Current Conceptualization and Operationalization of Adolescents’ Social Capital: A Systematic Review of Self-Reported Instruments

There is a great heterogeneity in the conceptualization and operationalization of social capital in empirical research targeting adolescents. There has not yet been an attempt to systematically map and psychometrically evaluate the existing instruments for measuring social capital that have been developed and validated for adolescent samples. The aim of this systematic review was to identify and evaluate the design and psychometric properties of self-reported instruments for social capital, specifically developed and validated for use among adolescents. The design of this study was a systematic review guided by the COSMIN methodology for systematic reviews of Patient Reported Outcome Measures. The search included six electronic databases and no time frame was applied. Twenty studies were identified as describing the development and validation of a social capital instrument for adolescent samples. The results reveal common denominators, but also great variation in the design and validation of the instruments. Adolescents were only involved in the development procedures of four instruments. There is a lack of social capital instruments that cover both the multidimensionality of social capital and contextual relevance in relation to adolescents. Careful examination of instruments should thus precede a decision when designing studies and further instrument development involving the target group is encouraged.


Introduction
Social capital refers to the sum of resources that individuals access through their social networks [1]. A growing amount of research has explored the relationship between social capital and adolescent mental health since the beginning of the new millennium [2]. Social capital has in this period been shown to be considered as a valuable contributor when explaining inequalities in adolescent health, mental health, and health behaviours [3][4][5][6][7]. Based on findings such as these together with the persistent adverse developments in adolescent mental health [8], which are disproportionately affecting girls [9], research findings on the positive relationship between social capital and adolescent mental health outcomes thus need to be translated into policies and evidence-based interventions aimed at strengthening social capital. There are, however, major challenges to be faced in order to achieve this, which are linked to the conceptualization of social capital and its operationalization for measurement in research and practice. Research has shown that social capital has seldom been comprehensively conceptualized for adolescents as a group distinct from adults [2]. Social networks for adolescents differ from those of adults [10], and the pitfalls of not recognizing this include underestimating the agency of adolescents, overemphasizing parental influence, and the misinterpretation of how adolescents define important dimensions and factors that could be used for conceptualization [11]. A limitation in the literature is thus the lack of understanding of social capital concerning the target group [12] and how this understanding could be operationalized in a self-reported measurement. Standardized and validated measures of social capital can play a critical role in the work of promoting mental health in adolescents [13].
The association between social capital and adolescent mental health has been investigated using cross-sectional data, mostly with one or two indicators of social capital and various measurements for mental health outcomes [5,[14][15][16]. Researchers and practitioners frequently choose items or sets of items that have not been developed to cover the multidimensionality of the concept and validated to assess self-reported social capital, but merely are linked to either social relationships, ties or networks [17]. Common sources of such items are, for example, the "Add Health Survey", The National Longitudinal Survey of Youths and The Health Behaviour of School-aged Children Survey (where only the 2001/2002 survey explicitly focused on social capital) [18]. It is understandably tempting to apply the concept of social capital to large, national and international data sets, or to draw items from these well-acknowledged surveys. However, concern can be raised about a lack of transparency in the development and validation procedures as well as the inclusiveness of adolescents' own voices during the process of development and psychometric validation, and in the perspectives representing the content of the end product. Social capital is a complex concept that also embodies multiple dimensions and constructs [19]. Bonding, bridging and linking social capital is generally used to distinguish between group contexts, reflecting social ties in homogenous groups, between heterogenous or cross-hierarchical groups [19]. Moreover, the structural dimension of social capital refers to the structure of networks, level of social participation and civic engagement, while cognitive social capital embodies how trust, reciprocity, sense of belonging and support are perceived by individuals [20]. Another distinction that is made frequently is a network or individual perspective, used to emphasize a collective force [20] or the resources that exist between individuals [1]. Using a single-dimension measurement to represent social capital may thus impair the usefulness of research findings. Similarly, researchers also encourage assessment of social capital in multiple relevant contexts such as family, school, peers, and neighbourhood/community when designing studies [21]. The school and peer context naturally overlap, since classmates constitute an important part of peer networks in adolescence. However, the school context is limited to physical location and time constraints and involves adults, while the peer context does not have those constraints [21]. By not considering the breadth of social capital, there is risk of missing out on important information that can contribute to a more comprehensive understanding of the relationship between social capital and adolescent mental health. It also raises the original question pertaining to psychometric validation: Are we measuring what we intend to measure?
The diversity in the conceptualization of social capital has led to a major debate within the field [22]. Some consider it to constitute a weakness, thus challenging the usefulness of the concept and the validity of the research findings. Others optimistically describe the diversity as vibrant [19], where various hypotheses are tested in order to provide a deeper understanding of the pathways between social capital and mental health [17]. What has become evident is that the lack of an agreed definition has led to great methodological heterogeneity in how social capital is operationalized for adolescents [2,5]. In summary, multi-dimensional and well-validated instruments for assessing social capital in adolescent samples should be more consequently used if the aim is to build an evidence base to be used to promote adolescent mental health. While the development and psychometric validation of new instruments for assessing social capital have been called for [17], it is first a necessity to investigate which instruments exist, what they measure and how they have been validated for adolescent samples. There has, to our knowledge, not been an attempt to systematically review and synthesize the evidence of self-report instruments for assessing social capital that have been developed and validated for use on adolescent samples. Similarly, little interest has been given to how adolescents have been involved in the development and validation processes of these instruments. To fill this gap, our current systematic review focuses on the evaluation of self-report instruments for assessing social capital among adolescents. We expect this systematic review to provide a direction for researchers, policymakers and practitioners on psychometrically validated measures of self-reported social capital in adolescent samples.

Design
The design of this study is a systematic review guided by the COSMIN methodology for systematic reviews of Patient Reported Outcome Measures (PROMs) [23]. This methodology provides a thorough description of how to evaluate the measurement properties once instruments are identified, as well as of the clearly defined steps from the initial search to the presentation. Steps 1-4 in the COSMIN manual concern the preparation and the performance of the literature search, as well as the selection of relevant studies, and Steps 5-7 concern the evaluation of the measurement properties of each instrument. A multidisciplinary research team within the disciplines of health science, public health and nursing and with expertise in social capital, mental health and youth studies was assembled to discuss the aim and identify research questions. The aim was formulated by clarifying the construct of interest (social capital), the population (adolescents), the type of instrument (self-report instruments developed and psychometrically validated for adolescent samples) and the measurement properties of interest. This process also included articulating the research questions.

Preparation and
The aim of this systematic review was thus to identify and evaluate the design and psychometric properties of instruments for assessing social capital specifically developed and validated for self-reporting among adolescents (10-19 years). The specific research questions were: (1) What are the dimensions, constructs and contexts of interest within the instruments? (2) In which ways have adolescents been involved in the development and validation process of the instrument? and (3) How have the instruments been validated in terms of the face and content validity; internal structure; reliability and responsiveness?

Criteria for Inclusion and Exclusion
The criteria for eligibility, which were thoroughly discussed between all the authors, were informed by the aim and the research questions. Studies eligible for inclusion were: instruments developed for or adapted to adolescent samples (10-19 years), explicit use of the term social capital in relation to instruments, description of the development and validation process, and explicit focus on adolescents as a group distinct from adults. Reasons for exclusion were: lack of included statement regarding item development or reference to the original source of items or instrument, lack of description of, or references to, a validation procedure that included adolescents, studies based on proxy reporting, such as by parents or others, and review articles.

Performance of the Search
Two librarians with expertise in search methodology were consulted on appropriate databases and search terms prior to performing the search. Based on this consultation and discussions between the authors, search words were tested in initial searches to assure accuracy and breadth. Keywords, titles and abstracts were then searched, accompanied by free text searches, in six electronic databases with a focus on the health and social sciences: PubMed, Scopus, Cumulated Index to Nursing and Allied Health Literature (CINAHL), PsycINFO, Sociological abstracts, and the Web of Science core collection. The final search words were: Adolescents OR Youth, Social Capital, Instrument (multiple synonyms combined with OR), development OR validat*. No time frame was applied, and articles published up until 8 February 2021 were included in the search. No restrictions were applied regarding the language of the publications, although an abstract available in English to enable initial inclusion was deemed necessary (see Appendix A for specifics). The systematic search was conducted in February 2021.

Study Selection
All identified studies from the searches were imported to EndNote. Endnote X9 (Clarivate, London, UK, 2021) was used to facilitate managing of citations and identification of duplicates. A randomized sample of titles and abstracts were reviewed by all the authors to triangulate assessments and cement criteria in the first step of the study selection. Any uncertainties that arose during the process led to discussions between the authors and joint decisions on how to proceed. The first author (MA) then took the lead on conducting the screening process, reviewing all titles and abstracts. If any uncertainties arose during the screening process (Figure 1), a full-text version of the article was retrieved and the methods and results section briefly reviewed to ascertain details of the development and validation procedures, followed by a decision of inclusion to next step or exclusion. The information in some of the articles was either too imprecise to determine the eligibility or there was a statement that more information was available upon request from the authors. Both these issues were addressed by contacting the authors of these articles for any additional information or data that could strengthen the screening process during the selection phase. Any remaining issues on eligibility were discussed between at least two of the authors to assure consensus. The screening process resulted in a total of 54 articles that were subjected to a full-text review. Two authors (MA, MN) separately reviewed all the articles in full text, followed by a discussion between all the authors on whether each article was eligible for inclusion. This procedure resulted in the inclusion of 20 articles that met the eligibility criteria.

Data Extraction
Three data extraction templates, inspired by those proposed in the COSMIN-methodology, were pre-developed to suit the aim and research questions and then used to systematically extract the data from each study. The first template was designed to extract study characteristics such as authors, publication year, journal of publication, country, aim and study design.
The second data extraction focused on answering the research question: what are the dimensions, constructs and contexts of interest within the instruments? Two strategies were applied to identify which dimensions and constructs of social capital that researchers had an explicit interest in measuring when developing their instrument, due to the diverse terminology used to describe social capital. First, any terms specified as dimensions or constructs were extracted from the studies. Second, these terms were interpreted into either a cognitive or structural dimension, as well as separated between bonding, bridging and linking forms of social capital [19]. The reason for this was to enhance the interpretability of the results.
The third data extraction focused on answering the research question: In which ways have adolescents been involved in the development and validation process of the instrument? We were also interested here in understanding the setting and sample that were included in the development and validation process. The template thus included the sample (size, mean age and range and gender), setting, pilot sample, adolescents included in the development and/or in face validity.

Data Extraction
Three data extraction templates, inspired by those proposed in the COSMIN-methodology, were pre-developed to suit the aim and research questions and then used to systematically extract the data from each study. The first template was designed to extract study characteristics such as authors, publication year, journal of publication, country, aim and study design.
The second data extraction focused on answering the research question: what are the dimensions, constructs and contexts of interest within the instruments? Two strategies were applied to identify which dimensions and constructs of social capital that researchers had an explicit interest in measuring when developing their instrument, due to the diverse terminology used to describe social capital. First, any terms specified as dimensions or constructs were extracted from the studies. Second, these terms were interpreted into either a cognitive or structural dimension, as well as separated between bonding, bridging

Evaluation of Measurement Properties of the Instruments (COSMIN Steps 5-7)
Two further data extraction templates were developed and used for the evaluation of measurement properties, inspired by the templates proposed in the COSMIN-methodology. The data extractions pertained to the research question: how have the instruments been validated in terms of face and content validity; internal structure; reliability and responsiveness? Each instrument was evaluated following Steps 5-7 in the COSMIN methodology, where Step 5 concerns content validity, Step 6 the internal structure and Step 7 involves reliability and responsiveness. All relevant information was systematically extracted from the articles into templates four and five following the aforementioned steps. The evaluation of content validation procedures, Step 5, resulted in the extraction of a number of subscales, their labels and total number of items accompanied by type of response options for the final versions of each instrument. Face and content validity evaluation identified whether the development of the instrument was theory explicit, if expert opinion was sought and if the target group was involved in the face validity procedures. The validation of internal structure of the instruments was evaluated in Step 6 involving structural validity, construct validity and concurrent validity or measurement invariance. The reliability and responsiveness in the studies were abstracted in Step 7.
The validity concerning comparison with a golden standard was inapplicable to this evaluation, due to there not being any golden standard for the measurement of social capital. Abstaining from this procedure is an option presented in the COSMIN methodology [23] where inconsistent validation procedures and differing thresholds applied for analyses are described as acceptable reasons.

Citation and Further Validation of the Included Instruments
The article for each instrument was retrieved via Google Scholar as an additional step, and then, by using the function "cited by", we explored whether the developed instruments had been included in any empirical studies where other validation procedures, in other contexts or age groups (still between 10-19 years old), had been conducted and reported. This step is not suggested in the COSMIN methodology [23], but this was carried out in order to determine whether any of the identified instruments had been tested for validity in other samples than in the original article, and if so, briefly present the findings. The Google Scholar search was undertaken on 1 June 2021.

Study Characteristics
Twenty studies were identified that described the development and validation of instruments for assessing social capital in adolescent samples from an initial count of 1956 hits in the databases (see Table 1). The included instruments were validated in twelve different countries: the USA (n = 5), China (n = 2), Iran (n = 2), and one each from Australia, Brazil, Burundi, Colombia, Japan, Lebanon, the Netherlands, Spain and the Turkish Republic of Northern Cyprus. The year of publication ranged from 2005 to 2019. Twelve of the 20 instruments had an original design of items, four were adaptations of existing instruments, one was a mixture of adaptation and original item development and two included items from existing survey questionnaires. It was not made clear in one study [24] whether the items were drawn directly from existing survey questionnaires or if they had been developed by the authors and inspired by existing survey items. Multiple conceptualizations of social capital were found in the included studies; however, after reviewing references and theoretical underpinnings, most references were traced back to social capital as conceptualized either by Bourdieu [1], Coleman [25] or Putnam [20].

Dimensions, Constructs and Contexts of Interest
The dimensions, constructs and contexts of the instruments are described in Table 2. The cognitive dimension of social capital was represented in all the included instruments, mostly through the constructs of trust and sense of belonging, but also by social cohesion [24,26,29,33,38,40,41], reciprocity [28,34,35,43], support [24,30,35,44] and others. The structural dimension was represented by composition or structure of networks [29,33,36,42], interaction, participation or engagement [29,30,33,35,41,43] or similar. Furthermore, bonding social capital could be attributed to all instruments, while bridging social capital was represented in ten instruments [27,28,30,35,36,39,42] and linking social capital was only represented in one instrument [36] where the diversity of friendships could be interpreted as such. Regarding the contexts of interest in the instruments, the family (n = 9), school (n = 8), peer (n = 10) and neighborhood or community (n = 11) contexts were represented. The extended family or extended networks could be attributed to four instruments, while the online context was only found explicitly in one. No instrument represented social capital in all four occurring contexts, namely the family, school, peer and neighborhood or community. Incorporating social capital in at least two of these contexts was most common; however, single contexts were also presented [24,26,33,34,44].

Adolescents Involvement in the Development and Validation
There was a diversity of characteristics within some samples, but mainly between studies, since the included studies had differing aims and were conducted in different settings and countries. Rural and urban samples of adolescents from different ethnic and socioeconomic backgrounds were represented. Sample sizes ranged between 39 and 6853 with a median of 718 adolescents. The age ranges of the adolescents included in the studies varied but could be distinguished between early (10-13 years), mid (14-16 years) and late adolescence (17-19 years) (see Table 3). Five studies covered the entire span of adolescent ages in their samples [33,35,36,38,39]. The sample was limited to late adolescence in three studies [26,32,44]. Six studies had a sample covering mid-late adolescence [24,30,31,[41][42][43] while only one study had a strict sample of mid adolescents [29]. The study by Hall [34] had a sample ranging from 6 to 16 years old, thus covering early to mid-adolescence. Lastly, four studies included only early adolescents in their validation study [27,28,37,40]. A large majority of the studies administered questionnaires for validation in a school setting. The exceptions distributed questionnaires in settings such as the household [24,35], at summer camps [36], both publicly and in school [39] and in two studies the setting was not specified [30,37]. Some studies described the setting and context further and specified, for example, art programme students [28], specific interest in sexual health [30], impoverished neighbourhoods [24], trauma exposure [34] and rural versus urban adolescents [29].  Adolescents were involved in the development phase of the instruments in only four studies. The involvement in three of them was described as target group interviews that were taken into consideration when items were decided upon [32,34,37]. The development of the questionnaire in the study by Onyx [39] was divided into two parts. Half of the items were drawn and adapted from an existing instrument and the other half was developed with help from adolescents. Involvement was described as active participation where a small group of adolescents had the leading role in developing questionnaire items. Focus groups interviews with the target group were conducted beforehand in the study by Harpham and colleagues [24], but there was no mention of linking this to the instrument development. Adolescent involvement for face validity purposes was more prevalent and described in 13 of the 20 studies. Face validity was mainly performed by having adolescents going through the questionnaire, pointing out any concerns followed by a discussion of the instrument in general. All face validity procedures were described as conducted with adolescents who were representative of the target group (Table 3).

Content Validity
The instrument characteristics and content validity are presented in Table 4. The Cosmin Methodology describes that one of the key components of content validity is a clear conceptual description of the construct of interest that guides the development of the instrument. The evaluation of this aspect therefore relied upon description of, and references to, the theoretical underpinnings guiding development (here referred to as theory explicit). Four instruments met the expectations that the development of the instrument was theory explicit, expert opinion was sought and the target group was involved in face validity procedures [29,36,40,42]. Face and content validation procedures were absent in four instruments [24,26,27,35]. The remaining 12 instruments were validated using either one or two of these procedures (Table 4). Following face and content validation, most studies described how the phrasing and wording of the items were adjusted.   Seven instruments were not validated at all through structural validity analyses (see Table 5). Different forms of factor analyses were the primary method of choice among those which were. Principal component analysis (PCA) was used to establish and reduce dimensions for two instruments [26,29]. Exploratory factor analysis (EFA) was applied in four instruments with originally developed items and was followed by confirmatory factor analysis (CFA) [30,32,33,40]. Three instruments were structurally validated using only EFA [27,31,43], two by only CFA [38,42] and one by the use of hierarchical factor analysis [39]. Studies where factor analyses were performed primarily reported factor loadings, eigenvalues and total variance explained. Item factor loadings greater than 0.5 were reported to be desirable, but there were exceptions of lower thresholds, as low as 0.27 [30]. Factors with eigenvalues above 1.0 were consistent for instruments where values were reported. Variance explained was reported for seven instruments, with 32.8% [31] to 74.0% of variance explained by the cognitive subscale in the instrument [29]. Additional results included reporting model fit indices such as the comparative fit index (CFI) and root-mean-square error of approximation (RMSEA). Reported CFI values for factor analyses were all above 0.85. Values of RMSEA below 0.08 (indicating good fit) were reported for five instruments [30,32,33,38,42], while two instruments had values of 0.08 or above (mediocre fit) [39,40].
The evaluation of construct validity by comparing hypotheses of relationship to results confirmed positive or inverse relationships with dependent variables in line with hypotheses (p-values < 0.01) in seven studies (Table 5). Six studies reported inconsistent results or partly confirmed hypotheses of relationship with dependent variables. A test of normality (>5% in sample), which was interpreted as an indicator of construct validity, was used in one study to assess item content relevance, which resulted in retention of all 18 items [30]. A comparison of factors with those of a previous study from which the instrument was partly adapted was performed in another study [39]. Their findings indicated both recurring and unique factors. Factors derived from a factor analysis were compared to a theoretical framework in one study and consistency was found, although caution concerning the results was recommended due to a relatively low sample size [42]. Five instruments were validated for concurrent validity or measurement invariance with varying results (Table 5).

Reliability and Responsiveness
Analyses of internal consistency and stability were used in the studies to assess reliability and responsiveness for the included social capital instruments. Cronbach's alpha was the main indicator of internal consistency and was used in 18 of the 20 instruments. However, the way of reporting varied between the value of the total scale and each subscale respectively. Four studies reported the value for the total scale, of which only one had a value below 0.70 [35]. Among instruments with Cronbach's alpha values reported for each subscale, the lowest values were found in the study by Almgren [26], ranging 0.28-0.86, and the highest in Takakura and colleagues' [43] study, with values of 0.92 and 0.94 for the two subscales. Five instruments were tested for stability via test-retest procedures. The instrument developed by Carrillo-Álvarez and colleagues [29] showed an intra-class coefficient of 0.86, while Kappa-coefficients ranging 0.64-0.97 were reported in the study by Paiva et al. [40] for the four subscales. Two studies reported stability via correlation coefficient, ranging 0.48-0.81 for subscales in one study [24,43] and 67% of the 37 items above 0.70 in the other [23]. Krasny and colleagues [36] compared mean scores with a control group and found no significant differences (see Table 5).

Citation and Further Validation of the Included Instruments
In total, the 20 included studies were cited 647 times (range 1-90) in Google Scholar. The articles were scanned for additional validation procedures, but only two of the twenty instruments had been used in other studies that included validation procedures. First, the instrument developed by Onyx and colleagues [39] was translated and validated in a sample of mid to late adolescents in Greece in the study by Koutra et al. [45]. Factor analysis produced a five-factor solution as opposed to the seven-factor solution in the original study, and four items were dropped from the original 34 items. The variance explained for the instrument was about 10 percentage points lower than in the original study and Cronbach's alpha ranged 0.53-0.73 for subscales. The other instrument was the one developed by Paiva and colleagues [40], which was used in three studies, all with the aim of translating and validating the instrument. These studies involved one Chinese version [46], validated for 10-12-year-olds; one Japanese version [47], validated for 10-15-year-olds; and one Persian version [48] validated for mid to late adolescents. More specifics can be found via references.

Discussion
This systematic review identified 20 instruments for assessing social capital that were developed and validated for adolescent samples. The thorough examination and evaluation of these instruments revealed some conceptual drawbacks and incoherence. First, about half of the studies did not provide explicit theoretical underpinnings described to guide the development of the instrument, the dimensions and constructs of interest and item generation. This can only be considered a vital flaw, since researchers within this field are encouraged to specify how they envisage the concept of social capital and build on existing theories to make findings more conceptually sound and empirically useful [12]. Second, the use of the term social cohesion to label constructs of interest and subscales was somewhat surprising given that social cohesion is often described as a broader concept, in which social capital acts as an important ingredient [19,49]. Lastly, some instruments included constructs or subscale labels that are not traditionally associated with social capital, but instead stem from theories on individual-level psychology, e.g., self-concept and self-efficacy. This raises concern about conceptual coherence in an already heterogenous environment of self-reported measures of social capital.
The common denominators for the included instruments were the inclusion of constructs representing the cognitive dimension, as well as bonding social capital. It is reasonable to suggest that trust, sense of belonging, reciprocity and support represent core constructs that should be included in some way when assessing social capital among adolescents based on our findings. There was, however, a great variation in the design of the questionnaires, number of items and subscales and validation procedures. Although a grey zone was encountered in the distinction between what counts as original item development and drawing on or adapting items from existing measures, twelve described procedures were interpreted as inclusive of original item development. This distinction was facilitated by information on whether data gathering was prospective in relation to the development or if items were drawn from retrospective data. Moreover, conducting a pilot study and including adolescents in face validity procedures also indicated original item development. Five instruments were adaptations of existing instruments for investigating social capital, all of which had previously been validated for adult samples. Of these, one instrument [27] did not undergo face validation procedures before the study was performed. If this truly is the case, it can be considered a weakness.
The instruments, with only one exception, targeted social capital in either one or several of the contexts of family, school, peers and neighbourhood/community. The exception [44], focussed on the online context, but since peer support was also the focus of that study, it is possible to attribute the peer context to this instrument as well. No instrument assessed social capital in all four contexts and only one instrument covered both the cognitive and structural dimensions together with bonding, bridging and linking social capital. This supports the conclusions that other researchers have made that it is unlikely for a single instrument to be able to cover the multidimensionality of the concept in adolescent samples [21]. However, the almost non-existence of linking social capital in these instruments may be an indication of its possible irrelevance for adolescents as a group and modest immediate effects on health outcomes, which at least may be true for early and mid-adolescence. Qualitative research has shown that older adolescents lack linking forms of social capital, which in the absence of strong social capital in the family, would help them navigate through structural systems and help them in their transition to adulthood [50].
The involvement of adolescents in the developmental phase of the instruments was infrequent and only seen in four studies. Only one of these incorporated adolescents as active agents in the item construction. This fact highlights the confidence that researchers have in previous research and a dominating view of theory as the main source of knowledge for operationalizing social capital. In order to stay true to what has historically been described as social capital, this approach may perhaps be motivated. However, the social environment and the ways in which adolescents socialize and communicate have undoubtedly changed over the past decades alongside societal advancements in mobility, communications and the economy. Arguably, so has also the relationship with adolescent health and wellbeing [51]. To maintain a possibility of discovering sociocultural specific indicators of social capital or new universal resources, we thus propose at least interviewing adolescents of the target group as part of the development procedure of new instruments. Due to the emphasis on consideration of sociocultural context when measuring social capital [12], involvement of adolescents in face validity procedures provides an important contribution in terms of validation. There is thus cause to question the validity of the instruments where such procedures were not reported. Nonetheless, some of these instruments comprised items drawn from existing surveys, where larger questionnaires covering multiple concepts may have been previously validated. Experience from this systematic review provides the insight that such procedures are more difficult to perform than in development and validation studies of strict instruments for assessing social capital with originally developed items.
A review of the results of all the evaluation steps revealed that the most comprehensive and transparent validation procedures were found for the Family Social Capital Questionnaire [29] and the Social Capital Questionnaire for Adolescent Students [40]. These provided content validity through explicit theory, expert opinion and target group opinion, and analyses for structural validity, internal consistency and stability. Most instruments adequately established face validity, structural and construct validity and internal consistency as described in the COSMIN methodology [23] although with varying levels of transparency. The instruments evaluated in this systematic review covered differing and multiple dimensions of social capital in varying contexts and settings. Not one instrument could be described as flawless, and all instruments were indeed validated for adolescent samples. We thus recommend that researchers and practitioners consider either of these when designing new studies where social capital is to be measured in a sample of adolescents. However, since many of the evaluated instruments were developed within a specific setting and with the intention of investigating the relationship between social capital and a certain outcome, this should be of guidance for such discussions. With reference to specific outcomes, the initial rationale for this systematic review was a search for validated instruments of social capital in order to investigate the relationship between social capital and adolescent mental health. In a hypothesized scenario, choosing an appropriate instrument would involve careful consideration of context, age, measurement properties and validation procedures, but also the original intention underlying instrument development. Six instruments were developed to assess social capital in its relationship with mental health outcomes [24,33,34,37,38,43]. These instruments were validated for different age groups through diverse procedures and would therefore need close examination before determining appropriateness. However, the two instruments with the most comprehensive procedures [29,40] were developed to assess social capital in general and, though with their own limitations, could therefore be regarded as universal in relation to outcomes.
If more researchers and practitioners are made aware of existing validated instruments, the usage of these may increase, which would enhance the comparability and transferability of findings. Even though the instruments dated back to 2005, other studies where any of these instruments had been used and validation procedures had been provided of their own were rare, as revealed by the citation investigation. This review also reveals the need for validated instruments that capture the multidimensionality of the concept, covering all four contexts of the family, school, peers and community. The difficulties of translating research findings into evidence-based policy and effective interventions will likely persist until these have been produced. Given the efforts of reducing mental health problems and promoting adolescent mental health worldwide, strengthening adolescent social capital should be a priority.

Limitations
Not all the steps described in the COSMIN methodology could be performed in this systematic review. The patient-reported outcome measures more often adhere to areas where a golden standard is present and where the outcome measure is more stringently defined, such as in the field of medicine. Nevertheless, we found the methodology to provide a robust framework that was fitting for the aim of our study. As mentioned, there are quite a few studies in which social capital has been assessed in adolescent samples that were not included in this review. Some of them provided an internal consistency analysis for their measure but were not eligible simply because of this. There is, however, a possibility that by adding databases covering other research fields, additional instruments could have been identified. We limited this risk by consulting experts on search strategies. Six of the included instruments were validated in samples that involved participants below the age of 10 or exceeded 19 years. Although individuals outside this range were a minority in each sample, it is necessary to mention since this may affect the results of the validation procedures.
All the authors were involved in the procedure of each step and any uncertainties that arose were discussed until consensus was reached. By having all authors initially review a randomized sample of studies independently followed by discussion, the reliability of the screening procedure was increased. While one author took the lead in performing the screening and evaluation, discussions preceded each step and were held along the way, and any issues were resolved amongst all authors. To further enhance the trustworthiness of this study, we contacted a number of authors in cases where issues were not resolved amongst the authors of this study and received multiple responses, for which we are grateful.

Conclusions
This systematic review identified 20 instruments for assessing social capital that were developed and validated for adolescent samples. The common denominators for the included instruments were the inclusion of constructs representing the cognitive dimension, as well as bonding social capital. Apart from these, there was great variation in the design of the questionnaires, number of items and subscales and validation procedures. There was no instrument that assessed social capital in the contexts of the family, school, peers and neighborhood or community. Adolescents were only involved in the development of four studies, which testifies to a predominantly theory-driven approach in the design of instruments for assessing social capital. Nevertheless, the theoretical underpinnings guiding the development procedures were poorly described for many of the instruments. Two instruments stood out in terms of their transparency and the adequate reporting of the validation procedures, both of which aimed to assess social capital in general and not in relation to a specific outcome.
Six instruments were developed to assess social capital in the relationship with mental health outcomes; however, these instruments were validated for different age groups through diverse procedures and would therefore need closer examination before determining their appropriateness in research and practice. Further validation work with the target group is proposed and the development and validation of new social capital instruments for adolescent samples is encouraged.

(142)
Searches in bold included and search hits presented in the Prisma Flowchart.