Derivation Dataset Research Articles

A Functional Genomic Screen to Identify Novel Genes Involved in Multiple Myeloma Tumour Development

Multiple myeloma (MM) is an incurable malignancy characterised by uncontrolled proliferation of plasma cells (PCs) in the bone marrow (BM). MM is a genetically heterogeneous disease with each patient's PCs harbouring unique genetic mutations, however the development of MM tumours is not only dependent on the underlying genetics but also on the selective pressures applied by the BM microenvironment. Hence, we hypothesise that identifying the dependencies which promote MM cell outgrowth in the BM microenvironment will allow for the identification of new drug targets. This project aims to use an in vivo functional screen, combining CRISPR-Cas9 gene editing, using a single guide RNA (sgRNA) library targeting thousands of genes, with a murine model of MM, to identify novel genes involved in MM tumour development. The Bassik human apoptosis and cancer CRISPR knockout library (Addgene #101926) was used to transduce MM cells with 31,324 unique sgRNAs targeting 3,015 genes and 1,500 control regions. The human MM cell line OPM2 was transduced with the Cas9 transgene (FuCas9GFP), followed by the sgRNA expression vector (pMCB320) generating OPM2-Cas9-sgRNA cells. These cells were then injected (5x10 5 cells/mouse, n=8 mice + 1 tumour-naive control) into the tibiae of immunodeficient NOD.Cg-Prkdc scid Il2rg tm1Wjl/SzJ (NSG) mice. Four weeks post-injection, the primary tumour within the injected tibia was isolated and sgRNA frequencies were assessed via next generation sequencing and compared with that of the initial library and the injected cells using the MAGeCK algorithm. To identify genes that play roles in MM tumour formation in patients, we assessed the expression of our identified genes in CD138-selected BM PCs from newly diagnosed MM patients (n=155) when compared with normal PCs from healthy donors (n=5) in microarray dataset E-MTAB-363 (ArrayExpress). Furthermore, to identify potential pathway involvement of our identified genes we performed gene set enrichment analysis (GSEA) using the University of California San Diego and Broad Institute developed GSEA software, comparing to Hallmark Gene Sets. Comparison of the in vitro cultured cells with the initial library identified 34 genes with undetectable (p<0.05) sgRNAs post in vitro culture, suggesting these genes play key roles in the survival of OPM2 cells. This list contained a number of genes known to control MM cell survival, such as IRF4 and CCND2. Additionally, sgRNAs targeting a further 115 genes were significantly depleted (p<0.05) in vitro and were completely undetectable in vivo, suggesting that these are implicated in cell proliferation and tumour formation in vivo. This gene list contained some already known MM drivers such as MYC, KRAS and DIS3. Furthermore, we identified 28 genes where CRISPR knockout had no significant effect in vitro but were critical dependencies for tumour development, with sgRNAs being undetectable (p<0.05) in vivo. This list contained genes known to be important for MM and BM stromal cell interaction such as ALCAM and SPP1. Notably, 25% of the genes identified as critical dependencies for OPM2 cells in vitro and/or in vivo are significantly upregulated >2-fold (p<0.05; limma) in MM patient PCs when compared with normal PCs (E-MTAB-363), suggesting these may play a role in tumour development in patients. Gene set enrichment analysis revealed that the genes that were essential in vitro were predominantly involved in cell survival related pathways such as cell cycle regulation (G2M checkpoint) and DNA replication/repair; however, the uniquely in vivo genes are predominantly involved in cell metabolism pathways such as MTORC1 signalling, oxidative phosphorylation and glycolysis. Overall, this study has successfully used a functional genetic screen to identify novel gene dependencies required for in vitro MM cell growth and in vivo MM tumour development. A secondary genetic screen will be performed in vivo to validate and refine this list of candidate genes. These validated genes will be further investigated using patient derived data sets to determine if they represent common vulnerabilities in patients and in multiple MM cell lines to investigate their specific mechanistic and pathway involvement. Identification and understanding the driving dependencies which promote MM cell outgrowth in the BM microenvironment will allow for the identification and development of new drug targets to improve patient outcomes.

Developing and Validating a Simple Risk Score for Patients with Acute Myocardial Infarction.

Mortality from acute myocardial infarction (AMI) remains substantial. The current study is aimed at developing a novel simple risk score for AMI. The Coronary Artery Tree description and Lesion EvaluaTion (CatLet) extended validation trial (ChiCTR2000033730) and the CatLet validation trial (ChiCTR-POC-17013536), both being registered with <ext-link ext-link-type="uri" xlink:href="http://chictr.org" xmlns:xlink="http://www.w3.org/1999/xlink">chictr.org</ext-link>, served as the derivation and validation datasets, respectively. Both datasets included 1,018 and 308 patients, respectively. They all suffered from AMI and underwent percutaneous intervention (PCI). The endpoint was 4-year all-cause death. Lasso regression analysis was used for covariate selection and coefficient estimation. Of 26 candidate predictor variables, the four strongest predictors for 4-year mortality were included in this novel risk score with an acronym of BACEF (serum alBumin, Age, serum Creatinine, and LVEF). This score was well calibrated and yielded an AUC (95% CI) statistics of 0.84 (0.80-0.87) in internal validation, 0.89 (0.83-0.95) in internal-external (temporal) validation, and 0.83 (0.77-0.89) in external validation. Notably, it outperformed the ACEF, ACEF II, GRACE scores with respect to 4-year mortality prediction. A simple risk score for 4-year mortality risk stratification was developed, extensively validated, and calibrated in patients with AMI. This novel BACEF score may be a useful risk stratification tool for patients with AMI.

Predicting wound complications following lower extremity revascularization

ObjectiveThe aim of this study was to create a simple risk score to identify factors associated with wound complications after infrainguinal revascularization. MethodsThe Veterans Affairs Surgical Quality Improvement Program national data set was queried from 2005 to 2021 to identify 22,114 patients undergoing elective open revascularization for peripheral arterial disease (claudication, rest pain, tissue loss) or peripheral aneurysm. Emergency and trauma cases were excluded. The data set was divided into a two-thirds derivation set and one-third validation set to create a risk prediction model. The primary end point was wound complication (wound dehiscence, superficial/deep wound surgical site infection). Eight independent risk factors for wound complications resulted from the model and were assigned whole number integer risk scores. Summary risk scores were collapsed into categories and defined as low (0-3 points), moderate (4-7 points), high (8-11 points), and very high (>12 points). ResultsThe wound complication rate in the derivation data set was 9.7% (n = 1428). Predictors of wound complication included age ≤73 (odds ratio [OR], 1.25; 95% confidence interval [CI], 1.08-1.46), body mass index ≥35 kg/m2 (OR, 1.99; 95% CI, 1.68-2.36), non-Hispanic White (vs others: OR, 1.48; 95% CI, 1.30-1.69), diabetes (OR, 1.23; 95% CI, 1.10-1.37), white blood cell count >9900/mm3 (OR, 1.18; 95% CI, 1.03-1.35), absence of coronary artery disease (OR, 1.27; 95% CI, 1.03-1.35), operative time >6 hours (OR, 1.20; 95% CI, 1.05-1.37), and undergoing a femoral endarterectomy in conjunction with bypass (OR, 1.34; 95% CI, 1.14-1.57). In both the derivation and validation sets, wound complications correlated with risk category. Among the defined categories in the derivation set, wound complication rates were 4.5% for low-risk patients, 8.5% for moderate-risk patients, 13.8% for high-risk patients, and 23.8% for very high-risk patients, with similar results for the internal validation data set. Operative indication did not independently associate with wound complications. Patients with wound complications had higher rates of reoperation and graft failure. ConclusionsThis risk prediction model uses easily obtainable clinical metrics that allow for informed discussion of wound complication risk for patients undergoing open infrainguinal revascularization.

Predicting Critical Speed of Railway Tracks Using Artificial Intelligence Algorithms

Motivated by concerns regarding safety and maintenance, the operational speed of a railway line must remain significantly below the critical speed associated with the track–ground system. Given the large number of track sections within a railway corridor that potentially need to be analyzed, the development of efficient predictive tools is of the utmost importance. Based on that, the problem can be analyzed in a few seconds instead of taking several hours of computational effort, as required by a numerical analysis. In this context, and for the first time, machine learning algorithms, namely artificial neural networks and support vector machine techniques, are applied to this particular issue. For its derivation, a reliable and robust dataset was developed by means of advanced numerical methodologies that were previously experimentally validated. The database is available as supplemental data and may be used by other researchers. Regarding the prediction process, the performance of both models was very satisfactory. From the results achieved, it is possible to conclude that the prediction tool is a novel and reliable approach for an almost instantaneous prediction of critical speed in a high number of track sections.

Optimization of A 12-lead ECG subset for Automated Early Left-Ventricular Activation Localization Approach based on Pace-Mapping Technology.

We previously developed an automated approach based on pace-mapping to localize early left-ventricular (LV) activation origin. To avoid a singular system, we require pacing from at least 2 more known sites than the number of ECG leads used. Fewer leads used means fewer pacing sites required. To identify an optimal minimal ECG-lead set for the automated approach. We used 1715 LV endocardial pacing sites to create derivation and testing datasets. The derivation dataset, consisting of 1012 known pacing sites pooled from 38 patients, was used to identify an optimal 3-lead set using random-forest regression (RFR), and a second 3-lead set using exhaustive search. The performance of these sets and the calculated Frank leads was compared within the testing dataset with 703 pacing sites pooled from 25 patients. The RFR yielded III, V1, and V4, while the exhaustive search identified leads II, V2 and V6. Comparison of these sets and the calculated Frank leads demonstrated similar performance when utilizing ≥ 5 known pacing sites. Accuracy improved with additional pacing sites, achieving mean accuracy of < 5 mm, after including up to 9 pacing sites when they were focused on a suspect area of ventricular activation origin (radius < 10 mm). The RFR identified the quasi-orthogonal leads set to localize the source of LV activation, minimizing the training set of pacing sites. Localization accuracy was high using these leads, and was not significantly different from using leads identified by exhaustive search, or empiric use of Frank leads.

A shapelet-based behavioral pattern extraction method for credit risk classification with behavior sparsity

In this paper, a shapelet-based behavioral pattern extraction method (Behavior2Shapelets) is proposed to address the behavior-sparsity issue in credit risk classification. In the proposed method, three steps are involved. First, by introducing the shapelets module, the proposed model can learn the behavioral patterns from the sparse behavioral features. Then, these extracted patterns are combined with a logistic regression model to construct an interpretable classifier. Finally, to enhance the robustness of the model, a dynamic strategy for threshold determination is used, which is based on the Kolmogorov-Smirnov statistic. The performance of the Behavior2Shapelets model is validated by using the Taiwan credit dataset and its seven derivative datasets. The empirical results demonstrate that the Behavior2Shapelets model outperforms the baseline methods in terms of classification accuracy and robustness. This highlights the feasibility of the proposed model as a solution to the behavior-sparsity problem in credit risk classification.

Development of a population-level prediction model for intensive care unit (ICU) survivorship and mortality in older adults: A population-based cohort study.

Given the growing utilization of critical care services by an aging population, development of population-level risk models which predict intensive care unit (ICU) survivorship and mortality may offer advantages for researchers and health systems. Our objective was to develop a risk model for ICU survivorship and mortality among community dwelling older adults. This was a population-based cohort study of 48,127 patients who were 50 years and older with at least one primary care visit between January 1, 2017, and December 31, 2017. We used electronic health record (EHR) data to identify variables predictive of ICU survivorship. ICU admission and mortality within 2 years after index primary care visit date were used to divide patients into three groups of "alive without ICU admission", "ICU survivors,"and "death." Multinomial logistic regression was used to identify EHR predictive variables for the three patient outcomes. Cross-validation by randomly splitting the data into derivation and validation data sets (60:40 split) was used to identify predictor variables and validate model performance using area under the receiver operating characteristics (AUC) curve. In our overall sample, 92.2% of patients were alive without ICU admission, 6.2% were admitted to the ICU at least once and survived, and 1.6% died. Greater deciles of age over 50 years, diagnoses of chronic obstructive pulmonary disorder or chronic heart failure, and laboratory abnormalities in alkaline phosphatase, hematocrit, and albumin contributed highest risk score weights for mortality. Risk scores derived from the model discriminated between patients that died versus remained alive without ICU admission (AUC = 0.858), and between ICU survivors versus alive without ICU admission (AUC = 0.765). Our risk scores provide a feasible and scalable tool for researchers and health systems to identify patient cohorts at increased risk for ICU admission and survivorship. Further studies are needed to prospectively validate the risk scores in other patient populations.

Derivation and validation of a new prediction model for sudden unexpected death in epilepsy based on a longitudinal prospective population-based cohort

ObjectiveWe conducted a population-based, prospective cohort study with a large sample size in Ningxia Province of the Northwest, a rural area in China, by developing a model to specifically assess risk factors of sudden unexpected death in epilepsy (SUDEP) in people with convulsive epilepsy by clinical variables. MethodsParticipants with convulsive epilepsy were recruited from January 1, 2008, to April 28, 2022, in rural Northwest China. They received regular assessments and management of epilepsy at the primary healthcare level and were followed up monthly. Information on the cause of death and relevant clinical details was obtained from death certificates or neurologist-conducted verbal autopsies. Survival analysis was employed to identify potential risk factors associated with SUDEP. ResultsFive variables were independently associated with SUDEP: generalized tonic-clonic seizures (GTCS) with ≥1 attack during the preceding month, GTCS with >3 attacks during the preceding year, body mass index (BMI) ≥24, age of onset ≤14 years, and duration >20 years. The area under receiver operator characteristic (ROC) curve (AUC) value (95% CI) of the model was 0.789 (0.735–0.843) in the derivation dataset and 0.830 (0.758–0.902) in the validation dataset. There was agreement between the observed and predicted probabilities of SUDEP. ConclusionsThis study establishes that high GTCS frequency, early age of onset, long duration of epilepsy, and being overweight are associated with an increased risk of SUDEP in individuals with convulsive epilepsy. The study also developed and validated a personalized prediction model to accurately assess the risk of SUDEP.

A nomogram to predict 28-day mortality in neonates with sepsis: a retrospective study based on the MIMIC-III database.

Sepsis is the second-leading cause of death in neonates. We established a predictive nomogram to identify critically ill neonates early and reduce the time to treatment. It is a retrospective case-control study based on the MIMIC-III database. The study population comprised 924 neonates diagnosed with sepsis. Neonates with sepsis included in the MIMIC-III database were enrolled, including 880 surviving neonates and 44 neonates who died. In the derivation dataset, stepwise regression and the Lasso algorithm were employed to select predictive variables, and the neonatal sequential organ failure assessment score (nSOFA) was calculated simultaneously. Bootstrap resampling was utilized to perform internal validation. The results indicated that the Lasso algorithm displayed superior discrimination, sensitivity, and specificity relative to stepwise regression and nSOFA scores. After 500 bootstrap resampling tests, the area under the receiver operating characteristic curve (AUC) of the Lasso algorithm was 0.912 (95% CI: 0.870-0.977). The nomogram based on the Lasso algorithm outperformed stepwise regression and nSOFA scores in terms of calibration and the clinical net benefit. This nomogram can assist in prognosticating neonatal severe sepsis and aid in guiding clinical practice while concurrently improving patient outcomes. The established nomogram revealed that jaundice, corticosteroid use, weight, serum calcium, inotropes and base excess are all important predictors of 28-day mortality in neonates with sepsis. This nomogram can facilitate the early identification of neonates with severe sepsis. However, it still requires further modification and external validation to make it widely available.

Machine Learning Model to Predict Graft Rejection After Kidney Transplantation

There are few predictive studies about early posttransplant outcomes taking into account baseline and posttransplant variables. The objective of this study was to create a predictive model for 30-day graft rejection using machine learning techniques. Retrospective study with 1255 patients undergoing transplant from living and deceased donors at a tertiary health service in Brazil. Recipient, donor, transplantation, and postoperative period data were collected from physical and electronic records. We split the data into derivation (training) and validation (test) datasets. Five supervised machine learning algorithms were developed with this subset of variables in the training set: Simple Logistic Regression, Lasso, Multilayer Perceptron, XGBoost, and Light GBM. There were 147 (12.48%) cases of graft rejection within 30 days of transplantation. The best model was XGBoost (accuracy, 0.839; receiver operating characteristic area under the curve, 0.715; precision, 0.900). The model showed that deceased donor transplantation, glomerulopathy as an underlying disease, and donor's use of vasoactive drugs had more than 20% importance as rejection risk factors. The variables with the greatest predictive values were thymoglobulin induction and delayed graft function. We fitted a machine learning model to predict 30-day graft rejection after kidney transplantation that reaches a higher accuracy and precision. Machine learning models could contribute to predicting kidney survival using nontraditional approaches.

Validation of the Conventional Trauma and Injury Severity Score and a Newly Developed Survival Predictive Model in Pediatric Patients with Blunt Trauma: A Nationwide Observation Study.

To date, there is no clinically useful prediction model that is suitable for Japanese pediatric trauma patients. Herein, this study aimed to developed a model for predicting the survival of Japanese pediatric patients with blunt trauma and compare its validity with that of the conventional TRISS model. Patients registered in the Japan Trauma Data Bank were grouped into a derivation cohort (2009-2013) and validation cohort (2014-2018). Logistic regression analysis was performed using the derivation dataset to establish prediction models using age, injury severity, and physiology. The validity of the modified model was evaluated by the area under the receiver operating characteristic curve (AUC). Among 11 predictor models, Model 1 and Model 11 had the best performance (AUC = 0.980). The AUC of all models was lower in patients with survival probability Ps < 0.5 than in patients with Ps ≥ 0.5. The AUC of all models was lower in neonates/infants than in other age categories. Model 11 also had the best performance (AUC = 0.762 and 0.909, respectively) in patients with Ps < 0.5 and neonates/infants. The predictive ability of the newly modified models was not superior to that of the current TRISS model. Our results may be useful to develop a highly accurate prediction model based on the new predictive variables and cutoff values associated with the survival mortality of injured Japanese pediatric patients who are younger and more severely injured by using a nationwide dataset with fewer missing data and added valuables, which can be used to evaluate the age-related physiological and anatomical severity of injured patients.

Deriving and validating an asthma diagnosis prediction model for children and young people in primary care.

Introduction: Accurately diagnosing asthma can be challenging. We aimed to derive and validate a prediction model to support primary care clinicians assess the probability of an asthma diagnosis in children and young people. Methods: The derivation dataset was created from the Avon Longitudinal Study of Parents and Children (ALSPAC) linked to electronic health records. Participants with at least three inhaled corticosteroid prescriptions in 12-months and a coded asthma diagnosis were designated as having asthma. Demographics, symptoms, past medical/family history, exposures, investigations, and prescriptions were considered as candidate predictors. Potential candidate predictors were included if data were available in ≥60% of participants. Multiple imputation was used to handle remaining missing data. The prediction model was derived using logistic regression. Internal validation was completed using bootstrap re-sampling. External validation was conducted using health records from the Optimum Patient Care Research Database (OPCRD). Results: Predictors included in the final model were wheeze, cough, breathlessness, hay-fever, eczema, food allergy, social class, maternal asthma, childhood exposure to cigarette smoke, prescription of a short acting beta agonist and the past recording of lung function/reversibility testing. In the derivation dataset, which comprised 11,972 participants aged <25 years (49% female, 8% asthma), model performance as indicated by the C-statistic and calibration slope was 0.86, 95% confidence interval (CI) 0.85-0.87 and 1.00, 95% CI 0.95-1.05 respectively. In the external validation dataset, which included 2,670 participants aged <25 years (50% female, 10% asthma), the C-statistic was 0.85, 95% CI 0.83-0.88, and calibration slope 1.22, 95% CI 1.09-1.35. Conclusions: We derived and validated a prediction model for clinicians to calculate the probability of asthma diagnosis for a child or young person up to 25 years of age presenting to primary care. Following further evaluation of clinical effectiveness, the prediction model could be implemented as a decision support software.

Optimal use of red cell volume distribution width-to-platelet ratio to exclude cirrhosis in patients with chronic hepatitis B

Background and aimsHepatitis B virus (HBV) infection is a major public health issue worldwide as it may cause serious liver diseases such as cirrhosis and hepatocellular carcinoma (HCC). Ruling out cirrhosis is important when treating chronic hepatitis B (CHB). The aim of this study was to compare the performance of the aspartate aminotransferase-to-platelet ratio index (APRI), fibrosis score based on four factors (FIB-4), and red cell volume distribution width-to-platelet ratio (RPR) in diagnosing liver fibrosis stages and to identify new cut-off values to rule out cirrhosis. MethodsBetween 2005 and 2020, 2182 eligible individuals who underwent liver biopsy were randomly assigned to derivation and validation cohorts in a 6:4 ratio. A grid search was applied to identify optimal cut-off values with a sensitivity of >90% and a negative predictive value (NPV) of at least 95%. ResultsOverall, 1309 individuals (175 patients with cirrhosis) were included in the derivation dataset, and 873 (117 patients with cirrhosis) were included in the validation cohort. The area under the receiver operating characteristic curve of RPR for diagnosing cirrhosis was 0.821, which was comparable to that of APRI (0.818, P = 0.7905) and FIB-4 (0.803, P = 0.2395). When applying an RPR of 0.06, cirrhosis was correctly identified with a sensitivity of 93.1% and an NPV of 97.1%, while it misclassified 12 of 175 (6.9%) patients in the derivation cohort. In the validation cohort, RPR had a sensitivity and NPV of 97.4% and 99.0%, respectively, and only misclassified 3 of 117 (2.6%) patients. Subgroup analysis indicated that the new RPR cut-off value performed more consistently than that of APRI and FIB-4 in all subgroups. ConclusionA recently established cut-off value for RPR (≤0.06) was validated and was more effective than APRI and FIB-4 in excluding patients with cirrhosis due to a higher sensitivity and NPV and a lower misclassification rate. This simple and dependable test could have significant clinical implications in identifying patients who require monitoring for portal hypertension-associated complications and screening for HCC, particularly in middle and primary healthcare settings.

Paying attention to cardiac surgical risk: An interpretable machine learning approach using an uncertainty-aware attentive neural network.

Machine learning (ML) is increasingly applied to predict adverse postoperative outcomes in cardiac surgery. Commonly used ML models fail to translate to clinical practice due to absent model explainability, limited uncertainty quantification, and no flexibility to missing data. We aimed to develop and benchmark a novel ML approach, the uncertainty-aware attention network (UAN), to overcome these common limitations. Two Bayesian uncertainty quantification methods were tested, generalized variational inference (GVI) or a posterior network (PN). The UAN models were compared with an ensemble of XGBoost models and a Bayesian logistic regression model (LR) with imputation. The derivation datasets consisted of 153,932 surgery events from the Australian and New Zealand Society of Cardiac and Thoracic Surgeons (ANZSCTS) Cardiac Surgery Database. An external validation consisted of 7343 surgery events which were extracted from the Medical Information Mart for Intensive Care (MIMIC) III critical care dataset. The highest performing model on the external validation dataset was a UAN-GVI with an area under the receiver operating characteristic curve (AUC) of 0.78 (0.01). Model performance improved on high confidence samples with an AUC of 0.81 (0.01). Confidence calibration for aleatoric uncertainty was excellent for all models. Calibration for epistemic uncertainty was more variable, with an ensemble of XGBoost models performing the best with an AUC of 0.84 (0.08). Epistemic uncertainty was improved using the PN approach, compared to GVI. UAN is able to use an interpretable and flexible deep learning approach to provide estimates of model uncertainty alongside state-of-the-art predictions. The model has been made freely available as an easy-to-use web application demonstrating that by designing uncertainty-aware models with innately explainable predictions deep learning may become more suitable for routine clinical use.

Deep learning-based model detects atrial septal defects from electrocardiography: a cross-sectional multicenter hospital-based study

Atrial septal defect (ASD) increases the risk of adverse cardiovascular outcomes. Despite the potential for risk mitigation through minimally invasive percutaneous closure, ASD remains underdiagnosed due to subtle symptoms and examination findings. To bridge this diagnostic gap, we propose a novel screening strategy aimed at early detection and enhanced diagnosis through the implementation of a convolutional neural network (CNN) to identify ASD from 12-lead electrocardiography (ECG). ECGs were collected from patients with at least one recorded echocardiogram at 3 hospitals from 2 continents (Keio University Hospital from July 2011 to December 2020, Brigham and Women's Hospital from January 2015 to December 2020, and Dokkyo Medical University Saitama Medical Center from January 2010 and December 2021). ECGs from patients with a diagnosis of ASD were labeled as positive cases while the remainder were labeled as negative. ECGs after the closure of ASD were excluded. After randomly splitting the ECGs into 3 datasets (50% derivation, 20% validation, and 30% test) with no patient overlap, a CNN-based model was trained using the derivation datasets from 2 hospitals and was tested on held-out datasets along with an external validation on the 3rd hospital. All eligible ECGs were used for derivation and validation whereas the earliest ECG for each patient was used for the test and external validation. The discrimination of ASD was assessed by the area under the receiver operating characteristic curve (AUROC). Multiple subgroups were examined to identify any heterogeneity. A total of 671,201 ECGs from 80,947 patients were collected from the 3 institutions. The AUROC for detecting ASD was 0.85-0.90 across the 3 hospitals. The subgroup analysis showed excellent performance across various characteristics Screening simulation using the model greatly increased sensitivity from 80.6% to 93.7% at specificity 33.6% when compared to using overt ECG abnormalities. A CNN-based model using 12-lead ECG successfully identified the presence of ASD with excellent generalizability across institutions from 2 separate continents. This work was supported by research grants from JST (JPMJPF2101), JSR corporation, Taiju Life Social Welfare Foundation, Kondou Kinen Medical Foundation, Research fund of Mitsukoshi health and welfare foundation, Tokai University School of Medicine Project Research and Internal Medicine Project Research, Secom Science and Technology Foundation, and Grants from AMED (JP23hma922012 and JP23ym0126813). This work was partially supported by One Brave Idea, co-funded by the American Heart Association and Verily with significant support from AstraZeneca and pillar support from Quest Diagnostics.

Optimal Estimation Inversion of Ionospheric Electron Density from GNSS-POD Limb Measurements: Part II-Validation and Comparison Using NmF2 and hmF2

A growing number of SmallSat/CubeSat constellations with high-rate (50–100 Hz) global navigation satellite system radio occultations (GNSS-RO) as well as low-rate (1 Hz) precise orbit determination (GNSS-POD) limb-viewing capabilities provide unprecedented spatial and temporal sampling rates for ionospheric studies. In the F-region electron density (Ne) retrieval process, instead of the conventional onion-peeling (OP) inversion, an optimal estimation (OE) inversion technique was recently developed using total electron content measurements acquired by GNSS-POD link. The new technique is applied to data acquired from the COSMIC-1, COSMIC-2, and Spire constellations. Although both OE and OP techniques use the Abel weighting function in Ne inversion, OE significantly differs in its performance, especially in the lower F- and E-regions. In this work, we evaluate and compare newly derived data sets using F2 peak properties with other space-based and ground-based observations. We determine the F2 peak Ne (NmF2) and its altitude (hmF2), and compare them with the OP-retrieved values. Good agreement is observed between the two techniques for both NmF2 and hmF2. In addition, we also utilize autoscaled F2 peak measurements from a number of worldwide Digisonde stations (∼30). The diurnal sensitivity and latitudinal variability of the F2 peak between the two techniques are carefully studied at these locations. Good agreement is observed between OE-retrieved NmF2 and Digisonde-measured NmF2. However, significant differences appear between OE-retrieved hmF2 and Digisonde-measured hmF2. During the daytime, Digisonde-measured hmF2 remains ∼25–45 km below the OE-retrieved hmF2, especially at mid and high latitudes. We also incorporate F-region Ne measurements from two incoherent scatter radar observations at high latitudes, located in the North American (Millstone Hill) and European (EISCAT at Tromso) sectors. The radar measurements show good agreement with OE-retrieved values. Although there are several possible sources of error in the ionogram-derived Ne profiles, our further analysis on F1 and F2 layers indicates that the low Digisonde hmF2 is caused by the autoscaled method, which tends to detect a height systematically below the F2 peak when the F1 layer is present.

Derivation and validation of a clinical prediction model for risk-stratification of children hospitalized with severe pneumonia in Bangladesh.

Children with severe pneumonia in low- and middle-income countries (LMICs) suffer from high rates of treatment failure despite appropriate World Health Organization (WHO)-directed antibiotic treatment. Developing a clinical prediction rule for treatment failure may allow early identification of high-risk patients and timely intervention to decrease mortality. We used data from two separate studies conducted at the Dhaka Hospital of the International Centre for Diarrheal Disease Research, Bangladesh (icddr,b) to derive and externally validate a clinical prediction rule for treatment failure of children hospitalized with severe pneumonia. The derivation dataset was from a randomized clinical trial conducted from 2018 to 2019, studying children aged 2 to 59 months hospitalized with severe pneumonia as defined by WHO. Treatment failure was defined by the persistence of danger signs at the end of 48 hours of antibiotic treatment or the appearance of any new danger signs within 24 hours of enrollment. We built a random forest model to identify the top predictors. The top six predictors were the presence of grunting, room air saturation, temperature, the presence of lower chest wall indrawing, the presence of respiratory distress, and central cyanosis. Using these six predictors, we created a parsimonious model with a discriminatory performance of 0.691, as measured by area under the receiving operating curve (AUC). We performed external validation using a temporally distinct dataset from a cohort study of 191 similarly aged children with severe acute malnutrition and pneumonia. In external validation, discriminatory performance was maintained with an improved AUC of 0.718. In conclusion, we developed and externally validated a parsimonious six-predictor model using random forest methods to predict treatment failure in young children with severe pneumonia in Bangladesh. These findings can be used to further develop and validate parsimonious and pragmatic prognostic clinical prediction rules for pediatric pneumonia, particularly in LMICs.

The estimation of coronary artery calcium thickness by computed tomography angiography based on optical coherence tomography measurements.

Optical coherence tomography (OCT) is recommended to be the most appropriate modality in assessing calcium thickness, however, it has limitations associated with infrared attenuation. Although coronary computed tomography angiography (CCTA) detects calcification, it has low resolution and hence not recommended to measure the calcium size. The aim of this study was to devise a simple algorithm to estimate calcium thickness based on the CCTA image. A total of 68 patients who had CCTA for suspected coronary artery disease and subsequently went on to have OCT were included in the study. 238 lesions of them divided into derivation and validation dataset at 2:1 ratio (47 patients with 159 lesions and 21 with 79, respectively) were analyzed. A new method was developed to estimate calcium thickness from the maximum CT density within the calcification and compared with calcium thickness measured by OCT. Maximum Calcium density and measured calcium-border CT density had a good correlation with a linear equation of y = 0.58x + 201 (r = 0.892, 95% CI 0.855-0.919, p < 0.001). The estimated calcium thickness derived from this equation showed strong agreement with measured calcium thickness in validation and derivation dataset (r2 = 0.481 and 0.527, 95% CI 0.609-0.842 and 0.497-0.782, p < 0.001 in both, respectively), more accurate than the estimation by full width at half maximum and inflection point method. In conclusion, this novel method provided the estimation of calcium thickness more accurately than conventional methods.

A Single Dermatome Clinical Prediction Rule for Independent Walking 1 Year After Spinal Cord Injury

ObjectiveTo derive and validate a simple, accurate CPR to predict future independent walking ability after SCI at the bedside that does not rely on motor scores and is predictive for those initially classified in the middle of the SCI severity spectrum. DesignRetrospective cohort study. Binary variables were derived, indicating degrees of sensation to evaluate predictive value of pinprick and light touch variables across dermatomes. The optimal single sensory modality and dermatome was used to derive our CPR, which was validated on an independent dataset. SettingAnalysis of SCI Model Systems dataset. ParticipantsIndividuals with traumatic SCI. The data of 3679 participants (N=3679) were included with 623 participants comprising the derivation dataset and 3056 comprising the validation dataset. InterventionsNot applicable. Main Outcome MeasuresSelf-reported ability to walk both indoors and outdoors. ResultsPinprick testing at S1 over lateral heels, within 31 days of SCI, accurately identified future independent walkers 1 year after SCI. Normal pinprick in both lateral heels provided good prognosis, any pinprick sensation in either lateral heel provided fair prognosis, and no sensation provided poor prognosis. This CPR performed satisfactorily in the middle SCI severity subgroup. ConclusionsIn this large multi-site study, we derived and validated a simple, accurate CPR using only pinprick sensory testing at lateral heels that predicts future independent walking after SCI.

Derivation and external validation of a clinical prognostic model identifying children at risk of death following presentation for diarrheal care.

Diarrhea continues to be a leading cause of death for children under-five. Amongst children treated for acute diarrhea, mortality risk remains elevated during and after acute medical management. Identification of those at highest risk would enable better targeting of interventions, but available prognostic tools lack validation. We used clinical and demographic data from the Global Enteric Multicenter Study (GEMS) to build clinical prognostic models (CPMs) to predict death (in-treatment, after discharge, or either) in children aged ≤59 months presenting with moderate-to-severe diarrhea (MSD), in Africa and Asia. We screened variables using random forests, and assessed predictive performance with random forest regression and logistic regression using repeated cross-validation. We used data from the Kilifi Health and Demographic Surveillance System (KHDSS) and Kilifi County Hospital (KCH) in Kenya to externally validate our GEMS-derived CPM. Of 8060 MSD cases, 43 (0.5%) children died in treatment and 122 (1.5% of remaining) died after discharge. MUAC at presentation, respiratory rate, age, temperature, number of days with diarrhea at presentation, number of people living in household, number of children <60 months old living in household, and how much the child had been offered to drink since diarrhea started were predictive of death both in treatment and after discharge. Using a parsimonious 2-variable prediction model, we achieved an area under the ROC curve (AUC) of 0.84 (95% CI: 0.82, 0.86) in the derivation dataset, and an AUC = 0.74 (95% CI 0.71, 0.77) in the external dataset. Our findings suggest it is possible to identify children most likely to die after presenting to care for acute diarrhea. This could represent a novel and cost-effective way to target resources for the prevention of childhood mortality.