A Scoping Review of the Literature on Child and/or Adolescent Health Microsimulation Models Authors: Claire de Oliveira, Yingkun Justin Wen, Michael Endale Mengesha, Filipa Sampaio Published: 2026 ABSTRACT AbstractBackgroundMicrosimulation models can be used for diverse purposes such as forecasting and undertaking economic evaluations. A review of the literature of existing child and/or adolescent microsimulation models has yet to be undertaken. The aim of this scoping review was to retrieve, synthesise and critically appraise the literature of child and/or adolescent health microsimulation models.MethodsA scoping review was conducted using PubMed, Embase, CINAHL, PsyclNFO, EconLit and Scopus. Additional searches (snowballing, Google searches, searches of specific journals and websites) were also undertaken to complement database searches. Relevant data on microsimulation models were extracted, and the quality of models was assessed using the Quality Assessment Reporting for Microsimulation Models checklist. Evidence was synthesised by scope of model, type of model and health outcome examined.ResultsForty-eight relevant microsimulation models were found. Most (n = 33) were developed specifically for children and/or adolescents, but there were a few (n = 15) broader models that examined these populations. Many microsimulation models were developed for the USA (n = 14); several were focused on select health conditions, such as obesity/overweight, cancer and human immunodeficiency virus (HIV). Just over half (n = 27) of the models found were classified as medium quality models, while the remaining ones were deemed high quality (n = 21). Nonetheless, several models lacked validation and, to a lesser extent, uncertainty analyses and generalisability.ConclusionsThere are a substantial number of child and/or adolescent health microsimulation models. However, future research should focus on developing models on other prevalent child/adolescent health conditions (e.g. allergies, ADHD) and for low- and middle-income settings.Supplementary InformationThe online version contains supplementary material available at https://doi.org/10.1007/s40258-026-01066-y. Key Points for Decision-Makers Microsimulation models are commonly used as ex-ante evaluation tools to examine potential behavioural and economic effects of interventions and/or policies and thus guide decision-making.This review sought to identify existing child and/or adolescent health microsimulation models to understand the work undertaken in this area as well as the main limitations of existing microsimulation models.This reviews’ findings, namely the methodological considerations and areas for future microsimulation model development, will be relevant for researchers and decision-makers looking to build robust child and/or adolescent health microsimulation models. Introduction The effects of childhood and adolescence poor health on later life outcomes are well known. A prior systematic review found that childhood circumstances affected not only individuals’ physical and mental health, but also their cognitive function and social relationships in adulthood [1]. For example, researchers have found a strong connection between birth weight and childhood health conditions and later life outcomes, such as educational attainment and labour market outcomes, using the 1958 longitudinal British birth cohort study data [2, 3]. Undertaking these analyses was made possible because of the use of longitudinal data that spanned many years, where individuals were followed over several decades. Unfortunately, long panels, such as the British birth cohort study, are not always available and can be costly to maintain over time. Microsimulation models can provide an attractive and cost-effective solution to this problem—they can simulate the behaviour of micro-agents, such as children or adolescents, over the lifetime and be used to understand the potential long-term behavioural and economic effects of health interventions or policies. Additionally, microsimulation models can be useful tools to help guide decision-making, particularly in instances where evidence is absent or limited. Previous work examining microsimulation models on mental health [4] found that most models were focused on the general adolescent and adult populations, with only one model specifically examining the mental health of children [5]. However, there are existing microsimulation models on child and adolescent health that have been used to ascertain the impact of child-related health policies [6] and to understand the causes and consequences of health behaviours, such as smoking, on adolescent health [7–9]. A comprehensive review of the literature is helpful to determine the existing microsimulation models that can be used to examine child and/or adolescent health. It is also important that microsimulation models produce robust and valid outputs that can be used for decision-making. Therefore, evaluating the quality of microsimulation models can help in this regard as it can provide methodological guidance for researchers looking to build similar models. The goal of this scoping review was to retrieve and synthesise the existing microsimulation models on child and/or adolescent health as well as to critically appraise the quality of these microsimulation models. Methods and Analysis Study Design Scoping reviews are useful reviews to describe the existing literature on a particular topic, while systematic reviews are used to answer a specific research question [10]. Given the objective of this study, a scoping review was deemed the most appropriate review to undertake [10]. In line with the recommended guidelines for scoping reviews, we followed the Joanna Briggs Institute Manual for Evidence Synthesis [10] and the Preferred Reporting Items for Systematic Reviews and Meta-Analyses extension for scoping reviews (PRISMA-ScR) guidelines [11]. The study protocol was registered at Open Science Framework (https://osf.io/a8txn/) and published in BMJ Open [12]. The scoping review followed the registered/published study protocol with very minor deviations (e.g., data synthesis was ultimately undertaken by the scope of model, the type of model and the health outcome examined in the model instead of just by the latter two). Eligibility Criteria The population, intervention, control, outcomes and study design (i.e., PICOS) criteria were used to guide the development of the search strategy and to inform the study inclusion and exclusion criteria. The population of interest included children and adolescents up to the age of 19, in line with child development stages [13] (newborns, infants, toddlers, pre-schoolers, school-aged children) and the WHO definition of adolescents).[14] All stand-alone microsimulation models (i.e., complete models that could be used for several purposes) were considered; studies and/or analyses, which employed microsimulation modelling techniques to answer, for example, one-off research questions, were not within scope. Microsimulation models that produced outputs for broader populations were considered as long as these outputs could be disaggregated for children and/or adolescents. There were no restrictions on the type of interventions or outcomes examined as long as the latter included health outcomes, such as physical health conditions/disorders, mental health/addiction conditions/disorders, and/or neurodevelopmental disorders. The comparison group criteria were not applicable in this context. The study design was limited to studies that employed a microsimulation model. Only original studies were considered, but snowballing was undertaken on any relevant related reviews in an effort to retrieve additional studies/microsimulation models that were not captured by the search. Study protocols and conference abstracts and/or proceedings were excluded as only completed studies describing microsimulation models were of interest. Search Strategy Six bibliographic databases—PubMed (until 23 September2024), Embase (until 18 September 2024), CINAHL (until 09 September 2024), PsyclNFO (until 11 September 2024), EconLit (until 09 September 2024), and Scopus (until 10 September 2024—were searched from database inception; an update of the search strategy was undertaken prior to manuscript submission in early June 2025. Studies published in English, French, German, Chinese, Portuguese and Swedish were considered, where relevant. In addition, snowballing (i.e., hand searches of references of key articles and reviews), Google searches (using Google and Google scholar) and searches on specific journal websites (e.g., the International Journal of Microsimulation) and other relevant websites (e.g., https://www.microsimulation.ac.uk/) were undertaken. The search was guided by search terms/strings organised by concepts (Table 1). A research librarian, in collaboration with the lead investigators (C.dO., F.S.), developed a structured search strategy, which reflected the inclusion and exclusion criteria; the research librarian then tailored the structured search strategy to each specific database. The full search strategies can be found in the appendix of the study protocol [12]. Table 1Concepts and search terms used to identify relevant studiesConceptSearch termsPopulationNewborns, infants, toddlers, pre-schoolers, school-aged children, tweens, teens, teenagers, adolescents, juvenile, youthIntervention/exposurePhysical health conditions/disorders (allergies, asthma, bronchitis, upper respiratory tract infection, cancer, hypertension, cerebral palsy, congenital heart disease, cystic fibrosis, diabetes, epilepsy, HIV, gastrointestinal disorders, inflammatory bowel disease, irritable bowel syndrome, juvenile arthritis, musculoskeletal disorders, urinary tract infections, urination disorders (anuria, enuresis, glycosuria, haematuria, oliguria, polyuria, proteinuria, incontinence, urinary retention), endocrine disorders, rheumatic disorders, haematological disorders, anaemia, growth disorders, psychosomatic pain, obesity, skin disorders and acne, nutrition disorders [child nutrition disorders, infant nutrition disorders]), mental health/addiction conditions/disorders (anxiety, depression, eating disorder [anorexia nervosa, bulimia nervosa, binge eating disorder], emotional disorder, post-traumatic stress disorder, PTSD, schizophrenia, psychosis, mood disorder, bipolar disorder, mania, obsessive-compulsive disorder, substance use disorder, smoking, alcohol use, drug use, drug misuse, cannabis use, gambling, suicidal ideation, self-harm, suicide), neurodevelopmental disorders (separation anxiety, attention deficit/hyperactivity disorder, conduct disorder, oppositional defiant disorder, child behaviour disorders, autism spectrum disorder, learning disabilities, developmental disabilities, intellectual disability, tic disorders, Tourette syndrome)ComparatorN/aOutcomeN/aStudy designMicrosimulation Concepts and search terms used to identify relevant studies Study Selection All relevant records were uploaded onto Rayyan, an online web application designed for reviews [15], where duplicate records were removed. Two independent reviewers screened the titles and abstracts (Y.W., R.W.) and additional reviewers (C.dO, F.S.) were brought in when there were disagreements. Following this step, the full text of all relevant records was reviewed for inclusion by two independent reviewers (Y.W., R.W.) and then organised by microsimulation model; where necessary, additional reviewers (C.dO, F.S) were consulted to confirm the eligibility of each microsimulation model. Data Extraction A data extraction form, based on the Cochrane good practice data extraction form[16] was developed and included the following elements, in line with a previous related review [4]: study information (author[s], year of publication, and country), aim of the microsimulation model, data source(s) employed in the microsimulation model and study population (e.g., newborns, infants, toddlers, pre-schoolers, school-aged children, adolescents), health outcomes examined in the model (physical, mental, neurodevelopmental and multiple), type of microsimulation model (static—models that evaluate the impact of policy changes on a population, without changing the population characteristics over time; dynamic—models that can evaluate the impact of policy changes but also simulate individual life events (e.g., aging, retirement) over time; and spatial—models that incorporate geography and can be either static or dynamic), description of the microsimulation model and methods employed, robustness checks and model adjustments, such as calibration (i.e., parameter adjustments made to ensure the model is capable of simulating the distributions of key variables), where applicable, validation (i.e., internal validation, where simulated outcomes are compared with actual outcomes, and external validation, where model forecasts are compared with other forecasts) and limitations of the microsimulation model. Two reviewers (Y.W., M.M.) extracted data on each microsimulation model independently and subsequently met to discuss and resolve any differences in the extracted data. A third reviewer (C.dO.) reviewed the extracted data. Quality Assessment The critical appraisal of publications (or, in this case, microsimulation models) is not mandatory in scoping reviews [11, 17]. However, it is important that microsimulation models are robust and of high quality [18] to encourage decision-makers to use these models and their outputs to inform policy. In this instance, assessing the quality of microsimulation models can provide valuable insights. To this end, the Quality Assessment Reporting for Microsimulation Models (QARMM) checklist [4] was employed to assess the quality of the retrieved microsimulation models. This checklist was specifically developed to ascertain the quality of microsimulation models in a previous review [4]. In short, the QARMM covers 6 criteria: (1) purpose of the model, (2) data, (3) transparency, (4) uncertainty, (5) validation and (6) generalisability. The first three criteria describe the model development and structure, while the last three criteria refer to the validity and scope of the outputs produced by the microsimulation model. Each criterion of the checklist is worth one point (though, in some instances, half points are attributed when a given criterion is not fully met). Table 2 provides details of the QARMM. The rating of each microsimulation model was done by two independent raters (Y.W., M.M.) who subsequently met to resolve any disagreements; a third reviewer (C.dO.) was consulted in instances where there were disagreements. Microsimulation models with a score of three or less were classified as low-quality models, those with a score greater than five were considered high quality models, while models with scores in between were deemed as medium quality models, in line with previous related work [4]. Table 2The Quality Assessment Reporting for Microsimulation Models (QARMM) checklistItemsPointsModel development and structurePurpose of the modelAre the objectives/goals of the model well defined?Is the target population described appropriately?1DataAre the data used in the model development representative of the population examined?Are the data sources informing parameter estimations provided?Are the parameters used to populate model frameworks specified?1TransparencyIs the model structure well described (e.g., are assumptions clear, are there choices for the user to make)?Is the time horizon of the model provided?Are all simulated strategies/scenarios specified and/or explained clearly?1Validity and scopeUncertaintyAre model uncertainties discussed?Are sensitivity analyses performed and reported?Is model calibration performed and reported, where required?1ValidationIs internal and/or external validation performed and reported?Is predictive validation performed and/or attempted?1GeneralisabilityIs model generalisability discussed?Are model limitations discussed?1 The Quality Assessment Reporting for Microsimulation Models (QARMM) checklist Purpose of the model Are the objectives/goals of the model well defined? Is the target population described appropriately? Data Are the data used in the model development representative of the population examined? Are the data sources informing parameter estimations provided? Are the parameters used to populate model frameworks specified? Transparency Is the model structure well described (e.g., are assumptions clear, are there choices for the user to make)? Is the time horizon of the model provided? Are all simulated strategies/scenarios specified and/or explained clearly? Uncertainty Are model uncertainties discussed? Are sensitivity analyses performed and reported? Is model calibration performed and reported, where required? Validation Is internal and/or external validation performed and reported? Is predictive validation performed and/or attempted? Generalisability Is model generalisability discussed? Are model limitations discussed? Data Synthesis Given the goal and the nature of the analysis, undertaking a meta-analysis was not within the scope of this review. Instead, a narrative synthesis was employed to synthesise the evidence [19], following an approach taken elsewhere [4]. The evidence was synthesised by the scope of the model (e.g., child- and/or adolescent-specific models versus broad models), type of model (e.g., static, dynamic, spatial) and health outcome examined. Results Microsimulation Model Selection Citations were merged across all academic databases (n = 10,572) and grey literature sources (n = 11); after duplicate records were removed (n = 4,379), 6,204 unique records remained for screening. All records found were in English. After the title and abstracts were reviewed, 253 full texts were assessed. Following full text screening and the addition of articles/microsimulation models from other sources (n = 2), 83 papers were ultimately deemed relevant for inclusion in the review [5–8, 20–63] (Fig. 1). Among the 83 papers, there were instances where some papers referred to the same microsimulation model. In particular, taken together, these 83 papers corresponded to 48 unique microsimulation models; in cases where more than one paper used the same microsimulation model, data extraction and quality assessment were undertaken on the original paper (i.e., the paper that provided the details around how the model was built). Furthermore, the microsimulation models could be grouped into distinct two groups: child and/or adolescent health-specific models (n = 33) [5–7, 20–49] and broad models/full population models that could be used to examine and report child and/or adolescent health outcomes (n = 15) [8, 50–63]. PRISMA flow diagram. PRISMA—Preferred Reporting Items for Systematic Reviews and Meta-Analyses. *These articles correspond to 48 unique microsimulation models The Childhood Obesity Intervention Cost-Effectiveness Study (CHOICES) model [27] was used in 11 distinct manuscripts captured in the review [27, 64–73]. Other frequently used microsimulation models included the IndiaSim [31] used in five studies [31, 74–77], the SimObesity model [52] used in four studies [52, 78–80], the Early Prevention of Obesity in CHildhood (EPOCH) model [29] used in four studies [29, 81–83], the Global Childhood Cancer (GCC) microsimulation model [47] used in three studies [47, 84, 85] and the Cost Effectiveness of Preventing AIDS Complications (CEPAC) used in three studies [23, 35, 86]. Other models were used in two studies; these included the Intellectual Disability Model (IDMOD) [60, 87], the Lifecourse Dynamic Microsimulation Model (LifeSim) [6, 88], the Lifetime Exposures and Asthma outcomes Projection (LEAP) model [55, 89], the Lymphatic Filariasis Simulation Model (LYMFASIM) [59, 90], the Population Health Model (POHEM) Neurological model [51, 91], the Modelling the Early Life-course (MELC) model [32, 92], the Pediatric Oncology Simulation Model for prevalence (POSIM-Prev) [34, 93], the Hall et al. model [28, 94], the McPherson et al. model [56, 95] , the Zeng et al. model [63, 96] and the Zhang et al. model [49, 97]. All remaining 31 models were each used in a single study. Microsimulation Model Characterisation Most microsimulations models were developed for the USA (n = 14) [5, 7, 8, 20, 24, 26–28, 30, 35, 39, 49, 50, 62] (Fig. 2 and Supplementary Table 1), followed by Canada (n = 6) [34, 38, 42, 51, 54, 55], Australia (n = 3) [29, 45, 60], and the UK (n = 3) [6, 52, 56], among other countries/areas. Five models were developed with a global focus [22, 37, 47, 48, 57]. The main aims of the microsimulation models were diverse and included, for example, simulating the incidence of a given disorder, estimating future costs/cost-savings associated with different interventions or undertaking an economic evaluation of a given intervention/strategy. Several data sources were employed in the development of the microsimulation models, including survey data, administrative/claims data, population statistics/census data and other published work. Some models made use of cross-sectional data [7, 21–26, 28, 31, 58], while other models employed longitudinal data [6, 20, 27, 29, 32, 39, 41, 49, 55]. Regarding the study populations, most microsimulation models examined child populations (n = 33) [5–7, 20–49]. The focus of the microsimulation models was mainly on obesity/overweight (n = 11) [20, 26–30, 39–41, 52, 56], followed by cancer (n = 8) [34, 36, 38, 42, 45, 47, 50, 62], mental disorders (such as conduct disorder, eating disorders, major depressive disorder) (n = 3) [5, 32, 49, 54, 60] and human immunodeficiency virus (HIV) (n = 3) [23, 35, 44] (Fig. 3). Four models looked at multiple outcomes [6, 21, 24, 37]. None of the models included in the review examined family spillover effects. Most microsimulation models were static (n = 29), while the remaining were dynamic models (n = 19); among the models examined, there were only two spatial microsimulation models, one static spatial model, the SimObesity [52], and one dynamic spatial model, a model developed by Sibils and colleagues (2024) [43]. Microsimulation models by country Microsimulation models by health outcome. HIV—human immunodeficiency virus Quality Assessment Just over half of the microsimulation models were classified as medium quality models (n = 27) [5, 7, 8, 20, 21, 25, 26, 28, 29, 31, 32, 35, 36, 40–42, 44, 46, 49, 50, 52, 54, 56, 59–61, 63]; the remaining were considered high quality models (n = 21) [6, 22–24, 27, 30, 33, 34, 37–39, 43, 45, 47, 48, 51, 53, 55, 57, 58, 62]. None of the microsimulation models examined were deemed to be of low quality, according to the QARMM. Generally, all models performed well on model development and structure (i.e., purpose of the model, data employed and transparency of the model), receiving full points (Table 3). Only one microsimulation model was not attributed full points for transparency [8]. Table 3Quality assessment scores of the microsimulation modelsStudy§Model development and structureValidity and scope of resultsTotalPurposeDataTransparencyUncertaintyValidationGeneralisabilityChild and/or adolescent-specific microsimulation models (n = 33)An [20]1110.5014.5Bowman et al. [21]111100.54.5Cavalcanti et al. [22]111110.55.5Chao el al. [7]1110.510.55.0Choi et al. [24]111110.55.5Ciaranello et al. [23]1110.5115.5Dimitrova et al. [25]111100.54.5Ferguson et al. [26]1110.510.55.0Gortmaker et al. [27]111110.55.5Hall et al. [28]111010.54.5Hayes et al. [29]1110*115.0Lim et al. [30]1111116.0Megiddo et al. [31]111100.54.5Milne et al. [32]1110.510.55.0Moncayo et al. [33]111110.55.5Moskalewicz et al. [34]1111116.0Neilan et al. [35]1110*0*0.53.5Palmer et al. [36]11110.50.55.0Pasricha et al. [37]1111116.0Pechlivanoglou et al. [38]111110.55.5Pryor et al. [39]1111116.0Rasella et al. [40]111100.54.5Shackleton et al. [41]1111015.0Shoukry et al. [42]1111015.0Sibils et al. [43]111110.55.5Siriruchatanon et al. [44]111100.54.5Skarda et al. [6]1110.5115.5Tan et al. [45]111110.55.5Verkleij et al. [46]111100.54.5Ward et al. (2019a) [47]111110.55.5Ward et al. (2019b) [5]111100.54.5Ward et al. (2022) [48]111110.55.5Zhang et al. [49]111100.54.5Broader microsimulation models that include children and/or adolescents (n = 15)Alagoz et al. [50]111010.54.5Amankwah et al. [51]111110.55.5Edwards and Clarke [52]1110115.0Etnel et al. [53]1111116.0Gallant and Lukeman [54]1110.500.54.0Lee et al. [55]1110.5115.5McPherson et al. [56]1110.50.50.54.5Meccanici et al. [57]111110.55.5Mertens et al. [58]1111116.0Plaisier et al. [59]1110.510.55.0Reddy et al. [8]110.50.510.54.5Schofield et al. [60]1110*014.0Van der Ploeg et al. [61]11100*0.53.5Yeh et al. [62]1111116.0Zeng et al. [63]111100.54.5§The original or seminal paper of each model is cited*Studies were given a score of zero on this item as they did not provide any evidence of undertaking uncertainty-related analyses and/or validation, where relevant, despite stating that they would undertake these in the future Quality assessment scores of the microsimulation models §The original or seminal paper of each model is cited *Studies were given a score of zero on this item as they did not provide any evidence of undertaking uncertainty-related analyses and/or validation, where relevant, despite stating that they would undertake these in the future The quality assessment scores were lower for the items of model validation, followed by uncertainty and generalisability. In particular, among the models that undertook validation, most undertook internal validation (only) (n = 12), followed by external validation (only) (n = 8). Among the studies that undertook internal validation, two studies also undertook face validation [34, 55]. In addition, there was one study that undertook backward validation (only) [7] , six studies that undertook both internal and external validation [6, 8, 45, 50, 52, 58] and one study that undertook internal, external and cross-validation [38]. With regards to uncertainty analyses, the most commonly used methods included Monte Carlo simulation and/or sensitivity analyses, including probabilistic sensitivity analyses. Further details can be found in Supplementary Table 1. There were 16 microsimulation models that did not undertake any sort of validation and thus were attributed zero points for this item [5, 20, 21, 25, 31, 35, 40–42, 44, 46, 49, 54, 60, 61, 63]. All models met the criteria of generalisability to some extent (either partially or fully). In terms of the microsimulation models’ limitations, some lacked the necessary data required for model parameters or used older data, some required making several assumptions in their development, some were unable to ascertain causality owing to the use of observational data and others lacked generalisability. Evidence Synthesis The evidence was synthesised by the scope of the model (child- and/or adolescent-specific models and broad models), type of model (static, dynamic) and health outcome examined. Microsimulation models varied in quality by scope of model. Regarding the type of model, there was a mix of medium and high quality models among both static and dynamic models. Regarding the health outcome, most cancer-specific models (n = 5/8) were classified as high-quality models [34, 38, 45, 47, 62]. The microsimulation models on obesity were mostly of medium quality (n = 8/11) [20, 26, 28, 29, 40, 41, 52, 56], with three (n = 3/11) classified as high quality models [27, 30, 39]. Mixed findings regarding quality were also found for the mental health and HIV microsimulation models. In sum, no clear pattern regarding quality could be established across the different dimensions used to synthesise the evidence. Discussion Tracking lifetime health outcomes in children and adolescents can be challenging as this requires following individuals over long periods of time. Microsimulation models can help in this regard. This scoping review sought to determine the existing microsimulation models that have been developed with a focus on child and/or adolescent health to understand how these models have been used for research purposes. Forty-eight unique child and/or adolescent health microsimulation models were found. Most microsimulation models were specifically developed to model child and/or adolescent health; however, there were also a few broader models that could be used to produce child and/or adolescent-specific outputs. The focus of most microsimulation models was on select childhood and adolescent conditions, such as obesity/overweight, cancer, HIV and mental disorders. A similar proportion of microsimulation models were deemed to be of medium and high quality. Most microsimulation models were specific to child and adolescent health, suggesting that there has been a recognised need to have population-specific models. On the basis of the studies retrieved, the most frequently used model was the Childhood Obesity Intervention Cost-Effectiveness Study (CHOICES) model [27], which simulates an open population through a non-parametric matching procedure, synthesising data from several sources, namely the USA 2010 Census tract data, the National Health and Nutrition Examination Survey, and the American Community Survey. This model was used in 11 distinct manuscripts captured in the review [27, 64–73]. Other frequently used microsimulation models included, for example, the Early Prevention of Obesity in Childhood (EPOCH) model [29], which was used in four studies [29, 81–83]. This microsimulation model was developed to determine whether early interventions for childhood obesity have an impact on body mass index z-scores at age 4 years and above, drawing on published data from reviews and the Longitudinal Study of Australian Children. Another model, used in three studies [47, 84, 85], was the Global Childhood Cancer (GCC) microsimulation model [47] whose purpose is to estimate childhood cancer incidence for 200 countries and territories worldwide, accounting for trends in population growth and urbanisation, geographical variation in cancer incidence and health system barriers that contribute to the under-diagnosis of childhood cancer. Finally, the Cost Effectiveness of Preventing AIDS Complications (CEPAC), a computer-based, state-transition, Monte Carlo simulation model of the progression and outcomes of HIV, was employed in three studies [23, 35, 86]. This model was initially developed for adults with two executable versions, one tailored to the USA and another for less developed countries but was then later developed for both paediatric and adolescent populations, leading to the creation of the CEPAC-Pediatric [23] and the CEPAC-Adolescent models [35], respectively. In some cases, many models were specifically built for a given purpose, usually the objective of the paper. Microsimulation models were classified as being of medium or high quality. The main elements where models were lacking were validation and, to a lesser extent, the use of uncertainty analyses. As mentioned by Eddy et al., ‘trust and confidence are critical to the success of health care models.’ To ensure this, these authors propose two main elements when building models: transparency (i.e., one can see how the model is built) and validation (i.e., one can assess how well the microsimulation model reproduces reality) [98]. While all models, except one [8], were able to deliver on the criterion of transparency, only 16 models reported undertaking any type of model validation [5, 20, 21, 25, 31, 35, 40–42, 44, 46, 49, 54, 60, 61, 63]. Validation involves evaluating the model structure, data sources employed, assumptions made and the outputs produced. It can take the form of verification or internal validation (where one checks the accuracy of coding and/or where the characteristics of the simulated cohort is compared with those of the observed cohort), cross validation (where one compares the results produced by the model with those of other models analysing the same or a similar problem), external validation (where one compares the model outputs to real world data) and predictive validation (where one compares model outputs with prospectively observed data) [98]. In some instances, undertaking validation can be challenging. For example, the PSY-SIM model, which was developed to study individuals with chronic psychotic disorders, was able to undertake internal validation but struggled to undertake external validation [99], given the lack of research on long-term outcomes among this patient population as well as the dearth of microsimulation models focused on schizophrenia or psychosis. This review’s findings are relevant to researchers and decision-makers looking to build robust child and/or adolescent health-specific microsimulation models to inform policy development and guide future health care delivery and service planning. In particular, the findings from this review provide researchers not only with a comprehensive description of microsimulation models in the area of child and/or adolescent health, but it also provides some methodological guidance in terms of relevant areas to tackle when developing new models. However, additional research is still needed. None of microsimulation models examined in this review focused on other prevalent child and adolescent health conditions, such as allergies, anxiety and ADHD. Researchers may want to consider this when developing future microsimulation models. Additionally, most microsimulation models were developed for high-income countries, such as Australia, Canada, the UK and the USA. Excluding nine studies [22, 23, 33, 37, 44, 47, 48, 53, 57], which used data from different countries and territories, there were only five microsimulation models developed for low- and middle-income settings (e.g. Albania [46], Brazil [59], Ethiopia [21], India [25] and Mozambique [43]). Future research should seek to address this gap in the literature. Moreover, future methodological priorities should focus on ensuring that validation and uncertainty analyses are undertaken when developing and building new microsimulation models. Undertaking validation is important to ensure model outputs are robust, thus building model credibility. In addition, robust outputs help increase users’ confidence in using these models to inform decision-making. Undertaking uncertainty analyses can help identify which factors (e.g., input data sampling errors, parameter estimates, or structural choices) have the largest impact on the simulated outputs, which may also be of interest to users. This scoping review has strengths. To the authors’ knowledge, this is the first scoping review examining microsimulation models developed to examine child and adolescent health. Moreover, although not required, this scoping review assessed the quality of the microsimulation models found, which can be helpful to guide the development of robust microsimulation models in the future. Notwithstanding, this scoping review has some limitations. Despite a comprehensive search, there is the possibility that this review may have missed relevant microsimulation models on child and/or adolescent health. The quality assessment checklist that was used to assess the quality of the models has not yet been validated and each criterion is weighted equally; nonetheless, it has been used in previous related research [4]. Finally, no patients and/or the public were involved in the review or interpretation of the findings. Conclusions This scoping review found a substantial number of microsimulation models that could be used to examine child and/or adolescent health. Although the microsimulation models were of medium or high quality, some models did not undertake any sort of validation. The findings of this review are relevant for researchers and decision-makers looking to build robust microsimulation models examining child and/or adolescent health in the future. Furthermore, more microsimulation models on child and/or adolescent health are required, particularly those with a focus on other prevalent health conditions among children and adolescents, such as allergies, anxiety and ADHD, as well as more models with a focus on low- and middle-income country settings. Supplementary Information Below is the link to the electronic supplementary material. Supplementary file1 (DOCX 159 KB) Acknowledgements We would like to thank the library services provided at Uppsala University for their help in developing the search strategy and Rebecca Wiechers for their assistance in reviewing titles, abstracts and full texts. Written permission has been obtained from both the library at Uppsala University and Rebecca Wiechers. Funding Open access funding provided by Uppsala University. Declarations Conflict of interest The authors do not report any competing interest in accordance with the International Committee of Medical Journal Editors’ requirements. Ethics Approval Not applicable. Consent to Participate Not applicable. Consent for Publication (from patients/participants) Not applicable. Availability of Data and Material Not applicable. Code Availability Not applicable. Authors’ Contributions Claire de Oliveira and Filipa Sampaio conceived and designed the study. Yingkun Justin Wen and Rebecca Wiechers screened the articles in Rayan; Yingkun Justin Wen and Michael Mengesha extracted the data and undertook quality assessment. Claire de Oliveira drafted the original draft of the manuscript. All authors provided comments and critical revisions on drafts of the manuscript. References Associated Data Supplementary Materials Supplementary file1 (DOCX 159 KB)