Developing a Scoring System Predicting Inpatient Mortality for Critically Ill Patients With ARDS

Introduction: Acute Respiratory Distress Syndrome (ARDS) is a life-threatening condition characterized by sudden lung edema and severe hypoxemia, often requiring ICU admission and mechanical ventilation. The Berlin Definition classifies ARDS by PaO₂/FiO₂ ratios into mild, moderate, and severe, with mortality rising from 27% in mild to 45% in severe cases. Despite advances in supportive care, mortality remains high. While general scores like APACHE II and qSOFA exist, they do not reflect the ARDS-specific pathophysiology. An ARDS-specific scoring tool could help guide clinical decisions and align care with patient goals.


Methods: We used the National Inpatient Sample (NIS) from 2016 to 2020 to identify adult patients diagnosed with ARDS (ICD-10 code J80). A multivariate logistic regression was performed using SPSS to identify ICD-10 codes most strongly associated with in-hospital mortality. Odds ratios were calculated using Euler’s constant (e^β). Statistically significant variables (p < 0.05) included acute kidney injury (AKI), viral pneumonia, systemic inflammatory response syndrome (SIRS)/sepsis, atrial fibrillation/flutter, brain dysfunction, and coronary artery disease (CAD). Each condition was treated as a binary variable and translated into a point-based scoring system by scaling each log(OR) relative to the smallest meaningful coefficient (CAD, log(OR) = 0.195).


Results: AKI showed the strongest mortality association (OR 2.38; p < 0.001), followed by viral pneumonia (OR 2.17), and SIRS/sepsis (OR 1.92). Brain dysfunction (OR 1.52) and atrial fibrillation/flutter (OR 1.51) were also significant, with CAD having a smaller yet meaningful association (OR 1.22). Points were assigned as follows: CAD (1), atrial fibrillation/flutter and brain dysfunction (2 each), SIRS/sepsis (3), and viral pneumonia and AKI (4 each).


Conclusions: We developed an ARDS-specific mortality risk score using nationally representative data. By quantifying key predictors of death, this tool can help clinicians identify high-risk patients and initiate early, informed goals-of-care conversations with families. Future prospective validation in hospital settings is necessary to assess its real-world utility and clinical impact.