Interpretable deep learning for the prediction of ICU admission likelihood and mortality of COVID-19 patients

Abstract

The global healthcare system is being overburdened by an increasing number of COVID-19 patients. Physicians are having difficulty allocating resources and focusing their attention on high-risk patients, partly due to the difficulty in identifying high-risk patients early. COVID-19 hospitalizations require specialized treatment capabilities and can cause a burden on healthcare resources. Estimating future hospitalization of COVID-19 patients is, therefore, crucial to saving lives. In this paper, an interpretable deep learning model is developed to predict intensive care unit (ICU) admission and mortality of COVID-19 patients. The study comprised of patients from the Stony Brook University Hospital, with patient information such as demographics, comorbidities, symptoms, vital signs, and laboratory tests recorded. The top three predictors of ICU admission were ferritin, diarrhoea, and alamine aminotransferase, and the top predictors for mortality were COPD, ferritin, and myalgia. The proposed model predicted ICU admission with an AUC score of 88.3% and predicted mortality with an AUC score of 96.3%. The proposed model was evaluated against existing model in the literature which achieved an AUC of 72.8% in predicting ICU admission and achieved an AUC of 84.4% in predicting mortality. It can clearly be seen that the model proposed in this paper shows superiority over existing models. The proposed model has the potential to provide tools to frontline doctors to help classify patients in time-bound and resource-limited scenarios.

Mortality Prediction Analysis among COVID-19 Inpatients Using Clinical Variables and Deep Learning Chest Radiography Imaging Features

“…Several prior studies also similarly reported improvements in performance when adding chest radiographic data to clinical prediction models [ 25 , 26 ], but with lower ROC-AUC values for predicting mortality or disease progression with their combined imaging and clinical models (0.73–0.83) than those of our mortality prediction model. When considering only non-imaging clinical data, our ROC-AUC performance metric of 0.87 for in-hospital COVID-19 mortality is identical to the value of 0.87 reported using the A-DROP clinical criteria in a different dataset [ 27 ] and mildly higher than several other deep learning models (ROC-AUC: 0.84) [ 28 ] or clinical scoring methods to quantify pneumonia severity, such as qSOFA (ROC-AUC: 0.73) and CRB-65 (ROC-AUC: 0.80) [ 27 ]; that said, machine learning models with higher ROC-AUC values have been reported [ 29 , 30 ].…”

Towards Improved XAI-Based Epidemiological Research into the Next Potential Pandemic

“…The attention approaches are inspired by human attention visual mechanisms, which use limited attention to quickly screen high value information from a large amount of information. This not only contributes to increase the prediction performance but is also efficient in gaining insight into information that is more critical to the model outputs instead of learning non-useful information [37,127,128].…”

Predicting medical device failure: a promise to reduce healthcare facilities cost through smart healthcare management

“…Furthermore, researchers have explored the possibility of using long short term memory (LSTM) and deep reinforcement learning to predict losses and cures of patients’ symptoms in the following few days after contracting the disease ( Kumar et al, 2021 ). Besides, the admission and mortality of COVID-19 patients are predicted using an interpretable DL model with an area under curve (AUC) of 88.3% ( Nazir & Ampadu, 2022 ). The application of AI is also recently used for privacy and security issues ( Hameed et al, 2021 ) and is widely used in smart and mobile healthcare ( Yamakoshi, Rolfe & Yamakoshi, 2021 ).…”