Symptom Networks, Developmental Pathways, and Intervention Evidence for Co-occurring Anxiety and Depression in Children and Adolescents: A Scoping Review

Abstract

Anxiety and depression frequently co-occur in children and adolescents, but studies of their joint presentation use different units of analysis. This scoping review examined how those approaches connect and where evidence remains limited. We searched PubMed, Web of Science Core Collection, Scopus, and APA PsycINFO for English-language studies available from 2010 to 11 May 2026. We mapped 205 eligible studies across eight non-exclusive empirical themes and interpreted their distribution from a developmental perspective. Study designs varied with the constructs examined. Among 52 symptom-network studies, 37 analysed cross-sectional data and 15 used longitudinal or trajectory cohort designs. Only 16 of 170 risk-factor studies also evaluated treatment outcomes. One symptom-network study linked network-derived findings to a neurobiological measure. Developmental and geographical coverage was uneven. Three studies were child-focused, and the United States and China accounted for 69.27% of the evidence. The descriptive Global South aggregate accounted for 5.85%. These patterns identify priorities for longitudinal within-person research, developmental stratification, and studies linking symptom models to biological measures and intervention mechanisms. Locally validated measures and research in underrepresented settings are needed to assess whether findings transfer across populations.

Introduction

Anxiety and depression are common and burdensome during childhood and adolescence. Global prevalence estimates are approximately 6.50% for anxiety disorders and 2.60% for depressive disorders (Polanczyk et al., 2015). In the US National Comorbidity Survey–Adolescent Supplement, lifetime prevalence was 31.90% for any anxiety disorder and 11.70% for major depressive disorder or dysthymia (Merikangas et al., 2010). Community data also show that anxiety and depressive disorders emerge and persist across childhood and adolescence (Costello et al., 2003). A health-record study of 1.7 million young people in Southern California documented increases in diagnoses between 2017 and 2021 (Xiang et al., 2024). Beyond diagnostic thresholds, co-occurring anxiety and depressive symptoms are associated with greater functional impairment and self-harm risk and may persist into adulthood (Balázs et al., 2013; Cummings et al., 2014; Garber & Weersing, 2010).

Several frameworks explain why anxiety and depression co-occur. Brady and Kendall (1992) distinguished sequential from concurrent comorbidity, while Clark and Watson (1991) emphasised shared negative affect. Cummings et al. (2014) proposed three pathways involving anxiety followed by depression, a shared diathesis associated with both disorders, and depression followed by anxiety. The Hierarchical Taxonomy of Psychopathology (HiTOP) and general psychopathology models place these conditions within broader patterns of symptom covariance (Caspi et al., 2014; Conway et al., 2019; Lahey et al., 2017). These frameworks address developmental ordering, shared vulnerability, and the structure of psychopathology at different levels. Their interpretation depends on the designs and measures available to connect these levels.

Across the period covered by this review, several research traditions proceeded largely in parallel. Epidemiological studies estimated prevalence (Ghandour et al., 2019) and population burden (Klaufus et al., 2022), while longitudinal studies traced joint symptom trajectories (Lallukka et al., 2019). Latent class and trajectory studies characterised developmental heterogeneity (Dai et al., 2024; de la Torre-Luque et al., 2020). Symptom-network models provided the conceptual and methodological basis for identifying bridge symptoms (Borsboom, 2017; Robinaugh et al., 2020), with recent empirical applications in adolescent samples (He et al., 2023; Konac et al., 2021; Tao et al., 2024). The treatment literature evaluated interventions in comorbid samples and examined whether comorbidity modified outcomes (Garber & Weersing, 2010; Milgram et al., 2025; Schniering & Rapee, 2020).

These traditions produce different outputs, including diagnostic groups, directional predictions, bridge symptoms, and latent subgroups. Diagnostic interviews identify comorbid disorders, continuous scales quantify correlated symptom burdens, and networks estimate relations among selected nodes. Samples range from schools and clinics to national cohorts and groups exposed to adversity. Differences in constructs and sampling therefore affect both the questions each study can answer and the comparability of its findings. Evidence mapping can identify where these approaches converge and where direct comparisons remain difficult.

Existing reviews have addressed complementary parts of this literature. Earlier narrative reviews integrated clinical co-occurrence, mechanisms, and treatment (Brady & Kendall, 1992; Garber & Weersing, 2010). Cummings et al. (2014) organised developmental and clinical evidence around a multiple-pathways model. Subsequent network, trajectory, and intervention research raises further questions about how these approaches connect. Wolpert et al. (2019) mapped interventions delivered without mental health professionals and separately evaluated their effectiveness. The present review addresses a broader question about the distribution and overlap of research approaches. Scoping review methods are suited to mapping heterogeneous evidence and clarifying its conceptual boundaries (Arksey & O'Malley, 2005; Levac et al., 2010). We used this design to examine evidence concentration and connections across empirical themes (Peters et al., 2020; Tricco et al., 2018). A comparison with related reviews is provided in Online Resource S6.

Terminological Conventions

We use co-occurrence as an umbrella term for anxiety and depression appearing within the same individual. Diagnostic comorbidity refers to both disorders meeting diagnostic thresholds, whereas symptom co-occurrence refers to their concurrent symptom burdens. Network connectivity describes estimated relations among selected nodes, including connections between anxiety and depression domains. Latent classes describe subgroups, and trajectories describe patterns over time. These operationalisations address different aspects of joint presentation.

Conceptual Framework

The multiple-pathways model of Cummings et al. (2014) informed the developmental interpretation of the evidence map. Its emphasis on temporal ordering complements the structural perspectives offered by HiTOP and the p factor and the symptom-level relations represented by network theory. We considered these perspectives when interpreting connections between empirical themes. Studies were not assigned directly to the three developmental pathways because theme membership alone does not establish developmental sequence.

For presentation, we grouped the eight empirical themes into four descriptive domains. Burden and correlates comprised prevalence or co-occurrence and risk or protective factors. Developmental patterns comprised longitudinal prediction and latent profiles or trajectories. Symptom and biological models comprised networks and neurobiological markers. Clinical evidence comprised severity or impairment and intervention outcomes. These domains organise the map for this review; they are distinct from the pathways proposed by Cummings et al. (2014). Theme definitions and boundary rules are provided in Online Resource S2.

The Present Review

The review combines descriptive mapping with developmental interpretation. We examine definitions, study characteristics, theme distributions, and overlap to identify which research approaches are connected and which remain separate. Four research questions guide the review:

• How do existing studies define and measure the co-occurrence of anxiety and depression in children and adolescents?

• What distributional characteristics do the included studies show by country or region, sample type, and study design?

• How is evidence distributed across the eight empirical themes, and how are those themes connected?

• What important evidence gaps remain in the field, and which directions should future research prioritise?

Methods

Overview

This review followed the Joanna Briggs Institute (JBI) framework for scoping reviews and PRISMA-ScR reporting guidance (Peters et al., 2020; Tricco et al., 2018). The review protocol specified the research questions, eligibility criteria, and charting fields and was deposited on the Open Science Framework on 11 May 2026 (registration z6qjb). An anonymised registration link is available at https://osf.io/z6qjb/overview?view_only=d7724ae886c24fafaa57aa4631d600e9. The PRISMA-ScR checklist and charting template are provided in Online Resources S3 and S4. This review used published literature and required no institutional review board approval, informed consent, or assent.

Design

Eligibility was structured using the Population, Concept, and Context framework. The population comprised children and adolescents, primarily aged 6–18 years. The concept was the joint occurrence or relationship of anxiety and depression. The context included any country, region, or service setting. The scoping design addressed the distribution of evidence across these dimensions.

Eligibility Criteria

Eligible reports described original empirical research involving children or adolescents aged 6–18 years. Broader age ranges were eligible when youth data could be extracted separately or when most participants fell within the target range. Anxiety and depression both had to be central variables. Eligible operationalisations included diagnostic comorbidity, symptom co-occurrence, network connectivity, latent classes or trajectories, directional prediction, shared risk or protective factors, and treatment outcomes. We excluded adult-only and university-student samples without separable youth data, case reports, protocols, reviews, meta-analyses, errata, scale-validation-only studies, and animal studies. Studies were also excluded when either anxiety or depression was incidental or subsumed under generic distress. Dissertations, theses, conference abstracts, non-peer-reviewed preprints, reports, book chapters, editorials, and commentaries were excluded because grey literature was not systematically searched. Complete peer-reviewed articles published online ahead of issue assignment were eligible when sufficient information was available for screening and charting. One relevant theoretical review was used only as contextual literature and was excluded from empirical counts.

Search Strategy

PubMed, Web of Science Core Collection, and Scopus were searched on 11 May 2026, which served as the common evidence cutoff. The protocol-specified APA PsycINFO search was completed via EBSCO on 21 August 2026, with an end date of 11 May 2026. All four databases covered the same evidence window, from 1 January 2010 to 11 May 2026, and applied the same English-language limit. Anxiety and depression terms were required in the title so that both conditions were central to each record. Publisher checks identified and removed three records first available after the evidence cutoff. Full database-specific strategies, fields, limits, run dates, hit counts, and reconciliation procedures are reported in Online Resource S5.

Study Selection

Records from all four databases were deduplicated before title and abstract screening and full-text assessment. The first author applied the same prespecified eligibility codebook at both stages. PubMed, Web of Science, and Scopus decisions were managed in Rayyan; only merged decisions were retained, so stage-specific agreement statistics could not be reconstructed. APA PsycINFO records were assessed using the same codebook. The first author rechecked all uncertain decisions and all provisionally included reports against the relevant source records or full texts before final classification.

Data Charting

The standardised charting form recorded bibliographic identity, country or region, sample source, age, sample size, design, instruments, operationalisation, analytic method, and key findings (Online Resource S4). Race and ethnicity were extracted using the categories, denominators, and ascertainment methods reported in each article. Extracted fields were checked against source records or full texts, with discrepancies resolved by re-examining the relevant material. Study-level data and source notes are provided in Online Resource S8. AI-assisted tools (Codex, OpenAI) supported source-text extraction and consistency checks. Classifications were based on the source articles; responsibility for the final coding and interpretation rests with the authors.

Primary study design and sample source were assigned using mutually exclusive categories (Table 1; Online Resource S4). The design codebook included an other or unclear category, although all included studies fell within the six substantive design categories. The sample source followed a primary-sample-first rule. Left-behind children or adolescents were defined as minors remaining in their home community while one or both parents migrated for work.

Table 1.

Summary characteristics of included studies (n = 205)

Characteristic	Category	n	% of included studies
Study design	Longitudinal/trajectory cohort	83	40.49
	Cross-sectional questionnaire survey	58	28.29
	Cross-sectional network-analysis study	36	17.56
	Randomised controlled or treatment trial	12	5.85
	Clinical comparison or case–control	10	4.88
	Treatment-outcome study or open trial	6	2.93
Sample source	School or community sample	123	60.00
	Clinical or help-seeking sample	44	21.46
	Population or national cohort	18	8.78
	Other or special sample	9	4.39
	Trauma- or adversity-exposed sample	8	3.90
	Left-behind children or adolescents	3	1.46
Country/region	United States	74	36.10
	China (including Macao SAR)	68	33.17
	Western Europe	33	16.10
	Global South (descriptive aggregate)	12	5.85
	Canada	7	3.41
	Australia and New Zealand	7	3.41
	Other East/Southeast Asia	2	0.98
	Not extracted or unclear	2	0.98
Developmental age band	Child-focused	3	1.46
	Adolescent-focused	78	38.05
	Mixed child-adolescent	78	38.05
	Broader age range or follow-up beyond 18 years	44	21.46
	Insufficient age-range information	2	0.98
Primary co-occurrence definition	Diagnostic comorbidity	32	15.61
	Symptom network or bridge-symptom definition	52	25.37
	Joint latent class, profile or trajectory definition	20	9.76
	Other operationalisations	76	37.07
	Concurrent scale cutoffs or symptom thresholds	23	11.22
	Treatment-response grouping	2	0.98

Developmental age bands described the primary analytic sample and relevant follow-up. Exact ranges were used when reported; source-defined developmental groups were retained with their reporting limitations (Online Resource S4). Primary operationalisation described the focal anxiety-depression construct, whereas non-exclusive themes described research approaches. Network nodes could be individual symptoms or symptom domains, but not solely whole-scale totals. Separate condition-specific trajectories could contribute to the latent theme without constituting a joint latent operationalisation.

Studies could receive multiple labels across eight empirical themes. To assess coding consistency, the corresponding author separately reclassified a random sample of 31 of the initial 153 studies (seed 20260609) using the same definitions. Overall percentage agreement across the 248 theme-level coding decisions was 87.50% (217/248). Per-theme Cohen’s κ ranged from 0.25 to 1.00; agreement was lowest for prevalence/co-occurrence (κ = 0.25) and risk/protective factors (κ = 0.37) (Online Resource S4). The first author then rechecked the discrepant and other boundary cases against the source records before the coding was finalised. The same definitions and boundary rules were applied to studies identified uniquely through PsycINFO. Intervention coding was limited to studies that evaluated treatment or prevention outcomes; studies that mentioned only possible intervention implications did not receive this label.

Synthesis Strategy

Synthesis combined descriptive mapping with developmental interpretation. We calculated counts and percentages for publication year, country or region, study design, sample source, operationalisation, and each empirical theme. Percentages used the 205 included studies as the denominator. Theme labels were non-exclusive, so their percentages could sum to more than 100%. Cross-tabulation matrices examined themes by study design, sample source, and country or region. Pairwise overlap was summarised by the shared study count and the Jaccard coefficient. The latter divides the shared count by the number of studies in the union of the two theme sets. These measures describe co-coding within reports and do not establish causal connections. A supplementary theme-by-operationalisation matrix examined how construct definitions related to thematic classification. Four descriptive domains organised the display, and developmental and structural frameworks informed its interpretation.

Methodological quality was not scored or used as an inclusion criterion because the review aimed to map a heterogeneous evidence base (Peters et al., 2020; Tricco et al., 2018). No formal stakeholder, patient, or public involvement consultation was undertaken.

Results

Study Selection

The four database searches identified 1,734 records: 561 from PubMed, 253 from Web of Science Core Collection, 335 from Scopus, and 585 from APA PsycINFO. After removing 787 duplicates, 947 unique records remained. Three records were verified as first available after 11 May 2026 and were removed before substantive screening. Of the 944 records screened, 468 proceeded to full-text assessment; 263 were excluded and 205 met the eligibility criteria (Fig. 1). The annual distribution of included studies is provided in Online Resource S1.

Fig. 1.

Study-selection flow diagram

Search Coverage and Sensitivity Analyses

Sensitivity analyses showed that relaxing the title-field restriction increased PubMed retrieval approximately fivefold, but the broader pool was not fully screened and therefore did not permit estimation of recall (Online Resource S5). Removing publication-type exclusions increased retrieval only slightly, from 561 to 566 records, whereas extending the anxiety and depression terms from title to title/abstract yielded 2,856 records. Five additional network–neurobiology candidates identified only in the broader pool were ineligible; the single eligible network–neurobiology study was already captured by the primary search.

Characteristics of Included Studies

The evidence base was concentrated in a small number of countries, designs, and settings (Table 1; Online Resources S7-S8). The United States contributed 74 studies and China, including Macao SAR, contributed 68; together they accounted for 69.27% of the included evidence. Western Europe contributed 33 studies, whereas the descriptive Global South aggregate comprised 12 (5.85%). Longitudinal or trajectory cohorts (83, 40.49%) and cross-sectional surveys (58, 28.29%) were the most common designs, and samples were mainly school- or community-based (123, 60.00%) or clinical- or help-seeking (44, 21.46%). All 18 records dated 2026 were complete, peer-reviewed articles available by the cutoff; none were preprints.

Race and ethnicity reporting varied in both terminology and detail (Online Resource S8). Articles reported racial categories, ethnic identities, or ethnic origins; some allowed multiple responses. Descriptions ranged from qualitative statements to category-specific counts or percentages, with differing denominators and ascertainment information. These characteristics were retained at the study level.

Operational definitions divided the literature into different analytic objects (Table 2). Symptom-network or bridge definitions were primary in 52 studies (25.37%), diagnostic comorbidity in 32 (15.61%), concurrent scale thresholds in 23 (11.22%), and joint latent class, profile, or trajectory definitions in 20 (9.76%). Two studies (0.98%) used treatment-response groupings. The remaining 76 (37.07%) used other operationalisations, including continuous scores, hybrid diagnostic and scale-based definitions, and condition-specific profiles. Instruments ranged from diagnostic assessments to symptom scales. Studies examining co-occurrence therefore did not necessarily estimate the same construct.

Table 2.

Primary operationalisations and selected measurement instruments (n = 205)

Domain	Operationalisation or instrument	n	% of included studies
Primary operationalisation	Diagnostic comorbidity	32	15.61
	Symptom network or bridge-symptom definition	52	25.37
	Joint latent class, profile or trajectory definition	20	9.76
	Other operationalisations	76	37.07
	Concurrent scale cutoffs or symptom thresholds	23	11.22
	Treatment-response grouping	2	0.98
Selected anxiety measure	GAD-7	49	23.90
	SCARED	21	10.24
	MASC	14	6.83
	RCADS	10	4.88
	ADIS	8	3.90
	SCAS	6	2.93
	DASS (including DASS-21)	6	2.93
	YSR	4	1.95
	K-SADS	3	1.46
Selected depression measure	PHQ-9	40	19.51
	CDI	30	14.63
	CES-D	22	10.73
	BDI	13	6.34
	RCADS	10	4.88
	MFQ/SMFQ	12	5.85
	CBCL	6	2.93
	PHQ-8	5	2.44
	DASS (including DASS-21)	5	2.44
	YSR	4	1.95
	K-SADS	3	1.46

Primary operationalisations are mutually exclusive; theme and selected measure categories are non-exclusive. Other operationalisations include continuous scores, hybrid definitions and condition-specific profiles. Selected measures refer to the youth analytic sample and include short or child versions within the named instrument family

Evidence Concentration Across Themes and Developmental Stages

Evidence was concentrated in risk, prevalence, and longitudinal research, with fewer studies addressing intervention outcomes or neurobiological markers (Fig. 2). Developmental coverage was also uneven, with adolescent and mixed-age samples predominating.

Fig. 2.

Evidence concentration by empirical theme and developmental age band. a, Counts and percentages for eight non-exclusive themes, ordered by frequency. Colours and marker shapes distinguish four descriptive domains used in this review. b, Developmental age bands. Percentages use the 205 included studies as the denominator

Risk or protective factors formed the largest theme (170 studies, 82.93%), followed by prevalence or co-occurrence (111, 54.15%), longitudinal prediction (96, 46.83%), and clinical severity or impairment (82, 40.00%; Fig. 2a). Symptom networks or bridge symptoms appeared in 52 studies (25.37%), latent profiles or trajectories in 27 (13.17%), intervention outcomes in 22 (10.73%), and neurobiological markers in 20 (9.76%). Developmental coverage was uneven: 3 studies (1.46%) were child-focused, 78 (38.05%) adolescent-focused, and 78 (38.05%) mixed child-adolescent. A further 44 (21.46%) covered broader age ranges or follow-up beyond 18 years; 2 (0.98%) lacked sufficient age-range information (Fig. 2b).

Design, Sample, and Geographic Patterning

Study design and sample source were aligned with the questions being asked (Figs. 3a and 3b). Longitudinal and trajectory evidence clustered in cohort designs (e.g., de la Torre-Luque et al., 2020). Among the 52 symptom-network studies, 37 analysed cross-sectional data, including 35 coded primarily as network studies and 2 as questionnaire surveys. The remaining 15 used longitudinal or trajectory cohort designs. Cross-sectional applications included He et al. (2023), Konac et al. (2021), and Tao et al. (2024). Treatment evidence was concentrated in clinical and trial designs (e.g., Schniering & Rapee, 2020). School or community samples supplied most prevalence, risk, prediction, and network research, while clinical samples contributed disproportionately to severity and treatment themes. National cohorts, trauma-exposed groups, and left-behind populations covered fewer themes.

Fig. 3.

Design, sample, and geographic patterning of the evidence base. a, Themes by primary study design. b, Themes by primary sample source. c, Themes by country or region. Circle size and colour encode counts, with exact counts printed inside non-zero cells. Studies may contribute to more than one theme

Geographical and developmental gaps cut across these design patterns (Fig. 3c). US studies more often addressed longitudinal, clinical, and treatment questions, whereas Chinese studies contributed heavily to school or community and symptom-network research. The descriptive Global South aggregate remained small, and the three child-focused studies were too few to determine whether patterns observed predominantly in adolescent or mixed-age samples also characterised childhood.

Cross-Theme Integration

Overlap was greatest among prevalence, risk, and longitudinal themes and smaller for intervention and biological themes (Fig. 4). Only 16 of 170 risk-factor studies (9.41%) also evaluated treatment outcomes. One of the 52 symptom-network studies linked network-derived depressive symptom burden with cortical thickness (Dobson et al., 2021). That cross-sectional analysis treated the biological measure separately from the network nodes. The 27 latent-profile or trajectory studies rarely overlapped with intervention, network, or biological research. Examples combining latent profiles with other approaches included Dai et al. (2024) and Wang and Jia (2026).

Fig. 4.

Pairwise overlap of empirical themes. The 28 theme pairs are ordered by Jaccard coefficient, calculated as the shared count divided by the number of studies in the union of the two sets. Labels report the coefficient and shared study count. The risk-and-treatment pair and the network-and-neurobiology pair are highlighted

Discussion

Principal Findings

The evidence map identifies an uneven relationship between research volume and integration. Prevalence, risk, and developmental prediction were extensively studied, but relatively few reports connected these approaches with biological measures or intervention outcomes. This matters because a correlate, a temporal predictor, and a modifiable mechanism provide different kinds of evidence. Progress depends on studies that test the links between them, alongside further descriptive work in populations that remain underrepresented.

The distribution of study designs helps explain this pattern. Symptom-network studies were concentrated in cross-sectional school or community samples, and intervention evidence was concentrated in clinical and trial settings. Biological studies formed a smaller, methodologically diverse group. These differences may shape apparent contrasts between themes because sampling, measurement, and follow-up determine which relationships can be examined. Repeated use of similar designs can strengthen evidence within a tradition while leaving its connections to other traditions uncertain.

The distinction between temporal sequence and shared vulnerability provides a useful interpretation of these gaps (Cummings et al., 2014). Repeated measurements can test whether anxiety precedes depression or vice versa, whereas concurrent associations alone cannot establish either sequence. Evaluating a shared vulnerability also requires separating its contribution from overlapping symptom content and other common causes. The concentration of studies in a theme therefore identifies available evidence without establishing support for a particular developmental pathway.

These sparse cross-level connections may partly reflect differences in study purpose and feasibility. However, the map cannot determine whether they arise from weak underlying relations, incompatible constructs, or hypotheses that have rarely been tested.

Conceptual Implications

The developmental, structural, and network perspectives address complementary questions. HiTOP and the p factor describe shared variation in psychopathology; the Cummings model addresses developmental progression; network theory represents relations among selected symptoms. Biological and contextual measures may help connect these accounts, but their role must be specified. For example, an exposure proposed as a common cause requires a different test from a biological consequence of persistent symptoms. Clarifying this role makes findings across approaches easier to compare.

This distinction changes how the large risk and longitudinal themes should be interpreted. A shared correlate can reflect common liability, correlated exposure, overlapping measurement content, or consequences of an earlier disorder. Directional prediction narrows these possibilities but does not, by itself, separate general psychopathology, internalising-specific variance, and disorder-specific processes. The mixed anxiety-to-depression, depression-to-anxiety, and bidirectional patterns are consistent with developmental heterogeneity and argue against a universal sequence. Stronger tests require repeated measurements that partition stable between-person differences from within-person change and explicitly report the developmental stage.

Network findings require a similar distinction between association and change. Cross-sectional bridge centrality describes symptom covariance across individuals and cannot establish what happens when a symptom changes within one individual (Robinaugh et al., 2020). Biological measures can enter a model as nodes, predictors, moderators, or mediators, each addressing a different question. Dobson et al. (2021) linked a network-derived symptom cluster to cortical thickness, providing an association for further investigation. A temporal mechanism would require repeated symptom and biological measurements with the direction of the proposed relationship specified in advance. Sensitivity to node selection, instruments, and sampling should also be examined before interpreting bridge symptoms as treatment targets.

Developmental specificity is an immediate test of this layered account. With only three child-focused studies, the network, trajectory, and risk patterns mapped here largely reflect adolescence or mixed-age samples and cannot be assumed to be invariant across childhood. A coherent next-step design would combine repeated symptom measurement with explicit developmental stratification, assign biological or contextual measures a prespecified role, and include an intervention component when modifiability is claimed. Not every study needs all components, but each should state which cross-level link it tests and which remains assumed. Such designs would preserve the distinct meanings of diagnostic comorbidity, symptom co-occurrence, network connectivity, and developmental trajectories while permitting the testing of their developmental interrelations.

Clinical and Preventive Implications

The co-occurrence documented across settings supports attention to both anxiety and depression in assessment, consistent with earlier clinical reviews (Garber & Weersing, 2010). The map also highlights functional impairment and self-harm as relevant clinical outcomes. It does not establish which screening instrument, threshold, or referral pathway performs best. Those decisions require evidence on diagnostic accuracy, treatment benefit, acceptability, and cost that was outside this synthesis.

The limited overlap between risk and intervention research identifies a specific problem for treatment development. Frequently reported correlates may be consequences of symptoms or markers of broader vulnerability, which would limit their usefulness as intervention targets. Transdiagnostic approaches offer one way to investigate shared processes (Garber & Weersing, 2010), but compatibility with a theory is insufficient to establish benefit. Trials can address this by measuring the proposed process repeatedly and testing whether its change precedes improvement in both anxiety and depression.

Because descriptive and intervention evidence came from different sample sources, transferability across settings cannot be assumed. Pragmatic or hybrid studies should test effectiveness, feasibility, and acceptability in the settings where screening or prevention would be implemented.

Under-Represented Populations and Equity

Limited representation affects both the geographical reach of the conclusions and which mechanisms appear central. Samples of left-behind youth, populations exposed to adversity, and national cohorts were concentrated in descriptive, risk, and network themes and were uncommon in intervention research. If exposures, service access, and symptom expression differ across settings, the observed bridge symptoms or developmental patterns may depend on context. The current evidence cannot distinguish contextual modification from differences in measurement and sampling.

Studies of left-behind children or adolescents illustrate the importance of context. Analyses of maltreatment classes and symptom profiles addressed different aspects of heterogeneity (Liu et al., 2026; Wang & Jia, 2026). Their findings can inform hypotheses about family conditions and symptom co-occurrence, but cross-sectional designs cannot establish temporal ordering or intervention response. Comparisons across migration histories, with attention to broader family and community conditions, would help determine which associations are context-specific.

The Global South aggregate is similarly a marker of evidence scarcity rather than a meaningful analytic population. Latin America, Africa, the Middle East, and South Asia differ in languages, health systems, measures, and research infrastructure (Patel et al., 2018). English-language and database restrictions may amplify the observed imbalance. Future studies should report countries separately, justify local constructs and thresholds, and include within-region comparisons where possible. Increasing the number of studies without improving construct validity and local interpretability would not resolve the underlying equity problem.

Measurement Heterogeneity and Cross-Cultural Equivalence

Measurement choices partly determine what a study means by co-occurrence. Diagnostic assessments identify disorders, continuous scales quantify symptom burden, person-centred models identify subgroups, and networks estimate relations among selected nodes. A latent model of environmental exposures answers a different question from a latent model of joint anxiety and depression symptoms. Distinguishing the analytic method from the construct it represents is therefore essential to interpreting differences across studies. Comparisons should specify which features are shared before treating apparently inconsistent findings as evidence of developmental or cultural variation.

Cross-cultural comparison adds another layer. Differences between countries or themes may reflect scientific priorities and exposure patterns, as well as instrument choice, translation, thresholds, node sets, and sampling frames. Because this review did not assess measurement invariance, it cannot attribute the observed distribution to culture or epidemiology. A useful strategy would combine a harmonised core of measures with locally validated items and test invariance before comparing scores (Stevanovic et al., 2017). It should also examine whether network or latent solutions remain stable under alternative operationalisations.

Strengths and Limitations

The review integrates 205 studies from four databases using explicit eligibility and coding rules. It distinguishes operationalisations and uses linked matrices to examine how themes vary with design, sample source, geography, and one another. This approach relates the distribution of evidence to the questions that different designs can address. The study-level table and source notes support examination of individual classifications.

Several limitations constrain interpretation. The search was restricted to English-language records in four databases and required both conditions in the title. It did not cover ERIC, CINAHL, the Cochrane Library, trial registries, dissertations, or systematic grey literature, so zero or sparse cells may be sensitive to coverage. Developmental classification was limited by incomplete age reporting and follow-up beyond age 18. Only 3 studies were child-focused, compared with 78 adolescent-focused and 78 mixed child-adolescent studies. Rayyan retained only merged screening decisions, preventing reconstruction of agreement statistics for individual screening stages. Theme labels were broad and intentionally non-exclusive, particularly for prevalence and risk, and their counts indicate evidence volume rather than quality, effect magnitude, or causal support. No formal assessment of risk of bias, synthesis of effect sizes, or analysis of measurement invariance was conducted. Geographical concentration and uncertain construct equivalence further limit generalisability. Race and ethnicity were reported when articles did so, but heterogeneous categories, ascertainment methods, and denominators precluded pooled estimates of representation. Geographical coverage cannot establish representation of racial or ethnic groups. Future evidence mapping should consider these characteristics alongside socioeconomic and cultural context.

Conclusions

This scoping review maps evidence on co-occurring anxiety and depression across complementary research approaches. The main gaps concern connections between risk and intervention research, symptom networks and biological measures, and evidence specific to childhood. Longitudinal studies with repeated symptom measurements and explicit developmental stratification could clarify these connections. Studies of proposed mechanisms should establish temporal ordering and, where modifiability is claimed, evaluate intervention effects. Locally validated measures and broader population coverage would help determine which findings transfer across settings. These priorities provide a basis for investigating developmental pathways and clinical relevance beyond descriptive overlap.

Authors’ Contributions

Rui Hu contributed to study conception and design, literature screening, data charting, primary coding, data analysis, repeated source verification, and drafting of the manuscript. Xuanyuan Chen contributed to study conception and design, supervision, separate verification of a random 20% theme-coding sample, interpretation of findings, and critical revision of the manuscript. Both authors reviewed and approved the final eligibility, coding, interpretation, and manuscript.

Funding

This research was supported by the Fundamental Research Funds for the Central Universities at Sichuan University (Project No. 2025 Internal Research-Public Administration 10).

Data Availability

The data supporting this review are provided in the article and the labelled worksheets of Online Resources S1–S8. The registered protocol and supporting project materials are available through the anonymous OSF view-only link: https://osf.io/z6qjb/overview?view_only=d7724ae886c24fafaa57aa4631d600e9.

Declarations

Ethical Approval

This study is a secondary scoping review of published literature and did not involve human participants or animals; ethics committee approval was not required.

Consent to Participate

Not applicable.

Consent for Publication

Not applicable.

Competing interests

The authors declare no competing interests.