Explainable artificial intelligence (XAI) for predicting the need for intubation in methanol-poisoned patients: a study comparing deep and machine learning models.

The need for intubation in methanol-poisoned patients, if not predicted in time, can lead to irreparable complications and even death. Artificial intelligence (AI) techniques like machine learning (ML) and deep learning (DL) greatly aid in accurately predicting intubation needs for methanol-poisoned patients. So, our study aims to assess Explainable Artificial Intelligence (XAI) for predicting intubation necessity in methanol-poisoned patients, comparing deep learning and machine learning models. This study analyzed a dataset of 897 patient records from Loghman Hakim Hospital in Tehran, Iran, encompassing cases of methanol poisoning, including those requiring intubation (202 cases) and those not requiring it (695 cases). Eight established ML (SVM, XGB, DT, RF) and DL (DNN, FNN, LSTM, CNN) models were used. Techniques such as tenfold cross-validation and hyperparameter tuning were applied to prevent overfitting. The study also focused on interpretability through SHAP and LIME methods. Model performance was evaluated based on accuracy, specificity, sensitivity, F1-score, and ROC curve metrics. Among DL models, LSTM showed superior performance in accuracy (94.0%), sensitivity (99.0%), specificity (94.0%), and F1-score (97.0%). CNN led in ROC with 78.0%. For ML models, RF excelled in accuracy (97.0%) and specificity (100%), followed by XGB with sensitivity (99.37%), F1-score (98.27%), and ROC (96.08%). Overall, RF and XGB outperformed other models, with accuracy (97.0%) and specificity (100%) for RF, and sensitivity (99.37%), F1-score (98.27%), and ROC (96.08%) for XGB. ML models surpassed DL models across all metrics, with accuracies from 93.0% to 97.0% for DL and 93.0% to 99.0% for ML. Sensitivities ranged from 98.0% to 99.37% for DL and 93.0% to 99.0% for ML. DL models achieved specificities from 78.0% to 94.0%, while ML models ranged from 93.0% to 100%. F1-scores for DL were between 93.0% and 97.0%, and for ML between 96.0% and 98.27%. DL models scored ROC between 68.0% and 78.0%, while ML models ranged from 84.0% to 96.08%. Key features for predicting intubation necessity include GCS at admission, ICU admission, age, longer folic acid therapy duration, elevated BUN and AST levels, VBG_HCO3 at initial record, and hemodialysis presence. This study as the showcases XAI's effectiveness in predicting intubation necessity in methanol-poisoned patients. ML models, particularly RF and XGB, outperform DL counterparts, underscoring their potential for clinical decision-making.

Global incidence and mortality trends of gastric cancer and predicted mortality of gastric cancer by 2035.

OBJECTIVE: To study the historical global incidence and mortality trends of gastric cancer and predicted mortality of gastric cancer by 2035. METHODS: Incidence data were retrieved from the Cancer Incidence in Five Continents (CI5) volumes I-XI, and mortality data were obtained from the latest update of the World Health Organization (WHO) mortality database. We used join-point regression analysis to examine historical incidence and mortality trends and used the package NORDPRED in R to predict the number of deaths and mortality rates by 2035 by country and sex. RESULTS: More than 1,089,000 new cases of gastric cancer and 769,000 related deaths were reported in 2020. The average annual percent change (AAPC) in the incidence of gastric cancer from 2003 to 2012 among the male population, South Korea, Japan, Malta, Canada, Cyprus, and Switzerland showed an increasing trend (P > 0.05); among the female population, Canada [AAPC, 1.2; (95%Cl, 0.5-2), P < 0.05] showed an increasing trend; and South Korea, Ecuador, Thailand, and Cyprus showed an increasing trend (P > 0.05). AAPC in the mortality of gastric cancer from 2006 to 2015 among the male population, Thailand [3.5 (95%cl, 1.6-5.4), P < 0.05] showed an increasing trend; Malta Island, New Zealand, Turkey, Switzerland, and Cyprus had an increasing trend (P > 0.05); among the male population aged 20-44, Thailand [AAPC, 3.4; (95%cl, 1.3-5.4), P < 0.05] showed an increasing trend; Norway, New Zealand, The Netherlands, Slovakia, France, Colombia, Lithuania, and the USA showed an increasing trend (P > 0.05). It is predicted that the mortality rate in Slovenia and France's female population will show an increasing trend by 2035. It is predicted that the absolute number of deaths in the Israeli male population and in Chile, France, and Canada female population will increase by 2035. CONCLUSION: In the past decade, the incidence and mortality of gastric cancer have shown a decreasing trend; however, there are still some countries showing an increasing trend, especially among populations younger than 45 years. Although mortality in most countries is predicted to decline by 2035, the absolute number of deaths due to gastric cancer may further increase due to population growth.

Syndromic Surveillance Tracks COVID-19 Cases in University and County Settings: Retrospective Observational Study.

Background: Syndromic surveillance represents a potentially inexpensive supplement to test-based COVID-19 surveillance. By strengthening surveillance of COVID-19-like illness (CLI), targeted and rapid interventions can be facilitated that prevent COVID-19 outbreaks without primary reliance on testing. Objective: This study aims to assess the temporal relationship between confirmed SARS-CoV-2 infections and self-reported and health care provider-reported CLI in university and county settings, respectively. Methods: We collected aggregated COVID-19 testing and symptom reporting surveillance data from Cornell University (2020-2021) and Tompkins County Health Department (2020-2022). We used negative binomial and linear regression models to correlate confirmed COVID-19 case counts and positive test rates with CLI rate time series, lagged COVID-19 cases or rates, and day of the week as independent variables. Optimal lag periods were identified using Granger causality and likelihood ratio tests. Results: In modeling undergraduate student cases, the CLI rate (P=.003) and rate of exposure to CLI (P<.001) were significantly correlated with the COVID-19 test positivity rate with no lag in the linear models. At the county level, the health care provider-reported CLI rate was significantly correlated with SARS-CoV-2 test positivity with a 3-day lag in both the linear (P<.001) and negative binomial model (P=.005). Conclusions: The real-time correlation between syndromic surveillance and COVID-19 cases on a university campus suggests symptom reporting is a viable alternative or supplement to COVID-19 surveillance testing. At the county level, syndromic surveillance is also a leading indicator of COVID-19 cases, enabling quick action to reduce transmission. Further research should investigate COVID-19 risk using syndromic surveillance in other settings, such as low-resource settings like low- and middle-income countries.

Patient-reported physical well-being predicts good long-term survival of hematopoietic stem cell transplantation.

AIM: This study aimed to explore the association between patient-reported items at different time points after hematopoietic stem cell transplantation (HSCT) and long-term survival. METHODS: We conducted a study with 144 allogeneic HSCT patients, following them for 5 years post-transplantation. Data from the Functional Assessment of Cancer Therapy-Bone Marrow Transplant (FACT-BMT) questionnaire were collected before transplantation and at 1, 3, 6, 12, 18, 36, and 60 months after transplantation. Demographic characteristics and survival status were also assessed. RESULTS: Among the 144 cases, the 5-year overall survival (OS), progression-free survival (PFS), non-relapse mortality (NRM), and graft-versus-host disease-free (GRFS) rates were 65%, 48%, 17%, and 36% respectively. Health-related quality of life (HRQOL) showed a fluctuating pattern over 5 years. Using a latent class mixed model, patients were classified into two groups based on their physical well-being (PWB) scores during the 60-month follow-up. Class 1 had initially lower PWB scores, which gradually increased over time. In contrast, Class 2 maintained higher PWB scores with slight increases over time. Kaplan-Meier survival analysis revealed that Class 1 had better OS (70.9% vs. 52.9%, p = 0.021), PFS (60.5% vs. 41.2%, p = 0.039), and GRFS (35.1% vs. 29.3%, p = 0.035) compared to Class 2. CONCLUSIONS: Patients who had higher initial PWB scores after HSCT demonstrated improved long-term survival outcomes. The PWB score could serve as a valuable predictor for the prognosis of HSCT.

Machine Learning Models for Prediction of Maternal Hemorrhage and Transfusion: Model Development Study.

BACKGROUND: Current postpartum hemorrhage (PPH) risk stratification is based on traditional statistical models or expert opinion. Machine learning could optimize PPH prediction by allowing for more complex modeling. OBJECTIVE: We sought to improve PPH prediction and compare machine learning and traditional statistical methods. METHODS: We developed models using the Consortium for Safe Labor data set (2002-2008) from 12 US hospitals. The primary outcome was a transfusion of blood products or PPH (estimated blood loss of ≥1000 mL). The secondary outcome was a transfusion of any blood product. Fifty antepartum and intrapartum characteristics and hospital characteristics were included. Logistic regression, support vector machines, multilayer perceptron, random forest, and gradient boosting (GB) were used to generate prediction models. The area under the receiver operating characteristic curve (ROC-AUC) and area under the precision/recall curve (PR-AUC) were used to compare performance. RESULTS: Among 228,438 births, 5760 (3.1%) women had a postpartum hemorrhage, 5170 (2.8%) had a transfusion, and 10,344 (5.6%) met the criteria for the transfusion-PPH composite. Models predicting the transfusion-PPH composite using antepartum and intrapartum features had the best positive predictive values, with the GB machine learning model performing best overall (ROC-AUC=0.833, 95% CI 0.828-0.838; PR-AUC=0.210, 95% CI 0.201-0.220). The most predictive features in the GB model predicting the transfusion-PPH composite were the mode of delivery, oxytocin incremental dose for labor (mU/minute), intrapartum tocolytic use, presence of anesthesia nurse, and hospital type. CONCLUSIONS: Machine learning offers higher discriminability than logistic regression in predicting PPH. The Consortium for Safe Labor data set may not be optimal for analyzing risk due to strong subgroup effects, which decreases accuracy and limits generalizability.

Insights into the performance of PREDICT tool in a large Mainland Chinese breast cancer cohort: a comparative analysis of versions 3.0 and 2.2.

BACKGROUND: PREDICT is a web-based tool for forecasting breast cancer outcomes. PREDICT version 3.0 was recently released. This study aimed to validate this tool for a large population in mainland China and compare v3.0 with v2.2. METHODS: Women who underwent surgery for nonmetastatic primary invasive breast cancer between 2010 and 2020 from the First Affiliated Hospital of Wenzhou Medical University were selected. Predicted and observed 5-year overall survival (OS) for both v3.0 and v2.2 were compared. Discrimination was compared using receiver-operator curves and DeLong test. Calibration was evaluated using calibration plots and chi-squared test. A difference greater than 5% was deemed clinically relevant. RESULTS: A total of 5424 patients were included, with median follow-up time of 58 months (IQR 38-89 months). Compared to v2.2, v3.0 did not show improved discriminatory accuracy for 5-year OS (AUC: 0.756 vs 0.771), same as ER-positive and ER-negative patients. However, calibration was significantly improved in v3.0, with predicted 5-year OS deviated from observed by -2.0% for the entire cohort, -2.9% for ER-positive and -0.0% for ER-negative patients, compared to -7.3%, -4.7% and -13.7% in v2.2. In v3.0, 5-year OS was underestimated by 9.0% for patients older than 75 years, and 5.8% for patients with micrometastases. Patients with distant metastases postdiagnosis was overestimated by 10.6%. CONCLUSIONS: PREDICT v3.0 reliably predicts 5-year OS for the majority of Chinese patients with breast cancer. PREDICT v3.0 significantly improved the predictive accuracy for ER-negative groups. Furthermore, caution is advised when interpreting 5-year OS for patients aged over 70, those with micrometastases or metastases postdiagnosis.

Construction of a preoperative nomogram model for predicting perineural invasion in advanced gastric cancer.

Objective: This study aimed to develop and validate a clinical and imaging-based nomogram for preoperatively predicting perineural invasion (PNI) in advanced gastric cancer. Methods: A retrospective cohort of 351 patients with advanced gastric cancer who underwent surgical resection was included. Multivariable logistic regression analysis was conducted to identify independent risk factors for PNI and to construct the nomogram. The performance of the nomogram was assessed using calibration curves, the concordance index (C-index), the area under the curve (AUC), and decision curve analysis (DCA). The disparity in disease-free survival (DFS) between the nomogram-predicted PNI-positive group and the nomogram-predicted PNI-negative group was evaluated using the Log-Rank test and Kaplan-Meier analysis. Results: Extramural vascular invasion (EMVI), Borrmann classification, tumor thickness, and the systemic inflammation response index (SIRI) emerged as independent risk factors for PNI. The nomogram model demonstrated a commendable AUC value of 0.838. Calibration curves exhibited excellent concordance, with a C-index of 0.814. DCA indicated that the model provided good clinical net benefit. The DFS of the nomogram-predicted PNI-positive group was significantly lower than that of the nomogram-predicted PNI-negative group (p < 0.001). Conclusion: This study successfully developed a preoperative nomogram model that not only effectively predicted PNI in gastric cancer but also facilitated postoperative risk stratification.

Defining Activity Thresholds Triggering a "Stand Hour" for Apple Watch Users: Cross-Sectional Study.

BACKGROUND: Sedentary behavior (SB) is one of the largest contributing factors increasing the risk of developing noncommunicable diseases, including cardiovascular disease and type 2 diabetes. Guidelines from the World Health Organization for physical activity suggest the substitution of SB with light physical activity. The Apple Watch contains a health metric known as the stand hour (SH). The SH is intended to record standing with movement for at least 1 minute per hour; however, the activity measured during the determination of the SH is unclear. OBJECTIVE: In this cross-sectional study, we analyzed the algorithm used to determine time spent standing per hour. To do this, we investigated activity measurements also recorded on Apple Watches that influence the recording of an SH. We also aimed to estimate the values of any significant SH predictors in the recording of a SH. METHODS: The cross-sectional study used anonymized data obtained in August 2022 from 20 healthy individuals gathered via convenience sampling. Apple Watch data were extracted from the Apple Health app through the use of a third-party app. Appropriate statistical models were fitted to analyze SH predictors. RESULTS: Our findings show that active energy (AE) and step count (SC) measurements influence the recording of an SH. Comparing when an SH is recorded with when an SH is not recorded, we found a significant difference in the mean and median AE and SC. Above a threshold of 97.5 steps or 100 kJ of energy, it became much more likely that an SH would be recorded when each predictor was analyzed as a separate entity. CONCLUSIONS: The findings of this study reveal the pivotal role of AE and SC measurements in the algorithm underlying the SH recording; however, our findings also suggest that a recording of an SH is influenced by more than one factor. Irrespective of the internal validity of the SH metric, it is representative of light physical activity and might, therefore, have use in encouraging individuals through various means, for example, notifications, to reduce their levels of SB.

Framework for Ranking Machine Learning Predictions of Limited, Multimodal, and Longitudinal Behavioral Passive Sensing Data: Combining User-Agnostic and Personalized Modeling.

BACKGROUND: Passive mobile sensing provides opportunities for measuring and monitoring health status in the wild and outside of clinics. However, longitudinal, multimodal mobile sensor data can be small, noisy, and incomplete. This makes processing, modeling, and prediction of these data challenging. The small size of the data set restricts it from being modeled using complex deep learning networks. The current state of the art (SOTA) tackles small sensor data sets following a singular modeling paradigm based on traditional machine learning (ML) algorithms. These opt for either a user-agnostic modeling approach, making the model susceptible to a larger degree of noise, or a personalized approach, where training on individual data alludes to a more limited data set, giving rise to overfitting, therefore, ultimately, having to seek a trade-off by choosing 1 of the 2 modeling approaches to reach predictions. OBJECTIVE: The objective of this study was to filter, rank, and output the best predictions for small, multimodal, longitudinal sensor data using a framework that is designed to tackle data sets that are limited in size (particularly targeting health studies that use passive multimodal sensors) and that combines both user agnostic and personalized approaches, along with a combination of ranking strategies to filter predictions. METHODS: In this paper, we introduced a novel ranking framework for longitudinal multimodal sensors (FLMS) to address challenges encountered in health studies involving passive multimodal sensors. Using the FLMS, we (1) built a tensor-based aggregation and ranking strategy for final interpretation, (2) processed various combinations of sensor fusions, and (3) balanced user-agnostic and personalized modeling approaches with appropriate cross-validation strategies. The performance of the FLMS was validated with the help of a real data set of adolescents diagnosed with major depressive disorder for the prediction of change in depression in the adolescent participants. RESULTS: Predictions output by the proposed FLMS achieved a 7% increase in accuracy and a 13% increase in recall for the real data set. Experiments with existing SOTA ML algorithms showed an 11% increase in accuracy for the depression data set and how overfitting and sparsity were handled. CONCLUSIONS: The FLMS aims to fill the gap that currently exists when modeling passive sensor data with a small number of data points. It achieves this through leveraging both user-agnostic and personalized modeling techniques in tandem with an effective ranking strategy to filter predictions.

Privacy-Preserving Federated Survival Support Vector Machines for Cross-Institutional Time-To-Event Analysis: Algorithm Development and Validation.

BACKGROUND: Central collection of distributed medical patient data is problematic due to strict privacy regulations. Especially in clinical environments, such as clinical time-to-event studies, large sample sizes are critical but usually not available at a single institution. It has been shown recently that federated learning, combined with privacy-enhancing technologies, is an excellent and privacy-preserving alternative to data sharing. OBJECTIVE: This study aims to develop and validate a privacy-preserving, federated survival support vector machine (SVM) and make it accessible for researchers to perform cross-institutional time-to-event analyses. METHODS: We extended the survival SVM algorithm to be applicable in federated environments. We further implemented it as a FeatureCloud app, enabling it to run in the federated infrastructure provided by the FeatureCloud platform. Finally, we evaluated our algorithm on 3 benchmark data sets, a large sample size synthetic data set, and a real-world microbiome data set and compared the results to the corresponding central method. RESULTS: Our federated survival SVM produces highly similar results to the centralized model on all data sets. The maximal difference between the model weights of the central model and the federated model was only 0.001, and the mean difference over all data sets was 0.0002. We further show that by including more data in the analysis through federated learning, predictions are more accurate even in the presence of site-dependent batch effects. CONCLUSIONS: The federated survival SVM extends the palette of federated time-to-event analysis methods by a robust machine learning approach. To our knowledge, the implemented FeatureCloud app is the first publicly available implementation of a federated survival SVM, is freely accessible for all kinds of researchers, and can be directly used within the FeatureCloud platform.

Moderate/severe biatrial dilation predicts adverse events after ablation in atrial fibrillation with heart failure.

AIMS: To retrospectively compare the long-term outcomes following atrial fibrillation (AF) ablation between heart failure (HF) with preserved ejection fraction (EF) (HFpEF) and reduced/mildly reduced EF (HFr-mrEF) patients, and to identify novel predictors of adverse clinical events. METHODS: In total, 1402 AF patients with HF who underwent successful ablation were consecutively enrolled. Adverse clinical events including all-cause death, HF hospitalization, and stroke were followed up. Cox proportional hazards models were used to assess the associations between clinical factors and events. Kaplan-Meier analysis was performed to estimate the cumulative incidences of these events. A receiver operating characteristic curve was used to test the ability of these predictors. RESULTS: During a follow-up period of 42 ± 15 months, 265 (18.9%) patients experienced adverse clinical events after ablation. The cumulative incidence of adverse clinical events was significantly higher in HFr-mrEF than in HFpEF (25.4% vs. 15.7%, P < 0.001), the similar tendency was observed on all-cause death (10.5% vs. 6.5%, P = 0.011) and HF hospitalization (17.2% vs. 10.1%, P < 0.001). After multivariate adjustment, non-paroxysmal AF [hazard ratio (HR) 1.922, 95% confidence interval (CI) 1.130-3.268, P = 0.016], LAD ≥ 45 mm (HR 2.197, 95% CI 1.206-4.003, P < 0.001), LVEF (HR 0.959, 95% CI 0.946-0.981, P < 0.001), and RAD ≥ 45 mm (HR 2.044, 95% CI 1.362-3.238, P < 0.001) remained the independent predictors for developing adverse clinical events. A predictive model performed with non-paroxysmal AF, LAD ≥ 45 mm and RAD ≥ 45 mm yielded an area under curve of 0.728 (95% CI 0.696-0.760, P < 0.001). CONCLUSIONS: AF patients with HFpEF had better long-term outcomes than those with HFr-mrEF, and moderate/severe biatrial dilation could predict adverse clinical events following catheter ablation in AF and HF patients.

A machine learning model to predict spontaneous vaginal delivery failure for term nulliparous women: An observational study.

OBJECTIVE: This study aims to construct and evaluate a model to predict spontaneous vaginal delivery (SVD) failure in term nulliparous women based on machine learning algorithms. METHODS: In this retrospective observational study, data on nulliparous women without contraindications for vaginal delivery with a singleton pregnancy ≥37 weeks and before the onset of labor from September 2020 to September 2021 were divided into a training set and a temporal validation set. Transperineal ultrasound was performed to collect angle of progression, head-perineum distance, subpubic arch angle, and their levator hiatal dimensions. The cervical length was measured via transvaginal ultrasound. The delivery methods were later recorded. Through LASSO regression analysis, indicators that can affect SVD failure were selected. Seven common machine learning algorithms were selected for model training, and the optimal algorithm was selected based on the area under the curve (AUC) to evaluate the effectiveness of the validation model. RESULTS: Four indicators related to SVD failure were identified through LASSO regression screening: angle of progression, cervical length, subpubic arch angle, and estimated fetal weight. The Gaussian NB algorithm was found to yield the highest AUC (0.82, 95% confidence interval [CI] 0.65-0.98) during model training, and hence it was chosen for verification with the temporal validation set, in which an AUC of 0.79 (95% CI 0.64-0.95) was obtained with accuracy, sensitivity, and specificity rates of 80.9%, 72.7%, and 75.0%, respectively. CONCLUSION: The Gaussian NB model showed good predictive effect, proving its potential as a clinical reference for predicting SVD failure of term nulliparous women before actual delivery.

Application of Chest CT Imaging Feature Model in Distinguishing Squamous Cell Carcinoma and Adenocarcinoma of the Lung.

Purpose: In situations where pathological acquisition is difficult, there is a lack of consensus on distinguishing between adenocarcinoma and squamous cell carcinoma from imaging images, and each doctor can only make judgments based on their own experience. This study aims to extract imaging features of chest CT, extract sensitive factors through logistic univariate and multivariate analysis, and model to distinguish between lung squamous cell carcinoma and lung adenocarcinoma. Methods: We downloaded chest CT scans with clear diagnosis of adenocarcinoma and squamous cell carcinoma from The Cancer Imaging Archive (TCIA), extracted 19 imaging features by a radiologist and a thoracic surgeon, including location, spicule, lobulation, cavity, vacuolar sign, necrosis, pleural traction sign, vascular bundle sign, air bronchogram sign, calcification, enhancement degree, distance from pulmonary hilum, atelectasis, pulmonary hilum and bronchial lymph nodes, mediastinal lymph nodes, interlobular septal thickening, pulmonary metastasis, adjacent structures invasion, pleural effusion. Firstly, we apply the glm function of R language to perform logistic univariate analysis on all variables to select variables with P < 0.1. Then, perform logistic multivariate analysis on the selected variables to obtain a predictive model. Next, use the roc function in R language to calculate the AUC value and draw the ROC curve, use the val.prob function in R language to draw the Calibrat curve, and use the rmda package in R language to draw the DCA curve and clinical impact curve. At the same time, 45 patients diagnosed with lung squamous cell carcinoma and lung adenocarcinoma through surgery or biopsy in the Radiotherapy Department and Thoracic Surgery Department of our hospital from 2023 to 2024 were included in the validation group. The chest CT features were jointly determined and recorded by the two doctors mentioned above and included in the validation group. The included image feature data are complete and does not require preprocessing, so directly entering statistical calculations. Perform ROC curves, calibration curves, DCA, and clinical impact curves in the validation group to further validate the predictive model. If the predictive model performs well in the validation group, further draw a nomogram to demonstrate. Results: This study extracted 19 imaging features from the chest CT scans of 75 patients downloaded from TCIA and finally selected 18 complete data for analysis. First, univariate analysis and multivariate analysis were performed, and a total of 5 variables were obtained: spicule, necrosis, air bronchogram Sign, atelectasis, pulmonary hilum and bronchial lymph nodes. After conducting modeling analysis with AUC = 0.887, a validation group was established using clinical cases from our hospital, Draw ROC curve with AUC = 0.865 in the validation group, evaluate the accuracy of the model through Calibrate calibration curve, evaluate the reliability of the model in clinical practice through DCA curve, and further evaluate the practicality of the model in clinical practice through clinical impact curve. Conclusion: It is possible to extract influential features from ordinary chest CT scans to determine lung adenocarcinoma and squamous cell carcinoma. The model we have set up performs well in terms of discrimination, accuracy, reliability, and practicality.

Development and validation of a nomogram for suicide attempts in patients with first-episode drug-naïve major depressive disorder.

Objective: The risk of suicide can be decreased by accurately identifying high-risk suicide groups and implementing the right interventions. The aim of this study was to develop a nomogram for suicide attempts (SA) in patients with first-episode drug-naïve (FEDN) major depressive disorder (MDD). Methods: This study undertook a cross-sectional analysis of 1,718 patients diagnosed with FEDN MDD, providing comprehensive clinical data from September 2016 to December 2018. Data on anthropometric and sociodemographic factors were gathered, and the severity of depression and anxiety was evaluated using the 17-item Hamilton Depression Scale (HAMD-17) and the Hamilton Anxiety Scale (HAMA), respectively. Additionally, thyroid hormone levels, lipid profile parameters, and fasting blood glucose (FBG) were measured. Suicide attempt (SA) history was verified based on an amalgamation of medical records, patient interviews, and family interviews. Participants were randomly divided into a training group (70%, n = 1,204) and a validation group (30%, n = 514). In the training group, LASSO analysis and multivariate regression were used to identify variables associated with SA. A nomogram was then constructed using the identified risk factors to estimate the likelihood of SA within the training group. To assess the accuracy, the area under the receiver operating characteristic curve (AUC) was utilized, and calibration plots were employed to evaluate calibration. Additionally, decision curve analysis (DCA) was performed to assess the precision of the model. Finally, internal validation was carried out using the validation group. Results: A practical nomogram has been successfully constructed, incorporating HAMD, HAMA, thyroid stimulating hormone (TSH), thyroid peroxidase antibody (TPOAb), and systolic blood pressure (SBP) parameters, to estimate the probability of SA in Chinese patients diagnosed with FEDN MDD. The pooled area under the ROC for SA risk in both the training and validation groups was found to be 0.802 (95% CI: 0.771 to 0.832) and 0.821 (95% CI: 0.774 to 0.868), respectively. Calibration analysis revealed a satisfactory correlation between the nomogram probabilities and the actual observed probabilities. The clinical applicability of the nomogram was confirmed through decision curve analysis. To enhance accessibility for clinicians and researchers, an online version of the nomogram can be accessed at https://doctorjunjunliu.shinyapps.io/dynnomapp/. Conclusions: We constructed and validated a nomogram for the early detection of FEDN MDD patients with a high risk of SA, thereby contributing to the implementation of effective suicide prevention programs.

Predictive Value of Jugulo-omohyoid Lymph Nodes in Lateral Lymph Node Metastasis of Papillary Thyroid Cancer.

BACKGROUND: Jugulo-omohyoid lymph nodes (JOHLN) metastasis has proven to be associated with lateral lymph node metastasis (LLNM). This study aimed to reveal the clinical features and evaluate the predictive value of JOHLN in PTC to guide the extent of surgery. METHODS: A total of 550 patients pathologically diagnosed with PTC between October 2015 and January 2020, all of whom underwent thyroidectomy and lateral lymph node dissection, were included in this study. RESULTS: Thyroiditis, tumor location, tumor size, extra-thyroidal extension, extra-nodal extension, central lymph node metastasis (CLNM), and LLMM were associated with JOHLN. Male, upper lobe tumor, multifocality, extra-nodal extension, CLNM, and JOHLN metastasis were independent risk factors from LLNM. A nomogram based on predictors performed well. Nerve invasion contributed the most to the prediction model, followed by JOHLN metastasis. The area under the curve (AUC) was 0.855, and the p-value of the Hosmer-Lemeshow goodness of fit test was 0.18. Decision curve analysis showed that the nomogram was clinically helpful. CONCLUSION: JOLHN metastasis could be a clinically sensitive predictor of further LLM. A high-performance nomogram was established, which can provide an individual risk assessment of LNM and guide treatment decisions for patients.

Predictive model for post-induction hypotension in patients undergoing transcatheter aortic valve implantation: a retrospective observational study.

PURPOSE: Post-induction hypotension (PIH) is an independent risk factor for prolonged postoperative stay and hospital death. Patients undergoing transcatheter aortic valve implantation (TAVI) are prone to develop PIH. This study aimed to develop a predictive model for PIH in patients undergoing TAVI. METHODS: This single-center retrospective observational study included 163 patients who underwent TAVI. PIH was defined as at least one measurement of systolic arterial pressure <90 mmHg or at least one incident of norepinephrine infusion at a rate >6 µg/min from anesthetic induction until 20 min post-induction. Multivariate logistic regression analysis was performed to develop a predictive model for PIH in patients undergoing TAVI. RESULTS: In total, 161 patients were analyzed. The prevalence of PIH was 57.8%. Multivariable logistic regression analysis showed that baseline mean arterial pressure ≥90 mmHg [adjusted odds ratio (aOR): 0.413, 95% confidence interval (95% CI): 0.193-0.887; p=0.023] and higher doses of fentanyl (per 1-µg/kg increase, aOR: 0.619, 95% CI: 0.418-0.915; p=0.016) and ketamine (per 1-mg/kg increase, aOR: 0.163, 95% CI: 0.062-0.430; p=0.002) for induction were significantly associated with lower risk of PIH. A higher dose of propofol (per 1-mg/kg increase, aOR: 3.240, 95% CI: 1.320-7.920; p=0.010) for induction was significantly associated with higher risk of PIH. The area under the curve (AUC) for this model was 0.802. CONCLUSION: The present study developed predictive models for PIH in patients who underwent TAVI. This model may be helpful for anesthesiologists in preventing PIH in patients undergoing TAVI.

Novel embedding model predicting the credit card's default using neural network optimized by harmony search algorithm and vortex search algorithm.

In today's banking and financial system, using a credit card has become indispensable. The credit card industry has existed due to a shift in consumer preferences and a rise in national economic growth. The number of issuing banks, card issuers, and transaction volumes has increased significantly. Nevertheless, owing to the growth in the number of transactions made with credit cards, both the total amount due and the rate of defaults on credit card loans have become issues that cannot be neglected. This issue must be resolved to ensure the continued and prosperous growth of the banking industry in the years to come. Currently, a few optimization algorithms-Whale optimization algorithm (WOA), Harmony Search (HS), Multi-verse optimization (MVO), and Vortex Search (VS)-have been used to achieve this purpose. However, because credit card default data is volatile and unequal, it is challenging for typical optimization algorithms to offer steady approaches with optimal performance. Studies have indicated that optimizing algorithms with suitable properties can significantly improve performance. To improve performance, some tuning was applied to the ANN. This study will assess twenty-three parameters, and the efficacy of all four approaches will be compared using ROC and AUC evaluations. The suggested model's performance is contrasted with a scenario where the classifiers were trained using original data. In contrast, the AUC values for VS-MLP were 0.7407 and 0.7271, while those for HS-MLP were 0.7074 and 0.6997. In the training and testing phases, AUC values of 0.7469 and 0.7329 from MVO-MLP and 0.72 and 0.7185 from WOA-MLP, respectively. The results show that the training accuracy of HS, VSA, MVO, and WOA are similar; MVO has the highest training accuracy. The credit card industry can benefit significantly from this methodology, which may help resolve default probabilities.

Improving the Prognostic Evaluation Precision of Hospital Outcomes for Heart Failure Using Admission Notes and Clinical Tabular Data: Multimodal Deep Learning Model.

BACKGROUND: Clinical notes contain contextualized information beyond structured data related to patients' past and current health status. OBJECTIVE: This study aimed to design a multimodal deep learning approach to improve the evaluation precision of hospital outcomes for heart failure (HF) using admission clinical notes and easily collected tabular data. METHODS: Data for the development and validation of the multimodal model were retrospectively derived from 3 open-access US databases, including the Medical Information Mart for Intensive Care III v1.4 (MIMIC-III) and MIMIC-IV v1.0, collected from a teaching hospital from 2001 to 2019, and the eICU Collaborative Research Database v1.2, collected from 208 hospitals from 2014 to 2015. The study cohorts consisted of all patients with critical HF. The clinical notes, including chief complaint, history of present illness, physical examination, medical history, and admission medication, as well as clinical variables recorded in electronic health records, were analyzed. We developed a deep learning mortality prediction model for in-hospital patients, which underwent complete internal, prospective, and external evaluation. The Integrated Gradients and SHapley Additive exPlanations (SHAP) methods were used to analyze the importance of risk factors. RESULTS: The study included 9989 (16.4%) patients in the development set, 2497 (14.1%) patients in the internal validation set, 1896 (18.3%) in the prospective validation set, and 7432 (15%) patients in the external validation set. The area under the receiver operating characteristic curve of the models was 0.838 (95% CI 0.827-0.851), 0.849 (95% CI 0.841-0.856), and 0.767 (95% CI 0.762-0.772), for the internal, prospective, and external validation sets, respectively. The area under the receiver operating characteristic curve of the multimodal model outperformed that of the unimodal models in all test sets, and tabular data contributed to higher discrimination. The medical history and physical examination were more useful than other factors in early assessments. CONCLUSIONS: The multimodal deep learning model for combining admission notes and clinical tabular data showed promising efficacy as a potentially novel method in evaluating the risk of mortality in patients with HF, providing more accurate and timely decision support.