This systematic review follows the guidelines outlined in the preferred reporting items for systematic reviews and meta-analyses (PRISMA) statement [14]. On October 1, 2023, a comprehensive literature search was conducted across four major databases: PubMed, Scopus, Web of Science, and Embase. Specific search terms were devised for each database, incorporating (“radius” OR “radial” OR “ulna” OR “ulnar” OR “forearm”) AND (“fracture*” OR “displace*”) AND (“sonograph*” OR “ultrasonograph*” OR “ultrasound” OR “POCUS”) AND (“pediatric*” OR “paediatric*” OR “child*” OR “neonat*” OR “infant*” OR “toddler*” OR “preschool” OR “pre-school” OR “juvenile” OR “young adult*”). Additionally, a thorough manual examination of references within the included studies ensured no relevant papers were inadvertently overlooked. The assessment process involved a detailed review of each article’s title, abstract, and/or full text. Two co-authors independently conducted this review, resolving uncertainties or ambiguities through consultation with a senior co-author. The AutoLit platform, developed by Nested Knowledge in St. Paul, Minnesota, USA, facilitated deduplication, screening, and data extraction.

All studies pertinent to the topic of interest, presenting at least one of the following diagnostic accuracy measures in pediatric patients (under 21 years old), were considered for inclusion: sensitivity, specificity, positive predictive value (PPV), negative predictive value (NPV), likelihood ratio (LR), diagnostic odds ratio (DOR), and area under the receiver operating characteristic curve (AUC). No constraints were imposed on publication date, country of origin, patient characteristics, reference standard type, or study design. Non-English literature, case reports, case series with fewer than five eligible patients, conference abstracts, editorial comments, and review articles were excluded from the study.

The quality assessment of diagnostic accuracy studies-2 (QUADAS-2) tool was applied to evaluate the quality of included studies [15]. The four primary domains of the QUADAS-2 tool, including patient selection, index test, reference standard, and flow and timing, underwent independent assessment for potential bias and concerns regarding applicability. Evaluations for each domain were based on specific criteria outlined in the tool, such as the representativeness of the study population, blinding of test results, and completeness of outcome data. Ratings of “low,” “high,” or “unclear” were assigned to each domain to determine the overall rigor and reliability of the evidence synthesis.

The primary analytical approach utilized a random effects diagnostic test accuracy (DTA) model, specifically the bivariate model developed by Reitsma et al. [16]. Summary receiver operating characteristic (SROC) curves were generated using this bivariate meta-analysis data. For visualization purposes, study-specific estimates were relatively weighted in SROC plots based on the weights within the random effects univariate DOR model. AUC and its confidence interval (CI) for each subgroup were calculated using a 2000 sample bootstrapping technique based on the bivariate model [17].

To assess heterogeneity, the I2 metric was employed following the approach by Holling et al. [18]. A significant level for heterogeneity was considered for I2 confidence intervals above 25%, leading to further sensitivity analyses through the DOR univariate meta-analysis to identify and re-analyze potential outliers.

The study also investigated the influence of various covariates on the reported rates using subgroup meta-analysis and meta-regression techniques. Covariates included training status and roles of the individuals involved, image acquisition methods (either four or six views), and the age of patients. Considering the variability in reporting metrics, with some studies reporting by patient number and others by the number of bones scanned, these aspects were treated as covariates for further subgroup analyses. Subgroup comparisons were made between studies reporting on a patient-wise basis vs. those reporting on a bone-wise basis. Separate subgroup analyses were also conducted for studies reporting diagnostic accuracy metrics for fractures in the radius and ulna.

Fagan plots and LR scattergrams were utilized to assess the clinical applicability of findings. Positive LRs above ten signified suitability for confirmation, whereas negative LRs below 0.1 indicated exclusion suitability. Fagan nomograms were constructed for pre-test prevalences of 25%, 50%, and 75%, based on the bivariate Reitsma model, as detailed by Zwinderman et al. [19].

Publication bias was scrutinized using an adaptation of Egger’s regression test for DTA meta-analysis, involving the analysis of funnel plot asymmetry with 2000 sample bootstrapping, as recommended by Noma et al. [20].

All statistical procedures were conducted using R (version 4.2.1, R Foundation for Statistical Computing, Vienna, Austria), utilizing packages such as “Mada,” “MVPBT” [21], “dmetatools” [17], “Metafor” [22], and “meta” [23].