design Considerations for Analysing Surveillance Data over Time
– Name: Language
  Label: Language
  Group: Lang
  Data: English
– Name: Author
  Label: Authors
  Group: Au
  Data: <searchLink fieldCode="AR" term="%22Butler%2C+Alexandra+E%2E%22">Butler, Alexandra E.</searchLink><br /><searchLink fieldCode="AR" term="%22Battista%2C+Kate%22">Battista, Kate</searchLink><br /><searchLink fieldCode="AR" term="%22Leatherdale%2C+Scott+T%2E%22">Leatherdale, Scott T.</searchLink><br /><searchLink fieldCode="AR" term="%22Meyer%2C+Samantha+B%2E%22">Meyer, Samantha B.</searchLink><br /><searchLink fieldCode="AR" term="%22Elliott%2C+Susan+J%2E%22">Elliott, Susan J.</searchLink><br /><searchLink fieldCode="AR" term="%22Majowicz%2C+Shannon+E%2E%22">Majowicz, Shannon E.</searchLink>
– Name: TitleSource
  Label: Source
  Group: Src
  Data: <searchLink fieldCode="SO" term="%22International+Journal+of+Social+Research+Methodology%22"><i>International Journal of Social Research Methodology</i></searchLink>. 2022 25(5):597-609.
– Name: Avail
  Label: Availability
  Group: Avail
  Data: Routledge. Available from: Taylor & Francis, Ltd. 530 Walnut Street Suite 850, Philadelphia, PA 19106. Tel: 800-354-1420; Tel: 215-625-8900; Fax: 215-207-0050; Web site: http://www.tandf.co.uk/journals
– Name: PeerReviewed
  Label: Peer Reviewed
  Group: SrcInfo
  Data: Y
– Name: Pages
  Label: Page Count
  Group: Src
  Data: 13
– Name: DatePubCY
  Label: Publication Date
  Group: Date
  Data: 2022
– Name: TypeDocument
  Label: Document Type
  Group: TypDoc
  Data: Journal Articles<br />Reports - Research
– Name: Subject
  Label: Descriptors
  Group: Su
  Data: <searchLink fieldCode="DE" term="%22Research+Design%22">Research Design</searchLink><br /><searchLink fieldCode="DE" term="%22Longitudinal+Studies%22">Longitudinal Studies</searchLink><br /><searchLink fieldCode="DE" term="%22Data+Analysis%22">Data Analysis</searchLink><br /><searchLink fieldCode="DE" term="%22Case+Studies%22">Case Studies</searchLink><br /><searchLink fieldCode="DE" term="%22Social+Science+Research%22">Social Science Research</searchLink><br /><searchLink fieldCode="DE" term="%22Public+Health%22">Public Health</searchLink><br /><searchLink fieldCode="DE" term="%22Public+Policy%22">Public Policy</searchLink><br /><searchLink fieldCode="DE" term="%22Foreign+Countries%22">Foreign Countries</searchLink><br /><searchLink fieldCode="DE" term="%22Youth%22">Youth</searchLink><br /><searchLink fieldCode="DE" term="%22Eating+Habits%22">Eating Habits</searchLink>
– Name: Subject
  Label: Geographic Terms
  Group: Su
  Data: <searchLink fieldCode="DE" term="%22Canada%22">Canada</searchLink>
– Name: DOI
  Label: DOI
  Group: ID
  Data: 10.1080/13645579.2021.1922804
– Name: ISSN
  Label: ISSN
  Group: ISSN
  Data: 1364-5579<br />1464-5300
– Name: Abstract
  Label: Abstract
  Group: Ab
  Data: Selection of appropriate study design and analytical methods is critical for producing robust research, as the design and analytical approach used can ultimately shape results and their interpretation. The objective of this research was to examine how findings from a large repeat cross-sectional data system compare to those from a sample of longitudinal data nested within the same data system, using milk and milk alternative consumption as a case study. Overall, the repeat cross-sectional and longitudinal findings were consistent, and this article highlights the strengths and limitations of both methodologies under examination. Application to social research and public health and policy are discussed.
– Name: AbstractInfo
  Label: Abstractor
  Group: Ab
  Data: As Provided
– Name: DateEntry
  Label: Entry Date
  Group: Date
  Data: 2023
– Name: AN
  Label: Accession Number
  Group: ID
  Data: EJ1367264
PLink https://search.ebscohost.com/login.aspx?direct=true&site=eds-live&db=eric&AN=EJ1367264
RecordInfo BibRecord:
  BibEntity:
    Identifiers:
      – Type: doi
        Value: 10.1080/13645579.2021.1922804
    Languages:
      – Text: English
    PhysicalDescription:
      Pagination:
        PageCount: 13
        StartPage: 597
    Subjects:
      – SubjectFull: Research Design
        Type: general
      – SubjectFull: Longitudinal Studies
        Type: general
      – SubjectFull: Data Analysis
        Type: general
      – SubjectFull: Case Studies
        Type: general
      – SubjectFull: Social Science Research
        Type: general
      – SubjectFull: Public Health
        Type: general
      – SubjectFull: Public Policy
        Type: general
      – SubjectFull: Foreign Countries
        Type: general
      – SubjectFull: Youth
        Type: general
      – SubjectFull: Eating Habits
        Type: general
      – SubjectFull: Canada
        Type: general
    Titles:
      – TitleFull: A Comparison of Repeat Cross-Sectional and Longitudinal Results from the COMPASS Study: Design Considerations for Analysing Surveillance Data over Time
        Type: main
  BibRelationships:
    HasContributorRelationships:
      – PersonEntity:
          Name:
            NameFull: Butler, Alexandra E.
      – PersonEntity:
          Name:
            NameFull: Battista, Kate
      – PersonEntity:
          Name:
            NameFull: Leatherdale, Scott T.
      – PersonEntity:
          Name:
            NameFull: Meyer, Samantha B.
      – PersonEntity:
          Name:
            NameFull: Elliott, Susan J.
      – PersonEntity:
          Name:
            NameFull: Majowicz, Shannon E.
    IsPartOfRelationships:
      – BibEntity:
          Dates:
            – D: 01
              M: 01
              Type: published
              Y: 2022
          Identifiers:
            – Type: issn-print
              Value: 1364-5579
            – Type: issn-electronic
              Value: 1464-5300
          Numbering:
            – Type: volume
              Value: 25
            – Type: issue
              Value: 5
          Titles:
            – TitleFull: International Journal of Social Research Methodology
              Type: main
ResultId 1