Machine learning models for predicting severe COVID-19 outcomes in hospitals

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

By applying AI techniques to a variety of pandemic-relevant data, artificial intelligence (AI) has substantially supported the control of the spread of the SARS-CoV-2 virus. Along with this, epidemiological machine learning studies of SARS-CoV-2 have been frequently published. While these models can be perceived as precise and policy-relevant to guide governments towards optimal containment policies, their black box nature can hamper building trust and relying confidently on the prescriptions proposed. This paper focuses on interpretable AI-based epidemiological models in the context of the recent SARS-CoV-2 pandemic. We systematically review existing studies, which jointly incorporate AI, SARS-CoV-2 epidemiology, and explainable AI approaches (XAI). First, we propose a conceptual framework by synthesizing the main methodological features of the existing AI pipelines of SARS-CoV-2. Upon the proposed conceptual framework and by analyzing the selected epidemiological studies, we reflect on current research gaps in epidemiological AI toolboxes and how to fill these gaps to generate enhanced policy support in the next potential pandemic.

Explainable artificial intelligence approaches for COVID-19 prognosis prediction using clinical markers

The COVID-19 influenza emerged and proved to be fatal, causing millions of deaths worldwide. Vaccines were eventually discovered, effectively preventing the severe symptoms caused by the disease. However, some of the population (elderly and patients with comorbidities) are still vulnerable to severe symptoms such as breathlessness and chest pain. Identifying these patients in advance is imperative to prevent a bad prognosis. Hence, machine learning and deep learning algorithms have been used for early COVID-19 severity prediction using clinical and laboratory markers. The COVID-19 data was collected from two Manipal hospitals after obtaining ethical clearance. Multiple nature-inspired feature selection algorithms are used to choose the most crucial markers. A maximum testing accuracy of 95% was achieved by the classifiers. The predictions obtained by the classifiers have been demystified using five explainable artificial intelligence techniques (XAI). According to XAI, the most important markers are c-reactive protein, basophils, lymphocytes, albumin, D-Dimer and neutrophils. The models could be deployed in various healthcare facilities to predict COVID-19 severity in advance so that appropriate treatments could be provided to mitigate a severe prognosis. The computer aided diagnostic method can also aid the healthcare professionals and ease the burden on already suffering healthcare infrastructure.

Identified Factors in COVID-19 Patients in Predicting Mortality

BACKGROUND

The coronavirus disease 2019 (COVID-19) pandemic has led to a significant increase in global mortality rates. Numerous studies have been conducted to identify the factors associated with mortality in COVID-19 cases. In these studies, overall mortality was evaluated in patients, and no distinction was made as ward or intensive care mortality.

AIM

This study aims to determine mortality-related factors in patients who died while in the ward. This could enable us to review the indications for intensive care hospitalization in possible pandemics.

MATERIALS AND METHOD

This retrospective study was conducted on a cohort of 237 patients who applied to our institution between January 2020 and December 2021 with the diagnosis of COVID-19. Demographic characteristics, length of stay, type of admission (emergency ward or outpatient clinic), presence of comorbidities, thoracic computerized tomography (CT) findings, and laboratory findings were extracted from the hospital database. The demographic and laboratory results of both deceased and recovered patients were compared.

RESULTS

While many demographic and laboratory findings were statistically significant in the initial analysis, multiple logistic regression analysis showed that decreased albumin levels (adjusted OR = 0.23, 95% CI = 0.09 - 0.57), increased troponin (adjusted OR = 1.03, 95% CI = 1.02 - 1.05), and procalcitonin (adjusted OR = 3.46, 95% CI = 1.04 - 11.47) levels and higher partial thromboplastin time (PTT) (adjusted OR = 1.18, 95% CI = 1.09 - 1.28) values, presence of diabetes mellitus (DM) in patients (adjusted OR = 2.18, 95% CI = 1.01 - 4.69, P = 0.047), and admission to hospital from the emergency department (adjusted OR = 5.15, 95% CI = 1.45 - 18.27, P = 0.011) were significantly associated with mortality when adjusted for age. When a predictive model is constructed with these variables, this model predicted mortality statistically significant (AUC = 0.904, 95% CI = 0.856 - 0.938, P < 0.001), with a sensitivity of 77.2% (95% CI, 67.8 - 85), a specificity of 91.2% (95% CI, 85.1 - 95.4), a positive predictive value (PPV) of 86.7% (95% CI, 72 - 85.3), and an negative predictive value (NPV) of 84.4% (95% CI, 79.4 - 89.6).

CONCLUSION

In this study, we may predict mortality among COVID-19-diagnosed patients admitted to the ward via this model which has the potential to provide guidance for reconsidering the indications for intensive care unit (ICU) admission.

Inflammation and Venous Thromboembolism in Hospitalized Patients with COVID-19

BACKGROUND

A link between inflammation and venous thromboembolism (VTE) in COVID-19 disease has been suggested pathophysiologically and clinically. The aim of this study was to investigate the association between inflammation and disease outcomes in adult hospitalized COVID-19 patients with VTE.

METHODS

This was a retrospective observational study, including quantitative and qualitative data collected from COVID-19 patients hospitalized at the Infectious Diseases Unit (IDU) of the University Hospital of Ioannina, from 1 March 2020 to 31 May 2022. Venous thromboembolism was defined as a diagnosis of pulmonary embolism (PE) and/or vascular tree-in-bud in the lungs. The burden of disease, assessed by computed tomography of the lungs (CTBoD), was quantified as the percentage (%) of the affected lung parenchyma. The study outcomes were defined as death, intubation, and length of hospital stay (LoS). A chi-squared test and univariate logistic regression analyses were performed in IBM SPSS 28.0.

RESULTS

After propensity score matching, the final study cohort included 532 patients. VTE was found in 11.2% of the total population. In patients with VTE, we found that lymphocytopenia and a high neutrophil/lymphocyte ratio were associated with an increased risk of intubation and death, respectively. Similarly, CTBoD > 50% was associated with a higher risk of intubation and death in this group of patients. The triglyceride-glucose (TyG) index was also linked to worse outcomes.

CONCLUSIONS

Inflammatory indices were associated with VTE. Lymphocytopenia and an increased neutrophil-to-lymphocyte ratio negatively impacted the disease's prognosis and outcomes. Whether these indices unfavorably affect outcomes in COVID-19-associated VTE must be further evaluated.

Development and validation of a prediction model for mechanical ventilation based on comorbidities in hospitalized patients with COVID-19

Background

Timely recognition of respiratory failure and the need for mechanical ventilation is crucial in managing patients with coronavirus disease 2019 (COVID-19) and reducing hospital mortality rate. A risk stratification tool could assist to avoid clinical deterioration of patients with COVID-19 and optimize allocation of scarce resources. Therefore, we aimed to develop a prediction model for early identification of patients with COVID-19 who may require mechanical ventilation.

Methods

We included patients with COVID-19 hospitalized in United States. Demographic and clinical data were extracted from the records of the Healthcare Cost and Utilization Project State Inpatient Database in 2020. Model construction involved the use of the least absolute shrinkage and selection operator and multivariable logistic regression. The model's performance was evaluated based on discrimination, calibration, and clinical utility.

Results

The training set comprised 73,957 patients (5,971 requiring mechanical ventilation), whereas the validation set included 10,428 (887 requiring mechanical ventilation). The prediction model incorporating age, sex, and 11 other comorbidities (deficiency anemias, congestive heart failure, coagulopathy, dementia, diabetes with chronic complications, complicated hypertension, neurological disorders unaffecting movement, obesity, pulmonary circulation disease, severe renal failure, and weight loss) demonstrated moderate discrimination (area under the curve, 0.715; 95% confidence interval, 0.709-0.722), good calibration (Brier score = 0.070, slope = 1, intercept = 0) and a clinical net benefit with a threshold probability ranged from 2 to 34% in the training set. Similar model's performances were observed in the validation set.

Conclusion

A robust prognostic model utilizing readily available predictors at hospital admission was developed for the early identification of patients with COVID-19 who may require mechanical ventilation. Application of this model could support clinical decision-making to optimize patient management and resource allocation.

Validation of a pre-established triage protocol for critically ill patients in a COVID-19 outbreak under resource scarcity: A retrospective multicenter cohort study

INTRODUCTION

In case of COVID-19 related scarcity of critical care resources, an early French triage algorithm categorized critically ill patients by probability of survival based on medical history and severity, with four priority levels for initiation or continuation of critical care: P1 -high priority, P2 -intermediate priority, P3 -not needed, P4 -not appropriate. This retrospective multi-center study aimed to assess its classification performance and its ability to help saving lives under capacity saturation.

METHODS

ICU patients admitted for severe COVID-19 without triage in spring 2020 were retrospectively included from three hospitals. Demographic data, medical history and severity items were collected. Priority levels were retrospectively allocated at ICU admission and on ICU day 7-10. Mortality rate, cumulative incidence of death and of alive ICU discharge, length of ICU stay and of mechanical ventilation were compared between priority levels. Calculated mortality and survival were compared between full simulated triage and no triage.

RESULTS

225 patients were included, aged 63.1±11.9 years. Median SAPS2 was 40 (IQR 29-49). At the end of follow-up, 61 (27%) had died, 26 were still in ICU, and 138 had been discharged. Following retrospective initial priority allocation, mortality rate was 53% among P4 patients (95CI 34-72%) versus 23% among all P1 to P3 patients (95CI 17-30%, chi-squared p = 5.2e-4). The cumulative incidence of death consistently increased in the order P3, P1, P2 and P4 both at admission (Gray's test p = 3.1e-5) and at reassessment (p = 8e-5), and conversely for that of alive ICU discharge. Reassessment strengthened consistency. Simulation under saturation showed that this two-step triage protocol could have saved 28 to 40 more lives than no triage.

CONCLUSION

Although it cannot eliminate potentially avoidable deaths, this triage protocol proved able to adequately prioritize critical care for patients with highest probability of survival, hence to save more lives if applied.