Smart Prediction Model For Unplanned Icu Transfer Based On Deep Learning Optimization : An Article Review

Abstract

Problem on units ICU already is problem which critical and already happened since long ago, for the ICU is one of the highest costs unit in hospitals, which made a system to predict activity on ICU is very demanding. COVID-19 shows the need for excellent time management in dealing with the abnormal flow of patients. Prediction of ICU transfer can be useful for patients and medical personnel to reduce medical cost and giving the time required by the nurses to prepare themselves for a huge patients flow. Reviews of related articles are carried out through the Google Scholar database. Screening then conducted based on identified article based on criteria eligibility. There are 7 final articles that assessed on a large scale data samples, method algorithm, and performance from the model which used on the article. Results obtained from this study which follow PRISM flow show a number of variable indicators that are commonly applied, namely: age, gender, liver function, blood pressure, pulse rate, temperature, respiratory rate, kgd and ECG data features. The best test results was achieved by research by Jonathan Rubin, et al due to the large number of varied data sets used, much more than other studies. This research also used adaptive boosting and gradient tree boosting approaches and evaluated with 4 main parameter that is accuracy, sensitivity, specificity, and AUC ROC. This study succeed in reaching performance evaluation model of 72.8% sensitivity, 76.3% specificity, 76.2% accuracy and 79.9% AUC ROC