Mapping immersive technologies in pediatric rehabilitation: a multi-database bibliometric analysis

Immersive technologies, including virtual, augmented, mixed, and extended reality, are increasingly investigated in pediatric rehabilitation, yet the literature remains dispersed across disciplines, technologies, clinical targets, and delivery settings. This study aimed to map publication and citation performance and the conceptual, intellectual, and social structure of immersive-technology research in pediatric rehabilitation across all settings from 2015 to 2024 and, as a secondary translational perspective, to examine the prominence of home-, remote-, and telerehabilitation-related terminology. Records were retrieved from Scopus, Web of Science, PubMed, and the Lens scholarly search platform and analyzed using bibliometric performance indicators and science-mapping techniques. A total of 388 publications were included. Publication output increased overall, with the United States of America, Italy, and China as the most productive countries. Virtual reality and game-based approaches dominated the field and were frequently associated with cerebral palsy and motor rehabilitation, whereas augmented and mixed reality were less represented. Multiple correspondence analysis explained 39.3% of total inertia across the first two dimensions and distinguished motor- and therapy-oriented, gamified, remote, cognitive, developmental, and emerging-technology orientations within the conceptual structure. Thematic evolution indicated diversification toward cognitive, gamified, and artificial-intelligence-linked applications, although these newer themes remained low in frequency. Author bibliographic-coupling and collaboration analyses identified distinct research communities, including domestically concentrated and internationally diverse groups. Home-, remote-, and telerehabilitation-related terminology appeared in 41 publications (10.6%), indicating that decentralized delivery remains a minority framing within the broader literature. Immersive-technology research in pediatric rehabilitation is expanding and diversifying, but bibliometric prominence should not be interpreted as evidence of clinical effectiveness. Further research should prioritize rigorous clinical evaluation, consistent reporting, standardized outcomes, and setting-specific implementation, including feasibility, usability, safety, and equity in home and community contexts.

Pediatric rehabilitation represents a critical and complex healthcare challenge, addressing a wide range of conditions that affect children’s physical, cognitive, emotional, and social development—increasingly beyond, as well as within traditional clinical environments (Sudhir and Sharath, 2023; Ramey et al., 2025). These include congenital disorders, neurological conditions, musculoskeletal impairments, and acquired injuries such as stroke, cerebral palsy, and developmental delays. The need for effective rehabilitation is especially significant during childhood and adolescence—periods of rapid brain development marked by synaptic pruning, myelination, and cortical reorganization (Ramey et al., 2025). From a clinical standpoint, disruptions during these stages are particularly critical, as they can lead to long-term impairments in motor function, communication, learning, and behavior, with lasting consequences for independence, participation, and life-course opportunities (Shonkoff et al., 2012).

The consequences of pediatric disability extend beyond the individual, affecting families, schools, and healthcare systems (Balbo and Bolano, 2024; Reichman et al., 2008). Children with disabilities or chronic health conditions who require rehabilitation often face challenges in academic achievement, peer relationships, and overall quality of life (Brekke et al., 2023). These difficulties arise through several interacting pathways. Functional impairments and ongoing treatment demands contribute to school absenteeism, reduced cognitive functioning, and lower aspirations, which constrain educational attainment; the effect is most pronounced for neurodevelopmental and neurological conditions (Brekke et al., 2023). Family resources further shape these outcomes: children of more highly educated parents tend to achieve better academic results, although this buffering is not equally protective across all conditions (Brekke et al., 2023). At the family level, the time, financial, physical, and emotional demands of raising a child with a disability place strain on the whole family system and are associated with poorer parental health and wellbeing, particularly for mothers and lower-resourced families (Balbo and Bolano, 2024; Reichman et al., 2008). These cumulative demands, together with restricted participation in school and community life, contribute to lower overall quality of life. Consistent with this, children undergoing rehabilitation are also at increased risk for secondary conditions such as anxiety, depression, and attention-deficit/hyperactivity disorder. Despite its importance, pediatric home rehabilitation remains under-resourced and unevenly accessible, particularly for children with mild or complex needs that may be misattributed to other developmental disorders (Ramey et al., 2025). Traditional clinic-based rehabilitation approaches are often limited by cost, accessibility, and engagement, especially for families in remote or underserved areas. This has led to growing interest in innovative, scalable interventions that can be delivered in home or community settings. In this paper, we use the term “immersive technologies” as a broad category encompassing extended reality (XR) —an umbrella term that itself includes virtual (VR), augmented (AR), and mixed reality (MR). These tools are increasingly explored as promising approaches in pediatric rehabilitation. They create engaging, interactive environments that enhance motivation, adherence, and personalization of therapy, helping overcome barriers in conventional pediatric home rehabilitation (Goldsworthy et al., 2023; Hess et al., 2025).

VR-based interventions have shown potential in improving motor functions, such as balance and gait, and cognitive domains, such as executive functioning and attention (Di Giusto et al., 2023; Shen et al., 2024). Immersive systems may also provide controlled, repeatable, and safe simulations of real-world tasks. However, these characteristics do not by themselves establish that immersive technologies are feasible or accessible for home use. Telehealth research suggests that XR may support presence, acceptability, and care processes, but implementation in home and community settings depends on factors such as device cost, connectivity, physical-space requirements, usability for children and caregivers, workflow integration, clinician and caregiver training, clinical validation, safety, and reimbursement (Ong et al., 2021; Dinh et al., 2023). Previous bibliometric studies have examined immersive technologies in pediatric and rehabilitation contexts. Kacmaz and Kaçmaz (2024) mapped VR and AR research in pediatrics and found that the literature was concentrated in areas such as surgical education or procedures, simulation technologies, and neurological conditions, with rehabilitation also emerging as an important research theme. Zhong et al. (2024) examined the more specific area of virtual reality in pediatric preoperative anxiety. These studies provide valuable insights into publication trends, applications, and geographical patterns across pediatric immersive-technology research. However, their scope was not restricted to rehabilitation or did not comprehensively examine the broader structure of immersive technologies within pediatric rehabilitation.

Related bibliometric studies have also examined virtual reality in rehabilitation. Guo et al. (2022) analyzed VR research in neurorehabilitation and identified cognitive rehabilitation and upper-limb motor rehabilitation as prominent areas, while Fan et al. (2023) mapped the broader development and emerging trends of VR rehabilitation. However, these studies primarily focused on VR and were not specifically designed to provide an integrated account of VR, AR, MR, and XR research across pediatric rehabilitation. To our knowledge, previous studies have not combined a pediatric-rehabilitation-focused, multi-database corpus with complementary analyses of publication and citation performance, conceptual themes, thematic evolution, intellectual linkages, and collaboration structures. A new bibliometric analysis is therefore warranted to clarify how this interdisciplinary field is organized, which technologies and rehabilitation targets predominate, how its themes have evolved, and how its research communities collaborate and draw upon shared knowledge bases.

Meta-analyses and systematic reviews, such as Hao et al. (2024), have examined condition- or intervention-specific evidence, including immersive rehabilitation for children with cerebral palsy. Calcaterra et al. (2025) and Sullivan et al. (2025) have also highlighted emerging interest in telerehabilitation, gamification, AI-supported interventions, and family-centered approaches relevant to home settings. These studies provide important clinical and implementation perspectives but do not replace a bibliometric analysis of the broader conceptual, intellectual, and social structure of immersive-technology research in pediatric rehabilitation.

Additionally, the current evidence base regarding immersive technologies in pediatric rehabilitation remains fragmented. Systematic reviews and meta-analyses report promising findings across several pediatric applications, including balance rehabilitation in cerebral palsy (AbdelMassih et al., 2024; Ceradini et al., 2024; Nordgård and Låg, 2021; Yang Z. et al., 2025; Ziab et al., 2022). Specifically, Ziab et al. (2022) analyzed 12 randomized controlled trials and reported a positive pooled effect of VR-based rehabilitation on static and functional balance in people with cerebral palsy, while emphasizing the limited number of well-conducted trials and the need for larger, more homogeneous samples. Small sample sizes, short intervention periods, heterogeneous protocols, and varied outcome measures therefore continue to limit generalizability and complicate the synthesis of findings. The literature is also heavily skewed toward adult populations, with relatively few pediatric-specific studies exploring the use of immersive technologies. Moreover, across health and medical-education XR research, the evidence is modality-skewed: VR receives the most attention, while AR and MR remain comparatively underexplored—a pattern that may also constrain modality diversity in pediatric rehabilitation research (Barteit et al., 2021). Within this broader field, home-based and remote rehabilitation remain underexplored translational contexts, particularly regarding implementation strategies, design requirements, feasibility, and long-term evaluation.

Accordingly, this study aimed to map publication and citation performance and the conceptual, intellectual, and social structure of immersive-technology research in pediatric rehabilitation from 2015 to 2024. A ten-year period (2015–2024) was considered optimal for identifying contemporary research trends while ensuring sufficient time for publication accumulation and citation maturation. The year 2015 marks the beginning of the rapid expansion of research enabled by commercially available immersive technologies, whereas 2024 was the latest fully indexed year at the time of the search. Excluding 2025 avoids bias arising from incomplete indexing and citation accrual, thereby enhancing the reliability and comparability of temporal bibliometric analyses across the entire study period. The analysis encompassed eligible studies across delivery settings. As a secondary translational perspective, we also examined the prominence of explicit home-, remote-, and telerehabilitation-related terminology within the broader corpus.

The study addressed the following research questions: What are the main publication and citation patterns in immersive-technology research for pediatric rehabilitation from 2015 to 2024?

How is the field conceptually, intellectually, and socially structured, and how have its research themes evolved during this period?

Data sources and search strategy

This study employed bibliometric methods to explore research trends in immersive technologies applied to pediatric rehabilitation. Publications were retrieved from four major scientific databases: Scopus, Web of Science, PubMed, and the Lens scholarly aggregator. These databases were selected because together they offer comprehensive coverage of multidisciplinary, biomedical, and emerging-technology research, making them well suited for a bibliometric analysis of immersive technologies in pediatric rehabilitation; their combined breadth and depth help ensure that the retrieved publications sufficiently represent the global research landscape for this topic. The initial search, covering Scopus, Web of Science, and PubMed, was conducted on 22 May 2025, and an additional search was performed in Lens on 30 May 2025 to broaden coverage across publishers and document types. The publication period was limited to 2015–2024. Only English-language articles were included, as English is the predominant language of indexed international publication in this field and to ensure consistent screening and reliable metadata extraction across the four sources. The 2015–2024 window was selected on both conceptual and methodological grounds. The start point corresponds to the period of widespread commercial availability and maturation of immersive-technology hardware (e.g., consumer virtual- and augmented-reality devices from around 2014–2016), after which pediatric rehabilitation research applying these technologies expanded substantially; earlier work is comparatively sparse and predates this diffusion. The end point was set at 2024 because, at the time the searches were conducted (May 2025), 2024 was the most recent fully indexed publication year. More recent records (2025 onward) were excluded because indexing and citation accrual for these years were still incomplete, which would have produced an artificial decline in the most recent counts and biased the trend and growth analyses. The resulting ten-year window captures a complete and recent phase of the field’s development while ensuring consistent, stable bibliographic coverage across all included years. The search strategy was structured around four conceptual domains: (1) immersive technologies (e.g., virtual, augmented, mixed, and extended reality), (2) rehabilitation and home-based care, (3) design, co-design, interventions, and services, and (4) the target population (children, youth, adolescents, young people). The design/co-design/intervention/services concept block was included to focus the retrieval on applied research involving the development, evaluation, or delivery of immersive therapeutic interventions and rehabilitation-related services, rather than publications that mentioned immersive technology and rehabilitation only conceptually or incidentally. While the core concepts remained consistent, each database required modifications to accommodate its specific query syntax. Supplementary Table S1 details the search strings used for each source.

Eligibility criteria

Inclusion criteria

Included publications were peer-reviewed journal articles and conference/proceedings papers focusing on immersive technologies (VR, AR, MR, or XR) applied to pediatric rehabilitation. Journal articles and conference/proceedings papers were included because the purpose of the study was to map the development and structure of the broader research field rather than to synthesize intervention effectiveness. Accordingly, eligibility was not restricted by clinical study design; randomized and non-randomized trials, observational studies, case studies, feasibility studies, usability evaluations, and technology-development studies could contribute to the bibliometric corpus when they met the topical and population criteria. Conference papers were retained because immersive rehabilitation is an interdisciplinary field spanning rehabilitation, engineering, computer science, and human–computer interaction, where relevant technological developments are frequently disseminated through conference proceedings. Inclusion in the bibliometric analysis should not be interpreted as indicating equivalent methodological quality or clinical evidential strength across publications. Publications were eligible regardless of delivery setting (e.g., clinic, laboratory, school, community, or home) as methodological and technological validation in supervised settings often precedes home deployment. Eligibility was not restricted to any specific diagnosis or clinical condition. Because the aim was to map the development and structure of the immersive-technology research field across pediatric rehabilitation as a whole, a condition-agnostic scope was adopted deliberately, so as to capture the full breadth of clinical targets rather than the literature of any single condition. This decision reflects the mapping objective of the study rather than an attempt to compare or pool findings across conditions. Implications for home-based and remote care were addressed in the interpretation of findings. Included publications focused on children and adolescents (ages 5–18 years), with studies eligible if participants fell within this range or if results for this age group were clearly reported. This range was selected because immersive-technology interventions typically require a level of motor control, cognitive comprehension, and ability to follow task instructions and tolerate head-mounted or screen-based systems that is more consistently established from around school age onward; studies in infants and toddlers (0–4 years) remain comparatively scarce and often rely on fundamentally different intervention paradigms.

Exclusion criteria

Records were excluded if they were: Published outside the 2015–2024 timeframe

Not in English

Unrelated to rehabilitation

Focused on technologies other than VR/AR/MR/XR

Focused on populations other than children/adolescents aged 5–18, or did not report pediatric results for this age group

Surveys or review articles

The time window of 2015–2024 was deliberately chosen to capture a 10-year period of research activity. Although searches were conducted in May 2025, publications from 2025 were excluded to ensure consistency across databases and to allow sufficient time for indexing completeness.

For the purposes of this bibliometric study, the scope of literature relevant to pediatric rehabilitation was operationalized broadly to include both core rehabilitation interventions and rehabilitation-adjacent therapeutic applications involving immersive technologies. Core rehabilitation interventions were those intended to improve, restore, maintain, or support physical, cognitive, communicative, psychological, behavioral, activity, or participation-related functioning in children with health conditions.

Rehabilitation-adjacent applications included therapeutic uses of immersive technology that addressed symptoms, distress, treatment tolerance, assessment, education, or engagement relevant to the wider pediatric rehabilitation and care pathway, even when optimization of longer-term functioning was not the primary outcome. These applications were retained because the purpose of the study was to map the broader interdisciplinary development and therapeutic use of immersive technologies relevant to pediatric rehabilitation rather than to evaluate rehabilitation effectiveness alone.

Procedural-distraction and acute pain-management studies were classified as supportive symptom-management applications. Psychological applications were included when immersive technology was used for therapeutic management of emotional, cognitive, or behavioral concerns. Assessment applications were included when they evaluated functions, symptoms, or capacities relevant to treatment planning, monitoring, or therapeutic decision-making. Educational applications were included when they directly supported pediatric patients, families, treatment participation, or rehabilitation-related care; studies concerned solely with professional education or clinical skills training without a pediatric therapeutic objective were excluded.

Publications focusing on vocational rehabilitation, employment training, drug addiction recovery, or other non-health domains were excluded.

Data collection and preprocessing

The initial search across Scopus, Web of Science, and PubMed yielded a total of 1,401 records (647 from Scopus, 535 from Web of Science, and 219 from PubMed). These records were imported into EndNote X9, where automatic deduplication was performed using the software’s built-in function, which compares fields such as titles, author(s), and publication years. This process resulted in a refined dataset of 817 unique records.

A supplementary search in Lens retrieved 419 additional records. The Lens records were exported and compared against the existing dataset (n = 817) using Microsoft Excel. To support accurate matching, the DOI and Title fields were normalized by converting text to lowercase and removing leading/trailing spaces. Duplicate detection was then performed using Excel functions (e.g., COUNTIF) to flag exact matches on normalized DOI and, where DOI was missing, on normalized Title. Potential near-duplicates (e.g., minor title variations) were resolved by manual inspection. After removing duplicate records already present in the 817-record dataset, 299 records were retained from Lens as unique additional records. When duplicate records were identified across databases, the most complete bibliographic record was retained as the primary record, while citation and cited-reference information were preferentially drawn from the version containing the most complete reference list. Bibliographic metadata were manually harmonized in a spreadsheet to consolidate obvious spelling, formatting, and naming variants across authors, institutions, affiliations, and countries. For example, author variants such as “Levac DE” and “Levac D″ were unified, and institutional and country names were standardized to prevent the same entity from being represented as multiple nodes during network and productivity analyses. A small number of records, primarily from PubMed, contained incomplete cited-reference metadata. Where full text or another reliable source was accessible, missing reference lists were manually retrieved, standardized, and added to the corresponding record. Records for which complete reference lists could not be obtained were retained in the dataset and continued to contribute to descriptive, productivity, thematic, and collaboration analyses; only the available cited-reference information was used in citation-based network analyses. No record was excluded solely because of missing cited-reference metadata. Comparable supplementation was generally not required for Scopus, Web of Science, or Lens records, which contained the cited-reference information needed for the planned analyses.

Screening was conducted independently by two reviewers using a standardized screening form developed for this study directly from the predefined inclusion and exclusion criteria (Section 2.2). The form was pilot-tested on a random subset of 30 records to check consistency of application and to clarify ambiguous criteria before full screening, after which minor wording refinements were made. Any discrepancies or uncertainties were resolved through discussion and consensus.

While inter-rater reliability was not formally quantified using Cohen’s kappa, discrepancies were infrequent and were resolved through discussion and consensus. A total of 388 records were included (n = 728 removed) in the final dataset. The overall screening and selection process is summarized in Figure 1. Bibliometric analyses were performed using the Bibliometrix package ((bibliometrix (RRID:SCR_023744)) in RStudio along with its web-based, point-and-click application Biblioshiny (Aria and Cuccurullo, 2017). The analyses addressed four key dimensions of bibliometric investigation aligned with the study’s research questions.

When key metadata fields required for analysis were missing or incomplete, we applied supplementary procedures to improve record completeness. For example, because PubMed exports may lack full cited-reference (CR) metadata, PubMed records were checked and missing reference lists were retrieved manually from full-text sources (when available) and recorded in a standardized format for downstream analyses.

The combined dataset from all four sources in CSV format was then imported into Biblioshiny, which automatically executed a final deduplication step as part of its data ingestion process, ensuring that all remaining duplicates were removed prior to analysis.

Descriptive and performance analysis

Descriptive and performance analyses quantify the growth, productivity, and citation impact of a research field and identify the main publication venues and contributing actors. This provides a baseline overview of how the literature has expanded over time and where influence and output are concentrated.

Basic bibliometric indicators were computed to describe publication trends over the selected period (2015–2024), including: Annual scientific production

Most prolific and influential authors, journals, and institutions

Country-level output and impact (both total and average citations)

Local impact metrics such as h-index and g-index for authors and sources

Co-occurrence and thematic analysis

Co-occurrence and thematic analyses are used to map the field’s conceptual structure by identifying how topics co-appear and how thematic emphases shift over time. These methods reveal dominant themes, emerging topics, and conceptual connections that are not visible from productivity and citation indicators alone. Prior to these analyses, author keywords were normalized using a manually constructed thesaurus to consolidate synonyms, lexical variants, and equivalent expressions into unified terms. This harmonized singular and plural forms, spelling variants and typographical errors, acronyms and their expansions, and terms denoting the same underlying concept. This normalization reduces fragmentation of conceptually equivalent terms and prevents artificial inflation of distinct keywords in the subsequent frequency, co-occurrence, and mapping analyses. The full thesaurus (variant terms → normalized term) is provided in Supplementary Table S2.

To explore conceptual structures and thematic concentrations within the dataset: Trend topic analysis: Visualized the temporal emergence and decline of key terms using median year and interquartile spread.

The conceptual structure was examined using MCA, implemented through the Conceptual Structure function in Biblioshiny (bibliometrix version 4.3.0) and applied to author keywords after thesaurus-based normalization (Supplementary Table S2). The 50 most frequent author keywords were included; lower-frequency terms outside this set were not entered. Hierarchical clustering with average linkage was then applied to the MCA factor coordinates, and the resulting hierarchy was cut into five clusters. The 50-term configuration was retained because it provided an interpretable representation of the conceptual structure with adequate variance captured by the first two dimensions, while the five-cluster solution yielded coherent and sufficiently distinct thematic groupings.

Thematic evolution and longitudinal mapping: The 2015–2024 observation period was divided using four cutting points (2016, 2018, 2020, and 2022), producing five consecutive and equal two-year intervals: 2015–2016, 2017–2018, 2019–2020, 2021–2022, and 2023–2024. Temporal slicing was used to examine changes in the conceptual structure of the field over time (Aria and Cuccurullo, 2017). The analysis was conducted with the number of words set to 250 and the minimum cluster frequency set to 5. Themes within each period were identified using the Walktrap clustering algorithm. The inclusion index was used to assess thematic continuity between consecutive periods, with the minimum weight-index threshold set to 0.1. Two-year intervals were selected pragmatically to balance temporal resolution with sufficient publication volume for co-word clustering within each period. Annual intervals would have produced sparse networks, particularly during the earlier years of the corpus, whereas longer intervals would have obscured shorter-term thematic changes. The two-year division also produced five equal and directly comparable periods across the ten-year study window. The resulting thematic links were used to examine the persistence, transformation, emergence, or disappearance of themes across consecutive periods.

Intellectual and social structure mapping

Science-mapping of intellectual and social structures examines how knowledge and collaboration are organized by analyzing networks among authors and cited references. This helps identify research communities, bridge authors, and the degree of fragmentation or integration in the field.

To examine the intellectual structure and collaborative dynamics: Author bibliographic-coupling analysis was performed at the author level, with authors linked according to overlap in the references cited by their publications. The author coupling network was normalized using the Salton similarity index and included up to 250 authors. Edges were weighted by coupling strength, isolated nodes were removed, and a minimum edge weight of one was applied. Communities were identified using the Walktrap algorithm implemented in Bibliometrix/Biblioshiny, with the minimum cluster-frequency parameter set to 2. Cluster impact was calculated using normalized global citation scores, and clusters were labelled using dominant author keywords. The resulting clusters were interpreted as descriptive groupings of authors with similar referencing patterns rather than fixed disciplinary boundaries.

Author collaboration networks were constructed from co-authorship data. Edges represented co-authorship relations, were normalized using the association-strength method, and were weighted by co-authorship frequency. Communities were detected using the Walktrap algorithm. The network was limited to a maximum of 300 author nodes, isolated nodes were removed, and a minimum edge weight of one co-authorship was applied. An automatic layout with a repulsion force of 0.5 was used for visualization, and papers with more than 20 authors were retained. Betweenness centrality was used to identify authors occupying bridging positions between otherwise distant groups. The resulting collaboration communities were interpreted as descriptive science-mapping structures under the selected parameter settings rather than definitive partitions of the field.

Interpretation of findings (sensemaking)

In line with calls for bibliometric studies to move beyond descriptive mapping, we adopted a structured sensemaking approach to interpret the outputs. Specifically, we (1) scanned the field using science-mapping outputs (e.g., trend topics, MCA dimensions, thematic evolution, and network structures), (2) sensed by developing integrative explanations that connect these patterns to clinical targets and delivery orientations relevant to pediatric rehabilitation, and (3) substantiated interpretations by explicitly linking each major implication in the Discussion to one or more reported bibliometric results (e.g., driver keywords, cluster labels, longitudinal linkages, or source growth patterns) (Öztürk et al., 2024; Lim & Kumar, 2024).

Descriptive and performance analysis

Annual scientific production

From 2015 to 2024, publications on XR (VR/AR/MR) in pediatric rehabilitation increased overall, with year-to-year fluctuations (Figure 2).

Sources and outlets

The dataset comprised 388 publications indexed in 220 source journals. Ranked by productivity and partitioned into three Bradford zones with approximately equal article counts, Zone 1 (core) contained 20 journals contributing 129 articles (≈33% of all publications and 9% of sources), Zone 2 comprised 65 journals with 131 articles (≈34%), and Zone 3 included the remaining 135 journals with 128 articles (≈33%). The distribution of journals by zone thus follows 20: 65: 135 (relative to the core 1: 3.25: 6.75) consistent with Bradford’s law with a compact core and increasing dispersion across Zones 2–3. The most prolific titles in Zone 1 were Games for Health Journal (16 articles), Journal of NeuroEngineering and Rehabilitation (14), Children (Basel) (10), Journal of Pediatric Rehabilitation Medicine (10), and Disability and Rehabilitation (9).

To complement the volume-based analyses, we assessed the local citation impact of journals in our corpus. Table 1 reports each source’s h-index and g-index computed on the local dataset; the h-index reflects consistent citation performance across a journal’s articles, whereas the g-index weights the influence of highly cited papers. In this set, the Journal of NeuroEngineering and Rehabilitation shows the highest h-index (9), while Games for Health Journal has the highest g-index (15); Journal of Pediatric Rehabilitation Medicine (h = 6, g = 10) and Children (Basel) (h = 5, g = 8) also rank prominently. As of May 2025, highly cited publications included Dumoulin et al. (2019)—a rehabilitation-adjacent pediatric emergency-department trial evaluating VR distraction during venipuncture and intravenous placement—cited 118 times, and Bortone et al. (2020)—a functioning-oriented rehabilitation study testing an immersive VR system with wearable haptics for pediatric upper-limb rehabilitation—cited 52 times. These publications represent different application categories within the corpus and should not be interpreted as clinically equivalent forms of rehabilitation.

Developmental Medicine & Child Neurology was the most locally cited source, receiving 559 citations. It was followed by Research in Developmental Disabilities (268) and Physical Therapy (242), both of which contributed substantial foundational literature. Other notably cited journals included the Journal of NeuroEngineering and Rehabilitation (186), Journal of Autism and Developmental Disorders (178), and Gait & Posture (150). Developmental Neurorehabilitation (146), Disability and Rehabilitation (141), Archives of Physical Medicine and Rehabilitation (140), and Pediatric Physical Therapy (131) also featured prominently.

Authors, countries and institutions

The activity view (Figure 3A) shows publication timing for these authors, with some exhibiting steady output across years and others showing bursts concentrated in specific periods. In this panel, bubble size is proportional to the number of articles published in a given year, while color intensity reflects total citations per year for those documents. The fractionalized frequency adjusts for co-authorship by dividing credit among coauthors, providing a fairer estimate of individual contribution within multi-author teams (Figure 3B). On this basis, Levac DE ranks highest (3.3), indicating sustained contribution across multiple papers. Several others—Wright FV, Biffi E, Monteiro CBDM—also show notable fractional output. Together, the two panels summarize who contributes most after adjusting for team size and when those contributions occurred.

Table 2 contrasts the consistency of citations across an author’s body of work (h-index) with the influence of their most highly cited papers (g-index). Most authors have similar h-index and g-index values; however, a few—such as Levac DE, Biffi E, and Monteiro CBDM—display a notably higher g-index compared to their h-index.

In Supplementary Figure S1, the top ten universities and research institutes that contribute most significantly to publication activity were highlighted. The time-series trend shows some institutions (e.g., University of Toronto, Stanford University) maintained steady leadership throughout, while others demonstrated more recent surges (Bloorview, IRCCS, Northeastern University).

Supplementary Figure S1A shows the institutions with the highest number of publications, led by the University of Toronto (55 articles), Stanford University School of Medicine (52), and the University of Washington (36). Supplementary Figure S1B illustrates the longitudinal trends from 2015 to 2024 of these leading institutions displaying an upward trajectory.

The country production analysis reveals that scientific output in the field is geographically diverse but unevenly distributed, with North America, Europe, and East Asia emerging as dominant hubs (Figure 4). Panel (A) of Figure 4 shows a world map illustrating the distribution of total publications across all active regions, while panel (B) presents a time-series line graph tracking publication output of major countries from 2015 to 2024. The United States of America (79 articles), Italy (44), and China (36) emerged as the most productive contributors. In Table 3, the left side column ranks countries by total citations, highlighting the dominance of high-output nations such as the United States of America, Italy, and Canada. The right-side column ranks countries by average article citations, where smaller producers like Germany, Ghana, and the Netherlands achieve higher per-article influence despite lower publication counts across the four sources. Average-citation rankings are sensitive to small denominators; therefore, smaller producers may rank highly due to one or a few highly cited papers.

Co-occurrence and thematic analysis

Trend topics and keyword frequency

The author keywords highlighted in pink boxes (Figure 5) show considerably lower overall frequency and are absent from the topmost frequent keywords (Table 4), yet they have demonstrated a notable increase in usage over time relative to their earlier presence. In this diagram, each horizontal line spans the first (Q1) to third quartile (Q3) years of occurrence for a given keyword, while the circle marks its median year of prominence (50th percentile). Circle size reflects the frequency of term use. Established themes such as virtual reality (Freq = 200, MedianYr = 2022), rehabilitation (Freq = 74, MedianYr = 2022), and cerebral palsy (Freq = 84, MedianYr = 2021) reflect sustained and long-standing research attention. By contrast, newer topics such as gamification (Freq = 5, MedianYr = 2024), extended reality (Freq = 5, MedianYr = 2024), and artificial intelligence (Freq = 5, MedianYr = 2023) have emerged more recently (2023–2024).

A total of 1,023 unique author keywords were identified in the dataset. Virtual reality was the most frequently used keyword, occurring 200 times (Table 4). Other related terms, such as augmented reality (28 occurrences), extended reality, and mixed reality, appeared with lower frequency.

Related terms that were part of the search string such as extended reality, and mixed reality appeared less and at the latter half in the data corpus while augmented reality (frequency = 28) although being part of the most frequent words, shows sporadic usage with no clear, consistent upward trend until the very last year (Table 5).

Frequently used keywords also included rehabilitation (74), cerebral palsy (84), children (53), serious games (24), and motor skills (10). In addition to cerebral palsy, more clinical terms such as autism (27) and hemiplegia (10) appeared within the top-ranked terms. Terms related to body functions like balance (16), executive function (12), upper limb (18), and gait (9) were also represented.

Terms related to behavioral and psychological outcomes like motivation (7), engagement (7), and anxiety (7) emerged but with relatively limited occurrences during the study period.

Factorial analysis

The factorial analysis revealed that the first two dimensions accounted for 39.3% of the total inertia (Dim 1 = 21.98%, Dim 2 = 17.34%), providing a moderate two-dimensional representation of the conceptual structure (Figure 6). In this Multiple Correspondence Analysis (MCA) map, keywords are plotted according to their coordinates on the first two dimensions. Terms located farther from the origin (0,0)—such as telerehabilitation, hemiplegia, and immersive virtual reality—act as drivers that pull the axes and clusters toward distinct thematic orientations, thereby defining the conceptual boundaries of the field. In contrast, terms located near the origin, including rehabilitation and children, are non-discriminatory: while foundational to the field, their ubiquity across the dataset limits their ability to differentiate thematic structures. MCA of author keywords yielded two dimensions with the following axis extremes: on Dim 1, negative extremes were telerehabilitation (−2.68), feedback (−2.40), exergaming (−2.33), hemiplegia (−2.30), therapy (−1.93), gamification (−1.73), and active video games (−1.66), while positive extremes were traumatic brain injury (+2.21), executive function (+1.63), immersive virtual reality (+1.41), training (+1.38), pediatric rehabilitation (+1.25), pediatrics (+1.06), and autism (+0.87). On Dim 2, negative extremes were physical therapy (−2.02), active video games (−1.83), intervention (−1.33), pediatrics (−1.24), traumatic brain injury (−1.18), and executive function (−0.93), whereas positive extremes were telerehabilitation (+2.65), motivation (+2.03), immersive virtual reality (+1.86), augmented reality (+1.67), feedback (+1.57), neurorehabilitation (+1.54), extended reality (+1.32), and artificial intelligence (+1.27). These extremes indicate that Dim 1 ranges from motor/therapy–gamified and remote topics to cognitive/developmental and immersivity topics, and Dim 2 ranges from applied/clinic–task terms to remote/engagement–advanced-technology terms.

Cluster analysis identified five distinct thematic groups: Core XR rehabilitation (rehabilitation, cerebral palsy, autism, children, gait, serious games and more)

Immersive technologies (augmented reality, neurorehabilitation, extended reality, immersive virtual reality)

Cognitive and developmental rehabilitation (executive function, pediatrics, attention, physical therapy, traumatic brain injury)

Gamified and device-based therapy (hemiplegia, active video games, Leap Motion, therapy, gamification, intervention)

Remote XR therapy (telerehabilitation, feedback, exergaming, motor learning, motivation)

Notably, representative documents were positioned mainly in Clusters 1 and 4, reflecting anchor works that embody these thematic orientations. Other clusters were defined primarily by keywords with non-discriminatory positioning near the factorial origin and therefore did not yield representative documents. Within Cluster 1 (Core XR rehabilitation), highly cited works include Dumoulin et al. (2019), which examined the use of virtual reality as a distraction tool during needle-related procedures in children, and Bortone et al. (2020), which investigated wearable haptic feedback combined with immersive virtual reality for exercise interventions in children with neuromotor impairments. In Cluster 4 (Gamified and device-based therapy), representative works include Kassee et al. (2017), which evaluated the home-based use of Nintendo Wii for the rehabilitation of children with hemiplegic cerebral palsy, and Chan-Víquez et al. (2023), which reported on the use of the video game Bootle Blast for home-based rehabilitation of hemiplegic children.

Thematic evolution and longitudinal mapping

Using co-word clustering and the inclusion index, the corpus was divided into five time slices (2015–2016, 2017–2018, 2019–2020, 2021–2022, 2023–2024) (Figure 7). In this thematic evolution map, rectangles represent thematic clusters identified through keyword co-occurrence analysis. Arrows indicate thematic continuity, drawn when clusters in consecutive periods share a significant proportion of keywords (measured by the inclusion index). The thickness of the arrows reflects the degree of continuity—stable themes, such as virtual reality and autism, appear across several slices with thicker connecting arrows, whereas some clusters like artificial intelligence and extended reality emerge later as new research fronts with thinner or no incoming arrows. 2015–2016: Topics included virtual reality and augmented reality with applications in cerebral palsy, upper limb, and gaming contexts.

2017–2018: Autism appeared as a new theme; clusters continued around virtual reality (e.g., upper limb, children). Augmented reality remained present.

2019–2020: Themes expanded to motor skills, balance, and executive function. Autism persisted. Augmented reality continued as a stable theme in the early slices.

2021–2022: Additional themes emerged, including artificial intelligence (linked to immersive virtual reality), exergaming, engagement, haptic feedback, muscle strength, pediatric rehabilitation, neurorehabilitation, and participatory design. Serious games continued alongside virtual reality. Augmented reality remained present with emphasis on upper limb.

2023–2024: Augmented reality persisted with links to social and communication skills. Extended reality appeared with links to IQ, memory, and physical activity. Exergaming remained present. Virtual reality continued as a central node with multiple linkages.

Intellectual and social structure

Author bibliographic coupling

Author bibliographic coupling analysis—in which authors are linked by the overlap in their cited references—was used to identify communities of researchers with similar referencing patterns. Communities were detected using the Walktrap algorithm, yielding 15 author clusters that together encompassed 88 papers. In Figure 8, the clusters are positioned according to network centrality on the x-axis (linkage to the broader author network) and mean normalized global citation score on the y-axis (citation impact; impact measure = global). Bubble colour indicates cluster membership, while bubble size reflects relative cluster size. Cluster labels were derived from dominant author keywords, with the accompanying confidence values indicating the strength of association between each keyword label and its cluster.

The map shows variation in the structural position and thematic orientation of the author communities. More central clusters were associated with terms such as rehabilitation, cerebral palsy, and balance, whereas more peripheral clusters included themes such as exergaming and telerehabilitation. These clusters should be interpreted as descriptive groupings of authors with similar referencing patterns rather than fixed disciplinary or thematic boundaries.

Game-mediated intervention terminology—including serious games, games, active video games, exergames, exergaming, and gamification —was identified in Clusters 2, 3, 5, 9, and 12. Across these clusters, 36 unique documents contained at least one game-related term:

Cluster 2 (n = 16). Papers describe motor/upper-limb rehabilitation tasks, performance/feedback, and physiotherapy-style training protocols using game-based interactions.

Cluster 3 (n = 2). Papers focus on serious-games platform implementations and home-based usage/feasibility for pediatric rehabilitation.

Cluster 5 (n = 12). Papers center on cerebral palsy and upper-limb/bimanual rehabilitation (e.g., HABIT-ILE), often using active video games/exergames and reporting usability/engagement alongside clinical outcomes.

Cluster 9 (n = 1). Paper reports at-home bimanual intensive training in virtual reality targeting hand/arm function.

Cluster 12 (n = 5). Papers describe wearable haptics/assistive technologies integrated with immersive/game-mediated tasks, with emphasis on motor control/kinematics and system prototypes.

Within these 36 documents, n = 10 were associated with home-based/remote/telerehabilitation terminology in the title, abstract, or keywords (by cluster: C2 = 3, C3 = 1, C5 = 5, C9 = 1, C12 = 0). To contextualize these cluster-level subsets and quantify delivery-setting terminology in the overall dataset, we screened the full corpus (n = 388) for home- and remote-mediated terms in the Title (TI), Abstract (AB), Author Keywords (DE), and Indexed Keywords (ID). Across the full corpus, explicit home/remote delivery terminology (e.g., home-based, telerehabilitation, remote, telemedicine, telehealth) appeared in n = 41 papers (10.6%) in the title, abstract, or keywords, indicating that such framing is present but not dominant. In the author bibliographic coupling subset (n = 88), n = 15 papers (17.0%) included home/remote delivery terminology, suggesting a higher concentration of remote-oriented framing within the more densely connected author communities captured by the coupling map. These 15 papers were distributed across clusters as follows: C2 = 5, C3 = 3, C5 = 6, and C9 = 1. Notably, these terms appeared not only in author keywords but also in titles, abstracts, and indexed keywords, reflecting varying degrees of emphasis across studies.

Author collaboration network

Using the co-authorship network settings described in Section 2.6, the analysis identified 47 collaboration communities among the authors retained in the 300-node network (Figure 9). The five largest communities in this view were Cluster 3 (18 authors), Cluster 12 (17), Cluster 1 (16), Cluster 5 (16), and Cluster 11 (15). At the network level, the authors with the highest betweenness centrality—indicating a bridging role between otherwise distant groups—were Monteiro CBDM, Biffi E, De Domenico C, De Araujo LV, Strazzer S, and Levac DE.

The analysis revealed that Clusters 3, 5, and 11 exhibited complete country-level homophily (Italy; share = 1.00; unique countries = 1), indicating fully domestic collaboration. In contrast, Cluster 12 demonstrated moderate diversity (Brazil; 0.737; 5), while Cluster 1 presented the most heterogeneous composition (Belgium; 0.273; 10). At the institutional level, Clusters 5 and 11 showed total affiliation homogeneity, each comprising authors from a single institution (IRCCS Centro Neurolesi Bonino Pulejo and IRCCS Fondazione Don Carlo Gnocchi, respectively). Cluster 12 had a notable concentration around the University of São Paulo (0.407; 17), followed by Cluster 3 (IRCCS Eugenio Medea; 0.583; 8). Cluster 1 was the most institutionally diverse (Université Catholique de Louvain; 0.192; 18).

Study content signals by cluster (based on titles and abstracts). To descriptively characterize the thematic focus of each author community, we aggregated and analyzed key terms from titles and abstracts within the top five clusters. Cluster 3. Clinical pediatric neurorehabilitation with cerebral palsy, upper-limb/bimanual training, frequent active video games/serious games, and clinic-to-home feasibility/usability reporting; outcomes often include motor performance, balance, and functional measures.

Cluster 1. Technology-enabled motor learning and upper-limb rehabilitation with task-oriented protocols, kinematic/clinical outcomes, and VR/AR components integrated with established therapy frameworks.

Cluster 5. Cerebral palsy and upper-limb/bimanual rehabilitation (e.g., HABIT-ILE) blending game-mediated tasks with physiotherapy routines; reports performance, engagement/usability, and functional outcomes.

Cluster 11. Mix of clinical implementations and task/protocol development (balance/mobility), instrumented assessments, and serious/exergame components; includes multi-session interventions.

Cluster 12. Instrumented systems—wearable/haptic/assistive integrations with immersive/game-mediated tasks—and studies capturing motor control/kinematics; spans prototype through pilot/controlled evaluations.

This bibliometric study synthesizes the evolving landscape of immersive technology (VR/AR/MR/XR) research in pediatric rehabilitation and discusses implications for home-based and remote delivery. We interpret the maps by explicitly linking each major implication to reported bibliometric outputs (e.g., thematic evolution linkages, MCA dimensions and clusters, and source growth patterns). It should be emphasized that bibliometric analyses describe patterns of scientific activity, thematic evolution, and scholarly collaboration, rather than providing direct evidence regarding the clinical effectiveness, quality, or superiority of immersive rehabilitation interventions.

Current trends, research clusters, and thematic directions

By analyzing publication trends, thematic structures, author networks and conceptual dynamics, the findings reveal a field that is expanding in volume while maturing in complexity and clinical relevance. This discussion elaborates on the key observations, their implications for the field, and actionable directions for future research and practice.

The Bradford analysis in our study indicates that publication activity in this domain is organized around a relatively compact set of core journals supported by an extensive multidisciplinary periphery. This distribution is characteristic of emerging interdisciplinary fields, where established clinical and rehabilitation outlets coexist alongside a wide range of technology-oriented and domain-specific venues. The observed patterns of publication growth further reinforce this structure: long-standing clinical and rehabilitation journals continue to serve as anchoring sources of scholarly output, while technologically focused and general-science journals show accelerated expansion, reflecting increasing interest in immersive and digitally mediated rehabilitation approaches. Citation indicators similarly highlight the coexistence of stable, high-impact clinical sources and rapidly evolving technology-driven channels. Overall, analysis of the citation and publication structure suggests that research on immersive technologies in pediatric rehabilitation draws simultaneously on mature clinical traditions and fast-moving technological innovation, underscoring the need to monitor both the established core and the expanding periphery to capture developments across the field (Chen et al., 2022; Poglitsch et al., 2024). From a clinical perspective, making this distinction explicit is useful for assessment and goal-setting, as it clarifies whether an XR intervention is primarily targeting motor function, cognitive and behavioral regulation, or a combination of both, and helps teams match specific platforms and tasks to individual profiles, priorities, and stages of rehabilitation. Previous bibliometric reviews on VR/AR in pediatric and healthcare contexts report patterns that closely mirror our findings. Similar to our Bradford analysis, these studies describe a small group of core clinical or pediatric journals supported by a wide multidisciplinary periphery of technology-, education-, and engineering-focused outlets, reflecting the inherently cross-sector nature of immersive technologies in rehabilitation. For example, Kacmaz and Kaçmaz (2024) found that VR/AR pediatric research is dispersed across numerous medical and technological journals.

Authorship is active but unevenly distributed: fractionalized counts highlight a small core of repeat contributors (e.g., Levac DE), while overall influence also reflects standout highly cited works (elevated g-index relative to h-index for several authors). Institutionally, leading centers (e.g., University of Toronto; Stanford University School of Medicine) sustain output with noticeable acceleration after 2020 among several sites (e.g., Bloorview, IRCCS, Northeastern University). Geographically, production is diverse but concentrated in North America, Europe, and East Asia, with the United States of America, Italy, and China as major producers; citation perspectives show volume-driven impact for high-output countries alongside high per-article influence from smaller producers (e.g., Germany, Ghana, Netherlands). Importantly, given the condition-agnostic scope of the search, the prominence of specific authors, institutions, and sources largely mirrors dominant XR application areas in this corpus (e.g., cerebral palsy/upper-limb rehabilitation) and should be interpreted as a reflection of research activity patterns rather than a ranked statement of clinical priorities. Together, these patterns suggest that the field’s evidence base is propelled by a mix of sustained institutional pipelines and selective, high-leverage contributions—reinforcing the value of cross-site collaboration and international partnerships for generalizability.

The evolution of immersive technologies, particularly VR, AR, and MR, reflects their increasing integration into clinically oriented applications. Earlier research frequently examined feasibility and preliminary therapeutic outcomes through relatively small exploratory studies. For example, Gagliardi et al. (2018) evaluated immersive VR-based gait training in children with cerebral palsy, illustrating the use of immersive technology in functioning-oriented pediatric rehabilitation. However, the early evidence was constrained by small samples and limited clinical validation. Post-2020, the field registered a significant surge in scholarly output, coinciding with global shifts in healthcare delivery associated with the COVID-19 pandemic (Wosik et al., 2020; Koonin et al., 2020).

The COVID pandemic, for example, accelerated interest in remote and home-oriented rehabilitation and increased research attention to technology-supported delivery (Pallavicini et al., 2022). This shift reflects the growing emphasis on decentralizing healthcare delivery to improve accessibility, particularly for pediatric populations with limited access to in-clinic care.

Emerging terms such as extended reality, artificial intelligence, and gamification appeared within the rehabilitation literature during 2023–2024, pointing to a possible shift toward more integrated and technologically sophisticated therapeutic approaches. However, these terms appear with relatively low frequency, and they should therefore be interpreted as emerging signals of future research activity rather than as mature or dominant areas within the current literature. Their recent appearance nonetheless reflects a growing interest in leveraging advanced digital tools to enhance therapeutic outcomes, particularly in pediatric and neurodevelopmental contexts.

For instance, AI-driven XR platforms are increasingly being explored for their capacity to deliver adaptive, real-time feedback during rehabilitation sessions. These systems can dynamically adjust therapeutic tasks based on patient performance, with the potential to improve both the precision and adaptability of interventions. Although not specific to pediatric rehabilitation, adjacent rehabilitation research illustrates how AI-enabled decision support systems can personalize rehabilitation routines for stroke survivors, offering therapists automated yet explainable suggestions for modifying treatment plans. Similarly, XR environments that incorporate Bayesian network models have been shown to predict learning and transfer outcomes in sensorimotor training, highlighting the potential of XR to support long-term skill retention. The integration of gamification into these platforms further enhances engagement and motivation, especially among children with disabilities. Gamified XR environments have been found to support personalized learning, social-emotional development, and cognitive rehabilitation, particularly when combined with AI and user-centered design principles (Power et al., 2024; Pimentel-Ponce et al., 2024). For example, collaborative gamification strategies tailored for children with autism have shown promising results in improving learning outcomes and reducing undesired behaviors. This apparent evolution from standalone technologies toward multi-dimensional, hybrid therapeutic modalities may suggest a growing clinical interest in XR and related technologies. It may reflect a broader interest in personalized, inclusive, and data-driven rehabilitation, in which AI, XR, and gamification could converge to support immersive and adaptive therapeutic approaches. Motor-oriented rehabilitation remains a foundational pillar, as evidenced by the consistent presence of terms such as “cerebral palsy,” “hemiplegia,” and “motor skills” across all time periods. These terms underscore the importance of XR in addressing physical impairments and improving motor function in children with neurological conditions (de Oliveira et al., 2021).

Our analysis also introduces a gradient that reflects differences in how rehabilitation approaches are implemented. This gradient aligns with the extremes of the MCA map (Dim 1/Dim 2), where on one end, terms related to structured clinical practice—such as physical therapy, intervention, and pediatrics—suggest more supervised, task-oriented applications typically delivered in controlled settings. On the other end, terms like telerehabilitation, motivation, augmented reality, extended reality, and artificial intelligence indicate a shift toward decentralized, technology-supported models that emphasize personalization, engagement, and flexible delivery. Taken together, the two dimensions illustrate that the field is evolving along both clinical and organizational lines: not only in the types of patient outcomes targeted (e.g., motor vs. cognitive or developmental), but also in the modes through which rehabilitation is delivered (e.g., clinic-based vs. remote and technology-enhanced). These patterns align with broader trends in digital health (Maggio et al., 2024; Yang X. et al., 2025), where remote modalities are increasingly used to extend access and continuity of care, and gamified or interactive environments are leveraged to enhance engagement and adherence. The conceptual analysis in our study also aligns broadly with the thematic groups emerging from the cluster results. A central rehabilitation-focused theme anchors the field, while a separate cluster captures work on immersive and emerging XR modalities. Additional groups reflect cognitive and developmental applications (Di Giusto et al., 2023), technology-supported motor training, and remote or engagement-oriented rehabilitation models. Together, these themes indicate that the field is diversifying across both clinical focus and delivery approaches. The thematic evolution map further shows that while VR and AR remain consistent core technologies, newer XR modalities are gradually taking on more specialized roles, extending into cognitive, social, and motor domains. Overall, the expanding thematic and modality landscape suggests a field moving beyond early single-technology exploration toward a broader, more integrated XR ecosystem within pediatric rehabilitation (Pimentel-Ponce et al., 2024).

The analysis of author networks and collaboration patterns revealed a field organized into distinct collaboration communities. The author network is fragmented into modular communities with limited inter-community ties in the plotted subset. Central figures such as Monteiro CBDM, Biffi E, De Domenico C, De Araujo LV, Strazzer S, and Levac DE act as intellectual hubs—high-betweenness “brokers” who bridge thematic clusters and facilitate knowledge diffusion across the field. This configuration carries strengths and risks: while central thought leaders can catalyze innovation and dissemination, reliance on a small set of connectors underscores the need for broader collaboration and capacity building (Xie et al., 2025; Gui et al., 2019). In particular, country- and institution-concentrated communities—e.g., C5 and C11 (Italy-dominated) and C3 (high Italian share)—suggest homophily that may streamline local pipelines but risks insular toolchains and limited external validity, whereas more heterogeneous communities spanning multiple institutions/countries—e.g., C1 (broadest country mix) and C12 (Brazil-anchored but multi-country)—resemble multi-site consortia better positioned for generalizable evidence. Clusters differ in orientation—some are technology-prototyping–oriented (e.g., device integration, haptics, kinematic capture, early pilots), while others are clinical-protocol–oriented (e.g., embedded protocols, standardized outcomes, longitudinal follow-up). Bridging these orientations could help move concepts from prototype to clinical embedding. Expanding interdisciplinary partnerships and fostering contributions from underrepresented regions and institutions could diversify perspectives and accelerate progress in XR-based pediatric rehabilitation.

These patterns are consistent with broader shifts reported in pediatric rehabilitation and digital health toward home-based and hybrid models, increased family involvement, and technology-mediated engagement. Clinically, we see a similar shift toward home-based and hybrid models of care, stronger expectations of family involvement, and a growing reliance on engaging, technology-mediated activities to sustain participation over time. Ensuring that XR research continues to track these practice-level trends is essential so that emerging interventions address concrete needs in children’s homes, schools, and community settings rather than remaining confined to experimental or highly resourced environments.

Design insights and future directions

The bibliometric patterns identified in this study suggest several design and implementation implications for advancing immersive technologies in pediatric rehabilitation, particularly with relevance to home-based and remote care. The following implementation considerations should be interpreted as informed perspectives derived from the observed publication patterns and the broader rehabilitation literature, rather than as direct empirical conclusions supported by the bibliometric analyses themselves. First, the continued prominence of gamification-related terms and clusters underscores the importance of engagement as a persistent design priority in pediatric therapy (Pimentel-Ponce et al., 2024; Iosa et al., 2022). In our analysis, this interpretation is reinforced by both the conceptual (MCA) patterns and the clustering results, which consistently position gamification-related terminology among the distinguishing features of the field and reveal a thematic group centered on technology-supported, game-based rehabilitation approaches. Children often require stimulating and interactive environments to sustain participation in repetitive, task-specific training. In this context, gamified interventions—ranging from commercial gaming consoles to serious games and custom therapeutic platforms—illustrate how playful mechanics can support adherence and participation. For example, programs such as HABIT-ILE and Bootle Blast have been used to motivate children with cerebral palsy and hemiplegia to adhere to therapy, highlighting the potential role of engagement-oriented design in supporting therapy routines.

Second, the increasing prominence of remote-oriented themes—such as telerehabilitation and exergaming—points to growing interest in hybrid and decentralized delivery models that integrate physical training with motivational and cognitive components. This trend is reflected in the conceptual structure, where terms like telerehabilitation, feedback, and exergaming occupy peripheral positions, and in the clustering results, which reveal a coherent thematic grouping focused on remote and engagement-based XR approaches. These developments suggest a broader shift toward home- and community-based rehabilitation solutions, accompanied by considerations related to user support, oversight, and system reliability. At the same time, the appearance of artificial-intelligence-related terminology indicates an emerging line of work exploring adaptive and personalized XR environments, although this remains an early trajectory rather than a well-established standard in the field (Kaelin et al., 2021; Tabassum et al., 2025; Halkiopoulos and Gkintoni, 2025). This interpretation is further aligned with the longitudinal mapping, where “artificial intelligence” appears as a later-emerging theme and is positioned in MCA alongside technology- and engagement-associated terms (e.g., extended reality/augmented reality and motivation), consistent with a shift toward technology-enabled personalization.

Third, interdisciplinary collaboration is becoming increasingly essential for the development of effective XR interventions. Journals such as Games for Health Journal and Scientific Reports, which have seen a marked rise in publication output over the study period, reflect the field’s growing appeal to researchers across technology, healthcare, and education (Ziker et al., 2021). Such collaborations are vital for addressing the complex, multifaceted challenges of designing XR systems that are not only clinically effective, but also user-friendly, scalable, and ethically sound. This is consistent with the Bradford core–periphery structure, in which clinically oriented outlets anchor the core of the field while technology-facing venues and proceedings channels contribute a growing share of recent output. Strengthening interdisciplinary collaboration is therefore likely to accelerate translation from prototype development to clinically embedded interventions.

At the same time, the feasibility of XR in real-world pediatric rehabilitation settings remains an ongoing consideration. While the bibliometric patterns highlight increasing attention to engagement and remote delivery, feasibility constraints are practice-level considerations that are not fully captured by publication metadata alone. In clinical practice, many children require active support from therapists or caregivers to initiate and sustain XR sessions, maintain motivation and attention, and regulate emotions when tasks are challenging or frustrating (Alrashidi et al., 2024). Implementing XR at home or in community contexts also entails practical demands related to hardware cost, equipment management, safe space and connectivity, caregiver burden, and the time and competence required from staff and families to deliver an adequate therapeutic “dose.” Designing XR systems that explicitly account for these constraints, with features such as simplified device setup, remote monitoring, built-in caregiver guidance, and realistic expectations about supervision and intensity, will be crucial for scalable and equitable deployment beyond well-resourced clinics.

Limitations

Although the study benefits from drawing on multiple databases, including Scopus, Web of Science, PubMed, and Lens, it is important to acknowledge some methodological and practical limitations that shaped the analysis. First, the coverage of each database is uneven: while Scopus and Web of Science are stronger in international journals, PubMed has a biomedical focus and often lacks full citation records, and Lens provides broad coverage but with variable metadata quality. These differences, together with our restriction to English-language publications, meant that certain fields and regions were better represented than others—particularly English-language and Western-published works—and may have introduced language bias, under-representing relevant work published in other languages or less-visible venues. Second, as explained in the Methods, the condition-agnostic scope of the search was an intentional design choice rather than a limitation of retrieval: it enables a field-level map across clinical targets but does not provide condition-specific depth for any single diagnosis.

Third, the mandatory design/co-design/intervention/services concept block increased search specificity but may have reduced sensitivity by excluding relevant publications that described their applications using alternative terminology without explicitly using design, co-design, intervention, or services in the searched fields.

The process of merging datasets also introduced challenges. Metadata across platforms is not standardized; journal titles, author names, and cited references frequently require harmonization, and despite careful cleaning, some inconsistencies and mismatches are inevitable. Deduplication posed another difficulty: even with DOI- and title-based matching, a small risk of residual duplicates or occasional misclassification remains. This could subtly affect citation networks and trend detection. There might also be temporal variations that further complicate the dataset, as databases differ in how quickly they index new publications, meaning that the most recent research was not consistently captured across sources. Moreover, because bibliometric databases are continuously updated, the results presented here represent a snapshot in time and may shift as coverage expands, and citation counts evolve. Together, these limitations highlight the interpretive caution required when synthesizing bibliometric data from multiple sources and query fields, while also underscoring the need for transparency in documenting the decisions made during search, screening, cleaning, and integration.

Finally, because this study relies on publication-level metadata, keywords, and abstracts, it does not fully represent several clinically important feasibility issues that arise when XR is used with children in routine care. These include the extent to which sessions depend on therapist or caregiver supervision, the support needed for sustaining motivation, attention, and emotional regulation, and organizational barriers such as time, staffing, and competence for delivering and monitoring technology-based rehabilitation. As a result, the prominence of XR in the literature should not be interpreted as evidence that these interventions are straightforward to implement or maintain in all pediatric services, particularly in resource-constrained or community-based settings.

Conclusion

The findings of this bibliometric analysis indicate that XR-based pediatric rehabilitation is a rapidly evolving and increasingly diverse research area, with expanding modalities, application themes, and clinical targets. Alongside this growth, the evidence base remains uneven, highlighting the need for more rigorous evaluation, improved reporting consistency, and greater standardization of outcomes and study designs to support meaningful comparison and translation.

Future progress will depend on interdisciplinary collaboration among researchers, clinicians, designers, and technology developers, with attention to feasibility and implementation across settings—including home-based rehabilitation—where issues such as usability, safety, data quality, and sustained engagement become critical. Equally important is an explicit focus on equity and accessibility, ensuring that XR interventions are adaptable to varied contexts and do not widen existing gaps in care. Overall, the trends and gaps identified in this study provide a practical reference point to align future work toward robust evidence generation, implementation readiness, and inclusive deployment in pediatric rehabilitation.