(1478) Characterizing Sepsis-Induced Hypotension Patients Who Benefit From an Early Vasopressor Strategy

Introduction: Intravenous fluids and vasopressors are used to achieve hemodynamic stability and improve end-organ perfusion in sepsis-induced hypotension and septic shock. The recent Crystalloid Liberal or Vasopressors Early Resuscitation in Sepsis (CLOVERS) trial compared a restrictive fluid (with early vasopressor use) to an early liberal fluid strategy and found no differences in mortality. Our investigation applies a novel machine learning approach to identify clinical characteristics in patients who would benefit from either strategy.

Methods: We performed a retrospective observational study of all encounters at Emory University Hospital from 2014-2021 who met Sepsis-3 criteria, received antibiotics within 6 hours from Sepsis-3 time, and ≥1L of crystalloid fluids. Patients receiving vasopressors prior to 3L of crystalloid fluids were considered the “restrictive fluid” (RF) group. Patients receiving 3L of crystalloid fluids prior to vasopressors (or no vasopressors) were labeled the “liberal fluid” (LF) group. Using in-hospital mortality as the outcome, we applied a doubly robust learner algorithm to estimate the conditional average treatment effects (CATE) of RF vs LF. Patients in the top 10% benefited most from fluids (“fluid responsive” [FR]). Patients in the bottom 10% benefited most from vasopressors (“vasopressor responsive”, [VR]).

Results: A total of N=28634 patient encounters met inclusion criteria (Training n=20,000; validation n=8634). In the validation cohort, 1753 were in the RF group, and 6881 in the LF group. The average treatment effect (ATE) of RF vs. LF was 0.01±0.07, indicating better treatment effect for LF overall. The ATE for the VR group was -0.11±0.04, while the ATE for FR group was 0.16±0.05. Compared to FR, VR had higher mean lactic acid (2.7±2.6 vs 2.6±3.1 mmol/L), blood urea nitrogen (40.1±31.8 vs 30.1±24.0 mg/dL), creatinine (3.1±3.7 vs 1.9±1.9 mg/dL), total bilirubin (2.4±5.7 vs 1.4±2.3) and lower bicarbonate (22.2±5.6 vs 23.2±6.3 mmol/L), all p< 0.001.

Conclusions: In sepsis induced hypotension and septic shock, patients with a higher severity of end organ injury may benefit from receiving early vasopressors.