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

Corticosteroids induce an early but limited decrease in IL-6 dependent pro-inflammatory responses in critically ill COVID-19 patients

BACKGROUND

Corticosteroids have become standard of care for COVID-19 but their effect on the systemic immune-inflammatory response has been little investigated.

METHODS

Multicenter prospective cohort, including critically ill COVID-19 patients between March and November 2020. C-reactive protein (CRP), lymphocyte count and fibrinogen levels were collected upon hospital admission before initiation of steroid treatment and at ICU admission, three days and seven days later, along with interleukin (IL)-6, IL-10 and tumor necrosis factor-alpha (TNF-α) plasma levels.

RESULTS

A hundred and fifty patients were included, 47 received corticosteroids, 103 did not. Median age was 62 [53-70], and 96 (65%) patients were mechanically ventilated. Propensity score matching rendered 45 well-balanced pairs of treated and non-treated patients, particularly on pre-treatment CRP levels. Using a mixed model, CRP (P=0.019), fibrinogen (P=0.003) and lymphocyte counts (P=0.006) remained lower in treated patients over ICU stay. Conversely, there was no significant difference over the ICU stay for Il-6 (P=0.146) and IL-10 (0.301), while TNF- α levels were higher in the treated group (P=0.013). Among corticosteroid-treated patients, CRP (P=0.012), fibrinogen (P=0.041) and lymphocyte count (P=0.004) over time were associated with outcome, whereas plasma cytokine levels were not.

CONCLUSIONS

Steroid treatment was associated with an early and sustained decrease in the downstream IL-6-dependent inflammatory signature but an increase in TNF-α levels. In corticosteroid-treated patients, CRP and lymphocyte count were associated with outcome, conversely to plasma cytokine levels. Further research on using these biomarker's kinetics to individualize immunomodulatory treatments is warranted.

Predicting COVID-19 severity: Challenges in reproducibility and deployment of machine learning methods

The increasing use of electronic health records (EHR) based computable phenotypes in clinical research is providing new opportunities for development of data-driven medical applications. Adopted widely in the United States and globally, EHRs facilitate systematic collection of patients' longitudinal information, which serves as one of the important foundations for artificial intelligence applications in medicine. Harmonization of input variables and outcome definitions is critically important for wider clinical applicability of artificial intelligence (AI) methodologies. In this review, we focused on Coronavirus Disease 2019 (COVID-19) severity machine learning prediction models and explored the pipeline for standardizing future disease severity model development using EHR information. We identified 2,967 studies published between 01/01/2020 and 02/15/2022 and selected 135 independent studies that had built machine learning prediction models to predict severity related outcomes of COVID-19 patients based on EHR data for the final review. These 135 studies spanning across 27 counties covered a broad range of severity related prediction outcomes. We observed substantial inconsistency in COVID-19 severity phenotype definitions among models in these studies. Moreover, there was a gap between the outcome of these models and clinician-recognized clinical concepts. Accordingly, we recommend that robust clinical input metrics, with outcome definitions which eliminate ambiguity in interpretation, to reduce algorithmic bias, mitigate model brittleness and improve generalizability of a universal model for COVID-19 severity. This framework can potentially be extended to broader clinical application.

The importance of predictors for in-hospital COVID-19 mortality changes over one month

BACKGROUND

Risk stratification enables care providers to make the proper clinical decision for the management of patients with COVID-19 infection. We aimed to explore changes in the importance of predictors for inpatient mortality of COVID-19 over one month.

METHODS

This research was a secondary analysis of data from in-hospital patients with COVID-19 infection. Individuals were admitted to four hospitals, New York, USA. Based on the length of hospital stay, 4370 patients were categorized into three mutually exclusive interval groups, day 1, day 2-7, and day 8-28. We measured changes in the importance of twelve confirmed predictors for mortality over one month, using principal component analysis.

RESULTS

On the first day of admission, there was a higher risk for organ dysfunction, particularly in elderly patients. On day 1, serum aspartate aminotransferase and sodium were also associated with an increased risk of mortality, while normal troponin opposes in-hospital death. With time, the importance of high aspartate aminotransferase and sodium concentrations decreases, while the variable quality of high troponin levels increases. Our study suggested the importance of maintaining normal blood pressure early in the management of patients. High serum concentrations of creatinine and C-reactive protein remain poor prognostic factors throughout the 28 days. The association of age with mortality increases with the length of hospital stay.

CONCLUSION

The importance of some patients' characteristics changes with the length of hospital stay. This should be considered in developing and deploying predictive models and the management of patients with COVID-19 infection.

The positive impact of COVID-19 on critical care: from unprecedented challenges to transformative changes, from the perspective of young intensivists

Over the past 2 years, SARS-CoV-2 infection has resulted in numerous hospitalizations and deaths worldwide. As young intensivists, we have been at the forefront of the fight against the COVID-19 pandemic and it has been an intense learning experience affecting all aspects of our specialty. Critical care was put forward as a priority and managed to adapt to the influx of patients and the growing demand for beds, financial and material resources, thereby highlighting its flexibility and central role in the healthcare system. Intensivists assumed an essential and unprecedented role in public life, which was important when claiming for indispensable material and human investments. Physicians and researchers around the world worked hand-in-hand to advance research and better manage this disease by integrating a rapidly growing body of evidence into guidelines. Our daily ethical practices and communication with families were challenged by the massive influx of patients and restricted visitation policies, forcing us to improve our collaboration with other specialties and innovate with new communication channels. However, the picture was not all bright, and some of these achievements are already fading over time despite the ongoing pandemic and hospital crisis. In addition, the pandemic has demonstrated the need to improve the working conditions and well-being of critical care workers to cope with the current shortage of human resources. Despite the gloomy atmosphere, we remain optimistic. In this ten-key points review, we outline our vision on how to capitalize on the lasting impact of the pandemic to face future challenges and foster transformative changes of critical care for the better.

Comparison of high-flow nasal therapy, noninvasive ventilation, and continuous positive airway pressure on outcomes in critically ill patients admitted for COVID-19 with acute respiratory failure

BACKGROUND

The optimal first-line noninvasive respiratory support (NIRS) to improve outcome in patients affected by COVID-19 pneumonia admitted to ICU is still debated.

METHODS

We conducted a retrospective study in seven French ICUs, including all adults admitted between July and December 2020 with documented SARS-CoV-2 acute respiratory failure (PaO2/FiO2<300 mmHg), and treated with either high-flow nasal therapy (HFNT) alone, noninvasive ventilation alone or in combination with HFNT (NIV), or continuous positive airway pressure alone or in combination with HFNT (CPAP). The primary outcome was NIRS failure at day 28, defined as the need for endotracheal intubation (ETI) or death without ETI.

RESULTS

Among the 355 patients included, 160 (45%) were treated with HFNT alone, 115 (32%) with NIV and 80 (23%) with CPAP. The primary outcome occurred in 65 (41%), 69 (60%), and 25 (31%) patients among those treated with HFNT alone, NIV, and CPAP, respectively (P<0.001). After univariate analysis, patients treated with CPAP had a trend for a lower incidence of the primary outcome, whereas patients treated with NIV had a significant higher incidence of the primary outcome, both compared to those treated with HFNT alone (unadjusted Hazard ratio 0.67; 95% CI [0.42-1.06], and 1.58; 95% CI [1.12-2.22]; P=0.09 and 0.008, respectively).

CONCLUSIONS

Among ICU patients admitted for severe COVID-19 pneumonia and managed with NIRS, the outcome seems to differ according to the initial chosen strategy. Prospective randomized controlled studies are warranted to identify the optimal strategy.

MedML: Fusing medical knowledge and machine learning models for early pediatric COVID-19 hospitalization and severity prediction

The COVID-19 pandemic has caused devastating economic and social disruption. This has led to a nationwide call for models to predict hospitalization and severe illness in patients with COVID-19 to inform the distribution of limited healthcare resources. To address this challenge, we propose a machine learning model, MedML, to conduct the hospitalization and severity prediction for the pediatric population using electronic health records. MedML extracts the most predictive features based on medical knowledge and propensity scores from over 6 million medical concepts and incorporates the inter-feature relationships in medical knowledge graphs via graph neural networks. We evaluate MedML on the National Cohort Collaborative (N3C) dataset. MedML achieves up to a 7% higher AUROC and 14% higher AUPRC compared to the best baseline machine learning models. MedML is a new machine learnig framework to incorporate clinical domain knowledge and is more predictive and explainable than current data-driven methods.

Development and Internal Validation of a New Prognostic Model Powered to Predict 28-Day All-Cause Mortality in ICU COVID-19 Patients-The COVID-SOFA Score

Background: The sequential organ failure assessment (SOFA) score has poor discriminative ability for death in severely or critically ill patients with Coronavirus disease 2019 (COVID-19) requiring intensive care unit (ICU) admission. Our aim was to create a new score powered to predict 28-day mortality. Methods: Retrospective, observational, bicentric cohort study including 425 patients with COVID-19 pneumonia, acute respiratory failure and SOFA score ≥ 2 requiring ICU admission for ≥72 h. Factors with independent predictive value for 28-day mortality were identified after stepwise Cox proportional hazards (PH) regression. Based on the regression coefficients, an equation was computed representing the COVID-SOFA score. Discriminative ability was tested using receiver operating characteristic (ROC) analysis, concordance statistics and precision-recall curves. This score was internally validated. Results: Median (Q1−Q3) age for the whole sample was 64 [55−72], with 290 (68.2%) of patients being male. The 28-day mortality was 54.58%. After stepwise Cox PH regression, age, neutrophil-to-lymphocyte ratio (NLR) and SOFA score remained in the final model. The following equation was computed: COVID-SOFA score = 10 × [0.037 × Age + 0.347 × ln(NLR) + 0.16 × SOFA]. Harrell’s C-index for the COVID-SOFA score was higher than the SOFA score alone for 28-day mortality (0.697 [95% CI; 0.662−0.731] versus 0.639 [95% CI: 0.605−0.672]). Subsequently, the prediction error rate was improved up to 16.06%. Area under the ROC (AUROC) was significantly higher for the COVID-SOFA score compared with the SOFA score for 28-day mortality: 0.796 [95% CI: 0.755−0.833] versus 0.699 [95% CI: 0.653−0.742, p < 0.001]. Better predictive value was observed with repeated measurement at 48 h after ICU admission. Conclusions: The COVID-SOFA score is better than the SOFA score alone for 28-day mortality prediction. Improvement in predictive value seen with measurements at 48 h after ICU admission suggests that the COVID-SOFA score can be used in a repetitive manner. External validation is required to support these results.

Dynamic of Serum TWEAK Levels in Critically Ill COVID-19 Male Patients

Although the number of cases and mortality of COVID-19 are seemingly declining, clinicians endeavor to establish indicators and predictors of such responses in order to optimize treatment regimens for future outbreaks of SARS-CoV-2 or similar viruses. Considering the importance of aberrant immune response in severe COVID-19, in the present study, we aimed to explore the dynamic of serum TNF-like weak inducer of apoptosis (TWEAK) levels in critically-ill COVID-19 patients and establish whether these levels may predict in-hospital mortality and if TWEAK is associated with impairment of testosterone levels observed in this population. The present single-center cohort study involved 66 men between the ages of 18 and 65 who were suffering from a severe type of COVID-19. Serum TWEAK was rising during the first week after admission to intensive care unit (ICU), whereas decline to baseline values was observed in the second week post-ICU admission (p = 0.032) but not in patients who died in hospital. Receiver-operator characteristics analysis demonstrated that serum TWEAK at admission to ICU is a significant predictor of in-hospital mortality (AUC = 0.689, p = 0.019). Finally, a negative correlation was found between serum TWEAK at admission and testosterone levels (r = −0.310, p = 0.036). In summary, serum TWEAK predicts in-hospital mortality in severe COVID-19. In addition, inflammatory pathways including TWEAK seem to be implicated in pathophysiology of reproductive hormone axis disturbance in severe form of COVID-19.

Clinical impact of ventilator-associated pneumonia in patients with the acute respiratory distress syndrome: a retrospective cohort study

Background

The clinical impact and outcomes of ventilator-associated pneumonia (VAP) have been scarcely investigated in patients with the acute respiratory distress syndrome (ARDS).

Methods

Patients admitted over an 18-month period in two intensive care units (ICU) of a university-affiliated hospital and meeting the Berlin criteria for ARDS were retrospectively included. The association between VAP and the probability of death at day 90 (primary endpoint) was appraised through a Cox proportional hazards model handling VAP as a delay entry variable. Secondary endpoints included (i) potential changes in the PaO₂/FiO₂ ratio and SOFA score values around VAP (linear mixed modelling), and (ii) mechanical ventilation (MV) duration, numbers of ventilator- and vasopressor-free days at day 28, and length of stay (LOS) in patients with and without VAP (median or absolute risk difference calculation). Subgroup analyses were performed in patients with COVID-19-related ARDS and those with ARDS from other causes.

Results

Among the 336 included patients (101 with COVID-19 and 235 with other ARDS), 176 (52.4%) experienced a first VAP. VAP induced a transient and moderate decline in the PaO₂/FiO₂ ratio without increase in SOFA score values. VAP was associated with less ventilator-free days (median difference and 95% CI, − 19 [− 20; − 13.5] days) and vasopressor-free days (− 5 [− 9; − 2] days) at day 28, and longer ICU (+ 13 [+ 9; + 15] days) and hospital (+ 11.5 [+ 7.5; + 17.5] days) LOS. These effects were observed in both subgroups. Overall day-90 mortality rates were 35.8% and 30.0% in patients with and without VAP, respectively (P = 0.30). In the whole cohort, VAP (adjusted HR 3.16, 95% CI 2.04–4.89, P < 0.0001), the SAPS-2 value at admission, chronic renal disease and an admission for cardiac arrest predicted death at day 90, while the COVID-19 status had no independent impact. When analysed separately, VAP predicted death in non-COVID-19 patients (aHR 3.43, 95% CI 2.11–5.58, P < 0.0001) but not in those with COVID-19 (aHR 1.19, 95% CI 0.32–4.49, P = 0.80).

Conclusions

VAP is an independent predictor of 90-day mortality in ARDS patients. This condition exerts a limited impact on oxygenation but correlates with extended MV duration, vasoactive support, and LOS.

Supplementary Information

The online version contains supplementary material available at 10.1186/s13613-022-00998-7.

Dynamic prediction of mortality in COVID-19 patients in the intensive care unit: A retrospective multi-center cohort study

Background

The COVID-19 pandemic continues to overwhelm intensive care units (ICUs) worldwide, and improved prediction of mortality among COVID-19 patients could assist decision making in the ICU setting. In this work, we report on the development and validation of a dynamic mortality model specifically for critically ill COVID-19 patients and discuss its potential utility in the ICU.

Methods

We collected electronic medical record (EMR) data from 3222 ICU admissions with a COVID-19 infection from 25 different ICUs in the Netherlands. We extracted daily observations of each patient and fitted both a linear (logistic regression) and non-linear (random forest) model to predict mortality within 24 h from the moment of prediction. Isotonic regression was used to re-calibrate the predictions of the fitted models. We evaluated the models in a leave-one-ICU-out (LOIO) cross-validation procedure.

Results

The logistic regression and random forest model yielded an area under the receiver operating characteristic curve of 0.87 [0.85; 0.88] and 0.86 [0.84; 0.88], respectively. The recalibrated model predictions showed a calibration intercept of −0.04 [−0.12; 0.04] and slope of 0.90 [0.85; 0.95] for logistic regression model and a calibration intercept of −0.19 [−0.27; −0.10] and slope of 0.89 [0.84; 0.94] for the random forest model.

Discussion

We presented a model for dynamic mortality prediction, specifically for critically ill COVID-19 patients, which predicts near-term mortality rather than in-ICU mortality. The potential clinical utility of dynamic mortality models such as benchmarking, improving resource allocation and informing family members, as well as the development of models with more causal structure, should be topics for future research.