View in EDS HTML Full Text PDF Full Text

Who Benefits from School-Based Teen Pregnancy Prevention Programs? Examining Multidimensional Moderators of Program Effectiveness across Four Studies

Saved in:
Bibliographic Details
Title: Who Benefits from School-Based Teen Pregnancy Prevention Programs? Examining Multidimensional Moderators of Program Effectiveness across Four Studies
Language: English
Authors: Vasilenko, Sara A. (ORCID 0000-0002-1773-8947), Odejimi, Omolola A., Glassman, Jill R., Potter, Susan C., Drake, Pamela M., Coyle, Karin K., Markham, Christine, Emery, Susan Tortolero, Peskin, Melissa F., Shegog, Ross, Addy, Robert C., Clark, Leslie F.
Source: Prevention Science. 2023 24(8):1535-1546.
Availability: Springer. Available from: Springer Nature. One New York Plaza, Suite 4600, New York, NY 10004. Tel: 800-777-4643; Tel: 212-460-1500; Fax: 212-460-1700; e-mail: customerservice@springernature.com; Web site: https://link.springer.com/
Peer Reviewed: Y
Page Count: 12
Publication Date: 2023
Sponsoring Agency: Office of Public Health and Science (DHHS), Office of Population Affairs
Contract Number: 5PHEPA0000030200
Document Type: Journal Articles
Reports - Research
Education Level: Junior High Schools
Middle Schools
Secondary Education
Descriptors: Program Effectiveness, Pregnancy, Prevention, Middle School Students, Early Adolescents, Sexuality, Student Characteristics
DOI: 10.1007/s11121-022-01423-y
ISSN: 1389-4986
1573-6695
Abstract: Recent research has suggested the importance of understanding for whom programs are most effective (Supplee et al., 2013) and that multidimensional profiles of risk and protective factors may moderate the effectiveness of programs (Lanza & Rhoades, 2012). For school-based prevention programs, moderators of program effectiveness may occur at both the individual and school levels. However, due to the relatively small number of schools in most individual trials, integrative data analysis across multiple studies may be necessary to fully understand the multidimensional individual and school factors that may influence program effectiveness. In this study, we applied multilevel latent class analysis to integrated data across four studies of a middle school pregnancy prevention program to examine moderators of program effectiveness on initiation of vaginal sex. Findings suggest that the program may be particularly effective for schools with USA-born students who speak another language at home. In addition, findings suggest potential positive outcomes of the program for individuals who are lower risk and engaging in normative dating or individuals with family risk. Findings suggest potential mechanisms by which teen pregnancy prevention programs may be effective.
Abstractor: As Provided
Entry Date: 2023
Accession Number: EJ1401542
Database: ERIC
Full text is not displayed to guests.
FullText Links:
  – Type: pdflink
    Url: https://content.ebscohost.com/cds/retrieve?content=AQICAHj0k_4E0hTGH8RJwT4gCJyBsGNe_WN95AvKlDbXJGqwxwFxiOB_CdyXvoAEW8olYINuAAAA4TCB3gYJKoZIhvcNAQcGoIHQMIHNAgEAMIHHBgkqhkiG9w0BBwEwHgYJYIZIAWUDBAEuMBEEDAEChcQvkt885hEJIgIBEICBmcTRYcap-lg8cfFOS6a1BpWAw-zhcWvEBtl0mBsKTCMboGsOLNx8Kl8EVKKeCs7KcNnQcOl9iusj7kgLXkyl_KssroK-epMVxdsJG90SKHKGpMt4pBiwV-bxGJv64s2bJi-wP8TSR4d7-_nZGeft1Ucwo-sbzkoLqyPkHRAP-X5M0cL6ygGZwKXU0WvOHD4ITi5o9L9UluAwtw==
Text:
  Availability: 1
  Value: <anid>AN0173822458;n9p01nov.23;2023Nov28.05:13;v2.2.500</anid> <title id="AN0173822458-1">Who Benefits from School-Based Teen Pregnancy Prevention Programs? Examining Multidimensional Moderators of Program Effectiveness Across Four Studies </title> <p>Recent research has suggested the importance of understanding for whom programs are most effective (Supplee et al., 2013) and that multidimensional profiles of risk and protective factors may moderate the effectiveness of programs (Lanza & Rhoades, 2012). For school-based prevention programs, moderators of program effectiveness may occur at both the individual and school levels. However, due to the relatively small number of schools in most individual trials, integrative data analysis across multiple studies may be necessary to fully understand the multidimensional individual and school factors that may influence program effectiveness. In this study, we applied multilevel latent class analysis to integrated data across four studies of a middle school pregnancy prevention program to examine moderators of program effectiveness on initiation of vaginal sex. Findings suggest that the program may be particularly effective for schools with USA-born students who speak another language at home. In addition, findings suggest potential positive outcomes of the program for individuals who are lower risk and engaging in normative dating or individuals with family risk. Findings suggest potential mechanisms by which teen pregnancy prevention programs may be effective.</p> <p>Keywords: Integrative data analysis; Latent class analysis; Teen pregnancy prevention programs; Sexual initiation</p> <p>Supplementary Information The online version contains supplementary material available at https://doi.org/10.1007/s11121-022-01423-y.</p> <p>Recent prevention science research has placed more attention on understanding for whom programs are most effective by examining moderators of program effectiveness due to the necessity of moving beyond a "one-size fits all" approach. Although primary analyses of prevention programs tend to focus on whether there was a significant effect of the intervention for the entire sample, subgroup analyses also play an important role in analyzing program effects. Subgroup analyses are used to explore potential reasons for lack of effects, as well as to determine whether an intervention worked more strongly for a subset of individuals (Supplee et al., [<reflink idref="bib23" id="ref1">23</reflink>]). The information from subgroup analyses can assist researchers in refining the underlying theory and/or design of interventions to ensure that these interventions can be optimally targeted or tailored for priority populations (Rothman, [<reflink idref="bib21" id="ref2">21</reflink>]; Rothwell, [<reflink idref="bib22" id="ref3">22</reflink>]). These types of analyses typically focus on multiple analyses of single items as moderators (Bloom & Michalopoulos, [<reflink idref="bib3" id="ref4">3</reflink>]). However, when multiple potential moderators are studied, there is a risk of a type I error due to multiple comparisons (Lanza & Rhoades, [<reflink idref="bib12" id="ref5">12</reflink>]). This issue could be addressed by including complex interaction terms; however, such analyses may be underpowered and may not most efficiently represent the groups of individuals that are prevalent in the population (Lanza & Rhoades, [<reflink idref="bib12" id="ref6">12</reflink>]). Thus, traditional approaches to subgroup analysis may make it difficult to fully understand the complex interplay of factors that moderate program effectiveness.</p> <p>One recent approach to addressing the complex, multidimensional nature of program effect modifiers is person-centered methods like LCA. LCA allows researchers to uncover profiles (latent classes) of individuals characterized by the interplay of multiple factors. This approach allows researchers to examine how multiple risk and protective factors cluster together to define distinct subject profiles within a sample (Vasilenko et al., [<reflink idref="bib25" id="ref7">25</reflink>]). In addition, it can serve as a form of dimension reduction in which the most relevant combinations of factors are examined, rather than all possible combinations of factors, leading to greater statistical power and the identification of clear, distinct subgroups (Lanza & Rhoades, [<reflink idref="bib12" id="ref8">12</reflink>]).</p> <p>One study by our research team applied LCA to explore more complex effect moderation using data from the replication evaluation of the middle-school teen pregnancy prevention (TPP) program, It's Your Game...Keep it Real (IYG) (Vasilenko et al., [<reflink idref="bib25" id="ref9">25</reflink>]). This study uncovered three distinct classes of youth defined by various combinations of levels of risk and protective factors for sexual behavior: Family Instability, Other Language Household, and Frequent Religious Attendance. Subsequent analyses found only a significant effect of the IYG program on reducing the prevalence of sexual initiation for the Family Instability class. This finding suggested a possible mechanism through which the program worked: providing key supports for those who were missing them due to disruptions to family life. It also suggested a novel way to evaluate program effectiveness: deciding a priori that the primary criterion for effectiveness might be better focused on particular high-risk subgroups for whom the program is most needed.</p> <p>While this study was one of the first to suggest the utility of using a latent class moderator to examine more nuanced subgroups of adolescents for whom a TPP program may be most effective, there are a number of issues that make it difficult to implement more broadly. First, larger sample sizes are needed to uncover a larger number of latent classes. Power calculations for LCA are complex, though recent research has provided some guidelines (Dziak et al., [<reflink idref="bib6" id="ref10">6</reflink>]). For example, a sample size of over 13,000 may be needed to have 80% power to detect a four-class model when five dichotomous indicators are used (Dziak et al., [<reflink idref="bib6" id="ref11">6</reflink>]). Although adding additional items or response options may improve the ability to uncover a larger number of classes in smaller samples, larger samples are generally needed to detect a larger number of latent classes, which means that potentially important subgroups may be missed in studies with samples typical of those used in TPP programs.</p> <p>In addition, standard LCA methods do not consider that individual-level factors are not independent and may cluster within higher-level units, such as schools. However, an extension of LCA, multilevel latent class analysis (MLCA), has been proposed to deal with the issue of within-school clustering of indicator variables (Henry & Muthen, [<reflink idref="bib9" id="ref12">9</reflink>]; Vermunt, [<reflink idref="bib26" id="ref13">26</reflink>]). MLCA allows researchers to account for these dependencies and determine how patterns of characteristics at the higher level (e.g., schools) may influence individuals' behavior (Henry & Muthen, [<reflink idref="bib9" id="ref14">9</reflink>]). Thus, an analysis using this approach on data from a prevention program assessing individuals nested within schools could uncover classes of characteristics at the individual level, as well as classes of schools based on probabilities of students within these schools reporting these same individual-level characteristics. This would allow researchers to examine the types of individuals and the types of schools for which a program works best. However, this approach may have limited applicability in TPP and other prevention programs in practice, as the number of higher-level units, such as schools, may be too small to uncover meaningful classes.</p> <p>Integrative data analysis (IDA) across multiple TPP studies provides a potential solution to these issues. IDA is a framework by which two or more individual studies are pooled into a single dataset for analysis (Curran & Hussong, [<reflink idref="bib4" id="ref15">4</reflink>]). Unlike other methods for investigating multiple studies, like meta-analysis, IDA pools raw data from individual participants (Hussong et al., [<reflink idref="bib10" id="ref16">10</reflink>]). By integrating data across multiple studies, researchers can assemble larger sample sizes, providing a higher likelihood of uncovering a greater number of latent classes that may more accurately represent the varying profiles of individuals within the population from which the samples are drawn. In addition, the larger number of schools in the integrated data allows using MLCA to construct latent class moderators at both the individual and school levels and examine how profiles at both these levels moderate program effectiveness. Note that we focus on data from individuals nested within schools to provide an example of the utility or combining MLCA with integrated data in a type of design that is common in TPP programs. However, MLCA can also be utilized with other types of nested data (e.g., individuals within community settings, daily measurements within individuals).</p> <p>In this study, we demonstrate MLCA with integrated using integrated data from four trials of a middle-school TPP (It's Your Game...Keep it Real) to examine multidimensional moderators of program effectiveness at the individual and school levels. Our analysis had the following aims:</p> <p></p> <ulist> <item> 1. To uncover latent classes of risk and protective factors for sexual initiation in middle school students at the individual and school levels.</item> <p></p> <item> 2. To test how membership in these classes at the individual and school levels moderated the effectiveness of the TPP programs on sexual initiation approximately 1 year post-intervention.</item> </ulist> <hd id="AN0173822458-2">Method</hd> <p></p> <hd id="AN0173822458-3">Participants and Procedure</hd> <p>Data for this study were from an integrative dataset from four randomized middle school pregnancy prevention trials that were replications or adaptations of the It's Your Game...Keep it Real program. These studies represent a convenience sample of trials the authors had access to with data sufficiently prepared for analysis. The studies include trials that found significant programs effects, as well as those that did not.</p> <hd id="AN0173822458-4">It's Your Game...Keep It Real (IYG)</hd> <p>It's Your Game...Keep It Real (IYG) is a skill and norm-based classroom curriculum with a parent component designed to promote abstinence and teach students about their bodies, healthy relationships, personal boundaries, and protecting themselves from pregnancy and STIs. IYG was evaluated in a cluster randomized controlled trial in ten middle schools in a large, urban, predominantly minority school district in Southeast Texas (Tortolero et al., [<reflink idref="bib24" id="ref17">24</reflink>]). Five schools were randomly assigned to receive the IYG intervention or the comparison condition, which received the district's standard health education curriculum. Seventh-grade students were recruited and followed through 9th grade. Baseline surveys were conducted in fall 2006 and spring 2007 and were completed by 907 students. The primary outcome was the proportion of lifetime sexual activity at the 9th-grade follow-up. Participants in the intervention condition were 50% less likely to initiate oral sex, 29% less likely to initiate vaginal sex, 66% less likely to initiate anal sex, and 36% less likely to initiate "any" type of sex over the 7th- to 9th-grade period.</p> <hd id="AN0173822458-5">IYG Texas Replication</hd> <p>To test the effectiveness of IYG in "real-world" settings, IYG was implemented by teachers in 20 urban and suburban middle schools in Southeast Texas from 2012 to 2015. IYG was evaluated using a group-randomized wait-list controlled effectiveness trial design, with 10 schools randomized to the intervention condition and 10 schools randomized to the comparison condition (standard care). The analytic sample comprised 1543 students (<emph>n</emph> = 804, intervention; <emph>n</emph> = 739, comparison) who were followed from baseline (seventh grade) to the 24-month follow-up (ninth grade). There were no significant differences in vaginal or oral sex initiation between study conditions at ninth grade follow-up (Peskin et al., [<reflink idref="bib17" id="ref18">17</reflink>]).</p> <hd id="AN0173822458-6">IYG South Carolina Replication</hd> <p>This IYG replication study took place from 2011 to 2014 and involved 24 rural middle schools (Potter et al., [<reflink idref="bib19" id="ref19">19</reflink>]). School staff in half (<reflink idref="bib12" id="ref20">12</reflink>) of the schools implemented the IYG curriculum. Youth completed baseline surveys in the fall of 7th grade prior to receiving the first 12 sessions of the curriculum. They received 12 additional sessions in 8th grade and then completed a follow-up survey (0 to 6 months post-program) in the spring of 8th grade. Participants completed a final follow-up survey 12 to 18 months post-program in the spring of 9th grade. The ninth-grade survey was completed by 1357 intervention and 1130 control students. The original studies' behavioral effects were not replicated in this study. There was no statistically significant effect on the initiation of vaginal sex between baseline and eighth grade. Significantly fewer students in the comparison condition reported initiating sex at ninth grade, relative to the intervention condition.</p> <hd id="AN0173822458-7">IYG Tech</hd> <p>The It's Your Game-Tech (IYG Tech) study evaluated a computer-based version of the IYG intervention delivered in the eighth grade. IYG-Tech was evaluated using a randomized, two-arm nested design among 19 schools in a large, urban school district in Southeast Texas from 2010 to 2012 (Peskin et al., [<reflink idref="bib18" id="ref21">18</reflink>]). Ten schools were randomized to the intervention condition, and 10 schools were randomized to the comparison condition (standard care). The final analytic sample included 1374 students. Although there was no significant difference in sexual risk behaviors between intervention and comparison group students in the ninth grade, post hoc analyses conducted among intervention students indicated delayed sexual initiation of vaginal initiation among students who received all or at least half of the curriculum compared to those who received less than half.</p> <hd id="AN0173822458-8">Integrated Dataset</hd> <p>The integrative dataset consisted of 8397 individuals across 73 schools. Because of our interest in examining sexual initiation, we restricted the sample to individuals who had not had sexual intercourse at baseline and had data on the outcome. Thus, our analytic sample consisted of 5972 students who had not previously engaged in sexual intercourse and had data on the outcome variable (44.7% male, 55.3% female; 22.5% Black, 47.2% Latinx, 4.4% multiracial, and 25.8% other race; <emph>M</emph> age at baseline = 14.17, SD = 0.81).</p> <hd id="AN0173822458-9">Measures</hd> <p></p> <hd id="AN0173822458-10">Harmonization Process</hd> <p>Integration of the datasets involved two major sets of decisions. First, after examining all the follow-up waves across each study, we made the decision to include outcome waves that occurred approximately 1 year after the final program content had been completed; these follow-ups ranged from 12 to 19 months after study completion. Second, we examined all potential study variables to assess whether the items were asked in a comparable way across all studies. Questions and response options were compared across all four studies, and variables were recoded and response options combined to create comparable measures. Details of the measures and recoding are presented in Appendix 1.</p> <hd id="AN0173822458-11">Indicators of Latent Class Membership</hd> <p>We included six indicators of risk and protective factors for sexual behavior, which were selected based on prior research suggesting they were associated with sexual initiation (Kirby & Lepore, [<reflink idref="bib11" id="ref22">11</reflink>]; Zimmer-Gembeck & Helfand, [<reflink idref="bib28" id="ref23">28</reflink>]) and their availability across the studies in the integrated dataset. The same variables were used to characterize profiles of individuals based on individual responses to the items and types of schools based on the prevalence of these items within students in schools. Missingness on the indicators range from 0.4 to 22%, with the latter due to the fact that the variable assessing whether another language was spoken in the home was not asked in the IYG-Tech study. We included this variable as our prior research has suggested it is conceptually important in the population studied and LCA uses full implementation maximum likelihood (FIML) to account for missing data. For the purposes of this demonstration, we considered the missingness on this variable missing at random, as the missingness was due to study design, rather than student characteristics, the IYG-Tech sample was drawn from the same geographic areas and population as two of the other studies, and we included other correlated variables, such as nation of birth, in the analysis. For the remainder of the indicator variables, there was no discernable pattern of missingness found so we assumed missing completely at random (MCAR). <emph>Other language household</emph> was an indicator of whether a participant spoke only English or another language in the home. <emph>Born in the USA</emph> measured whether the participant was born in the USA or in another country. <emph>Dating partner's age</emph> was a three-level indicator of whether the individual was in a romantic relationship, with recoded options including never date/do not usually date, only dating partners who were younger or less than 1 year older and had a partner who was more than 1 year older. <emph>Participant's grades</emph> indicated whether individuals reported getting mostly As, mostly Bs, mostly Bs and Cs, or lower. <emph>Family structure</emph> measured whether individuals lived with both parents or with only one parent/some other living arrangement. <emph>Alcohol use</emph> indicated whether the adolescent used alcohol, measured in the past year.</p> <hd id="AN0173822458-12">Outcome Variable</hd> <p>Our outcome was an indicator of sexual initiation, measuring whether individuals who had not reported ever engaging in vaginal sex at baseline reported doing so at the target follow-up.</p> <hd id="AN0173822458-13">Correlates of Class Membership</hd> <p>We examined group differences in class membership by race/ethnicity, gender, and age. Race was measured by four categories: Black, Latinx, multiracial, and other. Gender assessed whether participants reported being male or female. Age was measured as an integer measure of how old the participant was at baseline.</p> <hd id="AN0173822458-14">Statistical Analyses</hd> <p>We used MLCA (Henry & Muthen, [<reflink idref="bib9" id="ref24">9</reflink>]; Vermunt, [<reflink idref="bib26" id="ref25">26</reflink>]) to identify latent profiles of risk and protective factors influencing TPP program effectiveness at the individual (L1) and school (L2) levels based on the 6 indicators (see Appendix 2 for sample code). For each level, we considered models with 1–6 latent classes. Analyses were run in Latent Gold (Vermunt & Madgison, [<reflink idref="bib27" id="ref26">27</reflink>]). Although procedures for MLCA are largely similar to traditional LCA, model selection is more complex, based on the need to select appropriate models at the higher and lower levels when class structures and fit statistics at L1 differ based on the classes selected at L2. Thus, we used the three-step approach to model selection suggested by Lukočienė and colleagues (Lukočienė et al., [<reflink idref="bib15" id="ref27">15</reflink>]). While the two-step approach (Vermunt, [<reflink idref="bib26" id="ref28">26</reflink>]) has been commonly used, it ignores the dependency between the number of lower classes and the impact on the number of higher-level classes. The three-step approach considers that the decisions of number of classes at each level are not mutually independent. First, we selected the appropriate number of lower-level (in this case, individual) classes without considering the multilevel structure. Second, we ran a series of models which set the lower-level classes at the number suggested by step one to determine the appropriate model for the higher-level (in this case, schools). After selecting our higher-level model, we then ran a series of models with the number of higher-level classes set as determined in step two, to determine the optimal model for the lower level when the clustering at the higher level is considered. For all steps, we selected the model with the optimal number of latent classes based on fit statistics, interpretability, the conceptual separation between classes, class size (e.g., greater than 10% in each class) and selecting a model that was both parsimonious yet highlighted the utility of the method for this demonstration. We focused on Akaike information criteria (AIC) and Bayesian information criteria (BIC) to help select the appropriate number of latent classes at the lower level. In evaluating model fit, the AIC and the BIC perform the best when concurrently deciding the number of lower- and higher-level classes when indicators are categorical (Andrews & Currim, [<reflink idref="bib1" id="ref29">1</reflink>]; Fonseca & Cardoso, [<reflink idref="bib8" id="ref30">8</reflink>]; Lukociene et al., [<reflink idref="bib15" id="ref31">15</reflink>]).</p> <p>Next, we examined how class membership is associated with the outcome of sexual initiation using the three-step approach to covariates and the multilevel option. This approach involves (<reflink idref="bib1" id="ref32">1</reflink>) selecting an optimal latent class model, (<reflink idref="bib2" id="ref33">2</reflink>) saving the probabilities of membership in each class for each individual, and (<reflink idref="bib3" id="ref34">3</reflink>) running an outcome analysis, with individuals assigned to classes based on modal class membership. We examined how individual- and school-level risk classes moderated the effect of the intervention on sexual initiation, controlling for age, gender, race/ethnicity, and study in a logistic model accounting for the multilevel structure of the data.</p> <hd id="AN0173822458-15">Results</hd> <p></p> <hd id="AN0173822458-16">Model Selection and Class Interpretation</hd> <p>Using the steps suggested by the three-step approach to model selection in MLCA (Lukočienė et al., [<reflink idref="bib15" id="ref35">15</reflink>]), we first fit models with one through six classes at the individual level (Table 1). Based on BIC and AIC results which suggested a four-class model, we used a four-class solution to examine school-level clustering. The school-level results suggested a five- or six-class solution; however, the five- and six-class models each contained two classes with a very small number of schools (less than 10%), we chose the four-class model at L2, which had conceptually distinct classes, all class sizes above 10%, and had sufficient variety of classes for demonstration purposes. Then we set the number of higher-level classes at four and ran a series of models to select the lower-level (individual) classes (Table 1). Fit statistics suggested a six-class model; however, when the six-class model was selected at L1, one of the classes at L2 was very small (3% of schools), and thus we also examined the four- and five-class models. Based on our criteria emphasizing class size, parsimony, and interpretability, we selected a final model that had four school-level classes and four individual-level classes.</p> <p>Table 1 Fit statistics for models with 1–6 individual classes, based on four levels of school-level classes</p> <p> <ephtml> <table frame="hsides" rules="groups"><thead><tr><th align="left"><p># of Classes</p></th><th align="left"><p>BIC</p></th><th align="left"><p>AIC</p></th><th align="left"><p>AIC3</p></th><th align="left"><p>CAIC</p></th><th align="left"><p>SABIC</p></th><th align="left"><p>Entropy L2</p></th><th align="left"><p>Entropy L1</p></th></tr></thead><tbody><tr><td align="left"><p>1</p></td><td align="left"><p>46920.99</p></td><td align="left"><p>46847.35</p></td><td align="left"><p>46931.99</p></td><td align="left"><p>46886.04</p></td><td align="left"><p>46920.99</p></td><td align="left"><p>1.00</p></td><td align="left"><p>1.00</p></td></tr><tr><td align="left"><p>2</p></td><td align="left"><p>44475.58</p></td><td align="left"><p>44321.60</p></td><td align="left"><p>44498.58</p></td><td align="left"><p>44402.49</p></td><td align="left"><p>44475.58</p></td><td align="left"><p>0.92</p></td><td align="left"><p>0.80</p></td></tr><tr><td align="left"><p>3</p></td><td align="left"><p>44085.97</p></td><td align="left"><p>43851.65</p></td><td align="left"><p>44120.97</p></td><td align="left"><p>43974.75</p></td><td align="left"><p>44085.97</p></td><td align="left"><p>0.97</p></td><td align="left"><p>0.75</p></td></tr><tr><td align="left"><p>4</p></td><td align="left"><p>43880.72</p></td><td align="left"><p>43566.06</p></td><td align="left"><p>43927.72</p></td><td align="left"><p>43731.36</p></td><td align="left"><p>43880.72</p></td><td align="left"><p>0.98</p></td><td align="left"><p>0.69</p></td></tr><tr><td align="left"><p>5</p></td><td align="left"><p>43724.94</p></td><td align="left"><p>43329.94</p></td><td align="left"><p>43783.94</p></td><td align="left"><p>43537.45</p></td><td align="left"><p>43724.94</p></td><td align="left"><p>0.99</p></td><td align="left"><p>0.66</p></td></tr><tr><td align="left"><p>6</p></td><td align="left"><p>43679.21</p></td><td align="left"><p>43203.87</p></td><td align="left"><p>43750.21</p></td><td align="left"><p>43453.59</p></td><td align="left"><p>43679.21</p></td><td align="left"><p>0.99</p></td><td align="left"><p>0.60</p></td></tr></tbody></table> </ephtml> </p> <p>Results of the individual-level classes are presented in Table 2. <emph>Lower risk daters</emph> (30%) consisted of individuals who were very likely to speak only English and be born in the USA, be dating a partner who was likely to be the same age, receive As and Bs, and not drink alcohol. <emph>Single parent household</emph> (27%) was marked by high probabilities of not speaking another language at home and being born in the USA, dating the same age partner, receiving As and Bs, not using alcohol, and not living with both parents. <emph>Other language household</emph> (22%) included students who were highly likely to speak a language other than English at home and had the highest probability of being born outside the USA. They were likely to have no romantic relationship or one with a same age partner, unlikely to use alcohol and were likely to live with both parents. <emph>Higher individual risk</emph> (21%) consisted of individuals who were very likely to have had romantic relationships, most likely with an older partner, and engage in alcohol use.</p> <p>Table 2 Latent class prevalence and item-response probabilities for four-class model of risk and protective factors at the individual level (L1)</p> <p> <ephtml> <table frame="hsides" rules="groups"><thead><tr><th align="left" /><th align="left"><p>Lower risk daters</p></th><th align="left"><p>Single parent household</p></th><th align="left"><p>Other language household</p></th><th align="left"><p>Higher individual risk</p></th></tr></thead><tbody><tr><td align="left" /><td align="left"><p><bold>30%</bold></p></td><td align="left"><p><bold>27%</bold></p></td><td align="left"><p><bold>22%</bold></p></td><td align="left"><p><bold>21%</bold></p></td></tr><tr><td align="left" colspan="5"><p><italic>Item-response probabilities</italic></p></td></tr><tr><td align="left"><p><italic>Language</italic></p></td><td align="left" /><td align="left" /><td align="left" /><td align="left" /></tr><tr><td align="left"><p> Other language</p></td><td align="left"><p>0.03</p></td><td align="left"><p>0.07</p></td><td align="left"><p><bold>0.99</bold></p></td><td align="left"><p>0.42</p></td></tr><tr><td align="left"><p> English only</p></td><td align="left"><p><bold>0.97</bold></p></td><td align="left"><p><bold>0.93</bold></p></td><td align="left"><p>0.01</p></td><td align="left"><p><bold>0.58</bold></p></td></tr><tr><td align="left"><p><italic>Nativity</italic></p></td><td align="left" /><td align="left" /><td align="left" /><td align="left" /></tr><tr><td align="left"><p> Born outside USA</p></td><td align="left"><p>0.05</p></td><td align="left"><p>0.10</p></td><td align="left"><p>0.32</p></td><td align="left"><p>0.06</p></td></tr><tr><td align="left"><p> Born in USA</p></td><td align="left"><p><bold>0.95</bold></p></td><td align="left"><p><bold>0.90</bold></p></td><td align="left"><p><bold>0.68</bold></p></td><td align="left"><p><bold>0.94</bold></p></td></tr><tr><td align="left"><p><italic>Relationship status</italic></p></td><td align="left" /><td align="left" /><td align="left" /><td align="left" /></tr><tr><td align="left"><p> No relationship</p></td><td align="left"><p>0.00</p></td><td align="left"><p>0.26</p></td><td align="left"><p>0.35</p></td><td align="left"><p>0.05</p></td></tr><tr><td align="left"><p> Younger or same age</p></td><td align="left"><p><bold>0.65</bold></p></td><td align="left"><p><bold>0.56</bold></p></td><td align="left"><p><bold>0.44</bold></p></td><td align="left"><p><bold>0.44</bold></p></td></tr><tr><td align="left"><p> Older</p></td><td align="left"><p><bold>0.35</bold></p></td><td align="left"><p>0.17</p></td><td align="left"><p>0.20</p></td><td align="left"><p><bold>0.51</bold></p></td></tr><tr><td align="left"><p><italic>Grades</italic></p></td><td align="left" /><td align="left" /><td align="left" /><td align="left" /></tr><tr><td align="left"><p> Mostly Cs or lower</p></td><td align="left"><p>0.10</p></td><td align="left"><p>0.05</p></td><td align="left"><p>0.08</p></td><td align="left"><p>0.14</p></td></tr><tr><td align="left"><p> Mostly B and Cs</p></td><td align="left"><p>0.31</p></td><td align="left"><p>0.36</p></td><td align="left"><p>0.42</p></td><td align="left"><p><bold>0.50</bold></p></td></tr><tr><td align="left"><p> Mostly As and Bs</p></td><td align="left"><p><bold>0.59</bold></p></td><td align="left"><p><bold>0.59</bold></p></td><td align="left"><p><bold>0.50</bold></p></td><td align="left"><p><bold>0.36</bold></p></td></tr><tr><td align="left"><p><italic>Family structure</italic></p></td><td align="left" /><td align="left" /><td align="left" /><td align="left" /></tr><tr><td align="left"><p> Lives with both parents</p></td><td align="left"><p><bold>0.52</bold></p></td><td align="left"><p>0.36</p></td><td align="left"><p><bold>0.58</bold></p></td><td align="left"><p>0.38</p></td></tr><tr><td align="left"><p> Does not live with both parents</p></td><td align="left"><p>0.48</p></td><td align="left"><p><bold>0.64</bold></p></td><td align="left"><p>0.42</p></td><td align="left"><p><bold>0.62</bold></p></td></tr><tr><td align="left"><p><italic>Alcohol use</italic></p></td><td align="left" /><td align="left" /><td align="left" /><td align="left" /></tr><tr><td align="left"><p> No</p></td><td align="left"><p><bold>0.75</bold></p></td><td align="left"><p><bold>0.92</bold></p></td><td align="left"><p><bold>0.82</bold></p></td><td align="left"><p>0.22</p></td></tr><tr><td align="left"><p> Yes</p></td><td align="left"><p>0.25</p></td><td align="left"><p>0.08</p></td><td align="left"><p>0.18</p></td><td align="left"><p><bold>0.78</bold></p></td></tr></tbody></table> </ephtml> </p> <p>Probabilities greater than.5 for dichotomous and.35 for trichotomous variables have been bolded to facilitate interpretation</p> <p>Results of the school-level classes are presented in Table 3. Item response probabilities at the school level can be interpreted as an average of the probabilities of a particular response to an indicator from all schools within a given latent class, derived from the responses from students within each class. Thus, a higher response on a given indicator suggests that schools assigned to this class are more likely to have students reporting the response on the indicator. <emph>Lower risk schools</emph> (30%) consisted of schools with a majority of students speaking only English at home, being born in the USA, dating a same-age partner, and receiving mostly As and Bs. <emph>Higher risk schools</emph> (28%) consisted of schools with a high proportion of students not speaking another language at home and being born in the USA, dating a same-age partner or older partner, and receiving As and Bs or Bs and Cs. This class also had the highest prevalence of alcohol use. Slightly more than half of students in this type of school are estimated to not live with both their parents. <emph>Other language schools</emph> (25%) indicates schools that had a high prevalence of students who spoke another language at home, but had normative or low prevalence of other risk indicators. <emph>Mixed language, single parent schools</emph> (17%) was marked by most students speaking only English at home, but a substantial minority speaking another language at home. They were mostly at lower risk on all outcomes, but had more students not living with both parents than living with both parents.</p> <p>Table 3 Latent class prevalence and item-response probabilities for four-class model of risk and protective factors at the school level (L2)</p> <p> <ephtml> <table frame="hsides" rules="groups"><thead><tr><th align="left" /><th align="left"><p>Lower risk schools</p></th><th align="left"><p>Higher risk schools</p></th><th align="left"><p>Lower risk, other language schools</p></th><th align="left"><p>Mixed language, single parent schools</p></th></tr></thead><tbody><tr><td align="left" /><td align="left"><p><bold>30%</bold></p></td><td align="left"><p><bold>28%</bold></p></td><td align="left"><p><bold>25%</bold></p></td><td align="left"><p><bold>17%</bold></p></td></tr><tr><td align="left"><p>Lower risk daters</p></td><td align="left"><p>0.95</p></td><td align="left"><p>0.00</p></td><td align="left"><p>0.00</p></td><td align="left"><p>0.14</p></td></tr><tr><td align="left"><p>Single parent household</p></td><td align="left"><p>0.00</p></td><td align="left"><p>0.45</p></td><td align="left"><p>0.26</p></td><td align="left"><p>0.46</p></td></tr><tr><td align="left"><p>Other language household</p></td><td align="left"><p>0.03</p></td><td align="left"><p>0.00</p></td><td align="left"><p>0.64</p></td><td align="left"><p>0.30</p></td></tr><tr><td align="left"><p>Higher individual risk</p></td><td align="left"><p>0.02</p></td><td align="left"><p>0.55</p></td><td align="left"><p>0.11</p></td><td align="left"><p>0.11</p></td></tr><tr><td align="left"><p><italic>Language</italic></p></td><td align="left" /><td align="left" /><td align="left" /><td align="left" /></tr><tr><td align="left"><p> Other language</p></td><td align="left"><p>0.07</p></td><td align="left"><p>0.27</p></td><td align="left"><p><bold>0.69</bold></p></td><td align="left"><p>0.38</p></td></tr><tr><td align="left"><p> English only</p></td><td align="left"><p><bold>0.93</bold></p></td><td align="left"><p><bold>0.73</bold></p></td><td align="left"><p>0.31</p></td><td align="left"><p><bold>0.62</bold></p></td></tr><tr><td align="left"><p><italic>Nativity</italic></p></td><td align="left" /><td align="left" /><td align="left" /><td align="left" /></tr><tr><td align="left"><p> Born outside USA</p></td><td align="left"><p>0.06</p></td><td align="left"><p>0.08</p></td><td align="left"><p>0.24</p></td><td align="left"><p>0.15</p></td></tr><tr><td align="left"><p> Born in USA</p></td><td align="left"><p><bold>0.94</bold></p></td><td align="left"><p><bold>0.92</bold></p></td><td align="left"><p><bold>0.76</bold></p></td><td align="left"><p><bold>0.85</bold></p></td></tr><tr><td align="left"><p><italic>Relationship status</italic></p></td><td align="left" /><td align="left" /><td align="left" /><td align="left" /></tr><tr><td align="left"><p> No relationship</p></td><td align="left"><p>0.01</p></td><td align="left"><p>0.14</p></td><td align="left"><p>0.30</p></td><td align="left"><p>0.23</p></td></tr><tr><td align="left"><p> Younger or same age</p></td><td align="left"><p><bold>0.64</bold></p></td><td align="left"><p><bold>0.50</bold></p></td><td align="left"><p><bold>0.47</bold></p></td><td align="left"><p><bold>0.53</bold></p></td></tr><tr><td align="left"><p> Older</p></td><td align="left"><p><bold>0.35</bold></p></td><td align="left"><p><bold>0.36</bold></p></td><td align="left"><p>0.23</p></td><td align="left"><p>0.24</p></td></tr><tr><td align="left"><p><italic>Grades</italic></p></td><td align="left" /><td align="left" /><td align="left" /><td align="left" /></tr><tr><td align="left"><p> Mostly Cs or Lower</p></td><td align="left"><p>0.10</p></td><td align="left"><p>0.10</p></td><td align="left"><p>0.08</p></td><td align="left"><p>0.08</p></td></tr><tr><td align="left"><p> Mostly B and Cs</p></td><td align="left"><p>0.32</p></td><td align="left"><p><bold>0.43</bold></p></td><td align="left"><p>0.41</p></td><td align="left"><p>0.38</p></td></tr><tr><td align="left"><p> Mostly As and Bs</p></td><td align="left"><p><bold>0.58</bold></p></td><td align="left"><p><bold>0.47</bold></p></td><td align="left"><p><bold>0.51</bold></p></td><td align="left"><p><bold>0.54</bold></p></td></tr><tr><td align="left"><p><italic>Family structure</italic></p></td><td align="left" /><td align="left" /><td align="left" /><td align="left" /></tr><tr><td align="left"><p> Lives with both parents</p></td><td align="left"><p><bold>0.52</bold></p></td><td align="left"><p>0.37</p></td><td align="left"><p>0.50</p></td><td align="left"><p>0.45</p></td></tr><tr><td align="left"><p> Does not live with both parents</p></td><td align="left"><p>0.48</p></td><td align="left"><p><bold>0.63</bold></p></td><td align="left"><p>0.50</p></td><td align="left"><p><bold>0.55</bold></p></td></tr><tr><td align="left"><p><italic>Alcohol use</italic></p></td><td align="left" /><td align="left" /><td align="left" /><td align="left" /></tr><tr><td align="left"><p> No</p></td><td align="left"><p><bold>0.74</bold></p></td><td align="left"><p><bold>0.54</bold></p></td><td align="left"><p><bold>0.79</bold></p></td><td align="left"><p><bold>0.80</bold></p></td></tr><tr><td align="left"><p> Yes</p></td><td align="left"><p>0.26</p></td><td align="left"><p>0.46</p></td><td align="left"><p>0.21</p></td><td align="left"><p>0.20</p></td></tr></tbody></table> </ephtml> </p> <p>Probabilities greater than.5 for dichotomous and.35 for trichotomous variables have been bolded to facilitate interpretation</p> <p>Individuals were differentially represented in school-level classes based on their individual-level class membership (Table 3). Individuals in the lower risk daters class were primarily found in the lower risk schools class. The higher risk schools class contained students primarily from the single parent household and higher individual risk classes. The lower risk other language school class was predominately students from the other language household class, with a minority from the single parent household and higher individual risk classes. The mixed language, single parents' schools had a plurality of students from the single parent household class, but included students from all individual-level classes.</p> <hd id="AN0173822458-17">Program Effectiveness by Latent Class</hd> <p>Results examining class membership and program condition as predictors of initiation of vaginal sex (controlling for demographic factors and study) is shown in Table 4. We used the <emph>higher individual risk</emph> and <emph>higher risk schools</emph> as the reference groups, based on our prior research suggesting stronger effects in higher risk classes (Vasilenko et al., [<reflink idref="bib25" id="ref36">25</reflink>]). Individual-level class membership was associated with sexual initiation for those in the control group, with individuals in <emph>single parent household</emph> and <emph>other language household</emph> having lower rates of sexual initiation compared to the <emph>higher individual risk</emph> class. School-level classes were also associated with sexual initiation for individuals in the control group, with the <emph>single parent</emph> and <emph>lower risk</emph> schools having a lower prevalence compared to the <emph>higher risk schools</emph>. The treatment effect was not significant for the reference classes (<emph>higher individual risk</emph> and <emph>higher risk schools</emph>), and the effect of the intervention differed for both individual and school-level classes. Estimated prevalence of sexual initiation by treatment and school-level class derived from this model are presented in Fig. 1. As a follow-up test, we also tested whether the treatment effect was significant for each class by first changing the reference group for the individual classes, keeping the school-level classes constant, and then by changing the school-level classes, keeping the individual-level classes constant. At the individual-level, no classes had significant differences between the treatment and control group; however, the <emph>lower risk</emph> daters had a reduction in sexual initiation of 30%, and the <emph>single parent household</emph> had a reduction of 16%. For the school-level classes, there was a significant difference between the treatment and control group for the <emph>lower risk</emph>, <emph>other language</emph> schools, which was associated with a 23% decrease in sexual initiation.</p> <p>Table 4 Results of multinomial regression predicting of latent classes of consequences of vaginal sex initiation by gender</p> <p> <ephtml> <table frame="hsides" rules="groups"><thead><tr><th align="left" /><th align="left"><p>OR</p></th><th align="left"><p>CI Lower</p></th><th align="left"><p>CI Upper</p></th><th align="left"><p><italic>χ</italic><sup>2</sup></p></th></tr></thead><tbody><tr><td align="left"><p>Intercept</p></td><td char="." align="char"><p>0.94</p></td><td char="." align="char" /><td char="." align="char" /><td align="left"><p>0.07</p></td></tr><tr><td align="left"><p>L1 classes</p></td><td char="." align="char" /><td char="." align="char" /><td char="." align="char" /><td align="left"><p>63.67***</p></td></tr><tr><td align="left"><p> Lower risk daters</p></td><td char="." align="char"><p>0.70</p></td><td char="." align="char"><p>0.37</p></td><td char="." align="char"><p>1.32</p></td><td align="left" /></tr><tr><td align="left"><p> Single parent household</p></td><td char="." align="char"><p>0.25</p></td><td char="." align="char"><p>0.15</p></td><td char="." align="char"><p>0.41</p></td><td align="left" /></tr><tr><td align="left"><p> Other language household</p></td><td char="." align="char"><p>0.18</p></td><td char="." align="char"><p>0.12</p></td><td char="." align="char"><p>0.28</p></td><td align="left" /></tr><tr><td align="left"><p>L2 classes</p></td><td char="." align="char" /><td char="." align="char" /><td char="." align="char" /><td align="left"><p>425.27***</p></td></tr><tr><td align="left"><p> Lower risk schools</p></td><td char="." align="char"><p>0.51</p></td><td char="." align="char"><p>0.29</p></td><td char="." align="char"><p>0.89</p></td><td align="left" /></tr><tr><td align="left"><p> Lower risk, other language schools</p></td><td char="." align="char"><p>1.21</p></td><td char="." align="char"><p>0.99</p></td><td char="." align="char"><p>1.48</p></td><td align="left" /></tr><tr><td align="left"><p> Mixed language, single parent schools</p></td><td char="." align="char"><p>0.92</p></td><td char="." align="char"><p>0.75</p></td><td char="." align="char"><p>1.13</p></td><td align="left" /></tr><tr><td align="left"><p>Treatment effect</p></td><td char="." align="char" /><td char="." align="char" /><td char="." align="char" /><td align="left"><p>.14</p></td></tr><tr><td align="left"><p> Treatment X L1 class</p></td><td char="." align="char" /><td char="." align="char" /><td char="." align="char" /><td align="left"><p>26.00***</p></td></tr><tr><td align="left"><p> Treat X lower risk daters</p></td><td char="." align="char"><p>0.58</p></td><td char="." align="char"><p>0.23</p></td><td char="." align="char"><p>1.46</p></td><td align="left" /></tr><tr><td align="left"><p> Treat X single parent household</p></td><td char="." align="char"><p>0.79</p></td><td char="." align="char"><p>0.45</p></td><td char="." align="char"><p>1.39</p></td><td align="left" /></tr><tr><td align="left"><p> Treat X other language household</p></td><td char="." align="char"><p>1.23</p></td><td char="." align="char"><p>0.75</p></td><td char="." align="char"><p>2.02</p></td><td align="left" /></tr><tr><td align="left"><p>Treatment X L2 class</p></td><td char="." align="char" /><td char="." align="char" /><td char="." align="char" /><td align="left"><p>99.52***</p></td></tr><tr><td align="left"><p> Treat X lower risk schools</p></td><td char="." align="char"><p>1.88</p></td><td char="." align="char"><p>0.84</p></td><td char="." align="char"><p>4.20</p></td><td align="left" /></tr><tr><td align="left"><p> Treat X lower risk, other language schools</p></td><td char="." align="char"><p>0.61</p></td><td char="." align="char"><p>0.46</p></td><td char="." align="char"><p>0.81</p></td><td align="left" /></tr><tr><td align="left"><p> Treat X mixed language, single parent</p></td><td char="." align="char"><p>0.91</p></td><td char="." align="char"><p>0.64</p></td><td char="." align="char"><p>1.30</p></td><td align="left" /></tr></tbody></table> </ephtml> </p> <p>Analyses controlled for gender, race/ethnicity, age, and study. Higher individual risk and higher risk schools classes are used as the reference group *<emph>p</emph> <.05, **<emph>p</emph> <.01, ***<emph>p</emph> <.001</p> <p>Graph: Fig. 1Sexual initiation from baseline to follow-up based on school and individual-level latent class and treatment condition, based on results from multilevel logistic regression analysis. Results are adjusted for gender, race/ethnicity, age, and study, with higher individual risk and higher school risk as reference group. Asterisk indicates classes for which the difference between treatment and control group is significantly different in the full multilevel latent class model</p> <hd id="AN0173822458-18">Discussion</hd> <p></p> <hd id="AN0173822458-19">Effectiveness of Program Within Latent Classes</hd> <p>This study demonstrated the use of integrative data analysis and multilevel latent class analysis to examine how multidimensional profiles of risk and protective factors at the individual and school level moderated the effectiveness of TPP programs. We found evidence of heterogeneous profiles of risk at both the individual- and school-levels, and that both types of profiles may moderate the effectiveness of the program on sexual initiation. At the school level, we found the strongest effects of the program in schools marked by a high prevalence of students from a household where another language was spoken. Although not directly assessed with these indicators, the combination of high probabilities of speaking another language at home and lower probabilities of being born outside of the USA reported by individuals within these schools suggests that many adolescents in these schools may be second generation immigrants whose parents were born outside of the USA (largely from Latin American countries in this sample) but who themselves were born in the USA. Thus, these findings may be related to the immigrant paradox, in which individuals with later immigrant generation or greater acculturation have greater health-risk behaviors despite decreases in risk factors associated with health (Marks et al., [<reflink idref="bib16" id="ref37">16</reflink>]). In the area of adolescent sexual behavior, second-generation adolescents may have higher rates of sexual behavior and may be taxed with navigating their sexual behaviors in a context that is very different from the one in which they were raised by their immigrant parents (Lee & Hahm, [<reflink idref="bib13" id="ref38">13</reflink>]; Raffaelli et al., [<reflink idref="bib20" id="ref39">20</reflink>]). Thus, in schools with high percentages of second-generation immigrant adolescents, TPP programs may have a critically important role in demonstrating norms about sexuality within the U.S. context, leading to improved outcomes for individuals in these schools.</p> <p>Although not statistically significant based on post-hoc tests, there were differences at the individual-level which suggest more promising program effects were found for individuals with certain risk profiles. In particular, the program appeared to be more effective for individuals with a high probability of dating same-age partners, but otherwise relatively low prevalence of risk factors. It is possible that individuals with this kind of dating profile are at a stage where they are particularly responsive to socialization about sexuality and thus receive a greater program effect. In addition, the program appeared to be more effective for individuals with a relatively high probability of a family risk factor, not living with both parents, but not high prevalence of individual risk behaviors. This finding is consistent with prior research, which found effectiveness for one of the studies used in this analysis only for a class marked by family disruption (Vasilenko et al., [<reflink idref="bib25" id="ref40">25</reflink>]). Although this current study was not able to assess more serious levels of family disruption, taken together, these studies suggest that school-based TPP programs may be most effective for individuals who may have some level of family-based risk.</p> <p>While effects were beneficial to neutral for most classes, some classes had a higher prevalence of sexual initiation for the treatment compared to the control group. Post-hoc test revealed this difference was not significant for any class, suggesting neutral to positive impacts across studies. However, these results suggest further attention be paid to groups of individuals or schools who received neutral or negative program effects, to better understand why they may not receive benefits on sexual initiation and how the program influenced other behaviors. In particular, the <emph>lower risk</emph> schools showed higher initiation in the treatment compared to control group. While this is non-significant and may be an artifact of this particular model, it suggests a potential population for which the program may not be effective and which may require further program changes. However, these results should be interpreted as providing ways to understand and improve school-based TPP programs but not as evidence of a lack of efficacy of such programs. In addition, this study examined only one outcome, which is not the only measure of effectiveness of a program.</p> <hd id="AN0173822458-20">Challenges in Applications of MLCA</hd> <p>The approach used in this study, MLCA with integrated data across school-based studies, can provide a number of insights about school-based TPP programs that would not be possible with other approaches. First, it provides more information about the different types of individuals and schools in the populations studied. These characterizations can allow researchers to adapt programs to better address the needs of different groups of students. Second, this approach shows for which groups of individuals the program works, which may help researchers better understand null or small program effects. Finally, this approach may better help researchers understand potential processes by which their programs may be working in more holistic and nuanced ways. This study suggests a few potential mechanisms by which school-based middle school TPP programs may be effective, namely by providing knowledge and guidance for individuals in normative romantic relationships or experiencing family disruption and by creating particular norms about sexual behavior within schools, especially for second-generation immigrant adolescents who may be trying to reconcile socialization from their parents with American cultural messages. These types of insight may be harder to detect without using a method like MLCA that both examines multidimensional patterns of risk factors and potential moderators and accounts for the clustering of these factors at the individual and school levels.</p> <p>In this paper, we have demonstrated how MLCA can be used in one specific type of analysis with integrated data from multiple trials of a school-based prevention program. However, MLCA, with or without integrated data, could be useful in a number of other types of data relevant to prevention researchers. For example, a similar type of analysis could be conducted on other types of clustered data from prevention trials, such as data with individuals clustered within community settings. In addition, MLCA could be applied to data with multiple measurement occasions per person, such as daily diary or ecological momentary assessment data. Data from many diary or EMA studies may have a sufficiently large number of level 2 units (individuals) that this type of analysis could be conducted from a single study without needing to integrate data from several studies. As an example, MLCA from a daily diary study of sexual risk behaviors could uncover daily risk classes marked by the occurrence of different patterns of risk on a given day or occasion of sex, as well as person-level classes marked by broader patterns of risk behavior by individuals across multiple days. Thus, there is considerable potential for this type of analysis to be used with integrated data from school or community-based studies, as well as with intensive longitudinal data with multiple measures for each participant.</p> <p>However, there are many challenges to using MLCA on integrated data from multiple prevention trials, which both are limitations to our study and may pose difficulties in adopting the method. First, the biggest challenge was creating an integrated dataset that contains indicators and outcomes that have been asked in all studies. Although LCA uses FIML and thus can include cases with missing data, using items that have not been asked in all studies creates additional concerns about missing data, which should be considered in conducting this type of study (Enders, [<reflink idref="bib7" id="ref41">7</reflink>]; Little & Rubin, [<reflink idref="bib14" id="ref42">14</reflink>]). In addition, even when particular constructs are assessed in all studies, they may not be asked in identical ways or use identical response options. This creates both additional work in cleaning and recoding the data for analysis and may introduce error into models if individuals were answering questions in different ways across studies due to the different response options. When multi-item scales are used, multiple studies have delineated methods for pooling items and testing their invariance across studies (Curran & Hussong, [<reflink idref="bib4" id="ref43">4</reflink>]; Curran et al., [<reflink idref="bib5" id="ref44">5</reflink>]). In our case, we used only categorical indicators, which were able to be combined in ways that seemed to be conceptually similar. Ideally, a bridging study, in which some individuals complete both forms of a question to assess whether they are being answered in the same manner, could be used to validate the integrated measures (Bainte & Curran, [<reflink idref="bib2" id="ref45">2</reflink>]). However, due to the secondary analysis and demonstrational nature of this paper, we were not able to conduct a bridging study. Although we believe the differences between items were largely relatively minor and the benefits of being able to use this large, integrated dataset outweigh the negatives, this is a limitation to the current study. In addition to these methodological challenges, our use of secondary data across multiple studies meant we were limited to items that were asked across most or all studies, and thus we were not able to include all items that were theoretically or conceptually important, such as pubertal development, family processes, or peer attitudes or behaviors.</p> <p>Second, model-selection for MLCA is complex, which can lead to difficulty in selecting an optimal model. This is true in all types of LCA, as there are a variety of fit statistics and other metrics which can be used in selecting a model, in addition to interpretability and parsimony concerns. However, these issues are magnified in MLCA, as the L1 classes are dependent on the L2 classes. Thus, there are potentially a large number of potential models to consider, and decisions made at one level have a bearing on the classes that can be uncovered at the other. In this example, we chose a model that was interpretable and provided a good demonstration of the potential of the method for the purposes of this article. However, other models may also be justifiable, and thus this should not be considered the one true model, but one that provides a useful way of examining heterogeneity in these data. Researchers using this approach should conduct model selection carefully and deliberately and be explicit about the decisions they make at every stage of the model selection process.</p> <p>Finally, this integrated dataset, which combines data across multiple trials of a TPP program, has some limitations in terms of program evaluation and generalizability. Because we have data only from specific trials in particular areas, our sample may not be generalizable to all individuals and schools. In particular, the data have a higher proportion of Black and Latino students than the general population, making this research less generalizable to White or Asian adolescents. However, as these groups show higher rates of teen pregnancy these schools do represent the populations of youth more likely to receive teen pregnancy prevention programming provided in schools. In addition, this study only assessed middle school TPP programs, and we do not know if the effects would hold for programs delivered in different settings. Finally, this study only examined individual-level indicators of class membership, and future research should seek to understand school and community level contexts that individuals are embedded in to fully understand which contextual factors moderate program effectiveness.</p> <p>Despite these limitations and challenges, this study provides a demonstration of the potential of MLCA with integrated data for examining multidimensional moderators of program effectiveness at the individual and school levels. It suggests that effects of middle school-based TPP programs may be strongest in schools with higher proportions of second-generation immigrants and for students who are experiencing some level of family disruption or who engage in normative dating. These findings provide several potential mechanisms by which programs may be effective, which provide areas for future research and program modifications.</p> <hd id="AN0173822458-21">Funding</hd> <p>This project was funded by the Office of Population Affairs, Health and Human Services (Grant# 5 PHEPA 000003–02-00).</p> <hd id="AN0173822458-22">Declarations</hd> <p></p> <hd id="AN0173822458-23">Ethical Approval</hd> <p>All procedures performed in the original data collection involving human participants were in accordance with the ethical standards of the institution and with the 1964 Helsinki Declaration and its later amendments or comparable ethical standards. IRB approval for each of the studies was obtained from the either the University of Texas Institutional Review Board (IYG Texas Studies), ETR Associates Institutional Review Board (IYG SC), or the Institutional Review Board of the University of California San Francisco (DTL). In addition, the studies were approved by the relevant school districts' Office of Research and Accountability or Comprehensive Health Education Committee.</p> <hd id="AN0173822458-24">Informed Consent</hd> <p>Informed consent was obtained from participants.</p> <hd id="AN0173822458-25">Conflicts of Interest</hd> <p>The authors declare that they have no conflict of interest.</p> <hd id="AN0173822458-26">Supplementary Information</hd> <p>Below is the link to the electronic supplementary material.</p> <p>Graph: Supplementary file1 Appendix 1 Original items for study variables and recoded values. Appendix 2 Latent Gold Sample Code (DOCX 27 KB)</p> <hd id="AN0173822458-27">Publisher's Note</hd> <p>Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.</p> <ref id="AN0173822458-28"> <title> References </title> <blist> <bibl id="bib1" idref="ref29" type="bt">1</bibl> <bibtext> Andrews RL, Currim IS. A comparison of segment retention criteria for finite mixture logit models. Journal of Marketing Research. 2003; 40: 235-243. 10.1509/jmkr.40.2.235.19225</bibtext> </blist> <blist> <bibl id="bib2" idref="ref33" type="bt">2</bibl> <bibtext> Bainter SA, Curran PJ. Advantages of integrative data analysis for developmental research. Journal of Cognition and Development. 2015; 16: 1-10. 10.1080/15248372.2013.871721. 25642149</bibtext> </blist> <blist> <bibl id="bib3" idref="ref4" type="bt">3</bibl> <bibtext> Bloom HS, Michalopoulos C. When is the story in the subgroups?. Prevention Science. 2013; 14: 179-188. 10.1007/s11121-010-0198-x. 21279547</bibtext> </blist> <blist> <bibl id="bib4" idref="ref15" type="bt">4</bibl> <bibtext> Curran PJ, Hussong AM. Integrative data analysis: The simultaneous analysis of multiple data sets. Psychological Methods. 2009; 14: 81. 10.1037/a0015914. 19485623. 2777640</bibtext> </blist> <blist> <bibl id="bib5" idref="ref44" type="bt">5</bibl> <bibtext> Curran PJ, McGinley JS, Bauer DJ, Hussong AM, Burns A, Chassin L, Zucker R. A moderated nonlinear factor model for the development of commensurate measures in integrative data analysis. Multivariate Behavioral Research. 2014; 49: 214-231. 10.1080/00273171.2014.889594. 25960575. 4423418</bibtext> </blist> <blist> <bibl id="bib6" idref="ref10" type="bt">6</bibl> <bibtext> Dziak JJ, Lanza ST, Tan X. Effect size, statistical power, and sample size requirements for the bootstrap likelihood ratio test in latent class analysis. Structural Equation Modeling: A Multidisciplinary Journal. 2014; 21: 534-552. 10.1080/10705511.2014.919819. 25328371</bibtext> </blist> <blist> <bibl id="bib7" idref="ref41" type="bt">7</bibl> <bibtext> Enders, C. K. (2010). Applied missing data analysis. Guilford press.</bibtext> </blist> <blist> <bibl id="bib8" idref="ref30" type="bt">8</bibl> <bibtext> Fonseca JR, Cardoso MG. Mixture-model cluster analysis using information theoretical criteria. Intelligent Data Analysis. 2007; 11: 155-173. 10.3233/IDA-2007-11204</bibtext> </blist> <blist> <bibl id="bib9" idref="ref12" type="bt">9</bibl> <bibtext> Henry KL, Muthén B. Multilevel latent class analysis: An application of adolescent smoking typologies with individual and contextual predictors. Structural Equation Modeling. 2010; 17: 193-215. 10.1080/10705511003659342. 21057651. 2968712</bibtext> </blist> <blist> <bibtext> Hussong AM, Curran PJ, Bauer DJ. Integrative data analysis in clinical psychology research. Annual Review of Clinical Psychology. 2013; 9: 61-89. 10.1146/annurev-clinpsy-050212-185522. 23394226. 3924786</bibtext> </blist> <blist> <bibtext> Kirby, D. B, & Lepore, G. (2007). Sexual risk and protective factors factors affecting teen sexual behavior, pregnancy, childbearing and sexually transmitted disease: Which are important? Which can you change? Scotts Valley, CA: ETR Associates.</bibtext> </blist> <blist> <bibtext> Lanza ST, Rhoades BL. Latent class analysis: An alternative perspective on subgroup analysis in prevention and treatment. Prevention Science. 2012. 10.1007/s11121-011-0201-1. 22246429. 3372905</bibtext> </blist> <blist> <bibtext> Lee J, Hahm HC. Acculturation and sexual risk behaviors among Latina adolescents transitioning to young adulthood. Journal of Youth and Adolescence. 2010; 39: 414-427. 10.1007/s10964-009-9495-8. 20020189</bibtext> </blist> <blist> <bibtext> Little, R. J, & Rubin, D. B. (2019). Statistical analysis with missing data (Vol. 793). John Wiley & Sons.</bibtext> </blist> <blist> <bibtext> Lukočienė, O, Varriale, R, & Vermunt, J. K. (2010). 6. The simultaneous decision (s) about the number of lower-and higher-level classes in multilevel latent class analysis. Sociological Methodology, 40(1), 247–283.</bibtext> </blist> <blist> <bibtext> Marks AK, Ejesi K, García Coll C. Understanding the US immigrant paradox in childhood and adolescence. Child Development Perspectives. 2014; 8: 59-64. 10.1111/cdep.12071</bibtext> </blist> <blist> <bibtext> Peskin MF, Coyle KC, Anderson PM, Laris BA, Glassman JA, Franks HM, Thiel MA, Potter SC, Unti T, Edwards S, Johnson-Baker K, Cuccaro P, Diamond P, Markham CM, Emery ST. It's Your Game...Keep It Real!: Replication of an evidence-based HIV, STI, and teen pregnancy prevention program in southeast Texas. Journal of Primary Prevention. 2019; 40: 297-323. 10.1007/s10935-019-00549-0. 31028508</bibtext> </blist> <blist> <bibtext> Peskin MF, Shegog R, Markham CM, Thiel M, Baumler ER, Addy RC, Gabay EK, Tortolero Emery S. Efficacy of It's Your Game-Tech: A computer-based sexual health education program for middle school youth. Journal of Adolescent Health. 2015; 56: 515-521. 10.1016/j.jadohealth.2015.01.001</bibtext> </blist> <blist> <bibtext> Potter SC, Coyle KK, Glassman JR, Kershner S, Prince MS. It's Your Game... Keep It Real in South Carolina: A group randomized trial evaluating the replication of an evidence-based adolescent pregnancy and sexually transmitted infection prevention program. American Journal of Public Health. 2016; 106: S60-S69. 10.2105/AJPH.2016.303419. 27689496. 5049477</bibtext> </blist> <blist> <bibtext> Raffaelli, M, Kang, H, & Guarini, T. (2012). Exploring the immigrant paradox in adolescent sexuality: An ecological perspective.</bibtext> </blist> <blist> <bibtext> Rothman, A. J. (2013). Exploring connections between moderators and mediators: Commentary on subgroup analyses in intervention research. Prevention Science, 14(2), 189–192. https://doi.org/10.1007/s11121-012-0333-y</bibtext> </blist> <blist> <bibtext> Rothwell, P. M. (2005). Subgroup analysis in randomised controlled trials: importance, indications, and interpretation. The Lancet, 365(9454), 176–186. https://doi.org/10.1016/S0140-6736(05)17709-5</bibtext> </blist> <blist> <bibtext> Supplee LH, Kelly BC, MacKinnon DM, Barofsky MY. Introduction to the special issue: Subgroup analysis in prevention and intervention research. Prevention Science. 2013; 14: 107-110. 10.1007/s11121-012-0335-9. 23090721</bibtext> </blist> <blist> <bibtext> Tortolero SR, Markham CM, Peskin MF, Shegog R, Addy RC, Escobar-Chaves SL, Baumler ER. It's Your Game, Keep It Real: Delaying sexual behavior with an effective middle school program. Journal of Adolescent Health. 2010; 46: 169-179. 10.1016/j.jadohealth.2009.06.008,PMCID:PMC2818029</bibtext> </blist> <blist> <bibtext> Vasilenko SA, Glassman JR, Kugler KC, Peskin MF, Shegog R, Markham C, Tortolero Emory S, Coyle KK. Examining the effects of an adolescent pregnancy prevention program by risk profiles: A more nuanced approach to program evaluation. Journal of Adolescent Health. 2019; 64: 732-736. 10.1016/j.jadohealth.2018.12.003</bibtext> </blist> <blist> <bibtext> Vermunt JK. Multilevel latent class models. Sociological Methodology. 2003; 33: 213-239. 10.1111/j.0081-1750.2003.t01-1-00131.x</bibtext> </blist> <blist> <bibtext> Vermunt, J. K, & Magidson, J. (2016). Upgrade manual for Latent GOLD 5.1. Belmont, MA: Statistical Innovations.</bibtext> </blist> <blist> <bibtext> Zimmer-Gembeck, M. J, & Helfand, M. (2008). Ten years of longitudinal research on U.S. adolescent sexual behavior: Developmental correlates of sexual intercourse, and the importance of age, gender and ethnic background. Developmental Review, 28(2), 153–224. https://doi.org/10.1016/j.dr.2007.06.001</bibtext> </blist> </ref> <aug> <p>By Sara A. Vasilenko; Omolola A. Odejimi; Jill R. Glassman; Susan C. Potter; Pamela M. Drake; Karin K. Coyle; Christine Markham; Susan Tortolero Emery; Melissa F. Peskin; Ross Shegog; Robert C. Addy and Leslie F. Clark</p> <p>Reported by Author; Author; Author; Author; Author; Author; Author; Author; Author; Author; Author; Author</p> </aug> <nolink nlid="nl1" bibid="bib23" firstref="ref1"></nolink> <nolink nlid="nl2" bibid="bib21" firstref="ref2"></nolink> <nolink nlid="nl3" bibid="bib22" firstref="ref3"></nolink> <nolink nlid="nl4" bibid="bib12" firstref="ref5"></nolink> <nolink nlid="nl5" bibid="bib25" firstref="ref7"></nolink> <nolink nlid="nl6" bibid="bib26" firstref="ref13"></nolink> <nolink nlid="nl7" bibid="bib10" firstref="ref16"></nolink> <nolink nlid="nl8" bibid="bib24" firstref="ref17"></nolink> <nolink nlid="nl9" bibid="bib17" firstref="ref18"></nolink> <nolink nlid="nl10" bibid="bib19" firstref="ref19"></nolink> <nolink nlid="nl11" bibid="bib18" firstref="ref21"></nolink> <nolink nlid="nl12" bibid="bib11" firstref="ref22"></nolink> <nolink nlid="nl13" bibid="bib28" firstref="ref23"></nolink> <nolink nlid="nl14" bibid="bib27" firstref="ref26"></nolink> <nolink nlid="nl15" bibid="bib15" firstref="ref27"></nolink> <nolink nlid="nl16" bibid="bib16" firstref="ref37"></nolink> <nolink nlid="nl17" bibid="bib13" firstref="ref38"></nolink> <nolink nlid="nl18" bibid="bib20" firstref="ref39"></nolink> <nolink nlid="nl19" bibid="bib14" firstref="ref42"></nolink>
Header DbId: eric
DbLabel: ERIC
An: EJ1401542
AccessLevel: 3
PubType: Academic Journal
PubTypeId: academicJournal
PreciseRelevancyScore: 0
IllustrationInfo
Items – Name: Title
  Label: Title
  Group: Ti
  Data: Who Benefits from School-Based Teen Pregnancy Prevention Programs? Examining Multidimensional Moderators of Program Effectiveness across Four Studies
– Name: Language
  Label: Language
  Group: Lang
  Data: English
– Name: Author
  Label: Authors
  Group: Au
  Data: <searchLink fieldCode="AR" term="%22Vasilenko%2C+Sara+A%2E%22">Vasilenko, Sara A.</searchLink> (ORCID <externalLink term="http://orcid.org/0000-0002-1773-8947">0000-0002-1773-8947</externalLink>)<br /><searchLink fieldCode="AR" term="%22Odejimi%2C+Omolola+A%2E%22">Odejimi, Omolola A.</searchLink><br /><searchLink fieldCode="AR" term="%22Glassman%2C+Jill+R%2E%22">Glassman, Jill R.</searchLink><br /><searchLink fieldCode="AR" term="%22Potter%2C+Susan+C%2E%22">Potter, Susan C.</searchLink><br /><searchLink fieldCode="AR" term="%22Drake%2C+Pamela+M%2E%22">Drake, Pamela M.</searchLink><br /><searchLink fieldCode="AR" term="%22Coyle%2C+Karin+K%2E%22">Coyle, Karin K.</searchLink><br /><searchLink fieldCode="AR" term="%22Markham%2C+Christine%22">Markham, Christine</searchLink><br /><searchLink fieldCode="AR" term="%22Emery%2C+Susan+Tortolero%22">Emery, Susan Tortolero</searchLink><br /><searchLink fieldCode="AR" term="%22Peskin%2C+Melissa+F%2E%22">Peskin, Melissa F.</searchLink><br /><searchLink fieldCode="AR" term="%22Shegog%2C+Ross%22">Shegog, Ross</searchLink><br /><searchLink fieldCode="AR" term="%22Addy%2C+Robert+C%2E%22">Addy, Robert C.</searchLink><br /><searchLink fieldCode="AR" term="%22Clark%2C+Leslie+F%2E%22">Clark, Leslie F.</searchLink>
– Name: TitleSource
  Label: Source
  Group: Src
  Data: <searchLink fieldCode="SO" term="%22Prevention+Science%22"><i>Prevention Science</i></searchLink>. 2023 24(8):1535-1546.
– Name: Avail
  Label: Availability
  Group: Avail
  Data: Springer. Available from: Springer Nature. One New York Plaza, Suite 4600, New York, NY 10004. Tel: 800-777-4643; Tel: 212-460-1500; Fax: 212-460-1700; e-mail: customerservice@springernature.com; Web site: https://link.springer.com/
– Name: PeerReviewed
  Label: Peer Reviewed
  Group: SrcInfo
  Data: Y
– Name: Pages
  Label: Page Count
  Group: Src
  Data: 12
– Name: DatePubCY
  Label: Publication Date
  Group: Date
  Data: 2023
– Name: SourceSuprt
  Label: Sponsoring Agency
  Group: SrcSuprt
  Data: Office of Public Health and Science (DHHS), Office of Population Affairs
– Name: NumberContract
  Label: Contract Number
  Group: NumCntrct
  Data: 5PHEPA0000030200
– Name: TypeDocument
  Label: Document Type
  Group: TypDoc
  Data: Journal Articles<br />Reports - Research
– Name: Audience
  Label: Education Level
  Group: Audnce
  Data: <searchLink fieldCode="EL" term="%22Junior+High+Schools%22">Junior High Schools</searchLink><br /><searchLink fieldCode="EL" term="%22Middle+Schools%22">Middle Schools</searchLink><br /><searchLink fieldCode="EL" term="%22Secondary+Education%22">Secondary Education</searchLink>
– Name: Subject
  Label: Descriptors
  Group: Su
  Data: <searchLink fieldCode="DE" term="%22Program+Effectiveness%22">Program Effectiveness</searchLink><br /><searchLink fieldCode="DE" term="%22Pregnancy%22">Pregnancy</searchLink><br /><searchLink fieldCode="DE" term="%22Prevention%22">Prevention</searchLink><br /><searchLink fieldCode="DE" term="%22Middle+School+Students%22">Middle School Students</searchLink><br /><searchLink fieldCode="DE" term="%22Early+Adolescents%22">Early Adolescents</searchLink><br /><searchLink fieldCode="DE" term="%22Sexuality%22">Sexuality</searchLink><br /><searchLink fieldCode="DE" term="%22Student+Characteristics%22">Student Characteristics</searchLink>
– Name: DOI
  Label: DOI
  Group: ID
  Data: 10.1007/s11121-022-01423-y
– Name: ISSN
  Label: ISSN
  Group: ISSN
  Data: 1389-4986<br />1573-6695
– Name: Abstract
  Label: Abstract
  Group: Ab
  Data: Recent research has suggested the importance of understanding for whom programs are most effective (Supplee et al., 2013) and that multidimensional profiles of risk and protective factors may moderate the effectiveness of programs (Lanza & Rhoades, 2012). For school-based prevention programs, moderators of program effectiveness may occur at both the individual and school levels. However, due to the relatively small number of schools in most individual trials, integrative data analysis across multiple studies may be necessary to fully understand the multidimensional individual and school factors that may influence program effectiveness. In this study, we applied multilevel latent class analysis to integrated data across four studies of a middle school pregnancy prevention program to examine moderators of program effectiveness on initiation of vaginal sex. Findings suggest that the program may be particularly effective for schools with USA-born students who speak another language at home. In addition, findings suggest potential positive outcomes of the program for individuals who are lower risk and engaging in normative dating or individuals with family risk. Findings suggest potential mechanisms by which teen pregnancy prevention programs may be effective.
– 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: EJ1401542
PLink https://search.ebscohost.com/login.aspx?direct=true&site=eds-live&db=eric&AN=EJ1401542
RecordInfo BibRecord:
  BibEntity:
    Identifiers:
      – Type: doi
        Value: 10.1007/s11121-022-01423-y
    Languages:
      – Text: English
    PhysicalDescription:
      Pagination:
        PageCount: 12
        StartPage: 1535
    Subjects:
      – SubjectFull: Program Effectiveness
        Type: general
      – SubjectFull: Pregnancy
        Type: general
      – SubjectFull: Prevention
        Type: general
      – SubjectFull: Middle School Students
        Type: general
      – SubjectFull: Early Adolescents
        Type: general
      – SubjectFull: Sexuality
        Type: general
      – SubjectFull: Student Characteristics
        Type: general
    Titles:
      – TitleFull: Who Benefits from School-Based Teen Pregnancy Prevention Programs? Examining Multidimensional Moderators of Program Effectiveness across Four Studies
        Type: main
  BibRelationships:
    HasContributorRelationships:
      – PersonEntity:
          Name:
            NameFull: Vasilenko, Sara A.
      – PersonEntity:
          Name:
            NameFull: Odejimi, Omolola A.
      – PersonEntity:
          Name:
            NameFull: Glassman, Jill R.
      – PersonEntity:
          Name:
            NameFull: Potter, Susan C.
      – PersonEntity:
          Name:
            NameFull: Drake, Pamela M.
      – PersonEntity:
          Name:
            NameFull: Coyle, Karin K.
      – PersonEntity:
          Name:
            NameFull: Markham, Christine
      – PersonEntity:
          Name:
            NameFull: Emery, Susan Tortolero
      – PersonEntity:
          Name:
            NameFull: Peskin, Melissa F.
      – PersonEntity:
          Name:
            NameFull: Shegog, Ross
      – PersonEntity:
          Name:
            NameFull: Addy, Robert C.
      – PersonEntity:
          Name:
            NameFull: Clark, Leslie F.
    IsPartOfRelationships:
      – BibEntity:
          Dates:
            – D: 01
              M: 01
              Type: published
              Y: 2023
          Identifiers:
            – Type: issn-print
              Value: 1389-4986
            – Type: issn-electronic
              Value: 1573-6695
          Numbering:
            – Type: volume
              Value: 24
            – Type: issue
              Value: 8
          Titles:
            – TitleFull: Prevention Science
              Type: main
ResultId 1