LAUSR

Introduction/Hypothesis: Limitations in the rapid collection of patient-level data during public health emergencies remain a national strategic vulnerability. Over a 3-year period, we developed a consensus, electronic case report form (eCRF) and tested the feasibility of nationwide data collection for intensive care unit (ICU) patients with influenza. Methods: Twelve hospitals in the US implemented an observational protocol to collect data on ICU patients with laboratoryconfirmed influenza over 2-week periods. In year 1, data were collected retrospectively over hospital-identified timeframes. In years 2-3, data were collected prospectively during designated timeframes. In year 1, 80 clinical data elements were collected. In years 2-3, 151 clinical data elements were collected in a 2-tier system, with tier-1 data collected within 48-hours of admission and tier-2 data within 14 days. Descriptive statistics were calculated. Results: Over 3 years, data for 198 patients were collected across 12 sites (53% male, 91% adult). Year 1 data were retrospectively captured (n=74); year 2 (n=70) and year 3 (n=54) data were prospectively captured. 64% were mechanically ventilated, with consistency over the 3 years. 11% were on ECMO (11%, 4%, 19% by year). 90% received antiviral therapy within 48-hours (oseltamivir 98%). 87% received antibacterial drugs (93%, 82%, 85% by year). Antifungals were administered to 16% and corticosteroids to 41%. 38% were discharged home (32%, 50%, 30% by year). 21% remained in the hospital (19%, 19%, 28% by year). 16% died (20%, 9%, 20% by year). Conclusions: Data collection method and flu season timing varied over the 3-year period, but consistency of results across time suggest several lessons. A consensus eCRF for influenza has been developed that can be broadly applied with consistent data collection allowing for clinically relevant comparisons across time and location. There were important variations in treatment of lifethreatening influenza that could have affected outcomes over time; the reasons for treatment variance are not known and require study. For example, mechanical ventilation as supportive care was consistent throughout the 3-year period, while ECMO rates and treatment with corticosteroids varied. Next steps are automated data extraction and expanded data collection windows.