Playing video games is an integral part of adolescent culture, serving not only as entertainment but also as a platform for skill development and identity exploration [1]. For adolescents with behavioral or developmental needs, gaming provides structured environments to practice competencies that support personal and social growth. Therapeutic games can enhance emotional regulation, problem-solving, and executive function, making them valuable for interventions addressing these challenges [2]. Health care professionals increasingly explore how these skills can be applied in therapeutic contexts, particularly through digital interventions that expand access to psychological support [3]. However, further research is needed to understand how in-game learning effectively transfers to real-world applications and how game design supports this process.

A growing body of research examines the role of serious games in mental health interventions. Systematic reviews and meta-analyses have shown their effectiveness in addressing anxiety [4], depression [5,6], emotional regulation [7], and broader psychological well-being [8,9]. Serious games targeting adolescents show particular promise for fostering engagement and skill development among populations who may resist traditional forms of therapy [10]. Yet, existing reviews primarily evaluate outcomes rather than examining how their design rationale enables learning and behavioral change. Recent work has begun to highlight the need for design-focused evaluation frameworks in serious gaming interventions [11], including the integration of psychological, motivational, and contextual design principles [12]. Understanding these mechanisms is critical for developing scalable, evidence-based interventions that integrate meaningfully into health care practice.

Adolescents with behavioral or developmental challenges often engage deeply with video games, offering both opportunities for therapeutic skill development and challenges for implementation in health care contexts [13]. Many abilities cultivated through gameplay, such as executive functioning, emotional regulation, and social communication, align closely with health care objectives [14]. Serious games, distinct from entertainment games, are intentionally designed with educational or therapeutic intent to foster learning and behavioral change through structured feedback and interaction [15].

For adolescents experiencing anxiety, social stress, or executive dysfunction, game environments provide opportunities for cognitive rehearsal and emotional self-regulation [16]. The structured and controllable nature of gameplay can reduce stress and facilitate gradual exposure, particularly among individuals with autism spectrum disorder [17]. Integrating biofeedback features, such as heart rate monitoring, further enhances awareness of physiological states and supports self-management [3]. Multiplayer and cooperative formats extend the therapeutic potential by promoting communication and collaboration, aligning with broader educational and clinical goals [18].

From a motivational perspective, serious gaming, or the process of applying game-based interaction in structured interventions, benefits from self-determination theory, emphasizing autonomy, competence, and relatedness [19]. Well-designed games can support all 3 dimensions, fostering intrinsic motivation and engagement. Because players differ in what motivates them, adaptable and inclusive designs are essential [20]. Despite growing evidence for their value, the integration of serious games into health care remains limited. Participatory design, involving adolescents, caregivers, and clinicians, can enhance contextual relevance and therapeutic alignment [21]. In this sense, serious games can function as mediating tools that connect education, therapy, and everyday life, establishing conditions for what boundary-crossing theory [22] describes as transitions between distinct domains of practice.

Serious games designed for therapeutic or educational purposes often simulate or abstract real-world challenges to foster learning and behavioral change [23]. Understanding how these games enable the application of skills beyond gameplay remains complex [24,25]. Drawing on model-based reasoning, effective learning occurs when interactive representations balance abstraction with real-world fidelity, enabling players to engage in both conceptual exploration and situated practice [26,27]. Within this context, serious games can act as boundary objects or artifacts that mediate between distinct domains such as health care, education, and daily life [28].

Boundary crossing theory helps explain how serious games facilitate transitions between these domains. Akkerman and Bakker [22] identify 4 mechanisms of boundary crossing: identification, coordination, reflection, and transformation, which enable knowledge movement across contexts. In serious games, these mechanisms appear when players assume new roles, collaborate across disciplines, or reflect on in-game experiences to link them with real-world contexts [29]. Through these processes, players act as boundary crossers, integrating insights, perspectives, and strategies from simulated environments into their personal or professional realities.

Designing for boundary crossing requires balancing structure and flexibility. Overly standardized mechanics can constrain learning, while excessive abstraction risks disconnecting gameplay from therapeutic goals. Scaffolding methods, such as reflective dialogue, adaptive feedback, and structured debriefing, can strengthen the link between play and practice [30,31]. When aligned with health care objectives, serious games can thus serve as both learning tools and vehicles for systemic integration [32].

The ability to apply knowledge and skills from one context to another is central to the educational and therapeutic values of serious games [33,34]. In health care, successful transfer determines whether insights gained during gameplay lead to meaningful behavioral or cognitive changes in real-world settings [35,36]. Two complementary types of transfer are often distinguished: literal transfer, which involves the direct application of learned skills in similar contexts, and figural transfer, which requires abstraction and adaptation to new or unfamiliar situations [33]. While literal transfer supports practice through realism, figural transfer fosters generalizable learning and adaptability [34].

Boundary crossing and transfer are interdependent. Mechanisms such as reflection and coordination provide the cognitive scaffolding that allows experiences within a game to extend beyond its boundaries. Thus, serious games do not merely teach within a virtual system but prepare players to act beyond it.

The model of reality represented in a game further defines how transfer occurs. Each serious game embodies a representational model, causal, relational, procedural, or structural, that shapes how aspects of the real world are simplified or transformed during play. As research by Wenzler [37] notes, these models mediate the relationship between simulated and actual experiences, influencing both fidelity and interpretive depth. Complementing this, the Bogost [38] concept of procedural rhetoric highlights how the logic of rules and mechanics communicates meaning, shaping how players understand complex systems. Together, these frameworks explain how serious games model reality in ways that make transfer possible and meaningful.

For adolescents, such design choices are crucial. Games provide safe environments for exploring identity, practicing regulation, and testing social behaviors [39,40]. By combining representational fidelity with reflective abstraction, serious games can translate in-game learning into durable real-world competencies [41].

This systematic review examines how serious games in health care incorporate the constructs of transfer, boundary crossing, and models of reality within their design rationale and implementation. Unlike prior reviews that focus on outcomes, this study analyzes the conceptual coherence of design approaches, identifying how theoretical frameworks inform game development and affect the connection between simulated and real-world practice.

The study aims to clarify how design choices facilitate learning transfer, enable boundary-crossing conditions across educational and clinical domains, and represent reality through varying degrees of abstraction and fidelity.

We proposed the following research questions for this study:

Two reviewers independently screened the 2296 titles and abstracts using Rayyan (Qatar Computing Research Institute). Studies were excluded if they focused on simulation for data modeling, used games only as measurement tools (eg, eye-tracking or motor-tracking), and lacked peer review or methodological detail. Studies were included when they addressed a health care–related challenge affecting adolescents’ social, emotional, cognitive, or developmental functioning and used a serious game as an active learning or therapeutic tool. Because many included populations do not share a single diagnostic label, both health care and the target audience were interpreted more broadly. Disagreements were resolved through a structured consensus process and by including a third reviewer.

Following this stage, 47 papers were reviewed in full, resulting in 33 included studies. An overview of the key characteristics of the 33 included studies is provided in Table 1 (see Table 2 for the coding table). The excluded papers comprised 9 conference abstracts without accompanying full papers, 2 unavailable manuscripts, 2 available only in French, and 2 publications about the same artifact and context. A PRISMA 2020 flow diagram (Figure 1) summarizes the study selection process from 3997 initial records to 33 included studies.

Inter-rater reliability for the initial round of coding was κ=0.55, which reflects moderate agreement (κ=0.41‐0.60) [76,77]. Moderate agreement is acceptable in qualitative content analyses that require inferential judgment, particularly when coding implicit or theory-driven categories, such as transfer types, boundary-crossing mechanisms, and model typologies. These constructs were often not explicitly described, making some variability in coder interpretation expected.

Many games prioritized figural transfer (n=24), using metaphorical or symbolic representations to promote generalizable cognitive and emotional learning. Only 10 studies utilized literal transfer, replicating real-life tasks or environments. Examples include virtual reality supermarket training [53], which simulates grocery shopping to teach money management skills; Braingame Brian [64], which trains executive functions through structured gameplay; figural approaches, such as DEEP [71], which teaches emotional regulation through metaphorical underwater immersion; and iSocial [58], which develops social competence using virtual role-play, illustrating a focus on reflective learning rather than direct behavioral replication.

Despite these efforts, few studies incorporated explicit scaffolding or debriefing mechanisms to reinforce transfer, and longitudinal evaluations were rare. While short-term improvements in executive function and social skills were frequently measured, few studies investigated whether these abilities persisted or generalized beyond the intervention [72]. The overall dominance of figural transfer suggests that most games aim to develop reflective and conceptual understanding rather than direct behavioral transformation.

The distribution of learning mechanisms was uneven: reflection appeared in 22 studies, coordination in 13, identification in 9, and transformation in 3. This distribution shows that most games promote self-awareness and perspective taking but rarely enable systemic changes or real-world applications. For instance, Minecraft-based communities [78] provided safe spaces for adolescents to explore identity and social belonging, while DEEP [71] encouraged users to monitor and regulate emotions. In contrast, CHARISMA-virtual social training [51] exemplified stronger coordination and transformation, linking coaching-based gameplay with clinical follow-up and remote engagement. While participatory design processes frequently involved caregivers, therapists, or educators, few studies explicitly articulated a boundary-crossing framework. Consequently, most serious games functioned as standalone interventions, rather than integrated components within health care systems.

The conceptualization of models of reality varied significantly across the reviewed studies. Based on an analytical typology developed for this review, 4 types were identified: causal (n=12), procedural (n=7), relational (n=9), and structural (n=5). Causal and procedural models were most frequent overall (each appearing 19 times across primary and secondary codings), reflecting a focus on feedback-driven learning and stepwise behavioral training. Relational and structural models typically appeared as secondary logics, emphasizing social interaction and conceptual scaffolding. Common examples include virtual reality exposure therapy [72], which applied a causal–procedural model to replicate public speaking scenarios, and iSocial [58], which integrated relational and structural logics to foster social competence through collaborative virtual environments. In contrast, some games oversimplified real-world complexity, such as urban navigation simulators [74], which lacked the unpredictability representative of real-life settings.

A cross-analysis of model types and transfer forms revealed strong internal alignment between theoretical constructs and design intent. Causal–figural (n=9) and relational–figural (n=9) pairings dominated, representing enabling configurations that support reflective and metaphorical learning. Conversely, procedural–literal (n=4) and structural–literal (n=2) combinations were less common but corresponded to direct skill training and conceptual knowledge transfer.

These results indicate that most games intentionally (or implicitly) align their learning mechanisms with their representational logic: figural models enable reflective and identity-building experiences, while literal designs facilitate direct behavioral rehearsal. The absence of relational–literal pairings further underscores that social learning is almost exclusively represented symbolically in adolescent mental health games.

Stakeholder engagement was a consistent strength across many studies, ensuring relevance and usability through participatory design. For example, €UReka [54], a 3D game for financial literacy, involved multidisciplinary teams in tailoring design challenges, while virtual interactive gaze–based adaptive response technology [65] used real-time adaptive feedback to improve social communication. However, broader collaboration with policymakers and institutional stakeholders was rare, limiting scalability and integration into health care systems.

From a design perspective, these 2 forms of engagement can be understood as “design in the small” and “design in the large.” The former refers to participation in the artifact-level design (mechanics, interface, and content), whereas the latter concerns the contextual integration of the artifact within health care or educational systems. While most studies effectively address design in the small, few extend their focus to design in the large, leaving questions of implementation, sustainability, and policy alignment unresolved. Future research should therefore emphasize multilevel stakeholder engagement, involving decision-makers and regulatory partners to facilitate scaling from pilot projects to institutionalized practice [53].

This review highlights distinct but interrelated design tendencies: (1) a preference for figural transfer and reflective boundary mechanisms, (2) a predominance of causal–procedural and relational–structural model clusters, and (3) a strong alignment between representational logic and learning mechanisms. These findings suggest that serious games for adolescent mental health are predominantly designed as reflective learning tools, fostering emotional and cognitive awareness rather than direct behavioral transformation. However, their impact depends less on which configuration is used and more on how deliberately these design choices are made and articulated. When the mechanisms of transfer, boundary crossing, and the model of reality are explicitly aligned within the design process, they can actively enable boundary crossing conditions.

From a design perspective, this alignment operates across 2 levels: design in the small, focusing on the internal coherence of the artifact (mechanics, narrative, fidelity, and user experience), and design in the large, emphasizing how the artifact interacts with its broader institutional, educational, or clinical context. By making these theoretical dimensions explicit during both levels of design, developers and researchers can create serious games that not only model health care processes but also facilitate real-world transfer and systemic learning.

This review highlights the transformative potential of serious games in health care while identifying key areas for improvement. A major challenge is the need for a coherent design rationale that explicitly integrates transfer, boundary crossing, and models of reality into the process of game development. Without such a foundation, serious games risk remaining isolated interventions rather than scalable tools for systemic change. This risk is reflected in the reviewed studies by the limited occurrence of transformation mechanisms (3/33 studies), the predominance of figural transfer (24/33 studies), and the concentration of stakeholder engagement at the artifact level rather than at the level of institutional implementation. Recent meta-analyses confirm that although serious games show positive effects on engagement and short-term outcomes, long-term validation and theoretical consistency remain limited [79].

Among the reviewed studies, 24 employed figural transfer and 10 literal transfer, indicating that most serious games rely on symbolic or metaphorical learning rather than direct simulation. Figural transfer supports generalizable emotional and cognitive skills, whereas literal transfer offers task-specific realism. This emphasis on figural approaches suggests a preference for flexibility over precision but also highlights the need for scaffolding and debriefing mechanisms to reinforce learning. Drawing on educational and simulation-based research [80,81], future designs should integrate structured reflection and adaptive feedback to strengthen transfer beyond gameplay. This finding mirrors recent discussions in digital learning and simulation literature, emphasizing the importance of scaffolding mechanisms for durable learning outcomes [82,83].

Patterns in the data also reveal an alignment between transfer type and model typology: games employing causal or relational models tend to favor figural transfer. This finding suggests that representational logic and learning design interact closely, underscoring the importance of explicit design decisions to ensure coherence between game mechanics, learning goals, and therapeutic intent. This alignment is further illustrated by the absence of relational–literal pairings and by the concentration of figural transfer within causal and relational primary models (see Multimedia Appendix 2).

Boundary crossing was frequently observed but unevenly implemented. Reflection appeared in 22 studies, whereas transformation, representing deep cross-context learning, occurred in only 3 studies. Most games promoted perspective taking or self-awareness but stopped short of enabling collaborative learning or system-wide integration. This pattern corresponds with the low frequency of transformation as a boundary-crossing mechanism and suggests that learning is rarely designed to extend beyond the immediate intervention context. According to Akkerman and Bakker [29], transformation emerges when learning processes bridge distinct social or professional domains.

Explicitly defining intended boundary-crossing mechanisms within the design process could help serious games act as boundary objects, facilitating communication among clinicians, educators, and developers. This translational function positions games not only as therapeutic tools but also as mediators that enable shared understanding across sectors of health care.

Analysis of model types revealed a dominance of causal and procedural frameworks, with relational and structural models often appearing as secondary. These choices reflect a focus on mechanistic and process-oriented representations, supporting procedural learning that may limit the representation of social and ethical complexity when relational or structural models are not explicitly incorporated. This aligns with Bogost’s [38] notion of procedural rhetoric, where games express meaning through rules and system dynamics rather than narrative content.

While causal and procedural models effectively convey behavioral and process learning, they risk neglecting the interpersonal dimensions of health care. By contrast, relational and structural models, exemplified in iSocial [58] and CHARISMA-virtual social training [51], capture the social and contextual aspects of health practice. Integrating these model types can produce games that are both systematically rigorous and socially responsive, mirroring the real-world complexity of health care environments.

Stakeholder engagement was a consistent strength but remained concentrated at the artifact level. Across the reviewed studies, participatory involvement most frequently informed game mechanics, content, or usability, while engagement with organizational, policy, or stakeholders was rarely reported. Most studies prioritized design in the small through participatory methods involving therapists, educators, or caregivers, but few extended collaboration to policy, administration, or implementation levels. Expanding engagement to design in the large is essential for ensuring that serious games transition from prototypes to sustainable health care tools [84].

Across studies, serious games for adolescent health care predominantly support reflective and representational learning, emphasizing awareness and regulation rather than long-term behavioral transformation. However, the findings indicate that effectiveness depends less on any single design dimension than on their explicit alignment.

When transfer mechanisms, boundary-crossing strategies, and model logics are purposefully coordinated, serious games can actively enable boundary-crossing conditions. This allows learning to flow between digital environments, clinical practice, and educational systems. Such alignment strengthens both the internal validity of the game as a learning artifact and its external applicability as a health care intervention.

Based on the observed design patterns across the reviewed studies, these implications can be understood as a conceptual synthesis rather than a prescriptive model. In practice, this means designing serious games as multilevel systems that bridge theory and implementation, ensuring fidelity, feedback, and learner engagement, embedding cross-professional collaboration and evaluation, and aligning design with institutional and policy frameworks. With this approach, future developers and researchers can create serious games that not only demonstrate short-term efficacy but also achieve sustained, scalable, and clinically relevant impact in health care delivery.

This review advances serious game research by applying a structured multicriteria framework to examine design rationale, learning transfer, and real-world applicability. By analyzing 33 studies through the lenses of transfer, boundary crossing, and models of reality, this review provides one of the first quantified overviews of how theoretical constructs are applied in serious game design. The inclusion of typological and frequency analyses strengthens methodological transparency and highlights recurring design patterns.

However, several limitations should be acknowledged. The review is limited to published studies, introducing publication bias toward positive outcomes. While coding followed systematic criteria, some interpretive judgments were necessary to infer implicit design rationales. In addition, full PRISMA 2020 compliance and broader database coverage would enhance reproducibility. Finally, the scarcity of longitudinal studies constrains the understanding of sustained behavioral or cognitive impact. Despite these constraints, this review provides a solid foundation for future work. It shows that making design choices around transfer, boundary mechanisms, and models of reality explicit and aligned can help create more scalable, theoretically grounded health care games.

Future research should pursue longitudinal studies to evaluate how learning from serious games endures in real-world health care settings. Establishing standardized methods for assessing transfer, boundary crossing, and models of reality will improve comparability and design validation. Emphasis should be placed on explicit, theory-informed design alignment and interdisciplinary collaboration among developers, clinicians, and researchers. Broader stakeholder engagement, including policymakers, is vital for scaling interventions and integrating serious games into health care systems. These efforts can transform serious games from isolated innovations into sustainable, evidence-based tools that enhance learning, practice, and health care delivery.

Building on these insights, this review demonstrates the transformative potential of serious games in health care, particularly for individuals with developmental and behavioral needs [1,43]. By fostering skill acquisition, collaboration, and meaningful transfer between digital and real-world contexts, serious games offer promising tools for advancing emotional regulation, cognitive development, and social interaction. However, their scalability and long-term impact remain limited by inconsistently articulated design rationales, which hinder replicability and adaptation across health care settings [29,35].

By systematically examining learning transfer, boundary crossing, and models of reality, this review provides an evidence-based framework for understanding how design decisions influence learning outcomes and real-world applicability. The findings highlight that explicit alignment among these constructs supported by participatory co-design and interdisciplinary collaboration is essential for achieving sustainable health care innovation. Broadening stakeholder engagement to include clinicians, caregivers, policymakers, and end users can further strengthen contextual integration and scalability [64].

Future research should focus on developing standardized, theory-informed frameworks [33], implementing longitudinal and mixed methods evaluations [44], and establishing implementation guidelines that connect design in the small (artifact-level) with design in the large (systemic integration). Through such structured, evidence-based approaches, serious games can evolve from isolated prototypes into validated, scalable instruments that enhance clinical decision-making, therapeutic engagement, and professional development in modern health care. Future work should also focus on validating these design principles across larger datasets and contexts. Comparative and meta-analytic approaches could test how alignment among transfer, boundary crossing, and model typology predicts real-world outcomes, advancing serious game design as an evidence-based methodology in health care.

This review shows that serious games in adolescent health care predominantly support reflective and metaphorical learning, while rarely specifying how learning is expected to transfer beyond the game context. Making transfer mechanisms, boundary-crossing strategies, and underlying models of reality explicit during design could improve comparability, validation, and scalability of serious games in health care practice.