On average, time between admission and colonoscopy was between 2 and 3 days (median, 62.2 hours). In 3 out of 4 patients (77%), etiology of LGIB was confirmed; diverticular bleeding was most common (39%), followed distantly by hemorrhoidal bleeding (15%).
Compared with milder cases, patients with severe bleeding were more likely to stay in the ICU (49% vs. 19%; P less than .0001), have a blood transfusion (85% vs 36%; P less than .0001), and need to remain in the hospital for a longer period of time (6 days vs. 4 days; P = .0009). These findings exemplify the high level of resource utilization required for LGIB and show how severe bleeding dramatically compounds intensity of care.
Further analysis showed that none of the seven risk models were predictive across all outcomes; however, some predicted specific outcomes better than others. Leaders were the Glasgow-Blatchford score for blood transfusion (AUC 0.87; P less than .0001), the Oakland score for severe bleeding (AUC 0.74; P less than .0001), the Sengupta score for ICU stay (AUC 0.74; P less than .0001), and the Strate score for both recurrent bleeding during hospital stay (AUC 0.66; P = .0008) and endoscopic intervention (AUC 0.62; P = .01).
The investigators noted that the Glasgow-Blatchford score, which also is used in cases of UGIB, has previously demonstrated accuracy in predicting blood transfusion, as it did in the present study, suggesting that, “[i]n instances where there may be uncertainty of the origin of the bleeding, the Blatchford score may be a preferential choice of risk score.”
“Overall, we found that no singular score performed best across all the outcomes studied nor did any score have an extremely strong discriminatory power for any individual variable,” the investigators wrote, concluding that “... simpler and more powerful prediction tools are required for better risk stratification in LGIB.”
The investigators reported no financial support or conflicts of interest.
*This story was updated on Jan. 31, 2019.
SOURCE: Tapaskar N et al. Gastrointest Endosc. 2018 Dec 18. doi: 10.1016/j.gie.2018.12.011.