Selection criteria and eligibility process: The Population, Intervention, Comparison, Outcome (PICO) framework for healthy children aged between 5 and 17 years is provided in [Table/Fig-1].

To be included in the review, each study’s title or abstract had to contain at least one term from the search strategy. The terms included in the search strategy were “multidirectional reach tests,” “balance measurement among children,” “balance abilities among children,” “PRTs,” and “tools used for balance measurement.” Five electronic databases were searched for the review: PubMed/MedLine, SCOPUS, OVID-SP, Cochrane Library, and Physiotherapy Evidence Database. Additionally, Google Scholar was explored to gather supplementary information, and manual searches were conducted, including reviewing the references in identified publications.

Inclusion criteria: The inclusion criteria were as follows: (1) peer-reviewed research articles focusing on determining psychometric properties; (2) published from inception to November 2023; (3) focusing on children aged 5 to 17 years; (4) including healthy children as well as children suffering from any disease or disorder; (5) determining at least one of the psychometric properties, such as reliability, validity, responsiveness, etc.; and (6) written in English.

Exclusion criteria: The exclusion criteria were as follows: (1) studies exclusively focused on adults or the elderly population; (2) any intervention study using reach tests only as one of the outcome measures; (3) standalone book chapters and articles that were not available for open access to readers.

Search strategy and source of information: The search strategy for the present review involved the amalgamation of key terms and their synonyms. These terms were selected based on the research questions and were identified during the initial literature review of studies focusing on reach tests in children. The search terms included “children,” “paediatric,” “child,” “adolescents,” and “school-going children,” in association with “reliability,” “validity,” “psychometric properties,” and “clinimetric properties.” Types of studies included were “observational studies,” “cross-sectional studies,” “reliability studies,” “review studies,” and “clinical trials.” The search was conducted using Boolean operators ‘AND,’ ‘OR,’ and ‘NOT.’ The combinations used for the search strategy are mentioned in [Table/Fig-2].

Selection of studies: During the initial screening stage, studies were considered based on their titles and abstracts. In the second stage, approval was based on a full-text review. Articles discovered through manual searches were also included in the full-text screening stage. At each stage of the selection process, two independent reviewers screened and selected articles, resolving differences through discussion until consensus was reached. The agreement between the two reviewers before consensus discussions ranged from 94% at the abstract level to 92% at the full-text level.

The selected studies that met the inclusion criteria provided descriptive psychometric information, which was gathered and categorised based on measurement terms, objectives, age groups, scope/population, and psychometric properties. In a contemporary context, “validity” can be defined as the assurance that an inference or decision drawn from a measurement is appropriate. Ongoing assessment of validity is paramount, and it should be viewed as a unified concept.

Various dimensions of empirical evidence for validity, including concurrent, predictive, construct, known group/discriminative, convergent, and face validity, were sought. Concurrent validity assesses how effectively a measure aligns with a well-established test, often considered a standardised “gold standard” test; this data is usually collected simultaneously with the target measurement. Predictive validity is often described in terms of sensitivity and specificity. Sensitivity pertains to the ability of a measurement to identify individuals with a particular condition (e.g., children with impaired balance), while specificity concerns the capability to accurately distinguish individuals without the condition (e.g., typically developing children). The recommended benchmarks for sensitivity are >80%, and for specificity, >90%.

Construct validity is linked to the overall perceived validity of the measurement and is characterised by the theoretical foundation for employing the measurement, which often involves factor analysis. Known group validity examines whether a test can differentiate between a group of individuals known to have balance impairment and a group of typically developing individuals. Discriminative validity confirms that measures or tests that should not be related are indeed unrelated. Convergent validity measures the extent to which two constructs that should theoretically be related are, in fact, related. Known group validity, discriminative validity, and convergent validity are all considered subcategories of construct validity.

To evaluate validity, a comparison of score-level attributes or measurement constructs between the original and adapted versions is conducted, assessing whether the scale operates similarly across diverse populations (measurement invariance and differential item functioning). Face validity relates to the degree to which one or more individuals subjectively perceive that a questionnaire effectively addresses the concept it purports to measure.

Reliability represents the overall consistency of a measure, indicating how stable it is when repeated under consistent conditions. Initially, test-retest reliability examines the relative stability of an assessment over time, assessing how consistent the scores from the measurement tool are from one test administration to the next. Second, inter-rater reliability assesses the level of agreement between two raters. Third, internal consistency evaluates how effectively the items in the questionnaire measure the same underlying construct. Measures exceeding 0.80 are regarded as excellent, with a minimum acceptable value of 0.70.

Assessment of methodological quality: The quality of the selected articles was assessed using the GRADE methodology.

Please note that the GRADE criteria must be modified for the present review, as they are typically used for evaluating interventions and diagnostic tools [12]. For instance, randomised trials with no significant limitations are categorised as providing high-quality evidence, while observational studies lacking significant strengths or substantial limitations are considered to provide low-quality evidence. To determine the appropriate psychometric qualities, the COSMIN criteria were utilised.

A total of 944 articles were identified from different databases using the search terms “child$” AND “reach test$” AND (“psychometric” OR “validity” OR “reliability”) NOT “adults,” applying Boolean operators AND, OR, and NOT. Out of the 944 studies, 357 articles were retrieved from MedLine/PubMed, 124 from SCOPUS, 441 from OVID-SP, 16 from the Cochrane Library, and one from PEDro. Additionally, five articles were hand-picked that were not included in any of these databases.

From these articles, 629 studies were found to be duplicates. Furthermore, 113 articles were excluded because they were experimental, standalone, book chapters, or conference proceedings. Studies were also excluded due to reports that could not be retrieved (n=94) and records marked as ineligible by automation tools (n=73). This left a total of 35 articles that were screened for eligibility.

Out of these 35 articles, 15 were finally included, as data in nine studies did not pertain to the review, four articles lacked adequate detail for evaluation, and outcomes were not relevant in five studies and two as they were literature review and systematic review studies [Table/Fig-3].

[Table/Fig-3]:

Prisma flow diagram.

*Consider, if feasible to do so, reporting the number of records identified from each database or register searched (rather than the total number across all databases/registers)

**If automation tools were used, indicate how many records were excluded by a human and how many were excluded by automation tools

The characteristics of the included studies were compiled in [Table/Fig-4], in which the study design, population, sample size, age, aim of the study, outcome measures, and main findings were categorised according to the respective authors and publication years. The psychometric properties of the outcome measures and the quality of the studies were represented in [Table/Fig-5]. An adapted version of the COSMIN checklist for the representation of psychometric properties is depicted in [Table/Fig-6].

In total, 15 studies using the ‘reach test’ as one of the outcome measures were systematically reviewed, consisting of 14 cross-sectional studies and one clinical trial. Among the 15 included articles, 11 recruited typically developing or healthy children [13-22], one included children with traumatic brain injury [23], two involved children with Cerebral Palsy (CP) [24,25], one focused on children with balance impairment [26], and another targeted children with hearing impairment [27].

The reach tests used in the studies included the Forward Reach Test (FRT) [14,26,23], modified FRT (mFRT) [23], seated FRT (sFRT) [26,18], Sit and Reach Test (SRT) [15,16,21], Back-and-Saver Sit and Reach Test (BS-SRT) [13,20], and the Progressive Reach Test (PRT) [24,27].

For the FRT, face validity was compared between Typically Developing (TD) children and children with CP, revealing a significant difference with a p-value <0.05 for both forward and lateral directions. The concurrent validity for the lateral direction of the FRT was found to be between 88-100%, while for the forward direction, it was observed to be 86-88%. Content validity was established by correlating the values with the centre of pressure, which showed moderate to high correlations, with coefficients of correlation observed between 0.5 and 0.7 for both lateral and forward directions [22]. Another study determined the test-retest reliability, internal consistency, intrarater reliability, and inter-rater reliability of the FRT among children with balance impairments, which were depicted as ranging from 0.75 to 0.98, 0.70 to 0.80, 0.92 to 0.98, and 0.83 to 0.93, respectively [18]. Criterion validity and accuracy were also predicted, with inter-rater and test-retest reliability yielding a correlation coefficient of 0.84, and a significance level of less than 0.01 for accuracy, indicating excellent reliability [25]. The intrarater, inter-rater, accuracy, and responsiveness of the mFRT were evaluated for TD children and children with Traumatic Brain Injury (TBI), showing excellent reliability (ICC=0.90-0.99) and a significance level of 0.01 for accuracy. The Standard Error of Measurement (SEM) for responsiveness ranged from 0.90 to 1.4 [23].

A study determined the test-retest reliability and intrarater reliability of the seated FRT among TD children, with the intraclass correlation coefficient indicating moderate to good reliability, with values of ICC=0.40-0.75. Responsiveness was also assessed in terms of Standard Error of Mean (SEM), showing values between 1.58 and 9.38, with and without foot support [17].

The quality of the SRT and BS-SRT was evaluated by six authors, who recruited children aged 3 to 17 years. Among these six articles, four explored the psychometric properties of the SRT, while two evaluated the properties of the BS-SRT. The SRT is typically used to assess the flexibility of the hamstring muscles and/or the physical fitness of individuals. Criterion validity of the SRT was found to be moderate to good for children aged 11-15 years, with an R value ranging from 0.51 to 0.72, while it was found to be fair to moderate among children aged 6-17 years (R=0.37) when compared to the modified SRT [15]. All modifications of the SRT exhibited excellent reliability with significant accuracy (p-value <0.05). Specificity for this test was also determined, showing beta values ranging from 0.58 to 1.2 [13]. Responsiveness was calculated using SEM and effect size, showing an SEM of 5.5 for hamstring flexibility on both the right and left-sides [15], and 0.22 to 0.34 when comparing flexibility between children and adolescents [13]. A very small standard error of 0.04 was observed when comparing flexibility between males and females, with a small effect size of 0.08 [19].

The psychometric properties of the PRT, evaluated for children between two years to 11 years of age, demonstrated ‘moderate to good’ criterion validity with R-values ranging from 0.42 to 0.77. The measure also exhibited excellent construct validity (R=0.88) and excellent test-retest, intrarater, and inter-rater reliability, with ICC values greater than 0.75. Additionally, responsiveness was assessed using effect size, which yielded a large effect size of 0.95 [24]. The standard error of measurement (SEM) for the same outcome measure was calculated to be between 0.29 and 0.51 [27].

This review aims to investigate the psychometric attributes of reach tests used with children. To assess the psychometric qualities of the reach tests, eight different tests were identified. Overall, there was considerable variation among the studies in terms of the types of functional postural control tests (one system vs. multiple systems), the demographics, age ranges, and psychometric qualities examined in both healthy children and those with disabilities. The duration of the examinations, test procedures, and required equipment also varied significantly.

Test reliability was assessed at least once for each test, while measurement error, criterion validity, and construct validity were reported frequently, though not consistently [Table/Fig-5]. There remains a lack of research on responsiveness, sensitivity, specificity, internal consistency, accuracy, and test-retest reliability. Despite the fact that the FRT, PRT, SR, MSR, MFRT, PBS, and TUG were the most thoroughly studied, there is still limited evidence regarding their psychometric properties. Strong evidence were reached only for the FRT’s test-retest and inter-rater reliability of the FRT. Additional methodologically rigorous studies are likely to alter the estimates for all other psychometric properties.

It is postulated that postural control is a multisystemic construct [28], and its assessment instruments should address this. Test batteries were thought to somewhat align with this theoretical framework [Table/Fig-5]; nevertheless, the full multisystemic framework of postural control is covered by eight tests [28]. Understanding the underlying constructs of the tests, which establish the tests’ construct validity, is essential for identifying the postural control systems based on the type of task.

Construct validity was only examined for the FRT and PRT in these paediatric populations [29,30], and not for the Kids-BES Test [4,31]. The PBS does not evaluate reactive postural responses, whereas the FRT assesses the complete multisystemic framework [Table/Fig-5]. However, not all hypothetical systems were statistically converted into true dimensions [30]. The PBS appears to consist of two dimensions [30], and not every task within the PBS falls into the same dimensions [29]. Therefore, it is possible that other factors, such as the populations included, are at play, as these dimensions do not adequately capture the variety of systems required for postural control.

Previous exploratory work consistently demonstrated unidimensionality and task-specificity in studies involving healthy children [29,32,33] or heterogeneous pathological populations [29,34]. This was evidenced by correlations ≥0.70 for similar tasks (e.g., control of dynamics: TUG vs. SWOC [35], vs. FSST [36], and vs. FRT [25]) and small-sized correlations across different tasks (anticipatory, reactive, steady-state, and dynamic balance) [32,33]. While the various tasks appear to be highly related, depending on the task, a different dominant system is depicted by the small-sized significant correlations that represent criterion validity [Table/Fig-5]. The multisystemic nature of postural control [28] is supported, for example, by correlations in typically developing children between TUDS (control of dynamics) and FRT (orientation in space) (r=-0.32), or between BBW (control of dynamics) and SBST (anticipatory postural adjustments, APA) (r=-0.26) [37,38].

According to recent research, a child’s capacity to engage in postural control also depends on their developmental stage [32], which is supported by the availability of various age standards. Significant variations were observed, for instance, between the ages of 3, 4, 5, and 6 for BBW [37], and between 2.5 and 5 years for PBS [39], with significant six-month differences found. Therefore, postural regulation in healthy children is age- and task-specific.

On the other hand, in homogeneous diseased groups, the dimensionality examined through structural validity analysis varies [4,30,40]. The authors define the Functional Assessment of Balance (FAB) in children with Cerebral Palsy (CP) as having two dimensions: “static and quasi-dynamic balance function” and “stability in gait.” The One-Legged Stance (OLS) is included in the final dimension, but it appears to be more closely connected to the first than to the second [30]. It seems that the two dimensions are determined by the perceived difficulty or ease of the tasks, rather than the nature of the work for these children. Other preliminary research has demonstrated this shift in dimensionality: one dimension for the FAB with a modified item hierarchy in stroke patients [40,41], and four dimensions for the BESTest in stroke patients [40]. Therefore, item hierarchy and dimensionality are influenced by each specific pathology.

The results support the notion that the degree of the disease influences postural control performance. Children with mild motor deficits, such as hearing impairment [42], as well as those with severe motor deficits, such as traumatic brain injury with FSST [43], can be distinguished from healthy children. Additionally, children with varying functional levels, like those classified as Gross Motor Function Classification System (GMFCS) levels I-III (CP), also show differences, as indicated by PBS supporting the known-group validity of these tests [44,45] [Table/Fig-4]. In addition, functional postural control tests and GMFM total scores are more correlated in children with CP [25,44,46] than in more diverse groups, such as children with balance disorders [47]. This suggests that as movement disorders worsen, there is a stronger relationship between all motor components and postural control. Consequently, the severity of the underlying condition becomes more significant than the task-specificity observed in healthy individuals. Differences between the two groups should be noted.

The data supporting the good reliability of all other functional postural control tests is (very) weak, with the exception of the FRT, PRT, SR, MFRT, and PBS. This low evidence is primarily due to a high-risk of bias, although imprecision caused by small sample sizes and inconsistent results-evidenced by either conflicting outcomes or very wide confidence intervals-also contributed. More consistent results were found for children with severe movement disorders like CP or traumatic brain injuries, while more inconsistent results were observed for younger children [48]. This inconsistency may be attributed to the typical day-to-day variability in performance resulting from their developmental stage. Not every form of reliability was examined for every test, which is crucial if measurements must be repeated, shared among healthcare providers, or used to assess the efficacy of interventions [49-51]. While SEM-based measurement errors, derived from test-retest reliability, can be useful in interpreting the results of physiotherapeutic interventions, a modified score can only be ascribed in relation to the degree of error, which means that the change lacks clinical significance (responsiveness). Therefore, if the responsiveness for a given test is not established, the amount of evidence for measurement errors diminishes, according to COSMIN criteria. Currently, only the FRT [52], PRT, SR, MSR, MFRT, and TUG [53,54] have available records, and responsiveness has not been thoroughly studied. Either the Minimal Clinically Important Difference (MCID) using the anchor-based approach [52,53,55] or the Standardised Response Mean (SRM) using the distribution method [52,54,55] have been used to evaluate how responsive a test is. The anchor-based method allows a clinician or researcher to quickly determine whether a change is clinically meaningful, while the distribution method (calculation of SRM) only yields statistically significant results and does not allow for the clinical interpretation of test score changes [56,57]. Additionally, because the distribution approach relies on the population’s standard deviations, the results are not adequately generalisable [57].

In relation to the test’s comprehensiveness, administration times varied, ranging from less than five minutes for some individual tests to 30 minutes for the Kids-BESTest [31,58]. Longer administration times should not be viewed as a limitation. The children were given explanations and demonstrations of each test beforehand and were often allowed a practice run to become comfortable with the procedures. In this way, motor function was evaluated rather than cognitive capacity [36]. Unless otherwise indicated, functional postural control assessments are conducted barefoot to simulate the subject’s ability to balance as naturally as possible [58].

The comprehensive search query was used to methodically search three databases. Data extraction and bias risk assessment were conducted by two separate reviewers. By employing hand searching, the likelihood of overlooking potentially relevant articles was reduced. The degree of evidence for each test was determined using the COSMIN checklist, which is recommended for assessing techniques in psychometric studies. Consequently, it is true that a thorough interpretation of the findings from the included research is limited by the low-quality scores. Interpreting test findings is challenging due to the wide range of functional postural control tests available, which has led to various examinations of the different measurement characteristics of these tests.

Recommendations for future research: To draw firm conclusions about how well test scores reflect the dimensionality of the construct being measured, it is imperative that future research focuses on investigating the criterion validity of the most comprehensive test batteries using methodologically sound study designs. It is currently unclear whether the underlying dysfunctional neural pathways are connected to the multisystemic framework. Therefore, a comprehensive evaluation of criterion validity may help clarify if all systems are being addressed as theoretically recommended, ideally in conjunction with brain imaging methods. Both typically developing children and pathologically affected groups, ranging from mild to severe, should be included in this investigation, as they could all benefit from physiotherapeutic treatment planning related to postural control. Age disparities must also be considered. Additionally, emphasis should be placed on the responsiveness of structurally valid tests based on the anchor-based approach.

The conclusions drawn from the review indicate that FRT is the most commonly used “reach test” as an outcome measure in the majority of studies, and it possesses excellent psychometric properties. This is followed by the Sit and Reach Test and the Plank Reach Test among children. Further trials are recommended to establish the psychometric properties of reach tests in multiple directions for a better analysis of balance among children.