A Risk Model to Predict Hospital Mortality in Ventilated Patients With ARDS

Can healthy immune cells from a patient be frozen and used in the future to make adoptive cell therapies in the event of a cancer diagnosis?
Can a model using routinely collected data on ventilated patients with ARDS accurately predict mortality in these patients?

Among ventilated patients with acute respiratory distress syndrome (ARDS), use of a risk score based on routinely collected variables at the beginning of intensive care unit (ICU) admission and invasive mechanical ventilation (IMV) can predict mortality with moderate performance, according to study results published in the journal BMC Pulmonary Medicine.

Because there is great variability in mortality among patients with ARDS, and especially among those on IMV, investigators sought to develop a model for predicting mortality risk among this patient population.

Toward that end, they conducted a population-based, observational, retrospective study using data extracted from the Medical Information Mart for Intensive Care III (MIMIC-III) database, the MIMIC-IV database (v1.0), and the eICU Collaborative Research Database (eICU-CRD). MIMIC-III comprises unidentified health-related data of more than 60,000 ICU stays at Beth Israel Deaconess Medical Center (BIDMC), Boston, Massachusetts, between June 2001 and October 2012. MIMIC-IV includes BIDMC data from 2008 to 2019. The eICU-CRD is a multicenter database that comprises health data linked to more than 200,000 ICU encounters from 335 units at 208 hospitals throughout the US between 2014 and 2015.

All patients in the MIMIC-III, MIMIC-IV, and eICU-CRD databases who met the following criteria were included in the analysis: (1) at least 16 years of age; (2) diagnosis of ARDS in the initial 48 hours of ventilation; and (3) receipt of IMV for at least 48 consecutive hours. Only data from the first ICU admission of the initial hospitalization underwent analysis.

The researchers used least absolute shrinkage and selection operator (LASSO) followed by logistic regression to construct a predictive model using demographic, clinical, laboratory, comorbidity, and ventilation data within 24 hours of ICU admission and initiation of IMV. Data from a total of 1596 participants (535 individuals from MIMIC-III, 521 from MIMI-IV, and 540 from eICU) were included in the final cohort for analysis. The participants pooled from MIMIC-III and eICU (n=1075) were randomly divided into 1 of 2 cohorts: (1) the training cohort (70%; n=752), to develop the model; and (2) the internal validation cohort (30%; n=323), to test the performance of the model. Data from MIMIC-IV were used for external validation.

Results of the study showed that in the training cohort, the overall in-hospital mortality rate was 32.7%. In this cohort, 47.6% (358 of 752) of the participants developed severe ARDS within the initial 48 hours of ventilation.

From a total of 176 possible predictors, 9 independent predictive factors were included in the final model. Five of these variables were ascertained within the first 24 hours of ICU admission: (1) age (odds ratio [OR], 1.02; 95% CI, 1.01-1.03), (2) mean of respiratory rate (OR, 1.04; 95% CI, 1.01-1.08, (3) maximum of international normalized ratio (OR, 1.14; 95% CI, 1.03-1.31), (4) maximum of alveolo-arterial oxygen difference (OR, 1.002; 95% CI, 1.001-1.003), and (5) minimum of red blood cell distribution width (OR, 1.17; 95% CI, 1.09-1.27).

Additionally, 4 variables were measured within the first 24 hours after initiation of IMV: (1) mean of temperature (OR, 0.70; 95% CI, 0.57-0.86), (2) maximum of lactate (OR, 1.15; 95% CI, 1.09-1.22), (3) minimum of blood urea nitrogen (OR, 1.02; 95% CI, 1.01-1.03), and (4) minimum of white blood cell counts (OR, 1.03; 95% CI, 1.01-1.06).

The current model attained good discrimination (area under the curve [AUC], 0.77;
95% CI, 0.73-0.80) in the training cohort; however, the performance declined in the internal cohort (AUC, 0.75; 95% CI, 0.69-0.80) and in the external validation cohort (AUC, 0.70; 95% CI, 0.65-0.74), and demonstrated modest calibration.

The current analysis had several limitations. Because the study was retrospective in nature and observationally designed, selection bias, loss to follow-up, and the presence of confounding factors are unavoidable. Further, certain variables were excluded from the missing data (eg, albumin, hepatic function, and neutrophil-to-lymphocyte ratio), even though prior research has demonstrated that these variables might be associated with mortality among patients with ARDS.

The investigators concluded that further assessment of the current model is warranted, including additional prospective studies with more recent patient data.


Ye W, Li R, Liang H, et al. Development and validation of a clinical risk model to predict the hospital mortality in ventilated patients with acute respiratory distress syndrome: a population-based study. BMC Pulm Med. Published online July 11, 2022. doi:10.1186/s12890-022-02057-0