The Sheffield Scoring System: An 8-Year Descriptive Cohort Analysis of a Tool for Predicting the Need for Endoscopic Hemostatic Intervention in Pediatric Patients with Upper Gastrointestinal Bleeding at a Tertiary Hospital in Khon Kaen, Thailand Authors: Busara Charoenwat, Sumitr Sutra, Kaewjai Thepsuthammarat, Mike A. Thomson Published: 2026 ABSTRACT PurposeUpper gastrointestinal bleeding (UGIB) is life-threatening, and the gold standard for diagnosis and treatment is endoscopy, which is relatively invasive. Our purpose was to predict the need for endoscopy among children with UGIB using the Sheffield scoring system, characterize clinical and endoscopic findings among children with UGIB, and determine the predictive value of the Sheffield scoring system for endotherapeutic intervention.MethodsThis descriptive cohort study included 28 children aged 0–18 years who presented with UGIB between 2017 and 2024 and were assessed using the Sheffield scoring system. Clinical and endoscopic data were analyzed.ResultsAmong the 28 children, 21 (75.0%) and 7 (25.0%) respectively scored >8 (group 1) and <8 (group 2) in the Sheffield system. Endoscopic hemostasis was required for 22 (78.6%) patients, of whom 95.5% were in group 1. One child in Group 2 who underwent endoscopy had a vascular malformation with prior bleeding. The most prevalent underlying disease was portal hypertension (85.7%). The most significant endoscopic finding was esophageal varices (71.4%). The positive and negative predictive values were respectively 100.0% (95% confidence interval [CI], 83.9–100.0%) and 85.7% (95% CI, 42.1–99.6%). The sensitivity and specificity were respectively, 95.5% (95% CI, 77.2–99.9%) and 100.0% (95% CI, 54.1–100.0%).ConclusionThe Sheffield scoring system might be a useful tool for predicting the need for endoscopic hemostasis among children with UGIB. However, larger prospective studies are required for validation. INTRODUCTION Upper gastrointestinal bleeding (UGIB) is a common and potentially life-threatening condition in pediatric patients. Therefore, they might need to undergo immediate or emergency diagnostic and therapeutic hemostatic endoscopy. Although the actual burden of UGIB is yet to be accurately determined and the exact global incidence remains unknown, between 300–500 patients per year are annually diagnosed in the United Kingdom [12]. A case-crossover study found that that the annual incidence of UGIB in France [3] was 1–2 per 10,000 children and an 11-year retrospective endoscopic investigation found a 6.4% incidence of UGIB in the USA [4]. A multicenter, retrospective, cohort study found a mortality rate of approximately 2.07% that peaked within 7 days of admission. However, early endoscopy during admission was a protective factor [5]. Our institute has 24/7 on-call coverage, but only one pediatric endoscopist. Moreover, many low- and middle-income countries (LMICs) face a shortage of endoscopists who are available 24/7. Adult scoring systems to predict the need for endoscopic hemostatic intervention have been established. They include the pre-endoscopic Glasgow–Blatchford score (GBS) [6] and the Addenbrooke scoring system [7] as well as the post-endoscopic Rockall scoring system [8]. Furthermore, scoring models for adults with UGIB, such as the modified GBS [9] and AIMS65 [10], are being updated. However, adult scoring systems cannot be directly applied to the pediatric population because of differences in pathophysiology, hematological, and biological variables, and importantly, age-specific pathophysiology in children [167891011]. Therefore, we aimed to characterize the clinical presentation and endoscopic findings among children with UGIB at a tertiary hospital in a developing country. We also investigated the ability of the Sheffield scoring system to predict the need for endoscopic hemostatic intervention among children with acute UGIB. MATERIALS AND METHODS This study was conductedat the Gastrointestinal Endoscopy Center of Excellence at the Srinagarind Hospital, which is a referral and tertiary government hospital in Khon Kaen, Northeast Thailand. The International Review Board at the Center for Ethics in Human Research and Khon Kaen University Human Research Ethics Committee approved this study (HE 671161). The parents or guardians provided written, informed consent for the participation of their children in this study. The first patient was enrolled on January 1, 2017. Informed consent The first patient was enrolled on 1 January 2017. Written informed consent was obtained from the parents or legal guardians of all participants. The authors confirm that all procedures complied with relevant guidelines and regulations. Study design and population This descriptive cohort study analyzed data from all children aged ≤18 years who presented with UGIB and were admitted to Srinagarind Hospital between January 1, 2017 and March 31, 2024. We combined retrospective and prospective studies because UGIB is quite rare among children. The retrospective phase allowed sufficient data to accumulate over a prolonged period and increase statistical power. The prospective phase enabled standardized data collection and minimized missing or incomplete clinical variables. Medical records were retrospectively analyzed during the first period (January 1, 2017–January 31, 2022), and prospective data were analyzed exclusively during the second period (February 1, 2022-March 31, 2024). The Sheffield scoring system was applied retrospectively and prospectively to children with the intent to predict the need for endoscopic intervention and those who required esophagogastroduodenoscopy (EGD). Children with UGIB who were clinically managed without EGD were excluded because the objective of this study was to define the performance of the score in children who underwent endoscopy, rather than as a pre-endoscopic triage tool. This study included children aged ≤18 years with UGIB who presented at our center and underwent EGD. Patients with incomplete medical records, no EGD within 48 hours, and surgical conditions such as Meckel’s diverticulum were excluded. All participants with active bleeding or features associated with a high risk of bleeding (visible vessels, adherent clots, or varices) underwent endoscopic hemostatic intervention regardless of Sheffield scores. Data The demographic characteristics of the patients and the clinical data collected included the type (hematemesis and melena) and amount of UGIB, comorbidity status, medications, relevant clinical factors, poor perfusion, laboratory results, and the need for isotonic volume or blood product resuscitation, and Sheffield scores. Clinical management protocols were consistent between the retrospective and prospective study periods. Standard pre-endoscopic medical interventions for all patients with UGIB comprised an intravenous octreotide bolus of 1–2 mcg/kg followed by a maintenance dose of 1–2 mcg/kg/h until endoscopic hemostasis was successful or up to 5 days. In addition, 1 mg/kg of intravenous proton pump inhibitors were administered every 12 hours. Although the Sheffield system produces scores up to 24, the cutoff for intervention is ≥8; therefore, children with scores of ≥8 and <8 were assigned to groups 1 and 2. Statistical methods Continuous and categorical variables were respectively analyzed using Mann–Whitney U and chi square tests. Descriptive results are shown as means±standard deviations for numerical variables and as numbers (n) and percentages (%) for categorical data. Receiver operating characteristic (ROC) curves were calculated to determine average Sheffield scores. The predicted need for endoscopic hemostatic intervention was taken as the area under the ROC curve, together with 95% confidence intervals (CIs). Diagnostic accuracy, sensitivity, specificity, positive (PPV), and negative (NPV) predictive values were calculated using the cutoff defined by the ROC analysis. Data were analyzed using Stata version 10.1 (StataCorp LLC.). The statistical methods were reviewed by Kaewjai Thepsuthammarat at the Clinical Epidemiology Unit, Srinagarind Hospital, Faculty of Medicine, Khon Kaen University. RESULTS Baseline and clinical characteristics In total, 28 patients (male, 71.4%) were referred to the endoscopic unit during the study period. Sixteen and 12 children were respectively included in the retrospective and prospective analyses. The mean age of the patients was 4.9±3.4 years and 60.7% were aged <5 years. Hematemesis was evident in 26 (92.9%) patients. One (3.6%) patient had melena and another had hematemesis accompanied by melena (3.6%). The most common underlying disease was biliary atresia (BA; n=11), followed by extrahepatic portal vein thrombosis (EHPVT; n=6), congenital hepatic fibrosis (CHF; n=4), peptic ulcer disease (PUD; n=3), vascular malformation (n=2), and cirrhosis of unknown etiology and IgG4-sclerosing cholangitis with cirrhosis (n=1 each). Comorbidities consisted of portal hypertension (PHT) in 85.7% and non-PHT in 14.3% of the patients (Table 1). Table 1Demographic and clinical characteristics of the participantsCharacteristicsValue (n=28)Age (yr)<517 (60.7)≥511 (39.3)SexMale20 (71.4)Female8 (28.6)UGIB typeHematemesis26 (92.9)Melena1 (3.6)Hematemesis and melena1 (3.6)Underlying diseasesBA11 (39.3)EHPVT6 (21.4)CHF4 (14.3)PUD3 (10.7)Vascular malformation2 (7.1)Cirrhosis of unknown etiology1 (3.6)IgG4 sclerosing cholangitis with cirrhosis1 (3.6)Poor perfusionYes17 (60.7)No11 (39.3)Fluid bolus requiredYes15 (53.6)No13 (46.4)Blood transfusion requiredYes9 (32.1)No19 (67.9)Values are presented as number (%).BA: biliary atresia, CHF: congenital hepatic fibrosis, EHPVT: extrahepatic portal vein thrombosis, IgG4: immunoglobulin 4, PUD: peptic ulcer disease, UGIB: upper gastrointestinal bleeding. Demographic and clinical characteristics of the participants Values are presented as number (%). BA: biliary atresia, CHF: congenital hepatic fibrosis, EHPVT: extrahepatic portal vein thrombosis, IgG4: immunoglobulin 4, PUD: peptic ulcer disease, UGIB: upper gastrointestinal bleeding. Sheffield score in the study population Nearly all 21 (75.0%) and 7 (25.0%) patients with Sheffield scores ≥8 (group 1) and <8 (group 2) presented with hematemesis (90.5% and 90% in groups 1 and 2, respectively). PHT was detected in 24 patients, of whom 21 (87.5%) were in Group 1, and most had BA (47.6%). One patient without PHT had a vascular malformation and was classified into Group 1 (Supplementary Table 1). Endoscopic diagnosis and interventional procedures Among 22 (78.6%) patients who required endoscopic hemostatic intervention, 21 (95.5%) were in Group 1. The most common clinical presentation was hematemesis, followed by melena, and hematemesis with melena. The rate of therapeutic endoscopy was significantly higher in Group 1 that included patients with esophageal varices (EVs; n=20; 95.2%), than in Group 2. Ten and nine patients respectively had grades 2 and 3 EVs, and two had grade 3 EVs with vessels on vessels. The most prevalent endoscopic findings were EVs (71.4%), PUD (10.7%), portal hypertensive gastropathy (10.7%), and vascular malformations (7.2%). Among the patients who received endoscopic intervention, 17 (77.3%) and 3 (13.6%) were respectively treated by EVs ligation if they weighed ≥10 kg or were aged ≥1 year and by EV sclerosis if they weighed <10 kg or were aged <1 year. Blue rubber bleb nevus syndrome (9.1%) in two patients was treated using argon plasma cauterization (Table 2). Table 2Clinical characteristics of the participants according to endoscopyTherapeuticDiagnosticSheffield score≥8Sheffield score<8Sheffield score≥8Sheffield score<8UGIB typeHematemesis19 (67.9)1 (3.6)0 (0.0)6 (21.4)Melena1 (3.6)0 (0.0)0 (0.0)0 (0.0)Hematemesis and melena1 (3.6)0 (0.0)0 (0.0)0 (0.0)Endoscopic findingsEVs20 (71.4)0 (0.0)0 (0.0)0 (0.0)PUD0 (0.0)0 (0.0)0 (0.0)3 (10.7)PHG0 (0.0)0 (0.0)0 (0.0)3 (10.7)Vascular malformation1 (3.6)1 (3.6)0 (0.0)0 (0.0)Values are presented as number (%).EVs: Esophageal varices, PHG: portohypertensive gastropathy, PUD: peptic ulcer disease, UGIB: upper gastrointestinal bleeding. Clinical characteristics of the participants according to endoscopy Values are presented as number (%). EVs: Esophageal varices, PHG: portohypertensive gastropathy, PUD: peptic ulcer disease, UGIB: upper gastrointestinal bleeding. The sensitivity and specificity of a Sheffield score ≥8 for therapeutic endoscopic intervention were respectively, 95.5% (95% CI, 77.2–99.9%) and 100% (95% CI, 54.1–100.0%). The PPV and NPV were respectively 100.0% (95% CI, 83.9–100.0%) and 85.7% (95% CI, 42.1–99.6%; Fig. 1). ROC curves of Sheffield scores and endoscopic therapeutic intervention. ROC: receiver operating characteristic. DISCUSSION The most prevalent clinical presentation among patients with UGIB in this series was hematemesis (92.9%), which was consistent with the findings of studies in Europe and the USA [4111213]. Among the patients in Group 1, 90.5% presented with hematemesis. The most frequent underlying diseases among the patients with UGIB were consistent with the literature; PHT was the most prevalent, followed by BA, EHPVT, CHF, cirrhosis of unknown etiology, and IgG4-sclerosing cholangitis with cirrhosis. The major finding of therapeutic hemostatic endoscopy was EV that corresponded with PHT comorbidity. The etiology of UGIB varies with age and geographic distribution. The major etiology of UGIB is variceal bleeding in India [14], Asia, Sudan [15], and Iranm [1617], but non-variceal bleeding in Taiwan [18], Turkey [19], USA [4], regions of West Africa [20], and Saudi Arabia [21]. The most prevalent endoscopic finding among patients in countries with higher incomes was gastric erosion. This was associated with nonsteroidal anti-inflammatory drugs (NSAIDs) and infection with Helicobacter pylori. Most of the patients analyzed herein who had UGIB also had PHT (predominantly BA), and other underlying conditions. This etiological distribution reflects the characteristics of patients managed at our tertiary center that differed from that of patients in hospitals in high-income settings where nonvariceal bleeding is dominant. Consequently, the diagnostic performance of the Sheffield scoring system in our cohort might have been influenced by the high pre-test probability of endoscopic intervention and disease-specific characteristics of the study population. These factors should be considered when interpreting the findings, and caution is warranted when generalizing the results to populations with a predominance of nonvariceal bleeding. All children with UGIB were evaluated by EGD, indicating the significantly high quality of our endoscopic unit. The European Society of Gastrointestinal Endoscopy and the European Society for Pediatrics Gastroenterology Hepatology and Nutrition recommend that EGD should be applied within 12-24 hours in general, but as soon as possible when patients have unstable UGIB [22]. Although our institution has a 24/7 on-call coverage, only one pediatric endoscopist is available. Moreover, many hospitals and regions in LMICs have suboptimal endoscopy coverage. The predominant reasons are a shortage of pediatric endoscopists and pediatric-sized endoscopes, minimal or no assistance, and little or no financial support [12]. This suggests that training pediatric endoscopists to acquire such skills is urgently required [1]. The diagnostic yield of EGD in patients with UGIB was higher in the present, than in previous studies [91317222324]. A relatively low diagnostic yield of endoscopy among patients in Egypt was associated with the lack of a standardized interval between UGIB onset and endoscopy, because the distance between a referral hospital and a endoscopy center is distant [25]. The diagnostic yield of endoscopy is greater for EGD combined with colonoscopy (75.4%) than for EGD alone but only for clinically lower GIB (65%) [26]. The differences in the diagnostic yield of endoscopy for patients with UGIB might be attributed to variations in the analyzed population, geographic area, etiology, financial status, government referral policies, definitions of significant endoscopic findings and types of endoscopy [2026]. Adult scoring systems cannot be applied to children due to various physiological and pathophysiological factors. Risk factors have been analyzed and a scoring system was developed to predict the severity of UGIB in children [27]. However, this system classifies UGIB severity based on the International Classification of UGIB in adults. This is unsatisfactory for pediatric patients. The novel Sheffield scoring system uses pediatric parameters to predict the need for endoscopic hemostatic intervention to treat UGIB in children [2]. Among our study population, 22 patients underwent endoscopic hemostatic intervention and EVs were confirmed in 21 of them, who were all in group 1. The sensitivity was slightly equivalent to previous results [2]. However, the incidence of non-variceal bleeding was slightly higher [19] among most patients with UGIB and EVs (90–100% and 76.9%, respectively) [2028]. The high specificity and PPV were consistent with the predominance of EV and PHT as the underlying etiologies of UGIB in our cohort. Specificities and PPVs of 61.8–91.2% [22029] and 35.7–91.1% [21929], respectively, have been published. The NPV was lower (here than that found in previous studies (85.7%; 95% CI, 42.1–99.6%) vs. 88.6–94.8%) [21929]. Although diagnostic accuracy was high, our findings should be interpreted with caution because of the small sample, statistical uncertainty, and that external validation will be required before definitive conclusions can be achieved. One patient with a vascular malformation, a history of bleeding, and endoscopic confirmed high-risk bleeding generated a false-negative result. Endoscopic hemostatic intervention was required to prevent bleeding or rebleeding in this patient. This experience indicated that clinical judgment remains essential even when patients are classified by the scoring system as having low risk. This study had several limitations. The sample was relatively small and was derived from a single tertiary center, which might have limited the generalizability of the findings. The Sheffield scoring system was used only in children who underwent EGD. Patients with conservatively managed UGIB (without EGD) were excluded. This might have resulted in selection bias, and a possibly overestimated diagnostic performance of the scoring system. Accordingly, the findings should be interpreted as supportive evidence of score performance within a population that underwent endoscopy, rather than as validation of its use for pre-endoscopic decision-making. In addition, bleeding associated with PHT predominantly due to biliary atresia and related conditions was frequent in our cohort; this might have resulted in a high pre-test probability of endoscopic intervention. Moreover, although a combined retrospective–prospective design was selected to balance sample size considerations with data completeness, inherent limitations remain in the retrospective data collection. The study period was relatively short. Further large, prospective, multicenter studies over longer periods are required. With respect to the clinical presentation of UGIB herein, hematemesis was the most prevalent manifestation and the most common etiology in our patients was EVs. These findings were similar to those in patients in other LMICs. The Sheffield scoring system is a reliable tool to screen and predict the need for endoscopic hemostatic intervention. Local hospitals without endoscopists or sufficient pediatric endoscopic assistance could refer patients to our LMIC. Although we have 24/7 on-call coverage for all patients with UGIB who need endoscopy, this study included only one pediatric endoscopist (B.C.). However, we have a scheduled referral system and we cooperated with the Gastrointestinal Endoscopy Centre of Excellence, which did not affect the time to intervention. However, centers in LMICs often lack skilled pediatric endoscopists, assistance, and equipment. Therefore, training endoscopists to acquire specialized skills to improve the national framework, and expanding the number of centers of excellence is critical.