Area Under The ROC Curve Research Articles

Validation of serum non-invasive tests of liver fibrosis as prognostic markers of clinical outcomes in people with fatty liver disease in Australia.

The validity of non-invasive tests (NITs) of liver fibrosis for the prediction of liver and mortality outcomes in an Australian cohort is unknown. We aimed to verify the utility of available NITs to predict overall and cause-specific mortality and major adverse liver outcome (MALO). This was an analysis from the Crossroads 1 clinic sub-study of a randomly sampled adult cohort from regional Australia between 2001 and 2003. Baseline variables included demographic details, anthropometry, health and lifestyle data, and laboratory tests. Non-alcoholic fatty liver disease (NAFLD) and metabolic-(dysfunction) associated fatty liver disease (MAFLD) were defined by fatty liver index≥60 and other accepted criteria. Outcomes were defined by the International Statistical Classification of Diseases and Related Health Problems 10th Revision codes for linked hospitalization and death registry data. Available serum-based NITs were analyzed as predictors of overall, cardiovascular disease-related, and cancer-related mortality and MALO in those with fatty liver disease (FLD). In total, 1324 and 1444 participants were included for NAFLD and MAFLD analysis (prevalence 35.4% and 40.7%, respectively). There were 298 deaths (89 cardiovascular disease-related and 98 cancer-related) and 24 MALO over a median 19.7years of follow-up time. In both forms of FLD, fibrosis-4 index, Steatosis-Associated Fibrosis Estimator score, and Forns fibrosis score consistently had the highest area under the receiver operating characteristic curve (AUROC) for overall and cause-specific mortality, with AUROC>0.70 for each outcome. However, all had poor discriminatory ability for determining MALO in each FLD. Several liver fibrosis NITs perform similarly reasonably well in predicting the risk of mortality outcomes in those with FLD but are poorly discriminatory for MALO prediction.

Shear-Wave Elastography Improves Diagnostic Accuracy in Chronic Kidney Disease Compared to Conventional Ultrasound.

Non-invasive tests are increasingly demanded for diagnosing and prognostication of chronic kidney disease (CKD). Shear-wave elastography (SWE), an emerging technique for measuring tissue stiffness, shows promise for distinguishing between individuals with different stages of renal fibrosis. This study aimed to compare the diagnostic accuracy of two-dimensional SWE (2D-SWE) and conventional ultrasound for detecting CKD, employing renal biopsy as the gold standard. From May 2020 to October 2023, this prospective study included 30 healthy volunteers and 169 patients with CKD who had undergone 2D-SWE and conventional ultrasound of both kidneys. Cortical and medullary stiffness, cortical pixel intensity, renal length, parenchymal and cortical thickness, interlobar artery peak systolic velocity, end-diastolic velocity (EDV), and resistive index were measured. The diagnostic accuracy of 2D-SWE and conventional ultrasound was compared using the receiver operating characteristic curve (ROC) and Delong test. For diagnosing CKD, the area under the ROC (AUC) of cortical stiffness (0.96 [95% CI, 0.93, 0.99]) was significantly higher than that of all conventional ultrasound parameters, including EDV (0.78 [95% CI, 0.71, 0.86]) and cortical thickness (0.74 [95% CI, 0.67, 0.80]). The sensitivity of cortical stiffness (91%) was significantly higher than that of EDV (68%) and cortical thickness (53%). No significant difference was found in the specificity of cortical stiffness (96%) compared to that of EDV (79%) and cortical thickness (100%). Two-dimensional SWE showed higher diagnostic accuracy than that of conventional ultrasound for detecting CKD.

G4 & the balanced metric family – a novel approach to solving binary classification problems in medical device validation & verification studies

BackgroundIn medical device validation and verification studies, the area under the receiver operating characteristic curve (AUROC) is often used as a primary endpoint despite multiple reports showing its limitations. Hence, researchers are encouraged to consider alternative metrics as primary endpoints. A new metric called G4 is presented, which is the geometric mean of sensitivity, specificity, the positive predictive value, and the negative predictive value. G4 is part of a balanced metric family which includes the Unified Performance Measure (also known as P4) and the Matthews’ Correlation Coefficient (MCC). The purpose of this manuscript is to unveil the benefits of using G4 together with the balanced metric family when analyzing the overall performance of binary classifiers.ResultsSimulated datasets encompassing different prevalence rates of the minority class were analyzed under a multi-reader-multi-case study design. In addition, data from an independently published study that tested the performance of a unique ultrasound artificial intelligence algorithm in the context of breast cancer detection was also considered. Within each dataset, AUROC was reported alongside the balanced metric family for comparison. When the dataset prevalence and bias of the minority class approached 50%, all three balanced metrics provided equivalent interpretations of an AI’s performance. As the prevalence rate increased / decreased and the data became more imbalanced, AUROC tended to overvalue / undervalue the true classifier performance, while the balanced metric family was resistant to such imbalance. Under certain circumstances where data imbalance was strong (minority-class prevalence < 10%), MCC was preferred for standalone assessments while P4 provided a stronger effect size when evaluating between-groups analyses. G4 acted as a middle ground for maximizing both standalone assessments and between-groups analyses.ConclusionsUse of AUROC as the primary endpoint in binary classification problems provides misleading results as the dataset becomes more imbalanced. This is explicitly noticed when incorporating AUROC in medical device validation and verification studies. G4, P4, and MCC do not share this limitation and paint a more complete picture of a medical device’s performance in a clinical setting. Therefore, researchers are encouraged to explore the balanced metric family when evaluating binary classification problems.

Machine learning-based prediction of sarcopenia in community-dwelling middle-aged and older adults: findings from the CHARLS.

Sarcopenia is a prominent issue among aging populations and associated with poor health outcomes. This study aimed to examine the predictive value of questionnaire and biomarker data for sarcopenia, and to further develop a user-friendly calculator for community-dwelling middle-aged and older adults. We used two waves (2011 and 2013) of the China Health and Retirement Longitudinal Study (CHARLS) to predict sarcopenia, defined by the Asian Working Group for Sarcopenia 2019 criteria. We restricted the analytical sample to adults aged 45 or above (N = 2934). Five machine learning models were used to construct Q-based (only questionnaire variables), Bio-based (only biomarker variables), and combined (questionnaire plus biomarker variables) models. Area under the receiver operating characteristic curve (AUROC) was used for performance assessment. Temporal external validation was performed based on two datasets from CHARLS. Important predictors were identified by Shapley values and coefficients. Extreme gradient boosting (XGBoost), considering both questionnaire and biomarker characteristics, emerged as the optimal model, and its AUROC was 0.759 (95% CI: 0.747-0.771) at a decision threshold of 0.20 on the test set. Models also performed well on the external datasets. We found that cognitive function was the most important predictor in both Q-based and combined models, and blood urea nitrogen was the most important predictor in the Bio-based model. Other key predictors included education, haematocrit, total cholesterol, drinking, number of chronic diseases, and instrumental activities of daily living score. Our findings offer a potential for early screening and targeted prevention of sarcopenia among middle-aged and older adults in the community setting.

Prediction of New-onset Atrial Fibrillation in Patients with Hypertrophic Cardiomyopathy Using Plasma Proteomics Profiling.

Atrial fibrillation (AF) is the most common sustained arrhythmia among patients with hypertrophic cardiomyopathy (HCM), increasing symptom burden and stroke risk. We aimed to construct a plasma proteomics-based model to predict new-onset AF in patients with HCM and determine dysregulated signaling pathways. In this prospective, multi-center cohort study, we conducted plasma proteomics profiling of 4,986 proteins at enrollment. We developed a proteomics-based machine learning (ML) model to predict new-onset AF using samples from one institution (training set) and tested its predictive ability using independent samples from another institution (test set). We performed a survival analysis to compare the risk of new-onset AF among high- and low-risk groups in the test set. We performed pathway analysis of proteins significantly (univariable p<0.05) associated with new-onset AF using a false discovery rate (FDR) threshold of 0.001. The study included 284 patients with HCM (training set: 193, test set: 91). Thirty-seven (13%) patients developed AF during median follow-up of 3.2 years [25-75 percentile: 1.8-5.2]. Using the proteomics-based prediction model developed in the training set, the area under the receiver-operating-characteristic curve (AUC) was 0.89 (95% confidence interval 0.78-0.99) in the test set. In the test set, patients categorized as high-risk had a higher rate of developing new-onset AF (log-rank p=0.002). The Ras-MAPK pathway was dysregulated in patients who developed incident AF during follow-up (FDR<1.0×10-6). This is the first study to demonstrate the ability of plasma proteomics to predict new-onset AF in HCM and identify dysregulated signaling pathways.

Identifying cardiovascular disease risk in the U.S. population using environmental volatile organic compounds exposure: A machine learning predictive model based on the SHAP methodology

BackgroundCardiovascular disease (CVD) remains a leading cause of mortality globally. Environmental pollutants, specifically volatile organic compounds (VOCs), have been identified as significant risk factors. This study aims to develop a machine learning (ML) model to predict CVD risk based on VOC exposure and demographic data using SHapley Additive exPlanations (SHAP) for interpretability. MethodsWe utilized data from the National Health and Nutrition Examination Survey (NHANES) from 2011 to 2018, comprising 5098 participants. VOC exposure was assessed through 15 urinary metabolite metrics. The dataset was split into a training set (70 %) and a test set (30 %). Six ML models were developed, including Random Forest (RF), Light Gradient Boosting Machine (LightGBM), Decision Tree (DT), Extreme Gradient Boosting (XGBoost), Multi-Layer Perceptron (MLP), and Support Vector Machines (SVM). Model performance was evaluated using the Area Under the Receiver Operating Characteristic Curve (AUROC), accuracy, balanced accuracy, F1 score, J-index, kappa, Matthew's correlation coefficient (MCC), positive predictive value (PPV), negative predictive value (NPV), sensitivity (sens), specificity (spec) and SHAP was applied to interpret the best-performing model. ResultsThe RF model exhibited the highest predictive performance with an ROC of 0.8143. SHAP analysis identified age and ATCA as the most significant predictors, with ATCA showing a protective effect against CVD, particularly in older adults and those with hypertension. The study found a significant interaction between ATCA levels and age, indicating that the protective effect of ATCA is more pronounced in older individuals due to increased oxidative stress and inflammatory responses associated with aging. E-values analysis suggested robustness to unmeasured confounders. ConclusionsThis study is the first to utilize VOC exposure data to construct an ML model for predicting CVD risk. The findings highlight the potential of combining environmental exposure data with demographic information to enhance CVD risk prediction, supporting the development of personalized prevention and intervention strategies.

The 123 COVID SCORE: A simple and reliable diagnostic tool to predict in-hospital death in COVID-19 patients on hospital admission.

Patients hospitalized due to Coronavirus disease 2019 (COVID-19) are still burdened with high risk of death. The aim of this study was to create a risk score predicting in-hospital mortality in COVID-19 patients on hospital admission. Independent mortality predictors identified in multivariate logistic regression analysis were used to build the 123 COVID SCORE. Diagnostic performance of the score was evaluated using the area under the receiver-operating characteristic curve (AUROC). Data from 673 COVID-19 patients with median age of 70 years were used to build the score. In-hospital death occurred in 124 study participants (18.4%). The final score is composed of 3 variables that were found predictive of mortality in multivariate logistic regression analysis: (1) age, (2) oxygen saturation on hospital admission without oxygen supplementation and (3) percentage of lung involvement in chest computed tomography (CT). Four point ranges have been identified: 0-5, 6-8, 9-11 and 12-17, respectively corresponding to low (1.5%), moderate (13.4%), high (28.4%) and very high (57.3%) risk of in-hospital death. The 123 COVID SCORE accuracy measured with the AUROC was 0.797 (95% CI 0.757-0.838; p<0.0001) in the study population and 0.774 (95% CI 0.728-0.821; p<0.0001) in an external validation cohort consisting of 558 COVID-19 patients. The 123 COVID SCORE containing merely 3 variables: age, oxygen saturation, and percentage of lung involvement assessed with chest CT is a simple and reliable tool to predict in-hospital death in COVID-19 patients upon hospital admission.

Relationship Between White Blood Cell Count and Bacteremia Using Interval Likelihood Ratios in Hospitalized Patients.

Patients with bacteremia often have elevated white blood cell (WBC) and neutrophil counts, yet these alone are poor predictors of bacteremia. Data on the continuous relationship between WBC response and bacteremia are lacking. This study aims to characterize the relationship of WBC count, neutrophil percentage, and absolute neutrophil count (ANC) to bacteremia using interval likelihood ratios (ILRs) derived from a large sample of hospitalized patients. Retrospective cohort study in a large healthcare system from 2017 to 2018. This study included non-surgical inpatients who had at least one complete blood count (CBC) with differential up to 24hours after admission and a blood culture. Patients with immunosuppression and malignancy or who received antibiotics before negative blood cultures were excluded. Predictors were WBC count, ANC, and neutrophil percentage. The outcome was bacteremia. We compared test discrimination using the area under the receiver operating characteristics curve (AUROC). We calculated ILRs for bacteremia across test value intervals. As a practical example, we assumed a 5% pre-test probability of bacteremia and calculated the post-test probability for each interval. We compared this approach to a threshold approach using a threshold of 70% neutrophils. Of 25,776 patients with a CBC with differential and blood culture, 1160 had bacteremia. AUROC was highest for neutrophil percentage (0.74), followed by ANC (0.63) and WBC count (0.58). Probability of bacteremia increased exponentially from neutrophil percentage 80 to 100%. Odds of bacteremia varied 35-fold based on neutrophil percentage. A threshold approach with a cut-off of 70% significantly underestimated bacteremia risk at higher levels. ILRs offered a more discriminating approach to estimating the probability of bacteremia than a single threshold. Physicians assessing risk of bacteremia should pay attention to the magnitude of abnormality because very high and very low values have much stronger predictive power than dichotomized results.

Development and validation of a cardiovascular risk prediction model for Sri Lankans using machine learning.

Sri Lankans do not have a specific cardiovascular (CV) risk prediction model and therefore, World Health Organization(WHO) risk charts developed for the Southeast Asia Region are being used. We aimed to develop a CV risk prediction model specific for Sri Lankans using machine learning (ML) of data of a population-based, randomly selected cohort of Sri Lankans followed up for 10 years and to validate it in an external cohort. The cohort consisted of 2596 individuals between 40-65 years of age in 2007, who were followed up for 10 years. Of them, 179 developed hard CV diseases (CVD) by 2017. We developed three CV risk prediction models named model 1, 2 and 3 using ML. We compared predictive performances between models and the WHO risk charts using receiver operating characteristic curves (ROC). The most predictive and practical model for use in primary care, model 3 was named "SLCVD score" which used age, sex, smoking status, systolic blood pressure, history of diabetes, and total cholesterol level in the calculation. We developed an online platform to calculate the SLCVD score. Predictions of SLCVD score were validated in an external hospital-based cohort. Model 1, 2, SLCVD score and the WHO risk charts predicted 173, 162, 169 and 10 of 179 observed events and the area under the ROC (AUC) were 0.98, 0.98, 0.98 and 0.52 respectively. During external validation, the SLCVD score and WHO risk charts predicted 56 and 18 respectively of 119 total events and AUCs were 0.64 and 0.54 respectively. SLCVD score is the first and only CV risk prediction model specific for Sri Lankans. It predicts the 10-year risk of developing a hard CVD in Sri Lankans. SLCVD score was more effective in predicting Sri Lankans at high CV risk than WHO risk charts.

Comparison of the utility of SARC-F, SARC-CalF, and calf circumference as screening tools for sarcopenia in patients with osteoporosis.

Patients with osteoporosis who also have sarcopenia are at a high risk for falls and fractures. Early detection of sarcopenia is crucial for these patients. This study aimed to compare the effectiveness of SARC-F, SARC-CalF, and calf circumference (CC) as screening tools for sarcopenia in patients with osteoporosis. This cross-sectional study was retrospectively conducted on patients who attended the outpatient clinic for Osteoporosis and Sarcopenia at Kyoto Medical Center. Sarcopenia was determined based on low skeletal muscle mass and weak handgrip strength. Sensitivity and specificity analyses were conducted on SARC-F, SARC-CalF, and CC. The diagnostic utility of these three tools was compared using the receiver-operating characteristic (ROC) curves and the area under the ROC curves (AUC). A total of 225 patients (men/women: 33/192) with a median age of 69.0 years (interquartile range: 61.0‒75.0) were enrolled. The prevalence of sarcopenia was found to be 11.6%. CC had the highest sensitivity (80.8%), while SARC-F had the highest specificity (93.0%) for detecting sarcopenia. ROC analysis revealed that all three tools had significant potential for sarcopenia diagnosis, with SARC-CalF having the highest AUC compared to SARC-F and CC (0.753 vs. 0.619 and 0.700). A multivariate logistic regression, incorporating other confounders as explanatory variables, revealed that SARC-CalF was independently related to sarcopenia (odds ratio: 14.80, 95% confidence interval: 3.83-57.30, p < 0.001). In patients with osteoporosis, SARC-CalF is more effective in the early detection of sarcopenia than SARC-F and CC.

Improving Clinical Preparedness: Community Health Nurses and Early Hypoglycemia Prediction in Type 2 Diabetes Using Hybrid Machine Learning Techniques.

The aim of the study was to analyze the data of diabetic patients regarding warning signs of hypoglycemia to predict it at an early stage using various novel machine learning (ML) algorithms. Individual interviews with diabetic patients were conducted over 6 months to acquire information regarding their experience with hypoglycemic episodes. This information included warning signs of hypoglycemia, such as incoherent speech, exhaustion, weakness, and other clinically relevant cases of low blood sugar. Researchers used supervised, unsupervised, and hybrid techniques. In supervised techniques, researchers applied regression, while in hybrid classification ML techniques were used. In a 5-fold cross-validation approach, the prediction performance of seven models was examined using the area under the receiver operating characteristic curve (AUROC). We analyzed the data of 290 diabetic patients with low blood sugar episodes. Our investigation discovered that gradient boosting and neural networks performed better in regression, with accuracies of 0.416 and 0.417, respectively. In classification models, gradient boosting, AdaBoost, and random forest performed better overall, with AUC scores of 0.821, 0.814, and 0.821, individually. Precision values were 0.779, 0.775, and 0.776 for gradient boosting, AdaBoost, and random forest, respectively. AdaBoost and Gradient Boosting models, in particular, outperformed all others in predicting the probability of clinically severe hypoglycemia. These techniques enable community health nurses to predict hypoglycemia at an early stage and provide the necessary therapies to patients to prevent complications resulting from hypoglycemia.

Comparative analysis of SARC-F-EBM, Ishii test, and six other screening tools for sarcopenia in Chinese community-dwelling older adults: a cross-sectional diagnostic study

To assess the relative performance of simple screening methods for sarcopenia in Chinese community-dwelling older adults. Data of older adults aged ≥ 60 were collected through a cross-sectional investigation. Sarcopenia was defined according to the Asian Working Group for Sarcopenia 2019 criteria. The accuracy of screening methods was evaluated using sensitivity, specificity, receiver operating characteristic (ROC) curves and area under the ROC curves (AUC). The AUC value greater than 0.8 represented the good screening ability. A total of 918 older adults (44.3% men, mean age 70.4 ± 6.5 years) were included in this study. The overall prevalence rates of possible sarcopenia, confirmed sarcopenia, and severe sarcopenia were 59.5%, 12.8%, and 5.9%, respectively. In men, the SARC-F-EBM and Ishii tests indicated good screening capabilities for confirmed sarcopenia, with an AUC of 0.81 (95% CI: 0.77–0.85) and 0.80 (95% CI: 0.76–0.84), respectively. In women, the highest AUC was also achieved using the SARC-F-EBM at 0.79 (95% CI: 0.75–0.82), followed by the Ishii test at 0.77 (95% CI: 0.74–0.81), showing the moderate efficacy. A ranking diagram showed that SARC-F-EBM was most likely to be considered the best method for diagnosing sarcopenia in terms of AUC and sensitivity, regardless of sex. We recommend the SARC-F-EBM for sarcopenia screening in community-dwelling Chinese older adults when respondents are able to answer the questionnaire accurately; otherwise, the Ishii test consisting entirely of objective measures could be used.

Integrating machine learning and bioinformatics approaches for identifying novel diagnostic gene biomarkers in colorectal cancer.

This study aimed to identify diagnostic gene biomarkers for colorectal cancer (CRC) by analyzing differentially expressed genes (DEGs) in tumor and adjacent normal samples across five colon cancer gene-expression profiles (GSE10950, GSE25070, GSE41328, GSE74602, GSE142279) from the Gene Expression Omnibus (GEO) database. Intersecting identified DEGs with the module with the highest correlation to gene expression patterns of tumor samples in the gene co-expression network analysis revealed 283 overlapped genes. Centrality measures were calculated for these genes in the reconstructed STRING protein-protein interaction network. Applying LASSO logistic regression, eleven genes were ultimately recognized as candidate diagnostic genes. Among these genes, the area under the receiver operating characteristic curve (AUROC) values for nine genes (CDC25B, CDK4, IQGAP3, MMP1, MMP7, SLC7A5, TEAD4, TRIB3, and UHRF1) surpassed the threshold of 0.92 in both the training and validation sets. We evaluated the diagnostic performance of these genes with four machine learning algorithms: random forest (RF), support vector machines (SVM), artificial neural network (ANN), and gradient boosting machine (GBM). In the testing dataset (GSE21815 and GSE106582), the AUROC scores were greater than 0.95 for all of the machine learning algorithms, indicating the high diagnostic performance of the nine genes. Besides, these nine genes are also significantly correlated to twelve immune cells, namely Mast cells activated, Macrophages M0, M1, and M2, Neutrophils, T cells CD4 memory activated, T cells follicular helper, T cells CD8, T cells CD4 memory resting, B cells memory, Plasma cells, and Mast cells resting (P < 0.05). These results strongly suggest that all of the nine genes have the potential to serve as reliable diagnostic biomarkers for CRC.

Differentiate adrenal lipid-poor adenoma from nodular hyperplasia with CT quantitative parameters: a feasibility study.

To explore the potential of CT quantitative parameters in differentiating adrenal lipid-poor adenoma (LPA) from nodular hyperplasia and evaluate diagnostic performance. Patients with LPA or nodular hyperplasia who underwent contrast-enhanced CT before adrenalectomy were analyzed retrospectively. The study included 128 patients (83 with LPA and 45 with nodular hyperplasia). Each lesion's unenhanced attenuation, portal-venous phase attenuation (CTp), and the portal-venous phase attenuation of the abdominal aorta were evaluated. We subsequently calculated absolute enhancement [a lesion's portal-venous phase attenuation minus unenhanced attenuation (in HUs)], relative enhancement (absolute enhancement divided by unenhanced attenuation), and the relative enhancement ratio [(absolute enhancement divided by abdominal aorta's portal-venous phase attenuation) ×100%]. Lesion number and size were recorded. Volume was assessed by ITK-snap software and the ratio of lesion volume to ipsilateral adrenal volume (volume ratio) was determined. Intergroup differences were analyzed using Student's t-test and chi-squared test. Logistic regression models were developed, and receiver operating characteristic (ROC) curves were constructed to determine the area under the ROC curve (AUC), sensitivity, and specificity. The model's performance was then compared against radiologists' subjective assessments, and the inter- and intra-reader agreement values among radiologists were calculated. Portal-venous phase attenuation, volume ratio, and lesion number were independent predictors of LPA. The logistic regression model incorporating CTp, volume ratio, and lesion number achieved an AUC of 0.835, with a sensitivity of 73.5% and a specificity of 80.0%. The radiologists' diagnostic specificity and accuracy appeared to be inferior to the model. The inter-reader agreement among radiologists ranged from 0.082 to 0.535, and the intra-reader agreement of two radiologists were 0.734 and 0.583. The portal-venous phase CT demonstrated potential in distinguishing LPA from nodular hyperplasia. The diagnostic performance of the model integrating CTp, volume ratio, and lesion number outperformed radiologists in terms of variability and reproducibility.

Global change drives potential niche contraction and range shift of globally threatened African vulture

Human-induced global change poses an increasingly severe threat to biodiversity, with species having limited population sizes being particularly vulnerable. Mapping and modeling the distribution ranges of such species, along with detecting potential range shifts and contractions at both local and regional scales, are essential for developing effective conservation plans. Ruppell's vulture Gyps rueppelli, an ecologically important bird species native to Africa, is experiencing a rapid decline in its range. The purpose of this study is to map and model potential regional spatio-temporal distribution of Ruppell's vulture in Africa, alongside detecting the possibility of the species' range shifts and contractions. A total of 804 rarefied localities were identified where the Ruppell's vulture was the dominant bird species. This study employed the Maximum Entropy (MaxEnt) algorithm to perform species distribution modeling for the Ruppell's vulture. The modeling considered current climate conditions (1970s-2000s) as a baseline, along with two future climate change scenarios (Shared Socioeconomic Pathways: SSPs 245 and 585) for two future time periods (2050s and 2070s). The model's performance was evaluated by optimizing settings and examining the Area under the Receiver Operating Characteristic Curve (AUC-ROC). Among the 13 bioclimatic and anthropogenic variables included in the model, four (isothermality, cropland expansion, anthropogenic biomes, and urban expansion (in order of importance)) emerged as the most influential drivers of Ruppell's vulture regional distribution. All considered Species distribution models (SDMs) achieved high predictive performance, with AUC-ROC values exceeding 0.9.The model predicted a total of approximately 19,453 ha of suitable habitat for Ruppell's vultures in Africa, with East Africa identified as the most prominent region under the current climate scenario. Isothermality (38.8%) was the primary factor influencing Ruppell's vulture distribution, followed by agricultural expansion (29.9%) and anthropogenic biomes (7.2%) in the face of global change. The results reveal considerable future habitat loss (up to 61%) for Ruppell's vultures in the study area, alongside an eastward range shift (longitudinal axis) by the 2050s under projected climate change scenarios. These imply that Ruppell's vultures face imminent population decline and range shift due to significant habitat loss and climate change. Hence, prioritizing the development and implementation of a coordinated conservation program that incorporates captive breeding and assisted migration is critical to save this vulture species in its native African range.

Comparative Analysis of Machine Learning Models for Predicting Student Success in Online Programming Courses: A Study Based on LMS Data and External Factors

Early prediction of student performance in online programming courses is essential for implementing timely interventions to enhance academic outcomes. This study aimed to predict academic success by comparing four machine learning models: Logistic Regression, Random Forest, Support Vector Machine (SVM), and Neural Network (Multilayer Perceptron, MLP). We analyzed data from the Moodle Learning Management System (LMS) and external factors of 591 students enrolled in online object-oriented programming courses at the Universidad Estatal de Milagro (UNEMI) between 2022 and 2023. The data were preprocessed to address class imbalance using the synthetic minority oversampling technique (SMOTE), and relevant features were selected based on Random Forest importance rankings. The models were trained and optimized using Grid Search with cross-validation. Logistic Regression achieved the highest Area Under the Receiver Operating Characteristic Curve (AUC-ROC) on the test set (0.9354), indicating strong generalization capability. SVM and Neural Network models performed adequately but were slightly outperformed by the simpler models. These findings suggest that integrating LMS data with external factors enhances early prediction of student success. Logistic Regression is a practical and interpretable tool for educational institutions to identify at-risk students, and to implement personalized interventions.

A tri-light warning system for hospitalized COVID-19 patients: Credibility-based risk stratification for future pandemic preparedness

PurposeThe novel coronavirus pneumonia (COVID-19) has continually spread and mutated, requiring a patient risk stratification system to optimize medical resources and improve pandemic response. We aimed to develop a conformal prediction-based tri-light warning system for stratifying COVID-19 patients, applicable to both original and emerging variants. MethodsWe retrospectively collected data from 3646 patients across multiple centers in China. The dataset was divided into a training set (n = 1451), a validation set (n = 662), an external test set from Huoshenshan Field Hospital (n = 1263), and a specific test set for Delta and Omicron variants (n = 544). The tri-light warning system extracts radiomic features from CT (computed tomography) and integrates clinical records to classify patients into high-risk (red), uncertain-risk (yellow), and low-risk (green) categories. Models were built to predict ICU (intensive care unit) admissions (adverse cases in training/validation/Huoshenshan/variant test sets: n = 39/21/262/11) and were evaluated using AUROC ((area under the receiver operating characteristic curve)) and AUPRC ((area under the precision-recall curve)) metrics. ResultsThe dataset included 1830 men (50.2 %) and 1816 women (50.8 %), with a median age of 53.7 years (IQR [interquartile range]: 42–65 years). The system demonstrated strong performance under data distribution shifts, with AUROC of 0.89 and AUPRC of 0.42 for original strains, and AUROC of 0.77–0.85 and AUPRC of 0.51–0.60 for variants. ConclusionThe tri-light warning system can enhance pandemic responses by effectively stratifying COVID-19 patients under varying conditions and data shifts.

Risk Score for Predicting AKI from Contrast-Enhanced CT (Pre-CT AKI score): Training and Validation from Retrospective Cohort.

The lack of a recognized risk evaluation for Contrast-associated acute kidney injury (CA-AKI) after contrast-enhanced computed tomography (CECT) makes it challenging to counsel patients before the procedure. This study aims to identify the incidence of CA-AKI post CECT, assess the associated risk factors, develop and validate a predictive score. All adult patients who underwent CECT in 2018 to 2022 were included in the training cohort while those in 2023 formed the external validation cohort. Exclusions applied to patients with CKD stage 5, recent dialysis, or incomplete data. Multiple logistic regression was employed to identify risk factors. The area under the receiver operating characteristic curve (AUROC) was used to evaluate both internal and external validation. From 21,878 enrolled patients, 6,042 and 2,463 met the inclusion criteria for the training and validation cohorts with a mean eGFR of 86.0 (26.4) and 81.4 (27.6) mL/min/1.73 m2, respectively. In the training cohort, 492 patients (8.1%) developed CA-AKI, and 49 (0.8%) required dialysis. Independent risk factors for CA-AKI included male gender, clinical setting, hemoglobin levels of <10 g/dL, and baseline eGFR less than 90 mL/min/1.73 m2. The model, using a weighted integer score derived from these factors, exhibited an AUROC of 0.715 (95% CI: 0.692-0.743) in the training cohort and 0.706 (95% CI: 0.663-0.748) in the validation cohort. CECT can lead to CA-AKI in specific populations. The Pre-CT AKI risk score for CA-AKI following CECT demonstrated good discriminative power and can be easily applied in clinical practice.

Osteocalcin: may be a useful biomarker for early identification of rapidly progressive central precocious puberty in girls.

To assess serum osteocalcin (OC) as a potential biomarker for the early detection of rapidly progressive central precocious puberty (RP-CPP) in girls. Serum OC levels were quantified using enzyme-linked immunosorbent assays (ELISAs). In the retrospective analysis, receiver operating characteristic (ROC) curve analysis was employed to evaluate the ability of OC to identify RP-CPP. A prospective study and screening tests were utilized to assess the potential of OC for use in the early prediction of RP-CPP. Variable selection in the multivariate analysis was conducted using the Bayesian Information Criterion (BIC) and binary logistic regression was employed to construct the diagnostic prediction model. Girls with RP-CPP had significantly higher serum OC levels compared to girls with non-rapidly progressive central precocious puberty (NRP-CPP) (149.04±40.50 vs. 89.10±31.83 ng/mL, P < 0.001). The optimal OC cut-off point for differentiating RP-CPP from NRP-CPP was 107.05 ng/mL, the area under the ROC curve (AUC) was 0.90 (95%CI: 0.851-0.949; P < 0.001), with a sensitivity of 91.1% and specificity of 70.7%. The results of the prospective study indicated that changes in OC precede alterations in estradiol (E2) and bone age (BA). A diagnostic prediction model that includes duration of breast development, BA, OC, high-density lipoprotein cholesterol (HDL-C), and uterine length achieved an AUC of 0.961, with a sensitivity of 94.1% and specificity of 91.5% for the detection of RP-CPP. If OC is excluded from the model, the AUC decreases to 0.894, with sensitivity and specificity declining to 80.5% and 83.1%, respectively. Serum OC levels may serve as a promising biomarker for the early differentiation between RP-CPP and NRP-CPP in girls. The diagnostic prediction model that incorporates duration of breast development, BA, OC, HDL-C, and uterine length effectively identifies girls with RP-CPP.

Evaluation of Groundwater Potential Zones Using GIS‐Based Machine Learning Ensemble Models in the Gidabo Watershed, Ethiopia

AbstractThe main objective of this study is to map and evaluate groundwater potential zones (GWPZs) using advanced ensemble machine learning (ML) models, notably Random Forest (RF) and Support Vector Machine (SVM). GWPZs are identified by considering essential factors such as geology, drainage density, slope, land use/land cover (LULC), rainfall, soil, and lineament density. This is combined with datasets used for training and validating the RF and SVM models, which consisted of 75 potential sites (boreholes and springs), 22 non‐potential sites (bare lands and settlement areas), and 20 potential sites (water bodies). Each dataset is randomly partitioned into two sets: training (70%) and validation (30%). The model's performance is evaluated using the area under the receiver operating characteristic curve (AUC‐ROC). The AUC of the RF model is 0.91, compared to 0.88 for the SVM model. Both models classified GWPZs effectively, but the RF model performed slightly better. The classified GWPZ map shows that high GWPZs are typically located within water bodies, natural springs, low‐lying regions, and forested areas. In contrast, low GWPZs are primarily found in shrubland and grassland areas. This study is vital for decision‐makers as it promotes sustainable groundwater use and ensures water security in the studied area.