(1019) Clinical Trial Emulation as a Method for Causal Inference in Critical Care

Introduction: Real-world data (RWD), such as electronic health records (EHRs), have potential to augment critical care research through use of causal inference methods. Trial emulation is one such method where design features from randomized controlled trials (RCT) are incorporated to reduce systematic biases encountered in observational studies. This methodology is a good fit for critical care where clinician judgment is often a determinant of treatment received.

The CURE Drug Repurposing Collaboratory, in collaboration with a diverse group of stakeholders, developed a collection of openly available projects, the Edge Tool Suite, to convert proprietary EHR data into a common data model and facilitate observational research across multiple healthcare systems. We used data extracted through the Edge Tool Suite to compare the results of an observational trial emulation to the results of the RECOVERY trial of dexamethasone for treatment of COVID-19.

Methods: We recreated trial features within the observational data obtained from over 10,000 records of patients hospitalized with acute COVID-19 from the Edge Tool pilot site. Propensity score matching was used to match patients 2:1 without and with any recorded administration of dexamethasone. Covariates included demographic, comorbidity, and disease severity variables. Our primary outcome measure was 28-day mortality, with secondary outcome measure of hospital length of stay.

Results: In our trial emulation, use of dexamethasone was associated with significantly lower 28-day mortality among both those receiving oxygen alone (HR: 0.29, CI: 0.23-0.36, p < 0.001) and those on mechanical ventilation (HR: 0.26, CI: 0.20-0.34, p < 0.001), compared to those receiving no respiratory support. This finding aligns with the conclusions drawn from the RECOVERY trial.

Conclusions: This analysis suggests that EHR data can be used to emulate trails to the effectiveness of treatments within an RCT. Trial emulation can be a powerful tool to evaluate whether the results of RCT are present in real world treatment and may provide information on post-hoc analysis of treatment effectiveness within patient groups, or in hypotheses generation for future trials.