Abstract
BACKGROUND: Gastrointestinal bleeding (GIB) is a common complication of left ventricular assist device (LVAD) therapy accounting for frequent hospitalizations and high resource utilization. METHODS: We previously developed an endoscopic algorithm emphasizing upfront evaluation of the small bowel and minimizing low-yield procedures in LVAD recipients with GIB. We compared the diagnostic and therapeutic yield of endoscopy, health-care costs, and re-bleeding rates between conventional GIB management and our algorithm using chi-square, Fisher's exact test, Wilcoxon-Mann-Whitney, and Kaplan-Meier analysis. RESULTS: We identified 33 LVAD patients with GIB. Presentation was consistent with upper GIB in 20 (61%), lower GIB in 5 (15%), and occult GIB in 8 (24%) patients. Forty-one endoscopies localized a source in 23 (56%), resulting in 14 (34%) interventions. Algorithm implementation compared with our conventional cohort was associated with a 68% increase in endoscopic diagnostic yield (P< .01), a 113% increase in therapeutic yield (P= .01), a 27% reduction in the number of procedures per patient (P < .01), a 33% decrease in length of stay (P < .01), and an 18% reduction in estimated costs (P < .01). The same median number of red blood cell transfusions were used in the 2 cohorts, with no increase in re-bleeding events in the algorithm cohort (33.3%) compared with our conventional cohort (43.7%). CONCLUSIONS: Our endoscopic management algorithm for GIB in LVAD patients proved effective in reducing low-yield procedures, improving the diagnostic and therapeutic yield of endoscopy, and decreasing health-care resource utilization and costs, while not increasing the risk of a re-bleeding event.