This study is an overview of SRs and adheres to the Cochrane Handbook for SRs of Interventions [29] and the PRIOR (Preferred Reporting Items for Overviews of Reviews) statement [36]. The PRIOR checklist is reported in Multimedia Appendix 1 and the Sex and Gender Equity in Research (SAGER) guidelines [37], in Multimedia Appendix 2 (section 2). In addition, the search strategy and reporting follow the PRISMA-S (Preferred Reporting Items for Systematic Reviews and Meta-Analyses Literature Search Extension) [38] to ensure full transparency and reproducibility of search methods (see Multimedia Appendix 3).

The protocol for this overview was prospectively registered in PROSPERO (CRD42023485887) on December 5, 2023 (Multimedia Appendix 2 section 3), and later published in full [39]. The methods adhered to the predefined protocol and incorporated SAGER guidance where applicable. Reporting also followed PRISMA 2020 recommendations for SRs and overviews [40]. The full protocol can be downloaded from the Open Science Framework [41,42].

Neither patients nor the public were involved in designing or conducting this study. Therefore, no ethical approval was required for this overview. The analyzed data were open access.

The eligibility criteria were reported in detail in our protocol [39]. The inclusion criteria for this overview were based on the population, intervention, comparison, outcome, study type (PICOS) framework (Table 1).

Summary of inclusion and exclusion criteria used to identify SRs of digital behavior change interventions for the prevention of STIs and HIV. Criteria were defined following the Cochrane Handbook and the PRIOR guideline. All SRs were operationally defined as secondary research including primary clinical studies with explicit search strategies in ≥2 databases [29,43].

In this overview, a comprehensive search strategy was used, leveraging multiple key sources. Primary searches were conducted on the leading international databases MEDLINE (via PubMed), the Cochrane Database of SRs, Epistemonikos, and PsycINFO. Each database was searched independently using its native platform; no multidatabase platform (eg, EBSCOhost and Ovid) was used. To enhance the scope of our search, we conducted supplementary searches to identify any studies potentially overlooked by the primary search strategy or absent from the indexed databases. These supplementary efforts included (1) meticulous manual reviews of the references cited in the included studies, (2) examination of related SRs that shared at least one study with the reviews included in our study, and (3) additional records identified through various websites (n=7), organizations (n=1), citation searching (n=321), references provided by authors (n=1), and consultations with experts (n=28; Figure 1). Citation searching involved both backward citation screening (reference list checking) and forward citation tracking using Google Scholar. No additional search methods such as automated alerts, web-scraping tools, or application programming interface–based search retrieval were used.

Since this overview synthesizes published SRs rather than primary studies, study and trial registries (eg, ClinicalTrials.gov, International Clinical Trials Registry Platform) were not searched. Each source was rigorously scrutinized, with the date of the last search or consultation carefully documented to ensure the currency and relevance of the findings. Database-specific yields were MEDLINE (via PubMed; n=1113), Cochrane Library (n=148), PsycINFO (n=161), and Epistemonikos (n=1891), as detailed in Multimedia Appendix 2. Detailed methodologies and search strategies are available in Multimedia Appendix 2 (section 4).

The electronic search strategy (sections 5 and 6 in Multimedia Appendix 2) was developed and conducted under the supervision of an experienced librarian. The strategy did not undergo formal PRESS (Peer Review of Electronic Search Strategies) peer review. The search strategy was newly developed for this overview and was not adapted from any previous review. The first author performed the electronic search from the databases’ inception up to November 12, 2025, with no restrictions on publication date, language, or country of origin, following PRISMA-S recommendations to ensure transparent and reproducible reporting of search methods. No methodological search filters (eg, SR filters, randomized trial filters, and human-only filters) were applied beyond the predefined eligibility criteria. An earlier search completed on August 31, 2024, was rerun and updated in accordance with PRISMA-S guidance [38], to capture the most recent SRs before resubmission. Additionally, we manually searched the bibliographies of relevant reviews and the articles initially retrieved. Letters were also sent to authors and experts identified in the included and excluded studies during the screening stage to identify additional eligible studies. The eligibility criteria are listed in Table 1. Full search dates for each database, including the initial search (August 31, 2024) and rerun (November 12, 2025), are reported in Multimedia Appendix 2. Full search strategies for all databases, copied verbatim as executed, are provided in Multimedia Appendix 2.

Following deduplication and a pilot test of the inclusion and exclusion criteria, 2 independent reviewers (GDA and DLA) screened all titles, abstracts, and full-text articles for eligibility without knowledge of each other’s decisions. For the deduplication process, all records retrieved from database searches and supplementary sources were imported into Collaboratron (Epistemonikos). The software’s automated similarity-detection algorithm was used to identify duplicate entries, followed by manual verification by 2 independent reviewers (GDA and DLA) to ensure accuracy. This hybrid deduplication approach, combining automated and manual procedures, aligns with PRISMA-S recommendations for transparent management of search records. Records from electronic and bibliographic searches were stored and full text screening was conducted using Collaboratron by Epistemonikos [44]. Differences between the 2 reviewers (GDA and DLA) were resolved through discussion, and a third reviewer (SSC) was consulted when necessary. The list of studies excluded after the full text review, along with reasons for exclusions, is detailed in section 7 in Multimedia Appendix 2.

We reached out to 28 authors (up to 3 email attempts) to request additional information, particularly on gray literature such as conference presentations and reports (see the list of experts in the Contacting Experts section of section 4 in Multimedia Appendix 2).

The interrater reliability was assessed using Cohen kappa coefficient [45]. Two independent reviewers (GDA and DLA) evaluated the full-texts (section 8 in Multimedia Appendix 2).

We developed a data extraction tool in Microsoft Excel to obtain various study data recommended by Cochrane [46,47]. The data extracted from the SRs selected for the study was tested and calibrated by the team (section 9 in Multimedia Appendix 2). For this purpose, 1 author (GDA) created the spreadsheet and then extracted the data from 1 SR. Subsequently, 2 authors (GDA and DLA) independently extracted data from 3 SRs, and all authors provided feedback on whether the data elements were complete, and the extracted data were unambiguous. Once a consensus was reached through discussion, 1 author (GDA) created a data extraction manual for the spreadsheet, which can be found in Multimedia Appendix 2 (section 10).

Data items included SR characteristics, PICOS criteria, and variables related to DBCI (Textbox 1).

We performed critical appraisals of SRs using AMSTAR-2 [48], as outlined in our published protocol [39]. AMSTAR-2 consists of 16 items that assess the thoroughness of various aspects of a SR, such as the preparation process, literature search, study selection, data extraction, and analysis, as well as potential biases (eg, risk of bias, publication bias, or funding sources). Based on the type and number of weaknesses identified (ie, unmet items), the reviews were assigned a confidence rating: high, moderate, low, or critically low. Two authors (GDA and DLA) independently assessed all SRs using a spreadsheet (Microsoft Excel 2010) and reached consensus through discussion (section 11 in Multimedia Appendix 2).

To ensure the accuracy of the primary study outcome data and avoid overlap, we checked whether the included SRs shared overlapping primary studies. This was done by creating a citation matrix and calculating the overall corrected covered area (CCA) using Graphical Representation of Overlap for Overviews (GROOVE) [49]. The CCA quantifies the degree of overlap between primary studies included across SRs, calculated as the number of repeated primary studies (numerator) divided by the product of the total number of unique primary studies and the number of reviews, minus the total number of unique studies (denominator). The resulting value represents the proportion of shared evidence between reviews. A CCA of 0% to 5% indicates a slight overlap, 6% to 10% a moderate overlap, 11% to 15% a high overlap, and greater than 15% a very high overlap.”

From the 3643 records identified (3313 from electronic databases and 330 from bibliographic sources), 129 full-text articles were assessed for eligibility. Of these, 106 were excluded, leading to the inclusion of 23 SRs in this overview [56-78] (Figure 1). The interrater reliability for full-text screening of the initial 122 studies was robust (κ=0.88). An updated search conducted on November 12 yielded 378 additional records, of which 7 were assessed in full-text and 4 met the inclusion criteria. Additional studies were identified through supplementary methods, including website searches, citation searching, and expert consultations, ensuring a comprehensive review (section 4 in in Multimedia Appendix 2).

The 23 SRs included in the overview were published between 2014 and 2025, and covered 514 primary studies, including 410 randomized controlled trials (RCTs). Of these reviews, 9 focused exclusively on RCTs [56-58,66,69,70,75,77,78], 2 on non-RCTs [72,73], and 12 included both study types [59-65,67,68,71,74,76]. These reviews encompassed 129,481 participants, with individual studies ranging from 2662 to 27,704 participants. A detailed appraisal of methodological quality and overlap across the included SRs is presented at the end of this section to contextualize confidence in the synthesized evidence. The targeted populations included adolescents (n=14), youth (n=17), adults (n=15), men who have sex with men (n=14), LGBTQ+ individuals (n=10), and people diagnosed with sexually transmitted diseases other than HIV (n=10; sections 13, 14a, and 14b in Multimedia Appendix 2).

Most studies were conducted in high-income countries, including the United States, Portugal, and Chile, followed by upper-middle-income countries such as China and South Africa, and lower-middle–income countries such as India and Kenya. Country income levels were classified according to the World Bank Country and Lending Groups (FY2025, Atlas method, USD), which use Gross National Income per capita as the defining criterion. Classifications were verified as of November 10, 2025, based on the latest publicly available dataset [79]. All data are reported, and detailed characteristics of the included SRs are presented in section 13 in in Multimedia Appendix 2. The synthesis of study characteristics is summarized in Table 2.

The SAGER guidelines revealed that most studies did not include sex- or gender-disaggregated data, and significant gender differences were generally not reported. Only one study, Kamitani et al [63], mentioned transgender participants, but without a detailed analysis (section 12 in in Multimedia Appendix 2).

Confidence levels were determined according to the AMSTAR-2 criteria described in the Methods section, based on the number and severity of critical and noncritical weaknesses. Overall, confidence in the results was high in 21.7% (5/23) of the SRs, moderate in 13% (3/23), low in 17.4% (4/23), and critically low in 47.8% (11/23). The most common weaknesses identified were the absence of reported funding sources for primary studies (22/23, 95.7%), the lack of a full list of excluded studies (13/23, 56.5%), and the omission of a review protocol (10/23, 43.5%; section 16 in in Multimedia Appendix 2).

The overlap assessment revealed a minimal overlap in the 321 primary studies, most of which were cited only once in the 23 SRs, with a CCA of 2.70. Comparisons between pairs of SRs indicated that 82.6% (209 out of 253) had a low overlap (<5%), 8.7% (22 out of 253) had a moderate overlap (5% to <10%), 4% (10 out of 253) had a high overlap (10% to <15%), and 4.7% (12 out of 253) had a very high overlap (≥15%).

Given that 4 reviews (Bailey et al [56]; Burns et al [58]; Clarke et al [59]; and Mo et al [78]) provided explicit coding or mapping of BCTs, these were analyzed as a separate methodological subset (Table 3). The overlap assessment indicated consistently slight overlap (<5%) among them, confirming that this BCT-focused evidence base draws on distinct sets of primary studies. These findings are illustrated in sections 17 and 18 in in Multimedia Appendix 2.

This overview synthesized evidence from 23 SRs to evaluate the behavioral determinants, theoretical foundations, and effectiveness of DBCIs for preventing STIs and HIV. Consistent with the study aims, the findings indicate that DBCIs, particularly SMS reminders, mobile apps, and interactive web-based programs, can enhance engagement with sexual health services and increase STI and HIV testing. Cognitive determinants, such as HIV-related knowledge, motivation, and self-efficacy, also improved across numerous interventions. The most frequently identified and effective BCTs included goal setting, feedback on behavior, and prompts and cues. However, the high heterogeneity of intervention formats and the inconsistent reporting of BCTs across reviews limit the ability to draw definitive conclusions regarding the independent contribution of specific components. Taken together, these findings offer a clear synthesis of behavioral mechanisms across digital interventions and strengthen confidence in the patterns observed across reviews.

Only 4 reviews, Bailey et al [56], Burns et al [58], Clarke et al [59], and Mo et al [78], explicitly coded or mapped BCTs using standardized frameworks (BCTTv1 or TDF), allowing for a focused subset analysis. For the remaining reviews, behavioral mechanisms could only be inferred from narrative descriptions. The GROOVE overlap analysis confirmed minimal overlap across the 321 primary studies, indicating that the synthesized evidence draws from distinct and independent datasets. This methodological combination strengthens the validity of the synthesized findings and responds directly to recent calls for more mechanism-oriented evidence synthesis in digital health [80].

Overall, DBCIs show considerable promise in supporting STI and HIV prevention, particularly through timely reminders, personalized feedback, and interactive educational content. SMS-based interventions consistently improved clinic attendance and STI retesting, reinforcing the well-established role of prompts and cues in facilitating preventive behaviors. Web-based programs and mobile apps contributed to enhanced knowledge, motivation, and behavioral skills, aligning with established behavioral models emphasizing the importance of cognitive determinants. This overview highlights that cognitive determinants represent the most consistent pathways of change in DBCIs, clarifying how digital components influence prevention behaviors.

Behavioral outcomes such as condom use demonstrated mixed effects: Bailey et al [56] reported significant improvements, whereas Burns et al [58] found no notable impact. These differences likely reflect variations in populations, intervention content, MoDs, and follow-up durations, as well as the extent to which interventions incorporated explicit BCTs or theoretical frameworks. Findings from Mo et al [78] further highlight the inconsistency in how interventions implement and report behavior change strategies, particularly among adolescent populations. These discrepancies underscore that digital modalities may be more effective when they target motivational and self-regulatory determinants rather than complex interpersonal behaviors, a distinction that has been underexplored in prior reviews.

The limited and inconsistent application of behavioral theory across reviews is a central issue. Nearly half of the reviews did not reference any behavioral framework, despite strong evidence from broader health literature showing that theoretically grounded digital interventions are more effective and more interpretable. This gap constrains the field’s ability to identify mechanisms of action and optimize intervention design. By systematically examining these gaps, our overview provides conceptual clarity on how limited theoretical integration constrains interpretability, scalability, and optimization of digital interventions. Addressing these gaps is essential for moving the field toward theory-driven digital prevention, where mechanisms of action are explicitly linked to intervention components and outcomes.

The updated search added 4 recent SRs [75-78], which expand the evidence base with up-to-date findings on digital PrEP adherence support, smartphone-based HIV prevention tools, and BCT-coded adolescent interventions. However, their conclusions mirror earlier patterns: digital tools consistently improve testing and cognitive outcomes, whereas sustained behavioral and biological impacts remain inconsistent. By incorporating the most recent evidence available, including 4 SRs published in 2024-2025, this overview offers a timely and comprehensive picture of current digital prevention strategies and how they engage (or fail to engage) key behavioral mechanisms.

Our findings align with prior work demonstrating that digital interventions can effectively promote preventive health behaviors when grounded in behavioral theory and equipped with active engagement strategies. Studies by Simoni et al [23], Albarracín et al [34], and Thomas Craig et al [24] underscore the importance of behavioral mechanisms, self-regulation processes, and contextual tailoring, elements only partially reflected across the reviews included in this overview. Unlike previous syntheses, this overview integrates behavioral mechanisms, intervention content, and methodological quality to provide a more coherent understanding of how and why DBCIs work. This integration allows for a more granular understanding of intervention pathways than outcome-only syntheses can provide.

A key divergence from other health domains (eg, mental health and chronic disease management) is the limited integration of emerging digital technologies, such as AI-driven personalization, conversational agents, and virtual or augmented reality, in STI and HIV prevention. Similarly, sex- and gender-disaggregated analyses remain rare, with only 1 review explicitly addressing transgender populations. This raises concerns about equity and generalizability in digital sexual health research. This highlights a broader limitation in the STI and HIV prevention literature: the field has been slower than other digital health domains to adopt advanced technologies and rigorous behavioral frameworks, reducing its capacity to generate equitable and generalizable impact. Future digital prevention efforts must bridge this technological and conceptual gap to ensure that innovation translates into equitable public health impact. Extending beyond earlier reviews, this overview examines not only whether digital interventions work but also how and why they work, through explicit mapping of behavioral mechanisms across reviews.

This overview offers the most comprehensive behavioral synthesis to date of digital interventions for STI and HIV prevention, following the PRISMA-S (Multimedia Appendix 3). The use of AMSTAR-2 enabled robust appraisal of methodological quality, while GROOVE allowed quantification of overlap across primary studies. Together, the minimal overlap and explicit behavioral coding enhance confidence in the synthesized evidence.

However, several limitations should be acknowledged. First, conclusions depend on the quality and reporting of the included SRs, nearly half of which were rated critically low by AMSTAR-2. Second, BCT coding was absent or insufficient in most reviews, restricting the depth of mechanistic synthesis. Third, high intervention heterogeneity precluded meta-analytic pooling and limits comparability. Finally, reliance on SRs means that relevant primary studies not captured in those reviews may have been missed.

Despite these limitations, the overview provides a clear, methodologically grounded synthesis of how digital tools contribute to STI and HIV prevention and where future research should focus. This multilayered approach responds to recent calls for more rigorous, mechanism-oriented evidence synthesis capable of informing real-world decision-making and aligns with emerging frameworks such as evidence-based X, which emphasize the integration of mechanisms, context, and methodological rigor in digital health research [80].

DBCIs represent a promising and scalable strategy for strengthening STI and HIV prevention, particularly for improving testing behaviors and key cognitive determinants. Interventions incorporating goal setting, feedback on behavior, and prompts and cues show the most consistent positive effects, whereas outcomes related to condom use and biological measures remain mixed. The inconsistent application and reporting of behavioral theory across reviews limits both interpretability and scalability.

Building on prior work, this overview provides a novel contribution by integrating effectiveness evidence with a systematic mapping of BCTs and theoretical mechanisms, a perspective that has been largely absent from earlier syntheses. By identifying which active components are most consistently associated with beneficial outcomes and highlighting persistent reporting and methodological gaps, this study offers actionable guidance for designing next-generation digital interventions. These findings, which incorporate evidence updated through 2025, have pragmatic real-world implications for researchers, clinicians, and policymakers seeking to develop scalable, culturally responsive, and equity-focused digital prevention programs that meaningfully address the behavioral pathways that drive STI and HIV risk across diverse populations. Collectively, these insights offer a pathway for accelerating the development of digital public health tools that are both evidence-based and behaviorally informed.

This work was funded by the ANID, the Chilean government, the National Scientific and Technological Development Fund (FONDECYT) research project number 1250316. The funders had no role in the study design; data collection, analysis, or interpretation; the drafting of the report; or the decision to submit the paper for publication.

Data sharing is not applicable to this article as no new datasets were generated or analyzed during this study. All data used in this overview were extracted from previously published systematic reviews and are included in the manuscript and its supplementary materials.