Prediction of COVID-19 Hospital Length of Stay and Risk of Death Using Artificial Intelligence-Based Modeling

Abstract

Severe acute respiratory syndrome coronavirus 2 (SARS-CoV-2) is a highly infectious virus with overwhelming demand on healthcare systems, which require advanced predictive analytics to strategize COVID-19 management in a more effective and efficient manner. We analyzed clinical data of 2017 COVID-19 cases reported in the Dubai health authority and developed predictive models to predict the patient's length of hospital stay and risk of death. A decision tree (DT) model to predict COVID-19 length of stay was developed based on patient clinical information. The model showed very good performance with a coefficient of determination R2 of 49.8% and a median absolute deviation of 2.85 days. Furthermore, another DT-based model was constructed to predict COVID-19 risk of death. The model showed excellent performance with sensitivity and specificity of 96.5 and 87.8%, respectively, and overall prediction accuracy of 96%. Further validation using unsupervised learning methods showed similar separation patterns, and a receiver operator characteristic approach suggested stable and robust DT model performance. The results show that a high risk of death of 78.2% is indicated for intubated COVID-19 patients who have not used anticoagulant medications. Fortunately, intubated patients who are using anticoagulant and dexamethasone medications with an international normalized ratio of <1.69 have zero risk of death from COVID-19. In conclusion, we constructed artificial intelligence–based models to accurately predict the length of hospital stay and risk of death in COVID-19 cases. These smart models will arm physicians on the front line to enhance management strategies to save lives.

Factors Associated with Length of Hospital Stay among COVID-19 Patients in Saudi Arabia: A Retrospective Study during the First Pandemic Wave

“…Similar to our findings, Alwafi et al reported that hospitalized COVID-19 patients with end-stage renal diseases had a significantly higher risk of mortality and a significantly prolonged LoHS [ 13 ]. Mahboub et al found that diagnosis at admission, painkiller usage, ventilator intubation, azithromycin usage, antiviral usage, anti-inflammatory agent usage, vitamin C usage, urea test results, platelet count, haemoglobin levels, dimer levels, and potassium levels were significant predictors of LoHS [ 39 ]. Moreover, Wang et al reported that age and clinical grade were strongly related to the length of stay ( p < 0.01) and that a longer LoHS was associated with ≥45 years of age, severe illness, and admission to a provincial hospital [ 40 ].…”

Identifying Immunological and Clinical Predictors of COVID-19 Severity and Sequelae by Mathematical Modeling

“…Interestingly, stochastic modeling was previously used to model the human immune response to the yellow fever vaccine ( 9 ). Since COVID-19 is linked to immune response, modeling of the SARS-CoV-2 infection have been extensively published on different aspects of the disease, including the immune system using multiple ODEs to model immune cells, antibodies and cytokines ( 10 – 13 ), and on the clinical and radiological data ( 14 – 16 ). A few models on cytokine release syndrome in other diseases were also created ( 17 – 19 ).…”

Length of Stay Analysis of COVID-19 Hospitalizations Using a Count Regression Model and Quantile Regression: A Study in Bologna, Italy

“…From a management point of view, research in Dubai used AI-based modeling by using Decision Tree prediction models to forecast the LoS and risk of mortality accurately. In principle, these smart models might equip front-line clinicians with the tools they need to improve management techniques and save lives [ 19 ].…”