8%) and the outpatient cohort (25.8%) was statistically similar as well (P = .9) ( Fig. 2). Length of stay was significantly decreased in the <3-day group at 6.1 days (95% CI, 5.3-6.9) versus 10.3 days in the >3-day group (95% CI, 8.9-11.7) (P < .0001). Eight patients

had a length of stay > 20 days secondary to other comorbidities: 3 from the <3-day group and 5 from the >3-day group. These patients with a longer length of stay because of other comorbidities were excluded from the final results so data were more representative of length of stay because of OOGIB. We had analyzed the data both including and excluding these 8 patients. In both circumstances there was a significant difference in the length of stay between the two inpatient groups. We decided to take a conservative approach by excluding these outliers AZD2281 concentration to minimize any confounders. In this retrospective analysis of the use of VCE performed for OOGIB in both inpatients and outpatients, we demonstrated that the

early deployment of VCE results in a higher diagnostic yield and increased rate of therapeutic intervention. In turn, early deployment was associated with a significant reduction in length of stay, possibly associated with the increased intervention rate and reduction of the numbers of other procedures. learn more Statistically, the overall diagnostic yield of VCE was not different for the inpatient and outpatient populations (P = .054), even though the difference of yield between these two stood at 12.5%. This is likely because a significant proportion (37.5%; 54 of 144) of the these VCEs for inpatients was performed 3 days after admission, thus decreasing the overall yield for the inpatient population. This dilutional effect on the yield is supported by the statistical similarity for a positive yield between

the patients who received VCE 3 days after admission and those who had VCE done as outpatients. A significant increase was found in the diagnostic yield if VCE was performed within 3 days of admission for OOGIB. Detection of active bleeding by VCE declined progressively as days passed after admission (Fig. 4), consistent with the natural history of GI bleeding, which has a tendency to spontaneously cease with time. Presence of active bleeding or detection of angioectasia led to targeted interventions in all 3 groups. Overall, a significant increase in targeted interventions was performed for patients in the <3-day group commensurate with the overall higher diagnostic yield of VCE in this cohort. Non–small-bowel source of bleeding (stomach or colon) was noted to be higher in the inpatient (9%) than the outpatient (3.4%) population. This discrepancy may be explained by the fact that detection of vascular lesions may be subject to hemodynamic compromise and sedation use during the initial urgent endoscopic evaluation. Poor preparation of colon may be another contributing factor for missing significant findings during colonoscopy.