Prediction of in-hospital mortality among intensive care unit patients using modified daily Laboratory-based Acute Physiology Scores, version 2 (LAPS2)

Background

Mortality prediction for intensive care unit (ICU) patients frequently relies on single acuity measures based on ICU admission physiology without accounting for subsequent clinical changes.

Objectives

Evaluate novel models incorporating modified admission and daily, time-updating Laboratory-based Acute Physiology Scores, version 2 (LAPS2) to predict in-hospital mortality among ICU patients.

Research design

Retrospective cohort study.

Subjects

All ICU patients in five hospitals from October 2017 through September 2019.

Measures

We used logistic regression, penalized logistic regression, and random forest models to predict in-hospital mortality within 30 days of ICU admission using admission LAPS2 alone in patient-level and patient-day-level models, or admission and daily LAPS2 at the patient-day level. Multivariable models included patient and admission characteristics. We performed internal-external validation using four hospitals for training and the fifth for validation, repeating analyses for each hospital as the validation set. We assessed performance using scaled Brier scores (SBS), c-statistics, and calibration plots.

Results

The cohort included 13,993 patients and 120,101 ICU days. The patient-level model including the modified admission LAPS2 without daily LAPS2 had an SBS of 0.175 (95% CI 0.148-0.201) and c-statistic of 0.824 (95% CI 0.808-0.840). Patient-day-level models including daily LAPS2 consistently outperformed models with modified admission LAPS2 alone. Among patients with <50% predicted mortality, daily models were better calibrated than models with modified admission LAPS2 alone.

Conclusions

Models incorporating daily, time-updating LAPS2 to predict mortality among an ICU population perform as well or better than models incorporating modified admission LAPS2 alone.

Developing and Validating a Lung Cancer Risk Prediction Model: A Nationwide Population-Based Study

Lung cancer can be challenging to diagnose in the early stages, where treatment options are optimal. We aimed to develop 1-year prediction models for the individual risk of incident lung cancer for all individuals aged 40 or above living in Denmark on 1 January 2017. The study was conducted using population-based registers on health and sociodemographics from 2007-2016. We applied backward selection on all variables by logistic regression to develop a risk model for lung cancer and applied the models to the validation cohort, calculated receiver-operating characteristic curves, and estimated the corresponding areas under the curve (AUC). In the populations without and with previously confirmed cancer, 4274/2,826,249 (0.15%) and 482/172,513 (0.3%) individuals received a lung cancer diagnosis in 2017, respectively. For both populations, older age was a relevant predictor, and the most complex models, containing variables related to diagnoses, medication, general practitioner, and specialist contacts, as well as baseline sociodemographic characteristics, had the highest AUC. These models achieved a positive predictive value (PPV) of 0.0127 (0.006) and a negative predictive value (NPV) of 0.989 (0.997) with a 1% cut-off in the population without (with) previous cancer. This corresponds to 1.2% of the screened population experiencing a positive prediction, of which 1.3% would be incident with lung cancer. We have developed and tested a prediction model with a reasonable potential to support clinicians and healthcare planners in identifying patients at risk of lung cancer.

Impact of Mobile Monitoring Network Design on Air Pollution Exposure Assessment Models

Short-term mobile monitoring campaigns are increasingly used to assess long-term air pollution exposure in epidemiology. Little is known about how monitoring network design features, including the number of stops and sampling temporality, impacts exposure assessment models. We address this gap by leveraging an extensive mobile monitoring campaign conducted in the greater Seattle area over the course of a year during all days of the week and most hours. The campaign measured total particle number concentration (PNC; sheds light on ultrafine particulate (UFP) number concentration), black carbon (BC), nitrogen dioxide (NO2), fine particulate matter (PM2.5), and carbon dioxide (CO2). In Monte Carlo sampling of 7327 total stops (278 sites × 26 visits each), we restricted the number of sites and visits used to estimate annual averages. Predictions from the all-data campaign performed well, with cross-validated R2s of 0.51-0.77. We found similar model performances (85% of the all-data campaign R2) with ∼1000 to 3000 randomly selected stops for NO2, PNC, and BC, and ∼4000 to 5000 stops for PM2.5 and CO2. Campaigns with additional temporal restrictions (e.g., business hours, rush hours, weekdays, or fewer seasons) had reduced model performances and different spatial surfaces. Mobile monitoring campaigns wanting to assess long-term exposure should carefully consider their monitoring designs.

Prediction Models of Primary Membranous Nephropathy: A Systematic Review and Meta-Analysis

BACKGROUND

Several statistical models for predicting prognosis of primary membranous nephropathy (PMN) have been proposed, most of which have not been as widely accepted in clinical practice.

METHODS

A systematic search was performed in MEDLINE and EMBASE. English studies that developed any prediction models including two or more than two predictive variables were eligible for inclusion. The study population was limited to adult patients with pathologically confirmed PMN. The outcomes in eligible studies should be events relevant to prognosis of PMN, either disease progression or response profile after treatments. The risk of bias was assessed according to the PROBAST.

RESULTS

In all, eight studies with 1237 patients were included. The pooled AUC value of the seven studies with renal function deterioration and/or ESRD as the predicted outcomes was 0.88 (95% CI: 0.85 to 0.90; I² = 77%, p = 0.006). The paired forest plots for sensitivity and specificity with corresponding 95% CIs for each of these seven studies indicated the combined sensitivity and specificity were 0.76 (95% CI: 0.64 to 0.85) and 0.84 (95% CI: 0.80 to 0.88), respectively. All seven studies included in the meta-analysis were assessed as high risk of bias according to the PROBAST tool.

CONCLUSIONS

The reported discrimination ability of included models was good; however, the insufficient calibration assessment and lack of validation studies precluded drawing a definitive conclusion on the performance of these prediction models. High-grade evidence from well-designed studies is needed in this field.

Visual Monitoring of Fatty Acid Degradation during Green Tea Storage by Hyperspectral Imaging

The reduction in freshness during green tea storage leads to a reduction in its commercial value and consumer acceptance, which is thought to be related to the oxidation of fatty acids. Here, we developed a novel and rapid method for the assessment of green tea freshness during storage. Hyperspectral images of green tea during storage were acquired, and fatty acid profiles were detected by GC-MS. Partial least squares (PLS) analysis was used to model the association of spectral data with fatty acid content. In addition, competitive adaptive reweighted sampling (CARS) was employed to select the characteristic wavelengths and thus simplify the model. The results show that the constructed CARS-PLS can achieve accurate prediction of saturated and unsaturated fatty acid content, with residual prediction deviation (RPD) values over 2. Ultimately, chemical imaging was used to visualize the distribution of fatty acids during storage, thus providing a fast and nondestructive method for green tea freshness evaluation.

Modeling net energy partition patterns of growing-finishing pigs using nonlinear regression and artificial neural networks

The objectives of this study were to evaluate the net energy (NE) partition patterns of growing-finishing pigs at different growing stages and to develop the corresponding prediction models using nonlinear regression (NLR) and artificial neural networks (ANN). Twenty-four pigs with an initial body weight (BW) of ~30 kg were kept in metabolic cages and fed ad libitum and were moved into six respiration chambers in turns until ~90 kg. The NE partition patterns, i.e., NE for maintenance (NEm), NE retained as protein (NEp), and NE retained as lipid (NEl), were calculated based on indirect calorimetry and nitrogen balance techniques. The energy balance data collected through the animal trial was then randomly split into a training data set containing 75% of the samples and a testing data set containing the remaining 25% of the samples. The NLR models and a series of ANN models were established on the training data set to predict the metabolizable energy intake, NE intake, NEm, NEp, and NEl of pigs. The best-fitted ANN models were selected by 5-fold cross-validation in the training data set. The prediction performance of the best-fitted NLR and ANN models were compared on the testing data set. The results showed that the average NE intakes of pigs were 17.71, 23.25, 24.56, and 28.96 MJ/d in 30 to 45 kg, 45 to 60 kg, 60 to 75 kg, and 75 to 90 kg, respectively. The NEm and NEl (MJ/d) kept increasing as BW increased from 30 kg to 90 kg, while the NEp increased to its maximum value and then kept in a certain range of 4.64 to 4.88 MJ/d. The proportion of NEm for pigs at 30 to 90 kg stayed within the range of 42.0% to 48.6%, while the proportion of NEl kept increasing. For the prediction models built based on the animal trial, ANN models exhibited better performance than NLR models for all the target outputs. In conclusion, NE partition patterns changed in different growth stages of pigs, and ANN models are more flexible and powerful than NLR models in predicting the NE partition patterns of growing-finishing pigs.

The Trend of Risk for Cardiovascular Diseases During the Past Decade in Iran, Applying No-Lab and Lab-Based Prediction Models

Background

As a surrogate for all relevant risk factors, it is preferable to show trends in the mean cardiovascular disease(CVD) risk rather than to examine each risk factor trend separately.

Objectives

Using national representative data, this study aimed to determine the changes in the World Health Organization (WHO) CVD risk during the last decade considering both laboratory and non-laboratory risk scoring.

Methods

We used data from five rounds of the WHO STEPwise approach to surveillance surveys (2007-2016). In all, 62,076 (31,660 women) participants aged 40-65 years were included and their absolute CVD risk were calculated. The generalized linear model was performed to assess the trend of CVD risk in men and women, and also in diabetic and non-diabetic individuals.

Results

We showed significant declining trends in the mean CVD risk in the laboratory (from 10.5% to 8.8%) and non-laboratory (10.1% to 9.4%) models in men. In women, a significant reduction was observed in the laboratory-based model (from 8.4% to 7.8%). The laboratory model showed a greater decrease in men than women (P-for interaction < 0.001) and in diabetic patients (from 16.1% to 13.6%) than non-diabetic individuals (from 8.2% to 7%) (p-for interaction = 0.002). The proportion of high-risk individuals (risk ≥ 10%) decreased from 40% in 2007 to 31.5% in 2016 in men and from 29.8% to 26.1% in women based on the laboratory-model.

Conclusions

During the last decade, CVD risk had a significant decrease in men and women. The reduction was more evident in men and diabetic population. However, still, one-third of our population is considered high-risk.

Preoperative prediction of sonic hedgehog and group 4 molecular subtypes of pediatric medulloblastoma based on radiomics of multiparametric MRI combined with clinical parameters

Purpose

To construct a machine learning model based on radiomics of multiparametric magnetic resonance imaging (MRI) combined with clinical parameters for predicting Sonic Hedgehog (SHH) and Group 4 (G4) molecular subtypes of pediatric medulloblastoma (MB).

Methods

The preoperative MRI images and clinical data of 95 patients with MB were retrospectively analyzed, including 47 cases of SHH subtype and 48 cases of G4 subtype. Radiomic features were extracted from T1-weighted imaging (T1), contrast-enhanced T1 weighted imaging (T1c), T2-weighted imaging (T2), T2 fluid-attenuated inversion recovery imaging (T2FLAIR), and apparent diffusion coefficient (ADC) maps, using variance thresholding, SelectKBest, and Least Absolute Shrinkage and Selection Operator (LASSO) regression algorithms. The optimal features were filtered using LASSO regression, and a logistic regression (LR) algorithm was used to build a machine learning model. The receiver operator characteristic (ROC) curve was plotted to evaluate the prediction accuracy, and verified by its calibration, decision and nomogram. The Delong test was used to compare the differences between different models.

Results

A total of 17 optimal features, with non-redundancy and high correlation, were selected from 7,045 radiomics features, and used to build an LR model. The model showed a classification accuracy with an under the curve (AUC) of 0.960 (95% CI: 0.871-1.000) in the training cohort and 0.751 (95% CI: 0.587-0.915) in the testing cohort, respectively. The location of the tumor, pathological type, and hydrocephalus status of the two subtypes of patients differed significantly (p < 0.05). When combining radiomics features and clinical parameters to construct the combined prediction model, the AUC improved to 0.965 (95% CI: 0.898-1.000) in the training cohort and 0.849 (95% CI: 0.695-1.000) in the testing cohort, respectively. There was a significant difference in the prediction accuracy, as measured by AUC, between the testing cohorts of the two prediction models, which was confirmed by Delong's test (p = 0.0144). Decision curves and nomogram further validate that the combined model can achieve net benefits in clinical work.

Conclusion

The combined prediction model, constructed based on radiomics of multiparametric MRI and clinical parameters can potentially provide a non-invasive clinical approach to predict SHH and G4 molecular subtypes of MB preoperatively.

Improving in-season wheat yield prediction using remote sensing and additional agronomic traits as predictors

The development of accurate grain yield (GY) multivariate models using normalized difference vegetation index (NDVI) assessments obtained from aerial vehicles and additional agronomic traits is a promising option to assist, or even substitute, laborious agronomic in-field evaluations for wheat variety trials. This study proposed improved GY prediction models for wheat experimental trials. Calibration models were developed using all possible combinations of aerial NDVI, plant height, phenology, and ear density from experimental trials of three crop seasons. First, models were developed using 20, 50 and 100 plots in training sets and GY predictions were only moderately improved by increasing the size of the training set. Then, the best models predicting GY were defined in terms of the lowest Bayesian information criterion (BIC) and the inclusion of days to heading, ear density or plant height together with NDVI in most cases were better (lower BIC) than NDVI alone. This was particularly evident when NDVI saturates (with yields above 8 t ha^-1) with models including NDVI and days to heading providing a 50% increase in the prediction accuracy and a 10% decrease in the root mean square error. These results showed an improvement of NDVI prediction models by the addition of other agronomic traits. Moreover, NDVI and additional agronomic traits were unreliable predictors of grain yield in wheat landraces and conventional yield quantification methods must be used in this case. Saturation and underestimation of productivity may be explained by differences in other yield components that NDVI alone cannot detect (e.g. differences in grain size and number).

Prediction model for cognitive frailty in older adults: A systematic review and critical appraisal

Background

Several prediction models for cognitive frailty (CF) in older adults have been developed. However, the existing models have varied in predictors and performances, and the methodological quality still needs to be determined.

Objectives

We aimed to summarize and critically appraise the reported multivariable prediction models in older adults with CF.

Methods

PubMed, Embase, Cochrane Library, Web of Science, Scopus, PsycINFO, CINAHL, China National Knowledge Infrastructure, and Wanfang Databases were searched from the inception to March 1, 2022. Included models were descriptively summarized and critically appraised by the Prediction Model Risk of Bias Assessment Tool (PROBAST).

Results

A total of 1,535 articles were screened, of which seven were included in the review, describing the development of eight models. Most models were developed in China (n = 4, 50.0%). The most common predictors were age (n = 8, 100%) and depression (n = 4, 50.0%). Seven models reported discrimination by the C-index or area under the receiver operating curve (AUC) ranging from 0.71 to 0.97, and four models reported the calibration using the Hosmer-Lemeshow test and calibration plot. All models were rated as high risk of bias. Two models were validated externally.

Conclusion

There are a few prediction models for CF. As a result of methodological shortcomings, incomplete presentation, and lack of external validation, the models' usefulness still needs to be determined. In the future, models with better prediction performance and methodological quality should be developed and validated externally.

Systematic review registration

www.crd.york.ac.uk/prospero, identifier CRD42022323591.

Prediction of Ammonia Concentration in a Pig House Based on Machine Learning Models and Environmental Parameters

Accurately predicting the air quality in a piggery and taking control measures in advance are important issues for pig farm production and local environmental management. In this experiment, the NH₃ concentration in a semi-automatic piggery was studied. First, the random forest algorithm (RF) and Pearson correlation analysis were combined to analyze the environmental parameters, and nine input schemes for the model feature parameters were identified. Three kinds of deep learning and three kinds of conventional machine learning algorithms were applied to the prediction of NH₃ in the piggery. Through comparative experiments, appropriate environmental parameters (CO₂, H₂O, P, and outdoor temperature) and superior algorithms (LSTM and RNN) were selected. On this basis, the PSO algorithm was used to optimize the hyperparameters of the algorithms, and their prediction performance was also evaluated. The results showed that the R² values of PSO-LSTM and PSO-RNN were 0.9487 and 0.9458, respectively. These models had good accuracy when predicting NH₃ concentration in the piggery 0.5 h, 1 h, 1.5 h, and 2 h in advance. This study can provide a reference for the prediction of air concentrations in pig house environments.

Evaluating Al-based coagulants for drinking water facilities using Jar test and CCD/RSM analysis

This study evaluated Al-based coagulants for turbidity removal optimization in drinking water facility using Jar test and CCD/RSM analysis. The wide use of aluminum salts requires researching improved Al-based coagulants to reduce the treatment dosage. Eight polyaluminum chloride coagulants (PACl), i.e., Hyperlon 4064-PACl 2, Hyperlon 4393, 1757 X1, 1757 XL8- PACl 1, Ultrafloc 1406, Ultrafloc 3759, AlcoPAC 6, and AlcoPAC 1010 were first compared using a series of jar tests to determine the best candidate in removing the settled and filtered turbidity in water. The results showed that all PACls performed better than alum in removing water turbidity, but Hyperlon 4064 was the best. Then, the central composite design/response surface methodology (CCD/RSM) analysis was applied to Hyperlon 4064 to optimize dosage and pH to achieve the lowest final settled and filtered turbidity in the treated water, which were 21.7 mg/L, 7.53 and 27.95 mg/L, 7.91, respectively. Two quadratic models were generated by the CCD/RSM analysis with high correlations between the actual and predicted responses (R² = 0.9881 and 0.9809 for final settled turbidity and final filtered turbidity). The results from this study can provide useful information to the operating water treatment plants that use Al-based coagulants to remove turbidity in water.

Formation of Predictive-Based Models for Monitoring the Microbiological Quality of Beef Meat Processed for Fast-Food Restaurants

Consumption of raw or undercooked meat is responsible for 2.3 million foodborne illnesses yearly in Europe alone. The greater part of this illness is associated with beef meat, which is used in many traditional dishes across the world. Beneath the low microbiological quality of beef lies (pathogenic) bacterial contamination during primary production as well as inadequate hygiene operations along the farm-to-fork chain. Therefore, this study seeks to understand the microbiological quality pathway of minced beef processed for fast-food restaurants over three years using an artificial neural network (ANN) system. This simultaneous approach provided adequate precision for the prediction of a microbiological profile of minced beef meat as one of the easily spoiled products with a short shelf life. For the first time, an ANN model was developed to predict the microbiological profile of beef minced meat in fast-food restaurants according to meat and storage temperatures, butcher identification, and working shift. Predictive challenges were identified and included standardized microbiological analyses that are recommended for freshly processed meat. The obtained predictive models (an overall r² of 0.867 during the training cycle) can serve as a source of data and help for the scientific community and food safety authorities to identify specific monitoring and research needs.

Development and Validation of Novel Deep-Learning Models Using Multiple Data Types for Lung Cancer Survival

A well-established lung-cancer-survival-prediction model that relies on multiple data types, multiple novel machine-learning algorithms, and external testing is absent in the literature. This study aims to address this gap and determine the critical factors of lung cancer survival. We selected non-small-cell lung cancer patients from a retrospective dataset of the Taipei Medical University Clinical Research Database and Taiwan Cancer Registry between January 2008 and December 2018. All patients were monitored from the index date of cancer diagnosis until the event of death. Variables, including demographics, comorbidities, medications, laboratories, and patient gene tests, were used. Nine machine-learning algorithms with various modes were used. The performance of the algorithms was measured by the area under the receiver operating characteristic curve (AUC). In total, 3714 patients were included. The best performance of the artificial neural network (ANN) model was achieved when integrating all variables with the AUC, accuracy, precision, recall, and F1-score of 0.89, 0.82, 0.91, 0.75, and 0.65, respectively. The most important features were cancer stage, cancer size, age of diagnosis, smoking, drinking status, EGFR gene, and body mass index. Overall, the ANN model improved predictive performance when integrating different data types.

Modeling Physiological Predictors of Running Velocity for Endurance Athletes

Background: Properly performed training is a matter of importance for endurance athletes (EA). It allows for achieving better results and safer participation. Recently, the development of machine learning methods has been observed in sports diagnostics. Velocity at anaerobic threshold (V_AT), respiratory compensation point (V_RCP), and maximal velocity (V_max) are the variables closely corresponding to endurance performance. The primary aims of this study were to find the strongest predictors of V_AT, V_RCP, V_max, to derive and internally validate prediction models for males (1) and females (2) under TRIPOD guidelines, and to assess their machine learning accuracy. Materials and Methods: A total of 4001 EA (n_males = 3300, n_females = 671; age = 35.56 ± 8.12 years; BMI = 23.66 ± 2.58 kg·m^-2; VO_2max = 53.20 ± 7.17 mL·min^-1·kg^-1) underwent treadmill cardiopulmonary exercise testing (CPET) and bioimpedance body composition analysis. XGBoost was used to select running performance predictors. Multivariable linear regression was applied to build prediction models. Ten-fold cross-validation was incorporated for accuracy evaluation during internal validation. Results: Oxygen uptake, blood lactate, pulmonary ventilation, and somatic parameters (BMI, age, and body fat percentage) showed the highest impact on velocity. For V_AT R² = 0.57 (1) and 0.62 (2), derivation RMSE = 0.909 (1); 0.828 (2), validation RMSE = 0.913 (1); 0.838 (2), derivation MAE = 0.708 (1); 0.657 (2), and validation MAE = 0.710 (1); 0.665 (2). For V_RCP R² = 0.62 (1) and 0.67 (2), derivation RMSE = 1.066 (1) and 0.964 (2), validation RMSE = 1.070 (1) and 0.978 (2), derivation MAE = 0.832 (1) and 0.752 (2), validation MAE = 0.060 (1) and 0.763 (2). For V_max R² = 0.57 (1) and 0.65 (2), derivation RMSE = 1.202 (1) and 1.095 (2), validation RMSE = 1.205 (1) and 1.111 (2), derivation MAE = 0.943 (1) and 0.861 (2), and validation MAE = 0.944 (1) and 0.881 (2). Conclusions: The use of machine-learning methods allows for the precise determination of predictors of both submaximal and maximal running performance. Prediction models based on selected variables are characterized by high precision and high repeatability. The results can be used to personalize training and adjust the optimal therapeutic protocol in clinical settings, with a target population of EA.

Application of Soft-Computing Methods to Evaluate the Compressive Strength of Self-Compacting Concrete

This research examined machine learning (ML) techniques for predicting the compressive strength (CS) of self-compacting concrete (SCC). Multilayer perceptron (MLP), bagging regressor (BR), and support vector machine (SVM) were utilized for analysis. A total of 169 data points were retrieved from the various published articles. The data set was based on 11 input parameters, such as cement, limestone, fly ash, ground granulated blast-furnace slag, silica fume, rice husk ash, coarse aggregate, fine aggregate, superplasticizers, water, viscosity modifying admixtures, and one output with compressive strength of SCC. In terms of properly predicting the CS of SCC, the BR technique outperformed both the SVM and MLP models, as determined by the research results. In contrast to SVM and MLP, the coefficient of determination (R²) for the BR model was 0.95, whereas for SVM and MLP, the R² was 0.90 and 0.86, respectively. In addition, a k-fold cross-validation approach was adopted to check the accuracy of the employed models. The statistical measures mean absolute percent error, mean absolute error, and root mean square error ensure the validity of the model. Using sensitivity analysis, the influence of input factors on the intended CS of SCC was also explored. This analysis reveals that the highest contributing parameter towards the CS of SCC was cement with 16.2%, while rice husk ash contributed the least with 4.25% among all the input variables.

Artificial intelligence based prediction models for individuals at risk of multiple diabetic complications: A systematic review of the literature

AIM

The aim of this review is to examine the effectiveness of artificial intelligence in predicting multimorbid diabetes-related complications.

BACKGROUND

In diabetic patients, several complications are often present, which have a significant impact on the quality of life; therefore, it is crucial to predict the level of risk for diabetes and its complications.

EVALUATION

International databases PubMed, CINAHL, MEDLINE and Scopus were searched using the terms artificial intelligence, diabetes mellitus and prediction of complications to identify studies on the effectiveness of artificial intelligence for predicting multimorbid diabetes-related complications. The results were organized by outcomes to allow more efficient comparison.

KEY ISSUES

Based on the inclusion/exclusion criteria, 11 articles were included in the final analysis. The most frequently predicted complications were diabetic neuropathy (n = 7). Authors included from two to a maximum of 14 complications. The most commonly used prediction models were penalized regression, random forest and Naïve Bayes model neural network.

CONCLUSION

The use of artificial intelligence can predict the risks of diabetes complications with greater precision based on available multidimensional datasets and provides an important tool for nurses working in preventive health care.

IMPLICATIONS FOR NURSING MANAGEMENT

Using artificial intelligence contributes to a better quality of care, better autonomy of patients in diabetes management and reduction of complications, costs of medical care and mortality.

Homogeneous ensemble models for predicting infection levels and mortality of COVID-19 patients: Evidence from China

Background

Persistence of long-term COVID-19 pandemic is putting high pressure on healthcare services worldwide for several years. This article aims to establish models to predict infection levels and mortality of COVID-19 patients in China.

Methods

Machine learning models and deep learning models have been built based on the clinical features of COVID-19 patients. The best models are selected by area under the receiver operating characteristic curve (AUC) scores to construct two homogeneous ensemble models for predicting infection levels and mortality, respectively. The first-hand clinical data of 760 patients are collected from Zhongnan Hospital of Wuhan University between 3 January and 8 March 2020. We preprocess data with cleaning, imputation, and normalization.

Results

Our models obtain AUC = 0.7059 and Recall (Weighted avg) = 0.7248 in predicting infection level, while AUC=0.8436 and Recall (Weighted avg) = 0.8486 in predicting mortality ratio. This study also identifies two sets of essential clinical features. One is C-reactive protein (CRP) or high sensitivity C-reactive protein (hs-CRP) and the other is chest tightness, age, and pleural effusion.

Conclusions

Two homogeneous ensemble models are proposed to predict infection levels and mortality of COVID-19 patients in China. New findings of clinical features for benefiting the machine learning models are reported. The evaluation of an actual dataset collected from January 3 to March 8, 2020 demonstrates the effectiveness of the models by comparing them with state-of-the-art models in prediction.

Estimating 3-D whole-body composition from a chest CT scan

BACKGROUND

Estimating whole-body composition from limited region-computed tomography (CT) scans has many potential applications in clinical medicine; however, it is challenging.

PURPOSE

To investigate if whole-body composition based on several tissue types (visceral adipose tissue [VAT], subcutaneous adipose tissue [SAT], intermuscular adipose tissue [IMAT], skeletal muscle [SM], and bone) can be reliably estimated from a chest CT scan only.

METHODS

A cohort of 97 lung cancer subjects who underwent both chest CT scans and whole-body positron emission tomography-CT scans at our institution were collected. We used our in-house software to automatically segment and quantify VAT, SAT, IMAT, SM, and bone on the CT images. The field-of-views of the chest CT scans and the whole-body CT scans were standardized, namely, from vertebra T1 to L1 and from C1 to the bottom of the pelvis, respectively. Multivariate linear regression was used to develop the computer models for estimating the volumes of whole-body tissues from chest CT scans. Subject demographics (e.g., gender and age) and lung volume were included in the modeling analysis. Ten-fold cross-validation was used to validate the performance of the prediction models. Mean absolute difference (MAD) and R-squared (R² ) were used as the performance metrics to assess the model performance.

RESULTS

The R² values when estimating volumes of whole-body SAT, VAT, IMAT, total fat, SM, and bone from the regular chest CT scans were 0.901, 0.929, 0.900, 0.933, 0.928, and 0.918, respectively. The corresponding MADs (percentage difference) were 1.44 ± 1.21 L (12.21% ± 11.70%), 0.63 ± 0.49 L (29.68% ± 61.99%), 0.12 ± 0.09 L (16.20% ± 18.42%), 1.65 ± 1.40 L (10.43% ± 10.79%), 0.71 ± 0.68 L (5.14% ± 4.75%), and 0.17 ± 0.15 L (4.32% ± 3.38%), respectively.

CONCLUSION

Our algorithm shows promise in its ability to estimate whole-body compositions from chest CT scans. Body composition measures based on chest CT scans are more accurate than those based on vertebra third lumbar.

Development and Validation of Risk Prediction Models for Gestational Diabetes Mellitus Using Four Different Methods

Gestational diabetes mellitus (GDM), a common perinatal disease, is related to increased risks of maternal and neonatal adverse perinatal outcomes. We aimed to establish GDM risk prediction models that can be widely used in the first trimester using four different methods, including a score-scaled model derived from a meta-analysis using 42 studies, a logistic regression model, and two machine learning models (decision tree and random forest algorithms). The score-scaled model (seven variables) was established via a meta-analysis and a stratified cohort of 1075 Chinese pregnant women from the Northwest Women's and Children's Hospital (NWCH) and showed an area under the curve (AUC) of 0.772. The logistic regression model (seven variables) was established and validated using the above cohort and showed AUCs of 0.799 and 0.834 for the training and validation sets, respectively. Another two models were established using the decision tree (DT) and random forest (RF) algorithms and showed corresponding AUCs of 0.825 and 0.823 for the training set, and 0.816 and 0.827 for the validation set. The validation of the developed models suggested good performance in a cohort derived from another period. The score-scaled GDM prediction model, the logistic regression GDM prediction model, and the two machine learning GDM prediction models could be employed to identify pregnant women with a high risk of GDM using common clinical indicators, and interventions can be sought promptly.

Bioinformatics analysis identified immune infiltration, risk and drug prediction models of copper-induced death genes involved in salivary glands damage of primary Sjögren's syndrome

This study aimed to identify copper-induced death genes in primary Sjögren's syndrome (pSS) and explore immune infiltration, risk and drug prediction models for salivary glands (SGs) damage. The 3 datasets, including GSE40611, GSE23117, and GSE7451 from the Gene Expression Omnibus database were downloaded. The datasets were processed using the affy in R (version 4.0.3). In immune cells, copper-induced death genes were strongly expressed in "activated" dendritic cells (aDCs), macrophages and regulatory T cells (Treg). In immune functions, copper-induced death genes were strongly expressed in major histocompatibility complex (MHC) class I, human leukocyte antigen (HLA) and type I interferon (IFN) response. Correlation analysis showed that 5 genes including SLC31A1, PDHA1, DLD, ATP7B, and ATP7A were significantly correlated with immune infiltration. The nomogram suggested that the low expression of PDHA1 was significant for predicting the risk of pSS and the area under curve was 0.678. Drug model suggested that "Bathocuproine disulfonate CTD 00001350," "Vitinoin CTD 00007069," and "Resveratrol CTD 00002483" were the drugs most strongly associated with copper-induced death genes. In summary, copper-induced death genes are associated with SGs injury in pSS, which is worthy of clinicians' attention.

Should we adopt a prognosis-based approach to unexplained infertility?

The treatment of unexplained infertility is a contentious topic that continues to attract a great deal of interest amongst clinicians, patients and policy makers. The inability to identify an underlying pathology makes it difficult to devise effective treatments for this condition. Couples with unexplained infertility can conceive on their own and any proposed intervention needs to offer a better chance of having a baby. Over the years, several prognostic and prediction models based on routinely collected clinical data have been developed, but these are not widely used by clinicians and patients. In this opinion paper, we propose a prognosis-based approach such that a decision to access treatment is based on the estimated chances of natural and treatment-related conception, which, in the same couple, can change over time. This approach avoids treating all couples as a homogeneous group and minimizes unnecessary treatment whilst ensuring access to those who need it early.

Clinical Trial Considerations in Developing Treatments for Early Stages of Common, Chronic Kidney Diseases: A Scientific Workshop Cosponsored by the National Kidney Foundation and the US Food and Drug Administration

In the past decade, advances in the validation of surrogate end points for chronic kidney disease (CKD) progression have heightened interest in evaluating therapies in early CKD. In December 2020, the National Kidney Foundation sponsored a scientific workshop in collaboration with the US Food and Drug Administration (FDA) to explore patient, provider, and payor perceptions of the value of treating early CKD. The workshop reviewed challenges for trials in early CKD, including trial designs, identification of high-risk populations, and cost-benefit and safety considerations. Over 90 people representing a range of stakeholders including experts in clinical trials, nephrology, cardiology and endocrinology, patient advocacy organizations, patients, payors, health economists, regulators and policy makers attended a virtual meeting. There was consensus among the attendees that there is value to preventing the development and treating the progression of early CKD in people who are at high risk for progression, and that surrogate end points should be used to establish efficacy. Attendees also concluded that cost analyses should be holistic and include aspects beyond direct savings for treatment of kidney failure; and that safety data should be collected outside/beyond the duration of a clinical trial. Successful drug development and implementation of effective therapies will require collaboration across sponsors, patients, patient advocacy organizations, medical community, regulators, and payors.

Development of prediction models to select older RA patients with comorbidities for treatment with chronic low-dose glucocorticoids

OBJECTIVE

To develop prediction models for individual patient harm and benefit outcomes in elderly patients with rheumatoid arthritis (RA) and comorbidities treated with chronic low-dose glucocorticoid therapy or placebo.

METHODS

In the GLORIA trial, 451 RA patients aged 65+ were randomized to 2 years 5 mg/day prednisolone or placebo. Eight prediction models were developed from the dataset in a stepwise procedure based on prior knowledge. The first set of four models disregarded study treatment and examined general predictive factors. The second set of four models was similar but examined the additional role of low-dose prednisolone. In each set, two models focused on harm (1: occurrence of ≥ 1 adverse event of special interest (AESI); 2: number of AESIs per year) and two on benefit (3: early clinical response-disease activity; 4: lack of joint damage progression). Linear and logistic multivariable regression methods with backward selection were used to develop the models. The final models were assessed and internally validated with bootstrapping techniques.

RESULTS

Few variables were slightly predictive for one of the outcomes in the models, but none were of immediate clinical value. The quality of the prediction models was sufficient, the performance low to moderate: explained variance 12-15%, AUC 0.67-0.69.

CONCLUSION

Baseline factors are not helpful to select elderly RA patients for treatment with low-dose prednisolone given their low power to predict the chance of benefit or harm.

TRIAL REGISTRATION

https://clinicaltrials.gov; NCT02585258.