Predicting length of stay in hospitals intensive care unit using general admission features
| Autor: | Ahmed A. Eltahawi, Merhan A. Abd-Elrazek, Mohamed N. Abd-Elwhab, Mohamed H. Abd Elaziz |
|---|---|
| Rok vydání: | 2021 |
| Předmět: |
Bias variance tradeoff
business.industry Computer science 020209 energy Decision tree learning 020208 electrical & electronic engineering General Engineering 02 engineering and technology Engineering (General). Civil engineering (General) medicine.disease Intensive care unit World health law.invention Patient diagnosis law Machine learning Health care 0202 electrical engineering electronic engineering information engineering medicine Length of stay Performance measurement Medical team Medical emergency TA1-2040 business |
| Zdroj: | Ain Shams Engineering Journal, Vol 12, Iss 4, Pp 3691-3702 (2021) |
| ISSN: | 2090-4479 |
| DOI: | 10.1016/j.asej.2021.02.018 |
| Popis: | According to the World Health Organization (WHO), patient Length of Stay (LOS) in hospitals is an important performance measurement and monitoring indicator. Prolonged LOS in the Intensive Care Unit (ICU) may lead to consuming hospital resources, manpower, and equipment. Therefore, accurate prediction of patient LOS may aid the healthcare specialists to take medical decisions and allocate medical team and resources. As well, the patient and insurance companies may use this prediction to manage their budget. In this paper, a framework for predicting patient LOS in the ICU using different machine learning (ML) techniques is proposed. Unlike most of the previous studies, this study relies on general medical features collected on patient admission regardless of the patient diagnosis. This provide a broad scope and cover all patients making this approach general and easy to use. The prediction accuracy of the proposed approach was recorded to be very high and different for each ML technique. For example, the best prediction accuracy was achieved by fuzzy with accuracy reach 92%, while classification tree managed to achieve a prediction accuracy of 90% coming in the second place. |
| Databáze: | OpenAIRE |
| Externí odkaz: |
|