A Content Analysis of Self-report Child Anxiety Measures

Abstract

A clear understanding of the item content of psychological assessments is critical but often overlooked. This study describes the content overlap of seven commonly used and psychometrically validated measures of anxiety among children and adolescents. Symptom codes were created for all items across measures and items were sorted by these codes, which all fell into specific symptom categories. We conducted two analyses of all items: a “bottom-up” content categorization approach, which used symptom categories that were developed during this study, and a “top-down” DSM-5 categorization which mapped items onto symptoms of anxiety disorders in the DSM-5. Findings reveal a weak mean overlap across the included measures of youth anxiety. This suggests that the scope of anxiety measures should be carefully considered when designing studies, interpreting research, or assessing youth in clinical practice. Further research is needed to develop and establish a coding scheme for a more objective, comprehensive content analysis.

Introduction

Anxiety disorders are among the most common psychiatric disorders in children and adolescents both worldwide and in the United States [1]. More than 32% of adolescents in the United States suffer from anxiety by the time they reach the age of 18 [1]. Anxiety can significantly impact the quality of life of youth as it interferes with their school, cognitive, social, and family functioning [2–5]. Driven by the importance of measurement-based care and valid and reliable assessment in clinical research [6], there has been tremendous effort in the youth mental health field to accurately measure anxiety in children and adolescents. Numerous guidelines have been published to develop evidence-based assessment measures in research and clinical practice [7–9], and there are more than 20 measures with sound psychometrics that are aimed to assess anxiety in youth [7–9]. A meta-analytic review by Seligman et al. [10] examined the utility of commonly used child anxiety measures [i.e., Revised Children’s Manifest Anxiety Scale (RCMAS-2), State-Trait Anxiety Inventory for Children (STAI-C), Youth Self-Report (YSR)] and found that all three measures were effective in differentiating children with diagnosed anxiety disorders from those with no disorder. Furthermore, Etkin et al. [11] examined the quality of eight child anxiety measures based on several properties like validity, reliability, internal consistency, and extent of normative data. Across all criteria, most assessments showed “Good” to “Excellent” psychometric scores [11].

While it is important to have valid and reliable anxiety measures to use, extant measures are generally not uniform and differ from each other in terms of length, purpose and content. Yet, there has been minimal attention to these considerations in selecting evidence-based assessments, both generally and within childhood anxiety. For instance, Patient-Reported Outcomes Measurement Information System—Anxiety Short Form for Children and Adolescents (PROMIS) [12] has 13 items and is intended to quickly screen and track anxiety over treatment, whereas Multidimensional Anxiety Scale for Children (MASC) [13] and Screen for Child Anxiety Related Emotional Disorders (SCARED) [14] have 39 and 41 items, respectively, and are intended to assess anxiety across various subdomains (e.g., school avoidance, separation anxiety, etc.). Furthermore, some anxiety measures aim to capture general anxiety symptoms transdiagnostically, while others aim to cover the anxiety disorders classified in the Diagnostic and Statistical Manual of Mental Disorders, Fifth Edition [15] or through factor analyses. Because of the range of anxiety disorders defined in the DSM-5 [15] (social phobia, separation anxiety, generalized anxiety, specific phobias, etc.) and changes in anxiety presentations across development (i.e., more frequent specific fears among younger children and more frequent generalized anxiety and broad negative affect in adolescents), another possible point of discrepancy among child anxiety measures is which domains of anxiety they assess. Many measures have subscales to account for different anxiety disorders, but assessments vary in terms of how many items they assign to each disorder and which specific subscales they include. For instance, the MASC [13] has four subscales derived from factor analyses (physical symptoms, social anxiety, harm avoidance, and separation/panic), while the Youth Anxiety Measure for DSM-5 (YAM-5) [16] includes questions for each anxiety and related disorders classified by DSM-5. Furthermore, some measures such as SCARED [14] include special emphasis on school avoidance, whereas Revised Fear Survey Schedule for Children (FSSC-R) [17] focuses more on covering various stimuli for specific phobias. This raises the question of whether these childhood anxiety measures assess the same construct(s).

Given the possibilities for discrepancies among anxiety measures, there has been research into both overlap and differences of the content between anxiety measures since the 1990s [18, 19]. Keedwell and Snaith [18] examined over 200 studies to determine the most commonly used anxiety measures at the time and sorted the items from those scales into categories such as mood, cognition, behavioral, overarousal, and somatic. However, the study did not look at the degree of item overlap among different scales and rather examined what each scale measured, thus indicating the need for a more item-level approach to such analysis [18]. Another study examined the validity of the FSSC-R, the RCMAS, and the Modified State-Trait Anxiety Inventory for Children (STAIC-M) in terms of their ability to correctly distinguish between anxious and non-anxious children [19]. While the RCMAS and STAIC-M demonstrated their suitability in discriminating clinically anxious population, the FSSC-R did not [19]. Although this study is valuable in that it confirms the validity of two of the most commonly used measures for child anxiety, it does not investigate the convergent and divergent items between the three scales. Importantly, these reviews were conducted decades ago, and new child anxiety assessments have been developed since then that are now widely used (e.g., The Youth Anxiety Measure for DSM-5; YAM-5 [16], PROMIS—Anxiety [12]).

Most recently, Wall and Lee [20] showed not only that mean overlap among all the examined adult anxiety measures in their study was very low, but that on average only 29% of all possible anxiety symptoms were assessed by any given measure. This indicates that measures should not be considered interchangeable when it comes to assessing adult anxiety symptomatology given their many discrepancies [20]. In addition, Antony and Rowa [21] examined the criteria for evidence-based assessments of adult anxiety and identified areas in which current anxiety measures are lacking, such as overprotective behaviors, distress, functional impairment, quality of life, and the development and disorder course [21]. Antony and Rowa also laid out several reasons for the difficulty of establishing evidence-based assessment procedures for adult anxiety disorders, including the varied purposes of such assessments, the diversity of subjects being assessed (specifically when it comes to ethnicity), and the discrepancies that may occur if one administers several different measures at once to the same person [21]. Though these studies highlight the need for research into what anxiety measures are truly assessing at an item level, they focus exclusively on those designed for the adult population, indicating the need for similar analyses to be conducted on child anxiety measures.

To date, there have been studies examining item overlap between child measures on aggressive behaviors and obsessive-compulsive disorders, but not anxiety disorders [22–24]. In this study, we aim to examine self-report child anxiety measures in order to (1) assess the content overlap of self-report child anxiety measures both qualitatively and quantitatively, (2) identify the constructs that are over- or under-represented in self-report child anxiety measures, (3) examine the extent of DSM-5 representation in self-report child anxiety measures, and (4) provide a clear description of content assessed by different child anxiety measures.

Method

Measure Selection

To analyze the content overlap among self-report measures of nonspecific child anxiety, we followed methods employed in previous similar work [20, 22]. We first searched frequently used self-report measures of anxiety disorders among children and adolescents on Google Scholar. Search terms included “child” AND “anxiety” AND “measure” OR “scale.” Then, the initial list of measures was consulted with five clinicians and researchers with expertise in childhood anxiety. Inclusion criteria were that measures needed to: (a) be available in a self-report format; (b) be designed/adapted for child and adolescent populations; (c) measure the frequency and/or intensity of anxiety at the state and/or trait level; (d) be published and available in English; and (e) be frequently used in the field of psychology since publication date, as measured by number of citations according to Google Scholar; and (f) as rated by team of expert clinicians and researchers. For (e), we followed the method that was adopted by previous work on adult anxiety measures and mental health assessment [20, 25]. For (f), we compiled an initial list of the child anxiety measures and circulated it among clinical psychologists who specialize in pediatric anxiety. Only the measures which received full consensus from the team were considered to meet the criteria (f).

Exclusion criteria included (a) being clinician-rated and not available in a self-report format; (b) having a low citation count since publication date (i.e., each included measure had at least 30 citations); (c) having items that were too specific to one anxiety disorder (e.g., FSSC-R was excluded because it measured many different possible specific fears rather than broader anxiety symptoms); (d) being an adaptation of a measure that is already selected (e.g., RCADS is an adaptation of SCAS, so it was excluded while SCAS was included) and (e) measuring a small subsection of youth rather than a broad range (e.g., the Preschool Anxiety Scale was excluded because it focused on very young children rather than school-age children or adolescents, which was what all other measures focused on).

As a result, seven measures were evaluated: the 41-item Screen for Child Anxiety Related Emotional Disorders (SCARED) [14], the 44-item Spence Children’s Anxiety Scale (SCAS) [26], the 39-item Multidimensional Anxiety Scale for Children (MASC) [13], the 49-item Revised Children’s Manifest Anxiety Scale, Second Edition (RCMAS-2) [27], the 28-item Youth Anxiety Measure for DSM-5 (YAM-5) [16], the 13-item Patient-Reported Outcomes Measurement Information System—Anxiety Short Form for Children and Adolescents (PROMIS Anxiety) [12], and 16-items of the Youth Self-Report—Anxiety subscale (YSR) [28].

Item Selection and Categorization

A total of 230 items were included in the content analysis of the seven measures. No items were excluded from any of the measures. We categorized items using two approaches: (1) Content categorization and (2) DSM-5 categorization. MK and JC first independently coded each item for both categorization approaches. For items with discrepancies, MK, JC and AGG arrived at a consensus through discussion. For complex cases, MK, JC and AGG consulted with EL who previously conducted a similar study [20]. After piloting a coding system in the Content Categorization phase and coding one measure in the DSM Categorization phase, an interrater reliability coefficient of 0.76 was estimated across raters for DSM-5 Categorization, indicating acceptable agreement across raters [29]. This coefficient was generated for individual symptom codes, which were grouped under broader symptom categories (see more details below), the unit of analysis analyzed in this study. This suggests an even higher degree of reliability (i.e., reliability for symptom categories would necessarily be higher).

Content Categorization

Following Visontay et al. [22] and Wall and Lee’s [20] method, we first entered specific items of one measure (i.e., SCARED) into a spreadsheet. Then, we entered items of a different scale (e.g., SCAS) to thematically match the items that were previously entered. After reviewing each item, symptom codes were created to reflect each item. If a subsequent item captured the same symptom as a previously coded item, it was coded in an overlapping symptom, which was also considered for future items. For instance, an item on SCAS “I feel scared if I have to sleep on my own” is similar to the item on SCARED “I worry about sleeping alone”. In this case, both items are about anxiety around sleeping alone. Hence, we created a symptom code of “Sleep alone” and placed both items under this code. Symptom categories comprised of related symptom codes. For example, the item “I worry about sleeping alone” was coded as “sleep alone,” and the item “I am afraid to be alone in the house” was coded as “alone in the house,” but both were placed under the broader “separation anxiety” symptom category. While generating the symptom categories and entering the symptom codes into respective categories, we considered how the items have generally been grouped together in factor analyses and consulted with experts in the field. Furthermore, the items were categorized using the most straightforward and probable interpretation as this was collectively decided by the study team to achieve more streamlined process.

DSM-5 Categorization

Contrary to the previous method, we took a top-down approach. We first pulled all anxiety and related disorder criteria from the DSM-5 as well as OCD (because some measures were created when OCD was categorized as an anxiety disorder) [15]. Then, we created a category based on each criterion of anxiety disorder excluding criteria related to frequency of symptoms (e.g., one hour per day or more in OCD), clinically significant distress/impairment, and rule-out medical diagnoses. For example, criterion A of separation anxiety disorder describes various expressions of anxiety related to separation from attachment figures (e.g., recurrent excessive distress when anticipating or experiencing separation from home or from home major attachment figures, worry about losing major attachment figures or about possible harm to them). Hence, we made each of these descriptions a different symptom code (e.g., distress when anticipating separation, worry about losing attachment figure/harm to attachment figure). Afterwards, we placed each item under appropriate symptom code as in the Content Categorization approach. If an item did not assess any symptoms listed in DSM-5, it was placed under “Other” category. For more detailed procedure, please refer to the DSM-5 Coding manual created by the study authors (see Supplementary Material).

Compound and Idiosyncratic Items

Similar to Visontay et al. [22] and Wall and Lee [20], we encountered some items that could fall under more than one symptom code. In this case, we identified them as “compound” items and placed them under each matching symptom code. For instance, “My hands feel sweaty or cold” could fall under “sweating” and “chills or heat sensations” for the Content Categorization. In this case, we marked this item as “compound” and placed it twice under both symptom codes. During this process, we adopted a “benefit of the doubt” approach and leaned towards marking an item “compound” whenever an item could be categorized into more than one symptom code (i.e., counting a measure as assessing a construct to make conclusions about independence in item assessment across measures as conservative as possible).

Statistical Analysis

Like similar studies, the Jaccard Index was used to estimate the level of content overlap of the scales [30]. The Jaccard Index is calculated as s/(u1 + u2 + s) (s: shared number of items across two questionnaires, u1/u2: items unique to each scale) and its coefficients range from 0, indicating no overlap, to 1, indicating complete overlap [30]. There are no well-established guidelines for describing Jaccard Index coefficients, thus we followed the interpretive conventions of previous studies wherein coefficients of 0–0.19 are considered very weak, 0.20–0.39 weak, 0.40–0.59 moderate, 0.60–0.79 strong, and 0.80–1 very strong [30, 31]. The analyses were conducted using R [32] and the following packages: qgraph [33], ggplot2 [34], data.Table [35], reshape2 [36], psych [37], ade4 [38], and viridis [39]. The code was adapted from code provided by Fried in 2017 [30].

Results

In total, items were classified into 42 symptom categories (see Fig. 1). Only two symptom categories, “worry” and “social fear”, appeared across all seven scales, and 14 symptom categories were unique to a single scale (see Table 1). Moreover, 77% of symptom categories were measured by three or fewer of the seven scales. At least four of the seven scales assessed symptom categories related to social fear, worry, school-related anxiety, separation anxiety fear, indigestion/nausea, specific fear, physical shakiness, breathing difficulties, increased heart rate, nervous, and self-worth.

Fig. 1.

Each color reflects a scale and each number that starts with S denotes a symptom category. A solid circle means the scale contains that symptom category. An empty circle means the scale contains an item that can be placed under more than one category i.e., compound category

Table 1.

Number of symptoms that appear across scales

Number of symptoms	Scales	%
14	1	21.7
9	2	15.0
7	3	18.3
4	4	8.3
3	5	8.3
3	6	10.0
2	7	5.0

The Jaccard Index was 0.28 across all scales (on a range from 0 to 1), indicating a relatively weak mean overlap of scale items. Weak overlap was also generally demonstrated between any two scales with indices ranging from as low as 0.14, between the MASC and the PROMIS Anxiety Child, to a moderate 0.50 between the SCARED Child and the SCAS. Both the MASC and PROMIS Anxiety Child had the lowest mean overlap with all of the scales (0.28) while the SCARED Child had the highest mean overlap (0.40). See Table 2 for a full review of Jaccard Indices.

Table 2.

Jaccard Index between scales

	SCARED child	SCAS	MASC	RCMAS-2	YAM-5	PROMIS anxiety child	YSR
SCARED child	1	0.50	0.39	0.42	0.38	0.27	0.42
SCAS	0.50	1	0.37	0.25	0.48	0.24	0.39
MASC	0.39	0.37	1	0.27	0.23	0.14	0.26
RCMAS-2	0.42	0.25	0.27	1	0.23	0.29	0.38
YAM-5	0.38	0.48	0.23	0.23	1	0.35	0.32
PROMIS anxiety child	0.27	0.24	0.14	0.29	0.35	1	0.41
YSR	0.42	0.39	0.26	0.38	0.32	0.41	1
Mean overlap	0.40	0.37	0.28	0.31	0.35	0.28	0.36

The scales varied significantly at capturing different DSM-5 diagnostic categories. Separation anxiety disorder and social anxiety disorder related symptoms were most frequently and thoroughly measured among the seven scales, followed by panic disorder and generalized anxiety disorder. A single scale (YAM-5) assessed selective mutism. No scale measured diagnostic criteria from all of the DSM-5 anxiety diagnoses and some assessed only a few. For example, the SCARED Child did not capture any diagnostic symptoms related to selective mutism, specific phobia, agoraphobia, and obsessive-compulsive disorder (previously considered an anxiety disorder). See Table 3 for a full breakdown of each scale.

Table 3.

Scale diagnostic characteristics

	SCARED child	SCAS	MASC	RCMAS-2	YAM-5	PROMIS anxiety child	YSR
Symptoms categories captured	20	17	24	18	14	9	15
Unique symptom categories (n)	1	1	6	3	1	0	2
Scale captures X% of all 42 items	47.6	40.5	57.1	42.9	33.3	21.4	35.7
Scale captures X% of 8 core DSM-5 SAD symptoms	87.5	50.0	50.0	0	75.0	25.0	25.0
Scale captures X% of 1 core DSM-5 selective mutism symptoms	0	0	0	0	100	0	0
Scale captures X% of 3 core DSM-5 specific phobia symptoms	0	33.3	66.7	0	33.3	0	33.3
Scale captures X% of 3 core DSM-5 social anx. disorder symptoms	66.7	66.7	100	66.7	66.7	33.3	66.7
Scale captures X% of 16 core DSM-5 panic disorder symptoms	68.8	43.8	56.3	18.8	31.3	6.3	18.8
Scale captures X% of 6 core DSM-5 agoraphobia symptoms	0	50.0	16.7	0	66.7	16.7	0
Scale captures X% of 8 core DSM-5 GAD symptoms	25.0	12.5	37.5	75.0	25.0	25.0	25.0
Scale captures X% of 4 core DSM-5 OCD symptoms	0	75.0	25.0	0	0	0	25.0

A qualitative summary of the content assessed by each measure is provided in Table 4.

Table 4.

Suggested recommendations for clinicians researchers when selecting a self-report youth anxiety measure

Measure	Description
PROMIS anxiety child	Assesses general symptoms of anxiety briefly. May be best suited to measurement-based care when a clinician is less interested in specific symptom clusters related to anxiety but wants to monitor quickly through repeated assessments.
MASC	Assesses common categories and focuses more on avoidance and safety behaviors than other measures. Contains several unique items related to physiological symptoms of anxiety as well.
RCMAS-2	Assesses common categories of anxiety with an emphasis on worrying and social anxiety. Includes categories that measure general positive feelings as well.
YSR	Assesses common categories briefly. As it is a part of a broader measure, it may supplement the general clinical profile of youth but may be less informative in specialty anxiety settings.
SCAS	Assesses a broad range of categories including school-related anxiety and obsessions and compulsions, which are less frequently assessed in other measures.
YAM-5	Assesses categories that are compatible with the current DSM-5 categories.
SCARED child	Assesses a broad range of categories including school-related anxiety and avoidance. Contains two unique categories related to derealization.

Discussion

This study assessed the content overlap of commonly used self-reported measures of child anxiety. Additionally, analyses were done to determine frequently represented constructs and the extent to which DSM-5 diagnostic criteria are represented among these measures. Similar to others [20, 22], our analyses indicated low overlap across scales. While 14 item categories were unique to one out of seven analyzed scales, only two out of 42 item categories were consistently represented across all seven scales (“worry” and “social fear”). This discrepancy suggests that there is a lack of agreement among common, validated self-reported child anxiety measures, and that they are more incongruent than alike.

The minimal overlap among all scales demonstrates the importance of selecting a scale that measures exactly what one hopes to assess rather than assuming that multiple scales that have the same broad goal of assessing anxiety are equal. For instance, because the MASC and RCMAS-2 are around the same number of items, and both are self-reports meant to measure anxiety in children, one clinician may administer the MASC to a child while another clinician may administer the RCMAS-2, and they may assume they are equally measuring “anxiety.” However, our analyses show that the Jaccard Index between these two measures is just 0.27. Thus, the two clinicians would be learning very different things about that child at an item level, and the child may respond differently to both scales. In short, while scores between scales would exhibit a direct relationship, extant scales measure different facets of anxiety which should be interpreted accordingly.

The low Jaccard index among scales may also be explained by the lack of situational elements of anxiety in some scales that may be especially relevant among children and adolescents. Particularly, several scales did not include many items that fell into the “school-related anxiety” category found in this analysis. The MASC and RCMAS-2 included only one item (“I worry about getting called on in class”) that fit into this category, which does not capture many aspects of school anxiety, such as academic performance, separation from parents, or being accepted by peers. Other measures like the YSR and PROMIS Anxiety also only included one item relating to school (“I was afraid of going to school,” “[I] fear going to school”) that focuses on anxiety around going to school, unlike the MASC and RCMAS-2, but does not cover any specific elements of the school experience. Lastly, measures like the SCAS and SCARED Child include physical symptoms of anxiety while at school (e.g., “I get stomachaches at school”) and specific anxiety about school-related issues like tests (e.g., “I feel scared when I have to take a test”). While it is important to include various aspects of school that may cause a child anxiety, given that school is such an important element of the experiences of children and teens, it is clear that there is no clear agreement on the extent to which school experiences should be included in measures of child anxiety specifically. This is yet another example of the importance in considering measures based on the specific assessment question a clinician or researcher may have.

A third potential explanation for the weak mean overlap among all measures could relate to their diverse time frames. For instance, some scales like the SCAS and MASC-2 prompt the child to report on his/her symptoms at the moment of completing the assessment, without any mention of a specific time frame during which to self-evaluate. Conversely, other scales, such as the SCARED Child, ask the child to think retrospectively about his/her feelings within the past three months. The PROMIS Anxiety similarly instructs the child to reflect on a recent time frame, but in its case, it examines the past seven days. Evidently, there is not one universally agreed-upon time frame during which to evaluate the child’s symptoms of anxiety, nor is there a consensus on whether measures should assess symptoms retrospectively or at the time of their completion. Therefore, it could be that measures that frame children’s experiences within specific times contain more specific items, while those that do not have items that are more broad. This could influence the types of responses children give for different measures—perhaps a child may feel like generally he or she feels anxious most of the time, but not very much within the past three months, or even seven days. Thus, it is important for researchers to consider their desired time frame of assessment when administering self-report anxiety measures to children.

The weakest mean overlap between two measures was between the MASC and the PROMIS Anxiety Child (0.14). While part of the reason behind this lies in the brevity of the PROMIS Anxiety Child (13 items) and thus the decreased likelihood of overlap, it is also possible that the cause of the discrepancy is the breadth of symptoms covered by each scale’s items. The PROMIS Anxiety Child includes many items that fall into the broader categories we created, such as “worry,” “fear,” “anticipated fear,” and “nervous,” with items like “I felt worried,” “I felt scared,” and “I felt nervous.” Because the PROMIS Anxiety child is a shorter measure with items that were eliminated through an item response theory approach, it is likely that specific physiological symptoms of anxiety or situational aspects of anxiety fell out during the measure refinement process, as perhaps these components that are commonly assessed by other scales are not as central to the construct of “anxiety.”

On the other hand, the MASC includes the largest number of unique items, including ones such as “When I feel upset or scared, I let someone know right away” (coded as “sharing worries”) and “I keep my eyes open for danger” (coded as “hypervigilance”). Out of the 14 items that were unique to certain scales, six of them were included in the MASC. The MASC has been found to have a four-factor structure that includes physical symptoms, social anxiety, separation anxiety, and harm avoidance [13]. Thus, the discrepancy between the PROMIS Anxiety and MASC can be explained by the difference in emphasis on broader symptomology versus more specific categories, especially given that the PROMIS Anxiety does not include any items relating to physical symptoms or harm avoidance.

The scales also significantly differed based on the DSM-5 categorization. None of the scales captured all individual diagnostic criteria of the anxiety disorders in the DSM-5. Social anxiety disorder was most frequently measured by all of the scales, followed by panic disorder and generalized anxiety disorder. Separation anxiety disorder was also frequently measured by all scales except for RCMAS-2. Selective mutism was only measured by the YAM-5.

This incongruence between symptoms assessed in the scales and classified in the DSM-5 across the scales is anticipated as all of the scales except for the YAM-5 were developed before the DSM-5 [15]. Hence, the YAM-5 was the only scale that assessed all 7 anxiety disorders in the DSM-5 although it did not capture all of the diagnostic criteria within each anxiety disorder and did not include symptoms of OCD, which were included in measures developed during the DSM-IV/ICD-10 era.

Additionally, despite the central role of avoidance in prevailing psychological models of childhood anxiety, and its inclusion in several criteria for DSM-5 anxiety disorders, avoidance was rarely assessed across scales. In fact, avoidance was assessed substantially less frequently than several specific feared stimuli (e.g., social fears, separation fears) as well as physiological components of anxiety (e.g., shaking, stomachaches). The exclusion of avoidance from these scales underscores the need for clinicians and researchers to systematically assess avoidance during other aspects of assessment (e.g., using other scales or during a clinical interview). Future anxiety measure development should consider including avoidance as an essential feature.

To summarize, although all these questionnaires have been psychometrically validated, they may be measuring inherently different facets of anxiety. This suggests that users need to determine which construct they would like to examine before choosing a scale and do not use these measures interchangeably or interpret them interchangeably. To the extent that these measures are used to contribute to a diagnostic assessment in clinical settings, providers should also carefully consider the breadth of each scale when selecting it. If providers intend to use these scales for diagnostic codes [40], this issue becomes more complex and possibly warrants more guidelines on the selection of measures.

There are several study limitations. First, content analysis is inherently subjective and no firm or widely accepted methodological guidelines have been widely established (e.g., Wall and Lee, 2022; Visontay et al., 2019), and thus the validity of the system developed for this study might be questioned [20, 22]. This led to some decisions being made throughout coding that a third rater may disagree with. For instance, both raters in this study agreed that the item “I am afraid to be alone in the house” fit under the broader separation anxiety category, but another rater may see it as a symptom of generalized anxiety. Thus, a level of cautious discretion in interpreting these findings is needed, and it is important to hold in mind the subjective nature of the coding done. Though we used two independent coders who followed the same guidelines when categorizing the items and exhibited good interrater reliability, future research is warranted to establish a coding scheme for a more objective, comprehensive content analysis. Furthermore, while the interrater reliability coefficient for the analysis of individual symptom codes was 0.76, a coefficient for the broader symptom categories used in analyses was not calculated as the coding system was developed during the coding process. There was also no reliability analysis conducted for the content categorization as opposed to the DSM-5 categorization. This reiterates the need for caution when understanding and applying the results of this study.

Another limitation was that this study only focused on child-report measures and excluded parent-report measures. Parent-report measures were not included in this study, because the items in parent-report measures were largely similar to the counterpart items in child-report measures. Thus, it was determined that including parent-report measures would not provide substantial value to our analysis. Moreover, one benefit of excluding parent-report measures was that it allowed us to conduct a more in-depth review of all items included in several child-report measures without attempting to include few idiosyncratic items from parent-report measures. However, including parent-report measures in the analysis may potentially provide a broader overview of the content overlap among all child anxiety measures and is recommended for future research. Given the weak mean overlap among self-report measures of child anxiety that this study found, and the important implications this has on practical use among clinicians, future studies could also involve the objective determination of specific-use guidelines for these different assessments.

Summary

The content overlap among validated measures of child and adolescent anxiety in the field of psychological assessment is understudied. This study used quantitative and qualitative methods to examine the extent of content overlap across commonly used, valid self-report assessment of child anxiety. Results of this study suggest that there is a weak mean overlap among common measures of child and adolescent anxiety. Furthermore, only two symptoms appeared across all seven scales and 14 symptoms were unique to a single scale. 77% of symptom categories were measured by just three or fewer of the seven scales, again indicating the lack of agreement among the seven scales included in analysis. These findings suggest both clinicians and researchers should carefully consider item content when selecting a child anxiety assessment. More research is needed to apply this method to measures of other disorders in youth populations, and to develop a less subjective way of quantifying the difference between measures on an individual item basis.