Comparison of the Predictive Capacity of Oxygenation Parameters, Oxygenation Indices, and CURB-65 to Mortality, Mechanical Ventilation, and Vasopressor Support in Community-Acquired Pneumonia at Different Altitudes

Background: Populations residing at high altitudes display distinct physiological adaptations that are essential for understanding respiratory diseases. However, there is limited research on how these adaptations affect the assessment and prognosis of community-acquired pneumonia (CAP). Methods: A prognostic validation nested within a retrospective cohort was conducted on subjects with pneumonia admitted to two high-complexity institutions in Colombia at different altitudes above sea level. The receiver operating characteristic (ROC) curves were calculated for SaO ₂, PaO ₂, SpO ₂, A-a O ₂ gradient, a-A index, PaO ₂/FiO ₂, SpO ₂/FiO ₂, and the CURB-65 score to predict 30-day mortality, requirement for invasive mechanical ventilation (IMV), and need for vasopressor support. Results: 3467 were selected for analysis, with 73.7% (2557/3467) residing at high altitudes and 26.2% (910/3467) at low altitudes. The CURB-65 score ≥ 2 showed a performance in predicting mortality of 0.707 (95% CI: 0.653–0.761; p < 0.001) at low altitudes and 0.737 (95% CI: 0.709–0.765; p < 0.001) at high altitudes. The PaO ₂/FiO ₂ ≤ 300 showed a performance in predicting the need for IMV and vasopressor support of 0.734 (95% CI: 0.685–0.783; p < 0.001) and 0.724 (0.674–0.775; p < 0.001) at high altitudes, respectively. The SpO ₂/FiO ₂ ≤ 350 showed a performance in predicting the need for IMV of 0.679 (0.507–0.85; p < 0.001) at low altitudes. The A-a O ₂ gradient ≥ 10 showed a performance in predicting the need for vasopressor support of 0.686 (95% CI: 0.537–0.835; p = 0.06) at low altitudes. Conclusion: In patients with CAP at altitudes above 2500 m above sea level, PaO ₂/FiO ₂, SpO ₂/FiO ₂, and the A-a O ₂ gradient show a greater predictive capacity for 30-day mortality, need for IMV, and vasopressor requirements. The CURB-65 score showed a good predictive performance.