Receiver Operating Characteristic Area Under The Curve Research Articles

An efficient synthetic minority oversampling technique-based ensemble learning model to detect COVID-19 severity

The COVID-19 pandemic has highlighted the importance of accurately predicting disease severity to ensure timely intervention and effective allocation of healthcare resources, which can ultimately improve patient outcomes. This study aims to develop an efficient machine learning (ML) model based on patient demographic and clinical data. It utilizes advanced feature engineering techniques to reduce the dimensionality of dataset and address the issue of highly imbalanced data using synthetic minority oversampling technique (SMOTE). The study employs several ensemble learning models, including XGBoost, Random Forest, AdaBoost, voting ensemble, enhanced-weighted voting ensemble, and stack-based ensembles with support vector machine (SVM) and Gaussian Naïve Bayes as meta-learners, to develop the proposed model. The results indicate that the proposed model outperformed the top-performing models reported in previous studies. It achieved an accuracy of 0.978, sensitivity of 1.0, precision of 0.875, F1-score of 0.934, and receiver operating characteristic area under the curve (ROC-AUC) of 0.965. The study identified several features that significantly correlated with COVID-19 severity, which included respiratory rate (breaths per minute), c-reactive proteins, age, and total leukocyte count (TLC) count. The proposed approach presents a promising method for accurate COVID-19 severity prediction, which may prove valuable in assisting healthcare providers in making informed decisions about patient care.

The utility of qSOFA in assessing outcomes of patients with acute heart failure in a level II hospital: a single-center retrospective analysis

Abstract Funding Acknowledgements None. Introduction High all-cause mortality is associated with acute heart failure (AHF) hospitalization, both when the condition first develops and when it worsens. Although there are several risk models for patients with chronic heart failure (CHF), it is unknown whether these scoring systems may be applied to patients with ADHF to predict their chance of developing the condition. Purpose Acute Heart failure (AHF) patients benefit from accurate risk assessment. Those stratified with worse prognosis may benefit more from continuous and close monitoring, with subsequent prompt intervention in case of deterioration. Currently, there is a lack of a universally accepted risk score that can accurately predict outcomes in acute heart failure (AHF). Undeniably, sepsis and AHF share several clinical features. These similarities encourage the sepsis-to-AHF translation of SOFA and qSOFA scores. We compared different scoring systems used in sepsis and assessed their predictive accuracy. Methodology This retrospective study included all patients with acute heart failure admitted in a private teaching hospital between January 2014 and January 2023. This study included newly admitted adult patients with new-onset AHF or acute on chronic heart failure. Baseline qSOFA, SIRS, and SOFA values were determined. Receiver Operating Characteristics - Area under the Curve (ROC-AUC) analysis assessed these scores' ability to predict mortality, whereas chi-square and Fischer's exact test assessed secondary outcome risk. Results 545 patients were included in the study. The majority were males with a mean age of 65 years. Most patients in the study required oxygen support, were classified as NOHRIA A, and with NYHA Functional Class II. Renal and respiratory failures were prominent upon arrival. About 11% met ≥2 points in qSOFA. Therein, systolic BP of <100 mmHg was the most frequent criterion met. The primary endpoint, that is in-hospital mortality, occurred in 9.52% of patients. A positive qSOFA, defined as a score of 2 or more, was consistent in the majority of those meeting the primary endpoint (Chi2 = 264.698, p<0.001). The ROC analysis of qSOFA (AUC 0.949, 95% CI 0.924–0.974, accuracy of 73%), SOFA (AUC 0.935, 95% CI 0.887–0.983, accuracy of 72%), and SIRS criteria (AUC 0.939, 95% CI 0.917–0.961, accuracy of 72%) show no fundamental differences regarding the prediction of the primary endpoint. Patients with positive qSOFA criteria were more frequently affected by organ failure (OR 5.386, p <0.001), needed more vasopressors (OR 4.921, p < 0.001), notably with an ejection fraction <40% (OR 3.363, p <0.001), and required mechanical ventilator support more often (OR 3.794, p<0.001). Conclusion In conclusion, qSOFA used in Acute Heart Failure is a valuable tool in the early detection of in-hospital mortality and detection of worse outcomes, hence should be considered when making therapeutic strategies and decisions.AUC for Primary OutcomesSecondary Outcomes

Diuretic response after neonatal cardiac surgery: a report from the NEPHRON collaborative.

Multicenter early diuretic response (DR) analysis of single furosemide dosing following neonatal cardiac surgery is lacking to inform whether early DR predicts adverse clinical outcomes. We performed a retrospective cohort study utilizing data from the NEPHRON registry. Random forest machine learning generated receiver operating characteristic-area under the curve (ROC-AUC) and odds ratios for mechanical ventilation (MV) and respiratory support (RS). Prolonged MV and RS were defined using ≥ 90th percentile of observed/expected ratios. Secondary outcomes were prolonged CICU and hospital length of stay (LOS) and kidney failure (stage III acute kidney injury (AKI), peritoneal dialysis, and/or continuous kidney replacement therapy on postoperative day three) assessed using covariate-adjusted ROC-AUC curves. A total of 782 children were included. Cumulative urine output (UOP) metrics were lower in prolonged MV and RS patients, but DR poorly predicted prolonged MV (highest AUC 0.611, OR 0.98, sensitivity 0.67, specificity 0.53, p = 0.006, 95% OR CI 0.96-0.99 for cumulative 6-h UOP) and RS (highest AUC 0.674, OR 0.94, sensitivity 0.75, specificity 0.54, p < 0.001, 95% CI 0.91-0.97 UOP between 3 and 6h). Secondary outcome results were similar. DR had fair discrimination for kidney failure (AUC 0.703, OR 0.94, sensitivity 0.63, specificity 0.71, 95% OR CI 0.91-0.98, p < 0.001, cumulative 6-h UOP). Early DR poorly discriminated patients with prolonged MV, RS, and LOS in this cohort, though it may identify severe postoperative AKI phenotype. Future work is warranted to determine if early DR or late postoperative DR later, in combination with other AKI metrics, may identify a higher-risk phenotype.

Machine learning for predicting elective fertility preservation outcomes

This retrospective study applied machine-learning models to predict treatment outcomes of women undergoing elective fertility preservation. Two-hundred-fifty women who underwent elective fertility preservation at a tertiary center, 2019–2022 were included. Primary outcome was the number of metaphase II oocytes retrieved. Outcome class was based on oocyte count (OC): Low (≤ 8), Medium (9–15) or High (≥ 16). Machine-learning models and statistical regression were used to predict outcome class, first based on pre-treatment parameters, and then using post-treatment data from ovulation-triggering day. OC was 136 Low, 80 Medium, and 34 High. Random Forest Classifier (RFC) was the most accurate model (pre-treatment receiver operating characteristic (ROC) area under the curve (AUC) was 77%, and post-treatment ROC AUC was 87%), followed by XGBoost Classifier (pre-treatment ROC AUC 74%, post-treatment ROC AUC 86%). The most important pre-treatment parameters for RFC were basal FSH (22.6%), basal LH (19.1%), AFC (18.2%), and basal estradiol (15.6%). Post-treatment parameters were estradiol levels on trigger-day (17.7%), basal FSH (11%), basal LH (9%), and AFC (8%). Machine-learning models trained with clinical data appear to predict fertility preservation treatment outcomes with relatively high accuracy.

Construction of Risk Prediction Model of Type 2 Diabetic Kidney Disease Based on Deep Learning.

This study aimed to develop a diabetic kidney disease (DKD) prediction model using long short term memory (LSTM) neural network and evaluate its performance using accuracy, precision, recall, and area under the curve (AUC) of the receiver operating characteristic (ROC) curve. The study identified DKD risk factors through literature review and physician focus group, and collected 7 years of data from 6,040 type 2 diabetes mellitus patients based on the risk factors. Pytorch was used to build the LSTM neural network, with 70% of the data used for training and the other 30% for testing. Three models were established to examine the impact of glycosylated hemoglobin (HbA1c), systolic blood pressure (SBP), and pulse pressure (PP) variabilities on the model's performance. The developed model achieved an accuracy of 83% and an AUC of 0.83. When the risk factor of HbA1c variability, SBP variability, or PP variability was removed one by one, the accuracy of each model was significantly lower than that of the optimal model, with an accuracy of 78% (P<0.001), 79% (P<0.001), and 81% (P<0.001), respectively. The AUC of ROC was also significantly lower for each model, with values of 0.72 (P<0.001), 0.75 (P<0.001), and 0.77 (P<0.05). The developed DKD risk predictive model using LSTM neural networks demonstrated high accuracy and AUC value. When HbA1c, SBP, and PP variabilities were added to the model as featured characteristics, the model's performance was greatly improved.

A Machine Learning Application for Medical Image Analysis Using Deep Convolutional Neural Networks (CNNs) and Transfer Learning Models for Pneumonia Detection

Pneumonia is still a major problem in world health, causing a lot of illness and death, particularly among the most susceptible people. For care and therapy to be successful, a prompt and precise diagnosis is essential. Recent years have seen encouraging outcomes in medical image processing tasks when using deep learning methods, including Convolutional Neural Networks (CNNs). Using deep convolutional neural networks (CNNs) and transfer learning models, we provide a machine learning application for detecting pneumonia in medical photographs.Our suggested system makes use of convolutional neural network (CNN) designs that have already been trained using massive natural picture datasets, including VGG, ResNet, and DenseNet. We hope that by training these algorithms on a collection of chest X-ray pictures, we may modify their characteristics to better diagnose pneumonia. Furthermore, we investigate data augmentation methods to improve the models' generalizability, especially in cases when there is a lack of labelled data.Standard measures including precision, sensitivity, specificity, and area under the curve of the receiver operating characteristic (AUC-ROC) are used to assess the performance of the created models on a heterogeneous dataset of chest X-ray images.

A hybrid approach for optimizing software defect prediction using a gray wolf optimization and multilayer perceptron

PurposeSoftware defect prediction (SDP) is a critical aspect of software quality assurance, aiming to identify and manage potential defects in software systems. In this paper, we have proposed a novel hybrid approach that combines Gray Wolf Optimization with Feature Selection (GWOFS) and multilayer perceptron (MLP) for SDP. The GWOFS-MLP hybrid model is designed to optimize feature selection, ultimately enhancing the accuracy and efficiency of SDP. Gray Wolf Optimization, inspired by the social hierarchy and hunting behavior of gray wolves, is employed to select a subset of relevant features from an extensive pool of potential predictors. This study investigates the key challenges that traditional SDP approaches encounter and proposes promising solutions to overcome time complexity and the curse of the dimensionality reduction problem.Design/methodology/approachThe integration of GWOFS and MLP results in a robust hybrid model that can adapt to diverse software datasets. This feature selection process harnesses the cooperative hunting behavior of wolves, allowing for the exploration of critical feature combinations. The selected features are then fed into an MLP, a powerful artificial neural network (ANN) known for its capability to learn intricate patterns within software metrics. MLP serves as the predictive engine, utilizing the curated feature set to model and classify software defects accurately.FindingsThe performance evaluation of the GWOFS-MLP hybrid model on a real-world software defect dataset demonstrates its effectiveness. The model achieves a remarkable training accuracy of 97.69% and a testing accuracy of 97.99%. Additionally, the receiver operating characteristic area under the curve (ROC-AUC) score of 0.89 highlights the model’s ability to discriminate between defective and defect-free software components.Originality/valueExperimental implementations using machine learning-based techniques with feature reduction are conducted to validate the proposed solutions. The goal is to enhance SDP’s accuracy, relevance and efficiency, ultimately improving software quality assurance processes. The confusion matrix further illustrates the model’s performance, with only a small number of false positives and false negatives.

Fully automated CT imaging biomarkers for opportunistic prediction of future hip fractures.

Assess automated CT imaging biomarkers in patients who went on to hip fracture, compared with controls. In this retrospective case-control study, 6926 total patients underwent initial abdominal CT over a 20-year interval at one institution. A total of 1308 patients (mean age at initial CT, 70.5 ± 12.0 years; 64.4% female) went on to hip fracture (mean time to fracture, 5.2 years); 5618 were controls (mean age 70.3 ± 12.0 years; 61.2% female; mean follow-up interval 7.6 years). Validated fully automated quantitative CT algorithms for trabecular bone attenuation (at L1), skeletal muscle attenuation (at L3), and subcutaneous adipose tissue area (SAT) (at L3) were applied to all scans. Hazard ratios (HRs) comparing highest to lowest risk quartiles and receiver operating characteristic (ROC) curve analysis including area under the curve (AUC) were derived. Hip fracture HRs (95% CI) were 3.18 (2.69-3.76) for low trabecular bone HU, 1.50 (1.28-1.75) for low muscle HU, and 2.18 (1.86-2.56) for low SAT. 10-year ROC AUC values for predicting hip fracture were 0.702, 0.603, and 0.603 for these CT-based biomarkers, respectively. Multivariate combinations of these biomarkers further improved predictive value; the 10-year ROC AUC combining bone/muscle/SAT was 0.733, while combining muscle/SAT was 0.686. Opportunistic use of automated CT bone, muscle, and fat measures can identify patients at higher risk for future hip fracture, regardless of the indication for CT imaging. CT data can be leveraged opportunistically for further patient evaluation, with early intervention as needed. These novel AI tools analyse CT data to determine a patient's future hip fracture risk.

Integrated flood risk assessment in Hunza-Nagar, Pakistan: unifying big climate data analytics and multi-criteria decision-making with GIS

Floods are a widespread natural disaster with substantial economic implications and far-reaching consequences. In Northern Pakistan, the Hunza-Nagar valley faces vulnerability to floods, posing significant challenges to its sustainable development. This study aimed to evaluate flood risk in the region by employing a GIS-based Multi-Criteria Decision Analysis (MCDA) approach and big climate data records. By using a comprehensive flood risk assessment model, a flood hazard map was developed by considering nine influential factors: rainfall, regional temperature variation, distance to the river, elevation, slope, Normalized difference vegetation index (NDVI), Topographic wetness index (TWI), land use/land cover (LULC), curvature, and soil type. The analytical hierarchy process (AHP) analysis assigned weights to each factor and integrated with geospatial data using a GIS to generate flood risk maps, classifying hazard levels into five categories. The study assigned higher importance to rainfall, distance to the river, elevation, and slope compared to NDVI, TWI, LULC, curvature, and soil type. The weighted overlay flood risk map obtained from the reclassified maps of nine influencing factors identified 6% of the total area as very high, 36% as high, 41% as moderate, 16% as low, and 1% as very low flood risk. The accuracy of the flood risk model was demonstrated through the Receiver Operating Characteristics-Area Under the Curve (ROC-AUC) analysis, yielding a commendable prediction accuracy of 0.773. This MCDA approach offers an efficient and direct means of flood risk modeling, utilizing fundamental GIS data. The model serves as a valuable tool for decision-makers, enhancing flood risk awareness and providing vital insights for disaster management authorities in the Hunza-Nagar Valley. As future developments unfold, this study remains an indispensable resource for disaster preparedness and management in the Hunza-Nagar Valley region.

Prediction of diabetes progress based on machine learning approach

Uropathy is a serious chronic disease whose prevalence is increasing at an alarming rate. Early detection and prediction of diabetes in women is important because of the increased risk of diabetes-related complications during pregnancy. This study introduces machine learning models to assess the likelihood of diabetes in women. The importance of studying characteristics and improving prediction accuracy to understand the nuances of categorization. Specifically, for data preprocessing, experiments are conducted to solve the problem of missing values and outliers by replacing the zero values of certain features with the median values of the corresponding features. This step reduces the impact of less reliable data on model performance. As recognition models, Gaussian Naive Bayes (GNB), Support Vector Machine (SVM), and Random Forest (RF) are built. Performance analysis is performed along with a careful exploration of the hyperparameter space. Scores for Receiver Operating Characteristic - Area Under the Curve (ROC-AUC) are used to compare various models. Different features affect the classification to different degrees. The experimental findings indicate that the modified random forest model demonstrates superior prediction accuracy and robustness. These findings can assist physicians in predicting a patient's risk of developing diabetes earlier.

Hemodynamic Response to Exercise Training in Heart Failure With Reduced Ejection Fraction Patients.

Supervised exercise training decreases total and cardiac mortality and increases quality of life of heart failure with reduced ejection fraction (HFrEF) patients. However, response to training is variable from one patient to another and factors responsible for a positive response to training remain unclear. The aims of the study were to compare cardiac hemodynamic changes after an exercise training program in responders (R) versus non-responders (NR) HFrEF patients, and to compare different discriminators used to assess response to training. Seventy-six HFrEF patients (86% males, 57 ± 12 years) completed an exercise training program for 4 weeks. Patients underwent cardiopulmonary exercise testing (CPET) on a cycle ergometer before and after training. Cardiac hemodynamics were measured by impedance cardiography during CPET. The R and NR groups were classified using the median change in peak oxygen uptake (V̇O2peak). There were statistically significant differences in V̇O2peak (+35% vs. -1%, P < 0.0001) and in peaks of ventilation (+30% vs. +2%, P < 0.0001), cardiac output (COpeak) (+25% vs. +4%, P < 0.01), systolic blood pressure (+12% vs. +2%, P < 0.05), diastolic blood pressure (+9% vs. +4%, P < 0.05) and heart rate (+8% vs. +1%, P < 0.01) between R and NR after the training program. V̇O2peak was the best discriminator between R and NR (receiver operating characteristic (ROC) area under the curve (AUC) = 0.83, P < 0.0001), followed by COpeak (ROC AUC = 0.77, P < 0.0001). V̇O2peak is the best discriminator between HFrEF R and NR patients after the training program. Responders showed improvements in peak hemodynamic parameters. These results pave the way for other studies to determine how the individualization of exercise training programs and peak hemodynamic parameters potentially linked to a better positive response status.

Detecting anomalies in security cameras with 3D-convolutional neural network and convolutional long short-term memory

&lt;p&gt;This paper presents a novel deep learning-based approach for anomaly detection in surveillance films. A deep network that has been trained to recognize objects and human activity in movies forms the foundation of the suggested approach. In order to detect anomalies in surveillance films, the proposed method combines the strengths of 3D-convolutional neural network (3DCNN) and convolutional long short-term memory (ConvLSTM). From the video frames, the 3DCNN is utilized to extract spatiotemporal features,while ConvLSTM is employed to record temporal relationships between frames. The technique was evaluated on five large-scale datasets from the actual world (UCFCrime, XDViolence, UBIFights, CCTVFights, UCF101) that had both indoor and outdoor video clips as well as synthetic datasets with a range of object shapes, sizes, and behaviors. The results further demonstrate that combining 3DCNN with ConvLSTM can increase precision and reduce false positives, achieving a high accuracy and area under the receiver operating characteristic-area under the curve (ROC-AUC) in both indoor and outdoor scenarios when compared to cuttingedge techniques mentioned in the comparison.&lt;/p&gt;

Predictive value and potential association of PET/CT radiomics on lymph node metastasis of cervical cancer

Objective: Due to the information-rich nature of positron emission tomography/computed tomography (PET/CT) images, the authors hope to explore radiomics features that could distinguish metastatic lymph nodes (LNs) from hypermetabolic benign LNs, in addition to conventional indicators. Methods: PET/CT images of 106 patients with early-stage cervical cancer from 2019 to 2021 were retrospectively analyzed. The tumor lesions and LN regions of PET/CT images were outlined with SeeIt, and then radiomics features were extracted. The least absolute shrinkage and selection operator (LASSO) was used to select features. The final selected radiomics features of LNs were used as predictors to construct a machine learning model to predict LN metastasis. Results: The authors determined two morphological coefficient characteristics of cervical lesions (shape – major axis length and shape – mesh volume), one first order characteristics of LNs (first order – 10 percentile) and two gray-level co-occurrence matrix (GLCM) characteristics of LNs (GLCM – id and GLCM – inverse variance) were closely related to LN metastasis. Finally, a neural network was constructed based on the radiomic features of the LNs. The area under the curve of receiver operating characteristic (AUC-ROC) of the model was 0.983 in the training set and 0.860 in the test set. Conclusion: The authors constructed and demonstrated a neural network based on radiomics features of PET/CT to evaluate the risk of single LN metastasis in early-stage cervical cancer.

Machine learning to identify chronic cough from administrative claims data

Accurate identification of patient populations is an essential component of clinical research, especially for medical conditions such as chronic cough that are inconsistently defined and diagnosed. We aimed to develop and compare machine learning models to identify chronic cough from medical and pharmacy claims data. In this retrospective observational study, we compared 3 machine learning algorithms based on XG Boost, logistic regression, and neural network approaches using a large claims and electronic health record database. Of the 327,423 patients who met the study criteria, 4,818 had chronic cough based on linked claims–electronic health record data. The XG Boost model showed the best performance, achieving a Receiver-Operator Characteristic Area Under the Curve (ROC-AUC) of 0.916. We selected a cutoff that favors a high positive predictive value (PPV) to minimize false positives, resulting in a sensitivity, specificity, PPV, and negative predictive value of 18.0%, 99.6%, 38.7%, and 98.8%, respectively on the held-out testing set (n = 82,262). Logistic regression and neural network models achieved slightly lower ROC-AUCs of 0.907 and 0.838, respectively. The XG Boost and logistic regression models maintained their robust performance in subgroups of individuals with higher rates of chronic cough. Machine learning algorithms are one way of identifying conditions that are not coded in medical records, and can help identify individuals with chronic cough from claims data with a high degree of classification value.

A deep neural network for MRI spinal inflammation in axial spondyloarthritis.

To develop a deep neural network for the detection of inflammatory spine in short tau inversion recovery (STIR) sequence of magnetic resonance imaging (MRI) on patients with axial spondyloarthritis (axSpA). A total 330 patients with axSpA were recruited. STIR MRI of the whole spine and clinical data were obtained. Regions of interests (ROIs) were drawn outlining the active inflammatory lesion consisting of bone marrow edema (BME). Spinal inflammation was defined by the presence of an active inflammatory lesion on the STIR sequence. The 'fake-color' images were constructed. Images from 270 and 60 patients were randomly separated into the training/validation and testing sets, respectively. Deep neural network was developed using attention UNet. The neural network performance was compared to the image interpretation by a radiologist blinded to the ground truth. Active inflammatory lesions were identified in 2891 MR images and were absent in 14,590 MR images. The sensitivity and specificity of the derived deep neural network were 0.80 ± 0.03 and 0.88 ± 0.02, respectively. The Dice coefficient of the true positive lesions was 0.55 ± 0.02. The area under the curve of the receiver operating characteristic (AUC-ROC) curve of the deep neural network was 0.87 ± 0.02. The performance of the developed deep neural network was comparable to the interpretation of a radiologist with similar sensitivity and specificity. The developed deep neural network showed similar sensitivity and specificity to a radiologist with four years of experience. The results indicated that the network can provide a reliable and straightforward way of interpreting spinal MRI. The use of this deep neural network has the potential to expand the use of spinal MRI in managing axSpA.

TBscreen: A passive cough classifier for tuberculosis screening with a controlled dataset.

Recent respiratory disease screening studies suggest promising performance of cough classifiers, but potential biases in model training and dataset quality preclude robust conclusions. To examine tuberculosis (TB) cough diagnostic features, we enrolled subjects with pulmonary TB (N = 149) and controls with other respiratory illnesses (N = 46) in Nairobi. We collected a dataset with 33,000 passive coughs and 1600 forced coughs in a controlled setting with similar demographics. We trained a ResNet18-based cough classifier using images of passive cough scalogram as input and obtained a fivefold cross-validation sensitivity of 0.70 (±0.11 SD). The smartphone-based model had better performance in subjects with higher bacterial load {receiver operating characteristic-area under the curve (ROC-AUC): 0.87 [95% confidence interval (CI): 0.87 to 0.88], P < 0.001} or lung cavities [ROC-AUC: 0.89 (95% CI: 0.88 to 0.89), P < 0.001]. Overall, our data suggest that passive cough features distinguish TB from non-TB subjects and are associated with bacterial burden and disease severity.

Large-Scale Machine Learning Analysis Reveals DNA Methylation and Gene Expression Response Signatures for Gemcitabine-Treated Pancreatic Cancer.

Background: Gemcitabine is a first-line chemotherapy for pancreatic adenocarcinoma (PAAD), but many PAAD patients do not respond to gemcitabine-containing treatments. Being able to predict such nonresponders would hence permit the undelayed administration of more promising treatments while sparing gemcitabine life-threatening side effects for those patients. Unfortunately, the few predictors of PAAD patient response to this drug are weak, none of them exploiting yet the power of machine learning (ML). Methods: Here, we applied ML to predict the response of PAAD patients to gemcitabine from the molecular profiles of their tumors. More concretely, we collected diverse molecular profiles of PAAD patient tumors along with the corresponding clinical data (gemcitabine responses and clinical features) from the Genomic Data Commons resource. From systematically combining 8 tumor profiles with 16 classification algorithms, each of the resulting 128 ML models was evaluated by multiple 10-fold cross-validations. Results: Only 7 of these 128 models were predictive, which underlines the importance of carrying out such a large-scale analysis to avoid missing the most predictive models. These were here random forest using 4 selected mRNAs [0.44 Matthews correlation coefficient (MCC), 0.785 receiver operating characteristic-area under the curve (ROC-AUC)] and XGBoost combining 12 DNA methylation probes (0.32 MCC, 0.697 ROC-AUC). By contrast, the hENT1 marker obtained much worse random-level performance (practically 0 MCC, 0.5 ROC-AUC). Despite not being trained to predict prognosis (overall and progression-free survival), these ML models were also able to anticipate this patient outcome. Conclusions: We release these promising ML models so that they can be evaluated prospectively on other gemcitabine-treated PAAD patients.

Predicting delirium and the effects of medications in hospitalized COVID-19 patients using machine learning: A retrospective study within the Korean Multidisciplinary Cohort for Delirium Prevention (KoMCoDe).

Delirium is commonly reported from the inpatients with Coronavirus disease 2019 (COVID-19) infection. As delirium is closely associated with adverse clinical outcomes, prediction and prevention of delirium is critical. We developed a machine learning (ML) model to predict delirium in hospitalized patients with COVID-19 and to identify modifiable factors to prevent delirium. The data set (n = 878) from four medical centers was constructed. Total of 78 predictors were included such as demographic characteristics, vital signs, laboratory results and medication, and the primary outcome was delirium occurrence during hospitalization. For analysis, the extreme gradient boosting (XGBoost) algorithm was applied, and the most influential factors were selected by recursive feature elimination. Among the indicators of performance for ML model, the area under the curve of the receiver operating characteristic (AUROC) curve was selected as the evaluation metric. Regarding the performance of developed delirium prediction model, the accuracy, precision, recall, F1 score, and the AUROC were calculated (0.944, 0.581, 0.421, 0.485, 0.873, respectively). The influential factors of delirium in this model included were mechanical ventilation, medication (antipsychotics, sedatives, ambroxol, piperacillin/tazobactam, acetaminophen, ceftriaxone, and propacetamol), and sodium ion concentration (all p < 0.05). We developed and internally validated an ML model to predict delirium in COVID-19 inpatients. The model identified modifiable factors associated with the development of delirium and could be clinically useful for the prediction and prevention of delirium in COVID-19 inpatients.

Applying contrastive pre-training for depression and anxiety risk prediction in type 2 diabetes patients based on heterogeneous electronic health records: a primary healthcare case study.

Due to heterogeneity and limited medical data in primary healthcare services (PHS), assessing the psychological risk of type 2 diabetes mellitus (T2DM) patients in PHS is difficult. Using unsupervised contrastive pre-training, we proposed a deep learning framework named depression and anxiety prediction (DAP) to predict depression and anxiety in T2DM patients. The DAP model consists of two sub-models. Firstly, the pre-trained model of DAP used unlabeled discharge records of 85085 T2DM patients from the First Affiliated Hospital of Nanjing Medical University for unsupervised contrastive learning on heterogeneous electronic health records (EHRs). Secondly, the fine-tuned model of DAP used case-control cohorts (17491 patients) selected from 149596 T2DM patients' EHRs in the Nanjing Health Information Platform (NHIP). The DAP model was validated in 1028 patients from PHS in NHIP. Evaluation included receiver operating characteristic area under the curve (ROC-AUC) and precision-recall area under the curve (PR-AUC), and decision curve analysis (DCA). The pre-training step allowed the DAP model to converge at a faster rate. The fine-tuned DAP model significantly outperformed the baseline models (logistic regression, extreme gradient boosting, and random forest) with ROC-AUC of 0.91±0.028 and PR-AUC of 0.80±0.067 in 10-fold internal validation, and with ROC-AUC of 0.75 ± 0.045 and PR-AUC of 0.47 ± 0.081 in external validation. The DCA indicate the clinical potential of the DAP model. The DAP model effectively predicted post-discharge depression and anxiety in T2DM patients from PHS, reducing data fragmentation and limitations. This study highlights the DAP model's potential for early detection and intervention in depression and anxiety, improving outcomes for diabetes patients.

Machine learning-based prediction of medication refill adherence among first-time insulin users with type 2 diabetes

AimsThe prevalence of Type 2 Diabetes Mellitus (T2DM) is projected to be 7 % in 2030. Despite its need for long-term diabetes care, the adherence rate of injectable medications such as insulin is around 60 %, lower than the acceptable threshold of 80 %. This study aims to create classification models to predict insulin adherence among adult T2DM naïve insulin users. MethodsClinical data were extracted from Taipei Medical University Clinical Research Database (TMUCRD) from January 1st, 2004 to December 30th, 2020. A patient was regarded as adherent if his/her medication possession ratio (MPR) was at least 80 %. Seven domains of predictors were created, including demographics, baseline medications, baseline comorbidities, baseline laboratory data, healthcare resource utilization, index insulins, and the concomitant non-insulin T2DM medications. We built two Xgboost models for internal and external testing respectively. ResultsUsing a cohort of 4134 patients from Taiwan, our model achieved the Area Under the curve of the Receiver Operating Characteristic (AUROC) of the internal test was 0.782 and the AUROC of the external test was 0.771. the SHAP (SHapley Additive exPlanations) value showed that the number of prescribed medications, the number of outpatient visits, and laboratory data were predictive of future insulin adherence. ConclusionsThis is the first study to predict adherence among adult naïve insulin users. The developed model is a potential clinical decision support tool to identify possible non-adherent patients for healthcare providers to design individualized education plans.