In-Silico Approaches in Sports Epidemiology and their Contribution to the Research on Prevention through Physical Activity in Childhood

Introduction: Physical inactivity in childhood is a global public health concern. Traditional epidemiological approaches often fail to capture the complex interplay of behavioral, environmental, and social determinants. In-silico methods—computational modeling and simulations—offer a promising alternative to analyze and predict physical activity (PA) behaviors and intervention effects in youth populations.

Methods: This Rapid Review applied a systematic search strategy in Scopus and PubMed to identify peer-reviewed studies (published until May 2025) employing in-silico approaches in the context of physical activity promotion among children and adolescents. Screening and data extraction followed a structured eight-step framework aligned with Cochrane Rapid Review guidelines. Eligible studies included modeling techniques such as agent-based models (ABM), system dynamics models (SDM), and microsimulations.

Results: Out of 59 initial records, 24 studies met all inclusion criteria. Most studies applied ABM or SDM to simulate physical activity outcomes under varying policy or environmental conditions. Common themes included school-based interventions, spatial layout impacts, and peer network influences. While most models clearly stated their objectives, empirical validation and use of device-based activity data were lacking. The reviewed studies demonstrate potential but also highlight gaps in data integration, validation practices, and stakeholder engagement.

Discussion: In-silico approaches can substantially advance the design and evaluation of PA interventions in youth. However, their practical utility depends on stronger empirical grounding and participatory development. Future research should emphasize model transparency, validation with real-world data, and broader contextual diversity to ensure relevance and impact in public health policy and practice.

Key Words: Agent-Based Modeling, Physical Activity Promotion, Simulation Studies, Youth Health Behavior, Computational Public Health

Physical inactivity in childhood is a major public health concern. Promoting physical activity early helps prevent chronic diseases like obesity, type 2 diabetes, and cardiovascular conditions, and fosters long-term healthy behaviors (12, 17).

Traditional epidemiological methods often fall short in capturing the complex, reciprocal influences of individual, social, and environmental factors on physical activity. Particularly in developing and evaluating interventions, tools that address this multidimensionality are lacking. In-silico methods offer a promising alternative.

These computer-based simulation techniques allow researchers to analyze, predict, and visualize complex biological or behavioral processes in virtual environments. They span a wide range of approaches, including Artificial Intelligence (AI), Machine Learning (ML), Deep Learning (DL), imaging, statistical modeling, and data-driven simulations—leveraging modern computational power to support decision-making and simulate health trajectories (20, 31).

In public health, in-silico methods are increasingly used to evaluate interventions and guide policy—for example, in toxicity prediction, pathogen surveillance, and biomarker validation (23, 28, 29).

In sports epidemiology, especially in youth PA research, in-silico models help explore the interplay of behavioral, social, and environmental factors. They enable virtual testing of complex interventions before real-world implementation.

This Rapid Review aims to identify and synthesize the current evidence on the use of in-silico approaches within the field of sports epidemiology, specifically in relation to physical activity promotion and prevention strategies among children and adolescents (table 1). The objective is to provide an overview of existing modeling studies that apply computational methods to analyze, simulate, or predict movement-related health outcomes in public health contexts

This review is guided by the following key research questions:
- What kinds of studies exist that apply in-silico approaches in the context of physical activity and movement-related public health research in children and adolescents?
- Which specific in-silico methods are used in these studies, and how are they implemented within epidemiological and sports science frameworks?
- Where do research gaps exist in the current evidence base regarding the use of these methods in movement-related prevention strategies for young populations?

The questions will further inform a critical discussion on the potential benefits, opportunities, and limitations of in-silico methods in the context of physical activity promotion in childhood.

This Rapid Review was conducted following the eight-step methodology developed by the Competence Network Public Health COVID-19 (30), which is based on the guidance of the Cochrane Rapid Reviews Methods Group (33). This structured framework incorporates key developments in systematic evidence synthesis and comprises the following steps: (1) formulation of the research question, (2) literature search, (3) title, abstract, and full-text screening, (4) data extraction, (5) risk of bias assessment, (6) evidence synthesis, included within discussion: (7) dissemination, (8) updating.

Literature Search Strategy
A structured electronic literature search was carried out in April 2025 using the databases Scopus and MEDLINE (via PubMed). The search strategy aimed to identify relevant studies that applied in-silico modeling methods in the context of physical activity or movement-related public health interventions among children and adolescents.

The full search string is provided in Appendix I (see supplemental material online). Briefly, the search combined terms related to four main domains:
(1) in-silico modeling (e.g., “in silico”, “simulation model”),
(2) physical activity (e.g., “exercise behavior”, “movement behavior”),
(3) target population (e.g., “children”, “adolescents”), and
(4) context (e.g., “school”, “public health”).
Biomedical and molecular terms were excluded to increase specificity for behavioral and public health modeling (table 2).

Screening Process and Eligibility Criteria
Following a calibration phase involving double screening of 10 abstracts and 5 full texts by two reviewers (CN and KW), the remaining title/abstract and full-text screening was conducted by one reviewer (CN). A random sample of 20% of all records was independently reviewed by a second reviewer to ensure consistency and reliability of the selection process, in line with the recommendations of the Cochrane and Public Health COVID-19 networks (30). In cases of uncertainty or disagreement regarding eligibility, the reviewers engaged in brief discussion to reach consensus.

Studies were included based on the following predefined eligibility criteria. Only peer-reviewed articles and reviews published in English till Mai 2025 were included. 

Screening Results
A total of 43 studies were identified via the Scopus search (figure 1). The PubMed search yielded 16 additional records, all of which were duplicates. After deduplication, 43 unique studies remained and underwent structured abstract screening. Eligibility was assessed based on abstract content, with common exclusion reasons including lack of a defined target population and no mention of in-silico methods.
Full-text screening was required for 28 studies due to unclear eligibility. Three were excluded for lack of full-text access (36, 40, 45), and one (47) for focusing on exposure to particulate matter rather than physical activity interventions.

No further exclusions occurred during full-text screening. Kasman et al. (22), a study protocol without simulation results, was included due to its detailed and innovative modeling framework. Frerichs et al. (15) was also retained for its participatory modeling approach, despite lacking model execution or validation.
In total, 24 studies met all inclusion criteria and form the evidence base of this Rapid Review. A complete list of excluded studies with reasons for exclusion is available in Appendix II.

Data Extraction
Key characteristics of the included studies were extracted and are presented in table 4.

Risk of Bias Assessment
Given the methodological diversity of in-silico studies and the pragmatic needs of a Rapid Review, a simplified, domain-based assessment was applied, following criteria from public health simulation research. As proposed by key quality indicators included clarity of objectives, empirical validation, and transparency of assumptions (table 3). These formed the basis for assessing risk of bias across three domains:
- Model objectives: Was the simulation aim clearly defined (21)?
- Empirical validation: Was the model calibrated or tested against real-world data?
- Transparency: Were assumptions, parameters, and mechanisms fully described?

Each study was rated as low, high, or unclear risk of bias per domain. All studies clearly stated their objectives. However, most lacked formal empirical validation, limiting external validity. Only a few recent works incorporated validation or calibration (1, 24, 31), while most relied on theoretical inputs. Transparency was generally good, though details were sometimes sparse, particularly in early-stage studies (4, 15, 16, 22). Some models offered extensive reporting (24, 35); others omitted key information on equations, sources, or sensitivity analyses.

In summary, the included studies present a moderate risk of bias. All clearly defined their objectives (low risk), but none underwent empirical validation (high risk), and assumption reporting varied. This judgment aligns with adapted risk-of-bias criteria for public health modeling and highlights the need for improved calibration, transparency, and end-user involvement in future studies.

Evidence Synthesis
Due to the high heterogeneity of the models, a quantitative synthesis was not conducted. Instead, a qualitative evidence synthesis was performed (table 4).

What kinds of studies exist that apply in-silico approaches in the context of physical activity and movement-related public health research in children and adolescents?

The included studies (n=24) encompass a diverse range of in-silico modeling approaches applied to public health questions related to physical activity (PA) in children and adolescents. The majority employ agent-based modeling (ABM) or system dynamics modeling (SDM) to explore behavioral, environmental, and policy-related influences on PA.

Several studies use ABM to simulate individual-level movement behavior or intervention impacts, such as school design and layout (9, 19), environmental exposure (18), or peer and network dynamics (10, 15, 34). Others apply modeling to urban and transport planning (2, 3, 8) or to test dynamic responses to structural interventions in communities (6, 26).

Health impact and policy models are represented by studies like (13, 24, 25), the latter of which focuses on obesity and diabetes prevention at the population level. Several studies examine economic and health outcome trade-offs, including cost-effectiveness analyses (1, 16, 32), while (27) simulate PA-related behavior change mechanisms using artificial populations.

Some papers focus on technical or conceptual model design, including protocol-based frameworks (4, 5, 22) and recommender system development (35). A few studies explicitly integrate mental health, behavioral co-factors, or sociotechnical feedback loops, illustrating the multidimensional scope of modern in-silico applications (1, 7).

Notably, none of the models utilized device-based physical activity measures, such as accelerometer or wearable-derived data, despite their growing availability and importance in PA surveillance.

Across all studies, physical activity is framed within broader public health, behavioral science, or health system perspectives, with a consistent focus on children and adolescents as target populations.
Which specific in-silico methods are used in these studies, and how are they implemented within epidemiological and sports science frameworks?

All included studies apply in-silico modeling within public health, behavioral science, or epidemiology. Agent-Based Modeling (ABM) is the most common approach, followed by System Dynamics Modeling (SDM) and Microsimulation. ABMs simulate peer interactions, spatial behavior, and intervention pathways across diverse settings (2, 6, 8, 9, 10, 15, 18, 19, 22, 27, 34). SDMs address feedback loops and long-term trends in population-level studies (1,3, 26), while microsimulations assess policy impacts via individual-level transitions (13, 24, 32). Decision-analytic modeling supports cost-effectiveness analyses (1, 16, 32). Hybrid or conceptual models appear in exploratory work (4, 5, 21, 35).

Implementation contexts included:
- School/classroom settings: (6, 7, 9, 18)
- Urban/built environments: (2, 3, 8)
- Peer/social networks: (10, 15, 34)
- Health systems/surveillance datasets: (13, 24, 25)
- Digital/personalized health tools: (35)
- Mental health/behavioral factors: (1)

Most models used empirical or survey data, though validation varied. While some included calibration or comparison with real-world benchmarks (25, 32), many remained conceptual – highlighting a general lack of formal model validation.

Where do research gaps exist in the current evidence base regarding the use of these methods in movement-related prevention strategies for young populations?

The review identifies several recurring gaps in the current evidence base regarding in-silico approaches for physical activity and movement-related public health research in youth populations:

Validation Deficits
The majority of studies did not formally validate their simulation outputs against empirical intervention data. While some recent work incorporated calibration or internal validation (e.g. 1, 25, 32), many models remain untested beyond their hypothetical scenarios (e.g. 2, 9, 24, 27).

Limited Participatory Modeling
Very few studies engaged youth, educators, or community stakeholders in model development or interpretation. Exceptions such as Frerichs et al. (15) and de Mello Araújo et al. (10) illustrate participatory approaches, but these remain rare, reducing practical relevance and real-world applicability.

Geographic and Contextual Concentration
Most models are based on data and assumptions from high-income Western countries. There is limited representation of low- and middle-income contexts or culturally diverse populations, which limits the transferability of findings (e.g. (3) is one exception from Latvia).
Personalization and Dynamic Heterogeneity
Although some models account for individual differences (e.g., 16, 22, 35), adaptive, personalized, or longitudinally responsive modeling approaches are largely missing. There is a need for more dynamic simulations that reflect developmental, behavioral, or environmental change over time.

Policy Translation and Decision-Making Integration
Several studies estimate health or economic outcomes (1, 13, 16), but few are explicitly co-developed with policymakers or embedded into real-world planning processes. The integration of simulation outputs into policy workflows, especially in schools or municipal health departments, remains limited.

This Rapid Review synthesizes current in-silico modeling approaches targeting physical activity and movement-related prevention in children and adolescents. Among the 24 included studies, simulation methods – especially agent-based models (ABMs), system dynamics models (SDMs), and microsimulations – emerge as valuable tools in public health and epidemiology.

Applications span school-based movement (9), built environments (2, 3), policy scenarios (13, 24), and peer-based intervention diffusion (10, 15, 34). These methods help explore system dynamics, long-term effects, and ethically sensitive scenarios—especially relevant in youth-focused PA research, where experimental designs are limited.

However, key methodological gaps persist. Few studies included calibration or validation; most lacked empirical testing, limiting policy relevance (1, 25, 32). Many relied on simplified proxies (e.g., step counts), often omitting psychosocial or environmental factors.

Contextual diversity was limited; most models originated from high-income Western countries, with little attention to equity or community input. Participatory modeling – engaging youth, educators, or policymakers – remained rare, despite its value for relevance and uptake.

In-silico models show promise for evaluating interventions, informing school strategies, and designing active environments. Yet their full potential depends on stronger validation, behavioral depth, and stakeholder integration. Future research should co-develop models with end-users, align simulations with real-world data, and ensure transparency to strengthen their role in PA science.

Dissemination and Translation
To bridge the gap between modeling and practice, results will be shared via academic and professional channels, including the Deutsche Vereinigung für Sportwissenschaft (dvs; German Association for Sports Science), the Health Summit, and platforms such as the Leibniz ScienceCampus Karlsruhe and KIT Center Health Technologies. These efforts aim to foster interdisciplinary exchange and support real-world adoption of in-silico methods in prevention.

Given the rapid evolution of digital modeling and AI-driven simulation, regular updates of this review are advisable. A systematic update within 12–18 months could capture emerging research on machine learning for PA prediction, hybrid models using sensor data, studies from low- and middle-income countries, and participatory modeling approaches. Future updates may also apply formal quality frameworks (e.g., GRADE, ISPOR) and better link simulation to PA surveillance and implementation science—crucial steps to bridge the gap between theoretical models and real-world youth movement promotion.

A key insight—reflected in both the reviewed studies and broader PA research—is the need for high-quality, device-based data. As Ding and Ekelund (11) emphasize, accelerometer data have enhanced exposure assessment in PA epidemiology. Yet most models still rely on self-reported or aggregate inputs, limiting behavioral precision and dose–response accuracy.

Only a few studies used contextual data beyond theoretical assumptions, and none directly integrated device-based PA measures (9, 34). This gap reflects a structural barrier: without rich, representative input data, even sophisticated models risk oversimplifying youth movement behavior.

A forward-looking in-silico research agenda should incorporate validated physical activity data, reflect diverse populations, and prioritize models that are not only technically rigorous but also epidemiologically relevant and ethically responsible.

Implications and Limitations of In-Silico Modeling for Physical Activity Promotion
In-silico approaches offer strong potential for hypothesis generation, intervention design, and policy support in youth physical activity research. Their strength lies in simulating complex behavioral and environmental systems, enabling exploration of “what-if” scenarios and identification of strategic leverage points. This makes them valuable tools for early-stage decision-making and complements to traditional public health evaluation methods.

However, their impact is limited by insufficient empirical validation – especially for complex behaviors like physical activity. Without calibration against high-quality observational or longitudinal data, model reliability and generalizability remain constrained. Future studies must integrate validated, context-specific input and adopt rigorous validation procedures. Only then can in-silico methods evolve from exploratory tools to reliable instruments in evidence-based physical activity promotion.

Conflict of Interest
The authors have no conflict of interest.

Acknowledgements
Author Notes: We used ChatGPT-4 (GPT-4-Turbo, version: April 2025) to correct typographical errors and improve the grammar of the text. All errors are attributable to the authors.

Ethical Approval
Not applicable.

Funding
None.

This review highlights the growing relevance of in-silico methods for understanding and promoting physical activity among children. It addresses a critical gap in traditional sports epidemiology by emphasizing computational approaches to model complex health behaviors.

A structured rapid review was conducted using Scopus and Pub-Med to identify studies applying agent-based and system dynamics models in youth physical activity research. The approach follows Cochrane rapid review methodology, ensuring systematic screening and extraction.

The review identified 24 relevant studies, most of which used simulation to evaluate school-based and environmental interventions. Despite innovative modeling, many studies lacked empirical validation and sufficient data integration.

In-silico modeling holds promise for informing effective physical activity interventions in childhood. However, future work must prioritize empirical grounding, stakeholder engagement, and methodological transparency to enhance public health impact.