Swin-GA-RF: genetic algorithm-based Swin Transformer and random forest for enhancing cervical cancer classification.

Cervical cancer is a prevalent and concerning disease affecting women, with increasing incidence and mortality rates. Early detection plays a crucial role in improving outcomes. Recent advancements in computer vision, particularly the Swin transformer, have shown promising performance in image classification tasks, rivaling or surpassing traditional convolutional neural networks (CNNs). The Swin transformer adopts a hierarchical and efficient approach using shifted windows, enabling the capture of both local and global contextual information in images. In this paper, we propose a novel approach called Swin-GA-RF to enhance the classification performance of cervical cells in Pap smear images. Swin-GA-RF combines the strengths of the Swin transformer, genetic algorithm (GA) feature selection, and the replacement of the softmax layer with a random forest classifier. Our methodology involves extracting feature representations from the Swin transformer, utilizing GA to identify the optimal feature set, and employing random forest as the classification model. Additionally, data augmentation techniques are applied to augment the diversity and quantity of the SIPaKMeD1 cervical cancer image dataset. We compare the performance of the Swin-GA-RF Transformer with pre-trained CNN models using two classes and five classes of cervical cancer classification, employing both Adam and SGD optimizers. The experimental results demonstrate that Swin-GA-RF outperforms other Swin transformers and pre-trained CNN models. When utilizing the Adam optimizer, Swin-GA-RF achieves the highest performance in both binary and five-class classification tasks. Specifically, for binary classification, it achieves an accuracy, precision, recall, and F1-score of 99.012, 99.015, 99.012, and 99.011, respectively. In the five-class classification, it achieves an accuracy, precision, recall, and F1-score of 98.808, 98.812, 98.808, and 98.808, respectively. These results underscore the effectiveness of the Swin-GA-RF approach in cervical cancer classification, demonstrating its potential as a valuable tool for early diagnosis and screening programs.

Cardiac disease risk prediction using machine learning algorithms.

Heart attack is a life-threatening condition which is mostly caused due to coronary disease resulting in death in human beings. Detecting the risk of heart diseases is one of the most important problems in medical science that can be prevented and treated with early detection and appropriate medical management; it can also help to predict a large number of medical needs and reduce expenses for treatment. Predicting the occurrence of heart diseases by machine learning (ML) algorithms has become significant work in healthcare industry. This study aims to create a such system that is used for predicting whether a patient is likely to develop heart attacks, by analysing various data sources including electronic health records and clinical diagnosis reports from hospital clinics. ML is used as a process in which computers learn from data in order to make predictions about new datasets. The algorithms created for predictive data analysis are often used for commercial purposes. This paper presents an overview to forecast the likelihood of a heart attack for which many ML methodologies and techniques are applied. In order to improve medical diagnosis, the paper compares various algorithms such as Random Forest, Regression models, K-nearest neighbour imputation (KNN), Naïve Bayes algorithm etc. It is found that the Random Forest algorithm provides a better accuracy of 88.52% in forecasting heart attack risk, which could herald a revolution in the diagnosis and treatment of cardiovascular illnesses.

Improving the academic resilience of hospital nursing interns through a hybrid multi-criteria decision analysis model.

Purpose: To identify the main variables affecting the academic adaptability of hospital nursing interns and key areas for improvement in preparing for future unpredictable epidemics. Methods: The importance of academic resilience-related variables for all nursing interns was analyzed using the random forest method, and key variables were further identified. An importance-performance analysis was used to identify the key improvement gaps regarding the academic resilience of nursing interns in the case hospital. Results: The random forest showed that five items related to cooperation, motivation, confidence, communication, and difficulty with coping were the main variables impacting the academic resilience of nursing interns. Moreover, the importance-performance analysis revealed that three items regarding options examination, communication, and confidence were the key improvement areas for participating nursing interns in the case hospital. Conclusions: For the prevention and control of future unpredictable pandemics, hospital nursing departments can strengthen the link between interns, nurses, and physicians and promote their cooperation and communication during clinical practice. At the same time, an application can be created considering the results of this study and combined with machine learning methods for more in-depth research. These will improve the academic resilience of nursing interns during the routine management of pandemics within hospitals.

Machine learning based on biological context facilitates the identification of microvascular invasion in intrahepatic cholangiocarcinoma.

Intrahepatic cholangiocarcinoma (ICC) is a rare disease associated with a poor prognosis, primarily due to early recurrence and metastasis. An important feature of this condition is microvascular invasion (MVI). However, current predictive models based on imaging have limited efficacy in this regard. This study employed a random forest model to construct a predictive model for MVI identification and uncover its biological basis. Single-cell transcriptome sequencing, whole exome sequencing, and proteome sequencing were performed. The area under the curve of the prediction model in the validation set was 0.93. Further analysis indicated that MVI-associated tumor cells exhibited functional changes related to epithelial-mesenchymal transition and lipid metabolism due to alterations in the NF-kappa B and MAPK signaling pathways. Tumor cells were also differentially enriched for the IL-17 signaling pathway. There was less infiltration of SLC30A1+ CD8+ T cells expressing cytotoxic genes in MVI-associated ICC, whereas there was more infiltration of myeloid cells with attenuated expression of the MHC II pathway. Additionally, MVI-associated intercellular communication was closely related to the SPP1-CD44 and ANXA1-FPR1 pathways. These findings resulted in a brilliant predictive model and fresh insights into MVI.

Random forest model used to predict the medical out-of-pocket costs of hypertensive patients.

Background: Precise prediction of out-of-pocket (OOP) costs to improve health policy design is important for governments of countries with national health insurance. Controlling the medical expenses for hypertension, one of the leading causes of stroke and ischemic heart disease, is an important issue for the Japanese government. This study aims to explore the importance of OOP costs for outpatients with hypertension. Methods: To obtain a precise prediction of the highest quartile group of OOP costs of hypertensive outpatients, we used nationwide longitudinal data, and estimated a random forest (RF) model focusing on complications with other lifestyle-related diseases and the nonlinearities of the data. Results: The results of the RF models showed that the prediction accuracy of OOP costs for hypertensive patients without activities of daily living (ADL) difficulties was slightly better than that for all hypertensive patients who continued physician visits during the past two consecutive years. Important variables of the highest quartile of OOP costs were age, diabetes or lipidemia, lack of habitual exercise, and moderate or vigorous regular exercise. Conclusion: As preventing complications of diabetes or lipidemia is important for reducing OOP costs in outpatients with hypertension, regular exercise of moderate or vigorous intensity is recommended for hypertensive patients that do not have ADL difficulty. For hypertensive patients with ADL difficulty, habitual exercise is not recommended.

GATIS score for predicting the prognosis of rectal neuroendocrine neoplasms: A Chinese multicenter study of 12-year experience.

BACKGROUND: There is currently a shortage of accurate, efficient, and precise predictive instruments for rectal neuroendocrine neoplasms (NENs). AIM: To develop a predictive model for individuals with rectal NENs (R-NENs) using data from a large cohort. METHODS: Data from patients with primary R-NENs were retrospectively collected from 17 large-scale referral medical centers in China. Random forest and Cox proportional hazard models were used to identify the risk factors for overall survival and progression-free survival, and two nomograms were constructed. RESULTS: A total of 1408 patients with R-NENs were included. Tumor grade, T stage, tumor size, age, and a prognostic nutritional index were important risk factors for prognosis. The GATIS score was calculated based on these five indicators. For overall survival prediction, the respective C-indexes in the training set were 0.915 (95% confidence interval: 0.866-0.964) for overall survival prediction and 0.908 (95% confidence interval: 0.872-0.944) for progression-free survival prediction. According to decision curve analysis, net benefit of the GATIS score was higher than that of a single factor. The time-dependent area under the receiver operating characteristic curve showed that the predictive power of the GATIS score was higher than that of the TNM stage and pathological grade at all time periods. CONCLUSION: The GATIS score had a good predictive effect on the prognosis of patients with R-NENs, with efficacy superior to that of the World Health Organization grade and TNM stage.

Modelling the rapid detection of Carbapenemase-resistant Klebsiella pneumoniae based on machine learning and matrix-assisted laser desorption/ionization time-of-flight mass spectrometry.

In this study, 80 carbapenem-resistant Klebsiella pneumoniae (CR-KP) and 160 carbapenem-susceptible Klebsiella pneumoniae (CS-KP) strains detected in the clinic were selected and their matrix-assisted laser desorption/ionization time-of-flight mass spectrometry (MALDI-TOF MS) peaks were collected. K-means clustering was performed on the MS peak data to obtain the best "feature peaks", and four different machine learning models were built to compare the area under the ROC curve, specificity, sensitivity, test set score, and ten-fold cross-validation score of the models. By adjusting the model parameters, the test efficacy of the model is increased on the basis of reducing model overfitting. The area under the ROC curve of the Random Forest, Support Vector Machine, Logistic Regression, and Xgboost models used in this study are 0.99, 0.97, 0.96, and 0.97, respectively; the model scores on the test set are 0.94, 0.91, 0.90, and 0.93, respectively; and the results of the ten-fold cross-validation are 0.84, 0.81, 0.81, and 0.85, respectively. Based on the machine learning algorithms and MALDI-TOF MS assay data can realize rapid detection of CR-KP, shorten the in-laboratory reporting time, and provide fast and reliable identification results of CR-KP and CS-KP.

Enhancing flood mapping through ensemble machine learning in the Gamasyab watershed, Western Iran.

Floods are among the natural hazards that have seen a rapid increase in frequency in recent decades. The damage caused by floods, including human and financial losses, poses a serious threat to human life. This study evaluates two machine learning (ML) techniques for flood susceptibility mapping (FSM) in the Gamasyab watershed in Iran. We utilized random forest (RF), support vector machine (SVM), ensemble models, and a geographic information system (GIS) to predict FSM. The application of these models involved 10 effective factors in flooding, as well as 82 flood locations integrated into the GIS. The SVM and RF models were trained and tested, followed by the implementation of resampling techniques (RT) using bootstrap and subsampling methods in three repetitions. The results highlighted the importance of elevation, slope, and precipitation as primary factors influencing flood occurrence. Additionally, the ensemble model outperformed both the RF and SVM models, achieving an area under the curve (AUC) of 0.9, a correlation coefficient (COR) of 0.79, a true skill statistic (TSS) of 0.83, and a standard deviation (SD) of 0.71 in the test phase. The tested models were adapted to available input data to map the FSM across the study watershed. These findings underscore the potential of integrating an ensemble model with GIS as an effective tool for flood susceptibility mapping.

Improved mapping of heavy metals in agricultural soils using machine learning augmented with spatial regionalization indices.

The accurate spatial mapping of heavy metal levels in agricultural soils is crucial for environmental management and food security. However, the inherent limitations of traditional interpolation methods and emerging machine-learning techniques restrict their spatial prediction accuracy. This study aimed to refine the spatial prediction of heavy metal distributions in Guangxi, China, by integrating machine learning models and spatial regionalization indices (SRIs). The results demonstrated that random forest (RF) models incorporating SRIs outperformed artificial neural network and support vector regression models, achieving R2 values exceeding 0.96 for eight heavy metals on the test data. Hierarchical clustering for feature selection further improved the model performance. The optimized RF models accurately predicted the heavy metal distributions in agricultural soils across the province, revealing higher levels in the central-western regions and lower levels in the north and south. Notably, the models identified that 25.78 % of agricultural soils constitute hotspots with multiple co-occurring heavy metals, and over 6.41 million people are exposed to excessive soil heavy metal levels. Our findings provide valuable insights for the development of targeted strategies for soil pollution control and agricultural soil management to safeguard food security and public health.

Comparison between Three Radiomics Models and Clinical Nomograms for Prediction of Lymph Node Involvement in PCa Patients Combining Clinical and Radiomic Features.

PURPOSE: We aim to compare the performance of three different radiomics models (logistic regression (LR), random forest (RF), and support vector machine (SVM)) and clinical nomograms (Briganti, MSKCC, Yale, and Roach) for predicting lymph node involvement (LNI) in prostate cancer (PCa) patients. MATERIALS AND METHODS: The retrospective study includes 95 patients who underwent mp-MRI and radical prostatectomy for PCa with pelvic lymphadenectomy. Imaging data (intensity in T2, DWI, ADC, and PIRADS), clinical data (age and pre-MRI PSA), histological data (Gleason score, TNM staging, histological type, capsule invasion, seminal vesicle invasion, and neurovascular bundle involvement), and clinical nomograms (Yale, Roach, MSKCC, and Briganti) were collected for each patient. Manual segmentation of the index lesions was performed for each patient using an open-source program (3D SLICER). Radiomic features were extracted for each segmentation using the Pyradiomics library for each sequence (T2, DWI, and ADC). The features were then selected and used to train and test three different radiomics models (LR, RF, and SVM) independently using ChatGPT software (v 4o). The coefficient value of each feature was calculated (significant value for coefficient ≥ ±0.5). The predictive performance of the radiomics models and clinical nomograms was assessed using accuracy and area under the curve (AUC) (significant value for p ≤ 0.05). Thus, the diagnostic accuracy between the radiomics and clinical models were compared. RESULTS: This study identified 343 features per patient (330 radiomics features and 13 clinical features). The most significant features were T2_nodulofirstordervariance and T2_nodulofirstorderkurtosis. The highest predictive performance was achieved by the RF model with DWI (accuracy 86%, AUC 0.89) and ADC (accuracy 89%, AUC 0.67). Clinical nomograms demonstrated satisfactory but lower predictive performance compared to the RF model in the DWI sequences. CONCLUSIONS: Among the prediction models developed using integrated data (radiomics and semantics), RF shows slightly higher diagnostic accuracy in terms of AUC compared to clinical nomograms in PCa lymph node involvement prediction.

Application of Spatial Offset Raman Spectroscopy (SORS) and Machine Learning for Sugar Syrup Adulteration Detection in UK Honey.

Honey authentication is a complex process which traditionally requires costly and time-consuming analytical techniques not readily available to the producers. This study aimed to develop non-invasive sensor methods coupled with a multivariate data analysis to detect the type and percentage of exogenous sugar adulteration in UK honeys. Through-container spatial offset Raman spectroscopy (SORS) was employed on 17 different types of natural honeys produced in the UK over a season. These samples were then spiked with rice and sugar beet syrups at the levels of 10%, 20%, 30%, and 50% w/w. The data acquired were used to construct prediction models for 14 types of honey with similar Raman fingerprints using different algorithms, namely PLS-DA, XGBoost, and Random Forest, with the aim to detect the level of adulteration per type of sugar syrup. The best-performing algorithm for classification was Random Forest, with only 1% of the pure honeys misclassified as adulterated and <3.5% of adulterated honey samples misclassified as pure. Random Forest was further employed to create a classification model which successfully classified samples according to the type of adulterant (rice or sugar beet) and the adulteration level. In addition, SORS spectra were collected from 27 samples of heather honey (24 Calluna vulgaris and 3 Erica cinerea) produced in the UK and corresponding subsamples spiked with high fructose sugar cane syrup, and an exploratory data analysis with PCA and a classification with Random Forest were performed, both showing clear separation between the pure and adulterated samples at medium (40%) and high (60%) adulteration levels and a 90% success at low adulteration levels (20%). The results of this study demonstrate the potential of SORS in combination with machine learning to be applied for the authentication of honey samples and the detection of exogenous sugars in the form of sugar syrups. A major advantage of the SORS technique is that it is a rapid, non-invasive method deployable in the field with potential application at all stages of the supply chain.

L Test Subtask Segmentation for Lower-Limb Amputees Using a Random Forest Algorithm.

Functional mobility tests, such as the L test of functional mobility, are recommended to provide clinicians with information regarding the mobility progress of lower-limb amputees. Smartphone inertial sensors have been used to perform subtask segmentation on functional mobility tests, providing further clinically useful measures such as fall risk. However, L test subtask segmentation rule-based algorithms developed for able-bodied individuals have not produced sufficiently acceptable results when tested with lower-limb amputee data. In this paper, a random forest machine learning model was trained to segment subtasks of the L test for application to lower-limb amputees. The model was trained with 105 trials completed by able-bodied participants and 25 trials completed by lower-limb amputee participants and tested using a leave-one-out method with lower-limb amputees. This algorithm successfully classified subtasks within a one-foot strike for most lower-limb amputee participants. The algorithm produced acceptable results to enhance clinician understanding of a person's mobility status (>85% accuracy, >75% sensitivity, >95% specificity).

Soil organic carbon formation efficiency from straw/stover and manure input and its drivers: Estimates from long-term data in global croplands.

New soil organic carbon (SOC) formation in cropland from straw/stover or manure input is a vital source of SOC for climate change mitigation. However, location and variations in the efficiency, specifically the ratio of new SOC formation to organic C input (NCE), remain unquantified globally. In this study, the spatial variability of cropland NCE from straw/stover or manure input and explanatory factors were determined by analyzing 897 pairs of long-term field measurements from 404 globally distributed sites and by mapping grid-level cropland NCEs. The global NCE for paddy and upland averaged 13.8% (8.7%-25.1%, 5th-95th percentile) and 10.9% (6.8%-17.3%), respectively. The initial SOC and the clay content of soil, rather than temperature, were the most important factors regulating NCE. A parabola with an apex at approximately 17 g kg-1 between the initial SOC and NCE was resolved, and a positive correlation between soil clay content and NCE was observed. High-resolution mapping of the global NCE derived from manure/straw and insight into NCE dynamics provide a benchmark for diagnosing cropland soil C dynamics under climate change and identifying priority regions and actions for C management.

Texture-based speciation of otitis media-related bacterial biofilms from optical coherence tomography images using supervised classification.

Otitis media (OM), a highly prevalent inflammatory middle-ear disease in children worldwide, is commonly caused by an infection, and can lead to antibiotic-resistant bacterial biofilms in recurrent/chronic OM cases. A biofilm related to OM typically contains one or multiple bacterial species. OCT has been used clinically to visualize the presence of bacterial biofilms in the middle ear. This study used OCT to compare microstructural image texture features from bacterial biofilms. The proposed method applied supervised machine-learning-based frameworks (SVM, random forest, and XGBoost) to classify multiple species bacterial biofilms from in vitro cultures and clinically-obtained in vivo images from human subjects. Our findings show that optimized SVM-RBF and XGBoost classifiers achieved more than 95% of AUC, detecting each biofilm class. These results demonstrate the potential for differentiating OM-causing bacterial biofilms through texture analysis of OCT images and a machine-learning framework, offering valuable insights for real-time in vivo characterization of ear infections.

A novel coupling interpretable machine learning framework for water quality prediction and environmental effect understanding in different flow discharge regulations of hydro-projects.

Machine learning models (MLMs) have been increasingly used to forecast water pollution. However, the "black box" characteristic for understanding mechanism processes still limits the applicability of MLMs for water quality management in hydro-projects under complex and frequently artificial regulation. This study proposes an interpretable machine learning framework for water quality prediction coupled with a hydrodynamic (flow discharge) scenario-based Random Forest (RF) model with multiple model-agnostic techniques and quantifies global, local, and joint interpretations (i.e., partial dependence, individual conditional expectation, and accumulated local effects) of environmental factor implications. The framework was applied and verified to predict the permanganate index (CODMn) under different flow discharge regulation scenarios in the Middle Route of the South-to-North Water Diversion Project of China (MRSNWDPC). A total of 4664 sampling cases data matrices, including water quality, meteorological, and hydrological indicators from eight national stations along the main canal of the MRSNWDPC, were collected from May 2019 to December 2020. The results showed that the RF models were effective in forecasting CODMn in all flow discharge scenarios, with a mean square error, coefficient of determination, and mean absolute error of 0.006-0.026, 0.481-0.792, and 0.069-0.104, respectively, in the testing dataset. A global interpretation indicated that dissolved oxygen, flow discharge, and surface pressure are the three most important variables of CODMn. Local and joint interpretations indicated that the RF-based prediction model provides a basic understanding of the physical mechanisms of environmental systems. The proposed framework can effectively learn the fundamental environmental implications of water quality variations and provide reliable prediction performance, highlighting the importance of model interpretability for trustworthy machine learning applications in water management projects. This study provides scientific references for applying advanced data-driven MLMs to water quality forecasting and a reliable methodological framework for water quality management and similar hydro-projects.

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%.

Machine learning-based corrosion rate prediction of steel embedded in soil.

Predicting the corrosion rate for soil-buried steel is significant for assessing the service-life performance of structures in soil environments. However, due to the large amount of variables involved, existing corrosion prediction models have limited accuracy for complex soil environment. The present study employs three machine learning (ML) algorithms, i.e., random forest, support vector regression, and multilayer perception, to predict the corrosion current density of soil-buried steel. Steel specimens were embedded in soil samples collected from different regions of the Wisconsin state. Variables including exposure time, moisture content, pH, electrical resistivity, chloride, sulfate content, and mean total organic carbon were measured through laboratory tests and were used as input variables for the model. The current density of steel was measured through polarization technique, and was employed as the output of the model. Of the various ML algorithms, the random forest (RF) model demonstrates the highest predictability (with an RMSE value of 0.01095 A/m2 and an R2 value of 0.987). In light of the feature selection method, the electrical resistivity is identified as the most significant feature. The combination of three features (resistivity, exposure time, and mean total organic carbon) is the optimal scenario for predicting the corrosion current density of soil-buried steel.

Local ecological knowledge and perception as a strategy in the management of ecosystem services.

In designing and implementing initiatives to conserve biodiversity and ensure the flow of ecosystem services, it is crucial to understand the perspectives of communities living near protected areas. Improving conservation efforts may depend on analyzing socio-ecological factors and their impact on Local Ecological Knowledge (LEK) and perceptions of ecosystem services. We employed participatory methodologies with 80 farmers from agrarian settlements adjacent to protected areas in the Cerrado biome, Brazil, we quantified LEK and assessed perceptions of ecosystem services using an adaptation of the Q-methodology. We collected data on thirteen socio-ecological variables, including age, gender, farm size, education, engagement with conservation initiatives, and interactions with protected areas and Legal Reserves. Using artificial intelligence in a Random Forest (RF) modelling approach, we identified the most influential variables on LEK and perceptions. Our findings demonstrate that engagement in nature conservation and restoration initiatives, along with the use of native areas (protected and managed areas) significantly influence LEK levels within the farmers' communities. Farmers with full participation, from conception to implementation and evaluation of the initiatives, had a significantly higher LEK level (28.5 ± 13.0) compared to farmers without participation in those initiatives (11.4 ± 5.9). Farmers who used the cerrado for leisure and education (28.2 ± 21.2) had significantly higher LEK levels compared to farmers who do not attend or use the cerrado areas (13.5 ± 8.9) and those using areas of native vegetation for cattle raising (12.8 ± 6.8). These results highlight that, in addition to farmers' participation in conservation and restoration initiatives, the sustainable use of natural areas is fundamental to strengthen their local knowledge of ecosystem functioning. Furthermore, we found that the type of agroecosystem present on farms strongly? shapes farmers' perceptions of ecosystem services. Farmers perceive different ecosystem services depending on land use, indicating the need for tailored interventions for the planning and management of conservation areas. Farmers practicing soybean monoculture had significantly lower perception scores on ecosystem services (-5.1 ± 3.8) than to the other four evaluated groups. Overall, the study highlights the critical role of incorporating local knowledge and perceptions for the design of effective management strategies to increase ecosystem services provision and biodiversity conservation in areas adjacent to protected areas.

Machine Learning Identification of Modifiable Predictors of Patient Outcomes After Transcatheter Aortic Valve Replacement.

Background: Transcatheter aortic valve replacement (TAVR) is an important treatment option for patients with severe symptomatic aortic stenosis. It is important to identify predictors of excellent outcomes (good clinical outcomes, more time spent at home) after TAVR that are potentially amenable to improvement. Objectives: The purpose of the study was to use machine learning to identify potentially modifiable predictors of clinically relevant patient-centered outcomes after TAVR. Methods: We used data from 8,332 TAVR cases (January 2016-December 2021) from 21 hospitals to train random forest models with 57 patient characteristics (demographics, comorbidities, surgical risk score, lab values, health status scores) and care process parameters to predict the end point, a composite of parameters that designated an excellent outcome and included no major complications (in-hospital or at 30 days), post-TAVR length of stay of 1 day or less, discharge to home, no readmission, and alive at 30 days. We used recursive feature elimination with cross-validation and Shapley Additive Explanation feature importance to identify parameters with the highest predictive values. Results: The final random forest model retained 29 predictors (15 patient characteristics and 14 care process components); the area under the curve, sensitivity, and specificity were 0.77, 0.67, and 0.73, respectively. Four potentially modifiable predictors with relatively high Shapley Additive Explanation values were identified: type of anesthesia, direct movement to stepdown unit post-TAVR, time between catheterization and TAVR, and preprocedural length of stay. Conclusions: This study identified four potentially modifiable predictors of excellent outcome after TAVR, suggesting that machine learning combined with hospital-level data can inform modifiable components of care, which could support better delivery of care for patients undergoing TAVR.

Comparing causal random forest and linear regression to estimate the independent association of organisational factors with ICU efficiency.

PURPOSE: Parametric regression models have been the main statistical method for identifying average treatment effects. Causal machine learning models showed promising results in estimating heterogeneous treatment effects in causal inference. Here we aimed to compare the application of causal random forest (CRF) and linear regression modelling (LRM) to estimate the effects of organisational factors on ICU efficiency. METHODS: A retrospective analysis of 277,459 patients admitted to 128 Brazilian and Uruguayan ICUs over three years. ICU efficiency was assessed using the average standardised efficiency ratio (ASER), measured as the average of the standardised mortality ratio (SMR) and the standardised resource use (SRU) according to the SAPS-3 score. Using a causal inference framework, we estimated and compared the conditional average treatment effect (CATE) of seven common structural and organisational factors on ICU efficiency using LRM with interaction terms and CRF. RESULTS: The hospital mortality was 14 %; median ICU and hospital lengths of stay were 2 and 7 days, respectively. Overall median SMR was 0.97 [IQR: 0.76,1.21], median SRU was 1.06 [IQR: 0.79,1.30] and median ASER was 0.99 [IQR: 0.82,1.21]. Both CRF and LRM showed that the average number of nurses per ten beds was independently associated with ICU efficiency (CATE [95 %CI]: -0.13 [-0.24, -0.01] and -0.09 [-0.17,-0.01], respectively). Finally, CRF identified some specific ICUs with a significant CATE in exposures that did not present a significant average effect. CONCLUSION: In general, both methods were comparable to identify organisational factors significantly associated with CATE on ICU efficiency. CRF however identified specific ICUs with significant effects, even when the average effect was nonsignificant. This can assist healthcare managers in further in-dept evaluation of process interventions to improve ICU efficiency.