Predicting 90-day survival of patients with COVID-19: Survival of Severely Ill COVID (SOSIC) scores

Abstract

Background: Predicting outcomes of critically ill intensive care unit (ICU) patients with coronavirus-19 disease (COVID-19) is a major challenge to avoid futile, and prolonged ICU stays.

Methods: The objective was to develop predictive survival models for patients with COVID-19 after 1-to-2 weeks in ICU. Based on the COVID-ICU cohort, which prospectively collected characteristics, management, and outcomes of critically ill patients with COVID-19. Machine learning was used to develop dynamic, clinically useful models able to predict 90-day mortality using ICU data collected on day (D) 1, D7 or D14.

Results: Survival of Severely Ill COVID (SOSIC)-1, SOSIC-7, and SOSIC-14 scores were constructed with 4244, 2877, and 1349 patients, respectively, randomly assigned to development or test datasets. The three models selected 15 ICU-entry variables recorded on D1, D7, or D14. Cardiovascular, renal, and pulmonary functions on prediction D7 or D14 were among the most heavily weighted inputs for both models. For the test dataset, SOSIC-7's area under the ROC curve was slightly higher (0.80 [0.74-0.86]) than those for SOSIC-1 (0.76 [0.71-0.81]) and SOSIC-14 (0.76 [0.68-0.83]). Similarly, SOSIC-1 and SOSIC-7 had excellent calibration curves, with similar Brier scores for the three models.

Conclusion: The SOSIC scores showed that entering 15 to 27 baseline and dynamic clinical parameters into an automatable XGBoost algorithm can potentially accurately predict the likely 90-day mortality post-ICU admission (sosic.shinyapps.io/shiny). Although external SOSIC-score validation is still needed, it is an additional tool to strengthen decisions about life-sustaining treatments and informing family members of likely prognosis.

Keywords: Acute respiratory distress syndrome; COVID-19; Mechanical ventilation; Outcome; Predictive survival model.

Fig. 1

Flowchart of COVID-19 patient screening, inclusion, and assignment to the development and test datasets. ICU intensive care unit, SOSIC Survival of Severely Ill COVID score

Fig. 2

Variable weighting in the machine-learning models: Survival of Severely Ill COVID (SOSIC) scores on days 1, 7 and 14. A color gradient is used to show variable strength in the machine-learning models at different times, ranging from yellow for the highest preponderance input variables to progressively darker shade of purple for lower input variables. D day, BMI body mass index, ICU intensive care unit, MV mechanical ventilation, SOSIC Survival of Severely Ill COVID score, SOFA Sequential Organ-Failure Assessment, VAP ventilator-associated pneumonia

Fig. 3

Calibration and discrimination of the Survival of Severely Ill COVID (SOSIC)-1, SOSIC-7 and SOSIC-14 scores using the test dataset

Fig. 4

Graphic representation of the SOSIC-1, the SAPS II, and the SOFA performances in the A development and B the test datasets. SOSIC Survival of Severely Ill COVID score; SAPS II Simplified Acute Physiology Score II, SOFA Sequential Organ-Failure Assessment