{"title":"使用机器学习技术预测糖尿病患者ICU住院时间","authors":"Yuansi Hu, Ling Zheng, Jiacun Wang","doi":"10.1109/ICCSI55536.2022.9970666","DOIUrl":null,"url":null,"abstract":"Diabetes is a prevalent chronic disease that can result in serious damages to various organ systems gradually. Patients with diabetes in the intensive care unit (ICU) have poor health outcomes and require more intensive care with higher healthcare costs. To facilitate resource management of hospitals and to improve health outcomes of patients with diabetes, accurately estimating the length of stay at an early stage of ICU admissions is necessary. This study is aimed to predict the length of stay for patients with diabetes by applying machine learning techniques on clinical data available during the first 8 hours of ICU admissions. Two prediction tasks, the number of days in ICU and whether an ICU stay is long or short distinguished by the threshold 10 days, were explored. The neural network model achieved the best performance in predicting the number of days in ICU with a R2 value 0.3969 and a mean absolute error 1.94 days. The gradient boosting model is the best one to classify long and short ICU stays with an accuracy 0.8214. The results demonstrate that these two models are promising to estimate the length of stay at an early stage of ICU admissions for patients with diabetes.","PeriodicalId":421514,"journal":{"name":"2022 International Conference on Cyber-Physical Social Intelligence (ICCSI)","volume":"19 1","pages":"0"},"PeriodicalIF":0.0000,"publicationDate":"2022-11-18","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":"2","resultStr":"{\"title\":\"Predicting ICU Length of Stay for Patients with Diabetes Using Machine Learning Techniques\",\"authors\":\"Yuansi Hu, Ling Zheng, Jiacun Wang\",\"doi\":\"10.1109/ICCSI55536.2022.9970666\",\"DOIUrl\":null,\"url\":null,\"abstract\":\"Diabetes is a prevalent chronic disease that can result in serious damages to various organ systems gradually. Patients with diabetes in the intensive care unit (ICU) have poor health outcomes and require more intensive care with higher healthcare costs. To facilitate resource management of hospitals and to improve health outcomes of patients with diabetes, accurately estimating the length of stay at an early stage of ICU admissions is necessary. This study is aimed to predict the length of stay for patients with diabetes by applying machine learning techniques on clinical data available during the first 8 hours of ICU admissions. Two prediction tasks, the number of days in ICU and whether an ICU stay is long or short distinguished by the threshold 10 days, were explored. The neural network model achieved the best performance in predicting the number of days in ICU with a R2 value 0.3969 and a mean absolute error 1.94 days. The gradient boosting model is the best one to classify long and short ICU stays with an accuracy 0.8214. The results demonstrate that these two models are promising to estimate the length of stay at an early stage of ICU admissions for patients with diabetes.\",\"PeriodicalId\":421514,\"journal\":{\"name\":\"2022 International Conference on Cyber-Physical Social Intelligence (ICCSI)\",\"volume\":\"19 1\",\"pages\":\"0\"},\"PeriodicalIF\":0.0000,\"publicationDate\":\"2022-11-18\",\"publicationTypes\":\"Journal Article\",\"fieldsOfStudy\":null,\"isOpenAccess\":false,\"openAccessPdf\":\"\",\"citationCount\":\"2\",\"resultStr\":null,\"platform\":\"Semanticscholar\",\"paperid\":null,\"PeriodicalName\":\"2022 International Conference on Cyber-Physical Social Intelligence (ICCSI)\",\"FirstCategoryId\":\"1085\",\"ListUrlMain\":\"https://doi.org/10.1109/ICCSI55536.2022.9970666\",\"RegionNum\":0,\"RegionCategory\":null,\"ArticlePicture\":[],\"TitleCN\":null,\"AbstractTextCN\":null,\"PMCID\":null,\"EPubDate\":\"\",\"PubModel\":\"\",\"JCR\":\"\",\"JCRName\":\"\",\"Score\":null,\"Total\":0}","platform":"Semanticscholar","paperid":null,"PeriodicalName":"2022 International Conference on Cyber-Physical Social Intelligence (ICCSI)","FirstCategoryId":"1085","ListUrlMain":"https://doi.org/10.1109/ICCSI55536.2022.9970666","RegionNum":0,"RegionCategory":null,"ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":null,"EPubDate":"","PubModel":"","JCR":"","JCRName":"","Score":null,"Total":0}
Predicting ICU Length of Stay for Patients with Diabetes Using Machine Learning Techniques
Diabetes is a prevalent chronic disease that can result in serious damages to various organ systems gradually. Patients with diabetes in the intensive care unit (ICU) have poor health outcomes and require more intensive care with higher healthcare costs. To facilitate resource management of hospitals and to improve health outcomes of patients with diabetes, accurately estimating the length of stay at an early stage of ICU admissions is necessary. This study is aimed to predict the length of stay for patients with diabetes by applying machine learning techniques on clinical data available during the first 8 hours of ICU admissions. Two prediction tasks, the number of days in ICU and whether an ICU stay is long or short distinguished by the threshold 10 days, were explored. The neural network model achieved the best performance in predicting the number of days in ICU with a R2 value 0.3969 and a mean absolute error 1.94 days. The gradient boosting model is the best one to classify long and short ICU stays with an accuracy 0.8214. The results demonstrate that these two models are promising to estimate the length of stay at an early stage of ICU admissions for patients with diabetes.