Conceptual Clinical Variables Enhancing the Alvarado Score in Pediatric Appendicitis: Lessons for Artificial Intelligence Models

Conceptual Clinical Variables Enhancing the Alvarado Score in Pediatric Appendicitis: Lessons for Artificial Intelligence Models | Research Square window.SnipcartSettings = { analytics: { enabled: false } }; (function() { var accessVector = localStorage.getItem('access_vector') || ''; window.dataLayer = window.dataLayer || []; if (accessVector) { window.dataLayer.push({ user: { profile: { profileInfo: { snid: accessVector } } } }); } })(); (function(w,d,s,l,i){w[l]=w[l]||[];w[l].push({'gtm.start':new Date().getTime(),event:'gtm.js'});var f=d.getElementsByTagName(s)[0],j=d.createElement(s),dl=l!='dataLayer'?'&l='+l:'';j.async=true;j.src='https://www.googletagmanager.com/gtm.js?id='+i+dl;f.parentNode.insertBefore(j,f);})(window,document,'script','dataLayer','GTM-K279D39R'); Browse Preprints In Review Journals COVID-19 Preprints AJE Video Bytes Research Tools Research Promotion AJE Professional Editing AJE Rubriq About Preprint Platform In Review Editorial Policies Our Team Advisory Board Help Center Sign In Submit a Preprint Cite Share Download PDF Research Article Conceptual Clinical Variables Enhancing the Alvarado Score in Pediatric Appendicitis: Lessons for Artificial Intelligence Models PORNSRI THANACHATCHAIRATTANA This is a preprint; it has not been peer reviewed by a journal. https://doi.org/ 10.21203/rs.3.rs-8666814/v1 This work is licensed under a CC BY 4.0 License Status: Posted Version 1 posted You are reading this latest preprint version Abstract Background: Timely and accurate diagnosis of pediatric appendicitis remains challenging because of variable clinical presentations and limitations of existing scoring systems. Although the Alvarado score is widely used, its diagnostic performance in children is suboptimal, and reliance on imaging increases resource utilization and radiation exposure. Identification of objective, biologically meaningful predictors may improve early diagnosis and inform future artificial intelligence (AI)–based decision-support models. Methods: This retrospective study included children aged 2–18 years evaluated for suspected appendicitis between January 2014 and June 2024. Clinical characteristics, laboratory data, urinalysis results, imaging findings, operative notes, and histopathological reports were reviewed. Diagnostic performance of individual variables and the Alvarado score was assessed using sensitivity, specificity, predictive values, diagnostic odds ratios, and receiver operating characteristic analysis. A composite diagnostic framework based on seven clinical and laboratory factors was evaluated and compared with the Alvarado score alone. Results: A total of 375 patients were included, of whom 275 (73.3%) were diagnosed with appendicitis. Appendicitis was more prevalent in males and in children aged ≥6 years, with the highest incidence in preadolescence. The mean Alvarado score was significantly higher in the appendicitis group than in the non-appendicitis group (8 ± 2 vs 5 ± 2, p < 0.001). At a cut-off score of ≥7, the Alvarado score demonstrated a sensitivity of 77.1%, specificity of 71.0%, and an AUC of 74.05%. Seven factors—sex, age group, location of abdominal pain, white blood cell count, neutrophil percentage, urine ketone level ≥3+, and leukocyte esterase level—were significantly associated with appendicitis. The seven diagnostic conceptual factors achieved superior diagnostic performance (AUC 86.77%, 95% CI 82.56–90.96) compared with the Alvarado score alone. Urinary ketone levels ≥3+ showed a significant positive association with appendicitis, whereas leukocyte esterase demonstrated an inverse association. Conclusions: Integration of seven objective clinical and laboratory factors improves early diagnostic accuracy for pediatric appendicitis beyond the Alvarado score alone and supports more targeted use of ultrasonography in equivocal cases. Urinary ketones are a particularly useful adjunctive marker due to their physiological relevance and persistence after fluid resuscitation. Establishing structured, clinically meaningful data pipelines is essential for future development of reliable and explainable AI-assisted diagnostic tools. Pediatric appendicitis Alvarado score Urinary ketones Diagnostic accuracy Clinical prediction model Artificial intelligence Figures Figure 1 Background Pediatric appendicitis remains a common surgical emergency in which timely and accurate diagnosis is essential to prevent delayed treatment and unnecessary appendectomy. Despite advances in diagnostic strategies and growing interest in artificial intelligence (AI)–based decision support [1-11], reliable clinical predictors continue to play a central role in pediatric practice. The Alvarado score is one of the most widely used clinical scoring systems for suspected appendicitis [12]. Over time, several modifications and alternative tools—including the Pediatric Appendicitis Score, Appendicitis Inflammatory Response Score, Pediatric Appendicitis Risk Calculator, and the Biomarkers for the Diagnosis of Appendicitis in Pediatrics (BIDIAP) index [13-16]—have been developed to improve diagnostic accuracy and support early decision-making. However, none has consistently demonstrated high performance in pediatric populations. This limitation highlights the ongoing need to refine existing clinical variables rather than relying solely on new scoring frameworks or advanced technologies. Many components incorporated into appendicitis scoring systems are inherently subjective or variable, such as anorexia, nausea, pain localization, or a reported history of fever that may be transiently suppressed by antipyretic use at the time of presentation. These factors may reduce diagnostic reliability and complicate both clinical decision-making and the development of robust predictive or AI-based models. Consequently, objective and physiologically meaningful variables that can enhance diagnostic sensitivity warrant further investigation. Urinary ketones represent a potential adjunctive marker in this context. Ketogenesis reflects a catabolic metabolic state and occurs more rapidly in children than in adults, often within 12–24 hours vs 24 hours of reduced oral intake [17-19]. Urinary ketones may therefore serve as a low-cost and readily available adjunctive marker, even when serum ketone levels have normalized [19]. In the 2016 modification of the Alvarado score, ketonuria was incorporated as a surrogate marker for anorexia; however, the diagnostic contribution of specific urinary ketone cut-off levels has not been clearly established [12]. Given their low cost, rapid availability, and physiological relevance, urinary ketones may provide additional diagnostic value in pediatric appendicitis. Conventional inflammatory markers, such as white blood cell count and C-reactive protein (CRP), are routinely used in the evaluation of suspected appendicitis but lack reliable cut-off values and may remain normal in confirmed cases [20]. Other laboratory indices—lymphocyte-to-monocyte ratio [21], procalcitonin [17,21], and the systemic immune-inflammation index used in the BIDIAP model [14] — reflect systemic inflammatory responses rather than direct anatomical pathology. Although imaging plays a critical role in identifying anatomical abnormalities, excluding alternative diagnoses, and guiding operative planning [12,20,22,23], it should complement rather than replace sound clinical judgment, and unnecessary imaging should be avoided in pediatric populations [13]. The aims of this study were (1) to evaluate the diagnostic performance of urinary ketone cut-off levels at 2+ and 3+ within the Alvarado scoring system, and (2) to identify additional predictive factors that enhance the sensitivity and accuracy of the Alvarado score and may serve as meaningful variables for future next-generation AI research. Given that data quality and variable selection critically influence the learning process of AI models, establishing a robust clinical data pipeline is essential [3,10]. Method 1. Study design and setting This retrospective study included patients aged 2–18 years who were referred for surgical consultation by emergency physicians or pediatricians between January 2014 and June 2024. All patients underwent preoperative complete blood count testing and automated urinalysis. Patients were excluded if they had undergone interval appendectomy, had incomplete data, or were receiving ketogenic diets or medications known to affect ketone levels (e.g., high-dose vitamin C, levodopa, valproate, or non-steroidal anti-inflammatory drugs). None of the included patients had anorexia nervosa, bulimia, alcohol dependence, diabetes mellitus, cortisol deficiency, chronic kidney disease, or glycogen storage disease. Key clinical characteristics and basic laboratory parameters relevant to the diagnosis of appendicitis were grouped as predictive factors based on the components of the Alvarado score [12]. Age groups were defined according to disease incidence patterns and physiological developmental stages [2]. 2. Data collection Medical records were reviewed for each Alvarado score component, final diagnosis, operative findings, and histopathological reports. Patients managed non-operatively were considered to have a normal appendix if they showed spontaneous clinical improvement, remained asymptomatic at follow-up, or had imaging findings consistent with a normal appendix or an alternative diagnosis. The Alvarado score was routinely applied across all age groups. Body temperature was measured tympanically, and fever was defined as ≥37.5 °C. Both nausea and vomiting were included as positive scoring items. When discrepancies existed among medical records regarding symptom details, the predominant clinical assessment was used. Rebound tenderness included percussion tenderness and classic signs such as Blumberg, Rovsing, obturator, and Markle signs. A white blood cell count ≥11,000/mm³ and a neutrophil percentage ≥75% were considered positive laboratory thresholds. The duration of abdominal pain was excluded from analysis because of its marked unreliability in young children. Similarly, fasting duration and the timing of blood and urine examinations relative to symptom onset were not analyzed, as these factors could introduce significant variability and bias. 3. Urinalysis Automated urinalysis was performed using the Cobas® U601 and U411 analyzers (Roche™), with evaluation focused on urine specific gravity (sp.gr.), ketones (predominantly acetoacetate), and leukocyte esterase. A normal serum ketone level (1.0 mmol/L) is approximately equivalent to a urine ketone level of 2+ (corresponding to a serum level of approximately 1.5 mmol/L) [4,13]. Urine ketone levels ≥2+ were examined in a predefined subanalysis, while levels ≥3+ were defined as “meaningful ketonuria,” corresponding to an estimated serum ketone concentration of approximately 5 mmol/L (≈15 mg/dL), in accordance with the 2016 modification of the Alvarado scoring system [6]. 4. Surgical and pathological diagnosis Surgical decisions were based on clinical judgment and imaging findings consistent with appendicitis. Histopathological results were compared with intraoperative findings. The non-appendicitis group included surgical specimens demonstrating normal histology or lymphoid hyperplasia. 5. Statistical analysis Statistical analyses were performed using SPSS software version 18.0 (SPSS Inc., Chicago, IL, USA). Continuous variables with non-normal distributions were summarized as median and interquartile range (IQR), whereas categorical variables were expressed as percentages. Comparative analyses were conducted using the Student’s t test, Mann–Whitney U test, and Fisher’s exact test, as appropriate. Agreement and association were assessed using intraclass correlation coefficients, Pearson correlation, and Cramer’s V. Diagnostic odds ratios (dORs) with 95% confidence intervals (CIs) were calculated to assess the strength of individual predictors. Sensitivity, specificity, positive predictive value (PPV), negative predictive value (NPV), overall accuracy, and the area under the receiver operating characteristic curve (AUC) with 95% CIs were estimated using logistic regression models. A two-sided p value <0.05 was considered statistically significant. Results 1. Demographic and clinical characteristics A total of 375 patients were included in the analysis, of whom 275 were diagnosed with appendicitis (148 males and 127 females) and 100 with non-appendicitis (33 males and 67 females). Age distribution was categorized into preschool (2–5.9 years), childhood (6–9.9 years), preadolescence (10–12.9 years), early adolescence (13–15.9 years), and adolescence (16–17.9 years), as summarized in Table 1. [Here A4-size] Table 1 Demographic distribution in appendicitis and non-appendicitis groups Appendicitis was more prevalent among males (53.8%) and among children aged ≥6 years, with the highest incidence observed during preadolescence (before or early menarche in females). In contrast, non-appendicitis was more common among females (67%), particularly in the preschool and adolescent age groups (postmenarche). Across all patients, diagnosis other than appendicitis included non-detectable pathology (8.8%), other gastrointestinal conditions (10.4%), and non-gastrointestinal conditions (7.5%), as detailed in the legend of Table 2. [Here A4-size] Table 2 Distribution and diagnostic yield of imaging modalities, surgical and pathological diagnosis The duration of abdominal pain was significantly longer in the non-appendicitis group, with a median of 24 hours (IQR 12–72), compared with 12 hours (IQR 7–24) in the appendicitis group (p < 0.001). However, the diagnostic odds ratio (dOR) for pain duration was not statistically significant (dOR 0.99, 95% CI 0.99–1.00; p = 0.06; Table 3). [Here A4-size] Table 3 Diagnostic predictors for appendicitis Right lower quadrant pain was the predominant presenting symptom overall, whereas upper abdominal pain was more frequently observed in the non-appendicitis group (5%) than in the appendicitis group (0.4%, p = 0.002). The mean total Alvarado score was significantly higher in patients with appendicitis (8 ± 2) than in those without appendicitis (5 ± 2, p < 0.001). Most appendicitis cases fell within the “probable appendicitis” category (≥7 points) according to the original Alvarado scoring system [12]. At this cut-off, the sensitivity was 77.1%, specificity 71.0%, PPV 88.0%, NPV 53.0%, and overall accuracy 75.5% (Table 4). Table 4 Sensitivity, specificity, positive predictive values, negative predictive values, and accuracy Before surgical consultation, 82.9% of patients (311 of 375) were initially assessed as having “rule out appendicitis” or “acute abdominal pain.” Among these patients, 214 (68.9%) were ultimately confirmed to have appendicitis. 2. Leukocyte count and differentiation Total leukocyte count and neutrophil percentage were significantly higher in the appendicitis group than in the non-appendicitis group (white blood cell count: 16,431 ± 4,892 vs 11,640 ± 4,582 /mm³, p < 0.001; neutrophil percentage: 84% [IQR 78–89] vs 69% [IQR 59–80], p < 0.001). The sensitivity for a white blood cell count ≥12,000 /mm³ was 83.6% and increased to 90.9% when the threshold was lowered to ≥11,000 /mm³. A neutrophil percentage ≥75% yielded a sensitivity of 82.9%. The corresponding PPV(s) were 83.6%, 82.2%, and 85.7%, respectively. However, specificity and NPV(s) for all three parameters were below 65%, with overall diagnostic accuracy ranging from 76% to 79% —Table 4. 3. Urinalysis findings Urine ketone levels differed significantly between the appendicitis and non-appendicitis groups (p = 0.013). The strongest associations with appendicitis were observed at higher ketone levels, with dOR (95% CI) of 3.82 (1.43–10.19, p = 0.007) for 3+ ketonuria and 1.99 (1.12–3.54, p = 0.02) for 4+ ketonuria. When meaningful ketonuria (≥3+) was applied, Cramer’s V correlation coefficient was 0.159 (p = 0.024), indicating a weak but statistically significant association with the final diagnosis (Table 5). Table 5 Correlation between urine ketone and other factors Lowering the urine ketone threshold to ≥2+ increased sensitivity from 41.5% to 50.9%, but was accompanied by a slight reduction in PPV (from 82.0% to 80.5%) and mean dOR (from 2.12 to 2.01). Urine ketone levels showed a moderate correlation with urine sp.gr. (r = 0.346, p < 0.01); however, urine sp.gr., as a surrogate marker of dehydration, did not differ significantly between the appendicitis and non-appendicitis groups (p = 0.218). In contrast, leukocyte esterase demonstrated an inverse association with appendicitis, with dOR (95% CI) of 0.37 (0.15–0.95, p = 0.038) at 3+ and 0.28 (0.11–0.71, p = 0.008) at 4+ levels. 4. Diagnostic factors and ROC analysis At a cut-off score of 7, the Alvarado score demonstrated a high dOR of 8.24 (95% CI 4.92–13.79; p < 0.001), with an area under the AUC of 74.05% (95% CI 68.93–79.16) —Table 6. Table 6 ROC of various screening tools When seven diagnostic conceptual factors were combined, overall diagnostic performance improved substantially. Receiver operating characteristic (ROC) analysis of the combined model yielded an AUC of 86.77% (95% CI 82.56–90.96), superior to the predictive power of the individual Alvarado scores (AUC 80.79%, 95% CI 75.70–85.88) and of each conceptual factor (mean AUC ranged from 56.9-77.62). The ROC curves for the seven diagnostic conceptual factors are shown in Figure 1. Figure 1 ROC curve for combined seven diagnostic conceptual factors The diagnostic contributions of individual factors were as follows: 1. Sex: Female sex was associated with a lower prevalence of appendicitis, with a dOR (95% CI) of 0.42 (0.26–0.68; p < 0.001). 2. Age group: All age groups demonstrated statistically significant dOR (p < 0.05). The highest association was observed in the preadolescent group (10–12.9 years), with a dOR (95% CI) of 6.09 (2.18–17.04; p = 0.001). 3. Abdominal pain location: Upper abdominal pain was associated with a significantly lower likelihood of appendicitis (dOR 0.09, 95% CI 0.01–0.80; p = 0.031). In contrast, right lower quadrant pain showed a borderline association with appendicitis in cases consulted with suspected appendicitis (dOR 1.62, 95% CI 0.98–2.67; p = 0.059). 4. White blood cell count: White blood cell count demonstrated a dOR (95% CI) of 1.25 (1.17–1.33). For screening purposes, sensitivity was highest at a cut-off of 11,000/mm³ (90.9%) and decreased to 83.6% at 12,000/mm³. 5. Neutrophil percentage: As a continuous variable, neutrophil percentage showed a dOR (95% CI) of 1.08 (1.06–1.11). When dichotomized at the Alvarado score cut-off of ≥75%, the association strengthened, with a dOR (95% CI) of 2.80 (2.11–3.73). 6. Urine ketone level: Higher urine ketone levels were significantly associated with appendicitis. A ketone level of 3+ yielded a dOR (95% CI) of 3.82 (1.43–10.19; p = 0.007), while a level of 4+ showed a dOR (95% CI) of 1.99 (1.12–3.54; p = 0.02). The definition of meaningful ketonuria (≥3+) as applied in the Alvarado score also demonstrated diagnostic value, with a dOR (95% CI) of 2.12 (1.27–3.55; p = 0.004). 7. Leukocyte esterase: Increasing leukocyte esterase levels were inversely associated with appendicitis. The dOR (95% CI) was 0.37 (0.15–0.95; p = 0.038) at a level of 3+ and 0.28 (0.11–0.71; p = 0.008) at a level of 4+. 5. Imaging Among the 375 patients included in the study, 264 (70.4%) underwent diagnostic imaging, including plain abdominal radiography, ultrasonography (US), or computed tomography (CT) (Table 2). US was the most frequently used modality (n = 211, 56.3%). The overall imaging utilization rate was higher in the non-appendicitis group than in the appendicitis group (85% vs 65.1%). US contributed significantly to identifying the cause of abdominal pain, appendicitis vs non-appendicitis conditions (52.4% vs 67.0%, p = 0.012). US remained partially useful in cases with equivocal clinical findings, whereas CT was typically the final imaging modality used to establish the diagnosis. Despite routine application of the Alvarado score prior to surgical consultation, more than half of patients ultimately diagnosed with appendicitis required confirmatory imaging (65.1%), underscoring the continued role of imaging in clinical decision-making. 6. Surgical and pathological concordance Among patients who underwent appendectomy (n = 277), concordance between intraoperative surgical diagnoses and histopathological findings demonstrated a moderate level of agreement (κ = 0.461, p < 0.001) (Table 7). Agreement rates for rupture/gangrenous appendicitis, suppurative appendicitis, acute appendicitis, and non-appendicitis were 76.4%, 53.2%, 56.4%, and 11.1%, respectively. Discordance primarily occurred between macroscopic intraoperative assessments and histopathological grading. Excluding five positive exploratory operations (one Meckel’s diverticulitis and four ovarian masses or cysts), negative appendectomy occurred in 18 of 277 appendiceal specimens (6.5%). The pathological diagnoses of non-appendicitis, surgeons had intraoperatively diagnosed acute appendicitis in 88.9%. Of these, 15 out of 16 patients were ultimately found to have lymphoid hyperplasia on histopathological examination (Table 2). When histopathology confirmed acute appendicitis, surgeons classified 30.0% of cases as suppurative and 13.6% as rupture or gangrenous. Conversely, intraoperative diagnoses of acute appendicitis corresponded to 24.5% of histopathologically suppurative cases and 23.6% of rupture or gangrenous cases. Discussion Clinical scoring systems such as the Alvarado score continue to assist decision-making in suspected pediatric appendicitis but should not be applied in isolation. Laboratory parameters, particularly white blood cell count and neutrophil percentage, add diagnostic value but remain nonspecific. Previous studies have reported inconsistent associations between urinalysis findings and appendicitis [ 24 , 25 ]. While ketonuria has been incorporated as a surrogate marker for anorexia in some reports on Alvarado scoring systems [ 12 ], its independent diagnostic contribution remains incompletely defined. In the present study, integration of seven clinical and laboratory factors resulted in a substantially higher diagnostic performance than the Alvarado score alone, as reflected by a higher AUC. These findings support the use of early biological and clinical predictors to guide decisions regarding imaging and surgical consultation, particularly in equivocal presentations. 1. Refinement of urinary biomarkers When refining individual components of the Alvarado score, lowering the urine ketone cut-off from ≥ 3 + to ≥ 2 + increased sensitivity but reduced specificity, PPV, and dOR. These findings support the 2016 Alvarado modification, which defined urine ketone ≥ 3 + as a clinically meaningful diagnostic threshold [ 12 ]. Although urine sp.gr. may be influenced by hydration status—particularly in patients receiving intravenous fluids—urine ketone levels normalize more slowly and may remain interpretable even after resuscitation [ 19 ]. This persistence suggests that ketonuria may serve as a reliable adjunctive biomarker across different phases of patient evaluation. Leukocyte esterase, while not directly associated with appendicitis, remains clinically valuable for excluding alternative diagnoses such as urinary tract infection, urolithiasis, sexually transmitted infection, or autoimmune disease [ 25 ]. Similarly, urine pregnancy testing remains essential in adolescent females. In our cohort, pregnancy testing did not influence appendicitis diagnosis but successfully identified three pregnant patients prior to potential misdiagnosis, underscoring its role in safe clinical practice. 2. Contribution of the seven-factor model Conceptual clinical variables are clinically relevant parameters derived from structured patient characteristics, symptoms, physical findings, and basic laboratory data, reflecting real-world clinical reasoning rather than raw numerical values. The seven evaluated parameters—sex, age group, abdominal pain location, white blood cell count, neutrophil percentage, urine ketone ≥ 3+, and leukocyte esterase level—demonstrated greater discriminative ability than the Alvarado score alone. Appendicitis was less common in adolescent and preschool patients presenting for evaluation, whereas gynecologic conditions accounted for a substantial proportion of non-appendicitis diagnoses among females. In contrast, preadolescent patients, particularly before or early menarche, showed the highest likelihood of appendicitis. Diagnostic differentiation remains challenging in younger children because of atypical symptom presentation and limited cooperation. Although this study demonstrated longer pain duration in the non-appendicitis group and in complicated appendicitis cases with partial antibiotic exposure, the exact duration and location of pain, subjective nausea, and resolution of fever after antipyretic use may reduce the reliability of clinical assessment and contribute to diagnostic uncertainty. These limitations highlight the importance of conceptual clinical variables over raw symptom reporting, particularly when preparing structured data pipelines for clinical use or AI model training [ 5 ]. Consistent with prior literature, commonly used predictors across machine-learning studies include pain characteristics, tenderness, guarding, leukocyte indices, platelet count, hematocrit, hemoglobin, and CRP. A recent systematic review reported average AUC values exceeding 80% across models, with few outliers below 75% [ 1 ]. 3. Role of imaging and implications for AI development Although approximately 35% of appendicitis cases in our cohort were diagnosed clinically, US remained the first-line imaging modality, with CT reserved for equivocal cases, consistent with previous studies [ 12 , 23 , 25 ]. These findings reflect continued reliance on objective diagnostic evidence while emphasizing the need to balance diagnostic certainty with radiation exposure and resource utilization [ 12 , 22 , 23 , 26 , 27 ]. Early decision-making based on a combination of biological and clinical parameters may optimize imaging use and reduce unnecessary radiation. Separating AI training models for clinical diagnosis and radiographic interpretation may allow development of less complex and more reliable algorithms. Early diagnosis and timely intervention are associated with reduced rates of complicated appendicitis and shorter hospital stays. 4. Implications for AI and data pipelines The effectiveness of machine-learning and deep-learning models is strongly influenced by the quality of their input data and the structure of the training pipeline [ 10 ]. Although deep-learning models can autonomously learn complex patterns from unstructured data [ 8 ], their decision-making processes are often difficult to interpret, increasing the risk of misleading conclusions [ 7 ]. Conceptual clinical variables may serve as a bridge toward explainable AI, enabling transparency and traceability of decision-making [ 3 , 10 ]. However, explainable AI still requires models that align closely with clinician experience before responsibility can be appropriately shared [ 10 ]. Heterogeneous data, unnecessary variables, and high-cost parameters such as routine CT, magnetic resonant imaging (MRI), CRP, or ESR may not improve model performance and should be excluded from consensus data pipelines [ 10 ]. Data heterogeneity and variable acquisition protocols must be addressed through standardized preprocessing to enhance predictive accuracy [ 3 ]. Reports evaluating large language models as decision-support tools have demonstrated inconsistent recommendations and operational limitations, underscoring the need for rigorous validation before clinical integration [ 8 ]. A variety of algorithms—including random forests, artificial neural networks, convolutional neural networks, support vector machines, logistic regression, and extreme gradient boosting—have demonstrated superior performance compared with traditional clinical scoring systems [ 3 , 11 ]. No single model is universally optimal; selection of appropriate algorithms and careful hyperparameter tuning remain essential [ 9 , 11 ]. 5. Surgical–pathological concordance The moderate concordance observed between surgical and pathological diagnoses aligns with previous reports. Discrepancies may arise from sampling techniques, microscopic perforations, or fibrin coatings that obscure perforation sites. Although histopathology may reveal microscopic findings not evident intraoperatively, these differences rarely alter surgical management. Future AI developments incorporating advanced imaging analysis may allow preoperative identification of lymphoid hyperplasia and uncomplicated appendicitis amenable to non-operative management, improving patient selection and advancing personalized care. 6. Limitations This study has several limitations. Its retrospective design, single-center setting, and relatively small sample size may limit generalizability. However, the use of complete clinical data, defined urine ketone cut points (2 + and 3+), and seven diagnostic factors provides a structured framework for future model development. In contrast to many recent AI studies that rely on resampling-based training, non-histological diagnoses, or incomplete datasets [ 3 , 4 , 6 , 10 , 11 ], this study benefits from histopathological or radiographic confirmation and minimal missing data, strengthening internal validity. Nonetheless, this work should be regarded as an exploratory “sandbox” for variable selection; advances in imaging technology, evolving diagnostic workflows, and antibiotic-pretreated cases were not addressed and warrant evaluation in future multicenter studies. Conclusion Seven factors—sex, age group, location of abdominal pain, white blood cell count, neutrophil percentage, urine ketone level ≥ 3+, and leukocyte esterase level—collectively improved the accuracy of early diagnosis in pediatric appendicitis and enhanced decision-making for timely US in equivocal cases, outperforming the Alvarado score alone. Among these variables, urinary ketones are particularly valuable because their levels change more slowly following fluid resuscitation, allowing them to serve as a cost-effective and readily accessible indicator even after hydration. Artificial intelligence agents, similar to human clinicians, may generate differing interpretations when applied to complex clinical scenarios. Well-curated, multifaceted clinical data therefore provide a critical foundation for appropriately trained AI models [ 2 ]. Individualized AI approaches that separately enhance US or CT interpretation may optimize resource utilization more effectively than a single, unified machine-learning or deep-learning model that combines clinical, laboratory, and imaging data indiscriminately. Accurate classification of appendicitis severity should incorporate both intraoperative findings and histopathological confirmation. Macroscopically visible perforation identified by the surgeon and microscopic perforation detected by the pathologist remain the most reliable indicators of perforated appendicitis. Future AI developments focused on imaging analysis may enable preoperative exclusion of lymphoid hyperplasia and identification of uncomplicated appendicitis amenable to medical management, thereby improving patient selection and advancing precision care. Learning point The development of reliable AI models for complex clinical applications should begin with careful construction of precise, clinically meaningful datasets. Establishing an appropriate data pipeline before algorithm development is essential to achieving clinician-relevant models that can be trusted and effectively integrated into practice. Abbreviations AI Artificial intelligence BIDIAP Biomarkers for the Diagnosis of Appendicitis in Pediatrics Index CRP C-reactive protein Sp.gr. Specific gravity IQR Interquartile range dOR Diagnostic odds ratio CI Confidence interval PPV Positive predictive value NPV Negative predictive value AUC Area under the curve ROC Receiver operating characteristic US Ultrasonography CT Computed tomography MRI Magnetic resonant imaging ESR Erythrocyte sedimentation rate Declarations Ethics approval and consent to participate This study was approved by the Institutional Review Board of Mahidol University (Certificate of Approval No. COA. MURA2023/553). The requirement for informed consent was waived due to the retrospective nature of the study and the use of anonymized data. Clinical trial number Not applicable. Consent for publication Not applicable. Availability of data and materials The datasets used and/or analyzed during the current study are available from the corresponding author on reasonable request. Competing interests The authors declare that they have no competing interests. Funding This research received no specific grant from any funding agency in the public, commercial, or not-for-profit sectors. Authors’ contributions PT conceived and designed the study, acquired the data, performed the analysis, interpreted the results, drafted the manuscript, and substantively revised it. PT approved the final version of the manuscript for submission. Artificial intelligence–assisted tools were used for language editing only and did not influence the study design, data analysis, or interpretation. Acknowledgements The author would like to thank Ms. Suraida Aeesoa, Medical Statistician at the Department of Surgery, for assistance with the final data analysis. References Chekmeyan M, Liu SH. Artificial intelligence for the diagnosis of pediatric appendicitis: A systematic review. Am J Emerg Med. 2025 Jun;92:18-31. doi: 10.1016/j.ajem.2025.02.023. Epub 2025 Feb 17. PMID: 40048888. Shikha A, Kasem A. Decoding Pediatric Appendicitis: Unraveling Complexity with Artificial Intelligence and Evolving Management Insights [Internet]. Appendicitis - Current Insights. IntechOpen; 2024. Available from: http://dx.doi.org/10.5772/intechopen.1008318 Akbulut S, Kucukakcali Z, Colak C. Artificial intelligence in acute appendicitis: A comprehensive review of machine learning and deep learning applications. World J Gastroenterol. 2025 Nov 21;31(43):112000. doi: 10.3748/wjg.v31.i43.112000. PMID: 41358178; PMCID: PMC12678916. Marcinkevičs R, Sokol K, Paulraj A, Hilbert MA, Rimili V, Wellmann S, Knorr C, Reingruber B, Vogt JE, Reis Wolfertstetter P. External validation of predictive models for diagnosis, management and severity of pediatric appendicitis. Front Pediatr. 2025 Aug 29;13:1587488. doi: 10.3389/fped.2025.1587488. PMID: 40948509; PMCID: PMC12426264. Kendall J, Gaspar G, Berger D, Levman J. Machine Learning and Feature Selection in Pediatric Appendicitis. Tomography. 2025 Aug 13;11(8):90. doi: 10.3390/tomography11080090. PMID: 40863881; PMCID: PMC12390108. Marcinkevics R, Reis Wolfertstetter P, Wellmann S, Knorr C, Vogt JE. Using Machine Learning to Predict the Diagnosis, Management and Severity of Pediatric Appendicitis. Front Pediatr. 2021 Apr 29;9:662183. doi: 10.3389/fped.2021.662183. PMID: 33996697; PMCID: PMC8116489. Klimiene, Ugne & Marcinkevics, Ricards & Reis Wolfertstetter, Patricia & Ozkan, Ece & Paschke, Alyssia & Niederberger, David & Wellmann, Sven & Knorr, Christian & Vogt, Julia. (2022). Multiview Concept Bottleneck Models Applied to Diagnosing Pediatric Appendicitis. Ersahin K, Sanduleanu S, Thulasi Seetha S, Bremm J, Abbasli C, Zimmer C, Damer T, Kottlors J, Goertz L, Bruns C, Maintz D, Abdullayev N. From Bedside to Bot-Side: Artificial Intelligence in Emergency Appendicitis Management. Life (Basel). 2025 Sep 1;15(9):1387. doi: 10.3390/life15091387. PMID: 41010329; PMCID: PMC12470868. Chadaga K, Khanna V, Prabhu S, Sampathila N, Chadaga R, Umakanth S, Bhat D, Swathi KS, Kamath R. An interpretable and transparent machine learning framework for appendicitis detection in pediatric patients. Sci Rep. 2024 Oct 18;14(1):24454. doi: 10.1038/s41598-024-75896-y. Erratum in: Sci Rep. 2025 Jan 22;15(1):2841. doi: 10.1038/s41598-025-86494-x. PMID: 39424647; PMCID: PMC11489819. Erman A, Ferreira J, Ashour WA, Guadagno E, St-Louis E, Emil S, Cheung J, Poenaru D. Machine-learning-assisted Preoperative Prediction of Pediatric Appendicitis Severity. J Pediatr Surg. 2025 Jun;60(6):162151. doi: 10.1016/j.jpedsurg.2024.162151. Epub 2025 Jan 13. PMID: 39855986. Issaiy M, Zarei D, Saghazadeh A. Artificial Intelligence and Acute Appendicitis: A Systematic Review of Diagnostic and Prognostic Models. World J Emerg Surg. 2023 Dec 19;18(1):59. doi: 10.1186/s13017-023-00527-2. PMID: 38114983; PMCID: PMC10729387. Alvarado A. How to improve the clinical diagnosis of acute appendicitis in resource limited settings. World J Emerg Surg. 2016;11:16. doi: 10.1186/s13017-016-0071-8. PMID: 27118990; PMCID: PMC4845369. Gudjonsdottir J, Marklund E, Hagander L, Salö M. Clinical Prediction Scores for Pediatric Appendicitis. Eur J Pediatr Surg. 2021;31(3):252-260. doi: 10.1055/s-0040-1710534. Epub 2020 May 26. PMID: 32455443. Arredondo Montero J, Bardají Pascual C, Antona G, Ros Briones R, López-Andrés N, Martín-Calvo N. The BIDIAP index: a clinical, analytical and ultrasonographic score for the diagnosis of acute appendicitis in children. Pediatr Surg Int. 2023;39(1):175. doi: 10.1007/s00383-023-05463-5. PMID: 37038002; PMCID: PMC10085908. Türkmenoğlu Y, Kaçar A, Duras E, Kök S, Gözübüyük AA, Arat C. The Role of Alvarado and Pediatric Appendicitis Score in Acute Appendicitis in Children. J Pediatr Res. 2020;7(3):192-198. doi: 10.4274/jpr.galenos.2019.62582. Sağ S, Basar D, Yurdadoğan F, Pehlivan Y, Elemen L. Comparison of Appendicitis Scoring Systems in Childhood Appendicitis. Turk Arch Pediatr. 2022;57(5):532-537. doi: 10.5152/TurkArchPediatr.2022.22076. PMID: 36062441; PMCID: PMC9524470. Arredondo Montero J, Bronte Anaut M, Bardají Pascual C, Antona G, López-Andrés N, Martín-Calvo N. Alterations and diagnostic performance of capillary ketonemia in pediatric acute appendicitis: a pilot study. Pediatr Surg Int. 2022;39(1):44. doi: 10.1007/s00383-022-05332-7. PMID: 36495332; PMCID: PMC9741565. O'Donohoe PB, Kessler R, Beattie TF. Exploring the clinical utility of blood ketone levels in the emergency department assessment of paediatric patients. Emerg Med J. 2006;23(10):783-7. doi: 10.1136/emj.2006.035758. PMID: 16988307; PMCID: PMC2579600. Laffel L. Ketone bodies: a review of physiology, pathophysiology and application of monitoring to diabetes. Diabetes Metab Res Rev 1999;15:412–26. Stringer MD. Acute appendicitis. J Paediatr Child Health. 2017;53(11):1071-1076. doi: 10.1111/jpc.13737. Epub 2017 Oct 17. PMID: 29044790. Nissen M, Tröbs RB. The lymphocyte-to-monocyte ratio may distinguish complicated from non-complicated pediatric appendicitis: A retrospective study and literature review. Pediatr Neonatol. 2022;63(2):146-153. doi: 10.1016/j.pedneo.2021.08.018. Epub 2021 Oct 27. PMID: 34799285. Carpenter JL, Orth RC, Zhang W, Lopez ME, Mangona KL, Guillerman RP. Diagnostic performance of US for differentiating perforated from Nonperforated pediatric appendicitis: a prospective cohort study. Radiology 2017; 282: 835–41. doi: https:// doi. org/ 10. 1148/radiol. 2016160175 Rawolle T, Reismann M, Minderjahn MI, Bassir C, Hauptmann K, Rothe K, Reismann J. Sonographic differentiation of complicated from uncomplicated appendicitis. Br J Radiol. 2019;92(1099):20190102. doi: 10.1259/bjr.20190102. Epub 2019 May 29. PMID: 31112397; PMCID: PMC6636276. Biehl CM, Elliver M, Gudjonsdottir J, Salö M. Utility of Urine Dipstick Testing in Pediatric Appendicitis: Assessing its Role in Identifying Complicated Cases and Retrocecal Appendicitis. Eur J Pediatr Surg. 2024 Dec 19. doi: 10.1055/a-2490-1156. Epub ahead of print. PMID: 39701137. Chen CY, Zhao LL, Lin YR, Wu KH, Wu HP. Different urinalysis appearances in children with simple and perforated appendicitis. Am J Emerg Med. 2013 Nov;31(11):1560-3. doi: 10.1016/j.ajem.2013.06.027. Epub 2013 Sep 20. PMID: 24055480 Benabbas R, Hanna M, Shah J, Sinert R. Diagnostic Accuracy of History, Physical Examination, Laboratory Tests, and Point-of-care Ultrasound for Pediatric Acute Appendicitis in the Emergency Department: A Systematic Review and Meta-analysis. Acad Emerg Med. 2017;24(5):523-551. doi: 10.1111/acem.13181. PMID: 28214369. Kim MS, Kwon H-J, Kang KA, Do I-G, Park H-J, Kim EY, et al. Diagnostic performance and useful findings of ultrasound reevaluation for patients with equivocal CT features of acute appendicitis. Br J Radiol 2018; 91: 20170529 Tables Tables 1 to 7 are available in the supplementary files section Additional Declarations No competing interests reported. Supplementary Files Table1.docx Table2.docx Table3.docx Table4.docx Table5.docx Table6.docx Table7.docx Cite Share Download PDF Status: Posted Version 1 posted You are reading this latest preprint version Research Square lets you share your work early, gain feedback from the community, and start making changes to your manuscript prior to peer review in a journal. As a division of Research Square Company, we’re committed to making research communication faster, fairer, and more useful. We do this by developing innovative software and high quality services for the global research community. Our growing team is made up of researchers and industry professionals working together to solve the most critical problems facing scientific publishing. Also discoverable on Platform About Our Team In Review Editorial Policies Advisory Board Help Center Resources Author Services Accessibility API Access RSS feed Manage Cookie Preferences © Research Square 2026 | ISSN 2693-5015 (online) Privacy Policy Terms of Service Do Not Sell My Personal Information {"props":{"pageProps":{"initialData":{"identity":"rs-8666814","acceptedTermsAndConditions":true,"allowDirectSubmit":true,"archivedVersions":[],"articleType":"Research Article","associatedPublications":[],"authors":[{"id":583870554,"identity":"afa9a52f-58a2-4d1b-8872-cf6caadfbc5e","order_by":0,"name":"PORNSRI THANACHATCHAIRATTANA","email":"data:image/png;base64,iVBORw0KGgoAAAANSUhEUgAAAZAAAAAyAQMAAABI0h/eAAAABlBMVEX///8AAABVwtN+AAAACXBIWXMAAA7EAAAOxAGVKw4bAAAAsklEQVRIiWNgGAWjYBACxmYGNgYGAxsGBgkStaSRoAUIgFoYDpOghbmd+dnDHwXnE/tnNx98wFBjE02Ew9jMjXkMbifOuHMs2YDhWFpuA2EtDGbSDEAtDTdyzCQYGw4To4X9m+QPg3OJ80nQwmMmwWNwIHEDKVrKgX5JNt54Iy3ZIIEYvxj2H9/28McfO9l5N5IPPvhQY0OEFqgKRzCdQEg5CMhDaXtiFI+CUTAKRsEIBQBjRT3/y5eeFQAAAABJRU5ErkJggg==","orcid":"","institution":"Mahidol University","correspondingAuthor":true,"prefix":"","firstName":"PORNSRI","middleName":"","lastName":"THANACHATCHAIRATTANA","suffix":""}],"badges":[],"createdAt":"2026-01-22 07:54:33","currentVersionCode":1,"declarations":"","doi":"10.21203/rs.3.rs-8666814/v1","doiUrl":"https://doi.org/10.21203/rs.3.rs-8666814/v1","draftVersion":[],"editorialEvents":[],"editorialNote":"","failedWorkflow":false,"files":[{"id":101881440,"identity":"a5f4954d-45a4-4f82-a96c-d30e8ff16c78","added_by":"auto","created_at":"2026-02-04 15:12:10","extension":"png","order_by":1,"title":"Figure 1","display":"","copyAsset":false,"role":"figure","size":10150,"visible":true,"origin":"","legend":"\u003cp\u003eROC curve for combined seven diagnostic conceptual factors\u003c/p\u003e","description":"","filename":"1.png","url":"https://assets-eu.researchsquare.com/files/rs-8666814/v1/127d032a9cebbcf1eef3300e.png"},{"id":102749309,"identity":"ef820c22-aaaa-4908-9765-6c18a073b1bd","added_by":"auto","created_at":"2026-02-16 09:12:22","extension":"pdf","order_by":0,"title":"","display":"","copyAsset":false,"role":"manuscript-pdf","size":805978,"visible":true,"origin":"","legend":"","description":"","filename":"manuscript.pdf","url":"https://assets-eu.researchsquare.com/files/rs-8666814/v1/8fcc794d-1b5b-4f64-a064-9040e4e98662.pdf"},{"id":101790252,"identity":"05ad4969-9339-4416-9894-3ceec880485f","added_by":"auto","created_at":"2026-02-03 16:04:28","extension":"docx","order_by":2,"title":"","display":"","copyAsset":false,"role":"supplement","size":37150,"visible":true,"origin":"","legend":"","description":"","filename":"Table1.docx","url":"https://assets-eu.researchsquare.com/files/rs-8666814/v1/96082eb0987e3b8c1f64f330.docx"},{"id":101881100,"identity":"bae42d23-3805-42ae-b822-b3fdf4dab817","added_by":"auto","created_at":"2026-02-04 15:09:41","extension":"docx","order_by":3,"title":"","display":"","copyAsset":false,"role":"supplement","size":32812,"visible":true,"origin":"","legend":"","description":"","filename":"Table2.docx","url":"https://assets-eu.researchsquare.com/files/rs-8666814/v1/0c918cb5d2369e3df5b4f721.docx"},{"id":101790256,"identity":"2dc83c9c-ab3f-4902-bdeb-1385f81efe5f","added_by":"auto","created_at":"2026-02-03 16:04:29","extension":"docx","order_by":4,"title":"","display":"","copyAsset":false,"role":"supplement","size":32993,"visible":true,"origin":"","legend":"","description":"","filename":"Table3.docx","url":"https://assets-eu.researchsquare.com/files/rs-8666814/v1/23b372bb7d7421d2e0b81dad.docx"},{"id":101880664,"identity":"f0ec2ed9-fc9c-4425-9e91-9e5a9634dfa3","added_by":"auto","created_at":"2026-02-04 15:04:56","extension":"docx","order_by":5,"title":"","display":"","copyAsset":false,"role":"supplement","size":30208,"visible":true,"origin":"","legend":"","description":"","filename":"Table4.docx","url":"https://assets-eu.researchsquare.com/files/rs-8666814/v1/b3fa49e3bafcf9b9216c80e0.docx"},{"id":101790255,"identity":"a69f07d5-1a8f-4c57-830c-20a15d592533","added_by":"auto","created_at":"2026-02-03 16:04:28","extension":"docx","order_by":6,"title":"","display":"","copyAsset":false,"role":"supplement","size":29414,"visible":true,"origin":"","legend":"","description":"","filename":"Table5.docx","url":"https://assets-eu.researchsquare.com/files/rs-8666814/v1/5ad9552bb70ef3a4afd4b81f.docx"},{"id":101880553,"identity":"44090b9a-2d80-4d24-83d8-21890117d7d0","added_by":"auto","created_at":"2026-02-04 15:03:33","extension":"docx","order_by":7,"title":"","display":"","copyAsset":false,"role":"supplement","size":29804,"visible":true,"origin":"","legend":"","description":"","filename":"Table6.docx","url":"https://assets-eu.researchsquare.com/files/rs-8666814/v1/530143fd567c1d196a55f70f.docx"},{"id":101790253,"identity":"32ba1061-84d3-41bc-ac58-d9ca8cc0241a","added_by":"auto","created_at":"2026-02-03 16:04:28","extension":"docx","order_by":8,"title":"","display":"","copyAsset":false,"role":"supplement","size":30523,"visible":true,"origin":"","legend":"","description":"","filename":"Table7.docx","url":"https://assets-eu.researchsquare.com/files/rs-8666814/v1/fc649d6d2c79246a6e5dd75c.docx"}],"financialInterests":"No competing interests reported.","formattedTitle":"Conceptual Clinical Variables Enhancing the Alvarado Score in Pediatric Appendicitis: Lessons for Artificial Intelligence Models","fulltext":[{"header":"Background","content":"\u003cp\u003ePediatric appendicitis remains a common surgical emergency in which timely and accurate diagnosis is essential to prevent delayed treatment and unnecessary appendectomy. Despite advances in diagnostic strategies and growing interest in artificial intelligence (AI)\u0026ndash;based decision support [1-11], reliable clinical predictors continue to play a central role in pediatric practice.\u003c/p\u003e\n\u003cp\u003eThe Alvarado score is one of the most widely used clinical scoring systems for suspected appendicitis [12]. Over time, several modifications and alternative tools\u0026mdash;including the Pediatric Appendicitis Score, Appendicitis Inflammatory Response Score, Pediatric Appendicitis Risk Calculator, and the Biomarkers for the Diagnosis of Appendicitis in Pediatrics (BIDIAP) index [13-16]\u0026mdash;have been developed to improve diagnostic accuracy and support early decision-making. However, none has consistently demonstrated high performance in pediatric populations. This limitation highlights the ongoing need to refine existing clinical variables rather than relying solely on new scoring frameworks or advanced technologies.\u003c/p\u003e\n\u003cp\u003eMany components incorporated into appendicitis scoring systems are inherently subjective or variable, such as anorexia, nausea, pain localization, or a reported history of fever that may be transiently suppressed by antipyretic use at the time of presentation. These factors may reduce diagnostic reliability and complicate both clinical decision-making and the development of robust predictive or AI-based models. Consequently, objective and physiologically meaningful variables that can enhance diagnostic sensitivity warrant further investigation.\u003c/p\u003e\n\u003cp\u003eUrinary ketones represent a potential adjunctive marker in this context. Ketogenesis reflects a catabolic metabolic state and occurs more rapidly in children than in adults, often within 12\u0026ndash;24 hours vs 24 hours of reduced oral intake [17-19]. Urinary ketones may therefore serve as a low-cost and readily available adjunctive marker, even when serum ketone levels have normalized [19]. In the 2016 modification of the Alvarado score, ketonuria was incorporated as a surrogate marker for anorexia; however, the diagnostic contribution of specific urinary ketone cut-off levels has not been clearly established [12]. Given their low cost, rapid availability, and physiological relevance, urinary ketones may provide additional diagnostic value in pediatric appendicitis.\u003c/p\u003e\n\u003cp\u003eConventional inflammatory markers, such as white blood cell count and C-reactive protein (CRP), are routinely used in the evaluation of suspected appendicitis but lack reliable cut-off values and may remain normal in confirmed cases [20]. Other laboratory indices\u0026mdash;lymphocyte-to-monocyte ratio [21], procalcitonin [17,21], and the systemic immune-inflammation index used in the BIDIAP model [14]\u0026nbsp;\u0026mdash; reflect systemic inflammatory responses rather than direct anatomical pathology. Although imaging plays a critical role in identifying anatomical abnormalities, excluding alternative diagnoses, and guiding operative planning [12,20,22,23], it should complement rather than replace sound clinical judgment, and unnecessary imaging should be avoided in pediatric populations [13].\u003c/p\u003e\n\u003cp\u003eThe aims of this study were (1) to evaluate the diagnostic performance of urinary ketone cut-off levels at 2+ and 3+ within the Alvarado scoring system, and (2) to identify additional predictive factors that enhance the sensitivity and accuracy of the Alvarado score and may serve as meaningful variables for future next-generation AI research. Given that data quality and variable selection critically influence the learning process of AI models, establishing a robust clinical data pipeline is essential [3,10].\u003c/p\u003e"},{"header":"Method","content":"\u003cp\u003e\u003cstrong\u003e1.\u0026nbsp; \u0026nbsp;\u0026nbsp;Study design and setting\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eThis retrospective study included patients aged 2–18 years who were referred for surgical consultation by emergency physicians or pediatricians between January 2014 and June 2024.\u0026nbsp;All patients underwent preoperative complete blood count testing and automated urinalysis. Patients were excluded if they had undergone interval appendectomy, had incomplete data, or were receiving ketogenic diets or medications known to affect ketone levels (e.g., high-dose vitamin C, levodopa, valproate, or non-steroidal anti-inflammatory drugs). None of the included patients had anorexia nervosa, bulimia, alcohol dependence, diabetes mellitus, cortisol deficiency, chronic kidney disease, or glycogen storage disease.\u003c/p\u003e\n\u003cp\u003eKey clinical characteristics and basic laboratory parameters relevant to the diagnosis of appendicitis were grouped as predictive factors based on the components of the Alvarado score [12].\u0026nbsp;Age groups were defined according to disease incidence patterns and physiological developmental stages [2].\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003e2.\u0026nbsp; \u0026nbsp;\u0026nbsp;Data collection\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eMedical records were reviewed for each Alvarado score component, final diagnosis, operative findings, and histopathological reports. Patients managed non-operatively were considered to have a normal appendix if they showed spontaneous clinical improvement, remained asymptomatic at follow-up, or had imaging findings consistent with a normal appendix or an alternative diagnosis.\u003c/p\u003e\n\u003cp\u003eThe Alvarado score was routinely applied across all age groups. Body temperature was measured tympanically, and fever was defined as ≥37.5 °C. Both nausea and vomiting were included as positive scoring items. When discrepancies existed among medical records regarding symptom details, the predominant clinical assessment was used. Rebound tenderness included percussion tenderness and classic signs such as Blumberg, Rovsing, obturator, and Markle signs. A white blood cell count ≥11,000/mm³ and a neutrophil percentage ≥75% were considered positive laboratory thresholds.\u003c/p\u003e\n\u003cp\u003eThe duration of abdominal pain was excluded from analysis because of its marked unreliability in young children. Similarly, fasting duration and the timing of blood and urine examinations relative to symptom onset were not analyzed, as these factors could introduce significant variability and bias.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003e3.\u0026nbsp; \u0026nbsp;\u0026nbsp;Urinalysis\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eAutomated urinalysis was performed using the Cobas® U601 and U411 analyzers (Roche™), with evaluation focused on urine specific gravity (sp.gr.), ketones (predominantly acetoacetate), and leukocyte esterase. A normal serum ketone level (\u0026lt;0.5 mmol/L) corresponds to a urine ketone result of 1+, whereas hyperketonemia (\u0026gt;1.0 mmol/L) is approximately equivalent to a urine ketone level of 2+ (corresponding to a serum level of approximately 1.5 mmol/L) [4,13].\u003c/p\u003e\n\u003cp\u003eUrine ketone levels ≥2+\u0026nbsp;were examined in a predefined subanalysis, while levels ≥3+\u0026nbsp;were defined as “meaningful ketonuria,” corresponding to an estimated serum ketone concentration of approximately 5\u0026nbsp;mmol/L (≈15\u0026nbsp;mg/dL), in accordance with the 2016\u0026nbsp;modification of the Alvarado scoring system [6].\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003e4.\u0026nbsp; \u0026nbsp;\u0026nbsp;Surgical and pathological diagnosis\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eSurgical decisions were based on clinical judgment and imaging findings consistent with appendicitis. Histopathological results were compared with intraoperative findings. The non-appendicitis group included surgical specimens demonstrating normal histology or lymphoid hyperplasia.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003e5.\u0026nbsp; \u0026nbsp;\u0026nbsp;Statistical analysis\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eStatistical analyses were performed using SPSS software version 18.0 (SPSS Inc., Chicago, IL, USA). Continuous variables with non-normal distributions were summarized as median and interquartile range (IQR), whereas categorical variables were expressed as percentages. Comparative analyses were conducted using the Student’s t test, Mann–Whitney U test, and Fisher’s exact test, as appropriate. Agreement and association were assessed using intraclass correlation coefficients, Pearson correlation, and Cramer’s V.\u003c/p\u003e\n\u003cp\u003eDiagnostic odds ratios (dORs) with 95% confidence intervals (CIs) were calculated to assess the strength of individual predictors. Sensitivity, specificity, positive predictive value (PPV), negative predictive value (NPV), overall accuracy, and the area under the receiver operating characteristic curve (AUC) with 95% CIs were estimated using logistic regression models. A two-sided p value \u0026lt;0.05 was considered statistically significant.\u003c/p\u003e"},{"header":"Results","content":"\u003cp\u003e\u003cstrong\u003e1. \u0026nbsp; \u0026nbsp;Demographic and clinical characteristics\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eA total of 375 patients were included in the analysis, of whom 275 were diagnosed with appendicitis (148 males and 127 females) and 100 with non-appendicitis (33 males and 67 females). Age distribution was categorized into preschool (2\u0026ndash;5.9 years), childhood (6\u0026ndash;9.9 years), preadolescence (10\u0026ndash;12.9 years), early adolescence (13\u0026ndash;15.9 years), and adolescence (16\u0026ndash;17.9 years), as summarized in Table 1.\u0026nbsp;\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003e[Here A4-size] Table 1\u0026nbsp;\u003c/strong\u003eDemographic distribution in appendicitis and non-appendicitis groups\u003c/p\u003e\n\u003cp\u003eAppendicitis was more prevalent among males (53.8%) and among children aged \u0026ge;6 years, with the highest incidence observed during preadolescence (before or early menarche in females). In contrast, non-appendicitis was more common among females (67%), particularly in the preschool and adolescent age groups (postmenarche). Across all patients, diagnosis other than appendicitis included non-detectable pathology (8.8%), other gastrointestinal conditions (10.4%), and non-gastrointestinal conditions (7.5%), as detailed in the legend of Table 2.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003e[Here A4-size] Table 2\u0026nbsp;\u003c/strong\u003eDistribution and diagnostic yield of imaging modalities, surgical and pathological diagnosis\u003cstrong\u003e\u0026nbsp;\u0026nbsp;\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eThe duration of abdominal pain was significantly longer in the non-appendicitis group, with a median of 24 hours (IQR 12\u0026ndash;72), compared with 12 hours (IQR 7\u0026ndash;24) in the appendicitis group (p \u0026lt; 0.001). However, the diagnostic odds ratio (dOR) for pain duration was not statistically significant (dOR 0.99, 95% CI 0.99\u0026ndash;1.00; p = 0.06; Table 3).\u0026nbsp;\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003e[Here A4-size] Table 3\u0026nbsp;\u003c/strong\u003eDiagnostic predictors for appendicitis\u003c/p\u003e\n\u003cp\u003eRight lower quadrant pain was the predominant presenting symptom overall, whereas upper abdominal pain was more frequently observed in the non-appendicitis group (5%) than in the appendicitis group (0.4%, p = 0.002).\u003c/p\u003e\n\u003cp\u003eThe mean total Alvarado score was significantly higher in patients with appendicitis (8 \u0026plusmn; 2) than in those without appendicitis (5 \u0026plusmn; 2, p \u0026lt; 0.001). Most appendicitis cases fell within the \u0026ldquo;probable appendicitis\u0026rdquo; category (\u0026ge;7 points) according to the original Alvarado scoring system [12]. At this cut-off, the sensitivity was 77.1%, specificity 71.0%, PPV 88.0%, NPV 53.0%, and overall accuracy 75.5% (Table 4).\u0026nbsp;\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eTable 4\u003c/strong\u003e Sensitivity, specificity, positive predictive values, negative predictive values, and accuracy\u003c/p\u003e\n\u003cp\u003eBefore surgical consultation, 82.9% of patients (311 of 375) were initially assessed as having \u0026ldquo;rule out appendicitis\u0026rdquo; or \u0026ldquo;acute abdominal pain.\u0026rdquo; Among these patients, 214 (68.9%) were ultimately confirmed to have appendicitis.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003e2. \u0026nbsp; \u0026nbsp;Leukocyte count and differentiation\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eTotal leukocyte count and neutrophil percentage were significantly higher in the appendicitis group than in the non-appendicitis group (white blood cell count: 16,431 \u0026plusmn; 4,892 vs 11,640 \u0026plusmn; 4,582 /mm\u0026sup3;, p \u0026lt; 0.001; neutrophil percentage: 84% [IQR 78\u0026ndash;89] vs 69% [IQR 59\u0026ndash;80], p \u0026lt; 0.001).\u003c/p\u003e\n\u003cp\u003eThe sensitivity for a white blood cell count \u0026ge;12,000 /mm\u0026sup3; was 83.6% and increased to 90.9% when the threshold was lowered to \u0026ge;11,000 /mm\u0026sup3;. A neutrophil percentage \u0026ge;75% yielded a sensitivity of 82.9%. The corresponding PPV(s) were 83.6%, 82.2%, and 85.7%, respectively. However, specificity and NPV(s) for all three parameters were below 65%, with overall diagnostic accuracy ranging from 76% to 79% \u0026mdash;Table 4.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003e3. \u0026nbsp; \u0026nbsp;Urinalysis findings\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eUrine ketone levels differed significantly between the appendicitis and non-appendicitis groups (p = 0.013). The strongest associations with appendicitis were observed at higher ketone levels, with dOR (95% CI) of 3.82 (1.43\u0026ndash;10.19, p = 0.007) for 3+ ketonuria and 1.99 (1.12\u0026ndash;3.54, p = 0.02) for 4+ ketonuria. When meaningful ketonuria (\u0026ge;3+) was applied, Cramer\u0026rsquo;s V correlation coefficient was 0.159 (p = 0.024), indicating a weak but statistically significant association with the final diagnosis (Table 5).\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eTable 5\u0026nbsp;\u003c/strong\u003eCorrelation between urine ketone and other factors\u003c/p\u003e\n\u003cp\u003eLowering the urine ketone threshold to \u0026ge;2+ increased sensitivity from 41.5% to 50.9%, but was accompanied by a slight reduction in PPV (from 82.0% to 80.5%) and mean dOR (from 2.12 to 2.01).\u003c/p\u003e\n\u003cp\u003eUrine ketone levels showed a moderate correlation with urine sp.gr. (r = 0.346, p \u0026lt; 0.01); however, urine sp.gr., as a surrogate marker of dehydration, did not differ significantly between the appendicitis and non-appendicitis groups (p = 0.218). In contrast, leukocyte esterase demonstrated an inverse association with appendicitis, with dOR (95% CI) of 0.37 (0.15\u0026ndash;0.95, p = 0.038) at 3+ and 0.28 (0.11\u0026ndash;0.71, p = 0.008) at 4+ levels.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003e4. \u0026nbsp; \u0026nbsp;Diagnostic factors and ROC analysis\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eAt a cut-off score of 7, the Alvarado score demonstrated a high dOR of 8.24 (95% CI 4.92\u0026ndash;13.79; p \u0026lt; 0.001), with an area under the AUC of 74.05% (95% CI 68.93\u0026ndash;79.16) \u0026mdash;Table 6.\u003c/p\u003e\n\u003cp\u003e\u0026nbsp;\u003cstrong\u003eTable 6\u003c/strong\u003e ROC of various screening tools\u003c/p\u003e\n\u003cp\u003eWhen seven diagnostic conceptual factors were combined, overall diagnostic performance improved substantially. Receiver operating characteristic (ROC) analysis of the combined model yielded an AUC of 86.77% (95% CI 82.56\u0026ndash;90.96), superior to the predictive power of the individual Alvarado scores (AUC 80.79%, 95% CI 75.70\u0026ndash;85.88) and of each conceptual factor (mean AUC ranged from 56.9-77.62). The ROC curves for the seven diagnostic conceptual factors are shown in Figure 1.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eFigure 1\u003c/strong\u003e ROC curve for combined seven diagnostic conceptual factors\u003c/p\u003e\n\u003cp\u003eThe diagnostic contributions of individual factors were as follows:\u003c/p\u003e\n\u003cp\u003e1. \u003cstrong\u003eSex:\u003c/strong\u003e Female sex was associated with a lower prevalence of appendicitis, with a dOR (95% CI) of 0.42 (0.26\u0026ndash;0.68; p \u0026lt; 0.001).\u003c/p\u003e\n\u003cp\u003e2. \u003cstrong\u003eAge group:\u003c/strong\u003e All age groups demonstrated statistically significant dOR (p \u0026lt; 0.05). The highest association was observed in the preadolescent group (10\u0026ndash;12.9 years), with a dOR (95% CI) of 6.09 (2.18\u0026ndash;17.04; p = 0.001).\u003c/p\u003e\n\u003cp\u003e3. \u003cstrong\u003eAbdominal pain location:\u003c/strong\u003e Upper abdominal pain was associated with a significantly lower likelihood of appendicitis (dOR 0.09, 95% CI 0.01\u0026ndash;0.80; p = 0.031). In contrast, right lower quadrant pain showed a borderline association with appendicitis in cases consulted with suspected appendicitis (dOR 1.62, 95% CI 0.98\u0026ndash;2.67; p = 0.059).\u003c/p\u003e\n\u003cp\u003e4. \u003cstrong\u003eWhite blood cell count:\u003c/strong\u003e White blood cell count demonstrated a dOR (95% CI) of 1.25 (1.17\u0026ndash;1.33). For screening purposes, sensitivity was highest at a cut-off of 11,000/mm\u0026sup3; (90.9%) and decreased to 83.6% at 12,000/mm\u0026sup3;.\u003c/p\u003e\n\u003cp\u003e5. \u003cstrong\u003eNeutrophil percentage:\u003c/strong\u003e As a continuous variable, neutrophil percentage showed a dOR (95% CI) of 1.08 (1.06\u0026ndash;1.11). When dichotomized at the Alvarado score cut-off of \u0026ge;75%, the association strengthened, with a dOR (95% CI) of 2.80 (2.11\u0026ndash;3.73).\u003c/p\u003e\n\u003cp\u003e6. \u003cstrong\u003eUrine ketone level:\u003c/strong\u003e Higher urine ketone levels were significantly associated with appendicitis. A ketone level of 3+ yielded a dOR (95% CI) of 3.82 (1.43\u0026ndash;10.19; p = 0.007), while a level of 4+ showed a dOR (95% CI) of 1.99 (1.12\u0026ndash;3.54; p = 0.02). The definition of meaningful ketonuria (\u0026ge;3+) as applied in the Alvarado score also demonstrated diagnostic value, with a dOR (95% CI) of 2.12 (1.27\u0026ndash;3.55; p = 0.004).\u003c/p\u003e\n\u003cp\u003e7. \u003cstrong\u003eLeukocyte esterase:\u003c/strong\u003e Increasing leukocyte esterase levels were inversely associated with appendicitis. The dOR (95% CI) was 0.37 (0.15\u0026ndash;0.95; p = 0.038) at a level of 3+ and 0.28 (0.11\u0026ndash;0.71; p = 0.008) at a level of 4+.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003e5. \u0026nbsp; \u0026nbsp;Imaging\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eAmong the 375 patients included in the study, 264 (70.4%) underwent diagnostic imaging, including plain abdominal radiography, ultrasonography (US), or computed tomography (CT) (Table 2). US was the most frequently used modality (n = 211, 56.3%). The overall imaging utilization rate was higher in the non-appendicitis group than in the appendicitis group (85% vs 65.1%).\u003c/p\u003e\n\u003cp\u003eUS contributed significantly to identifying the cause of abdominal pain, appendicitis vs non-appendicitis conditions (52.4% vs 67.0%, p = 0.012). US remained partially useful in cases with equivocal clinical findings, whereas CT was typically the final imaging modality used to establish the diagnosis.\u0026nbsp;\u003c/p\u003e\n\u003cp\u003eDespite routine application of the Alvarado score prior to surgical consultation, more than half of patients ultimately diagnosed with appendicitis required confirmatory imaging (65.1%), underscoring the continued role of imaging in clinical decision-making.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003e6. \u0026nbsp; \u0026nbsp;Surgical and pathological concordance\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eAmong patients who underwent appendectomy (n = 277), concordance between intraoperative surgical diagnoses and histopathological findings demonstrated a moderate level of agreement (\u0026kappa; = 0.461, p \u0026lt; 0.001) (Table 7). Agreement rates for rupture/gangrenous appendicitis, suppurative appendicitis, acute appendicitis, and non-appendicitis were 76.4%, 53.2%, 56.4%, and 11.1%, respectively. Discordance primarily occurred between macroscopic intraoperative assessments and histopathological grading.\u003c/p\u003e\n\u003cp\u003eExcluding five positive exploratory operations (one Meckel\u0026rsquo;s diverticulitis and four ovarian masses or cysts), negative appendectomy occurred in 18 of 277 appendiceal specimens (6.5%). The pathological diagnoses of non-appendicitis, surgeons had intraoperatively diagnosed acute appendicitis in 88.9%. Of these, 15 out of 16 patients were ultimately found to have lymphoid hyperplasia on histopathological examination (Table 2).\u0026nbsp;\u003c/p\u003e\n\u003cp\u003eWhen histopathology confirmed acute appendicitis, surgeons classified 30.0% of cases as suppurative and 13.6% as rupture or gangrenous. Conversely, intraoperative diagnoses of acute appendicitis corresponded to 24.5% of histopathologically suppurative cases and 23.6% of rupture or gangrenous cases.\u003c/p\u003e"},{"header":"Discussion","content":"\u003cp\u003eClinical scoring systems such as the Alvarado score continue to assist decision-making in suspected pediatric appendicitis but should not be applied in isolation. Laboratory parameters, particularly white blood cell count and neutrophil percentage, add diagnostic value but remain nonspecific. Previous studies have reported inconsistent associations between urinalysis findings and appendicitis [\u003cspan citationid=\"CR24\" class=\"CitationRef\"\u003e24\u003c/span\u003e, \u003cspan citationid=\"CR25\" class=\"CitationRef\"\u003e25\u003c/span\u003e]. While ketonuria has been incorporated as a surrogate marker for anorexia in some reports on Alvarado scoring systems [\u003cspan citationid=\"CR12\" class=\"CitationRef\"\u003e12\u003c/span\u003e], its independent diagnostic contribution remains incompletely defined.\u003c/p\u003e \u003cp\u003eIn the present study, integration of seven clinical and laboratory factors resulted in a substantially higher diagnostic performance than the Alvarado score alone, as reflected by a higher AUC. These findings support the use of early biological and clinical predictors to guide decisions regarding imaging and surgical consultation, particularly in equivocal presentations.\u003c/p\u003e\n\u003ch3\u003e1. Refinement of urinary biomarkers\u003c/h3\u003e\n\u003cp\u003eWhen refining individual components of the Alvarado score, lowering the urine ketone cut-off from \u0026ge;\u0026thinsp;3\u0026thinsp;+\u0026thinsp;to \u0026ge;\u0026thinsp;2\u0026thinsp;+\u0026thinsp;increased sensitivity but reduced specificity, PPV, and dOR. These findings support the 2016 Alvarado modification, which defined urine ketone\u0026thinsp;\u0026ge;\u0026thinsp;3\u0026thinsp;+\u0026thinsp;as a clinically meaningful diagnostic threshold [\u003cspan citationid=\"CR12\" class=\"CitationRef\"\u003e12\u003c/span\u003e]. Although urine sp.gr. may be influenced by hydration status\u0026mdash;particularly in patients receiving intravenous fluids\u0026mdash;urine ketone levels normalize more slowly and may remain interpretable even after resuscitation [\u003cspan citationid=\"CR19\" class=\"CitationRef\"\u003e19\u003c/span\u003e]. This persistence suggests that ketonuria may serve as a reliable adjunctive biomarker across different phases of patient evaluation.\u003c/p\u003e \u003cp\u003eLeukocyte esterase, while not directly associated with appendicitis, remains clinically valuable for excluding alternative diagnoses such as urinary tract infection, urolithiasis, sexually transmitted infection, or autoimmune disease [\u003cspan citationid=\"CR25\" class=\"CitationRef\"\u003e25\u003c/span\u003e]. Similarly, urine pregnancy testing remains essential in adolescent females. In our cohort, pregnancy testing did not influence appendicitis diagnosis but successfully identified three pregnant patients prior to potential misdiagnosis, underscoring its role in safe clinical practice.\u003c/p\u003e\n\u003ch3\u003e2. Contribution of the seven-factor model\u003c/h3\u003e\n\u003cp\u003eConceptual clinical variables are clinically relevant parameters derived from structured patient characteristics, symptoms, physical findings, and basic laboratory data, reflecting real-world clinical reasoning rather than raw numerical values. The seven evaluated parameters\u0026mdash;sex, age group, abdominal pain location, white blood cell count, neutrophil percentage, urine ketone\u0026thinsp;\u0026ge;\u0026thinsp;3+, and leukocyte esterase level\u0026mdash;demonstrated greater discriminative ability than the Alvarado score alone. Appendicitis was less common in adolescent and preschool patients presenting for evaluation, whereas gynecologic conditions accounted for a substantial proportion of non-appendicitis diagnoses among females. In contrast, preadolescent patients, particularly before or early menarche, showed the highest likelihood of appendicitis.\u003c/p\u003e \u003cp\u003eDiagnostic differentiation remains challenging in younger children because of atypical symptom presentation and limited cooperation. Although this study demonstrated longer pain duration in the non-appendicitis group and in complicated appendicitis cases with partial antibiotic exposure, the exact duration and location of pain, subjective nausea, and resolution of fever after antipyretic use may reduce the reliability of clinical assessment and contribute to diagnostic uncertainty. These limitations highlight the importance of conceptual clinical variables over raw symptom reporting, particularly when preparing structured data pipelines for clinical use or AI model training [\u003cspan citationid=\"CR5\" class=\"CitationRef\"\u003e5\u003c/span\u003e].\u003c/p\u003e \u003cp\u003eConsistent with prior literature, commonly used predictors across machine-learning studies include pain characteristics, tenderness, guarding, leukocyte indices, platelet count, hematocrit, hemoglobin, and CRP. A recent systematic review reported average AUC values exceeding 80% across models, with few outliers below 75% [\u003cspan citationid=\"CR1\" class=\"CitationRef\"\u003e1\u003c/span\u003e].\u003c/p\u003e\n\u003ch3\u003e3. Role of imaging and implications for AI development\u003c/h3\u003e\n\u003cp\u003eAlthough approximately 35% of appendicitis cases in our cohort were diagnosed clinically, US remained the first-line imaging modality, with CT reserved for equivocal cases, consistent with previous studies [\u003cspan citationid=\"CR12\" class=\"CitationRef\"\u003e12\u003c/span\u003e, \u003cspan citationid=\"CR23\" class=\"CitationRef\"\u003e23\u003c/span\u003e, \u003cspan citationid=\"CR25\" class=\"CitationRef\"\u003e25\u003c/span\u003e]. These findings reflect continued reliance on objective diagnostic evidence while emphasizing the need to balance diagnostic certainty with radiation exposure and resource utilization [\u003cspan citationid=\"CR12\" class=\"CitationRef\"\u003e12\u003c/span\u003e, \u003cspan citationid=\"CR22\" class=\"CitationRef\"\u003e22\u003c/span\u003e, \u003cspan citationid=\"CR23\" class=\"CitationRef\"\u003e23\u003c/span\u003e, \u003cspan citationid=\"CR26\" class=\"CitationRef\"\u003e26\u003c/span\u003e, \u003cspan citationid=\"CR27\" class=\"CitationRef\"\u003e27\u003c/span\u003e].\u003c/p\u003e \u003cp\u003eEarly decision-making based on a combination of biological and clinical parameters may optimize imaging use and reduce unnecessary radiation. Separating AI training models for clinical diagnosis and radiographic interpretation may allow development of less complex and more reliable algorithms. Early diagnosis and timely intervention are associated with reduced rates of complicated appendicitis and shorter hospital stays.\u003c/p\u003e\n\u003ch3\u003e4. Implications for AI and data pipelines\u003c/h3\u003e\n\u003cp\u003eThe effectiveness of machine-learning and deep-learning models is strongly influenced by the quality of their input data and the structure of the training pipeline [\u003cspan citationid=\"CR10\" class=\"CitationRef\"\u003e10\u003c/span\u003e]. Although deep-learning models can autonomously learn complex patterns from unstructured data [\u003cspan citationid=\"CR8\" class=\"CitationRef\"\u003e8\u003c/span\u003e], their decision-making processes are often difficult to interpret, increasing the risk of misleading conclusions [\u003cspan citationid=\"CR7\" class=\"CitationRef\"\u003e7\u003c/span\u003e]. Conceptual clinical variables may serve as a bridge toward explainable AI, enabling transparency and traceability of decision-making [\u003cspan citationid=\"CR3\" class=\"CitationRef\"\u003e3\u003c/span\u003e, \u003cspan citationid=\"CR10\" class=\"CitationRef\"\u003e10\u003c/span\u003e]. However, explainable AI still requires models that align closely with clinician experience before responsibility can be appropriately shared [\u003cspan citationid=\"CR10\" class=\"CitationRef\"\u003e10\u003c/span\u003e].\u003c/p\u003e \u003cp\u003eHeterogeneous data, unnecessary variables, and high-cost parameters such as routine CT, magnetic resonant imaging (MRI), CRP, or ESR may not improve model performance and should be excluded from consensus data pipelines [\u003cspan citationid=\"CR10\" class=\"CitationRef\"\u003e10\u003c/span\u003e]. Data heterogeneity and variable acquisition protocols must be addressed through standardized preprocessing to enhance predictive accuracy [\u003cspan citationid=\"CR3\" class=\"CitationRef\"\u003e3\u003c/span\u003e]. Reports evaluating large language models as decision-support tools have demonstrated inconsistent recommendations and operational limitations, underscoring the need for rigorous validation before clinical integration [\u003cspan citationid=\"CR8\" class=\"CitationRef\"\u003e8\u003c/span\u003e].\u003c/p\u003e \u003cp\u003eA variety of algorithms\u0026mdash;including random forests, artificial neural networks, convolutional neural networks, support vector machines, logistic regression, and extreme gradient boosting\u0026mdash;have demonstrated superior performance compared with traditional clinical scoring systems [\u003cspan citationid=\"CR3\" class=\"CitationRef\"\u003e3\u003c/span\u003e, \u003cspan citationid=\"CR11\" class=\"CitationRef\"\u003e11\u003c/span\u003e]. No single model is universally optimal; selection of appropriate algorithms and careful hyperparameter tuning remain essential [\u003cspan citationid=\"CR9\" class=\"CitationRef\"\u003e9\u003c/span\u003e, \u003cspan citationid=\"CR11\" class=\"CitationRef\"\u003e11\u003c/span\u003e].\u003c/p\u003e\n\u003ch3\u003e5. Surgical–pathological concordance\u003c/h3\u003e\n\u003cp\u003eThe moderate concordance observed between surgical and pathological diagnoses aligns with previous reports. Discrepancies may arise from sampling techniques, microscopic perforations, or fibrin coatings that obscure perforation sites. Although histopathology may reveal microscopic findings not evident intraoperatively, these differences rarely alter surgical management. Future AI developments incorporating advanced imaging analysis may allow preoperative identification of lymphoid hyperplasia and uncomplicated appendicitis amenable to non-operative management, improving patient selection and advancing personalized care.\u003c/p\u003e\n\u003ch3\u003e6. Limitations\u003c/h3\u003e\n\u003cp\u003eThis study has several limitations. Its retrospective design, single-center setting, and relatively small sample size may limit generalizability. However, the use of complete clinical data, defined urine ketone cut points (2\u0026thinsp;+\u0026thinsp;and 3+), and seven diagnostic factors provides a structured framework for future model development.\u003c/p\u003e \u003cp\u003eIn contrast to many recent AI studies that rely on resampling-based training, non-histological diagnoses, or incomplete datasets [\u003cspan citationid=\"CR3\" class=\"CitationRef\"\u003e3\u003c/span\u003e, \u003cspan citationid=\"CR4\" class=\"CitationRef\"\u003e4\u003c/span\u003e, \u003cspan citationid=\"CR6\" class=\"CitationRef\"\u003e6\u003c/span\u003e, \u003cspan citationid=\"CR10\" class=\"CitationRef\"\u003e10\u003c/span\u003e, \u003cspan citationid=\"CR11\" class=\"CitationRef\"\u003e11\u003c/span\u003e], this study benefits from histopathological or radiographic confirmation and minimal missing data, strengthening internal validity. Nonetheless, this work should be regarded as an exploratory \u0026ldquo;sandbox\u0026rdquo; for variable selection; advances in imaging technology, evolving diagnostic workflows, and antibiotic-pretreated cases were not addressed and warrant evaluation in future multicenter studies.\u003c/p\u003e"},{"header":"Conclusion","content":"\u003cp\u003eSeven factors\u0026mdash;sex, age group, location of abdominal pain, white blood cell count, neutrophil percentage, urine ketone level\u0026thinsp;\u0026ge;\u0026thinsp;3+, and leukocyte esterase level\u0026mdash;collectively improved the accuracy of early diagnosis in pediatric appendicitis and enhanced decision-making for timely US in equivocal cases, outperforming the Alvarado score alone. Among these variables, urinary ketones are particularly valuable because their levels change more slowly following fluid resuscitation, allowing them to serve as a cost-effective and readily accessible indicator even after hydration.\u003c/p\u003e \u003cp\u003eArtificial intelligence agents, similar to human clinicians, may generate differing interpretations when applied to complex clinical scenarios. Well-curated, multifaceted clinical data therefore provide a critical foundation for appropriately trained AI models [\u003cspan citationid=\"CR2\" class=\"CitationRef\"\u003e2\u003c/span\u003e]. Individualized AI approaches that separately enhance US or CT interpretation may optimize resource utilization more effectively than a single, unified machine-learning or deep-learning model that combines clinical, laboratory, and imaging data indiscriminately.\u003c/p\u003e \u003cp\u003eAccurate classification of appendicitis severity should incorporate both intraoperative findings and histopathological confirmation. Macroscopically visible perforation identified by the surgeon and microscopic perforation detected by the pathologist remain the most reliable indicators of perforated appendicitis. Future AI developments focused on imaging analysis may enable preoperative exclusion of lymphoid hyperplasia and identification of uncomplicated appendicitis amenable to medical management, thereby improving patient selection and advancing precision care.\u003c/p\u003e \u003cp\u003e \u003cstrong\u003eLearning point\u003c/strong\u003e \u003cp\u003eThe development of reliable AI models for complex clinical applications should begin with careful construction of precise, clinically meaningful datasets. Establishing an appropriate data pipeline before algorithm development is essential to achieving clinician-relevant models that can be trusted and effectively integrated into practice.\u003c/p\u003e \u003c/p\u003e"},{"header":"Abbreviations","content":"\u003cdiv class=\"DefinitionList\"\u003e \u003cdiv class=\"DefinitionListEntry\"\u003e \u003cdiv class=\"Term\"\u003eAI\u003c/div\u003e \u003cdiv class=\"Description\"\u003e \u003cp\u003eArtificial intelligence\u003c/p\u003e \u003c/div\u003e \u003c/div\u003e \u003cdiv class=\"DefinitionListEntry\"\u003e \u003cdiv class=\"Term\"\u003eBIDIAP\u003c/div\u003e \u003cdiv class=\"Description\"\u003e \u003cp\u003eBiomarkers for the Diagnosis of Appendicitis in Pediatrics Index\u003c/p\u003e \u003c/div\u003e \u003c/div\u003e \u003cdiv class=\"DefinitionListEntry\"\u003e \u003cdiv class=\"Term\"\u003eCRP\u003c/div\u003e \u003cdiv class=\"Description\"\u003e \u003cp\u003eC-reactive protein\u003c/p\u003e \u003c/div\u003e \u003c/div\u003e \u003cdiv class=\"DefinitionListEntry\"\u003e \u003cdiv class=\"Term\"\u003eSp.gr.\u003c/div\u003e \u003cdiv class=\"Description\"\u003e \u003cp\u003eSpecific gravity\u003c/p\u003e \u003c/div\u003e \u003c/div\u003e \u003cdiv class=\"DefinitionListEntry\"\u003e \u003cdiv class=\"Term\"\u003eIQR\u003c/div\u003e \u003cdiv class=\"Description\"\u003e \u003cp\u003eInterquartile range\u003c/p\u003e \u003c/div\u003e \u003c/div\u003e \u003cdiv class=\"DefinitionListEntry\"\u003e \u003cdiv class=\"Term\"\u003edOR\u003c/div\u003e \u003cdiv class=\"Description\"\u003e \u003cp\u003eDiagnostic odds ratio\u003c/p\u003e \u003c/div\u003e \u003c/div\u003e \u003cdiv class=\"DefinitionListEntry\"\u003e \u003cdiv class=\"Term\"\u003eCI\u003c/div\u003e \u003cdiv class=\"Description\"\u003e \u003cp\u003eConfidence interval\u003c/p\u003e \u003c/div\u003e \u003c/div\u003e \u003cdiv class=\"DefinitionListEntry\"\u003e \u003cdiv class=\"Term\"\u003ePPV\u003c/div\u003e \u003cdiv class=\"Description\"\u003e \u003cp\u003ePositive predictive value\u003c/p\u003e \u003c/div\u003e \u003c/div\u003e \u003cdiv class=\"DefinitionListEntry\"\u003e \u003cdiv class=\"Term\"\u003eNPV\u003c/div\u003e \u003cdiv class=\"Description\"\u003e \u003cp\u003eNegative predictive value\u003c/p\u003e \u003c/div\u003e \u003c/div\u003e \u003cdiv class=\"DefinitionListEntry\"\u003e \u003cdiv class=\"Term\"\u003eAUC\u003c/div\u003e \u003cdiv class=\"Description\"\u003e \u003cp\u003eArea under the curve\u003c/p\u003e \u003c/div\u003e \u003c/div\u003e \u003cdiv class=\"DefinitionListEntry\"\u003e \u003cdiv class=\"Term\"\u003eROC\u003c/div\u003e \u003cdiv class=\"Description\"\u003e \u003cp\u003eReceiver operating characteristic\u003c/p\u003e \u003c/div\u003e \u003c/div\u003e \u003cdiv class=\"DefinitionListEntry\"\u003e \u003cdiv class=\"Term\"\u003eUS\u003c/div\u003e \u003cdiv class=\"Description\"\u003e \u003cp\u003eUltrasonography\u003c/p\u003e \u003c/div\u003e \u003c/div\u003e \u003cdiv class=\"DefinitionListEntry\"\u003e \u003cdiv class=\"Term\"\u003eCT\u003c/div\u003e \u003cdiv class=\"Description\"\u003e \u003cp\u003eComputed tomography\u003c/p\u003e \u003c/div\u003e \u003c/div\u003e \u003cdiv class=\"DefinitionListEntry\"\u003e \u003cdiv class=\"Term\"\u003eMRI\u003c/div\u003e \u003cdiv class=\"Description\"\u003e \u003cp\u003eMagnetic resonant imaging\u003c/p\u003e \u003c/div\u003e \u003c/div\u003e \u003cdiv class=\"DefinitionListEntry\"\u003e \u003cdiv class=\"Term\"\u003eESR\u003c/div\u003e \u003cdiv class=\"Description\"\u003e \u003cp\u003eErythrocyte sedimentation rate\u003c/p\u003e \u003c/div\u003e \u003c/div\u003e \u003c/div\u003e"},{"header":"Declarations","content":"\u003cp\u003e\u003cstrong\u003eEthics approval and consent to participate\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eThis study was approved by the Institutional Review Board of Mahidol University (Certificate of Approval No. COA. MURA2023/553).\u003c/p\u003e\n\u003cp\u003eThe requirement for informed consent was waived due to the retrospective nature of the study and the use of anonymized data.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eClinical trial number\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eNot applicable.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eConsent for publication\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eNot applicable.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eAvailability of data and materials\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eThe datasets used and/or analyzed during the current study are available from the corresponding author on reasonable request.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eCompeting interests\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eThe authors declare that they have no competing interests.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eFunding\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eThis research received no specific grant from any funding agency in the public, commercial, or not-for-profit sectors.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eAuthors’ contributions\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003ePT conceived and designed the study, acquired the data, performed the analysis, interpreted the results, drafted the manuscript, and substantively revised it. PT approved the final version of the manuscript for submission.\u003c/p\u003e\n\u003cp\u003eArtificial intelligence–assisted tools were used for language editing only and did not influence the study design, data analysis, or interpretation.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eAcknowledgements\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eThe author would like to thank Ms. Suraida Aeesoa, Medical Statistician at the Department of Surgery, for assistance with the final data analysis.\u003c/p\u003e"},{"header":"References","content":"\u003col\u003e\n\u003cli\u003eChekmeyan M, Liu SH. Artificial intelligence for the diagnosis of pediatric appendicitis: A systematic review. Am J Emerg Med. 2025 Jun;92:18-31. doi: 10.1016/j.ajem.2025.02.023. Epub 2025 Feb 17. PMID: 40048888.\u003c/li\u003e\n\u003cli\u003eShikha A, Kasem A. Decoding Pediatric Appendicitis: Unraveling Complexity with Artificial Intelligence and Evolving Management Insights [Internet]. Appendicitis - Current Insights. IntechOpen; 2024. Available from: http://dx.doi.org/10.5772/intechopen.1008318\u003c/li\u003e\n\u003cli\u003eAkbulut S, Kucukakcali Z, Colak C. Artificial intelligence in acute appendicitis: A comprehensive review of machine learning and deep learning applications. World J Gastroenterol. 2025 Nov 21;31(43):112000. doi: 10.3748/wjg.v31.i43.112000. PMID: 41358178; PMCID: PMC12678916.\u003c/li\u003e\n\u003cli\u003eMarcinkevičs R, Sokol K, Paulraj A, Hilbert MA, Rimili V, Wellmann S, Knorr C, Reingruber B, Vogt JE, Reis Wolfertstetter P. External validation of predictive models for diagnosis, management and severity of pediatric appendicitis. Front Pediatr. 2025 Aug 29;13:1587488. doi: 10.3389/fped.2025.1587488. PMID: 40948509; PMCID: PMC12426264.\u003c/li\u003e\n\u003cli\u003eKendall J, Gaspar G, Berger D, Levman J. Machine Learning and Feature Selection in Pediatric Appendicitis. Tomography. 2025 Aug 13;11(8):90. doi: 10.3390/tomography11080090. PMID: 40863881; PMCID: PMC12390108.\u003c/li\u003e\n\u003cli\u003eMarcinkevics R, Reis Wolfertstetter P, Wellmann S, Knorr C, Vogt JE. Using Machine Learning to Predict the Diagnosis, Management and Severity of Pediatric Appendicitis. Front Pediatr. 2021 Apr 29;9:662183. doi: 10.3389/fped.2021.662183. PMID: 33996697; PMCID: PMC8116489.\u003c/li\u003e\n\u003cli\u003eKlimiene, Ugne \u0026amp; Marcinkevics, Ricards \u0026amp; Reis Wolfertstetter, Patricia \u0026amp; Ozkan, Ece \u0026amp; Paschke, Alyssia \u0026amp; Niederberger, David \u0026amp; Wellmann, Sven \u0026amp; Knorr, Christian \u0026amp; Vogt, Julia. (2022). Multiview Concept Bottleneck Models Applied to Diagnosing Pediatric Appendicitis.\u003c/li\u003e\n\u003cli\u003eErsahin K, Sanduleanu S, Thulasi Seetha S, Bremm J, Abbasli C, Zimmer C, Damer T, Kottlors J, Goertz L, Bruns C, Maintz D, Abdullayev N. From Bedside to Bot-Side: Artificial Intelligence in Emergency Appendicitis Management. Life (Basel). 2025 Sep 1;15(9):1387. doi: 10.3390/life15091387. PMID: 41010329; PMCID: PMC12470868.\u003c/li\u003e\n\u003cli\u003eChadaga K, Khanna V, Prabhu S, Sampathila N, Chadaga R, Umakanth S, Bhat D, Swathi KS, Kamath R. An interpretable and transparent machine learning framework for appendicitis detection in pediatric patients. Sci Rep. 2024 Oct 18;14(1):24454. doi: 10.1038/s41598-024-75896-y. Erratum in: Sci Rep. 2025 Jan 22;15(1):2841. doi: 10.1038/s41598-025-86494-x. PMID: 39424647; PMCID: PMC11489819.\u003c/li\u003e\n\u003cli\u003eErman A, Ferreira J, Ashour WA, Guadagno E, St-Louis E, Emil S, Cheung J, Poenaru D. Machine-learning-assisted Preoperative Prediction of Pediatric Appendicitis Severity. J Pediatr Surg. 2025 Jun;60(6):162151. doi: 10.1016/j.jpedsurg.2024.162151. Epub 2025 Jan 13. PMID: 39855986.\u003c/li\u003e\n\u003cli\u003eIssaiy M, Zarei D, Saghazadeh A. Artificial Intelligence and Acute Appendicitis: A Systematic Review of Diagnostic and Prognostic Models. World J Emerg Surg. 2023 Dec 19;18(1):59. doi: 10.1186/s13017-023-00527-2. PMID: 38114983; PMCID: PMC10729387.\u003c/li\u003e\n\u003cli\u003eAlvarado A. How to improve the clinical diagnosis of acute appendicitis in resource limited settings. World J Emerg Surg. 2016;11:16. doi: 10.1186/s13017-016-0071-8. PMID: 27118990; PMCID: PMC4845369. \u003c/li\u003e\n\u003cli\u003eGudjonsdottir J, Marklund E, Hagander L, Sal\u0026ouml; M. Clinical Prediction Scores for Pediatric Appendicitis. Eur J Pediatr Surg. 2021;31(3):252-260. doi: 10.1055/s-0040-1710534. Epub 2020 May 26. PMID: 32455443.\u003c/li\u003e\n\u003cli\u003eArredondo Montero J, Bardaj\u0026iacute; Pascual C, Antona G, Ros Briones R, L\u0026oacute;pez-Andr\u0026eacute;s N, Mart\u0026iacute;n-Calvo N. The BIDIAP index: a clinical, analytical and ultrasonographic score for the diagnosis of acute appendicitis in children. Pediatr Surg Int. 2023;39(1):175. doi: 10.1007/s00383-023-05463-5. PMID: 37038002; PMCID: PMC10085908.\u003c/li\u003e\n\u003cli\u003eT\u0026uuml;rkmenoğlu Y, Ka\u0026ccedil;ar A, Duras E, K\u0026ouml;k S, G\u0026ouml;z\u0026uuml;b\u0026uuml;y\u0026uuml;k AA, Arat C. The Role of Alvarado and Pediatric Appendicitis Score in Acute Appendicitis in Children. J Pediatr Res. 2020;7(3):192-198. doi: 10.4274/jpr.galenos.2019.62582.\u003c/li\u003e\n\u003cli\u003eSağ S, Basar D, Yurdadoğan F, Pehlivan Y, Elemen L. Comparison of Appendicitis Scoring Systems in Childhood Appendicitis. Turk Arch Pediatr. 2022;57(5):532-537. doi: 10.5152/TurkArchPediatr.2022.22076. PMID: 36062441; PMCID: PMC9524470. \u003c/li\u003e\n\u003cli\u003eArredondo Montero J, Bronte Anaut M, Bardaj\u0026iacute; Pascual C, Antona G, L\u0026oacute;pez-Andr\u0026eacute;s N, Mart\u0026iacute;n-Calvo N. Alterations and diagnostic performance of capillary ketonemia in pediatric acute appendicitis: a pilot study. Pediatr Surg Int. 2022;39(1):44. doi: 10.1007/s00383-022-05332-7. PMID: 36495332; PMCID: PMC9741565.\u003c/li\u003e\n\u003cli\u003eO\u0026apos;Donohoe PB, Kessler R, Beattie TF. Exploring the clinical utility of blood ketone levels in the emergency department assessment of paediatric patients. Emerg Med J. 2006;23(10):783-7. doi: 10.1136/emj.2006.035758. PMID: 16988307; PMCID: PMC2579600. \u003c/li\u003e\n\u003cli\u003eLaffel L. Ketone bodies: a review of physiology, pathophysiology and application of monitoring to diabetes. Diabetes Metab Res Rev 1999;15:412\u0026ndash;26.\u003c/li\u003e\n\u003cli\u003eStringer MD. Acute appendicitis. J Paediatr Child Health. 2017;53(11):1071-1076. doi: 10.1111/jpc.13737. Epub 2017 Oct 17. PMID: 29044790.\u003c/li\u003e\n\u003cli\u003eNissen M, Tr\u0026ouml;bs RB. The lymphocyte-to-monocyte ratio may distinguish complicated from non-complicated pediatric appendicitis: A retrospective study and literature review. Pediatr Neonatol. 2022;63(2):146-153. doi: 10.1016/j.pedneo.2021.08.018. Epub 2021 Oct 27. PMID: 34799285.\u003c/li\u003e\n\u003cli\u003eCarpenter JL, Orth RC, Zhang W, Lopez ME, Mangona KL, Guillerman RP. Diagnostic performance of US for differentiating perforated from Nonperforated pediatric appendicitis: a prospective cohort study. Radiology 2017; 282: 835\u0026ndash;41. doi: https:// doi. org/ 10. 1148/radiol. 2016160175\u003c/li\u003e\n\u003cli\u003eRawolle T, Reismann M, Minderjahn MI, Bassir C, Hauptmann K, Rothe K, Reismann J. Sonographic differentiation of complicated from uncomplicated appendicitis. Br J Radiol. 2019;92(1099):20190102. doi: 10.1259/bjr.20190102. Epub 2019 May 29. PMID: 31112397; PMCID: PMC6636276.\u003c/li\u003e\n\u003cli\u003eBiehl CM, Elliver M, Gudjonsdottir J, Sal\u0026ouml; M. Utility of Urine Dipstick Testing in Pediatric Appendicitis: Assessing its Role in Identifying Complicated Cases and Retrocecal Appendicitis. Eur J Pediatr Surg. 2024 Dec 19. doi: 10.1055/a-2490-1156. Epub ahead of print. PMID: 39701137.\u003c/li\u003e\n\u003cli\u003eChen CY, Zhao LL, Lin YR, Wu KH, Wu HP. Different urinalysis appearances in children with simple and perforated appendicitis. Am J Emerg Med. 2013 Nov;31(11):1560-3. doi: 10.1016/j.ajem.2013.06.027. Epub 2013 Sep 20. PMID: 24055480\u003c/li\u003e\n\u003cli\u003eBenabbas R, Hanna M, Shah J, Sinert R. Diagnostic Accuracy of History, Physical Examination, Laboratory Tests, and Point-of-care Ultrasound for Pediatric Acute Appendicitis in the Emergency Department: A Systematic Review and Meta-analysis. Acad Emerg Med. 2017;24(5):523-551. doi: 10.1111/acem.13181. PMID: 28214369.\u003c/li\u003e\n\u003cli\u003eKim MS, Kwon H-J, Kang KA, Do I-G, Park H-J, Kim EY, et al. Diagnostic performance and useful findings of ultrasound reevaluation for patients with equivocal CT features of acute appendicitis. Br J Radiol 2018; 91: 20170529\u003c/li\u003e\n\u003c/ol\u003e"},{"header":"Tables","content":"\u003cp\u003eTables 1 to 7 are available in the supplementary files section\u003c/p\u003e"}],"fulltextSource":"","fullText":"","funders":[],"hasAdminPriorityOnWorkflow":false,"hasManuscriptDocX":true,"hasOptedInToPreprint":true,"hasPassedJournalQc":"","hasAnyPriority":false,"hideJournal":true,"highlight":"","institution":"","isAcceptedByJournal":false,"isAuthorSuppliedPdf":false,"isDeskRejected":"","isHiddenFromSearch":false,"isInQc":false,"isInWorkflow":false,"isPdf":false,"isPdfUpToDate":true,"isWithdrawnOrRetracted":false,"journal":{"display":true,"email":"[email protected]","identity":"researchsquare","isNatureJournal":false,"hasQc":true,"allowDirectSubmit":true,"externalIdentity":"","sideBox":"","snPcode":"","submissionUrl":"/submission","title":"Research Square","twitterHandle":"researchsquare","acdcEnabled":true,"dfaEnabled":false,"editorialSystem":"","reportingPortfolio":"","inReviewEnabled":false,"inReviewRevisionsEnabled":true},"keywords":"Pediatric appendicitis, Alvarado score, Urinary ketones, Diagnostic accuracy, Clinical prediction model, Artificial intelligence","lastPublishedDoi":"10.21203/rs.3.rs-8666814/v1","lastPublishedDoiUrl":"https://doi.org/10.21203/rs.3.rs-8666814/v1","license":{"name":"CC BY 4.0","url":"https://creativecommons.org/licenses/by/4.0/"},"manuscriptAbstract":"\u003cp\u003e\u003cstrong\u003eBackground:\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eTimely and accurate diagnosis of pediatric appendicitis remains challenging because of variable clinical presentations and limitations of existing scoring systems. Although the Alvarado score is widely used, its diagnostic performance in children is suboptimal, and reliance on imaging increases resource utilization and radiation exposure. Identification of objective, biologically meaningful predictors may improve early diagnosis and inform future artificial intelligence (AI)–based decision-support models.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eMethods:\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eThis retrospective study included children aged 2–18 years evaluated for suspected appendicitis between January 2014 and June 2024. Clinical characteristics, laboratory data, urinalysis results, imaging findings, operative notes, and histopathological reports were reviewed. Diagnostic performance of individual variables and the Alvarado score was assessed using sensitivity, specificity, predictive values, diagnostic odds ratios, and receiver operating characteristic analysis. A composite diagnostic framework based on seven clinical and laboratory factors was evaluated and compared with the Alvarado score alone.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eResults:\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eA total of 375 patients were included, of whom 275 (73.3%) were diagnosed with appendicitis. Appendicitis was more prevalent in males and in children aged ≥6 years, with the highest incidence in preadolescence. The mean Alvarado score was significantly higher in the appendicitis group than in the non-appendicitis group (8 ± 2 vs 5 ± 2, p \u0026lt; 0.001). At a cut-off score of ≥7, the Alvarado score demonstrated a sensitivity of 77.1%, specificity of 71.0%, and an AUC of 74.05%.\u003c/p\u003e\n\u003cp\u003eSeven factors—sex, age group, location of abdominal pain, white blood cell count, neutrophil percentage, urine ketone level ≥3+, and leukocyte esterase level—were significantly associated with appendicitis. The seven diagnostic conceptual factors achieved superior diagnostic performance (AUC 86.77%, 95% CI 82.56–90.96) compared with the Alvarado score alone. Urinary ketone levels ≥3+ showed a significant positive association with appendicitis, whereas leukocyte esterase demonstrated an inverse association.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eConclusions:\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eIntegration of seven objective clinical and laboratory factors improves early diagnostic accuracy for pediatric appendicitis beyond the Alvarado score alone and supports more targeted use of ultrasonography in equivocal cases. Urinary ketones are a particularly useful adjunctive marker due to their physiological relevance and persistence after fluid resuscitation. Establishing structured, clinically meaningful data pipelines is essential for future development of reliable and explainable AI-assisted diagnostic tools.\u003c/p\u003e","manuscriptTitle":"Conceptual Clinical Variables Enhancing the Alvarado Score in Pediatric Appendicitis: Lessons for Artificial Intelligence Models","msid":"","msnumber":"","nonDraftVersions":[{"code":1,"date":"2026-02-03 16:04:19","doi":"10.21203/rs.3.rs-8666814/v1","editorialEvents":[{"type":"communityComments","content":0}],"status":"published","journal":{"display":true,"email":"[email protected]","identity":"researchsquare","isNatureJournal":false,"hasQc":true,"allowDirectSubmit":true,"externalIdentity":"","sideBox":"","snPcode":"","submissionUrl":"/submission","title":"Research Square","twitterHandle":"researchsquare","acdcEnabled":true,"dfaEnabled":false,"editorialSystem":"","reportingPortfolio":"","inReviewEnabled":false,"inReviewRevisionsEnabled":true}}],"origin":"","ownerIdentity":"e42a9d54-b028-4b33-aa8b-7813e5d98c35","owner":[],"postedDate":"February 3rd, 2026","published":true,"recentEditorialEvents":[],"rejectedJournal":[],"revision":"","amendment":"","status":"posted","subjectAreas":[],"tags":[],"updatedAt":"2026-02-16T07:11:04+00:00","versionOfRecord":[],"versionCreatedAt":"2026-02-03 16:04:19","video":"","vorDoi":"","vorDoiUrl":"","workflowStages":[]},"version":"v1","identity":"rs-8666814","journalConfig":"researchsquare"},"__N_SSP":true},"page":"/article/[identity]/[[...version]]","query":{"redirect":"/article/rs-8666814","identity":"rs-8666814","version":["v1"]},"buildId":"XKTyCvWXoU3ODBz1xrDgd","isFallback":false,"isExperimentalCompile":false,"dynamicIds":[84888],"gssp":true,"scriptLoader":[]}