2897 Prognostic Factors for Identifying Etiology of Subacute Drops of Hemoglobin in Hospitalized Patients Without Overt Gastrointestinal Bleeding

Abstract

INTRODUCTION: Studies have shown hospitalized patients have a decrease in hemoglobin (Hb) concentration with no overt blood loss. GI consultation and endoscopy are routinely requested for a decrease in Hb. In this study, we retrospectively reviewed hospitalized patients who had a subacute drop in hemoglobin without evidence of GI bleeding. Our aim was to identify prognostic factors that could be predictive of endoscopic findings. METHODS: Electronic health records were obtained to identify patients admitted to our hospital from October 2016 to May 2018 that had undergone an endoscopic procedure during their hospitalization. Next, we excluded those with an admitting diagnosis of a gastrointestinal bleed (hemorrhage, melena via ICD10 code). We then manually identified those with a subacute drop in Hb, defined as a drop in Hb ≥ 3 from baseline admission. This resulted in a cohort of 113 patients. Hb changes for each patient were examined and the maximum drop in Hb was categorized into four groups: minimal drop was defined as 3-3.5 mg/dL, moderate drop was 3.6–5.5 mg/dL, high drop 5.6-7.5 mg/dL and very high drop >7.5 mg/dl. Endoscopic procedure outcomes for each patient were analyzed and grouped into either category “bleed” or “no bleed”. “Bleed” was categorized as an ulcer, frank bleeding, varices and/or neoplasm which are lesions that can be the cause of drop in Hb. Normal endoscopy, gastritis, duodenitis, and/or polyps were categorized as “no bleed” which are unlikely the cause in drop in Hb. RESULTS: Sixty-two patients (56%) belonged to the "no-bleed group" and fifty-one patients (44%) to the “bleed” group. Patients with minimal Hb drop were found to be predominantly in the “no-bleed” group, 85% compared to 15% for the “bleed” group. There were equal numbers of "bleed" and "no bleed” patients in both the moderate and high Hb drop groups. For very high Hb drop >7.5, the number of patients in the "bleed" group were greater than "non-bleed" group, 71% vs 29%. Comorbidities such as HTN, Type 2 Diabetes, CAD, Liver disease, CVA, GERD, and Kidney disease were also considered in association with a subacute drop in Hb and outcomes of bleed vs. no bleed. Using multivariate modeling, patients with Diabetes or those receiving packed RBCs were found to have a statistically significant predictor of "bleed", P-value 0.047 and 0.044, respectively. CONCLUSION: For the majority of patients, the primary reason for a subacute drop in hemoglobin is multifactorial and cannot be unambiguously attributed to GI bleed.