Do Predictors in Upper GI Bleeding Include A Weekend Time of Admission?
There have been marked evolutions in the management of patients with nonvariceal upper gastrointestinal bleeding (NVUGIB), and more specifically peptic ulcer bleeding, over the past 5 to 7 years that include the introduction of efficacious therapies such as endoscopic hemostasis with a single modality (preferably not just the sole injection of epinephrine), or a combination of methods, followed by high-dose intravenous proton pump inhibition. Regardless, early risk stratification remains a cornerstone of appropriate management and the explicit use of validated scoring schemes is encouraged.

Many prognosticating scales exist to predict mortality in patients with NVUGIB and work is ongoing in this area. The Rockall score is probably the most widely known risk-stratification tool for upper-gastrointestinal hemorrhage and has been validated in numerous health care settings as an accurate predictive tool of both rebleeding and death. The complete Rockall score makes use of both clinical and endoscopic criteria with a score ranging from 0 to 11 points, with higher scores indicating higher risk. A predictive clinical Rockall score (ie, the score before endoscopy) also can be calculated solely on the basis of clinical variables at presentation. The current issue of Clinical Gastroenterology and Hepatology adds to the growing body of knowledge with 3 studies that examine specific determinants of dying from UGIB including the presence of Helicobacter pylori, and the time of admission to the hospital (weekdays vs weekends).

Most predictive models of death in NVUGIB include a combination of patient characteristics and endoscopic findings, with or without additional therapeutic-related factors. Chiu et al, in the current issue, report on a large single-center experience. A total of 3854 patients were eligible for inclusion in both a derivation and a subsequent validation cohort. Of note, the group specifically focused their analysis on patients with high-risk endoscopic lesions requiring endoscopic therapy. This Asian center is known to provide care with great expertise and in fact 23.9% of patients included in the cohort were randomized in now-completed endoscopic or pharmacotherapeutic clinical trials. Overall, 7.1% of these high-risk patients died and significant predictors of death were age (>70 y), the presence of comorbidity, more than one listed comorbidity, hematemesis at presentation, systolic blood pressure lower than 100 mm Hg, in-hospital bleeding, rebleeding, and a need for surgery. Interestingly, an H pylori-related ulcer was associated significantly with a lower risk of mortality, a finding that may be attributable to an enhanced acid-suppressive effect of proton pump inhibitors (PPIs) in this patient subgroup, or to an associated younger age among such patients according to Chiu et al. Yet, one probably cannot completely exclude confounding attributable to an unfavorable influence of anti-inflammatory, aspirin-related, or idiopathic ulcers. Chiu et al showed that the performance of their predictive scale was numerically superior than that of the Rockall score, and also suggested that the nonendoscopic components of the model may be useful as a pre-endoscopic predictive tool. Perhaps surprising was the lack of significant influence attributable to the method of hemostasis and use of high-dose PPIs. Yet, the study spanned many years of evolving practice and may have been contaminated in part by the ongoing randomized trials, a mix of patients starting to bleed in an out-patient versus an in-patient setting, 2 distinct groups of patients that may each be associated with differing determinants of death, and perhaps was statistically underpowered. Regardless, the data presented by Chiu et al are clinically useful in allowing for an improved stratification of patients at risk of dying that should guide clinical management.

The current issue also contains 2 very interesting large cohort studies by Ananthakrishnan et al and Shaheen that assessed the importance of time of admission to the hospital in an attempt to identify a possible "weekend" effect leading to poorer outcomes, as has been suggested previously in gastroenterologic and other conditions.

Both analyses suggested that weekend admissions for NVUGIB or peptic ulcer bleeding are associated with excess mortality with similar respective odds ratios of 1.21 (95% confidence interval [CI], 1.09-1.35) and 1.08 (95% CI, 1.02-1.15), yielding in one study an unadjusted number needed to harm of 143 (ie, one excess death for a weekend admission for every 143 discharges). Patients admitted during the weekend also had a greater risk of death on subsequent days.

Additional predictors of mortality also were noted that included patient-level characteristics (older age, white race, nonprivate health insurance, an increasing number of comorbid conditions, endoscopic or surgical interventions, and transferred patients) as well as institutional determinants (mortality was higher in the northeastern United States, and urban teaching or nonteaching centers).

Importantly, the overall mortalities noted in these studies were similar and quite low at 3.1% to 3.2%. The study by Shaheen assessed a range of study periods (focusing on peptic ulcer bleeding), further showing a significant decrease in mortality over time with patients admitted during the later years of the study (2000-2005 vs 1993-1999) at lower risk of death (odds ratio, 0.75; 95% CI, 0.71-0.79). Although some regional registries have not found a significant decline over time, other single time point estimates obtained by recent national registries in Canada and the United Kingdom appear to support the former conclusions.

In the Shaheen study, weekday admissions also resulted in shorter, cheaper hospital stays, although both studies showed a delay in the performance of endoscopy for weekend admissions. Patients with a primary diagnosis of acute variceal hemorrhage also showed a weekend effect with a lower use of early endoscopy, but no survival difference when these patients were admitted to nonteaching hospitals.

The Nationwide Inpatient Sample is the largest all-payer database of national hospital discharges (∼8 million/y). It represents a 20% stratified random sample of nonfederal acute-care hospitals in the United States. Although both studies extracted data from the Nationwide Inpatient Sample database, small but important differences exist in the study methodologies with regards to year of sampling, and patient selection criteria, to name a few, that can explain some differences in secondary conclusions. It is true that these types of large administrative databases typically may be influenced by unrecognized confounding (especially lingering uncertainty about disease severity differences between patients admitted on weekdays vs weekends), however, these data are quite suggestive of a weekend effect. It may be attributable to many considerations including patients presenting later in the course of the disease, or systems issues (such as the availability and quality of supportive care, with disparities in staffing pattern on weekends, for physicians, nurses, and other support staff as noted previously; the availability of an on-call GI bleed team or endoscopy laboratory staff; and the presence of trainee programs).

The lack of impact attributable to endoscopy or PPI therapy on mortality may be attributable to the absence of specific data on the type of endoscopic hemostasis performed or pharmacotherapy received (with also older management practices captured for a large part of the data used in the Shaheen study). However, Ananthakrishnan et al did note that within the smaller group of patients who received endoscopic intervention, there was no difference in mortality between weekend and weekday admissions, indirectly suggesting some benefit attributable to endoscopy. This study also confirmed the results of previous randomized trials suggesting that early endoscopy, although exerting no direct effect on patient outcomes, has been associated with shorter and cheaper lengths of stay.

What can we, as clinicians, take back to our practices from this plethora of probing new data? First, that clinical and endoscopic determinants are powerful predictors of subsequent mortality, emphasizing the importance of early risk stratification. Second, that the mortality, contrary to what often is stated, may in fact have come down in recent years from the traditional 10% quoted figure. Third, that this improvement likely is owing to enhanced resuscitative and supportive measures (perhaps in the form of selected institutional resources) that may not be as readily available on weekends, although it also is likely that endoscopic therapy has an important role to play, as does profound acid suppression, based on summary randomized controlled trials data. Fourth, that the optimal timing of the diagnostic/therapeutic endoscopy requires further study, however, at the very least, early endoscopy (≤24 h) results in shorter and cheaper hospital stays. These realizations, obtained from both randomized controlled trials and real-life data sources, should help convince clinicians and decision makers alike that adequate resources need to be provided to allow for competent resuscitation, risk stratification, early endoscopy, the availability of timely skilled endoscopic intervention, and appropriate PPI therapy—all of which should be coordinated through a collaborative multidisciplinary group that includes primary care physicians, emergency room physicians, internists, intensivists, surgeons, gastroenterologists, and, more recently, interventional radiologists in selected cases.