Super Learner Research Articles

Association of statin use with outcomes of patients admitted with COVID-19: an analysis of electronic health records using superlearner

ImportanceStatin use prior to hospitalization for Coronavirus Disease 2019 (COVID-19) is hypothesized to improve inpatient outcomes including mortality, but prior findings from large observational studies have been inconsistent, due in part to confounding. Recent advances in statistics, including incorporation of machine learning techniques into augmented inverse probability weighting with targeted maximum likelihood estimation, address baseline covariate imbalance while maximizing statistical efficiency.ObjectiveTo estimate the association of antecedent statin use with progression to severe inpatient outcomes among patients admitted for COVD-19.Design, setting and participantsWe retrospectively analyzed electronic health records (EHR) from individuals ≥ 40-years-old who were admitted between March 2020 and September 2022 for ≥ 24 h and tested positive for SARS-CoV-2 infection in the 30 days before to 7 days after admission.ExposureAntecedent statin use—statin prescription ≥ 30 days prior to COVID-19 admission.Main outcomeComposite end point of in-hospital death, intubation, and intensive care unit (ICU) admission.ResultsOf 15,524 eligible COVID-19 patients, 4412 (20%) were antecedent statin users. Compared with non-users, statin users were older (72.9 (SD: 12.6) versus 65.6 (SD: 14.5) years) and more likely to be male (54% vs. 51%), White (76% vs. 71%), and have ≥ 1 medical comorbidity (99% vs. 86%). Unadjusted analysis demonstrated that a lower proportion of antecedent users experienced the composite outcome (14.8% vs 19.3%), ICU admission (13.9% vs 18.3%), intubation (5.1% vs 8.3%) and inpatient deaths (4.4% vs 5.2%) compared with non-users. Risk differences adjusted for labs and demographics were estimated using augmented inverse probability weighting with targeted maximum likelihood estimation using Super Learner. Statin users still had lower rates of the composite outcome (adjusted risk difference: − 3.4%; 95% CI: − 4.6% to − 2.1%), ICU admissions (− 3.3%; − 4.5% to − 2.1%), and intubation (− 1.9%; − 2.8% to − 1.0%) but comparable inpatient deaths (0.6%; − 1.3% to 0.1%).Conclusions and relevanceAfter controlling for confounding using doubly robust methods, antecedent statin use was associated with minimally lower risk of severe COVID-19-related outcomes, ICU admission and intubation, however, we were not able to corroborate a statin-associated mortality benefit.

Biohydrogen from food waste: Modeling and estimation by machine learning based super learner approach

This study demonstrated the application of a hybrid Bayesian algorithm (BA) and support vector regression (SVR) as a potential super-learner tool (BA-SVR) to predict biohydrogen production from food waste-originated feedstocks. The novelty of the present approach, as compared to the existing response surface methodology (RSM), includes (i) hybridization of BA with SVR for modeling of biohydrogen production and minimization of biomethane formation, (ii) performance evaluation and comparison of the developed BA-SVR models with the existing RSM models based on the several indicators such as coefficient of determination (R2), relative error (RE), mean absolute error (MAE), mean absolute percentage error (MAPE), and root mean square error (RMSE), (iii) analysis of the robustness of the model and (iv) testing generalization ability. The calculated values of these indicators suggested that the proposed super leaner models demonstrated better performance predicting the biohydrogen and biomethane (products) responses than those using the existing RSM models - as reported in Rafieenia et al. 2019 [45]. The estimated low errors for biohydrogen: MAE = 0.5919, RMSE = 0.592, MAPE = 11.1387; for biomethane: MAE = 0.2681, RMSE = 0.2688, MAPE = 0.3708, signifie the reliable model predictions. The BA-SVR model also provided high adj R2 (>0.99 for both biohydrogen and biomethane), indicating an excellent fitting of the model. Concerning the MAPE, the proposed BA-SVR models for both the biohydrogen and biomethane responses showed superior performances (as compared to the RSM models) with a performance enhancement of 64.16% and 98.81%, respectively.

Development and evaluation of a risk algorithm predicting alcohol dependence after early onset of regular alcohol use.

Likelihood of alcohol dependence (AD) is increased among people who transition to greater levels of alcohol involvement at a younger age. Indicated interventions delivered early may be effective in reducing risk, but could be costly. One way to increase cost-effectiveness would be to develop a prediction model that targeted interventions to the subset of youth with early alcohol use who are at highest risk of subsequent AD. A prediction model was developed for DSM-IV AD onset by age 25 years using an ensemble machine-learning algorithm known as 'Super Learner'. Shapley additive explanations (SHAP) assessed variable importance. Respondents reporting early onset of regular alcohol use (i.e. by 17 years of age) who were aged 25 years or older at interview from 14 representative community surveys conducted in 13 countries as part of WHO's World Mental Health Surveys. The primary outcome to be predicted was onset of life-time DSM-IV AD by age 25 as measured using the Composite International Diagnostic Interview, a fully structured diagnostic interview. AD prevalence by age 25 was 5.1% among the 10 687 individuals who reported drinking alcohol regularly by age 17. The prediction model achieved an external area under the curve [0.78; 95% confidence interval (CI) = 0.74-0.81] higher than any individual candidate risk model (0.73-0.77) and an area under the precision-recall curve of 0.22. Overall calibration was good [integrated calibration index (ICI) = 1.05%]; however, miscalibration was observed at the extreme ends of the distribution of predicted probabilities. Interventions provided to the 20% of people with highest risk would identify 49% of AD cases and require treating four people without AD to reach one with AD. Important predictors of increased risk included younger onset of alcohol use, males, higher cohort alcohol use and more mental disorders. A risk algorithm can be created using data collected at the onset of regular alcohol use to target youth at highest risk of alcohol dependence by early adulthood. Important considerations remain for advancing the development and practical implementation of such models.

CHURN ANALISIS PADA DATA PELANGGAN TELEKOMUNIKASI MENGGUNAKAN ENSEMBLE LEARNING

Intense competition in broadband services will create high opportunities for consumers to switch providers, such as conditions that arise in competition for SMS, telephone, and internet services. The churn rate is the percentage of consumers who stop subscribing to the service. Ideally, this churn percentage is only 5% – 10%, and if it exceeds this figure, it indicates the company's inability to retain customers. A high churn rate indicates a decline in the cellular operator's market share and affects the company's revenue. Based on these problems, it is necessary to analyze the churn behavior of broadband subscribers to determine the dissatisfaction factors of cellular telecommunications consumers. Then predictions are made for customers who tend to churn from provider companies and determine the characteristics of churn and stay customers. The ensemble method is used to detect churn, which consists of several methods, including random forest, boosting, and super learner. Random Forest is proven to produce the best classification method with an excellent ability to predict customer churn, which is 80.1%, with an average usage time of 3 years.

Personalized online ensemble machine learning with applications for dynamic data streams.

In this work we introduce the personalized online super learner (POSL), an online personalizable ensemble machine learning algorithm for streaming data. POSL optimizes predictions with respect to baseline covariates, so personalization can vary from completely individualized, that is, optimization with respect to subject ID, to many individuals, that is, optimization with respect to common baseline covariates. As an online algorithm, POSL learns in real time. As a super learner, POSL is grounded in statistical optimality theory and can leverage a diversity of candidate algorithms, including online algorithms with different training and update times, fixed/offline algorithms that are not updated during POSL's fitting procedure, pooled algorithms that learn from many individuals' time series, and individualized algorithms that learn from within a single time series. POSL's ensembling of the candidates can depend on the amount of data collected, the stationarity of the time series, and the mutual characteristics of a group of time series. Depending on the underlying data-generating process and the information available in the data, POSL is able to adapt to learning across samples, through time, or both. For a range of simulations that reflect realistic forecasting scenarios and in a medical application, we examine the performance of POSL relative to other current ensembling and online learning methods. We show that POSL is able to provide reliable predictions for both short and long time series, and it's able to adjust to changing data-generating environments. We further cultivate POSL's practicality by extending it to settings where time series dynamically enter and exit.

A Demonstration of Modified Treatment Policies to Evaluate Shifts in Mobility and COVID-19 Case Rates in US Counties.

Mixed evidence exists of associations between mobility data and coronavirus disease 2019 (COVID-19) case rates. We aimed to evaluate the county-level impact of reducing mobility on new COVID-19 cases in summer/fall of 2020 in the United States and to demonstrate modified treatment policies to define causal effects with continuous exposures. Specifically, we investigated the impact of shifting the distribution of 10 mobility indexes on the number of newly reported cases per 100,000 residents 2 weeks ahead. Primary analyses used targeted minimum loss-based estimation with Super Learner to avoid parametric modeling assumptions during statistical estimation and flexibly adjust for a wide range of confounders, including recent case rates. We also implemented unadjusted analyses. For most weeks, unadjusted analyses suggested strong associations between mobility indexes and subsequent new case rates. However, after confounder adjustment, none of the indexes showed consistent associations under mobility reduction. Our analysis demonstrates the utility of this novel distribution-shift approach to defining and estimating causal effects with continuous exposures in epidemiology and public health.

A hybrid super ensemble learning model for the early-stage prediction of diabetes risk

Diabetes mellitus has become a rapidly growing chronic health problem worldwide. There has been a noticeable increase in diabetes cases in the last two decades. Recent advances in ensemble machine learning methods play an important role in the early detection of diabetes mellitus. These methods are both faster and less costly than traditional methods. This study aims to propose a new super ensemble learning model to enable an early diagnosis of diabetes mellitus. Super learner is a cross-validation-based approach that makes better predictions by combining prediction results of more than one machine learning algorithm. The proposed super learner model was created with four base-learners (logistic regression, decision tree, random forest, gradient boosting) and a meta learner (support vector machines) as a result of a case study. Three different dataset were used to measure the robustness of the proposed model. Chi-square was determined as an optimal feature selection technique from five different techniques, and also hyper-parameter settings were made with GridSearch. Finally, the proposed new super learner model achieved to obtain the best accuracy results in the detection of Diabetes mellitus compared to the base-learners for the early-stage diabetes risk prediction (99.6%), PIMA (92%), and diabetes 130-US hospitals (98%) dataset, respectively. This study revealed that super learner algorithms can be effectively used in the detection of diabetes mellitus. Also, obtaining of the high and convincing statistical scores shows the robustness of the proposed super learner model.

Efficient thyroid disorder identification with weighted voting ensemble of super learners by using adaptive synthetic sampling technique

<abstract><p>There are millions of people suffering from thyroid disease all over the world. For thyroid cancer to be effectively treated and managed, a correct diagnosis is necessary. In this article, we suggest an innovative approach for diagnosing thyroid disease that combines an adaptive synthetic sampling method with weighted average voting (WAV) ensemble of two distinct super learners (SLs). Resampling techniques are used in the suggested methodology to correct the class imbalance in the datasets and a group of two SLs made up of various base estimators and meta-estimators is used to increase the accuracy of thyroid cancer identification. To assess the effectiveness of our suggested methodology, we used two publicly accessible datasets: the KEEL thyroid illness (Dataset1) and the hypothyroid dataset (Dataset2) from the UCI repository. The findings of using the adaptive synthetic (ADASYN) sampling technique in both datasets revealed considerable gains in accuracy, precision, recall and F1-score. The WAV ensemble of the two distinct SLs that were deployed exhibited improved performance when compared to prior existing studies on identical datasets and produced higher prediction accuracy than any individual model alone. The suggested methodology has the potential to increase the accuracy of thyroid cancer categorization and could assist with patient diagnosis and treatment. The WAV ensemble strategy computational complexity and the ideal choice of base estimators in SLs continue to be constraints of this study that call for further investigation.</p></abstract>

A SuperLearner-enforced approach for the estimation of treatment effect in pediatric trials.

Randomized Clinical Trials (RCT) represent the gold standard among scientific evidence. RCTs are tailored to control selection bias and the confounding effect of baseline characteristics on the effect of treatment. However, trial conduction and enrolment procedures could be challenging, especially for rare diseases and paediatric research. In these research frameworks, the treatment effect estimation could be compromised. A potential countermeasure is to develop predictive models on the probability of the baseline disease based on previously collected observational data. Machine learning (ML) algorithms have recently become attractive in clinical research because of their flexibility and improved performance compared to standard statistical methods in developing predictive models. This manuscript proposes an ML-enforced treatment effect estimation procedure based on an ensemble SuperLearner (SL) approach, trained on historical observational data, to control the confounding effect. The REnal SCarring Urinary infEction trial served as a motivating example. Historical observational study data have been simulated through 10,000 Monte Carlo (MC) runs. Hypothetical RCTs have been also simulated, for each MC run, assuming different treatment effects of antibiotics combined with steroids. For each MC simulation, the SL tool has been applied to the simulated observational data. Furthermore, the average treatment effect (ATE), has been estimated on the trial data and adjusted for the SL predicted probability of renal scar. The simulation results revealed an increased power in ATE estimation for the SL-enforced estimation compared to the unadjusted estimates for all the algorithms composing the ensemble SL.

Decision tree models for the estimation of geo-polymer concrete compressive strength.

The green concretes industry benefits from utilizing gel to replace parts of the cement in concretes. However, measuring the compressive strength of geo-polymer concretes (CSGPoC) needs a significant amount of work and expenditure. Therefore, the best idea is predicting CSGPoC with a high level of accuracy. To do this, the base learner and super learner machine learning models were proposed in this study to anticipate CSGPoC. The decision tree (DT) is applied as base learner, and the random forest and extreme gradient boosting (XGBoost) techniques are used as super learner system. In this regard, a database was provided involving 259 CSGPoC data samples, of which four-fifths of is considered for the training model and one-fifth is selected for the testing models. The values of fly ash, ground-granulated blast-furnace slag (GGBS), Na2SiO3, NaOH, fine aggregate, gravel 4/10 mm, gravel 10/20 mm, water/solids ratio, and NaOH molarity were considered as input of the models to estimate CSGPoC. To evaluate the reliability and performance of the decision tree (DT), XGBoost, and random forest (RF) models, 12 performance evaluation metrics were determined. Based on the obtained results, the highest degree of accuracy is achieved by the XGBoost model with mean absolute error (MAE) of 2.073, mean absolute percentage error (MAPE) of 5.547, Nash-Sutcliffe (NS) of 0.981, correlation coefficient (R) of 0.991, R2 of 0.982, root mean square error (RMSE) of 2.458, Willmott's index (WI) of 0.795, weighted mean absolute percentage error (WMAPE) of 0.046, Bias of 2.073, square index (SI) of 0.054, p of 0.027, mean relative error (MRE) of -0.014, and a20 of 0.983 for the training model and MAE of 2.06, MAPE of 6.553, NS of 0.985, R of 0.993, R2 of 0.986, RMSE of 2.307, WI of 0.818, WMAPE of 0.05, Bias of 2.06, SI of 0.056, p of 0.028, MRE of -0.015, and a20 of 0.949 for the testing model. By importing the testing set into trained models, values of 0.8969, 0.9857, and 0.9424 for R2 were obtained for DT, XGBoost, and RF, respectively, which show the superiority of the XGBoost model in CSGPoC estimation. In conclusion, the XGBoost model is capable of more accurately predicting CSGPoC than DT and RF models.

Mapping phosphorus sorption and availability in California vineyard soils using an ensemble of machine learning models

AbstractSpatial variability of soil P is tied to pedogenic state factors and management practices in cultivated soils. The distribution of P availability and sorption was predictively mapped in the Napa and Lodi American Viticulture Areas in California. We tested three machine learning algorithms, Random Forest (RF), Extreme Gradient Boosting (XGB), and Cubist, as well as two super learner ensembles of base models, model stacking and model averaging. Pedons (n = 141) were analyzed for Olsen P and phosphorus‐sorption index (PSI), aggregated by depth weighted average (0–30 cm and 30–100 cm) and intersected with rasters of environmental predictors to model Olsen P and PSI. Base models (RF, XGB, and Cubist) performed well for PSI prediction (R2 = .68–.73), but less well for Olsen P (R2 = .46–.56). For ensembles, model averaging was selected for PSI at 0–30 cm (R2 = .77) and model stacking was selected for PSI at 30–100 cm (R2 = .74). For Olsen P, model averaging was selected for 0–30 cm (R2 = .42), and model stacking for 30–100 cm (R2 = .52). Predictions (30‐m) highlight regional trends in P‐sorption capacity and Olsen P reflective of differences in pedogenic controls on P dynamics. Predictions were strong for PSI, and less robust for Olsen P. Fe/Al‐(hydr)oxides control P sorption in weathered soils, whereas management influences Olsen P. Because the spatial variability of Fe/Al‐(hydr)oxides is tied to pedogenic state factors, P‐sorption capacity lends itself to environmental correlation mapping owing to the pedological underpinnings of digital soil mapping.

Forecasting domestic waste generation during successive COVID-19 lockdowns by Bidirectional LSTM super learner neural network

Accurate prediction of domestic waste generation is a challenging task for municipalities to implement sustainable waste management strategies. In the present study, domestic waste generation in the Kingdom of Bahrain, representing a Small Island Developing State (SIDS) case study, has been investigated during successive COVID-19 lockdowns due to the pandemic in 2020. Temporal trends of daily domestic waste generation between 2019 and 2020 and their statistical analyses exhibited remarkable variations highlighting the impact of consecutive COVID-19 lockdowns on domestic waste generation. Machine learning has great potential for predicting solid waste generation rates, but only a few studies utilized deep learning approaches. The state-of-the-art Bidirectional Long Short-Term Memory (BiLSTM) network model as a deep learning method is applied to forecast daily domestic waste data in 2020. Bayesian optimization algorithm (BOA) was hybridized with BiLSTM to generate a super learner approach. The performance of the BOA-BiLSTM super learner model was further compared with the statistical ARIMA model. Performance indicators of the developed models using ARIMA and BiLSTM showed that the latter yielded superior performance for short-term forecasts of domestic waste generation. The MAE, RMSE, MAPE, and R2 were 47.38, 60.73, 256.43, and 0.46, respectively, for the ARIMA model, compared to 3.67, 12.57, 0.24, and 0.96, respectively, for the BiLSTM model. Additionally, the relative errors for the BiLSTM model were lower than those of the ARIMA model. This study highlights that the BiLSTM can be a reliable forecasting tool for solid waste management policymakers during public health emergencies.

Occupations on the map: Using a super learner algorithm to downscale labor statistics.

Detailed and accurate labor statistics are fundamental to support social policies that aim to improve the match between labor supply and demand, and support the creation of jobs. Despite overwhelming evidence that labor activities are distributed unevenly across space, detailed statistics on the geographical distribution of labor and work are not readily available. To fill this gap, we demonstrated an approach to create fine-scale gridded occupation maps by means of downscaling district-level labor statistics, informed by remote sensing and other spatial information. We applied a super-learner algorithm that combined the results of different machine learning models to predict the shares of six major occupation categories and the labor force participation rate at a resolution of 30 arc seconds (~1x1 km) in Vietnam. The results were subsequently combined with gridded information on the working-age population to produce maps of the number of workers per occupation. The super learners outperformed (n = 6) or had similar (n = 1) accuracy in comparison to best-performing single machine learning algorithms. A comparison with an independent high-resolution wealth index showed that the shares of the four low-skilled occupation categories (91% of the labor force), were able to explain between 28% and 43% of the spatial variation in wealth in Vietnam, pointing at a strong spatial relationship between work, income and wealth. The proposed approach can also be applied to produce maps of other (labor) statistics, which are only available at aggregated levels.

The Conditional Super Learner.

Using cross validation to select the best model from a library is standard practice in machine learning. Similarly, meta learning is a widely used technique where models previously developed are combined (mainly linearly) with the expectation of improving performance with respect to individual models. In this article we consider the Conditional Super Learner (CSL), an algorithm that selects the best model candidate from a library of models conditional on the covariates. The CSL expands the idea of using cross validation to select the best model and merges it with meta learning. We propose an optimization algorithm that finds a local minimum to the problem posed and proves that it converges at a rate faster than O_p(n^-1/4). We offer empirical evidence that: (1) CSL is an excellent candidate to substitute stacking and (2) CLS is suitable for the analysis of Hierarchical problems. Additionally, implications for global interpretability are emphasized.

A novel automated SuperLearner using a genetic algorithm-based hyperparameter optimization

Industrial revolution 4.0 has pushed industries worldwide to use machine learning (ML) models to address real-world engineering problems. The industry generally faces two main challenges in ML applications: the lack of skilled data scientists and the cost of obtaining large labeled datasets. These challenges need to be addressed to unlock the full potential of ML. In this work, a novel automated SuperLearner (SL) model using a genetic algorithm (AutoSL-GA) based hyperparameter optimization (HPO) is introduced to address the aforementioned challenges and assist scientists and engineers. Detailed comparisons are performed between AutoSL-GA, SuperLearner using Bayesian-based HPO (AutoSL-BO), and another well-known automated machine learning (AutoML) algorithm called Tree-based pipeline optimization tool (TPOT). Six different benchmark datasets were used to compare the performance and computational times of the models. AutoSL-GA resulted in higher performance with lower computational time than other models for all six benchmark datasets. Finally, a sensitivity analysis for dataset size was performed, in which AutoSL-GA also outperformed the other models across the dataset sizes while consuming the least computational resources.

Coupling Process-Based Models and Machine Learning Algorithms for Predicting Yield and Evapotranspiration of Maize in Arid Environments

Crop yield prediction is critical for investigating the yield gap and potential adaptations to environmental and management factors in arid regions. Crop models (CMs) are powerful tools for predicting yield and water use, but they still have some limitations and uncertainties; therefore, combining them with machine learning algorithms (MLs) could improve predictions and reduce uncertainty. To that end, the DSSAT-CERES-maize model was calibrated in one location and validated in others across Egypt with varying agro-climatic zones. Following that, the dynamic model (CERES-Maize) was used for long-term simulation (1990–2020) of maize grain yield (GY) and evapotranspiration (ET) under a wide range of management and environmental factors. Detailed outputs from three growing seasons of field experiments in Egypt, as well as CERES-maize outputs, were used to train and test six machine learning algorithms (linear regression, ridge regression, lasso regression, K-nearest neighbors, random forest, and XGBoost), resulting in more than 1.5 million simulated yield and evapotranspiration scenarios. Seven warming years (i.e., 1991, 1998, 2002, 2005, 2010, 2013, and 2020) were chosen from a 31-year dataset to test MLs, while the remaining 23 years were used to train the models. The Ensemble model (super learner) and XGBoost outperform other models in predicting GY and ET for maize, as evidenced by R2 values greater than 0.82 and RRMSE less than 9%. The broad range of management practices, when averaged across all locations and 31 years of simulation, not only reduced the hazard impact of environmental factors but also increased GY and reduced ET. Moving beyond prediction and interpreting the outputs from Lasso and XGBoost, and using global and local SHAP values, we found that the most important features for predicting GY and ET are maximum temperatures, minimum temperature, available water content, soil organic carbon, irrigation, cultivars, soil texture, solar radiation, and planting date. Determining the most important features is critical for assisting farmers and agronomists in prioritizing such features over other factors in order to increase yield and resource efficiency values. The combination of CMs and ML algorithms is a powerful tool for predicting yield and water use in arid regions, which are particularly vulnerable to climate change and water scarcity.

A new machine learning algorithm to predict veno-arterial ECMO implantation after post-cardiotomy low cardiac output syndrome

BackgroundPost-cardiotomy low cardiac output syndrome (PC-LCOS) is a life-threatening complication after cardiac surgery involving a cardiopulmonary bypass (CPB). Mechanical circulatory support with veno-arterial membrane oxygenation (VA-ECMO) may be necessary in the case of refractory shock. The objective of the study was to develop a machine-learning algorithm to predict the need for VA-ECMO implantation in patients with PC-LCOS. Patients and methodsPatients were included in the study with moderate to severe PC-LCOS (defined by a vasoactive inotropic score (VIS) > 10 with clinical or biological markers of impaired organ perfusion or need for mechanical circulatory support after cardiac surgery) from two university hospitals in Paris, France. The Deep Super Learner, an ensemble machine learning algorithm, was trained to predict VA-ECMO implantation using features readily available at the end of a CPB. Feature importance was estimated using Shapley values. ResultsBetween January 2016 and December 2019, 285 patients were included in the development dataset and 190 patients in the external validation dataset. The primary outcome, the need for VA-ECMO implantation, occurred respectively, in 16% (n = 46) and 10% (n = 19) in the development and the external validation datasets. The Deep Super Learner algorithm achieved a 0.863 (0.793−0.928) ROC AUC to predict the primary outcome in the external validation dataset. The most important features were the first postoperative arterial lactate value, intraoperative VIS, the absence of angiotensin-converting enzyme treatment, body mass index, and EuroSCORE II. ConclusionsWe developed an explainable ensemble machine learning algorithm that could help clinicians predict the risk of deterioration and the need for VA-ECMO implantation in moderate to severe PC-LCOS patients.

Predicting cumulative lead (Pb) exposure using the Super Learner algorithm

Chronic lead (Pb) exposure causes long term health effects. While recent exposure can be assessed by measuring blood lead (half-life 30 days), chronic exposures can be assessed by measuring lead in bone (half-life of many years to decades). Bone lead measurements, in turn, have been measured non-invasively in large population-based studies using x-ray fluorescence techniques, but the method remains limited due to technical availability, expense, and the need for licensing radioactive materials used by the instruments. Thus, we developed prediction models for bone lead concentrations using a flexible machine learning approach--Super Learner, which combines the predictions from a set of machine learning algorithms for better prediction performance. The study population included 695 men in the Normative Aging Study, aged 48 years and older, whose bone (patella and tibia) lead concentrations were directly measured using K-shell-X-ray fluorescence. Ten predictors (blood lead, age, education, job type, weight, height, body mass index, waist circumference, cumulative cigarette smoking (pack-year), and smoking status) were selected for patella lead and 11 (the same 10 predictors plus serum phosphorus) for tibia lead using the Boruta algorithm. We implemented Super Learner to predict bone lead concentrations by calculating a weighted combination of predictions from 8 algorithms. In the nested cross-validation, the correlation coefficients between measured and predicted bone lead concentrations were 0.58 for patella lead and 0.52 for tibia lead, which has improved the correlations obtained in previously-published linear regression-based prediction models. We evaluated the applicability of these prediction models to the National Health and Nutrition Examination Survey for the associations between predicted bone lead concentrations and blood pressure, and positive associations were observed. These bone lead prediction models provide reasonable accuracy and can be used to evaluate health effects of cumulative lead exposure in studies where bone lead is not measured.

Exploring the Influencing Factors in Identifying Soil Texture Classes Using Multitemporal Landsat-8 and Sentinel-2 Data

Soil texture is a key soil property driving physical, chemical, biological, and hydrological processes in soils. The rapid development of remote sensing techniques shows great potential for mapping soil properties. This study highlights the effectiveness of multitemporal remote sensing data in identifying soil textural class by using retrieved vegetation properties as proxies of soil properties. The impacts of sensors, modeling resolutions, and modeling techniques on the accuracy of soil texture classification were explored. Multitemporal Landsat-8 and Sentinel-2 images were individually acquired at the same time periods. Three satellite-based experiments with different inputs, i.e., Landsat-8 data, Sentinel-2 data (excluding red-edge parameters), and Sentinel-2 data (including red-edge parameters) were conducted. Modeling was carried out at three spatial resolutions (10, 30, 60 m) using five machine-learning (ML) methods: random forest, support vector machine, gradient-boosting decision tree, categorical boosting, and super learner that combined the four former classifiers based on the stacking concept. In addition, a novel SHapley Addictive Explanation (SHAP) technique was introduced to explain the outputs of the ML model. The results showed that the sensors, modeling resolutions, and modeling techniques significantly affected the prediction accuracy. The models using Sentinel-2 data with red-edge parameters performed consistently best. The models usually gave better results at fine (10 m) and medium (30 m) modeling resolutions than at a coarse (60 m) resolution. The super learner provided higher accuracies than other modeling techniques and gave the highest values of overall accuracy (0.8429), kappa (0.7611), precision (0.8378), recall rate (0.8393), and F1-score (0.8398) at 30 m with Sentinel-2 data involving red-edge parameters. The SHAP technique quantified the contribution of each variable for different soil textural classes, revealing the critical roles of red-edge parameters in separating loamy soils. This study provides comprehensive insights into the effective modeling of soil properties on various scales using multitemporal optical images.

Decision support by machine learning systems for acute management of severely injured patients: A systematic review.

IntroductionTreating severely injured patients requires numerous critical decisions within short intervals in a highly complex situation. The coordination of a trauma team in this setting has been shown to be associated with multiple procedural errors, even of experienced care teams. Machine learning (ML) is an approach that estimates outcomes based on past experiences and data patterns using a computer-generated algorithm. This systematic review aimed to summarize the existing literature on the value of ML for the initial management of severely injured patients.MethodsWe conducted a systematic review of the literature with the goal of finding all articles describing the use of ML systems in the context of acute management of severely injured patients. MESH search of Pubmed/Medline and Web of Science was conducted. Studies including fewer than 10 patients were excluded. Studies were divided into the following main prediction groups: (1) injury pattern, (2) hemorrhage/need for transfusion, (3) emergency intervention, (4) ICU/length of hospital stay, and (5) mortality.ResultsThirty-six articles met the inclusion criteria; among these were two prospective and thirty-four retrospective case series. Publication dates ranged from 2000 to 2020 and included 32 different first authors. A total of 18,586,929 patients were included in the prediction models. Mortality was the most represented main prediction group (n = 19). ML models used were artificial neural network ( n = 15), singular vector machine (n = 3), Bayesian network (n = 7), random forest (n = 6), natural language processing (n = 2), stacked ensemble classifier [SuperLearner (SL), n = 3], k-nearest neighbor (n = 1), belief system (n = 1), and sequential minimal optimization (n = 2) models. Thirty articles assessed results as positive, five showed moderate results, and one article described negative results to their implementation of the respective prediction model.ConclusionsWhile the majority of articles show a generally positive result with high accuracy and precision, there are several requirements that need to be met to make the implementation of such models in daily clinical work possible. Furthermore, experience in dealing with on-site implementation and more clinical trials are necessary before the implementation of ML techniques in clinical care can become a reality.