Hospital Readmission and Length-of-Stay Prediction Using an Optimized Hybrid Deep Model

Future Internet, Journal Year: 2023, Volume and Issue: 15(9), P. 304 - 304

Published: Sept. 6, 2023

Hospital readmission and length-of-stay predictions provide information on how to manage hospital bed capacity the number of required staff, especially during pandemics. We present a hybrid deep model called Genetic Algorithm-Optimized Convolutional Neural Network (GAOCNN), with unique preprocessing method predict length stay for patients various conditions. GAOCNN uses one-dimensional convolutional layers stay. The parameters are optimized via genetic algorithm. To show performance proposed in conditions, we evaluate under three healthcare datasets: Diabetes 130-US hospitals dataset, COVID-19 MIMIC-III dataset. diabetes dataset has both stay, while datasets just include Experimental results that model’s accuracy was 97.2% diabetic patients. Furthermore, prediction 89%, 99.4%, 94.1% diabetic, COVID-19, ICU patients, respectively. These confirm superiority compared existing methods. Our findings offer platform managing funds resources diseases.

Language: Английский

---

Machine learning-based prediction of length of stay (LoS) in the neonatal intensive care unit using ensemble methods

Neural Computing and Applications, Journal Year: 2024, Volume and Issue: 36(23), P. 14433 - 14448

Published: May 7, 2024

Abstract Neonatal medical data holds critical information within the healthcare industry, and it is important to analyze this effectively. Machine learning algorithms offer powerful tools for extracting meaningful insights from of neonates improving treatment processes. Knowing length hospital stay in advance very managing resources, personnel, costs. Thus, study aims estimate infants treated Intensive Care Unit (NICU) using machine algorithms. Our conducted a two-class prediction long short-term lengths utilizing unique dataset. Adopting hybrid approach called Classifier Fusion-LoS, involved two stages. In initial stage, various classifiers were employed including classical models such as Logistic Regression, ExtraTrees, Random Forest, KNN, Support Vector Classifier, well ensemble like AdaBoost, GradientBoosting, XGBoost, CatBoost. Forest yielded highest validation accuracy at 0.94. subsequent Voting Classifier—an method—was applied, resulting increasing 0.96. method outperformed existing studies terms accuracy, both neonatal-specific other general research. While estimation offers into potential suitability incubators NICUs, which are not universally available every city, patient admission, plays pivotal role delineating protocols patients. Additionally, research provides crucial management planning beds, equipment,

Language: Английский

---

Examining Patients Length of Stay Estimation with Explainable Artificial Intelligence Methods

Published: Jan. 1, 2024

Language: Английский

---

Mechanical ventilation intervention based on machine learning from vital signs monitoring: a scoping review

Measurement Science and Technology, Journal Year: 2023, Volume and Issue: 34(6), P. 062001 - 062001

Published: March 3, 2023

Abstract Asynchronous breathing (AB) during mechanical ventilation (MV) may lead to a detrimental effect on the patient’s condition. Due massive amount of data displayed in large ICU, machine learning algorithm (MLA) was proposed extensively extract patterns within multiple continuous-in-time vital signs, determine which are variables that will predict AB, intervene MV as an early warning system, and finally replace highly demand clinician’s cognition. This study reviews MLA for prediction detection models from signs monitoring intervention. Publication development intervention based support clinicians’ decision-making process extracted three electronic academic research databases Web Science Core Collection (WoSCC), ScienceDirect, PUBMED Central February 2023. 838 papers extracted. There 14 review papers, while 25 related pass with quality assessments (QA). Few studies have been published considered VS along parameters waveforms Vital is not only predictor developed MLA. Most suggested developing direct requires more concern pre-processing real-time avoid false positive than itself.

Language: Английский

---

An IoT-Based CNN Model for Patients in ICU Beds During the COVID-19 Outburst

Published: July 28, 2023

Language: Английский

---

Machine learning: Predicting hospital length of stay in patients admitted for lupus flares

Lupus, Journal Year: 2023, Volume and Issue: 32(12), P. 1418 - 1429

Published: Oct. 1, 2023

Background Although rare, severe systemic lupus erythematosus (SLE) flares requiring hospitalization account for most of the direct costs SLE care. New machine learning (ML) methods may optimize care by predicting which patients will have a prolonged hospital length stay (LOS). Our study uses approach to predict LOS in admitted and assesses features prolong LOS. Methods sampled 5831 from National Inpatient Sample Database 2016–2018 collected 90 demographics comorbidity features. Four models were built (XGBoost, Linear Support Vector Machines, K Nearest Neighbors, Logistic Regression) LOS, their performance was evaluated using multiple metrics, including accuracy, receiver operator area under curve (ROC-AUC), precision-recall (PR- AUC), F1-score. Using highest-performing model (XGBoost), we assessed feature importance our input Shapley value explanations (SHAP) rank impact on Results XGB performed best with ROC-AUC 0.87, PR-AUC 0.61, an F1 score 0.56, accuracy 95%. The significant “the need central line,” “acute dialysis,” renal failure.” Other top include those related infectious comorbidities. Conclusion results consistent established literature showed promise ML over traditional predictive analyses, even rare rheumatic events such as flare hospitalizations.

Language: Английский

---

An efficient random forest algorithm-based telemonitoring framework to predict mortality and length of stay of patients in ICU

Multimedia Tools and Applications, Journal Year: 2023, Volume and Issue: 83(17), P. 50581 - 50600

Published: Nov. 7, 2023

Language: Английский

---

Modeling the optimization of COVID-19 pooled testing: How many samples can be included in a single test?

Informatics in Medicine Unlocked, Journal Year: 2022, Volume and Issue: 32, P. 101037 - 101037

Published: Jan. 1, 2022

This study tries to answer the crucial question of how many biological samples can be optimally included in a single test for COVID-19 pooled testing.It builds novel theoretical model which links local population tested region, number test, "attitude" toward resource cost saving and time taken as well corresponding function function, together. The numerical simulation results are then used formulate function. Finally, loss minimized is constructed optimal calculated.In example, we consider region 1 million needs infection COVID-19. solution calculates 4.254 when given weight 50% under probability 10%. Other combinations also presented.As see our results, at 10%, setting (in integer level) [4,6] reasonable wide range subjective attitude between costs. Therefore, current practice, 5-mixed would sound better than commonly 10-mixed samples.

Language: Английский

---

Oversampling techniques for predicting COVID-19 patient length of stay

2021 IEEE International Conference on Big Data (Big Data), Journal Year: 2022, Volume and Issue: unknown, P. 5253 - 5262

Published: Dec. 17, 2022

COVID-19 is a respiratory disease that caused global pandemic in 2019. It highly infectious and has the following symptoms: fever or chills, cough, shortness of breath, fatigue, muscle body aches, headache, new loss taste smell, sore throat, congestion runny nose, nausea vomiting, diarrhea. These symptoms vary severity; some people with many risk factors have been known to lengthy hospital stays die from disease. In this paper, we analyze patients' electronic health records (EHR) predict severity their infection using length stay (LOS) as our measurement severity. This an imbalanced classification problem, shorter LOS rather than longer one. To combat synthetically create alternate oversampled training data sets. Once data, run it through Artificial Neural Network (ANN), which during its hyperparameters tuned by bayesian optimization. We select model best F1 score then evaluate discuss it.

Language: Английский

---

LoSNet: A Tailored Deep Neural Network Framework for Precise Length of Stay Prediction in Disease-Specific Hospitalization

Procedia Computer Science, Journal Year: 2024, Volume and Issue: 235, P. 2599 - 2608

Published: Jan. 1, 2024

During the COVID-19 pandemic, healthcare sector faced unprecedented challenges in effectively managing hospital resources. A crucial aspect of resource planning and allocation is ability to predict expected length a patient's stay. Detecting whether patient requires extended hospitalization or shorter stay becomes vital for efficient utilization. This paper aims build deep learning-based analytical model named LoSNet that predicts each at time admission Hospital. The early prediction requirement would aid professionals optimizing utility beds other In this direction, compares various machine-learning models including Random Forest, Decision Tree, Logistic Regression, Naïve Bayes with customized neural network model. dataset used analysis includes ground truth on 3,18,438 patients' categorized into eleven classes such as 0-10 days being one class, 11-20 another so more than 100 days. methodology employed study involves data collection, transformation, training LoSNet, no attention mechanisms. results indicate impressive performance over random classifier, cross-entropy loss 1.531 an accuracy 0.408 predicting durations 11-class classification setup, highlighting framework's effectiveness management.

Language: Английский

---