Simons Ashoka Early Career Fellow Ashoka University
Disclosure information not submitted.
Introduction: Immunoparalysis is an emerging concept, characterized by increased susceptibility to infections due to a persistent anti-inflammatory response. This is important in ICU disorders, such as ARDS, where dysregulated physiology is common and compounded by interventions such as mechanical ventilation and steroids. An earlier study in a small cohort demonstrated increased susceptibility in ARDS to late onset ventilator associated pneumonia. However, till date, clear evidence of this increased susceptibility remains lacking. We leverage the MIMIC-IV database in a non ideal, pragmatic fashion to explore inherent susceptibility of ARDS patients to infections in the immediate resolution period, and attempt identification of phenotypic heterogeneity driving this phenomenon
Methods: 1. In a retrospective analysis using the MIMIC-IV database, mechanically ventilated ICU admissions were chosen, with ARDS patients identified using ICD codes. ARDS resolution was pragmatically selected as return of PO2/fiO2(P/F) ratio to above 300 (t= 0) 2. Matched non-ARDS admissions were identified using Nearest neighbour propensity score matching with caliper thresholds on demographics and co-morbidities after chronological synchronisation (2192 admissions each). 4 sub groups were also identified 3. A 7-day right censored window was used for time to event analysis. Interventions such as antibiotic use, steroid use, mechanical ventilation, and RRT were matched in a second step using a logistic regression approach.
Results: 1. A Markov Chain Monte Carlo approach demonstrated significantly increased susceptibility of the ARDS group, validated using log rank test 2. Kaplan-Meier analysis adjusted using inverse probability of treatment weighting (IPTW) for four subgroups showed statistically significant differences (p < 0.005), with one showing a less pronounced difference (p = 0.01)
Conclusions: Our results build on previous evidence on Ventilator Associated Pneumonia (VAP) in ARDS, whilst demonstrating its extension into the immediate resolution phase, using EHR data. Our current efforts focus on external validation which is showing similar trend, identification of the predominant causative organisms, and specific host factors contributing to each subgroup. This would be important to understand the heterogeneity of ARDS.
