Weekend Effect on Clinical Outcomes of Acute Lower Gastrointestinal Bleeding: A Large Multicenter Cohort Study in Japan

Introduction: Weekend admissions showed increased mortality in several medical conditions. This study aimed to examine the weekend effect on acute lower gastrointestinal bleeding (ALGIB) and its mortality and other outcomes. Methods: This retrospective cohort study (CODE BLUE-J Study) was conducted at 49 Japanese hospitals between January 2010 and December 2019. In total, 8,120 outpatients with acute hematochezia were enrolled and divided into weekend admissions and weekday admissions groups. Multiple imputation (MI) was used to handle missing values, followed by propensity score matching (PSM) to compare outcomes. The primary outcome was mortality; the secondary outcomes were rebleeding, length of stay (LOS), blood transfusion, thromboembolism, endoscopic treatment, the need for interventional radiology, and the need for surgery. Colonoscopy and computed tomography (CT) management were also evaluated. Results: Before PSM, there was no significant difference in mortality (1.3% vs. 0.9%, p = 0.133) between weekend and weekday admissions. After PSM with MI, 1,976 cases were matched for each admission. Mortality was not significantly different for weekend admissions compared with weekday admissions (odds ratio [OR] 1.437, 95% confidence interval [CI] 0.785–2.630; p = 0.340). No significant difference was found with other secondary outcomes in weekend admissions except for blood transfusion (OR 1.239, 95% CI 1.084–1.417; p = 0.006). Weekend admission had a negative effect on early colonoscopy (OR 0.536, 95% CI 0.471–0.609; p < 0.001). Meanwhile, urgent CT remained significantly higher in weekend admissions (OR 1.466, 95% CI 1.295–1.660; p < 0.001). Conclusion: Weekend admissions decrease early colonoscopy and increase urgent CT but do not affect mortality or other outcomes except transfusion.

Patients with ALGIB may be more susceptible to weekend admission than those with UGIB because colonoscopy requires bowel preparation.Even in ALGIB, the early colonoscopy rate was low on weekends, but there was no difference in mortality [25].However, there is only one report on the weekend effect on ALGIB outcomes [25], and the weekend effect on ALGIB remains unknown.The report did not consider detailed clinical information such as symptoms, vital signs, laboratory tests, or medication history.Therefore, clinical information such as body mass index and nonsteroidal antiinflammatory drugs (NSAIDs) related to ALGIB mortality was not also considered [26].
We hypothesized that weekend admission for ALGIB would be associated with high mortality, given the clinical course of ALGIB, and would affect other outcomes.Therefore, we evaluated the weekend effect on the mortality rate and other outcomes of ALGIB using largescale data from a Japanese multicenter cohort study that collected a great amount of clinical information.We also evaluated the weekend effect on colonoscopy and computed tomography (CT) management.

Materials and Methods
This multicenter retrospective cohort study was conducted between January 2010 and December 2019 at 49 centers across Japan.The data collection period was set to begin in January 2010 because many hospitals have implemented electronic medical records, which reduces the amount of missing information.And the data collection period was set to December 2019 because the study protocol was created in January 2020.The 10-year time period was set because there was consensus at several meetings that the above time period was necessary to enroll at least 100 cases per institution in this study.This allowed hospitals with fewer beds to participate in the study.Endoscopy and CT were available for 24 h in all the hospitals.However, not all of the participating hospitals were tertiary hospitals, as some of the smaller hospitals in Japan can provide a 24 h colonoscopy and CT.Details of this study have already been reported in the COlonic DivErticular Bleeding Leaders Update Evidence from multicenter Japanese Study (CODE BLUE-J Study) [27,28].It was decided that the CODE BLUE-J Study would collect at least 10,000 cases, and it was agreed at several meetings that this would require collecting at least 100 cases per institution.This study received ethical approval from the relevant ethics committees and institutional review boards, and an opt-out method was included in all participating hospitals (online suppl.Table 1; for all online suppl.material, see https://doi.org/10.1159/000533744).A total of 10,342 patients (≥20 years) admitted due to acute hematochezia were reviewed.Among them, 8,120 patients were targeted, excluding those diagnosed with UGIB, for the first admission (Fig. 1).They were divided into weekend admissions and weekday admissions, and outcomes were compared between the two groups.Weekend admission was set from Friday at midnight to Sunday at midnight, with reference to previous studies [1,4,6,8,10,20,24].

Clinical Data Assessment
We collected data on hospital type (academic or nonacademic), hospital size (large or small), age, sex, lifestyle factor, and Charlson Comorbidity Index (CCI) [29], which is widely used and has been validated for use in gastrointestinal bleeding research [30][31][32].Data on drug use within 30 days of admission (NSAIDs, antiplatelets, and anticoagulants), Eastern Cooperative Oncology Group performance status (PS) [33], symptoms, vital signs, relevant laboratory tests ordered, and diagnosis were also evaluated.Academic hospitals were defined as national centers or university hospitals.A large hospital was described as a hospital with more than 765 beds, which is the median number of beds in this study.Shock was defined as a systolic blood pressure of ≤100 mm Hg or a heart rate of ≥100 beats/min.The diagnosis was made mainly based on findings from the initial and second endoscopies after excluding other diseases by combining colonoscopy findings with those of other imaging tests, such as CT, small bowel endoscopy (capsule or balloon endoscopy), or upper gastrointestinal endoscopy.

Outcomes
In this study, the primary outcome was mortality.The secondary outcomes were rebleeding, length of stay (LOS), blood transfusion, thromboembolism, endoscopic treatment, need for interventional radiology, and surgery requirement.In addition, we evaluated the performance of CT, urgent CT (performed within 1 h after admission) [27], enhanced CT, colonoscopy, and early colonoscopy (performed within 24 h after admission).Mortality was defined as in-hospital mortality from any causes.All secondary outcomes were included during the admission.Rebleeding was defined as significant amounts of fresh, bloody, or winecolored stools [30,31].Transfusion was defined as a red blood cell transfusion.Thromboembolism included acute coronary syndrome, cerebrovascular disease, and pulmonary embolism/ deep vein thrombosis.Endoscopic treatment, interventional radiology, and surgery were included from any point during admission.

Statistical Analyses
The categorical variables were compared using a χ 2 test or Fisher's exact test.LOS, which has a non-normal distribution, was compared using the Wilcoxon rank-sum test.At least one missing data value was found in 35% (2,842/8,120) of all cases.However, there were no missing data values in diagnoses and outcomes.
Multiple imputation (MI) was used to account for missing values.MI is a procedure that replaces missing values with other reliable values by creating MI patterns to avoid the bias caused by missing values, and it often gives more reliable results than a completed case analysis [34].MI was performed using the chained equation method under the missing-at-random assumption [35].All outcomes and clinical data, excluding the diagnosis, were included in the imputation models, and 20 imputed datasets were created.Propensity scores were calculated using logistic regression analysis, and scores were estimated using the following variables: hospital type, hospital size, age, gender, body mass index, drinking, smoking, CCI, NSAIDs, antiplatelets, anticoagulants, PS, syncope, abdominal pain, fever, diarrhea, melena, blood pressure, heart rate, shock, albumin, white blood cells, hemoglobin, blood urea nitrogen, creatinine, platelet, INR of prothrombin time, and C-reactive protein.Propensity score matching (PSM) was then performed within the 20 imputed datasets, using nearest neighbor matching and caliper widths of 0.2 of the pooled standard deviations.The results for each dataset were combined using Rubin's rules [36].The odds ratio (OR) of the outcomes was estimated from the logistic regression models.The difference in LOS was tested using a generalized linear model.
The sensitivity analyses were performed on the primary outcome.The first was a complete case analysis because the MI method was used.In the complete case analysis, the propensity score was measured with the same variables used for PSM after MI.Then PSM was performed using nearest neighbor matching and caliper widths of 0.2 of the pooled standard deviations.The second was an analysis in which the etiologies were added to the variables when calculating the propensity score in the main analysis.
Finally, we changed the definitions of weekends and weekdays as sensitivity analyses, with the first defining weekend and weekday as Friday through Saturday and Sunday through Thursday, respectively, and the second defining weekend and weekday as Saturday through Sunday and Monday through Thursday, respectively.It was noticed that the weekend started with the close of business on Friday rather than the midnight, as this timing affected the number of staff and examinations.Since a late Friday admission was presumed to be less likely to receive a colonoscopy than a late Sunday admission because of pretreatment, we compared the Friday through Saturday admissions and the Sunday through Thursday admissions.This sensitivity analysis was recommended previously [25].However, our data categorized only admission days on weekdays; therefore, the time of admission is unknown.Since a weekday may be affected by midnight versus the close of the business on Friday, Saturday through Sunday admissions were also compared with Monday through Thursday admissions, excluding Friday admissions.Since this was not our first study of the CODE BLUE-J Study, the sample size was not calculated in this study, and the sample size collected in the first study was used.
Statistical significance was set at p < 0.05.All analyses were performed using R software version 4.1.3(The R Foundation for Statistical Computing, Vienna, Austria).MI was carried out using the mice package (v3.13.0).After MI, PSM was performed using the MatchThem package (v1.0.0).PSM for complete cases was performed using the MatchIt package (v4.9.9).

Results
A total of 1,977 patients were admitted on weekends, and 6,143 patients were admitted on weekdays.The characteristics of weekend admission and weekday admission are shown in Table 1.The amount of missing data regarding maximum drinking history was 15.2% (1,233/8,120).Although weekend admissions had a significantly higher number of patients over 70 years of age and patients with abdominal pain, shock, and elevated white blood cell counts (p = 0.039, 0.003, 0.018, and 0.009, respectively), absolute standardized differences were less than 0.1 in all variables between the two groups.The final  2. Approximately 60% of patients with ALGIB were diverticular bleeding in both weekend and weekday admissions.In weekend admissions, ischemic colitis was significantly higher (p < 0.001), and inflammatory bowel disease and radiation colitis were significantly lower (p < 0.001 and 0.015, respectively).

Before PSM
The outcomes before PSM are shown in Table 3. Weekend and weekday mortality rates were 1.3% and 0.9%, respectively.Mortality was not significantly different for weekend admissions compared with weekday admissions (OR 1.475, 95% confidence interval [CI] 0.923-2.358;p = 0.104).Blood transfusion rates between weekend and weekday admissions were 31.3% and 28.4%, respectively.The blood transfusion rate was significantly higher for weekend admissions compared with weekday admissions (OR, 1.148 [95% CI, 1.028-1.281];p = 0.014).

Sensitivity Analysis
In the complete case analysis, there were 1,306 weekend admissions and 3,972 weekday admissions.The characteristics and etiologies of the complete cases are shown in online supplementary Tables 2, 3, respectively.The outcomes of the sensitivity analysis are demonstrated in Table 4. Mortality rates before and after PSM were not significantly different for weekend admissions compared to weekday admissions.
In the next sensitivity analysis, bleeding etiologies were included in the calculation of the propensity score.The mortality was not significantly higher for weekend admissions (OR, 1.461 [95% CI, 0.796-2.680];p = 0.301).Furthermore, the sensitivity analyses with different definitions for weekends and weekdays showed similar results.Therefore, all sensitivity analyses did not show a significant effect of weekend admission on mortality.

Discussion
The weekend effect on clinical outcomes of ALGIB was evaluated using data from a large multicenter study in Japan.There was no weekend effect on mortality and other outcomes except blood transfusion.Furthermore, in weekend admissions, emergency CT was frequently performed, and early colonoscopy was avoided but did not affect many outcomes.
In general, critically ill patients are more likely to be admitted during the weekend.Mildly ill patients refrain from visiting the hospital until Monday.In this study as well, more severely ill patients were admitted on weekends than on weekdays.In addition, many older people were admitted on weekends.However, the results did not change before and after PSM, and weekend admission did not affect mortality.Li et al. [25] reported that the OR for mortality in ALGIB was 0.89 (95% CI, 0.65-1.23) in weekend admission compared to weekday admission.An exploratory study of ALGIB mortality factors in a large (n = 1,198) prospective observational study in Italy found that the OR of mortality for weekend admission compared to weekday admission was 1.59 (95% CI, 0.73-3.45)[37], and weekend Values are the number (%).Others included anal condyloma, anal laceration or fissure, colorectal laceration, colorectal perforation, Cronkhite-Canada syndrome, diverticulitis, drug-induced ulcer, fistula or penetration into colorectum, graft-versus-host disease, hematoma, Henoch-Schönlein purpura, intussusception, Kaposi's sarcoma, mucosal bleeding, mucosal lymphoid hyperplasia, mucosal prolapse syndrome, pseudoaneurysm and stoma-related bleeding.

Weekend Effect on Mortality of ALGIB
Dig Dis 2023;41:890-899 DOI: 10.1159/000533744 admission was not a significant risk factor for mortality in ALGIB.These reports support the results of our study, and the effect of weekend admission on ALGIB mortality was considered negligible.
The primary outcome was mortality in the present study.Certainly, because mortality was considered to be situated downstream of rebleeding [38] and the mortality rate was much lower than rebleeding rate, rebleeding may be suitable as the primary outcome.However, it is more difficult to identify rebleeding than mortality in retrospective design, and ascertainment bias can occur in the assessment of rebleeding events.Therefore, we set mortality as the primary outcome in the present study.
Early colonoscopy was avoided during weekend admissions, probably because of the lack of staff and the need for bowel preparation on weekends.However, it did not affect mortality.This may be because, unlike UGIB, lower gastrointestinal bleeding has a high rate of spontaneous hemostasis of 66-92.5% [39][40][41] and can contribute to a low mortality rate.Therefore, it is speculated that ALGIB often does not necessarily require hemostasis, and it may be more important to perform a colonoscopy during hospitalization than to perform an early colonoscopy.In fact, it has been reported that colonoscopy in patients with ALGIB suppresses mortality [25].In this study, weekend admissions had a high colonoscopy rate of about 90%, similar to weekday admissions.This higher colonoscopy rate may have resulted in higher diagnosis rates and lower mortality rates in both weekend and weekday admissions, with no difference in mortality rates.Therefore, an early colonoscopy may be less effective for ALGIB.In fact, three meta-analyses including randomized controlled trials [38,42,43] found that patients did not show any improvement in ALGIB.However, Shiratori et al. [44] reported that early colonoscopy was particularly beneficial for patients with a shock index ≥1 or performance status ≥3 at presentation.Therefore, there may be certain populations that may benefit from early colonoscopy.However, the results of this study also suggest that early colonoscopy is not required even in the special situation of weekend admission.
The early colonoscopy rate in weekend admissions was low, whereas the urgent CT rate in weekend admissions was high.In ALGIB, CT contributes to the diagnosis and search for a bleeding site, but enhanced CT has side effects such as renal damage, and CT is not recommended as the initial diagnosis method in Japan [18].In Japan, colonoscopy is recommended for the diagnosis and search for bleeding sites, while CT is recommended as an alternative when colonoscopy is difficult to perform.On the other hand, UK guidelines recommend CT angiography as the first line for shocks and other unstable circulatory objects [15].In the USA, it is recommended to perform CT before surgery or angiography [19].On weekends, CT may be preferred because, unlike colonoscopy, it requires no pretreatment and is easier to perform.Since there are many urgent CTs on the weekend, it is speculated that the indication of early colonoscopy is often judged by these urgent CT results on the weekend.This is likely because the stigmata of recent hemorrhage identification rate by colonoscopy is high when the source of bleeding is found from urgent CT [43].The detection rate was higher with colonoscopy following CT for vascular lesions (35.7 vs. 20.6%,p = 0.01), leading to more endoscopic therapies (34.9 vs. 13.4%,p < 0.01) [45].Furthermore, early colonoscopy may not be necessary in some cases because the diagnosis can be confirmed by urgent CT alone.Therefore, it was speculated that the necessity of early colonoscopy was judged from the CT findings, and different strategies were used for weekends and weekdays but did not affect many outcomes.These results suggest that urgent CT on weekends can manage ALGIB without changing the clinically important outcome and that early colonoscopy is not always necessary on weekends.
The weekend effect was observed only in blood transfusions.Blood transfusions in ALGIB were performed according to evidence of the effectiveness of restricted blood transfusions in UGIB.As in the USA and Europe, restricted blood transfusion with hemoglobin 7 g/ dL is also recommended in Japan [15,18,19].At present, based on these findings, blood transfusion is performed at the discretion of the attending physician in the clinical setting.However, since a number of blood transfusions were consistently performed during the weekend after PSM, the weekend admission may affect the attending physician's judgment.As previously reported [25], blood transfusions were frequently performed on weekdays.However, the report did not consider the reason for more blood transfusions on weekday admissions, even though there were few patients with shock on weekday admissions.Furthermore, multivariate analysis was not performed to clarify the factors for blood transfusion requirements.Therefore, it seemed unclear whether there were fewer blood transfusions during weekend admissions than during weekday admissions.The reason for the frequent transfusions on weekdays may be better access to transfusions on weekdays than on weekends.As with UGIB, it was reported that blood transfusions were slightly more likely to occur during the weekend [20].Therefore, we think that even patients with ALGIB are likely to have blood transfusions on weekends.
There were several limitations in this study.First, since this study was a retrospective cohort study, selection bias was unavoidable.Second, although the confounding factors were well balanced by PSM, the unmeasured confounders could not be balanced between the two groups.Third, participating facilities were able to handle endoscopy and CT 24 h per day, so the results could not be applied to facilities that could not handle them 24 h per day.Fourth, the weekend admissions could not include Friday night because admission time was unknown.Therefore, two sensitivity analyses were performed, but the results did not change.Fifth, some weekdays included holidays, which could have affected the results.Sixth, this study was conducted only in Japan and may be difficult to generalize to other countries.Finally, our study did not have a sufficient sample size.In PSM, 1,976 subjects were matched for each group.The mortality rate was 0.009 in the weekday group as the reference, and the OR was 1.437 (Table 3).The power was calculated as 0.174 under a type I error rate of 0.05, and 11,779 subjects were required in each group to reject the null hypothesis.Further studies Weekend Effect on Mortality of ALGIB Dig Dis 2023;41:890-899 DOI: 10.1159/000533744 will require a sufficient sample size.Nevertheless, this study has several strengths.Unlike administrative data, the data used in this study have high validity for aspects such as diagnosis and procedure.The study is unique in that no other study has evaluated the weekend effect on ALGIB with such a large amount of clinical data and the weekend effect on CT management.
In conclusion, weekend admission did not affect the mortality and many other outcomes of ALGIB.Furthermore, on weekends, early colonoscopy may not always be necessary if an urgent CT is performed and ALGIB is properly managed.

Table 1 .
Characteristics of patients with ALGIB urgently admitted on weekends and weekdays

Table 2 .
Etiologies of ALGIB in weekend and weekday admissions

Table 3 .
Clinical and management outcomes between weekend and weekday admissionsValues are the number (%).LOS is expressed as medians and IQR.CI, confidence interval; CT, computed tomography; IQR, interquartile range; IVR, interventional radiology; LOS, length of stay; OR, odds ratio; MI, multiple imputation; PSM, propensity score matching.† Linear regression was performed for the comparison of LOS between the two groups and demonstrated as a coefficient (95% confidence interval).p value <0.05 was considered to indicate statistical significance.

Table 4 .
Sensitivity analyses for mortality of ALGIB