Abstract 14717: Identifying Predictors of Red Cell Distribution Width at Admission and Changes in RDW During Heart Failure Hospitalization

Introduction: The causes of risk-associated elevations in red cell distribution width (RDW) are unknown, but RDW strongly predicts adverse outcomes in heart failure (HF) patients. Initial RDW level and the change in RDW (ΔRDW) were previously shown to predict hospital length of stay, 30-day readmission, and 30-day mortality. Hypothesis: Risk factors and diagnoses predict initial RDW and in-hospital ΔRDW of HF patients. Methods: Patients (N=6,414) were studied if they had an inpatient stay for the primary diagnosis of HF during 2004-2013, had RDW measured within 24 hours after admission, and had a second RDW tested within 24 hours prior to discharge. ΔRDW was calculated as the difference between the two RDW measurements. All patients were aged 65 years or older and were discharged alive and not to hospice. Results: In linear regression, predictors of initial RDW included age, being ever diagnosed with 15 comorbidities (Table), and diagnosis in the prior month with 8 comorbidities. Those 24 factors explained 10.5% of the variation in RDW (model F-statistic= 33.3, p<0.001). In contrast, predictors of ΔRDW were sex, length of stay, initial red blood cell count, mean corpuscular volume, RDW, mean corpuscular hemoglobin concentration, sodium, calcium, potassium, medications administered in-hospital (e.g., ACE-I, inotropes, and vasopressors), and diagnosis ever or within the last month of various comorbidities (Table). Those factors explained 13.9% of the variation in ΔRDW (model F-statistic= 35.9, p<0.001). Conclusions: Predictors of initial RDW and ΔRDW in HF patients clustered among diagnoses of cardiovascular, pulmonary, renal, and psychological diseases, and (as expected) hematological/bleeding conditions. These findings may aid in determining the cause of RDW elevations and in personalizing medical care using common, inexpensive, electronic laboratory data. With most variance in RDW and ΔRDW still unexplained, discovery of other predictors is needed.