View in EDS HTML Full Text PDF Full Text

Unpacking Trauma Exposure Risk Factors and Differential Pathways of Influence: Predicting Postwar Mental Distress in Bosnian Adolescents

Saved in:
Bibliographic Details
Title: Unpacking Trauma Exposure Risk Factors and Differential Pathways of Influence: Predicting Postwar Mental Distress in Bosnian Adolescents
Language: English
Authors: Layne, Christopher M., Olsen, Joseph A., Baker, Aaron, Legerski, John-Paul, Isakson, Brian, Pasalic, Alma, Durakovic-Belko, Elvira, Dapo, Nermin, Campara, Nihada, Arslanagic, Berina, Saltzman, William R., Pynoos, Robert S.
Source: Child Development. Jul-Aug 2010 81(4):1053-1076.
Availability: Wiley-Blackwell. 350 Main Street, Malden, MA 02148. Tel: 800-835-6770; Tel: 781-388-8598; Fax: 781-388-8232; e-mail: cs-journals@wiley.com; Web site: http://www.wiley.com/WileyCDA/
Peer Reviewed: Y
Page Count: 24
Publication Date: 2010
Document Type: Journal Articles
Reports - Research
Descriptors: Intervention, Adolescents, Factor Analysis, War, Posttraumatic Stress Disorder, Predictor Variables, Mental Disorders, Stress Variables, At Risk Persons, Influences, Comparative Analysis, Causal Models, Adjustment (to Environment), Coping, Counseling Techniques, Counseling Effectiveness, Foreign Countries, Child Development
Geographic Terms: Bosnia and Herzegovina
DOI: 10.1111/j.1467-8624.2010.01454.x
ISSN: 0009-3920
Abstract: Methods are needed for quantifying the potency and differential effects of risk factors to identify at-risk groups for theory building and intervention. Traditional methods for constructing war exposure measures are poorly suited to "unpack" differential relations between specific types of exposure and specific outcomes. This study of 881 Bosnian adolescents compared both "common factor-effect indicator" (using exploratory factor analysis) versus "composite causal-indicator" methods for "unpacking" dimensions of war exposure and their respective paths to postwar adjustment outcomes. The composite method better supported theory building and most intervention applications, showing how multitiered interventions can enhance treatment effectiveness and efficiency in war settings. Used together, the methods may unpack the elements and differential effects of "caravans" of risk and promotive factors that co-occur across development.
Abstractor: As Provided
Entry Date: 2010
Accession Number: EJ890275
Database: ERIC
Full text is not displayed to guests.
FullText Links:
  – Type: pdflink
    Url: https://content.ebscohost.com/cds/retrieve?content=AQICAHj0k_4E0hTGH8RJwT4gCJyBsGNe_WN95AvKlDbXJGqwxwEn1JbKuKdrWYWbLKKoCS1-AAAA4TCB3gYJKoZIhvcNAQcGoIHQMIHNAgEAMIHHBgkqhkiG9w0BBwEwHgYJYIZIAWUDBAEuMBEEDDcWkSWQ4kAzPRa8kQIBEICBmXyWSdmGAfpkkJXLn5wSpDlDqVfNqhXmn8KEp_1MZx2FTV_rG-zXvgEX_RPjv2V9crQfllYCberYSYL33xe3gYVtDEupogvLpCBoYV4Yacx6sjsiPGKLkX99x2gSP7kSFvo8OAo8HM5czgZMk-2UNeSQRE38TBEhaWrAOBCvfrawGabNGtH0RP15QH2J43zzt5IqjAfzB90f9w==
Text:
  Availability: 1
  Value: <anid>AN0052214229;cdv01jul.10;2019Jun03.14:50;v2.2.500</anid> <title id="AN0052214229-1">Unpacking Trauma Exposure Risk Factors and Differential Pathways of Influence: Predicting Postwar Mental Distress in Bosnian Adolescents. </title> <p>Methods are needed for quantifying the potency and differential effects of risk factors to identify at‐risk groups for theory building and intervention. Traditional methods for constructing war exposure measures are poorly suited to "unpack" differential relations between specific types of exposure and specific outcomes. This study of 881 Bosnian adolescents compared both common factor–effect indicator (using exploratory factor analysis) versus composite causal–indicator methods for "unpacking" dimensions of war exposure and their respective paths to postwar adjustment outcomes. The composite method better supported theory building and most intervention applications, showing how multitiered interventions can enhance treatment effectiveness and efficiency in war settings. Used together, the methods may unpack the elements and differential effects of "caravans" of risk and promotive factors that co‐occur across development.</p> <p>In the aftermath of wars and other catastrophic events, questions concerning whom and how to help—that is, <emph>how to best identify at‐risk groups and allocate scarce mental health resources</emph>—confront every practitioner and agency that strives to promote the recovery of affected youths and families. Efforts to address these questions have contributed to our growing knowledge base by improving the methodological rigor of disaster research ([<reflink idref="bib49" id="ref1">49</reflink>]), building theory by identifying the correlates and sequelae of war exposure (e.g., [<reflink idref="bib18" id="ref2">18</reflink>]; [<reflink idref="bib40" id="ref3">40</reflink>]), and field‐testing interventions (e.g., [<reflink idref="bib31" id="ref4">31</reflink>]). These studies have identified a broad range of types of distress in war‐exposed youths, including posttraumatic stress disorder (PTSD), depression and other internalizing disorders, externalizing behaviors, impaired school performance, and disturbances in personality formation and moral development (see [<reflink idref="bib3" id="ref5">3</reflink>]; [<reflink idref="bib19" id="ref6">19</reflink>]; [<reflink idref="bib20" id="ref7">20</reflink>]; [<reflink idref="bib54" id="ref8">54</reflink>]).</p> <p>Although the literature identifies a general dose–response relation between the magnitude of war exposure and youth postwar distress, studies are providing growing evidence of <emph>differential relations</emph> between specific trauma‐related risk factors and specific posttraumatic adjustment outcomes ([<reflink idref="bib29" id="ref9">29</reflink>])<emph>.</emph> In their review, [<reflink idref="bib20" id="ref10">20</reflink>] note that for at least half of war‐exposed youths, PTSD reactions decrease on their own in the postwar aftermath, whereas other youths continue to meet full PTSD diagnostic criteria or experience delayed‐onset stress reactions. The persistence of these reactions is linked to both the degree of severity of initial trauma exposure as well as the presence of postwar family stressors, such as maternal dysfunction and poverty. Such findings underscore the influential roles played not only by the specific <emph>types</emph> of trauma and loss to which youths are exposed but also of associated moderators and mediators of youth postwar adjustment (see [<reflink idref="bib42" id="ref11">42</reflink>]). For example, evidence that having one's parent killed and the disappearances of loved ones are especially traumatogenic ([<reflink idref="bib54" id="ref12">54</reflink>]) suggests that war‐related causal risk factors may differ significantly in their potency. In addition, a gender‐moderated effect is evident in findings that war‐exposed girls generally manifest higher levels of anxiety and mood symptoms than boys, whereas boys manifest higher rates of externalizing behavior ([<reflink idref="bib50" id="ref13">50</reflink>]; [<reflink idref="bib54" id="ref14">54</reflink>]). The importance of mediators is underscored by evidence that different war‐related causal risk factors may exert their effects on postwar outcomes through different pathways of influence. For example, [<reflink idref="bib19" id="ref15">19</reflink>] report that PTSD symptoms in refugee youths are more strongly linked to earlier war trauma and subsequent resettlement strain, whereas depression symptoms are more strongly linked to recent stressful life events and current maternal mental health. Taken together, these studies constitute important early progress in understanding the mechanisms and processes that underlie the etiology and clinical course of youth psychosocial maladjustment following war‐related trauma and loss.</p> <p>However, the progress made in the war and disaster literature in identifying the pathways through which specific types of trauma exposure differentially influence posttraumatic psychosocial adjustment is impeded by four methodological practices. As discussed by [<reflink idref="bib47" id="ref16">47</reflink>], [<reflink idref="bib48" id="ref17">48</reflink>] and [<reflink idref="bib3" id="ref18">3</reflink>], these practices include: (a) the inappropriate use of the <emph>common latent factor with effect indicators model</emph> (i.e., factor analysis) to derive the dimensionality of trauma exposure measures; (b) inadequate event sampling in scale construction (e.g., selecting a best <emph>sample</emph> of "internally consistent" events vs. a comprehensive event <emph>census</emph>), including the practice of importing existing trauma exposure measures into new settings that differ markedly from those for which the measures were created; (c) the use of total‐scale summative scoring (e.g., creating a unit‐weighted "total war exposure" score that intermixes many different types of trauma); and (d) the use of scale reliability indices (e.g., Cronbach's alpha; [<reflink idref="bib15" id="ref19">15</reflink>]) that are incompatible with the psychometric assumptions or particular applications of the measurement model being used.</p> <p>Consequently, although the literature provides compelling evidence that war exposure is generally harmful to youth adjustment, its ability to explain <emph>which</emph> types of exposure are most harmful, for <emph>which</emph> outcomes, for <emph>whom, how</emph> they transmit their effects, and <emph>how</emph> this knowledge should guide intervention, is much more limited ([<reflink idref="bib36" id="ref20">36</reflink>]). Accordingly, although noting a general dose–response relation between trauma exposure and mental health, [<reflink idref="bib3" id="ref21">3</reflink>] offer the caveat that different types of trauma relate differentially to PTSD and other outcomes, such that some traumatic events have only moderate or negligible long‐term effects whereas others are devastating. The authors note that the literature is largely unable to explain how or why this occurs. Such shortcomings recently led an APA task force to conclude that although some posttrauma <emph>indicators of risk</emph> have been identified, no method exists for accurately predicting whether specific youths will recover on their own or instead require some intervention. The task force called for the development of "well‐validated risk assessment tools that can be feasibly implemented <emph>in diverse settings and for diverse traumatic events and that will help identify the high‐risk youth and families who are in need of clinical services</emph>" ([<reflink idref="bib2" id="ref22">2</reflink>]).</p> <p>Developing risk assessment methods and tools for quantifying the potency and differential effects of specific types of trauma exposure that are valid and useful across diverse settings carries great promise for advancing the field, especially through building theory, enhancing risk screening and intervention, and guiding social policy ([<reflink idref="bib24" id="ref23">24</reflink>]). Specifically, a method that aggregates into the same test dimension <emph>equifinal</emph> causal risk factors (i.e., factors that lead to the same causal consequences; [<reflink idref="bib53" id="ref24">53</reflink>]) will help to quantify the differential effects of specific types of risk factors on specific outcomes. This practice can build theory by clarifying the etiological significance of specific types of trauma and loss for specific outcomes and the differential pathways of influence that interconnect them. In turn, clarifying <emph>who is at risk for what outcome</emph> will improve risk screening and triage by assigning groups with different types of trauma and loss (e.g., traumatic bereavement) to intervention modules that target those types of distress for which members are at greatest risk (e.g., maladaptive grief reactions). Equally valuable, knowledge of key pathways of influence can improve the effectiveness and efficiency of interventions by identifying "high‐value" targets for intervention. This will guide treatments in addressing not only "endpoint" distress—related <emph>outcomes</emph> (e.g., PTSD, depression, traumatic grief), but also <emph>key causal contributors to</emph> those outcomes (e.g., trauma and loss reminders).</p> <p>Given these advantages, we propose that it is a propitious time to reevaluate the ways in which war exposure is measured. Much can be gained by supplementing the traditional test construction method of factor analysis and related approaches (e.g., [<reflink idref="bib18" id="ref25">18</reflink>]) that "unpack" trauma exposure into dimensions that <emph>model the structure of event co‐occurrences</emph>, with a method that "unpacks" exposure into dimensions that <emph>model the occurrences of specific types of trauma and loss and their differential links to psychosocial consequences</emph>. As part of this reevaluation, candidate methods for constructing measures of trauma exposure should be evaluated according to their promise for strengthening research methodology, theory, and intervention. Specifically, a method promotes methodological rigor through proper model specification, improved model fit, increased variance explained, strengthened causal inference, and greater clarification of the "real‐world" structure of relations between variables ([<reflink idref="bib44" id="ref26">44</reflink>]). A method furthers theory building by enhancing the description, explanation, and prediction of ways in which stress‐related disorders originate, persist, abate, recur, lead to dysfunction, and are counteracted by promotive and protective factors within and across specific developmental periods. Last, a method enhances intervention by improving the accuracy of risk detection and triage, identifying appropriate treatment components and modalities (e.g., classroom vs. small group treatment; Layne et al., 2008), and prioritizing intervention foci, such as targeting family versus peer versus adult mentor social support in a postwar psychosocial program ([<reflink idref="bib34" id="ref27">34</reflink>]).</p> <hd id="AN0052214229-2">Some Essential Details: The Common Factor–Effect Indicator and Composite–Causal Indicator Mod...</hd> <p>Developing a method for accurately predicting who is at risk for what outcomes will involve evaluating two fundamentally different approaches to conceptualizing, measuring, and modeling trauma exposure. These are the <emph>common latent factor with effect indicators model</emph> (i.e., reflective–indicator model; [<reflink idref="bib38" id="ref28">38</reflink>]), which emphasizes the <emph>structure of co‐occurrence among events</emph>, and the <emph>composite variable with causal indicators model</emph> (i.e., formative‐indicator model; [<reflink idref="bib23" id="ref29">23</reflink>]), which emphasizes the <emph>structure of event occurrences and their respective consequences</emph> (e.g., personal income and education both contribute to socioeconomic status). [<reflink idref="bib39" id="ref30">39</reflink>] propose that the two models can be differentiated in four ways: (a) Effect indicators <emph>manifest or reflect</emph> the underlying construct, whereas causal indicators <emph>define</emph> the composite. (b) Effect indicators (assuming equivalent reliability) are <emph>interchangeable</emph>, whereas each (nonredundant) causal indicator is a <emph>unique constitutive element of</emph> the composite. (c) <emph>Covariation</emph> among effect indicators is both expectable and desirable due to the influence of a shared common factor, whereas causal indicators are not necessarily expected to covary (indeed, high covariation may reflect excessive redundancy). And (d) effect indicators, but not causal indicators, are expected to <emph>share a common set of causal precursors and causal consequences</emph> because they emanate from a common causal origin. The models can also be rationally differentiated by considering whether a change in the <emph>common factor</emph> necessarily precedes a change in its "effect" indicators (e.g., a siege <emph>causes</emph> a city to be surrounded and shelled), or conversely, whether a change in the "causal" indicators necessarily precedes a change in the resulting <emph>composite</emph> (surrounding and shelling a city <emph>causes or creates</emph> a siege).</p> <p>The <emph>common factor–effect indicator</emph> model (Figure 1A) derives from classical test theory and undergirds many traditional test construction tools, including exploratory factor analysis (EFA), confirmatory factor analysis (CFA), and Cronbach's alpha ([<reflink idref="bib15" id="ref31">15</reflink>]; see also [<reflink idref="bib6" id="ref32">6</reflink>]). The common factor model, often paired with alpha, has been in use for over 20 years to derive "internally consistent" dimensions of trauma exposure (e.g., [<reflink idref="bib18" id="ref33">18</reflink>]). A core assumption of the model is that the indicators are <emph>homogeneous</emph> (i.e., share a <emph>common causal origin</emph>; [<reflink idref="bib7" id="ref34">7</reflink>]). Arrows leading from the latent common factor to each "effect" indicator place the locus of causation within the factor, signifying that it explains the variances of the indicators. The proportion of each indicator's variance explained by the common factor variance is denoted by its squared factor loading (). The model‐implied covariance between any two indicators is reproduced by multiplying their factor loadings (; [<reflink idref="bib38" id="ref35">38</reflink>]), signifying that <emph>indicators covary because, and to the extent that, they are jointly caused by causal processes underlying the common factor</emph>. <emph>U</emph> denotes each item's uniqueness as comprised of both unsystematic (i.e., random) error variance and systematic variance not shared with the common factor (e.g., method variance). Although alpha assumes both item homogeneity (a strict assumption) and tau equivalence (equivalent factor loadings, a stricter assumption), it tests neither assumption but is instead a simple function of the number of test items and the average interitem bivariate correlation ([<reflink idref="bib55" id="ref36">55</reflink>]).</p> <p>Graph: 1 (A) A common latent factor–effect indicators model. (B) A composite–causal indicators model. (C) Differentiated "real‐world" causal structure between four causal risk factors and two outcomes. (D) A composite–causal indicators model of the variables in (C). (E) A common latent factor–effect indicators model of the variables in (C). (F) Incorrect causal inference regarding the variables in (C) produced by statistical artifact: Dropping friend killed forces both life threat indicators to act as its proxies. (G) A hierarchical confirmatory common latent factor–effect indicators model. (H) A bifactor confirmatory common latent factor–effect indicators model.</p> <p>In contrast, the <emph>composite–causal indicator model</emph> (Figure 1B; [<reflink idref="bib23" id="ref37">23</reflink>]) does not assume indicator homogeneity. Rather, each of the indicators contributes to the composite (denoted by arrows leading from the indicators to the composite), placing the locus of causation in the indicators. Because each indicator is an independent contributor to the composite, covariation or "internal consistency" among indicators (Figure 1B, paths a–f) is of peripheral interest ([<reflink idref="bib16" id="ref38">16</reflink>]). The error variance is specified at 0, signifying that composite values are a linear combination of the indicators. The composite thus mediates the predictive effects of its indicators on outcomes and constrains each indicator to have the same proportional influence on all outcomes (e.g., Indicator 1 exerts equivalent predictive effects on Outcomes A and B in Figure 1B; [<reflink idref="bib21" id="ref39">21</reflink>]). The model is flexible: Causal risk factors that are equifinal in relation to specified outcomes but known a priori to differ in their respective potencies can be aggregated into <emph>fixed weight composites</emph> (in which weights are fixed to different magnitudes to model differential effects); alternatively, risk factors of unknown potencies can be aggregated into <emph>unknown weight composites</emph>. Correctly specified composites function as "unitary entities" in relation to other variables (e.g., causal indicators theorized to be equifinal should relate in similar ways to their theorized causal consequences; [<reflink idref="bib59" id="ref40">59</reflink>]). Accordingly, equifinal causal risk factors (e.g., father's death by shelling, aunt's death by suicide, brother's death by car accident) may exert similar effects on a specified set of outcomes (PTSD, traumatic grief) <emph>irrespective</emph> of whether they share a common origin, co‐occur, covary, or manifest "internal consistency." Thus, <emph>when the aim is to model links between event occurrences and their respective consequences</emph>, risk factors that <emph>are</emph> equifinal in reference to focal outcomes may be assigned to the same composite, whereas risk factors that co‐occur, covary, and manifest internal consistency, but are <emph>not</emph> equifinal, may be assigned to different composites. Given its focus on modeling consequences, the composite approach is well suited for modeling not only relations between "real‐world" objective events (e.g., paths from war‐time losses to postwar hardships) but also between objective events and subjective experiences (e.g., paths between trauma and distress). The degree of discrimination used in combining events into dimensions depends on the desired degree of specificity in measuring their differential effects.</p> <hd id="AN0052214229-3">Contrasting the Two Approaches: Conceptualizing, Measuring, and Modeling Trauma Exposure</hd> <p>A comparison of the common factor versus composite models should be qualified by two caveats. First, neither is entirely theory driven nor entirely data driven: Although strongly empirical, the common factor approach requires rational judgment pertaining to such activities as selecting extraction and rotation algorithms, specifying factor‐loading cutoff values, and factor interpretation. Further, the composite approach should be guided by both theory and empirical evidence concerning links between causal risk factors and their consequences. Thus, the approaches may be conceptualized as lying on opposing ends of a bipolar continuum ranging from greater emphasis on empirical evidence to greater emphasis on theory.</p> <p>A second caveat is that the strengths and weaknesses of each model depend on the application. Consider the "real‐world" differentiated model in Figure 1C, in which all risk factors are equifinal for PTSD symptoms, but only <emph>father killed</emph> and <emph>close friend killed</emph> are equifinal for traumatic grief. The utility of each approach rests on which portion of the model—the <emph>structure of relations between co‐occurring risk factors</emph>, versus the <emph>structure of the occurrence of specific causal risk factors and their differential consequences—</emph>is of primary interest. Because the common factor model draws exclusively on the <emph>covariances between risk factor indicators</emph> (paths a–f), it is useful for aggregating clusters of traumatic events that co‐occurred sufficiently to produce coherent factors. This focus on the structure <emph>among</emph> risk factors gives the common factor model three potential strengths. First, each indicator is a risk marker for the other indicators. Second, all indicators are <emph>general</emph> risk markers for all outcomes ([<reflink idref="bib2" id="ref41">2</reflink>]). Third, the model can "unpack" complex events (e.g., wars) into often‐meaningful clusters of co‐occurring events (e.g., sieges). However, because it is designed to extract "best internally consistent"<emph>samples</emph> of covarying events, the model may drop events that <emph>occur, but do not systematically co‐occur</emph>, that would be captured via event census ([<reflink idref="bib48" id="ref42">48</reflink>]). Indeed, the only conditions under which the common factor model is likely to produce "clean" equifinal test dimensions are when all equifinal events (e.g., all types of traumatic bereavement) happen to covary strongly, and all <emph>non</emph>equifinal events (bereavement vs<emph>.</emph> material loss) do not. Such conditions are unlikely to emerge in war‐related settings, although there may be rare exceptions (cf. [<reflink idref="bib40" id="ref43">40</reflink>]).</p> <p>Figure 1D presents a composite model of the relations in Figure 1C. Given its emphasis on causes and consequences, including differential relations (see paths g–l), this approach conducts a <emph>census</emph> of trauma exposure types for two reasons. First, each (nonredundant) causal indicator is a constitutive element of the composite and thus indispensable ([<reflink idref="bib6" id="ref44">6</reflink>]). Second, events that occur but do not <emph>co‐occur and covary</emph> with other risk factors may nevertheless exert potent influences (see Figure 1C, paths j and l from <emph>father killed</emph>, which does not covary). The "data input" used by the composite model thus centers on <emph>paths linking risk factors to their theorized consequences</emph> (Figure 1C, paths g–l). The composite model is thus well suited not only to aggregate equifinal risk factors into common dimensions, but also to assign causally diverse events to different composites to facilitate the detection of differential relations.</p> <hd id="AN0052214229-4">Model–Application Mismatch: Using the Common Factor Model to Unpack Risk Factor–Consequence L...</hd> <p>Given its long‐standing use (see [<reflink idref="bib3" id="ref45">3</reflink>]; [<reflink idref="bib26" id="ref46">26</reflink>]), potential drawbacks that may arise when the common factor model is used <emph>in applications focusing on the causal consequences of war exposure</emph> should be considered. A model that "packs" types of war exposure into dimensions based on their covariance structure (Figure 1C, paths a–f) may indeed be ill suited to "unpack" the differential causal effects <emph>of</emph> specific types of exposure on specific consequences (paths g–l). In such applications, any of six conditions may create "model–application mismatches" that dilute, distort, and obscure "real‐world" differential relations. These include: (a) The aim is to detect differential relations between specific types of causal risk factors and specific outcomes. (b) Events that co‐occur are <emph>not</emph> equifinal (Figure 1C, paths g–j). (c) Events that do not co‐occur <emph>are</emph> equifinal (paths g, h, and <emph>k</emph>). (d) Ambiguity in test scores (produced by combining different causal risk factors that exert different effects) is costly. (e) <emph>Sampling</emph>, rather than conducting a <emph>census</emph> of, trauma types will create costly false negatives (e.g., youths whose fathers are killed are not detected), costly false positives (youths needing only general support receive specialized services), or lead to incorrect conclusions about causal relations. And (f) researchers wish to create variables and assessment tools that have predictive utility at the individual case level and are transportable across diverse settings and diverse traumatic events ([<reflink idref="bib2" id="ref47">2</reflink>]).</p> <p>Figure 1E depicts a scenario that may arise when the common factor model is used to unpack the "real‐world" consequences of war exposure (as found in Figure 1C) and contains three major misspecification errors. These include the <emph>violation of the assumption of homogeneity</emph>, the <emph>loss of differential relations</emph>, and the <emph>exclusion of an influential causal risk factor</emph>. The first error, <emph>violation of indicator homogeneity</emph>, arises because <emph>Siege Exposure</emph> is modeled as the <emph>cause</emph> of shooting and shelling, when instead <emph>these acts define and initiate a siege and ceasing them will cause it to end</emph> ([<reflink idref="bib39" id="ref48">39</reflink>]). This fact repudiates the assumption of indicator homogeneity and the associated need to aggregate (or disaggregate) indicator‐level risk factors into test dimensions on the basis of their covariance structure. Why should causal risk factors (including those that exert equifinal effects) necessarily co‐occur or covary if they do not share a common causal origin? Indeed, modeling war exposure as a homogeneous latent common factor is incompatible with PTSD diagnostic criteria because the etiology of the disorder lies in observable <emph>indicator‐level</emph> Criterion A1 events, such as mortar shell attacks, rather than unobservable "latent" causal processes. The indicators thus do not <emph>share</emph> a common origin with outcomes (e.g. "latent" sieges do not cause shelling on one hand <emph>and</emph> PTSD on the other); rather, <emph>indicators are etiological risk factors for PTSD</emph> (<emph>shelling</emph> can cause PTSD). Causally misspecified common factor models may thus lead to the logical fallacy of <emph>affirming the consequent</emph>: (a) If test items are homogeneous, then they load a common factor and show evidence of internal consistency. (b) This set of indicators loads a common factor and produces a high alpha value. (c) <emph>Therefore, these indicators are homogeneous and share the same causal consequences</emph> ([<reflink idref="bib39" id="ref49">39</reflink>]). Thus, well‐intentioned researchers may create a test of "siege exposure" and attempt to model, assess, or treat its "sequelae."</p> <p>The second specification error found in Figure 1E is the <emph>loss of differentiated relations</emph> (see Figure 1C, paths k and l). Sieges and other campaigns often generate clusters of co‐occurring events (Figure 1C, paths b, d, and e) that a common factor approach will aggregate. However, covariation between types of war exposure does not reflect indicator homogeneity, but rather, <emph>perceived event co‐occurrence</emph>s. "Packing" types of war exposure into dimensions on the basis of their covariance structure is thus likely to intermix causally diverse (i.e., nonequifinal) risk factors into the same dimension. Such causally diverse dimensions are especially likely to emerge in inherently complex war and disaster settings where many different types of trauma and loss co‐occur. Causally diverse test dimensions are of concern for two reasons. First, creating summative ("unit‐weighted") aggregate variables to model or assess the "sequelae" of a given dimension of war exposure implies that <emph>all indicators are equifinal and equally potent</emph> causal risk factors for all its modeled sequelae<emph>.</emph> For example, summative scoring of the model in Figure 1E implies that both life‐threat events and <emph>friend killed</emph> are <emph>equivalent and interchangeable</emph> causal risk factors for both PTSD and traumatic grief ([<reflink idref="bib39" id="ref50">39</reflink>]). Second, pooling the variances of causally diverse risk factors into the common factor variance (thereby increasing <emph>intradimension</emph> causal diversity) decreases their sensitivity to detecting the differential effects of different types of risk factors. Indeed, major discrepancies may arise between the <emph>causal effects</emph> of risk factors in the "real world" and the <emph>magnitudes of their factor loadings</emph>. In Figure 1C, the effects of the life‐threat items on traumatic grief are negligible, whereas <emph>friend killed</emph> and <emph>father killed</emph> may exert moderate and potent effects, respectively. However, the covariances among indicators (paths a–f) may produce large loadings for both life‐threat items and moderate and negligible loadings for <emph>friend killed</emph> and <emph>father killed</emph>, respectively. The common factor variance may thus consist mostly of variance extracted from causally weak indicators. This may both reduce the strength of paths from the common factor to outcomes (Figure 1E, path b) and lead to erroneous conclusions regarding the potencies of the respective indicators or of that dimension.</p> <p>The third misspecification error in Figure 1E is the <emph>loss of a potent causal risk factor</emph> (<emph>father killed</emph>) due to its lack of covariation with other risk factors (Figure 1C, paths a, c, and f). Because it extracts variables based on their covariance structure, the common factor model is susceptible to dropping two types of events. These include <emph>events that occur but do not systematically co‐occur</emph> with other modeled risk factors (e.g., random or unrelated events), and <emph>events whose measurement is compromised by statistical artifact</emph> (e.g., restricted variance due to floor effects). The exclusion of causal risk factors may reduce proportions of explained variance (reducing predictive precision) and bias model parameters in ways that overestimate the vulnerability of exposed subgroups (due to "unexplained" elevations in distress scores; [<reflink idref="bib48" id="ref51">48</reflink>]). Omitting causal risk factors also compromises test validity by reducing the correspondence between a test and the "real‐world" structure among variables it purports to measure ([<reflink idref="bib8" id="ref52">8</reflink>]) and can decrease the sensitivity of screening instruments in ways that lead to false negatives and poor matches of individuals to treatments ([<reflink idref="bib24" id="ref53">24</reflink>]; [<reflink idref="bib47" id="ref54">47</reflink>]). Omitted causal risk factors can also create spurious statistical artifacts. Figure 1F depicts how distortions of the "real‐world" scenario in Figure 1C can lead to faulty conclusions about causal relations. Statistically independent <emph>father killed</emph> does not covary, whereas high endorsement rates for <emph>friend killed</emph> produce a ceiling effect that artificially restricts its covariance with both life‐threat items <emph>even though these three types of events co‐occur in that setting</emph>. The exclusion of <emph>friend killed</emph> forces both life‐threat items to act as its proxies. The resulting biased model parameters may undermine theory building by <emph>under</emph>estimating the etiological significance of traumatic death in relation to traumatic grief, while <emph>over</emph>estimating that of life threat ([<reflink idref="bib48" id="ref55">48</reflink>]). The model may also impede intervention by increasing false negatives ("high‐risk" bereaved youths are not detected) and false positives (youths exposed to low‐potency risk factors receive specialized treatment components).</p> <hd id="AN0052214229-5">Appropriate Use of the Common Factor–Effect Indicator Model to Clarify Differential Relations</hd> <p>Notwithstanding these caveats, we emphasize the utility of the common factor model for clarifying differential relations among multidimensional constructs whose causal structure is compatible with indicator homogeneity (e.g., [<reflink idref="bib25" id="ref56">25</reflink>]). These include both the <emph>hierarchical</emph> and the <emph>bifactor</emph> confirmatory factor analytic models (Figures 1G and 1H; [<reflink idref="bib10" id="ref57">10</reflink>]). Both models allow a priori model specification of hypothesized dimensions and carry a lower risk of bias in model parameters than EFA ([<reflink idref="bib37" id="ref58">37</reflink>]). The latent structure in both models consists of a general factor and one or more specific components for different subsets of items. Although "specific" components are often used to model potentially confounding methodological effects, they can also model multidimensional structures among "real‐world" variables. In contrast to traditional common factor models that contain multiple <emph>separate</emph> factors, the hierarchical and bifactor models each allows an item to reflect both general characteristics shared with all of the other indicators and characteristics unique to specific item subsets. The design of the bifactor model ensures that the latent‐specific components are uncorrelated with one another and with the general factor. These features make the models useful for studying the breadth and specificity of complex constructs, including <emph>convergent validity</emph> (via the general factor) and <emph>discriminant validity</emph> (via the specific components) in multifaceted conceptual domains. The models permit testing of measurement and structural hypotheses relating to differential relations while addressing potential collinearity among latent components and controlling for broader shared variance. Although the hierarchical and bifactor models are difficult to distinguish on purely empirical grounds, each embodies a distinctive theoretical structure that may be plausible within a given setting. Moreover, the common factor approach can also, like the composite–causal indicator approach, prioritize the clarification of structural relations between constructs over internal consistency. [<reflink idref="bib37" id="ref59">37</reflink>] demonstrate that the validity of common factor models often increases when indicators are selected (even those of "poor" psychometric quality) that clearly demarcate construct domains and boundaries between constructs. Conversely, selecting only highly correlated indicators in the interests of maximizing internal consistency may lead to test bias and distorted estimates of the construct's relations with other constructs.</p> <hd id="AN0052214229-6">A Comparison of the Common Factor and Composite‐Based Approaches in Postwar Bosnian Youth</hd> <p> <bold>Study rationale. </bold> Given the potential benefits of each approach, as well as potential problems arising from model–application mismatches, this study compared the performance of a common factor–effect indicator model (using EFA) versus a composite–causal indicator model. The study focused on the potential of each model to (a) build theory capable of explaining differential risk factor–outcome relations, (b) predict which groups are at risk for specific types of distress, and (c) increase the effectiveness and efficiency of interventions. Because low endorsement rates restricted the variances of many war exposure items (e.g., personal injury), the composite approach relied only on theory and prior knowledge to assign specific types of trauma to exposure dimensions. Further, because the cross‐sectional design precluded causal inference, we investigated only the <emph>predictive effects</emph> of risk factors (including theorized mediators) on outcomes. Last, given space limitations and our focus on "unpacking" the covariance structure and consequences of risk factors, no gender effects were tested.</p> <p> <bold>Study setting. </bold> The 1991–1995 civil war in the Former Yugoslavia exemplifies the brutal nature of modern armed conflict and its stressful aftermath. The 1992–1995 Bosnian conflict led to the traumatic deaths of close to 100,000 people, approximately 41% of whom were civilians, and the displacement of 1,800,000 people, many of whom faced years of subsequent hardship as internally displaced persons or war refugees ([<reflink idref="bib1" id="ref60">1</reflink>]). The war led to widespread traumatic deaths, physical injury, traumatic disappearances, major disruptions in social institutions, and brutal mistreatment through acts of assault, torture, rape, and starvation inflicted on detained civilians and prisoners of war. Many military strategies directly targeted civilians, including sieges of cities and protected enclaves, ethnic cleansing campaigns, and genocidal operations. These campaigns generated widespread exposure to direct life threat, the witnessing of traumatic deaths and serious injury, deprivations in basic necessities, exposure to the elements, and the forced exodus of entire regions in refugee flights ([<reflink idref="bib51" id="ref61">51</reflink>]).</p> <p> <bold>Study design and hypotheses. </bold> Data were collected in the fall of 2000 (nearly 5 years postwar) in a screening survey used in a postwar psychosocial program for war‐exposed adolescents ([<reflink idref="bib31" id="ref62">31</reflink>]). The survey was administered in nine heavily war‐exposed secondary schools in Central Bosnia and was based on a developmental model of childhood traumatic stress ([<reflink idref="bib52" id="ref63">52</reflink>]). Survey items measured a broad range of traumatic experiences, theorized mediators of youth postwar adjustment including trauma reminders and adversities ([<reflink idref="bib35" id="ref64">35</reflink>]), and distress‐related outcomes. Because the common factor approach models the <emph>structure of event co‐occurrences</emph>, we predicted that EFA‐derived factors would reflect major military strategies employed in the Bosnian conflict (e.g., sieges, ethnic cleansing; Hypothesis 1) and would substantively differ in their composition compared with the rationally derived composite dimensions (Hypothesis 2). We also theorized that highly "traumatogenic" war‐time risk factors (e.g., life threat, traumatic death) would exert direct predictive effects, whereas the predictive effects of less‐traumatogenic war‐time risk factors would be fully mediated by postwar trauma reminders and adversities. We thus predicted that the <emph>partially mediated</emph> version of each model would fit better than its <emph>fully mediated</emph> (indirect effects only) counterpart (Hypothesis 3). Last, we predicted that the composite–causal indicator model would show more differentiated relations between theorized causal risk factors (including <emph>prewar trauma</emph> and different types of <emph>war‐time trauma</emph>), theorized mediators (including <emph>postwar adversities</emph> and <emph>trauma reminders</emph>), and outcomes (including <emph>posttraumatic stress</emph>, <emph>depressive</emph>, <emph>and traumatic grief reactions</emph>; and <emph>functional impairment</emph>) than the common factor model (Hypothesis 4).</p> <hd id="AN0052214229-7">Method</hd> <p></p> <hd id="AN0052214229-8">Measures</hd> <p>All measures were reviewed for cultural appropriateness and forward‐ and back‐translated by doctoral students in psychology at the University of Sarajevo. Measures were then combined into a self‐report screening survey and administered to classrooms by trained school counselors. All distress measures except the OQ Somatization Subscale used a standard 5‐point Likert‐type scale measuring the frequency with which each symptom was experienced during the last month ranging from 0 (<emph>none of the time</emph>) to 4 (<emph>most of the time</emph>). All trauma and stressful life event exposure measures used a dichotomous scale ranging from 0 (<emph>event did not occur within the specified time frame</emph>) to 1 (<emph>event occurred at least once within the time frame</emph>). Prior to modeling, we transformed the observed score on each scale or subscale by dividing by the total number of items to form an <emph>item‐level average score</emph> (). Calculating the item‐level average enhanced interpretation by benchmarking the summative scale score against the original item‐level metric (averages of Likert scale scores ranged from 0 to 4; averages of exposure scale scores ranged from 0 to 1) while controlling for between‐scale differences in scale length and scale variance.</p> <hd id="AN0052214229-9">Participants</hd> <p>Participants consisted of (<emph>N</emph> = 881) students attending nine Central Bosnian secondary schools during the 2000–2001 school year who were screened for their appropriateness for a UNICEF‐sponsored trauma and grief‐focused psychosocial program ([<reflink idref="bib31" id="ref65">31</reflink>]). The sample consisted of 55% (<emph>n </emph>=<emph> </emph>480) girls and 44% (<emph>n </emph>=<emph> </emph>390) boys (1% not reported), ranging from 13 to 19 years old (, <emph>SD</emph> = 1.10) and from the 1st to the 3rd year in school (, <emph>SD</emph> = 0.82). We randomly assigned approximately half of the participants to one of two subsamples. We used Subsample 1 for model construction, and Subsample 2 to replicate (without further modification) the model parameters constructed using Subsample 1. The two subsamples did not significantly differ (<emph>p </emph>< .05) in age, sex, grade in school, or school affiliation.</p> <hd id="AN0052214229-10">Measures</hd> <p> <emph>Prewar Trauma Exposure Index</emph> ([<reflink idref="bib27" id="ref66">27</reflink>]) is a 10‐item self‐report measure of exposure to a range of prewar traumatic experiences. Items were generated from field research and from adapting items from the Stressful Life Events Screening Questionnaire (SLESQ; [<reflink idref="bib22" id="ref67">22</reflink>]). Items are measured on a yes–no scale ("Before the war, did you ever have a life‐threatening illness?"). Items 1–9 (the 10th is open‐ended) were summed into a composite score (total sample).</p> <p> <emph>War Trauma Screening Inventory</emph> (WTSI; [<reflink idref="bib33" id="ref68">33</reflink>]) is a 49‐item self‐report measure of exposure to different types of war‐related trauma and loss. Items were drawn from Bosnian field research and from a literature review (e.g., [<reflink idref="bib40" id="ref69">40</reflink>]). Items were analyzed in two ways: For the composite–causal indicator approach, the first author drew on proposed "common denominator" dimensions of trauma exposure (e.g., [<reflink idref="bib26" id="ref70">26</reflink>]; [<reflink idref="bib52" id="ref71">52</reflink>]) to divide items into eight content domains consisting of <emph>Direct Exposure</emph> (11 items; total sample), <emph>Witnessing Violence</emph> (11 items,), <emph>Life Threat</emph> (5 items,), <emph>Traumatic Death</emph> (7 items,), <emph>Harm to Loved Ones</emph> (5 items,), <emph>Threat to Loved Ones</emph> (3 items,), <emph>Loss and Displacement</emph> (7 items,), and <emph>Separation From Loved Ones</emph> (2 items,). Table 1 presents a breakdown of WTSI items by the eight domains. For the common factor–effect indicator method, five war exposure subscales were derived using EFA (see Table 2) consisting of <emph>Siege Exposure</emph> (18 items including 1 reverse‐scored item, total sample), <emph>Expulsion and Displacement</emph> (9 items including 2 reverse‐scored items,), <emph>Detention and Maltreatment</emph> (9 items,), <emph>Harm to or Death of Loved One</emph> (7 items,), and <emph>Personal Assault or Injury</emph> (2 items,).</p> <p>1  
Composite–Causal Indicator‐Derived Dimensions of War Trauma Exposure</p> <p> <ephtml> <table><thead valign="bottom"><tr><th>Direct Exposure</th><th>Witnessing Violence</th><th>Life Threat</th><th>Traumatic Death<sup>a</sup></th><th>Harm to Loved Ones</th><th>Threat to Loved Ones</th><th>Loss and Displacement</th><th>Separation from Loved Ones</th></tr></thead><tbody valign="top"><tr><td>In serious accident
Life‐threatening cold
Deprived of food/water 
Exposed to disaster 
Physically tortured 
Taken prisoner 
Unwanted sexual contact
Was physically injured
Victim of serious crime
Became seriously ill
Was physically assaulted</td><td>Saw dead body
 Saw severely injured person
 Saw serious accident
 Saw property destruction 
 Saw someone being killed or injured 
 Touched/carried wounded or killed person 
 Saw distraught person 
 Witnessed assault 
 Witnessed person taken prisoner or abducted
 Witnessed torture
 Witnessed loved one being harmed or killed</td><td>Grenade/bomb exploded nearby 
 Bullet came close 
 Strongly believed would be hurt or killed 
 Verbally threatened with harm/death 
 Enemy soldiers entered home</td><td>Close friend killed 
 Mother killed
 Sibling killed 
 Extended family member killed 
 Loved one committed suicide 
 Loved one killed in violence not due to war 
Father killed</td><td>Loved one tortured 
 Loved one assaulted 
 Loved one taken prisoner or detained
 Loved one seriously injured 
 Loved one had life‐threatening illness</td><td>Loved one had very dangerous duties
 Loved one threatened with harm/death
 Loved one went missing</td><td> Left country due to war
 Forced from home
 Left village/town because of war 
 Changed to new school because of war 
 Lived in refugee center
 Home badly damaged 
 Lost beloved pet</td><td>Separated from loved one when greatly feared for their safety 
Separated from both caregivers for month or longer</td></tr></tbody></table> </ephtml> </p> <p>1 <sups>a</sups>The item <emph>loved one died of natural causes</emph> was dropped from the composite–causal indicator‐derived dimensions due to its ambiguity as a traumatic stressor.</p> <p>2  
Percent Item Endorsement (%), Factor Loadings (Fi), Communalities (h2), and Factor Correlations for Maximum Likelihood Factors Extraction and Promax Rotation on War Exposure Inventory Items</p> <p> <ephtml> <table><thead valign="bottom"><tr><th>Item (type of trauma exposure during war)</th><th>%</th><th><italic>F</italic><sub>1</sub><sup>a</sup></th><th><italic>F</italic><sub>2</sub></th><th><italic>F</italic><sub>3</sub></th><th><italic>F</italic><sub>4</sub></th><th><italic>F</italic><sub>5</sub></th><th><italic>h</italic><sup>2</sup></th></tr></thead><tbody valign="top"><tr><td>Bomb landed so close could have been hurt/killed</td><td>56</td><td><bold>0.77</bold></td><td>−0.02</td><td>−0.14</td><td>−0.02</td><td>0.02</td><td>0.54</td></tr><tr><td>Saw the body of someone who had been killed</td><td>45</td><td><bold>0.76</bold></td><td>−0.23</td><td>0.12</td><td>−0.10</td><td>−0.00</td><td>0.58</td></tr><tr><td>Bullet came so close could have been hurt/killed</td><td>53</td><td><bold>0.75</bold></td><td>0.02</td><td>−0.07</td><td>−0.02</td><td>0.09</td><td>0.56</td></tr><tr><td>Saw someone who was severely injured</td><td>56</td><td><bold>0.73</bold></td><td>−0.12</td><td>0.11</td><td>0.04</td><td>−0.10</td><td>0.57</td></tr><tr><td>Witnessed a serious accident</td><td>20</td><td><bold>0.69</bold></td><td>−0.06</td><td>−0.03</td><td>0.05</td><td>0.24</td><td>0.59</td></tr><tr><td>Witnessed the massive destruction of property</td><td>51</td><td><bold>0.68</bold></td><td>0.05</td><td>−0.01</td><td>−0.04</td><td>0.02</td><td>0.45</td></tr><tr><td>Witnessed someone being killed/severely injured</td><td>21</td><td><bold>0.65</bold></td><td>−0.15</td><td>0.35</td><td>−0.01</td><td>−0.09</td><td>0.61</td></tr><tr><td>Was in serious accident (car crash, bad fall, mine)</td><td>05</td><td><bold>0.65</bold></td><td>−0.04</td><td>−0.16</td><td>0.03</td><td>0.17</td><td>0.44</td></tr><tr><td>A close personal friend was killed</td><td>34</td><td><bold>0.57</bold></td><td>0.04</td><td>−0.15</td><td>0.19</td><td>−0.03</td><td>0.42</td></tr><tr><td>Touched/carried seriously injured or killed person</td><td>11</td><td><bold>0.55</bold></td><td>−0.18</td><td>0.13</td><td>0.01</td><td>0.22</td><td>0.44</td></tr><tr><td>Experienced life‐threatening cold</td><td>14</td><td><bold>0.53</bold></td><td>0.07</td><td>0.09</td><td>−0.04</td><td>−0.00</td><td>0.32</td></tr><tr><td>Witnessed people in extreme distress (distraught)</td><td>72</td><td><bold>0.51</bold></td><td>−0.16</td><td>0.11</td><td>0.38</td><td>−0.21</td><td>0.54</td></tr><tr><td>Life‐threatening deprivations of food or water</td><td>13</td><td><bold>0.48</bold></td><td>0.12</td><td>0.07</td><td>0.01</td><td>0.13</td><td>0.34</td></tr><tr><td>Strongly believed would be seriously hurt or killed</td><td>51</td><td><bold>0.45</bold></td><td>0.05</td><td>0.17</td><td>0.23</td><td>−0.12</td><td>0.42</td></tr><tr><td>Exposed to major natural or man‐made disaster</td><td>12</td><td><bold>0.43</bold></td><td>0.19</td><td>0.02</td><td>−0.08</td><td>0.25</td><td>0.32</td></tr><tr><td>Loved one had very dangerous duties</td><td>79</td><td><bold>0.41</bold></td><td>0.07</td><td>−0.29</td><td>0.32</td><td>−0.13</td><td>0.38</td></tr><tr><td>Witnessed someone being physically assaulted</td><td>20</td><td><bold>0.40</bold></td><td>0.13</td><td>0.32</td><td>0.03</td><td>−0.11</td><td>0.38</td></tr><tr><td>Left country because of the war</td><td>18</td><td>−<bold>0.51</bold></td><td><bold>0.56</bold></td><td>0.05</td><td>0.03</td><td>0.43</td><td>0.62</td></tr><tr><td>Forced to leave home because of war</td><td>48</td><td>0.02</td><td><bold>0.91</bold></td><td>0.16</td><td>−0.03</td><td>−0.07</td><td>0.90</td></tr><tr><td>Forced to leave village/town because of war</td><td>37</td><td>−0.28</td><td><bold>0.90</bold></td><td>0.07</td><td>0.06</td><td>0.08</td><td>0.86</td></tr><tr><td>Forced to change to new school because of the war</td><td>49</td><td>−0.01</td><td><bold>0.74</bold></td><td>0.09</td><td>0.07</td><td>0.17</td><td>0.64</td></tr><tr><td>Lived in a collective refugee center during the war</td><td>11</td><td>−0.15</td><td><bold>0.62</bold></td><td>0.04</td><td>0.10</td><td>0.25</td><td>0.49</td></tr><tr><td>Home seriously damaged in the war</td><td>35</td><td>0.26</td><td><bold>0.51</bold></td><td>0.07</td><td>−0.04</td><td>−0.04</td><td>0.38</td></tr><tr><td>Lost a beloved pet because of war</td><td>26</td><td>0.22</td><td><bold>0.47</bold></td><td>0.03</td><td>0.06</td><td>−0.07</td><td>0.34</td></tr><tr><td>Mother killed during war</td><td>01</td><td>−0.07</td><td>−<bold>0.46</bold></td><td>0.35</td><td>0.23</td><td>0.30</td><td>0.42</td></tr><tr><td>Brother or sister killed during war</td><td>34</td><td>0.21</td><td>−<bold>0.43</bold></td><td>−0.02</td><td>0.17</td><td>0.27</td><td>0.29</td></tr><tr><td>Was physically tortured</td><td>02</td><td>−0.04</td><td>−0.01</td><td><bold>0.76</bold></td><td>−0.01</td><td>0.38</td><td>0.86</td></tr><tr><td>Taken prisoner or confined in detention camp</td><td>03</td><td>−0.14</td><td>0.11</td><td><bold>0.73</bold></td><td>0.09</td><td>0.23</td><td>0.71</td></tr><tr><td>Was verbally threatened with harm or death</td><td>05</td><td>−0.02</td><td>−0.08</td><td><bold>0.70</bold></td><td>0.13</td><td>0.09</td><td>0.55</td></tr><tr><td>Witnessed someone being abducted/taken prisoner</td><td>22</td><td>0.25</td><td>0.16</td><td><bold>0.70</bold></td><td>−0.06</td><td>−0.32</td><td>0.66</td></tr><tr><td>Enemy soldiers forcibly entered residence</td><td>12</td><td>−0.06</td><td>0.30</td><td><bold>0.66</bold></td><td>0.15</td><td>−0.08</td><td>0.65</td></tr><tr><td>Witnessed someone being tortured</td><td>08</td><td>0.38</td><td>0.13</td><td><bold>0.55</bold></td><td>−0.20</td><td>−0.06</td><td>0.53</td></tr><tr><td>Loved one verbally threatened with harm or death</td><td>11</td><td>0.03</td><td>0.20</td><td><bold>0.53</bold></td><td>0.40</td><td>−0.26</td><td>0.64</td></tr><tr><td>Witnessed a loved one being harmed or killed</td><td>09</td><td>0.20</td><td>−0.08</td><td><bold>0.50</bold></td><td>0.19</td><td>0.23</td><td>0.56</td></tr><tr><td>Loved one was physically assaulted</td><td>14</td><td>0.03</td><td>0.11</td><td><bold>0.49</bold></td><td>0.44</td><td>−0.25</td><td>0.56</td></tr><tr><td>Loved one was tortured</td><td>08</td><td>−0.09</td><td>−0.05</td><td>0.49</td><td><bold>0.54</bold></td><td>0.09</td><td>0.60</td></tr><tr><td>Loved one missing, greatly feared for their safety</td><td>20</td><td>−0.02</td><td>0.12</td><td>0.18</td><td><bold>0.59</bold></td><td>0.05</td><td>0.49</td></tr><tr><td>Loved one taken prisoner/held in detention camp</td><td>20</td><td>−0.14</td><td>0.10</td><td>0.35</td><td><bold>0.51</bold></td><td>0.11</td><td>0.50</td></tr><tr><td>Extended family member killed</td><td>47</td><td>0.16</td><td>0.03</td><td>−0.17</td><td><bold>0.51</bold></td><td>−0.12</td><td>0.34</td></tr><tr><td>Loved one was seriously physically injured</td><td>37</td><td>0.19</td><td>−0.02</td><td>−0.09</td><td><bold>0.45</bold></td><td>0.19</td><td>0.34</td></tr><tr><td>Loved one committed suicide</td><td>05</td><td>−0.05</td><td>−0.09</td><td>0.04</td><td><bold>0.44</bold></td><td>0.12</td><td>0.20</td></tr><tr><td>Loved one died in accident or community violence</td><td>08</td><td>0.01</td><td>0.08</td><td>0.01</td><td><bold>0.43</bold></td><td>0.27</td><td>0.33</td></tr><tr><td>Was touched in a way that I did not want</td><td>02</td><td>0.04</td><td>0.09</td><td>0.06</td><td>0.07</td><td><bold>1.07</bold><sup>b</sup></td><td>1.24</td></tr><tr><td>Seriously injured during war</td><td>04</td><td>0.28</td><td>−0.05</td><td>0.18</td><td>−0.07</td><td><bold>0.59</bold></td><td>0.55</td></tr><tr><td><bold>*Nonloading items (all loadings < 0.40):</bold></td></tr><tr><td>Father killed during war</td><td>10</td><td>−0.04</td><td>−0.16</td><td>0.04</td><td>0.33</td><td>−0.04</td><td>0.10</td></tr><tr><td>Separated from loved one when greatly feared for their safety</td><td>60</td><td>0.32</td><td>0.28</td><td>−0.14</td><td>0.25</td><td>0.06</td><td>0.34</td></tr><tr><td>Loved one died of natural causes</td><td>47</td><td>0.22</td><td>0.24</td><td>−0.22</td><td>0.16</td><td>−0.15</td><td>0.21</td></tr><tr><td>Loved one had life‐threatening physical illness</td><td>14</td><td>0.27</td><td>0.05</td><td>−0.05</td><td>0.08</td><td>0.05</td><td>0.10</td></tr><tr><td>Was the victim of a serious crime</td><td>06</td><td>0.02</td><td>0.25</td><td>0.39</td><td>0.01</td><td>0.16</td><td>0.32</td></tr><tr><td>Separated from primary caregivers for ≥1 month</td><td>13</td><td>0.15</td><td>0.29</td><td>−0.02</td><td>−0.02</td><td>0.32</td><td>0.23</td></tr><tr><td>Became seriously ill</td><td>10</td><td>0.23</td><td>0.12</td><td>−0.06</td><td>−0.05</td><td>0.32</td><td>0.18</td></tr><tr><td>Was physically assaulted</td><td>03</td><td>0.34</td><td>−0.08</td><td>0.38</td><td>−0.06</td><td>0.26</td><td>0.44</td></tr><tr><td valign="top">Promax factor correlations</td><td><italic>F</italic><sub>1</sub></td><td>1.00</td><td> </td><td> </td><td> </td><td> </td><td> </td></tr><tr><td><italic>F</italic><sub>2</sub></td><td>.18</td><td>1.00</td><td> </td><td> </td><td> </td><td> </td></tr><tr><td><italic>F</italic><sub>3</sub></td><td>.27</td><td>.21</td><td>1.00</td><td> </td><td> </td><td> </td></tr><tr><td><italic>F</italic><sub>4</sub></td><td>.14</td><td>.04</td><td>.27</td><td>1.00</td><td> </td><td> </td></tr><tr><td><italic>F</italic><sub>5</sub></td><td>.37</td><td>.31</td><td>.21</td><td>.13</td><td>1.00</td><td> </td></tr></tbody></table> </ephtml> </p> <p>2 <sups>a</sups>Factor labels: <emph>F</emph><subs>1</subs> = siege exposure; <emph>F</emph><subs>2</subs> = expulsion and displacement; <emph>F</emph><subs>3</subs> = detention and maltreatment; <emph>F</emph><subs>4</subs> = harm to or death of loved one; <emph>F</emph><subs>5</subs> = direct personal assault or injury. Primary factor loadings for each variable are bolded (decision rule: complexly loaded items with both loadings ≥ 0.50 load on both factors). <sups>b</sups>Heywood case (model estimates a negative residual variance) linked to low item variance ([<reflink idref="bib12" id="ref72">12</reflink>]).</p> <p> <emph>Postwar Trauma Exposure Inventory</emph> (PWTEI; [<reflink idref="bib27" id="ref73">27</reflink>], [<reflink idref="bib28" id="ref74">28</reflink>]) is a 13‐item self‐report measure of exposure to nonwar‐related traumatic events. Items were adapted from the SLESQ and from field research in postwar Bosnia. Items include general traumatic events ("Since the war, has a loved one attempted suicide?") as well as events characteristic of the Bosnian conflict ("Since the war, has a loved one who disappeared during the war been declared dead?"). The PWTEI contains four content domains that were rationally derived by the first author, including <emph>Direct Exposure</emph> (6 items, total sample), <emph>Witnessing</emph> (2 items,), <emph>Harm to Loved Ones</emph> (3 items,), and <emph>Traumatic Death</emph> (2 items,).</p> <p> <emph>Trauma Reminders Screening Inventory</emph> (TRSI; [<reflink idref="bib32" id="ref75">32</reflink>]) is a 17‐item self‐report measure of the frequency of respondents' exposure to reminders of war‐related traumatic events during the past month and associated functional impairment. Using theory and EFA, items were partitioned into <emph>general trauma reminders</emph> (8 items; e.g., news of war atrocities, political instability), <emph>interpersonal trauma reminders</emph> (4 items; e.g., friends, family), and <emph>trauma reminder‐related interference</emph> (3 items; e.g., impaired social relationships). Because the TRSI measures <emph>perceptions</emph> of being reminded (suggesting a common underlying causal process), alphas were calculated for each subscale (full‐sample general α = .90, interpersonal α = .83, interference α = .85). We modeled general and interpersonal reminders as postwar mediating variables, and reminder‐related interference as an outcome.</p> <p> <emph>UCLA Reaction Index–Revised</emph> (RI–R; [<reflink idref="bib57" id="ref76">57</reflink>]) as used in this study is a 17‐item self‐report scale of symptom frequency during the previous month. Items correspond to the 17 <emph>DSM–IV</emph> PTSD criteria. The total scale has shown good internal consistency (α<emph> </emph>= .87), criterion‐referenced validity in relation to a range of distress measures (.30 to.70), and 2‐week test–retest reliability (.75) in postwar Bosnian youth ([<reflink idref="bib34" id="ref77">34</reflink>]). Full sample, full‐scale α = .87.</p> <p> <emph>Depression Self‐Rating Scale</emph> ([<reflink idref="bib5" id="ref78">5</reflink>]) is an 18‐item self‐report measure of depressive symptoms. The original 3‐point scale (for children) was modified to a 5‐point frequency scale to increase developmental appropriateness and sensitivity to clinical change. The adapted version has shown acceptable internal consistency (α = .85), criterion‐referenced validity (.37 to.62), and test–retest reliability (.64) in postwar Bosnian youth ([<reflink idref="bib34" id="ref79">34</reflink>]). Full sample, full‐scale α = .89 in this study.</p> <p> <emph>Youth Outcome Questionnaire: Somatic Symptoms Subscale</emph> ([<reflink idref="bib58" id="ref80">58</reflink>]) is an 8‐item self‐report measure of somatic distress (e.g., "I have headaches or feel dizzy") in children and adolescents. Items are measured on a 5‐point frequency scale ranging from 0 (<emph>never or almost never</emph>) to 4 (<emph>almost always or always</emph>). Full sample, full‐scale alpha α = .80 in this study.</p> <p> <emph>Postwar Adversities Index</emph> ([<reflink idref="bib30" id="ref81">30</reflink>]) is a 25‐item self‐report inventory of stressful life events and circumstances during the past 6 months. Items were rationally divided into two content domains by two independent raters who showed substantial interrater agreement using a dichotomous scale (κ = .79). Domains were <emph>postwar existential adversities</emph> (16 items; short on pocket money) and <emph>interpersonal adversities</emph> (6 items; living with people who fight). Items were coded according to valence (positive/neutral/negative life event) and independence/dependence by six independent raters on a 5‐point scale. Raters showed excellent interrater reliability (α = .92) on valence and acceptable reliability (α = .79) on event independence. Chronic and acute events ([<reflink idref="bib41" id="ref82">41</reflink>]) were both included. Because this study focused on adverse outcomes, only <emph>negative‐independent</emph> life events were included; three events rated as <emph>dependent</emph> (pregnancy, getting into fights, and drug use) were dropped from analysis.</p> <hd id="AN0052214229-11">Procedure</hd> <p>To minimize potential stigmatization of individual students, school counselors selected entire classrooms with the highest prevalence rates of severely war‐exposed students. Details regarding classroom selection, consent, and survey administration procedures are provided elsewhere ([<reflink idref="bib31" id="ref83">31</reflink>]). Measures and procedures were approved by an ad hoc institutional review board (IRB) formed of Bosnian mental health professionals and a UNCEF officer, and by the IRB of Brigham Young University.</p> <hd id="AN0052214229-12">Results</hd> <p>In a first step, we used Mplus 5.1 ([<reflink idref="bib46" id="ref84">46</reflink>]) to perform EFA on the full‐sample covariance matrix formed by the 49 WTSI items. Maximum likelihood factors extraction was used in combination with promax (oblique) rotation. This procedure adjusts for missing values on observed variables and is suitable for binary‐ordered categorical indicators ([<reflink idref="bib45" id="ref85">45</reflink>]). After several preliminary extractions to estimate the number of factors and verify the absence of multicollinearity and the factorability of the correlation matrix, five factors were extracted. Table 2 presents item endorsement rates (in %), the rotated factor pattern matrix, communality (<emph>h</emph><sups>2</sups>) values, and a suggested label for interpreting each factor. Given the dichotomous scaling of the WTSI and low (< 10%) endorsement rates for many items, a liberal criterion (explained variance) was used to denote the primary factor loadings. The factors varied in quality as indicated by the number of variables with primary loadings and the sizes of the loadings. Factors were labeled <emph>Siege Exposure</emph> (Factor 1), <emph>Expulsion & Displacement</emph> (Factor 2), <emph>Detention & Maltreatment</emph> (Factor 3), <emph>Harm to or Death of Loved One</emph> (Factor 4), and <emph>Direct Personal Assault or Injury</emph> (Factor 5). Although interpretable, the solution poorly accounted for many items as denoted by low communalities and the loss of eight items (15% of total) due to low loadings. Among dropped items were events with low (<emph>mother killed</emph> = 1%), moderate (<emph>father killed</emph> = 10%), and comparatively high endorsement rates (<emph>physically separated from loved one</emph> = 60%) that correlated weakly with the retained items (full correlation matrix available from first author). Consistent with Hypothesis 1, the five EFA‐derived factors reflected idiosyncratic features of the Bosnian war, including sieges, ethnic cleansing, and detention and maltreatment ([<reflink idref="bib51" id="ref86">51</reflink>]).</p> <p>In a second step, we evaluated the degree of convergence between the trauma exposure dimensions derived using the common factor versus the composite approach. The approaches were compared at two levels: (a) visually inspecting the item contents of the factors and composites (i.e., comparing Table 1 with Table 2), and (b) examining correlations between the respective subscale scores (a comparison matrix of factor/composite item overlaps and intercorrelations is available from the first author). In support of Hypothesis 2, the two approaches showed poor to moderate convergence at both levels of analysis and were consistent with our concerns regarding the <emph>within‐dimension</emph> diversity of war exposure dimensions derived via the common factor model. Visual inspection revealed that <emph>Siege Exposure</emph> (Factor 1), <emph>Detention & Maltreatment</emph> (Factor 3), and <emph>Harm to or Death of Loved One</emph> (Factor 4) each contained a broad range of trauma types, such that their constituent items were broadly dispersed across the composite dimensions. For example, the item contents of Factor 1 (<emph>Siege Exposure</emph>) were distributed across the <emph>Direct Exposure</emph>, <emph>Witnessing Violence</emph>, <emph>Life Threat</emph>, <emph>Traumatic Death</emph>, <emph>Threat to Loved Ones</emph>, and <emph>Loss and Displacement</emph> composites. Further, the contents of Factor 3 (<emph>Detention & Maltreatment</emph>) were distributed across <emph>Direct Exposure, Witnessing Violence, Life Threat</emph>, and <emph>Harm to Loved Ones</emph>. The greatest degree of convergence was found in the overlap between <emph>Expulsion & Displacement</emph> (Factor 2) and <emph>Loss and Displacement</emph> (Composite 7). The inverse loadings of <emph>mother killed</emph> and <emph>sibling killed</emph> on <emph>Expulsion & Displacement</emph> were unexpected. A second point of convergence was found in the overlap between the two items loading <emph>Direct Personal Assault or Injury</emph> (Factor 5) and <emph>Direct Exposure</emph> (Composite 1). However, because <emph>Direct Exposure</emph> contained six items not found in Factor 5, Factor 5 was a small subset of Composite 1.</p> <p>The general lack of convergence between the two approaches observed in our visual inspection was reflected in a similar lack of differentiation in their subscale intercorrelations. The strongest points of convergence were found between <emph>Siege Exposure</emph> (Factor 1) and <emph>Witnessing Violence</emph> (Composite 2) (<emph>r </emph>= .89), and between <emph>Expulsion & Displacement</emph> (Factor 2) and <emph>Loss and Displacement</emph> (Composite 7) (<emph>r </emph>= .97). More generally, however, the factor–composite correlation matrix showed little clear evidence of between‐method convergence between specific factors and composites, with the great majority of correlations falling in the low to moderate range (.10 < <emph>r </emph>< .60). For example, the Factor 3 (<emph>Detention & Maltreatment</emph>) subscale score correlated between <emph>r </emph>= .40 and.60 (<emph>p </emph>< .01) with those of the <emph>Direct Exposure</emph>, <emph>Witnessing Violence</emph>, <emph>Life Threat</emph>, and <emph>Harm to Loved Ones</emph> composites. In general, most factor subscale scores correlated indiscriminately with most composite subscale scores—an unsurprising finding given the high degree of dispersion of factor item contents across composites, and vice versa.</p> <p>In a third step, we used Amos 16.0 ([<reflink idref="bib56" id="ref87">56</reflink>]) to evaluate the fit of both the common factor and composite models in Subsample 1, and to evaluate the fit of the (unmodified) models in Subsample 2. We formed our models using observed variable path analysis—a variant of SEM conducive to modeling composite variables ([<reflink idref="bib9" id="ref88">9</reflink>]). Given the exploratory nature of the study and the lack of guiding theory regarding relations between <emph>specific</emph> types of trauma and loss exposure, mediators, and outcomes, we used a conservative "repeated pruning" model‐building strategy for both the composite and common factor method subsamples: We first created a semisaturated model that contained all possible paths linking (a) prewar trauma to each theorized mediator and each outcome, (b) each war‐time exposure dimension to each theorized mediator and each outcome, and (c) each theorized mediator to each outcome. We then tested the model and deleted all nonsignificant paths with <emph>p </emph>> .10. We then retested the model and deleted all remaining nonsignificant paths with <emph>p </emph>> .10. In a final pass, we retested the model and deleted all nonsignificant paths with <emph>p </emph>> .05.</p> <p>The composite–causal indicator model for Subsample 1 (see Figure 2) had good fit, χ<sups>2</sups>(<reflink idref="bib79" id="ref89">79</reflink>) = 75.73, <emph>p </emph>= .58; normed fix index (NFI) = .98, comparative fit index (CFI) = 1.0, root mean square error of approximation (RMSEA) = 0 (90% confidence interval [CI] = 0–0.02), Hoelter (.01) = 676; and explained 55%, 46%, 27%, and 28% of the variance in PTSD symptoms, reminder‐related functional impairment, depression, and somatic symptoms, respectively. Contrary to expectation, <emph>Harm to Loved Ones</emph> and <emph>Threat to Loved Ones</emph> did not exert significant predictive effects. Consistent with Hypothesis 3, the partially mediated model fit significantly better than its fully mediated counterpart, χ<sups>2</sups>(<reflink idref="bib8" id="ref90">8</reflink>) = 55.04, <emph>p </emph>< .01, indicating that prewar and some "traumatogenic" war‐time exposures exerted direct predictive effects. Direct paths linked <emph>prewar trauma</emph> to PTSD symptoms, somatic symptoms, and functional impairment; war‐time <emph>traumatic bereavement</emph> to PTSD symptoms and impairment; and war‐time <emph>witnessing violence</emph> to impairment. Postwar <emph>direct exposure</emph> had a direct path to impairment, and postwar <emph>witnessing violence</emph> had direct paths to both impairment and depression.</p> <p>Graph: 2 Observed variable path analysis model of composite–causal indicator‐derived dimensions of trauma exposure.</p> <p>The common factor–effect indicator model for Subsample 1 (see Figure 3), also had good fit χ<sups>2</sups>(<reflink idref="bib60" id="ref91">60</reflink>) = 57.71, <emph>p </emph>= .56; NFI = .98, CFI = 1.0, RMSEA = 0 (90% CI = 0–0.03), Hoelter (.01) = 705; and explained 55%, 46%, 28%, and 29% of the variances in PTSD symptoms, reminder‐related functional impairment, depression symptoms, and somatic symptoms, respectively. Contrary to expectation, neither <emph>Detention & Maltreatment</emph> nor <emph>Personal Assault or Injury</emph> (derived from Factors 3 and 5, respectively) exerted significant predictive effects. Consistent with Hypothesis 3, the partially mediated model fit significantly better than the fully mediated model, χ<sups>2</sups>(<reflink idref="bib6" id="ref92">6</reflink>) = 43.79, <emph>p </emph>< .01. As in the composite model, <emph>prewar</emph> trauma had three direct paths to PTSD symptoms, somatic symptoms, and functional impairment. Notably, only <emph>Harm to or Death of Loved One</emph> (Factor 4) exerted a direct predictive effect (to depression) whereas the predictive effects of all other war‐time variables were fully mediated. <emph>Postwar Direct Exposure</emph> had a direct path to reminder‐related functional impairment, and <emph>Postwar Harm to Loved Ones</emph> had a direct path to PTSD symptoms.</p> <p>Graph: 3 Observed variable path analysis model of common factor–effect indicator‐derived dimensions of trauma exposure.</p> <p>In a fourth step, given the susceptibility of "repeated pruning" to sample‐idiosyncratic over‐fitting ([<reflink idref="bib9" id="ref93">9</reflink>]), the goodness of fit of both models derived using Subsample 1 was tested (without further modification) in Subsample 2. The composite model had generally good fit, χ<sups>2</sups>(<reflink idref="bib79" id="ref94">79</reflink>) = 159.58, <emph>p </emph>< .01; NFI = .95, CFI = .97, RMSEA = .05 (90% CI = 0.04–0.06), Hoelter (.01) = 292. The composite model retained (at a significance level of <emph>p</emph> < .05) 28 (68%) of the 41 significant paths identified in Subsample 1 and explained 50%, 53%, 33%, and 30% of the variances in PTSD symptoms, reminder‐related functional impairment, depression symptoms, and somatic symptoms, respectively. The common factor model also had generally good fit, χ<sups>2</sups>(<reflink idref="bib60" id="ref95">60</reflink>) = 115.93, <emph>p </emph>< .01; NFI = .96, CFI = .98, RMSEA = .05 (90% CI = 0.03–0.06), Hoelter (.01) = 320. The common factor model retained 26 (72%) of the 36 paths in Subsample 1 and explained 51%, 53%, 33%, and 31% of the variances in PTSD symptoms, functional impairment, depression symptoms, and somatic symptoms, respectively. Reduced power (due to halving the sample), as well as restricted predictor variance, likely played roles in reducing the number of significant paths in both Subsample 2 models: In the composite model, 6 of 13 nonsignificant paths had predictor means < 0.10, and 10 of 13 had means < 0.15. In the common factor model, 4 of 10 nonsignificant paths had predictor means < 0.10, and 5 of 10 had means < 0.15.</p> <p>In a fifth step, we tested Hypothesis 4 (beginning with Subsample 1) using effects decomposition ([<reflink idref="bib9" id="ref96">9</reflink>]) to rank prewar, war‐time, and postwar trauma risk factors by their total predictive effects on each outcome. Total effects for each risk factor were calculated by summing its direct effects (via direct paths) and indirect effects (via mediated paths). The top and bottom halves of Table 3 contain the risk factor rankings for the composite‐ and common factor‐based approaches, respectively. Consistent with Hypothesis 4, the rankings for the composite model showed more differentiation than the common factor model between war‐time trauma exposure and outcomes. Four war‐time variables (<emph>Life Threat</emph>, <emph>Traumatic Death</emph>, <emph>Witnessing Violence</emph>, and <emph>Loss and Displacement</emph>) appeared in the top six rankings across outcomes in the composite model. In contrast, three war‐time variables (<emph>Siege Exposure</emph>, <emph>Expulsion & Displacement</emph>, and <emph>Harm to or Death of Loved Ones</emph>) appeared in the top six rankings for the common factor model; <emph>Siege Exposure</emph> was the dominant risk factor for all outcomes. In a Subsample 2 replication analysis, the risk factor rankings in both models remained relatively stable (rankings not shown). Five war‐time variables (the four from Subsample 1 plus <emph>Separation From Loved Ones</emph>) appeared in the top six rankings of the composite model. The common factor model contained the three war‐time risk factors found in Subsample 1; <emph>Siege Exposure</emph> was again the dominant risk factor. <emph>Prewar Trauma</emph> was a predictor for all outcomes in both the composite and common factor models, in both Subsamples 1 and 2.</p> <p>3  
Rankings (by Predictive Effect) of Trauma Exposure Risk Factors and Postwar Mediators on Posttraumatic Stress, Depression, Somatic Symptoms, and Trauma Reminder‐Related Functional Impairment</p> <p> <ephtml> <table><thead valign="bottom"><tr><th>Composite–Causal Indicator Model of War Trauma Exposure: Risk Factor Rankings by Outcome<sup>a</sup></th></tr><tr><th>Trauma exposure risk factor ranking<sup>b</sup></th><th>Posttraumatic stress 
(<italic>R</italic><sup>2</sup> = 55%)
<italic>Risk factor</italic> (total) [direct] {indirect} effect</th><th>Depression (<italic>R</italic><sup>2</sup> = 27%)
<italic>Risk factor</italic> (total) [direct] {indirect} effect</th><th>Somatic distress (<italic>R</italic><sup>2</sup> = 28%)
<italic>Risk factor</italic> (total) [direct] {indirect} effect</th><th>Functional impairment (<italic>R</italic><sup>2</sup> = 46%)
<italic>Risk factor</italic> (total) [direct] {indirect} effect</th></tr></thead><tbody valign="top"><tr><td>1</td><td><italic>War‐Time Life Threat</italic> (.26) [.00] {.26}</td><td><italic>War‐Time Life Threat</italic> (.16) [.00] {.16}</td><td><italic>Prewar Trauma</italic> (.18) [.12] {.06}</td><td><italic>War‐Time Life Threat</italic> (.23) [.00] {.23}</td></tr><tr><td>2</td><td><italic>Prewar Trauma</italic> (.16) [.12] {.04}</td><td><italic>War‐Time Loss & Displacement</italic> (.09) [.00] {.09}</td><td><italic>War‐Time Life Threat</italic> (.16) [.00] {.16}</td><td><italic>Prewar Trauma</italic> (.14) [.11] {.03}</td></tr><tr><td>3</td><td><italic>War‐Time Traumatic Death</italic> (.10) [.08] {.03}<sup>c</sup></td><td><italic>Prewar Trauma</italic> (.07) [.00] {.07}</td><td><italic>War‐Time Loss & Displacement</italic> (.07) [.00] {.07}</td><td><italic>War‐Time Witnessing</italic> (.13) [.08] {.05}</td></tr><tr><td>4</td><td><italic>War‐Time</italic> Witnessing (.07) [.00] {.07}</td><td><italic>Postwar Harm to Loved Ones</italic> (.07) [.00] {.07}</td><td><italic>Postwar Harm to Loved Ones</italic> (.05) [.00] {.05}</td><td><italic>Postwar Direct Trauma Exposure</italic> (.11) [.09] {.02}</td></tr><tr><td>5</td><td><italic>War‐Time Loss & Displacement</italic> (.06) [.00] {.06}</td><td><italic>Postwar Direct Trauma Exposure</italic> (.04) [.00] {.04}</td><td><italic>War‐Time Witnessing Violence</italic> (.03) [.00] {.03}</td><td><italic>War‐Time Traumatic Death</italic> (.05) [.00] {.05}</td></tr><tr><td>6</td><td><italic>Postwar Harm to Loved Ones</italic> (.05) [.00] {.05}</td><td><italic>War‐Time Witnessing</italic> (.02) [.00] {.02}</td><td><italic>Postwar Direct Trauma Exposure</italic> (.02) [.00] {.02}</td><td><italic>War‐Time Loss & Displacement</italic> (.04) [.00] {.04}</td></tr><tr><td>Postwar mediator ranking<sup>d,e</sup></td><td><italic>Mediators</italic> [direct effect]</td><td><italic>Mediators</italic> [direct effect]</td><td><italic>Mediators</italic> [direct effect]</td><td><italic>Mediators</italic> [direct effect]</td></tr><tr><td>1</td><td><italic>General Trauma Reminders</italic> [.41]</td><td><italic>Interpersonal Adversities</italic> [.24]</td><td><italic>Existential Adversities</italic> [.21]</td><td><italic>General Trauma Reminders</italic> [.31]</td></tr><tr><td>2</td><td><italic>Interpersonal Trauma Reminders</italic> [.24]</td><td><italic>Existential Adversities</italic> [.23]</td><td><italic>General Trauma Reminders</italic> [.20]</td><td><italic>Interpersonal Trauma Reminders</italic> [.30]</td></tr><tr><td>3</td><td><italic>Interpersonal Adversities</italic> [.17]</td><td><italic>Interpersonal Trauma Reminders</italic> [.17]</td><td><italic>Interpersonal Trauma Reminders</italic> [.15]</td><td><italic>Existential Adversities</italic> [.11]</td></tr><tr><td>4</td><td><italic>Existential Adversities</italic> [.14]</td><td><italic>General Trauma Reminders</italic> [.13]</td><td><italic>Interpersonal Adversities</italic> [.13]</td><td><italic>Interpersonal Adversities</italic> [.09]</td></tr></tbody></table> </ephtml> </p> <p>3  
Rankings (by Predictive Effect) of Trauma Exposure Risk Factors and Postwar Mediators on Posttraumatic Stress, Depression, Somatic Symptoms, and Trauma Reminder‐Related Functional Impairment</p> <p> <ephtml> <table><thead valign="bottom"><tr><th>Common Factor–Effect Indicator Model of War Trauma Exposure: Risk Factor Rankings by Outcome<sup>a</sup></th></tr><tr><th>Trauma exposure risk factor ranking<sup>b</sup></th><th>Posttraumatic stress (<italic>R</italic><sup>2</sup> = 55%)
<italic>Risk factor</italic> (total) [direct] {indirect} effect</th><th>Depression (<italic>R</italic><sup>2</sup> = 28%)
<italic>Risk factor</italic> (total) [direct] {indirect} effect</th><th>Somatic distress (<italic>R</italic><sup>2</sup> = 29%)
<italic>Risk factor</italic> (total) [direct] {indirect} effect</th><th>Functional impairment (<italic>R</italic><sup>2</sup> = 46%)
<italic>Risk factor</italic> (total) [direct] {indirect} effect</th></tr></thead><tbody valign="top"><tr><td>1</td><td><italic>War‐Time Siege Exposure</italic> (.28) [.00] {.28}</td><td><italic>War‐Time Siege Exposure</italic> (.18) [.00] {.18}</td><td><italic>War‐Time Siege Exposure</italic> (.18) [.00] {.18}</td><td><italic>War‐Time Siege Exposure</italic> (.24) [.00] {.24}</td></tr><tr><td>2</td><td><italic>Prewar Trauma</italic> (.15) [.10] {.04}</td><td><italic>War‐Time Expulsion & Displacement</italic> (.13) [.00] {.13}</td><td><italic>Prewar Trauma</italic> (.18) [.12] {.05}</td><td><italic>Prewar Trauma</italic> (.15) [.12] {.03}</td></tr><tr><td>3</td><td><italic>War‐Time Expulsion & Displacement</italic> (.11) [.00] {.11}</td><td><italic>Prewar Trauma</italic> (.07) [.00] {.07}</td><td><italic>War‐Time Expulsion & Displacement</italic> (.11) [.00] {.11}</td><td><italic>Postwar Direct Trauma Exposure</italic> (.10) [.09] {.01}</td></tr><tr><td>4</td><td><italic>Postwar Harm to Loved Ones</italic> (.10) [.05] {.04}</td><td><italic>Postwar Harm to Loved One</italic> (.07) [.00] {.07}</td><td><italic>Postwar Harm to Loved Ones</italic> (.05) [.00] {.05}</td><td><italic>War‐Time Expulsion & Displacement</italic> (.08) [.00] {.08}</td></tr><tr><td>5</td><td><italic>Postwar Direct Trauma Exposure</italic> (.03) [.00] {.03}</td><td><italic>Postwar Direct Trauma Exposure</italic> (.04) [.00] {.04}</td><td><italic>Postwar Direct Trauma Exposure</italic> (.02) [.00] {.02}</td><td><italic>Postwar Harm to Loved One</italic> (.03) [.00] {.03}</td></tr><tr><td>6</td><td><italic>War‐Time Harm to or Death of Loved One</italic> (.01) [.00] {.01}</td><td><italic>War‐Time Harm to or Death of Loved One</italic> (−.09) [−.08] {−.01}<sup>f</sup></td><td><italic>War‐Time Harm to or Death of Loved One</italic> (.00) [.00] {.00}</td><td><italic>War‐Time Harm to or Death of Loved One</italic> (.02) [.00] {.02}</td></tr><tr><td>Postwar mediator ranking<sup>g</sup></td><td><italic>Mediators</italic> [direct effect]</td><td><italic>Mediators</italic> [direct effect]</td><td><italic>Mediators</italic> [direct effect]</td><td><italic>Mediators</italic> [direct effect]</td></tr><tr><td>1</td><td><italic>General Trauma Reminders</italic> [.33]</td><td><italic>Existential Adversities</italic> [.24]</td><td><italic>Existential Adversities</italic> [.21]</td><td><italic>General Trauma Reminders</italic> [.33]</td></tr><tr><td>2</td><td><italic>Interpersonal Trauma Reminders</italic> [.30]</td><td><italic>Interpersonal Adversities</italic> [.22]</td><td><italic>General Trauma Reminders</italic> [.20]</td><td><italic>Interpersonal Trauma Reminders</italic> [.30]</td></tr><tr><td>3</td><td><italic>Existential Adversities</italic> [.12]</td><td><italic>Interpersonal Trauma Reminders</italic> [.18]</td><td><italic>Interpersonal Trauma Reminders</italic> [.15]</td><td><italic>Existential Adversities</italic> [.12]</td></tr><tr><td>4</td><td><italic>Interpersonal Adversities</italic> [.08]</td><td><italic>General Trauma Reminders</italic> [.13]</td><td><italic>Interpersonal Adversities</italic> [.13]</td><td><italic>Interpersonal Adversities</italic> [.08]</td></tr></tbody></table> </ephtml> </p> <ulist> <item>3 <emph>Note</emph>. <emph>R</emph><sups>2</sups> = proportion of variance explained by model.</item> <item>4 <sups>a</sups>Predictive effects were calculated using standardized models in which all outcomes and mediators were treated as continuous variables. <sups>b</sups>Six highest‐ranking trauma exposure risk factors (by total predictive effect) for each outcome. <sups>c</sups>Direct + indirect effects versus total effects do not always correspond due to rounding error. <sups>d</sups><emph>R</emph><sups>2</sups> values for mediators in composite–causal indicator model = 24% for <emph>General Trauma Reminders</emph>, 12% for <emph>Interpersonal Trauma Reminders</emph>, 27% for <emph>Existential Adversities</emph>, and 19% for <emph>Interpersonal Adversities</emph>. <sups>e</sups>All four of the modeled putative postwar mediators are presented, ranked by direct predictive effect on the outcome, for both models. <sups>f</sups>Negative association suggests a suppressor effect<emph>. </emph><sups>g</sups>%<emph>R</emph><sups>2</sups> values for mediators in common factor–effect indicator = 19% for <emph>General Trauma Reminders</emph>, 7% for <emph>Interpersonal Trauma Reminders</emph>, 28% for <emph>Existential Adversities</emph>, and 20% for <emph>Interpersonal Adversities</emph>.</item> </ulist> <hd id="AN0052214229-13">Discussion</hd> <p>This study addressed the question of how to conceptualize and model the multidimensional features of war exposure in ways that best facilitate theory building, strengthen methodological rigor, and improve risk screening and intervention. We presented two different approaches to modeling trauma exposure, the traditionally used <emph>common factor with effect indicators model</emph> and the <emph>composite–causal indicators model</emph>, and discussed their respective strengths in promoting these three aims. We focused on the core assumption of indicator homogeneity (i.e., are trauma exposure indicators <emph>causal effects</emph> of a common factor, or instead <emph>contributors</emph> to a composite?) and its far‐reaching implications for how war exposure is conceptualized, measured, studied, and treated. The central question is: <emph>Should specific types of trauma and loss be aggregated into dimensions according to their covariance structure, or instead, according to their consequences?</emph> Given its widespread use via EFA and CFA, we reviewed limitations of the common factor model for applications that aim to model differential links between the occurrence of specific types of traumatic events and specific causal consequences. We posed a caveat: Applications that violate the assumption of indicator homogeneity (shelling cities <emph>causes</emph> sieges, not the reverse) can create "model–application" mismatches. Such mismatches can impede a call to move beyond identifying general "markers of risk" that only denote <emph>who</emph> may need <emph>some form</emph> of mental health services, to instead creating tools that can accurately predict <emph>who</emph> may need <emph>which</emph> types of mental health treatment components across different war and disaster settings ([<reflink idref="bib2" id="ref97">2</reflink>]).</p> <p>We further proposed that a <emph>composite</emph>‐based approach that "packs" indicators into equifinal war exposure dimensions may better elucidate the differential effects of specific types of trauma on specific outcomes. A composite approach conceptualizes covariance between types of trauma exposure as reflecting <emph>perceptions of event co‐occurrence within specific settings</emph>. Shedding unnecessary constraints linked to homogeneity (e.g., retaining only items that load a common factor and that increase "internal consistency") encourages an increased focus on equifinality through aggregating <emph>specific types</emph> of causal risk factors (e.g., traumatic bereavement) according to their common causal effects (PTSD, traumatic grief), and on identifying key <emph>mediators</emph> of those effects (trauma reminders, poverty, family strife).</p> <p>We compared the common factor–effect indicator model (derived through EFA) to a rationally derived composite–causal indicator model in a split‐sample replication design. Results generally supported the four study hypotheses and illustrated the relative strengths (and some potential drawbacks) of the two methods. The common factor model extracted clusters of co‐occurring events that reflected idiosyncratic features of how the Bosnian conflict was waged (e.g., sieges, ethnic cleansing). The choice of test construction method is thus consequential, as the two methods derived dimensions of war exposure that differed substantially in their elements and relations with theorized mediators and outcomes. Both methods produced good model fit and explained substantial proportions of variance in outcomes. The partially mediated models fit better than their fully mediated counterparts, providing support for the hypothesis that "highly traumatogenic" prewar and war‐time risk factors would exert direct predictive effects. However, most predictive effects were transmitted via postwar mediators (trauma reminders and adversities). The composite–causal indicator model produced clearer evidence of differential relations linking prewar and war‐time trauma to postwar outcomes. In contrast, the common factor–effect indicator model showed less differentiation, as illustrated by the marked diversity in types of war exposure loading the Siege factor and its dominance as the primary risk factor for all outcomes.</p> <p>We now discuss the implications of these findings by addressing five questions relating to theory building, strengthening methodological rigor, and improving risk screening, triage, and intervention.</p> <hd id="AN0052214229-14">Study Implications: Five Questions Regarding Theory Building, Research Methods, and Intervent...</hd> <p> <bold>Question 1: Which measurement approach is preferable for modeling war exposure?. </bold> The good fit of the Subsample 2 composite model is consistent with the assertion that specific types of trauma and loss (e.g., traumatic deaths) may generate similar consequences (e.g., PTSD) without necessarily sharing the same causal origin (i.e., processes underlying a common factor that <emph>cause</emph> the indicators to occur and co‐occur) or manifesting "evidence" of indicator homogeneity. Our results suggest that which method is preferable depends on which part of the model—the <emph>structure of co‐occurrences among traumatic events</emph>, versus the <emph>structure of paths that differentially link the occurrences of specific types of trauma and loss to specific causal consequences</emph>—is of primary interest. The "big picture" focus of the common factor model makes it well suited for "first‐pass" questions that arise in the <emph>situation analysis</emph> phase of crisis response: "How was the war waged—which clusters of events co‐occurred?" and "Who is at (global) risk due to the accumulation of co‐occurring risk factors?" In contrast, the composite–causal indicator model is more useful for addressing deeper questions concerning <emph>which types of trauma and loss place which subgroups at risk for which outcomes via which pathways of influence</emph>. Such issues are more relevant to the needs assessment, risk screening/triage, case conceptualization, intervention planning, program evaluation, and prevention phases of response ([<reflink idref="bib29" id="ref98">29</reflink>]) given that they address questions such as: "How should we prioritize trauma types according to severity?""How can we use limited mental health resources most wisely?" and "Which mediating pathways should our interventions target and seek to interrupt?"</p> <p>Our results also raise a caveat: "Packing" trauma exposure risk factors into dimensions based on an incorrect assumption regarding their causal origin (i.e., indicator homogeneity) may undermine efforts to "unpack" and clarify their differential relations with their respective causal consequences. Understanding that inter‐item covariances <emph>reflect event co‐occurrences, not item homogeneity</emph>, is the key to interpreting our puzzling results: The inverse loadings of <emph>mother killed</emph> and <emph>sibling killed</emph> on <emph>Expulsion & Displacement</emph> (Factor 2) reflect the grim strategy of ethnic cleansing—to frighten "undesirable" ethnic groups into fleeing a region by killing those who remain behind. The inverse loading of <emph>left country because of the war</emph> on <emph>Siege</emph> (Factor 1) reflects the difficulties inherent in escaping a siege. The failure of <emph>father killed</emph> to load any factor reflects the widespread conscription of fathers, who were exposed to lethal risks at the front lines that were largely independent of the risks to which their wives and children were exposed. Thus, a serious model–application mismatch may arise when the "real‐world" structure among event types is incompatible with the assumptions of indicator homogeneity and of indices often used to gauge homogeneity (e.g., alpha; see [<reflink idref="bib55" id="ref99">55</reflink>]). Although dropping inversely loaded items may drive up estimates of "internal consistency," it reduces test validity by reducing the correspondence between the test and the structure among variables it purports to measure ([<reflink idref="bib7" id="ref100">7</reflink>], [<reflink idref="bib8" id="ref101">8</reflink>]). These results illustrate how complex patterns of "real‐life" nonhomogeneous events can generate complex factor structures that faithfully mirror them (e.g., <emph>Expulsion & Displacement</emph> reflects both being driven from one's home and having one's mother and siblings <emph>not</emph> killed).</p> <p> <bold>Question 2: Are some types of trauma exposure more potent causal risk factors than others, and if so, can they be prioritized?. </bold>Table 3 illustrates the potential utility of effects decomposition via SEM for identifying who is at risk, for what outcomes, via which pathways. It is striking that all high‐ranking risk factors consisted of events (life threat, traumatic death) consistently identified as traumatogenic ([<reflink idref="bib26" id="ref102">26</reflink>]). These findings suggest that different types of trauma and loss "specialize," to a meaningful extent, in the types of risks they convey and the pathways through which they transmit their effects, permitting them to be ranked by total effect in reference to focal outcomes. This differentiation‐focused method may help the field move beyond identifying "at‐risk" individuals using ambiguous general markers of risk ([<reflink idref="bib2" id="ref103">2</reflink>]) to instead stratifying them according to their risks for specific outcomes given their exposures to specific risk factors.</p> <p> <bold>Question 3: How can we build our knowledge base in ways that help us develop risk assessment tools that can accurately predict who will need what types of help, across diverse settings and traumatic events?. </bold> The composite‐based method may hold three important advantages over the common factor approach for theory building and intervention applications. First, it promotes the construction of content‐valid assessment instruments <emph>that are tailored to specific settings</emph> because its aim is to capture a census rather than a "best internally consistent sample" of trauma exposure. Second, it unpacks settings into <emph>common denominator constructs</emph> (e.g., life threat, traumatic death) that are more likely to be consistent in their composition and causal effects across diverse settings than event‐specific variables (e.g., <emph>Siege</emph>). "Packaging" concepts at the more abstract level of <emph>dimensions of trauma types</emph> enhances the transportability of variables, measures, and findings across diverse settings ([<reflink idref="bib2" id="ref104">2</reflink>]). Thus, one does not typically import (at least unmodified) measures across settings but rather, a general conceptual framework of exposure dimensions to serve as a test construction template. "Event tailoring" an exposure measure is guided by a two‐step algorithm: <emph>(a) Did exposure to a dimension (e.g., life threat) occur? (b) If yes, how was it manifest in this setting? Generate or modify as many items as needed</emph> (e.g., using qualitative rapid needs assessment; see [<reflink idref="bib4" id="ref105">4</reflink>]) <emph>to maximize content, construct, and ecological validity</emph>. Third, "packaging" findings across studies into a common currency of exposure dimensions will help the field learn more efficiently across diverse events by facilitating cross‐pollinating discourse between sub‐literatures (e.g., war, terrorism, hurricanes) that have been unnecessarily compartmentalized by setting‐specific, event‐bound measures and variables ([<reflink idref="bib26" id="ref106">26</reflink>]). This will increase the growth rate, generalizability, and practical utility of the knowledge base while permitting meta‐analytic studies that treat event type or setting as a moderator (e.g., life threat during torture or rape may be more traumatogenic, or may produce more distressing secondary adversities or trauma reminders, than life threat during a hurricane; [<reflink idref="bib49" id="ref107">49</reflink>]).</p> <p> <bold>Question 4: How can these methods promote the study of development in high‐risk settings?. </bold> Findings that prewar trauma exerted direct predictive effects 5 years postwar underscores the potency of adverse childhood events in even highly stressful contexts, and suggests that <emph>when</emph> traumatic events occur may be as consequential as <emph>what</emph> occurs. Moreover, the complementary strengths of the common factor and composite‐based models suggest that both models, used in tandem, can "unpack" complex events and their aftermath ([<reflink idref="bib11" id="ref108">11</reflink>]). Examples include <emph>unpacking risk factor caravans</emph> and <emph>resource caravans</emph>. These caravans respectively consist of clusters of co‐occurring risk and promotive/protective factors (e.g., social support) that "travel" with their host across development ([<reflink idref="bib29" id="ref109">29</reflink>]). The common factor and related methods are well suited to unpack configurations of risk factors and resources that co‐occur within and across developmental stages. Conversely, the composite approach is better suited to unpack the differential relations between individual caravan elements and their respective consequences. Such a dual‐pronged approach can support not only secondary and tertiary interventions that seek to reduce psychopathology and dysfunction, but also strength‐based public health approaches that emphasize prevention and positive development through the orchestration of resource gain cycles. This dual approach can also help to disentangle both the complex array of risk factors theorized to <emph>cause</emph> developmental trauma disorder—a constellation of cognitive, affective, somatic, and behavioral problems now under consideration for <emph>DSM–V</emph> ([<reflink idref="bib13" id="ref110">13</reflink>])—as well as its structure. The composite model broadly applies ([<reflink idref="bib17" id="ref111">17</reflink>]): "Latent" sieges are no more a <emph>cause</emph> of shelling youth than physical abuse causes beatings, emotional abuse causes screaming, and community violence causes stabbings.</p> <p>The common factor and composite methods can also be used to study differential risks for exposure to types of trauma as a function of developmental stage. Because adolescents often gather supplies in war settings and have social networks populated with peers and adult friends ([<reflink idref="bib34" id="ref112">34</reflink>]), they may be at higher risk than children for many types of exposure, including life threat, witnessing injury or death, and having a close friend killed or injured. In contrast, children and adolescents may be at equivalent risk for loss of a parent. Differences in child versus adolescent reactions to war exposure may thus reflect differential exposure <emph>to</emph> specific types of trauma, differential effects <emph>of</emph> those exposures, and maturational and other contextual effects (Pat‐Horenczyk et al., 2009; [<reflink idref="bib52" id="ref113">52</reflink>]; [<reflink idref="bib54" id="ref114">54</reflink>]).</p> <p> <bold>Question 5: How can these findings improve assessment and intervention?. </bold> The limited utility of the common factor model for clinical applications is illustrated by the question, "What does knowing that siege exposure is the primary risk factor for every outcome really tell us about how to help war‐exposed youth?"<emph>Co‐occurring</emph> risk factors (e.g., all indicators loading <emph>Siege</emph>) are risk markers for one another, and for general "at risk" status ([<reflink idref="bib2" id="ref115">2</reflink>]); thus, they help to identify <emph>who may need help</emph>. However, such risk markers provide very little guidance regarding <emph>how, where, when, and why to help</emph>. To the extent that it disperses equifinal risk factors (e.g., types of traumatic death) across different test dimensions, aggregates causally diverse risk factors into the same dimension (e.g., <emph>Siege</emph>), or drops "uncooperative" critical items (e.g., <emph>father killed</emph>), the common factor method can undermine assessment in four ways. First, causally complex exposure dimensions are ambiguous: Clinicians do not diagnose or treat clients for siege exposure per se. Such "risk factors" are too complex and vague in their makeup to serve any purpose other than general "markers of risk" ([<reflink idref="bib2" id="ref116">2</reflink>]). Second, ambiguous test scores increase the risk for false negatives and false positives ([<reflink idref="bib24" id="ref117">24</reflink>]): A youth who is devastated by witnessing his father's death may score 3 on a war exposure inventory (and be overlooked), whereas a youth exposed to many less severe events may score 10. Third, because its "data input" is the structure of covariances <emph>among risk factors</emph>, the common factor model can create "context‐specific" variables (e.g., <emph>Siege</emph>) and tests that reflect idiosyncratic ways in which risk factors co‐occurred within a given setting and that have limited transportability across settings ([<reflink idref="bib2" id="ref118">2</reflink>]).</p> <p>A fourth problem is that causally diverse variables dilute, blur, and obscure, rather than clarify, differential relations between causes and consequences. <emph>Siege Exposure</emph> is the primary risk factor for all outcomes in this study <emph>precisely because it is a complex composite of risk factors (or risk markers) for all outcomes</emph>. Our capacity to accurately predict who will need services ([<reflink idref="bib2" id="ref119">2</reflink>]) hinges on our ability to differentiate between the potencies, pathways of influence, and consequences of specific causal risk factors. Failure to discriminate adversely affects all "downstream" activities that require knowledge of <emph>who is at risk for what outcomes via which pathways of influence</emph> to make good evidence‐based clinical decisions. Models in which most risk factors relate indiscriminately to most outcomes make elimination of <emph>nonsignificant</emph> pathways difficult. This reduction in model parsimony can both increase false‐negative rates (by not ruling out those <emph>not</emph> at high risk for specific outcomes) and reduce efficiency (as specialized treatment components are assigned unnecessarily). For example, our results suggest that siege‐exposed youths are at highest risk for all outcomes and thus are of higher priority than all other groups and should receive broad‐spectrum treatments targeting all outcomes.</p> <p>In contrast, better differentiated composite–causal indicator models may promote cost‐effective, public health‐based interventions ([<reflink idref="bib29" id="ref120">29</reflink>]). Because subgroups most likely need services for those outcomes for which their types of exposure place them at risk, differentiated models can balance the <emph>level of need</emph> (groups at higher risk receive more specialized services), <emph>effectiveness</emph> (necessary services are provided), and <emph>efficiency</emph> (no unnecessary specialized services are provided). Differentiated models help to identify the specific types of trauma and loss that exert the greatest causal influence on each outcome and their respective pathways of influence. This knowledge will shed much‐needed light on the key determinants of the clinical course of postwar distress and help to identify, prioritize, and triage youths and families to components that target their difficulties. A major advantage of differentiated models is their capacity to identify "<emph>high‐value</emph>"<emph>targets for intervention within the broader social and physical ecology in ways that facilitate assessment‐driven, theoretically grounded, and evidence‐based treatment tailoring at the individual case level</emph>. Such an approach will encourage wellness, prevention, and intervention efforts that systematically address youth's specific life circumstances (e.g., orchestrating personal resource gain cycles to build "stress resistance‐enhancing" resource caravans) rather than assigning treatment components to individuals solely on the basis of clinical cutoff scores on "end point" distress and dysfunction measures (Layne, Beck, et al., 2009). Subgroups exposed only to "indirect effect" (fully mediated) types of trauma may receive substantial and perhaps sufficient benefit from supportive skills‐based components that interrupt key mediators and mitigate key moderators ([<reflink idref="bib36" id="ref121">36</reflink>]). Evidence that existential adversities (financial hardships) and trauma reminders (destroyed buildings) mediate the links between war exposure and postwar distress blurs the traditional distinction between "material" versus "psychosocial" postwar interventions: Rebuilding infrastructure to improve access to basic necessities, employment, housing, and schools is also a potent psychosocial intervention—a major added benefit for the same investment ([<reflink idref="bib42" id="ref122">42</reflink>]). Further, evidence that highly traumatogenic events exert both indirect <emph>and</emph> direct effects suggests ways to introduce supplemental specialized components for high‐risk groups for which mediator‐focused interventions may be insufficient. For example, Trauma and Grief Component Therapy ([<reflink idref="bib31" id="ref123">31</reflink>]) is a multitiered school‐based intervention that includes "Tier 1" classroom components targeting key mediators (e.g., coping with trauma reminders) and moderators (e.g., recruiting social support), and "Tier 2" group‐based components targeting highly traumatogenic events (trauma and grief processing). A program evaluation ([<reflink idref="bib14" id="ref124">14</reflink>]; [<reflink idref="bib31" id="ref125">31</reflink>]) found that Tier 1 components were widely disseminated at low cost, producing substantial benefit with low iatrogenic risk. The Tier 2 condition produced improvement rates comparable to rigorously controlled treatment efficacy trials.</p> <hd id="AN0052214229-15">Study Limitations</hd> <p>This study used a self‐report screening index of life events ([<reflink idref="bib43" id="ref126">43</reflink>]) and did not include measures of loss reminders (a theorized mediator; [<reflink idref="bib35" id="ref127">35</reflink>]) and grief reactions; both measures are included in a forthcoming study. The cross‐sectional retrospective design precludes both causal inference (as event reporting may be confounded with current psychological functioning) and the longitudinal analysis of adjustment trajectories ([<reflink idref="bib29" id="ref128">29</reflink>]). The extended time frame (5 years postwar) created a window in which other risk factors (especially postwar trauma) could occur and compete for explained variance, allowing perhaps only the most traumatogenic or impactful war‐time risk factors to exert significant predictive effects. The time frame also sampled youths 8–14 years old at the <emph>end</emph> of the war—a group less likely to be severely exposed than older adolescents. This may have led to low endorsement rates for some event types (e.g., physical injury) that prevented their predictive effects from being adequately tested. In addition, the purposive sampling of classes with a high prevalence of severely war‐exposed students limits the generalizability of findings to more severely exposed youths residing in the region. Moreover, the variance decomposition method used to rank risk factors assumes that all nontrivial variables are included in the model; omitting influential variables may alter parameter estimates and risk factor rankings ([<reflink idref="bib9" id="ref129">9</reflink>]). Further study is also needed to evaluate the divergence in performance between the two approaches, as well as the generalizability and utility of the eight‐dimensional framework used in the composite approach, in other settings and applications. Last, although evidence for its utility is accumulating ([<reflink idref="bib11" id="ref130">11</reflink>]), critics (e.g., [<reflink idref="bib59" id="ref131">59</reflink>]) question the applicability of the composite model; for example, different research teams could produce different sets of indicators purporting to measure the same composite. We concur that this is possible but assert that the caveat applies with equal force to the common factor model, which may produce "researcher‐idiosyncratic" discrepancies arising from item wordings, guiding theory, and so forth.</p> <ref id="AN0052214229-16"> <title> Footnotes </title> <blist> <bibl id="bib1" idref="ref60" type="bt">1</bibl> <bibtext> Financial support was provided by UNICEF Bosnia & Herzegovina, the BYU Family Studies Center, the David M. Kennedy Center for International Studies, the Bing Fund, and Tony Bennett. The authors thank Preston Finley, James Wu, Ann Anderson, and Jenifer Alonso for assisting with manuscript preparation. Special thanks to Theresa Betancourt, Alan Steinberg, Sarah Ostrowski, Harolyn Belcher, and two anonymous reviewers for their generous and insightful comments on previous drafts of this manuscript.</bibtext> </blist> </ref> <ref id="AN0052214229-17"> <title> References </title> <blist> <bibtext> Ahmetasevic, N. (2007, June 20). Bosnia's book of the dead. Justice Report, Balkan Investigative Reporting Network. Retrieved May 10, 2010, from <ulink href="http://birn.eu.com/en/88/10/3377/">http://birn.eu.com/en/88/10/3377/</ulink>.</bibtext> </blist> <blist> <bibl id="bib2" idref="ref22" type="bt">2</bibl> <bibtext> APA Presidential Task Force on Posttraumatic Stress Disorder and Trauma in Children and Adolescents. (2008). Children and trauma: Update for mental health professionals. Washington, DC: American Psychological Association.</bibtext> </blist> <blist> <bibl id="bib3" idref="ref5" type="bt">3</bibl> <bibtext> Barenbaum, J., Ruchkin, V., & Schwab‐Stone, M. (2004). The psychosocial aspects of children exposed to war: Practice and policy initiatives. Journal of Child Psychology and Psychiatry, 45, 41 – 62.</bibtext> </blist> <blist> <bibl id="bib4" idref="ref105" type="bt">4</bibl> <bibtext> Betancourt, T. S., Speelman, L., Onyango, G., & Bolton, P. (2009). A qualitative study of mental health problems among children displaced by war in Northern Uganda. Transcultural Psychiatry, 46, 238 – 256.</bibtext> </blist> <blist> <bibl id="bib5" idref="ref78" type="bt">5</bibl> <bibtext> Birelson, P. (1981). The validity of depressive disorder in childhood and development of a self‐rating scale: A research project. Journal of Clinical Psychology and Psychiatry, 22, 73 – 88.</bibtext> </blist> <blist> <bibl id="bib6" idref="ref32" type="bt">6</bibl> <bibtext> Bollen, K., & Lennox, R. (1991). Conventional wisdom on measurement: A structural equation perspective. Psychological Bulletin, 110, 305 – 314.</bibtext> </blist> <blist> <bibl id="bib7" idref="ref34" type="bt">7</bibl> <bibtext> Borsboom, D., Mellenbergh, G. J., & Van Heerden, J. (2003). The theoretical status of latent variables. Psychological Review, 110, 203 – 219.</bibtext> </blist> <blist> <bibl id="bib8" idref="ref52" type="bt">8</bibl> <bibtext> Borsboom, D., Mellenbergh, G. J., & Van Heerden, J. (2004). The concept of validity. Psychological Review, 111, 1061 – 1071.</bibtext> </blist> <blist> <bibl id="bib9" idref="ref88" type="bt">9</bibl> <bibtext> Byrne, B. (2001). Structural equation modeling with AMOS: Basic concepts, applications, and programming. New York: Psychology Press.</bibtext> </blist> <blist> <bibtext> Chen, F. F., West, S. G., & Sousa, K. H. (2006). A comparison of bifactor and second‐order models of quality of life. Multivariate Behavioral Research, 41, 189 – 224.</bibtext> </blist> <blist> <bibtext> Coltman, T., Devinney, T. M., Midgley, D. F., & Venaik, S. (2008). Formative versus reflective measurement models: Two applications of formative measurement. Journal of Business Research, 61, 1250 – 1262.</bibtext> </blist> <blist> <bibtext> Comrey, A. L., & Lee, H. B. (1992). A first course in factor analysis (2nd ed.). Hillsdale, NJ: Erlbaum.</bibtext> </blist> <blist> <bibtext> Courtois, C. A., & Gold, S. N. (2009). The need for inclusion of psychological trauma in the professional curriculum: A call to action. Psychological Trauma: Theory, Research, Practice, and Policy, 1, 3 – 23.</bibtext> </blist> <blist> <bibtext> Cox, J., Davies, D. R., Burlingame, G. M., Campbell, J. E., Layne, C. M., & Katzenbach, R. J. (2007). Effectiveness of a trauma/grief‐focused group intervention: A qualitative study with war‐exposed Bosnian adolescents. International Journal of Group Psychotherapy, 57, 319 – 345.</bibtext> </blist> <blist> <bibtext> Cronbach, L. J. (1951). Coefficient alpha and the internal structure of tests. Psychometrika, 16, 297 – 334.</bibtext> </blist> <blist> <bibtext> Diamantopoulos, A., Riefler, P., & Roth, K. P. (2008). Advancing formative measurement models. Journal of Business Research, 61, 1203 – 1218.</bibtext> </blist> <blist> <bibtext> Dodge, K. A., Greenberg, M. T., Malone, P. S., & Conduct Problems Prevention Research Group (2008). Testing an idealized dynamic cascade model of the development of serious violence in adolescence. Child Development, 79, 1907 – 1927.</bibtext> </blist> <blist> <bibtext> Dyregrov, A., Gupta, L., Gjestad, R., & Mukanoheli, E. (2000). Trauma exposure and psychological reactions to genocide among Rwandan children. Journal of Traumatic Stress, 13, 3 – 21.</bibtext> </blist> <blist> <bibtext> Entholt, K., & Yule, W. (2006). Practitioner review: Assessment and treatment of refugee children and adolescents who have experienced war‐related trauma. Journal of Child Psychology and Psychiatry, 47, 1197 – 1210.</bibtext> </blist> <blist> <bibtext> Fayyad, J., Karam, E., Karam, A. N., Tabet, C. C., & Ghosn, M. B. (2004). PTSD in children and adolescents following war. In R. R. Silva (Ed.), Posttraumatic stress disorders in children and adolescents (pp. 306 – 352). New York: W.W. Norton.</bibtext> </blist> <blist> <bibtext> Franke, G. R., Preacher, K. J., & Rigdon, E. E. (2008). Proportional structural effects of formative indicators. Journal of Business Research, 61, 1229 – 1237.</bibtext> </blist> <blist> <bibtext> Goodman, L. A., Corcoran, C., Turner, K., Yuan, N., & Green, B. L. (1998). Assessing traumatic event exposure: General issues and preliminary findings for the Stressful Life Events Screening Questionnaire. Journal of Traumatic Stress, 11, 521 – 542.</bibtext> </blist> <blist> <bibtext> Grace, J. B., & Bollen, K. A. (2008). Representing general theoretical concepts in structural equation models: The role of composite variables. Environmental and Ecological Statistics, 15, 191 – 213.</bibtext> </blist> <blist> <bibtext> Kraemer, H. C., Kazdin, A. E., Offord, D. R., Kessler, R. C., Jensen, P. S., & Kupfer, D. J. (1999). Measuring the potency of risk factors for clinical or policy significance. Psychological Methods, 4, 257 – 271.</bibtext> </blist> <blist> <bibtext> Krueger, R. F. (1999). The structure of common mental disorders. Archives of General Psychiatry, 56, 921 – 926.</bibtext> </blist> <blist> <bibtext> La Greca, A. M., Silverman, W. K., Vernberg, E. M., & Roberts, M. C. (2002). Helping children cope with disasters and terrorism. Washington, DC: APA.</bibtext> </blist> <blist> <bibtext> Layne, C. M. (1999a). Pre‐War Trauma Exposure Index. Unpublished psychological test, University of California, Los Angeles.</bibtext> </blist> <blist> <bibtext> Layne, C. M. (1999b). Post‐War Trauma Exposure Index. Unpublished psychological test, University of California, Los Angeles.</bibtext> </blist> <blist> <bibtext> Layne, C. M., Beck, C. J., Rimmasch, H., Southwick, J. S., Moreno, M. A., & Hobfoll, S. E. (2009). Promoting "resilient" posttraumatic adjustment in childhood and beyond: "Unpacking" life events, adjustment trajectories, resources, and interventions. In D. Brom, R. Pat‐Horenczyk, & J. Ford (Eds.). Treating traumatized children: Risk, resilience, and recovery (pp. 13 – 47). New York: Routledge.</bibtext> </blist> <blist> <bibtext> Layne, C. M., & Djapo, N. (1999). Post‐War Adversities Index. Unpublished psychological test, University of California, Los Angeles.</bibtext> </blist> <blist> <bibtext> Layne, C. M., Saltzman, W. R., Poppleton, L., Burlingame, G. M., Pasalic, A., Durakovic, E., et al. (2008). Effectiveness of a school‐based group psychotherapy program for war‐exposed adolescents: A randomized controlled trial. Journal of the American Academy of Child and Adolescent Psychiatry, 47, 1048 – 1062.</bibtext> </blist> <blist> <bibtext> Layne, C. M., Steinberg, A. M., & Pynoos, R. S. (1998). Trauma Reminders Screening Inventory. Unpublished psychological test, University of California, Los Angeles.</bibtext> </blist> <blist> <bibtext> Layne, C. M., Stuvland, R., Saltzman, W. R., Djapo, N., & Pynoos, R. S. (1999). War Trauma Exposure Index. Unpublished psychological test, University of California, Los Angeles.</bibtext> </blist> <blist> <bibtext> Layne, C. M., Warren, J. S., Hilton, S., Lin, D., Pasalic, A., Fulton, J., et al. (2009). Measuring adolescent perceived support amidst war and disaster: The Multi‐Sector Social Support Inventory. In B. K. Baker (Ed.), Adolescents and war: How youth deal with political violence (pp. 145 – 176). New York: Oxford University Press.</bibtext> </blist> <blist> <bibtext> Layne, C. M., Warren, J. S., Saltzman, W. R., Fulton, J. B., Steinberg, A. M., & Pynoos, R. S. (2006). Contextual influences on posttraumatic adjustment: Retraumatization and the roles of revictimization, posttraumatic adversities, and distressing reminders. In L. A. Schein, H. I. Spitz, G. M. Burlingame, P. R. Muskin, & S. Vargo (Eds.), Psychological effects of catastrophic disasters: Group approaches to treatment (pp. 235 – 286). New York: Haworth.</bibtext> </blist> <blist> <bibtext> Layne, C. M., Warren, J. S., Watson, P. J., & Shalev, A. Y. (2007). Risk, vulnerability, resistance, and resilience: Toward an integrative conceptualization of posttraumatic adaptation. In M. J. Friedman, T. M. Keane, & P. A. Resnick (Eds.), Handbook of PTSD: Science and practice (pp. 497 – 520). New York: Guilford.</bibtext> </blist> <blist> <bibtext> Little, T. D., Lindenberger, U., & Nesselroade, J. R. (1999). On selecting indicators for multivariate measurement and modeling with latent variables: When "good" indicators are bad and "bad" indicators are good. Psychological Methods, 4, 192 – 211.</bibtext> </blist> <blist> <bibtext> MacDonald, R. P. (1999). Test theory: A unified treatment. Mahwah, NJ: Erlbaum.</bibtext> </blist> <blist> <bibtext> MacKenzie, S. B., Podsakoff, P. M., & Jarvis, C. B. (2005). The problem of measurement model misspecification in behavioral and organizational research and some recommended solutions. Journal of Applied Psychology, 90, 710 – 730.</bibtext> </blist> <blist> <bibtext> Macksoud, M. S., & Aber, J. L. (1996). The war experiences and psychosocial development of children in Lebanon. Child Development, 67, 70 – 88.</bibtext> </blist> <blist> <bibtext> Masten, A. S., Neemann, J., & Andenas, S. (1994). Life events and adjustment in adolescents: The significance of event independence, desirability, and chronicity. Journal of Research on Adolescence, 4, 71 – 97.</bibtext> </blist> <blist> <bibtext> Miller, K. E., & Rasmussen, A. (2010). War exposure, daily stressors, and mental health in conflict and post‐conflict settings: Bridging the divide between trauma‐focused and psychosocial frameworks. Social Science & Medicine, 70, 7 – 16.</bibtext> </blist> <blist> <bibtext> Monroe, S. M. (2008). Modern approaches to conceptualizing and measuring human life stress. Annual Review of Clinical Psychology, 4, 33 – 52.</bibtext> </blist> <blist> <bibtext> Mulaik, S. A. (2009). Linear causal modeling with structural equations. New York: Taylor & Francis.</bibtext> </blist> <blist> <bibtext> Muthén, B. O. (2002). Beyond SEM: General latent variable modeling. Behaviormetrika, 29, 81 – 117.</bibtext> </blist> <blist> <bibtext> Muthén, B. O., & Muthén, L. (2009). Mplus 5.1. Los Angeles, CA: Author.</bibtext> </blist> <blist> <bibtext> Netland, M. (2001). Assessment of exposure to political violence and other potentially traumatizing events. Journal of Traumatic Stress, 14, 311 – 326.</bibtext> </blist> <blist> <bibtext> Netland, M. (2005). Event‐list construction and treatment of exposure data in research on political violence. Journal of Traumatic Stress, 18, 507 – 517.</bibtext> </blist> <blist> <bibtext> Norris, F. H., Galea, S., Friedman, M. J., & Watson, P. J. (Eds.). (2006). Methods for disaster mental health research. New York: Guilford.</bibtext> </blist> <blist> <bibtext> Pat‐Horenczyk, R., Qasrawi, R., Lesack, R., Haj‐Yahia, M., Peled, O., Shaheen, M., et al. (2009). Posttraumatic symptoms, functional impairment, and coping among adolescents on both sides of the Israeli–Palestinian conflict: A cross‐cultural approach. Applied Psychology: An International Review, 58, 688 – 708.</bibtext> </blist> <blist> <bibtext> Powell, S., & Duraković‐Belko, E. (Eds.). (2002). Sarajevo 2000: The psychosocial consequences of war. Results of empirical research from the territory of former Yugoslavia. Retrieved May 10, 2010, from <ulink href="http://psih.org/2000e.pdf">http://psih.org/2000e.pdf</ulink>.</bibtext> </blist> <blist> <bibtext> Pynoos, R. S., Steinberg, A. M., & Wraith, R. (1995). A developmental model of childhood traumatic stress. In D. Cicchetti & D. J. Cohen (Eds.), Developmental psychopathology: Vol. 2. Risk, disorder, and adaptation. Wiley Series on personality processes (pp. 72 – 95). Oxford, England: Wiley.</bibtext> </blist> <blist> <bibtext> Richters, J. E. (1997). The Hubble hypothesis and the developmentalist's dilemma. Development and Psychopathology, 9, 193 – 229.</bibtext> </blist> <blist> <bibtext> Shaw, J. A. (2003). Children exposed to war/terrorism. Clinical Child and Family Psychology Review, 6, 237 – 246.</bibtext> </blist> <blist> <bibtext> Sijtsma, K. (2009). On the use, the misuse, and the very limited usefulness of Cronbach's alpha. Psychometrika, 74, 107 – 120.</bibtext> </blist> <blist> <bibtext> SPSS. (2009). Amos 16.0 (Software). Chicago: Author.</bibtext> </blist> <blist> <bibtext> Steinberg, A. M., Brymer, M., Decker, K., & Pynoos, R. S. (2004). The UCLA PTSD Reaction Index. Current Psychiatry Reports, 6, 96 – 100.</bibtext> </blist> <blist> <bibtext> Wells, M., Burlingame, G., Lambert, M., Hoag, M., & Hope, C. (1996). Conceptualization and measurement of patient change during psychotherapy: Development of the Outcome Questionnaire and the Youth Outcome Questionnaire. Psychotherapy, 33, 275 – 283.</bibtext> </blist> <blist> <bibtext> Wilcox, J. B., Howell, R. D., & Breivik, E. (2008). Questions about formative measurement. Journal of Business Research, 61, 1219 – 1228.</bibtext> </blist> </ref> <aug> <p>By Christopher M. Layne; Joseph A. Olsen; Aaron Baker; John‐Paul Legerski; Brian Isakson; Alma Pašalić; Elvira Duraković‐Belko; Nermin Đapo; Nihada Ćampara; Berina Arslanagić; William R. Saltzman and Robert S. Pynoos</p> <p>Reported by Author; Author; Author; Author; Author; Author; Author; Author; Author; Author; Author; Author</p> </aug> <nolink nlid="nl1" bibid="bib49" firstref="ref1"></nolink> <nolink nlid="nl2" bibid="bib18" firstref="ref2"></nolink> <nolink nlid="nl3" bibid="bib40" firstref="ref3"></nolink> <nolink nlid="nl4" bibid="bib31" firstref="ref4"></nolink> <nolink nlid="nl5" bibid="bib19" firstref="ref6"></nolink> <nolink nlid="nl6" bibid="bib20" firstref="ref7"></nolink> <nolink nlid="nl7" bibid="bib54" firstref="ref8"></nolink> <nolink nlid="nl8" bibid="bib29" firstref="ref9"></nolink> <nolink nlid="nl9" bibid="bib42" firstref="ref11"></nolink> <nolink nlid="nl10" bibid="bib50" firstref="ref13"></nolink> <nolink nlid="nl11" bibid="bib47" firstref="ref16"></nolink> <nolink nlid="nl12" bibid="bib48" firstref="ref17"></nolink> <nolink nlid="nl13" bibid="bib15" firstref="ref19"></nolink> <nolink nlid="nl14" bibid="bib36" firstref="ref20"></nolink> <nolink nlid="nl15" bibid="bib24" firstref="ref23"></nolink> <nolink nlid="nl16" bibid="bib53" firstref="ref24"></nolink> <nolink nlid="nl17" bibid="bib44" firstref="ref26"></nolink> <nolink nlid="nl18" bibid="bib34" firstref="ref27"></nolink> <nolink nlid="nl19" bibid="bib38" firstref="ref28"></nolink> <nolink nlid="nl20" bibid="bib23" firstref="ref29"></nolink> <nolink nlid="nl21" bibid="bib39" firstref="ref30"></nolink> <nolink nlid="nl22" bibid="bib55" firstref="ref36"></nolink> <nolink nlid="nl23" bibid="bib16" firstref="ref38"></nolink> <nolink nlid="nl24" bibid="bib21" firstref="ref39"></nolink> <nolink nlid="nl25" bibid="bib59" firstref="ref40"></nolink> <nolink nlid="nl26" bibid="bib26" firstref="ref46"></nolink> <nolink nlid="nl27" bibid="bib25" firstref="ref56"></nolink> <nolink nlid="nl28" bibid="bib10" firstref="ref57"></nolink> <nolink nlid="nl29" bibid="bib37" firstref="ref58"></nolink> <nolink nlid="nl30" bibid="bib51" firstref="ref61"></nolink> <nolink nlid="nl31" bibid="bib52" firstref="ref63"></nolink> <nolink nlid="nl32" bibid="bib35" firstref="ref64"></nolink> <nolink nlid="nl33" bibid="bib27" firstref="ref66"></nolink> <nolink nlid="nl34" bibid="bib22" firstref="ref67"></nolink> <nolink nlid="nl35" bibid="bib33" firstref="ref68"></nolink> <nolink nlid="nl36" bibid="bib12" firstref="ref72"></nolink> <nolink nlid="nl37" bibid="bib28" firstref="ref74"></nolink> <nolink nlid="nl38" bibid="bib32" firstref="ref75"></nolink> <nolink nlid="nl39" bibid="bib57" firstref="ref76"></nolink> <nolink nlid="nl40" bibid="bib58" firstref="ref80"></nolink> <nolink nlid="nl41" bibid="bib30" firstref="ref81"></nolink> <nolink nlid="nl42" bibid="bib41" firstref="ref82"></nolink> <nolink nlid="nl43" bibid="bib46" firstref="ref84"></nolink> <nolink nlid="nl44" bibid="bib45" firstref="ref85"></nolink> <nolink nlid="nl45" bibid="bib56" firstref="ref87"></nolink> <nolink nlid="nl46" bibid="bib79" firstref="ref89"></nolink> <nolink nlid="nl47" bibid="bib60" firstref="ref91"></nolink> <nolink nlid="nl48" bibid="bib11" firstref="ref108"></nolink> <nolink nlid="nl49" bibid="bib13" firstref="ref110"></nolink> <nolink nlid="nl50" bibid="bib17" firstref="ref111"></nolink> <nolink nlid="nl51" bibid="bib14" firstref="ref124"></nolink> <nolink nlid="nl52" bibid="bib43" firstref="ref126"></nolink>
Header DbId: eric
DbLabel: ERIC
An: EJ890275
AccessLevel: 3
PubType: Academic Journal
PubTypeId: academicJournal
PreciseRelevancyScore: 0
IllustrationInfo
Items – Name: Title
  Label: Title
  Group: Ti
  Data: Unpacking Trauma Exposure Risk Factors and Differential Pathways of Influence: Predicting Postwar Mental Distress in Bosnian Adolescents
– Name: Language
  Label: Language
  Group: Lang
  Data: English
– Name: Author
  Label: Authors
  Group: Au
  Data: <searchLink fieldCode="AR" term="%22Layne%2C+Christopher+M%2E%22">Layne, Christopher M.</searchLink><br /><searchLink fieldCode="AR" term="%22Olsen%2C+Joseph+A%2E%22">Olsen, Joseph A.</searchLink><br /><searchLink fieldCode="AR" term="%22Baker%2C+Aaron%22">Baker, Aaron</searchLink><br /><searchLink fieldCode="AR" term="%22Legerski%2C+John-Paul%22">Legerski, John-Paul</searchLink><br /><searchLink fieldCode="AR" term="%22Isakson%2C+Brian%22">Isakson, Brian</searchLink><br /><searchLink fieldCode="AR" term="%22Pasalic%2C+Alma%22">Pasalic, Alma</searchLink><br /><searchLink fieldCode="AR" term="%22Durakovic-Belko%2C+Elvira%22">Durakovic-Belko, Elvira</searchLink><br /><searchLink fieldCode="AR" term="%22Dapo%2C+Nermin%22">Dapo, Nermin</searchLink><br /><searchLink fieldCode="AR" term="%22Campara%2C+Nihada%22">Campara, Nihada</searchLink><br /><searchLink fieldCode="AR" term="%22Arslanagic%2C+Berina%22">Arslanagic, Berina</searchLink><br /><searchLink fieldCode="AR" term="%22Saltzman%2C+William+R%2E%22">Saltzman, William R.</searchLink><br /><searchLink fieldCode="AR" term="%22Pynoos%2C+Robert+S%2E%22">Pynoos, Robert S.</searchLink>
– Name: TitleSource
  Label: Source
  Group: Src
  Data: <searchLink fieldCode="SO" term="%22Child+Development%22"><i>Child Development</i></searchLink>. Jul-Aug 2010 81(4):1053-1076.
– Name: Avail
  Label: Availability
  Group: Avail
  Data: Wiley-Blackwell. 350 Main Street, Malden, MA 02148. Tel: 800-835-6770; Tel: 781-388-8598; Fax: 781-388-8232; e-mail: cs-journals@wiley.com; Web site: http://www.wiley.com/WileyCDA/
– Name: PeerReviewed
  Label: Peer Reviewed
  Group: SrcInfo
  Data: Y
– Name: Pages
  Label: Page Count
  Group: Src
  Data: 24
– Name: DatePubCY
  Label: Publication Date
  Group: Date
  Data: 2010
– Name: TypeDocument
  Label: Document Type
  Group: TypDoc
  Data: Journal Articles<br />Reports - Research
– Name: Subject
  Label: Descriptors
  Group: Su
  Data: <searchLink fieldCode="DE" term="%22Intervention%22">Intervention</searchLink><br /><searchLink fieldCode="DE" term="%22Adolescents%22">Adolescents</searchLink><br /><searchLink fieldCode="DE" term="%22Factor+Analysis%22">Factor Analysis</searchLink><br /><searchLink fieldCode="DE" term="%22War%22">War</searchLink><br /><searchLink fieldCode="DE" term="%22Posttraumatic+Stress+Disorder%22">Posttraumatic Stress Disorder</searchLink><br /><searchLink fieldCode="DE" term="%22Predictor+Variables%22">Predictor Variables</searchLink><br /><searchLink fieldCode="DE" term="%22Mental+Disorders%22">Mental Disorders</searchLink><br /><searchLink fieldCode="DE" term="%22Stress+Variables%22">Stress Variables</searchLink><br /><searchLink fieldCode="DE" term="%22At+Risk+Persons%22">At Risk Persons</searchLink><br /><searchLink fieldCode="DE" term="%22Influences%22">Influences</searchLink><br /><searchLink fieldCode="DE" term="%22Comparative+Analysis%22">Comparative Analysis</searchLink><br /><searchLink fieldCode="DE" term="%22Causal+Models%22">Causal Models</searchLink><br /><searchLink fieldCode="DE" term="%22Adjustment+%28to+Environment%29%22">Adjustment (to Environment)</searchLink><br /><searchLink fieldCode="DE" term="%22Coping%22">Coping</searchLink><br /><searchLink fieldCode="DE" term="%22Counseling+Techniques%22">Counseling Techniques</searchLink><br /><searchLink fieldCode="DE" term="%22Counseling+Effectiveness%22">Counseling Effectiveness</searchLink><br /><searchLink fieldCode="DE" term="%22Foreign+Countries%22">Foreign Countries</searchLink><br /><searchLink fieldCode="DE" term="%22Child+Development%22">Child Development</searchLink>
– Name: Subject
  Label: Geographic Terms
  Group: Su
  Data: <searchLink fieldCode="DE" term="%22Bosnia+and+Herzegovina%22">Bosnia and Herzegovina</searchLink>
– Name: DOI
  Label: DOI
  Group: ID
  Data: 10.1111/j.1467-8624.2010.01454.x
– Name: ISSN
  Label: ISSN
  Group: ISSN
  Data: 0009-3920
– Name: Abstract
  Label: Abstract
  Group: Ab
  Data: Methods are needed for quantifying the potency and differential effects of risk factors to identify at-risk groups for theory building and intervention. Traditional methods for constructing war exposure measures are poorly suited to "unpack" differential relations between specific types of exposure and specific outcomes. This study of 881 Bosnian adolescents compared both "common factor-effect indicator" (using exploratory factor analysis) versus "composite causal-indicator" methods for "unpacking" dimensions of war exposure and their respective paths to postwar adjustment outcomes. The composite method better supported theory building and most intervention applications, showing how multitiered interventions can enhance treatment effectiveness and efficiency in war settings. Used together, the methods may unpack the elements and differential effects of "caravans" of risk and promotive factors that co-occur across development.
– Name: AbstractInfo
  Label: Abstractor
  Group: Ab
  Data: As Provided
– Name: DateEntry
  Label: Entry Date
  Group: Date
  Data: 2010
– Name: AN
  Label: Accession Number
  Group: ID
  Data: EJ890275
PLink https://search.ebscohost.com/login.aspx?direct=true&site=eds-live&db=eric&AN=EJ890275
RecordInfo BibRecord:
  BibEntity:
    Identifiers:
      – Type: doi
        Value: 10.1111/j.1467-8624.2010.01454.x
    Languages:
      – Text: English
    PhysicalDescription:
      Pagination:
        PageCount: 24
        StartPage: 1053
    Subjects:
      – SubjectFull: Intervention
        Type: general
      – SubjectFull: Adolescents
        Type: general
      – SubjectFull: Factor Analysis
        Type: general
      – SubjectFull: War
        Type: general
      – SubjectFull: Posttraumatic Stress Disorder
        Type: general
      – SubjectFull: Predictor Variables
        Type: general
      – SubjectFull: Mental Disorders
        Type: general
      – SubjectFull: Stress Variables
        Type: general
      – SubjectFull: At Risk Persons
        Type: general
      – SubjectFull: Influences
        Type: general
      – SubjectFull: Comparative Analysis
        Type: general
      – SubjectFull: Causal Models
        Type: general
      – SubjectFull: Adjustment (to Environment)
        Type: general
      – SubjectFull: Coping
        Type: general
      – SubjectFull: Counseling Techniques
        Type: general
      – SubjectFull: Counseling Effectiveness
        Type: general
      – SubjectFull: Foreign Countries
        Type: general
      – SubjectFull: Child Development
        Type: general
      – SubjectFull: Bosnia and Herzegovina
        Type: general
    Titles:
      – TitleFull: Unpacking Trauma Exposure Risk Factors and Differential Pathways of Influence: Predicting Postwar Mental Distress in Bosnian Adolescents
        Type: main
  BibRelationships:
    HasContributorRelationships:
      – PersonEntity:
          Name:
            NameFull: Layne, Christopher M.
      – PersonEntity:
          Name:
            NameFull: Olsen, Joseph A.
      – PersonEntity:
          Name:
            NameFull: Baker, Aaron
      – PersonEntity:
          Name:
            NameFull: Legerski, John-Paul
      – PersonEntity:
          Name:
            NameFull: Isakson, Brian
      – PersonEntity:
          Name:
            NameFull: Pasalic, Alma
      – PersonEntity:
          Name:
            NameFull: Durakovic-Belko, Elvira
      – PersonEntity:
          Name:
            NameFull: Dapo, Nermin
      – PersonEntity:
          Name:
            NameFull: Campara, Nihada
      – PersonEntity:
          Name:
            NameFull: Arslanagic, Berina
      – PersonEntity:
          Name:
            NameFull: Saltzman, William R.
      – PersonEntity:
          Name:
            NameFull: Pynoos, Robert S.
    IsPartOfRelationships:
      – BibEntity:
          Dates:
            – D: 01
              M: 01
              Type: published
              Y: 2010
          Identifiers:
            – Type: issn-print
              Value: 0009-3920
          Numbering:
            – Type: volume
              Value: 81
            – Type: issue
              Value: 4
          Titles:
            – TitleFull: Child Development
              Type: main
ResultId 1