(1307) Characterizing Elderly Sepsis Patients With COPD and Anemia: A Learning Health System Initiative

Introduction: COPD, especially comorbid with anemia, increases mortality risk in patients with sepsis. We characterized anemia in COPD patients at least 65 years old who underwent Surviving Sepsis Campaign guided intensive care in a Five-Star Medicare Hospital between January 1, 2020 and February 29, 2024.

Methods: Clinicodemographic data within 24h of presentation was extracted from EMRs under IRB exemption. Pathway to ICU from ED was direct or indirect within < 72h from general care or telemetry units. Anemia was defined as hemoglobin [Hb] concentration of < 12 g/dL for females and < 13 g/dL for males. Gold staging was missing for some patients, so COPD severity was estimated using Red Cell Index (RCI). Youden Index defined optimal mortality cut points for Hb across sepsis levels. Continuous vs. categorical variables respectively were summarized as mean (SD) vs. proportions (%) compared using ANOVA vs. chi square test. A two-tailed p< 0.05 was considered significant controlling for COVID-19, renal replacement and RCI.

Results: Among candidate sepsis (n=907) and septic shock (n=1638) patients, 120 and 232 respectively met inclusion criteria. Isomorphic demographics and comorbidities across sepsis levels permitted pooled statistical summary including age 77 +/- 8 years across females (48%) and males (52%) distributed by Whites (95%), Blacks (2%) and other races (3%) exhibiting hypertension (84%), heart failure (54%), coronary artery disease (49%), iron deficiency anemia (44%), renal failure (43%), diabetes (39%) and obesity (28%). Presentation anemia was 67% (sepsis) and 72% (septic shock) (p=0.39). Mortality risk cut point for Hb (g/dL) was 10.8 (sepsis) vs. 8.3 (septic shock) (p < 0.001). Cut points cumulatively predicted 82% of patients who died. There were no intergroup differences in RBC transfusion rates (17%) or hospital length of stay (9.2 +/- 7.0) days. Hospital mortality was for sepsis (9%) vs. septic shock (24%) (p < 0.001).

Conclusions: Although present initiative was not designed to unravel mechanisms of anemia, learning highly predictive mortality cut points derived from real-world evidence generated using nationally recognized standards of practice warrants prospective confirmation. Despite relatively small sample, strength of findings derives from demographic and polymorbidity homogeneity.