Gradient-boosted Model Research Articles

Developing Model-Agnostic Meta-Learning Enabled Lightbgm Model Asthma Level Prediction in Smart Healthcare Modeling

Millions of people across the world suffer from the chronic respiratory condition known as asthma. Predicting the severity of asthma based on a variety of personal and environmental characteristics might yield useful information for preventative measures. LightGBM model is a gradient-boosted model with the potential for great accuracy, but it requires careful hyperparameter adjustment to reach its full potential. Common tuning techniques have a hard time generalizing to new data distributions. The dataset was used, and its many subsets were used to represent various demographics and geographic areas. LightGBM was configured with hyperparameters, trained on a sample dataset, and then verified for each job. To quickly adjust to new tasks, the MAML method sought to find the optimal values for its hyperparameters. After the meta-training step was complete, the generalizability of the hyperparameters was tested on new data. After including MAML for hyperparameter adjustment, the Light- GBM model showed a gain of 7% in accuracy, coming in at 98.5%. Predictions of severe asthma had a crucially high 97.8 percent degree of accuracy. The model’s recall rate for severe asthma levels was 97.4%, demonstrating its capacity to reliably detect and anticipate important instances. An F1-score of 97.1%, a metric that averages the accuracy and recall of a model, is indicative of good overall performance.. For gradient-boosted model applications like asthma level prediction, MAML provides a viable path for hyperparameter adjustment. Although there are obstacles to be overcome, this method has the potential to greatly improve the flexibility and precision of predictive healthcare models. More effective implementations and a wider range of applications can be explored in future studies.

Explainable Machine Learning Models to Predict Gibbs-Donnan Effect During Ultrafiltration and Diafiltration of High-Concentration Monoclonal Antibody Formulations.

Evaluating the Gibbs-Donnan and volume exclusion effects during protein ultrafiltration and diafiltration (UF/DF) is crucial in biopharmaceutical process development to precisely control the concentration of the drug substance in the final formulation. Understanding the interactions between formulation excipients and proteins under these conditions requires a domain-specific knowledge of molecular-level phenomena. This study developed gradient boosted tree models to predict the Gibbs-Donnan and volume exclusion effects for amino acids and therapeutic monoclonal antibodies using simple molecular descriptors. The models' predictions were interpreted by information gain and Shapley additive explanation (SHAP) values to understand the modes of action of the antibodies and excipients and to validate the models. The results translated feature effects in machine learning to real-world molecular interactions, which were cross-referenced with existing scientific literature for verification. The models were validated in pilot-scale manufacturing runs of two antibody products requiring high levels of concentration. By only requiring a molecule's biophysicochemical descriptors and process conditions, the proposed models provide an in silico alternative to conventional UF/DF experiments to accelerate process development and boost process understanding of the underlying molecular mechanisms through rational interpretation of the models.

Addressing bias in bagging and boosting regression models.

As artificial intelligence (AI) becomes widespread, there is increasing attention on investigating bias in machine learning (ML) models. Previous research concentrated on classification problems, with little emphasis on regression models. This paper presents an easy-to-apply and effective methodology for mitigating bias in bagging and boosting regression models, that is also applicable to any model trained through minimizing a differentiable loss function. Our methodology measures bias rigorously and extends the ML model's loss function with a regularization term to penalize high correlations between model errors and protected attributes. We applied our approach to three popular tree-based ensemble models: a random forest model (RF), a gradient-boosted model (GBT), and an extreme gradient boosting model (XGBoost). We implemented our methodology on a case study for predicting road-level traffic volume, where RF, GBT, and XGBoost models were shown to have high accuracy. Despite high accuracy, the ML models were shown to perform poorly on roads in minority-populated areas. Our bias mitigation approach reduced minority-related bias by over 50%.

Gradient boosted decision trees reveal nuances of auditory discrimination behavior.

Animal psychophysics can generate rich behavioral datasets, often comprised of many 1000s of trials for an individual subject. Gradient-boosted models are a promising machine learning approach for analyzing such data, partly due to the tools that allow users to gain insight into how the model makes predictions. We trained ferrets to report a target word's presence, timing, and lateralization within a stream of consecutively presented non-target words. To assess the animals' ability to generalize across pitch, we manipulated the fundamental frequency (F0) of the speech stimuli across trials, and to assess the contribution of pitch to streaming, we roved the F0 from word token to token. We then implemented gradient-boosted regression and decision trees on the trial outcome and reaction time data to understand the behavioral factors behind the ferrets' decision-making. We visualized model contributions by implementing SHAPs feature importance and partial dependency plots. While ferrets could accurately perform the task across all pitch-shifted conditions, our models reveal subtle effects of shifting F0 on performance, with within-trial pitch shifting elevating false alarms and extending reaction times. Our models identified a subset of non-target words that animals commonly false alarmed to. Follow-up analysis demonstrated that the spectrotemporal similarity of target and non-target words rather than similarity in duration or amplitude waveform was the strongest predictor of the likelihood of false alarming. Finally, we compared the results with those obtained with traditional mixed effects models, revealing equivalent or better performance for the gradient-boosted models over these approaches.

Validation of a novel algorithm with a high specificity in ruling out MDS.

A previously published web-based App using Gradient-boosted models (GBMs) of eight laboratory parameters was established by Oster et al. to facilitate diagnosis or exclusion of myelodysplastic syndromes (MDS) in patients. To validate their algorithm, we compared 175 anemic patients with MDS diagnosis from our German MDS Registry with 1378 non-MDS anemic patients who consulted various specialties in the Düsseldorf university hospital. Based on hemoglobin level, leukocyte and platelet count, mean corpuscular volume, absolute neutrophil count, absolute monocyte count, glucose and creatinine, plus the patients' gender and age, we could not reproduce a high negative predictive value (NPV), but confirmed a useful specificity of 90.9% and a positive predictive value (PPV) of 77.1%. 1192 of 1378 controls were correctly categorized as "probably not MDS (pnMDS)" patients. A total of 65 patients were wrongly classified as "probable MDS (pMDS)," of whom 48 had alternative explanations for their altered laboratory results. In a second analysis, we included 29 patients with chronic myelomonocytic leukemia (CMML) resulting in only one label as possible MDS, suggesting that highly proliferative bone marrow disorders are correctly excluded. The possibility of reliably excluding MDS from differential diagnosis based on peripheral blood lab work appears to be attractive for patients and physicians alike while the confirmation of MDS diagnosis still requires a bone marrow biopsy.

Emissions and mitigation potential of endocrine disruptors during outdoor exercise: Fate, transport, and implications for human health

The endocrine system is responsible for secreting and controlling hormones crucial in regulating key body activities. However, endocrine disruptors or endocrine-disrupting chemicals (EDCs) can harm human health and well-being by interfering with this complex process. This report seeks to assess the present state of understanding about endocrine disruptors in China, including their origins, impacts, and obstacles, and to provide actionable recommendations for reducing exposure and mitigating negative effects. Strong negative correlations between ANOE and rural ecological compensation (REC) and a negative correlation between ANOE and forest coverage (FC) were found in this analysis of the relationships between agricultural nitrous oxide emissions (ANOE), agricultural methane emissions (AME), and land use and land cover variables (LUPC). Just as LUPC is significantly inversely related to FC, AME is positively related. The team uses a gradient-boosted model (GBM) with a Gaussian loss function and fine-tunes the model's parameters to achieve optimal performance and reliable prediction results. With a relative relevance score of 90.36 for ANOE and 67.64 for AME, the analysis shows that LUPC is the most important factor in influencing emission levels. This study aims to increase knowledge of endocrine disruptors' potential advantages and disadvantages in outdoor exercise. The study aims to aid in preventing and managing many diseases and disorders caused by hormonal imbalances or disruptions by examining the origins, effects, and potential mitigation of these substances during outdoor activity. Safe and healthful outdoor exercise is promoted by the study's efforts to discover and implement effective and sustainable solutions to decrease emissions and exposure to endocrine disruptors. This comprehensive study aims to promote a healthier and more sustainable environment for individuals engaging in outdoor exercise by synthesizing current knowledge, providing practical recommendations, and emphasizing the importance of awareness and action.

Ecological Factors Driving Tree Diversity across Spatial Scales in Temperate Forests, Northeast China

Species diversity has been shown to be influenced by environmental factors, but the mechanism underlying their relationship remains unclear across spatial scales. Based on field investigation data collected from 3077 sample plots in temperate forest ecosystems, we compared tree species richness, evenness and dominance at 10 km × 10 km, 30 km × 30 km and 90 km × 90 km spatial scales. Then, we detected the scale dependence of changes in tree species composition on climate, topography and forest structure using variation partitioning and quantified their contribution to tree diversity with gradient–boosted models (GBMs) and fitted their relationships. The magnitude of tree richness, evenness and dominance significantly increased with spatial scale. Ecological factors jointly accounted for 24.3%, 26.5% and 38.5% of the variation in tree species composition at the three spatial scales, respectively. The annual mean temperature had a strong impact on tree richness, evenness and dominance and peaked at an intermediate scale. Tree evenness and dominance increased with the variation of temperature but had upper and lower limits. Tree richness obviously increased with annual precipitation on multiple scales and decreased with annual sunshine duration at large spatial scales. Tree richness, evenness and dominance obviously increased with the variation in elevation and diameter at breast height at large scales and small scales, respectively. Tree dominance decreased with tree height at a small scale. The dependence of tree diversity on ecological factors increased with spatial scales. Furthermore, different factors exert various controls on tree diversity at different spatial scales, representing a comprehensive mechanism regulating tree diversity.

Prediction of Liver Steatosis and Fibrosis Based on Clinical Variables Using a Large National Survey Database.

Vibration-controlled transient elastography (VCTA) and controlled attenuation parameter (CAP) are used more frequently to diagnose liver fibrosis and steatosis among nonalcoholic fatty liver disease patients. However, limited robust data are available on the clinical variables strongly related to these disorders and who needs to be referred for screening. We used the National Health and Nutritional Examination Survey 2017-2018 database to identify the clinical predictors strongly related to liver steatosis and advanced fibrosis. Baseline comparisons among these groups were made based on widely accepted cutoffs. Linear and logistic regressions were performed to identify the associations between the clinical variables and liver steatosis and fibrosis. We used adaptive lasso regression, gradient-boosted model, and decision trees to determine clinical variables strongly related to these outcomes. A Naïve Byes classifier and decision trees were used to calculate the predicted probabilities of liver steatosis and fibrosis. 32% of our population had evidence of liver steatosis using 294 dB/m as a cutoff. An increase in age, serum triglyceride, and body mass index were associated with a statistically significant increase in liver steatosis; in contrast, females had statistically significantly lower values for liver steatosis by 15 points in the multivariable linear regression model. Serum LDL, smoking, and systolic and diastolic blood pressure are poorly associated with liver steatosis in the adaptive lasso regression. On the other hand, sex, tobacco use, metabolic energy expenditure, and serum triglyceride are the least associated with liver fibrosis based on decision tree analysis and a gradient-boosted model. In decision trees, people with a body mass index above 30 and HbA1c above 5.7 have a 72% likelihood of liver steatosis compared to 14% for people with a body mass index below 30. On the other hand, people with a body mass index above 41 have a 38% likelihood of liver fibrosis. Body mass index, hemoglobin A1c, serum triglyceride level, sex, and age could provide a good prediction for liver steatosis, while body mass index, blood pressure, platelet counts, hemoglobin A1c, serum LDL, or HDL are highly associated with liver fibrosis and should be used as an initial screening tool prior referral for VCTE/CAP.

Machine Learning Analysis of Physical Activity Data to Classify Postural Dysfunction.

Machine learning (ML) analysis of biometric data in non-controlled environments is underexplored. To evaluate whether ML analysis of physical activity data can be employed to classify whether individuals have postural dysfunction in middle-aged and older individuals. A 1 week period of physical activity was measured by a waist-worn uni-axial accelerometer during the 2003-2004 National Health and Nutrition Examination Survey sampling period. Features of physical activity along with basic demographic information (42 variables) were paired with ML models to predict the success or failure of a standard 30 s modified Romberg test during which participants had their eyes closed and stood upon a 3-inch compliant surface. Model performance was evaluated by area under the receiver operating characteristic curve (AUC-ROC), balanced accuracy, and F1-score. The cohort was comprised of 1625 participants ≥40 years (median age 61, IQR 51-71). Approximately half (47%) were diagnosed with postural dysfunction having failed the binarized (pass/fail) scoring mechanism of the modified Romberg exam. Five ML models were trained on the classification task, achieving AUC values ranging from 0.67 to 0.73. The support vector machine (SVM) and a gradient-boosted model, XGBoost, achieved the highest AUC of 0.73 (SD 0.71-0.75). Age was the most important variable for SVM classification, followed by four features that evaluated accelerometer counts at various thresholds, including those delineating total, moderate, and moderate-vigorous activity. ML analysis of accelerometer-derived physical activity data to classify postural dysfunction in middle-aged and older individuals is feasible in real-world environments such as the home. 3 Laryngoscope, 133:3529-3533, 2023.

MaxEnt brings comparable results when the input data are being completed; Model parameterization of four species distribution models.

Species distribution models (SDMs) are practical tools to assess the habitat suitability of species with numerous applications in environmental management and conservation planning. The manipulation of the input data to deal with their spatial bias is one of the advantageous methods to enhance the performance of SDMs. However, the development of a model parameterization approach covering different SDMs to achieve well-performing models has rarely been implemented. We integrated input data manipulation and model tuning for four commonly-used SDMs: generalized linear model (GLM), gradient boosted model (GBM), random forest (RF), and maximum entropy (MaxEnt), and compared their predictive performance to model geographically imbalanced-biased data of a rare species complex of mountain vipers. Models were tuned up based on a range of model-specific parameters considering two background selection methods: random and background weighting schemes. The performance of the fine-tuned models was assessed based on recently identified localities of the species. The results indicated that although the fine-tuned version of all models shows great performance in predicting training data (AUC > 0.9 and TSS > 0.5), they produce different results in classifying out-of-bag data. The GBM and RF with higher sensitivity of training data showed more different performances. The GLM, despite having high predictive performance for test data, showed lower specificity. It was only the MaxEnt model that showed high predictive performance and comparable results for identifying test data in both random and background weighting procedures. Our results highlight that while GBM and RF are prone to overfitting training data and GLM over-predict nonsampled areas MaxEnt is capable of producing results that are both predictable (extrapolative) and complex (interpolative). We discuss the assumptions of each model and conclude that MaxEnt could be considered as a practical method to cope with imbalanced-biased data in species distribution modeling approaches.

Using a gradient boosted model for case ascertainment from free-text veterinary records

Case ascertainment for prevalence and incidence studies from veterinary clinical data poses a major challenge because medical notes are not consistently structured or complete. Using natural language processing (NLP) and machine learning, this study aimed to obtain accurate case recognition for feline upper respiratory tract infections (primarily caused by viruses such as feline herpes virus (FHV-1) and feline calici virus (FCV), and bacteria such as Chlamydophila felis, Mycoplasma felis and Bordetella bronchiseptica using retrospective electronic veterinary records from the Royal Society for Prevention of Cruelty to Animals, Queensland (RSPCA Qld). Data cleaning and NLP on eight years of free-text veterinary records from RSPCA Queensland was carried out to derive text-based predictors. The NLP steps included sorting records by length of stay, vectorising, tokenising and spell checking against a bespoke veterinary database. A gradient boosted model (GBM) was trained to predict the probability of each animal having a diagnosis of upper respiratory infection. A manually annotated dataset was used for training the algorithm to learn dominant patterns between predictors (frequencies of n-grams) and responses (manual binary case classification). The GBM’s performance was tested against an out of sample validation dataset, and model agnostics were used to interrogate the model’s learning process. The GBM used patient-level frequencies of 1250 unique n-grams as predictor variables and was able to predict the probability of cases in the validation dataset with an accuracy of 0.95 (95% CI 0.92, 0.97) and F1 score of 0.96. Predictors that exerted the highest influence on the model included frequencies of “doxycycline”, “flu”, “sneezing”, “doxybrom” and “ocular”. The trained GBM was deployed on the full dataset spanning eight years, comprising 60,258 clinical entries. The prevalence in the full dataset was predicted to be 23.59%, which is in line with domain expertise from practicing veterinarians at the shelter. Case ascertainment is a crucial step for further epidemiological study of cat flu. Ultimately, this tool can be extended to other clinical procedures, conditions, and diseases such as intensive care treatment due to snake bites and tick paralysis, physical injuries such as orthopaedic fractures or chest injuries and labour-intensive infectious diseases like parvovirus, canine cough, and ringworm, all of which require prolonged quarantine and care.

Abstract 15358: Comparative Performance of Recurrent Heart Failure Prediction Models: Incorporating Social Determinants and Applying Machine Learning

Introduction: Improving performance of predictive algorithms for 30-d readmission following heart failure (HF) hospitalization may enhance clinical management decisions. Applying novel data (social determinants of health; SDOH) and novel methods (machine learning) have been proposed. We compared performance to predict 30-d HF readmission using models with and without SDOH and under traditional and machine learning techniques. Hypothesis and Methods: We used data on 27,071 admissions within Emory Healthcare (mean age: 66 , 48% female, 51% black). We categorized 69 covariates as patient-level clinical (e.g., diastolic blood pressure), patient-level demographic (e.g. gender, age), and zip code SDOH (e.g. percent unemployed). Logistic regression under generalized linear model (GLM) and gradient boosted model (GBM) frameworks were applied to test all combinations of covariate sets. We evaluated performance using 5-fold cross-validation calculating area under the curve (AUC; discrimination) and mean absolute error (MAE; calibration) statistics. Results: 30-d readmission occurred in 17.7% of hospital admissions. Under GLM, model performance was similar across all covariate sets (AUCs .500 - .502). Under GBM, model performance measures were higher than GLM but similar across covariate sets (.500-.664). Notably, SDOH improved model discrimination (AUC .606 clinical only, AUC .664 clinical+SDOH) in the full dataset (“in-sample”) runs. Conclusions: Patient clinical data provided the richest information for prediction of HF readmission, with incremental gains through the addition of area SDOH data such as zip code level unemployment. Furthermore, these gains were only realized in GBM highlighting the marginal utility in flexible modeling methods. While non-clinical data sources can provide a rich understanding of the social mechanisms producing disease, their utility for clinical prediction may be outcome-specific.

MACHINE LEARNING PREDICTION OF IN-HOSPITAL DISEASE PROGRESSION IN COMMUNITY-ACQUIRED PNEUMONIA: DERIVATION AND VALIDATION OF CLINICAL DETERIORATION IN PNEUMONIA (CDIP)

MACHINE LEARNING PREDICTION OF IN-HOSPITAL DISEASE PROGRESSION IN COMMUNITY-ACQUIRED PNEUMONIA: DERIVATION AND VALIDATION OF CLINICAL DETERIORATION IN PNEUMONIA (CDIP)

Abstract 1428: DNA methylation predicts early onset of primary tumor in patients with Li-Fraumeni syndrome

Abstract Background: Li-Fraumeni syndrome (LFS) is an autosomal dominant cancer predisposition syndrome. Approximately 80% of individuals with LFS harbor a germline TP53 pathogenic variant rendering them susceptible to a wide spectrum of early-onset malignancies. A comprehensive surveillance regimen termed the ‘Toronto Protocol’, has recently been adopted for early tumor detection, demonstrating significant improvement in survival among TP53 pathogenic variant carriers. However, the protocol’s “one-size-fits-all” approach fails to consider an individual patient's risk of cancer. We built a machine learning model that predicts early-onset of primary tumors in LFS by estimating the probability of cancer onset before the age of six years, leveraging patient peripheral blood leukocyte methylation profiles. Methods: We built a gradient-boosted tree model to predict the probability of cancer onset before the age of six using methylation data from 288 TP53 pathogenic variant carriers. An external test set of 82 TP53 pathogenic variant carriers was used to validate our model. To increase the signal-to-noise ratio, methylation probes associated with TP53 status were retained and probes associated with aging were removed. In our study, we were primarily interested in minimizing the false negative rate (i.e. to reduce the number of patients who developed cancer before the age of six but were undetected by our algorithm. Findings: We correctly predicted whether the first tumor will occur before the age of six with an accuracy of 79% in our external test set. Importantly, our model classified 90% of the patients that developed cancer prior to the age of six correctly. In addition, 81% of the individuals without cancer in the external test set were predicted correctly. Interpretation: Our tool provides additional value to clinicians in stratifying patients into low- or high-risk groups of developing early-onset malignancies, and helps inform rational use of clinical surveillance tools for early cancer detection, with the ultimate aim to improve overall patient outcomes. Citation Format: Vallijah Subasri, Benjamin Brew, Lauren Erdman, Tanya Guha, Jordan R. Hansford, Elizabeth Cairney, Carol Portwine, Christine Elser, Jonathan L. Finlay, Kim E. Nichols, Wendy Kohlmann, Noa Alon, Ana Novokmet, Ledia Brunga, Anita Villani, Kelvin C. de Andrade, Payal P. Khincha, Sharon A. Savage, Joshua D. Schiffman, David Malkin, Anna Goldenberg. DNA methylation predicts early onset of primary tumor in patients with Li-Fraumeni syndrome [abstract]. In: Proceedings of the American Association for Cancer Research Annual Meeting 2022; 2022 Apr 8-13. Philadelphia (PA): AACR; Cancer Res 2022;82(12_Suppl):Abstract nr 1428.

Who will respond to intensive PTSD treatment? A machine learning approach to predicting response prior to starting treatment

Despite the established effectiveness of evidence-based PTSD treatments, not everyone responds the same. Specifically, some individuals respond early while others respond minimally throughout treatment. Our ability to predict these trajectories at baseline has been limited. Predicting which individuals will respond to a certain type of treatment can significantly reduce short- and long-term costs and increase the ability to preemptively match individuals with treatments to which they are most likely to respond. In the present study, we examined whether veterans’ responses to a 3-week Cognitive Processing Therapy-based intensive PTSD treatment program could be accurately predicted prior to the first session. Using a sample of 432 veterans, and a wide range of demographic and clinical data collected during intake, we assessed six machine learning and statistical methods and their ability to predict fast and minimal responders prior to treatment initiation. For fast response classification, gradient boosted models (GBM) had the highest AUC-PR (0.466). For minimal response classification, elastic net (EN) had the highest mean CV AUC-PR (0.628). Using the best performing classifiers, we were able to predict both fast and minimal responders prior to starting treatment with relatively high AUC-ROC of 0.765 (GBM) and 0.826 (EN), respectively. These results may inform treatment modifications, although the accuracy may not be sufficient for clinicians to base inclusion/exclusion decisions entirely on the classifiers. Future research should evaluate whether these classifiers can be expanded to predict to which treatment type(s) an individual is most likely to respond based on various clinical, circumstantial, and biological features.

The relationship between transthoracic echocardiography and mortality in adult patients with multiple organ dysfunction syndrome: analysis of the MIMIC-III database.

BackgroundFor patients with multiple organ dysfunction syndrome (MODS), timely assessment of the condition and real-time adjustment of the treatment plan are of critical importance. To this end, transthoracic echocardiography (TTE) is widely used in clinical practice, but whether TTE can improve the short-term prognosis of MODS patients is currently unclear.MethodsWe extracted data of patients from the Medical Information Mart for Intensive Care III (MIMIC-III) database and included cases according to inclusion and exclusion criteria. The primary endpoint was the mortality within 28 days after admission to the intensive care unit (ICU), and multi-variate regression analysis was conducted to infer the risk factors associated death within 28 days after ICU admission. Double robust analysis was conducted to investigate the correlation between TTE and the endpoint. A gradient-boosted model (GBM) was constructed to calculate the propensity score (PS) of patients received TTE, so as to reduce the difference of variates between these two groups.ResultsA total of 13,844 MODS cases were included and were divided into a TTE group and a non-TTE group. There were 5,022 cases (36.28%) in the TTE group, 2,416 (48.10%) of whom were female; and 8,822 (63.72%) in the non-TTE group, 4,129 of whom (46.80%) were female. The covariates that contributed significantly to PS included chronic heart failure (CHF), creatine kinase (CK), troponin, partial pressure of oxygen (PO2), and sequential organ failure assessment (SOFA) scores. Double robust analysis showed that within 28 days after ICU admission, the TTE group had lower risk of death when compared with the non-TTE group. The adjusted odds ratio (OR) value of TTE for 28 days death was 0.73 [95% confidence interval (CI): 0.65–0.82; P<0.001]. The other 3 models had similar results, suggesting that conduct TTE for patients with MODS in ICU was associated with lower risk of 28 days mortality.ConclusionsTTE can reduce the mortality of MODS patients in the ICU.

The effects of midazolam or propofol plus fentanyl on ICU mortality: a retrospective study based on the MIMIC-IV database.

BackgroundPatients in the intensive care unit (ICU) are often under stress and fail to cooperate well with invasive treatments. Analgesia and sedation are of great significance for reducing the suffering of patients and ensuring the application and effectiveness of treatment. For better clinical choice, we aimed to explore the effect of the combination of propofol + fentanyl or midazolam + fentanyl on the short-term prognosis of hospitalized patients in the ICU.MethodsAccording to the inclusion and exclusion criteria, we retrospectively included patients in the MIMIC-IV database receiving midazolam + fentanyl or propofol + fentanyl analgesic and sedative treatment using Structured Query Language (SQL) to extract clinical data from the MIMIC-IV database. The primary endpoint was the death rate within 28 days after the patient was admitted to the ICU. Doubly robust estimation was used to infer the relationship between sedation and analgesia and 28 days outcome. The gradient boosted model (GBM) was used to estimate the propensity score (PS) of the patient’s sedation and analgesia program, PS was used as the weight, and the inverse probabilities weighting (IPW) model was used to generate a weighted cohort.ResultsIn total, 4,188 cases were included, with 2,174 (51.9%) in the propofol group and 2,014 (48.1%) in the midazolam group. In the PS score matching cohort, the 28-day mortality of patients in the midazolam group was 30.8%, and the 28-day mortality of patients in the propofol group was 25.5%. The adjusted odds ratio (OR) value was 1.421 [95% confidence interval (CI): 1.118–1.806, P<0.001]. Patients in the propofol group did not use vasoactive drugs for a longer period of time than the midazolam group, and patients in the propofol group received significantly more fluids than those in the midazolam group in the first three days after admission to the ICU.ConclusionsCompared with midazolam combined with fentanyl, propofol combined with fentanyl for sedation and analgesia can reduce the risk of short-term death in ICU patients.

Machine Learning Simulation of Land Cover Impact on Surface Urban Heat Island Surrounding Park Areas

The urban heat island effect has been studied extensively by many researchers around the world with the process of urbanization coming about as one of the major culprits of the increasing urban land surface temperatures. Over the past 20 years, the city of Dallas, Texas, has consistently been one of the fastest growing cities in the United States and has faced rapid urbanization and great amounts of urban sprawl, leading to an increase in built-up surface area. In this study, we utilize Landsat 8 satellite images, Geographic Information System (GIS) technologies, land use/land cover (LULC) data, and a state-of-the-art methodology combining machine learning algorithms (eXtreme Gradient Boosted models, or XGBoost) and a modern game theoretic-based approach (Shapley Additive exPlanation, or SHAP values) to investigate how different land use/land cover classifications impact the land surface temperature and park cooling effects in the city of Dallas. We conclude that green spaces, residential, and commercial/office spaces have the largest impacts on Land Surface Temperatures (LST) as well as the Park’s Cooling Intensity (PCI). Additionally, we have found that the extent and direction of influence of these categories depends heavily on the surrounding area. By using SHAP values we can describe these interactions in greater detail than previous studies. These results will provide an important reference for future urban and park placement planning to minimize the urban heat island effect, especially in sprawling cities.

Machine learning approaches improve risk stratification for secondary cardiovascular disease prevention in multiethnic patients

ObjectivesIdentifying high-risk patients is crucial for effective cardiovascular disease (CVD) prevention. It is not known whether electronic health record (EHR)-based machine-learning (ML) models can improve CVD risk stratification compared with...

Prediction of Major Adverse Events After Endovascular Aneurysm Repair Using a Machine Learning Model

Predicting major adverse events (MAE), including type I and III endoleaks (ELs), after endovascular aneurysm repair (EVAR) is essential for clinical decision-making. No clinical decision instruments can reliably predict for MAE after EVAR. We developed a machine learning model (MLM) for predicting MAE within 3 years after elective EVAR using preoperative demographic information and aneurysm geometry. MLMs infer complex interaction without prespecification, possibly increasing the prediction accuracy compared with logistic regression models (LRMs). Patients with intact infrarenal abdominal aortic aneurysms who had undergone EVAR with preoperative M2S reconstruction from 2010 to 2017 at a single institution were analyzed. The MAEs included aneurysm- or intervention-related death within 30 days, aneurysm rupture, the development of type I or III ELs, greater than three reinterventions, or EVAR explantation. Patients with <3 years of follow-up were excluded. Seventy-four preoperative predictors were evaluated. Of these, 30 were geometric factors derived from M2S models. The remaining factors were demographic. The performance of LRMs, random forest models (RFMs), and gradient-boosted models (GBMs) were evaluated using a 75% and 25% training and testing split, fivefold cross-validation, and receiver operating characteristic performance metric. A total of 250 patients were included (mean age, 71.9 ± 8.6 years; 86% male). The abdominal aortic aneurysm diameter was 56.0 ± 10.6 mm, and the infrarenal neck length was 24.6 ± 4.5 mm. Of the 250 patients, 40 (16%) had experienced MAE: 11 had died within 30 days, 3 had experienced aneurysm rupture, 10 had developed a type I EL, 10 had developed a type III EL, 10 had required graft explantation, and 4 had undergone four or more reinterventions. The most significant 45 predictive variables are shown in the Fig. All discrimination and calibration metrics favored the MLM over the LRM. The area under the curve for the LRM was 0.70 compared with 0.83 for the RFM and 0.77 for the GBM. The LRM classification accuracy was 68% compared with 84% for the RFM and GBM. The GBM showed superior calibration performance (Hosmer-Lemeshow P = .201). The removal of the geometric parameters significantly degraded the performance for all the models (likelihood ratio P < .001). The MLM displayed superior discrimination and calibration compared with the LRM in predicting MAE for the first 3 years after elective EVAR, possibly owing to the complex interaction terms in the predictive factors. The anatomic characteristics were more influential than the demographic data in predicting for MAE. Our MLM might predict for MAE after EVAR and thereby aid in selecting patients for EVAR, nonoperative observation, or open repair.