1389: Predictors of Poor Outcomes in Pediatric Sepsis-Induced Ards: Prospective Cohort Study From an Lmic

Introduction: Pediatric ARDS (pARDS), often triggered by sepsis, carries a high mortality burden, particularly in resource-limited settings. While 75% of pediatric ARDS cases are sepsis-related, most predictive data derive from high-income countries. We hypothesized that simple clinical and laboratory markers collected early during ICU admission could predict poor outcomes in children with sepsis-induced ARDS in a lower-middle-income country (LMIC) setting. Methods: This was a 3-year prospective cohort study (2021–2023) conducted in the PICU at Aga Khan University Hospital, Karachi, Pakistan. Children aged 1 month to 18 years with sepsis were enrolled and monitored for ARDS development and mortality. Demographic data, severity scores (PELOD-2, VIS), and laboratory parameters (platelet count, creatinine, lactate) were recorded. ARDS was diagnosed based on PALICC criteria. Logistic regression was used to identify independent predictors of ARDS development and mortality. Results: Out of 128 septic patients, 44.2% developed ARDS within 96 hours, and overall mortality was 23.4%. Non-survivors were more likely to be immunocompromised (30% vs. 13.3%, p=0.03). Significant predictors of mortality included higher VIS at 24 hours (0.84 vs. 0.28, p< 0.001), elevated PELOD-2 scores (≥4 in 21.4% of non-survivors vs. 8.6% survivors), thrombocytopenia (117 vs. 193×103/µL, p=0.044), elevated creatinine (1.70 vs. 0.87 mg/dL, p=0.015), and increased lactate levels (5.11 vs. 3.67 mmol/L, p=0.007). Higher FiO2 requirements were also observed in non-survivors (p=0.023). On multivariate analysis, only lactate on Day 1 independently predicted ARDS development (OR 1.21, p=0.018). Conclusions: In LMIC PICUs, simple early markers—especially lactate, platelet count, and VIS—can aid in identifying children at high risk for ARDS and mortality. These low-cost indicators can guide early escalation in care, improve triage, and help validate context-specific severity scoring tools in resource-constrained settings.