Predictors of intubation and mortality in COVID-19 patients: a retrospective study

Abstract

Background: Estimating the risk of intubation and mortality among COVID-19 patients can help clinicians triage these patients and allocate resources more efficiently. Thus, here we sought to identify the risk factors associated with intubation and intra-hospital mortality in a cohort of COVID-19 patients hospitalized due to hypoxemic acute respiratory failure (ARF).

Results: We included retrospectively a total of 187 patients admitted to the subintensive and intensive care units of the University Hospital "Maggiore della Carità" of Novara between March 1st and April 30th, 2020. Based on these patients' demographic characteristics, early clinical and laboratory variables, and quantitative chest computerized tomography (CT) findings, we developed two random forest (RF) models able to predict intubation and intra-hospital mortality. Variables independently associated with intubation were C-reactive protein (p < 0.001), lactate dehydrogenase level (p = 0.018) and white blood cell count (p = 0.026), while variables independently associated with mortality were age (p < 0.001), other cardiovascular diseases (p = 0.029), C-reactive protein (p = 0.002), lactate dehydrogenase level (p = 0.018), and invasive mechanical ventilation (p = 0.001). On quantitative chest CT analysis, ground glass opacity, consolidation, and fibrosis resulted significantly associated with patient intubation and mortality. The major predictors for both models were the ratio between partial pressure of arterial oxygen and fraction of inspired oxygen, age, lactate dehydrogenase, C-reactive protein, glycemia, CT quantitative parameters, lymphocyte count, and symptom onset.

Conclusions: Altogether, our findings confirm previously reported demographic, clinical, hemato-chemical, and radiologic predictors of adverse outcome among COVID-19-associated hypoxemic ARF patients. The two newly developed RF models herein described show an overall good level of accuracy in predicting intra-hospital mortality and intubation in our study population. Thus, their future development and implementation may help not only identify patients at higher risk of deterioration more effectively but also rebalance the disproportion between resources and demand.

Keywords: COVID-19; Factor risk; Intubation; Mortality; Quantitative computerized tomography; Random forest.

Fig. 1

Quantitative CT analysis of a 74-year-old male COVID-19 patient. a Non-contrast chest CT on admission, showing a characteristic bilateral and subpleural ground glass opacity (GGO). b Well-aerated parenchyma segmented semi-automatically by a 3D slicer; the blue area is the result of the subtraction of the all parenchyma (HU − 1100; − 250) with GGO (− 700; − 250) + consolidation (− 250; + 150). c The GGO area, obtained by semi-automatic segmentation (HU − 700; − 250). d Manual segmentation of consolidation areas (HU − 250; + 150)

Fig. 2

Important predictor variables (importance > 50) in the random forest (RF) models for intra-hospital mortality (a) and intubation prediction (b)

Fig. 3

AUC of the RF models. a) The RF model balanced accuracy in predicting intra-hospital mortality is 0.89 (κ = 0.72; AUC = 0.73). b The RF model balanced accuracy in predicting intubation is 0.9 (κ = 0.75; AUC = 0.74)

Fig. 4

RF model performance without CT parameters. a The RF model balanced accuracy in predicting intra-hospital mortality is 0.75 (κ = 0.1). b The RF model balanced accuracy in predicting intubation is 0.69 (κ = 0.29)