Introduction
Upper gastrointestinal (UGI) bleeding is a frequent cause of emergency admission and
hospitalization. In the United States, its incidence is 1/1000 individuals per year
and it accounts for more than 100 000 hospital admissions per year at a cost of more
than 2 billion USD [1]
[2]. In the UK, the incidence of non-variceal UGI bleeding was reported to be 50 – 150/100
000 [3]. In the United States, the costs for variceal and non-variceal UGI bleeding were
estimated to vary between 3400 and 23 300 USD [4]
[5].
The severity of UGI bleeding is highly variable; severe episodes account for a small
proportion of all bleeding [3]. Several validated scores, based on clinical and endoscopic findings, attempt to
predict the rebleeding risk [6]
[7]. UGI bleeding with a low rebleeding risk can be safely managed on an outpatient
basis, allowing a reduced duration of hospital stay and decreased associated costs
[6]
[7]. The Glasgow-Blatchford Bleeding score (GBS, [Table 1]) predicts the need for surgery, blood transfusion and any endoscopic hemostasis
[8]
[9]
[10]
[11]. Its main advantage over other scores available for selecting patients at low risk
of rebleeding is that it can be computed in a straightforward manner, without UGI
endoscopy. A GBS of 0 has been shown to accurately predict a very low risk of requiring
a clinical intervention, defined as blood transfusion, endoscopic hemostasis or surgical
management for bleeding control [8].
Table 1
Glasgow-Blatchford score.
|
GBS
|
|
Blood urea (mmol/l)
|
|
|
< 6.5
|
0
|
|
6.5 – 7.9
|
2
|
|
8 – 9.9
|
3
|
|
10 – 25
|
4
|
|
> 25
|
6
|
|
Hemoglobin: men (g/l)
|
|
|
> 129
|
0
|
|
120 – 129
|
1
|
|
100 – 119
|
3
|
|
< 100
|
6
|
|
Hemoglobin: women (g/l)
|
|
|
> 119
|
0
|
|
100 – 119
|
1
|
|
< 100
|
6
|
|
Systolic blood pressure (mmHg)
|
|
|
> 109
|
0
|
|
100 – 109
|
1
|
|
90 – 99
|
2
|
|
< 90
|
3
|
|
Other markers
|
|
|
BPM ≥ 100
|
1
|
|
Melena
|
1
|
|
Syncope
|
2
|
|
Liver disease[*]
|
2
|
|
Heart failure[*]
|
2
|
Abbreviations: BPM, beats per minute; GBS, Glasgow-Blatchford Bleeding Score.
Total score is obtained by adding the value of all items.
Adapted from Stanley AJ et al [8].
* Based on past medical history, clinical or laboratory evidence.
We tested the hypothesis that, by implementing the GBS, selecting low-risk GI bleeders
and safely managing them as outpatients, costs and duration of hospital stay can be
reduced.
Patients and methods
Consecutive outpatients admitted to the Emergency Department of the University Hospital
of Geneva (Geneva, Switzerland) with UGI bleeding were prospectively screened for
inclusion in the study. Patients over 18 years of age with UGI bleeding defined as
hematemesis or coffee ground emesis or with melena were eligible for inclusion. Exclusion
criteria were pregnancy and hematochezia (to avoid all lower gastrointestinal bleeding,
acknowledging that severe UGI bleeding may present with hematochezia). Vital signs,
clinical history, physical examination and hemoglobin levels were obtained in the
Hospital Emergency Department, at admission, to calculate the GBS. The study was divided
into two consecutive phases:
-
During the “observational” study phase (October 2009 – August 2010), apart from a
systematic calculation of the GBS, routine local clinical practice was unchanged:
all patients received proton pump inhibitor (PPI) therapy and underwent an UGI endoscopy
during the 12 h following hospital admission, under midazolam sedation or with orotracheal
intubation for unstable patients. This was performed in the endoscopy unit or in the
Emergency Department in the case of patient instability or when endoscopies were done
during on-call hours. The physician responsible for the patient in the Emergency Department
decided whether to discharge or to admit the patient without interference from the
investigators.
-
During the “interventional” study phase (January 2011 – January 2012), patients with
a GBS of 0 were not admitted to the hospital and they received an appointment for
an ambulatory UGI endoscopy during the following 48 h. If the GBS was > 0, patient
management was identical to the observational study phase.
Vital parameters and hemoglobin levels were recorded and supportive measures such
as drug administration, blood transfusion and resuscitation, were administered by
the physician in charge (i. e., physician of the Emergency Department relayed by the
physician in charge on the ward) and prospectively monitored. Endoscopic hemostasis
during emergent UGI endoscopy was performed if deemed indicated by the endoscopist.
All emergent endoscopies were reviewed daily by a panel of ≥ 2 senior endoscopists.
In the case of rebleeding, the decision to perform a surgical operation or not was
taken by the patient and the medical team (physician in charge, endoscopist, surgeon
and radiologist).
Follow-up was performed at 7 and 30 days by one of the investigators (MG) using hospital
charts supplemented by face-to-face visits or phone calls. Data prospectively collected
included vital signs, the presence of major comorbidities, results of blood analysis,
endoscopic findings, PPI administration, number of red blood cell units transfused,
rebleeding events, readmission, surgery and death. Hospital costs were calculated
for patients with a GBS of 0 during each study phase. According to Swiss public healthcare
tariffs, cost calculations included all real costs for the first 24 h, calculated
using the 2013 TARMED reimbursement rates plus, in the case of a hospital stay longer
than 24 h, a daily package of 686 EUR.
The study protocol was approved by the hospital ethics committee (Geneva University
Hospital IRB) and it was registered in a publically accessible registry (clinicaltrials.gov
#NCT01029626); all patients gave written informed consent before inclusion in the
study.
Statistical analysis
Categorical and continuous variables were described as percentages and by their median
and range, respectively. Comparisons between groups were performed using Fisher’s
exact test for categorical data and the Mann – Whitney test for continuous variables.
All P tests were two-sided, and P values < 0.05 were considered to be statistically significant. All analyses were
performed on an intention-to-treat basis including the 208 patients. Sample-size calculation
was based on the assumption that duration of hospital stay for patients admitted with
UGI bleeding would be shortened by 25 % in the interventional study phase compared
with the observational study phase. On the basis of 0.8 power, 104 patients were required
in each study phase to detect a significant difference (P = 0.05). Analyses were performed using Prism version 6.00 for Windows (GraphPad Software,
San Diego, CA, USA).
Results
Patients
Two-hundred and eight consecutive consenting patients with UGI bleeding were included
in this study ( [Fig. 1]). Nineteen patients were excluded in each study phase because they were unable (n = 12)
or refused (n = 15) to give consent, or because blood urea level was not measured
(n = 11). Follow-up at 7 and 30 days was obtained for 208 (100 %) and 198 (95 %) patients,
respectively. At 30 days, 7 (3.4 %) patients had died and three (1.4 %) patients were
lost to follow-up.
Fig. 1 Enrollment and outcome. Patients were enrolled in two consecutive phases: during
the first, observational, phase, all patients had upper gastrointestinal endoscopy;
during the second, interventional, phase, patients with a Glasgow-Blatchford Bleeding
Score (GBS) of 0 were discharged with an appointment for upper gastrointestinal endoscopy
during the following 48 h.
Patients’ characteristics at baseline were similar for the two study phases ([Table 2]). As expected, compared with patients who presented with a GBS > 0, those with a
GBS of 0 were younger, had lower blood urea levels, higher hemoglobin levels and a
lower heart rate during each study phase ( [Table 2]). The distribution of GBS was similar for the two study phases, with 15 (14 %) and
11 (11 %) patients having a GBS of 0 during the observational and interventional study
phase, respectively (P = 0.530) ( [Fig. 2]).
Table 2
Patients’ characteristics at baseline[1].
|
Observational study phase
(n = 104)
|
Interventional study phase
(n = 104)
|
|
GBS = 0
(n = 15)
|
GBS > 0
(n = 89)
|
P-value
|
GBS = 0[2]
(n = 11)
|
GBS > 0[2]
(n = 93)
|
P-value
|
|
Age (years)
|
35 (21 – 91)
|
69 (20 – 99)
|
0.0001
|
33 (20 – 43)
|
71 (25 – 96)
|
0.001
|
|
Men, n (%)
|
11 (73)
|
68 (76)
|
NS
|
10 (91)
|
58(62)
|
NS
|
|
Urea (mmol/L)
|
5 (1.8 – 6.1)
|
10 (2 – 49.6)
|
0.0001
|
4 (2.6 – 5.8)
|
11 (2.4 – 38.2)
|
0.001
|
|
Hb (g/L)
|
150 (127 – 160)
|
98 (38 – 185)
|
0.0001
|
150 (136 – 174)
|
86 (46 – 156)
|
0.001
|
|
SBP (mmHg)
|
135 (110 – 160)
|
120 (90 – 170)
|
NS
|
130 (110 – 145)
|
120 (70 – 170)
|
NS
|
|
BPM
|
75 (61 – 93)
|
96 (61 – 150)
|
0.048
|
69 (53 – 99)
|
89 (59 – 150)
|
0.001
|
|
BPM > 100, n (%)
|
0
|
35 (39)
|
0.001
|
0
|
31 (33)
|
0.030
|
|
Melena, n (%)
|
0
|
65 (73)
|
0.0001
|
0
|
71 (76)
|
0.0001
|
|
Syncope, n (%)
|
0
|
6 (7)
|
NS
|
0
|
11 (12)
|
NS
|
|
Liver disease, n (%)
|
0
|
33 (37)
|
0.049
|
0
|
24 (26)
|
NS
|
|
Heart failure, n (%)
|
0
|
4 (5)
|
NS
|
0
|
3 (3)
|
NS
|
|
GBS
|
0
|
9 (1 – 17)
|
0.0001
|
0
|
11 (1 – 17)
|
0.001
|
Abbreviations: BPM, beats per minute; GBS, Glasgow-Blatchford Bleeding Score; NS,
non-significant; SBP, systolic blood pressure.
1 Values are medians with ranges in parentheses except otherwise stated.
2
P > 0.05 for all comparisons with the corresponding subgroup of patients in the observational
study phase.
Fig. 2 Distribution of the number of patients per Glasgow-Blatchford Bleeding Score during
the observational and interventional study phases.
Endoscopic findings
[Table 3] summarizes the endoscopic findings in each study phase. Endoscopic diagnoses in
patients with a GBS of 0 were non-severe lesions, mostly Los Angeles grade A esophagitis,
Forrest III duodenal ulcer or Mallory-Weiss tears. None of these patients had variceal
bleeding. During the observational study phase, lesions listed as “other” in patients
with a GBS of 0 consisted of gastric erosions (n = 1) and during the interventional
study phase, they consisted of Mallory-Weiss tears (n = 5) and gastric erosions (n = 1).
Table 3
Endoscopic diagnosis[1].
|
Observational study phase
(n = 104)
|
Interventional study phase
(n = 104)
|
|
Diagnosis
|
GBS = 0
(n = 15)
|
GBS > 0
(n = 89)
|
P-value
|
GBS = 0
(n = 11)
|
GBS > 0
(n = 93)
|
P-value
|
|
No lesions
|
8 (53)
|
10 (11)
|
0.0005
|
3 (27)
|
10 (11)
|
NS
|
|
Blood in upper digestive tract
|
1 (7)
|
25 (28)
|
NS
|
0
|
26 (28)
|
NS
|
|
EV or GV
|
0
|
21 (24)
|
0.037
|
0
|
16 (17)
|
NS
|
|
Reflux esophagitis
|
6 (40)
|
22 (25)
|
NS
|
1 (9)
|
28 (30)
|
NS
|
|
PHT gastropathy
|
0
|
19 (21)
|
NS
|
0
|
6 (6)
|
NS
|
|
Gastric ulcer
|
0
|
22 (25)
|
0.036
|
0
|
33 (35)
|
0.015
|
|
Duodenal ulcer
|
0
|
22 (25)
|
0.036
|
1 (9)
|
24 (26)
|
NS
|
|
Other lesion[2]
|
1 (7)
|
10 (11)
|
NS
|
6 (54)
|
9 (10)
|
0.001
|
Abbreviations: EV, esophageal varices; GBS, Glasgow-Blatchford Bleeding Score; GV,
gastric varices; NS, non-significant; PHT, portal hypertension.
1 Values are numbers of patients with percentages in parentheses; values may add to
more than 100 % as some patients had several diagnoses.
2 Mallory-Weiss tears, gastrointestinal stromal tumor, gastric arterio-venous ectasia,
angiodysplasia, Dieulafoy’s lesion.
Outcomes, hospital stay and costs
In both study phases, none of the patients with a GBS of 0 received any clinical intervention
(apart from one patient who received a hemoclip [QuickClip, Olympus, Tokyo, Japan]
for a small and superficial gastric erosion during the interventional study phase;
this intervention was considered to be not needed by two senior endoscopists after
reviewing the case on the following day). In contrast, 27 % and 39 % of patients with
a GBS > 0 received endoscopic hemostasis during the observational and interventional
study phases, respectively. As expected, non-endoscopic outcomes (red blood cell transfusions
and administration of PPI) were markedly different for patients with a GBS of 0 versus
> 0, in each study phase ([Table 4]). No significant difference in clinical outcomes was observed between study phases
for patients with a GBS of 0 and for those with a GBS > 0 ([Table 4]), apart from the duration of hospital stay of patients with a GBS of 0: this decreased
by 68 % during the interventional study phase, from 19 h (5 – 148 h) to 6 h (1 – 13 h)
(P = 0.010). Concomitantly, in patients with a GBS of 0, median hospitalization costs
decreased by 34 %, from 1272 EUR (553 – 4296 EUR) to 845 EUR (336 – 1441 EUR ) (P = 0.002). None of the patients in the group with a GBS of 0 were readmitted, presented
rebleeding or required a repeat endoscopy. All patients with a GBS of 0 were alive
at the 30-day follow-up ([Fig. 3]).
Table 4
Clinical outcome[1].
|
Observational study phase
(n = 104)
|
Interventional study phase
(n = 104)
|
|
GBS = 0
(n = 15)
|
GBS > 0
(n = 89)
|
P-value
|
GBS = 0
(n = 11)
|
GBS > 0
(n = 93)
|
P-value
|
|
Clinical intervention (transfusion, surgery or hemostasis), n (%)
|
0
|
57 (64)
|
0.0001
|
1 (9)
|
63 (68)
|
0.0001
|
|
Hemostasis, n (%)
|
0
|
24 (27)
|
0.019
|
1 (9)
|
36 (39)
|
NS
|
|
Surgery, n (%)
|
0
|
2 (2)
|
NS
|
0
|
1 (1)
|
NS
|
|
Death at 30 days, n (%)
|
0
|
5 (6)
|
NS
|
0
|
2 (2)
|
NS
|
|
Blood transfusion, n (%)
|
0
|
54 (61)
|
0.0001
|
0
|
59 (63)
|
0.0001
|
|
PPI duration (days)
|
1 (0 – 2)
|
3 (0 – 6)
|
0.0004
|
0
|
3 (0 – 3)
|
0.001
|
|
Hospital stay duration (h)
|
19 (5 – 148)
|
189 (5 – 816)
|
0.0001
|
6 (1 – 13)[2]
|
207 (7 – 1035)
|
0.0001
|
Abbreviations: GBS, Glasgow-Blatchford Bleeding Score; NS, non-significant; PPI, proton
pump inhibitors.
1 Values are medians with ranges in parentheses except otherwise stated.
2
P = 0.010 for comparison with the observational study phase; all other comparisons
with the observational study phase showed no statistical significance.
Fig. 3 Number of patients needing an intervention (i. e., endoscopic hemostasis, blood transfusion
or surgery) for each category of Glasgow-Blatchford Bleeding Score during the observational
(A) and interventional (B) study phases.
Discussion
This study was aimed at defining whether applying the GBS score to patients presenting
at an Emergency Department with UGI bleeding may alter the hospital stay and costs
by selecting patients who will or will not need an intervention (surgery, blood transfusion
or endoscopic hemostasis). We found that duration of hospital stay and costs were
drastically reduced in low-risk UGI bleeders by implementing a policy of outpatient
management.
All of our patients with a GBS of 0 were alive at 30 days without having presented
rebleeding or requiring repeat endoscopy. This is in line with the large multicentric
study by Stanley et al. which showed that, among a total of 228 patients who presented
with a GBS of 0, none had presented rebleeding or had needed repeat endoscopy during
follow-up [8]. These authors also reported that the introduction of the GBS into their clinical
practice was associated with a reduced proportion of low-risk patients admitted to
the hospital, down to 32 % (a formal protocol to discharge patients with a GBS of
0 was not used in that study).
We selected a risk score that can be computed before endoscopy to maximize benefits
in terms of hospital costs and capacity overload in the Emergency Department. Among
three risk scores available for assessing the severity of UGI bleeding before performing
endoscopy, we selected the GBS because several studies have demonstrated its superiority
over the Rockall score in predicting the need for clinical intervention, and the GBS
is simpler to calculate than the “artificial neural network” [12]. The GBS is promptly and easily calculated, allowing an immediate decision to perform
UGI endoscopy or to discharge the patient as soon as hemoglobin and urea blood levels
are available. We experienced difficulties in implementing a systematic measurement
of urea blood level in the Emergency Department of our hospital, which partly accounts
for the relatively long duration of hospital stay observed in some of our patients
who had a GBS of 0. This may also be the case in other hospitals as Stanley et al.
did not obtain urea blood levels for 4 % of their patients (similar to the 4.4 % observed
in our screened patients) [8]. Nevertheless, based on the results of the present study, the GBS has now been formally
implemented in our hospital algorithm for the management of outpatients who present
with UGI bleeding.
To the best of our knowledge, all other studies that used pre-endoscopy risk scores
to select low-risk patients for a non-admission policy were observational:
-
In a retrospective audit of two hospitals, Robins et al. showed that discharging selected
UGI bleeders without endoscopy was safe: severity of UGI bleeding was calculated based
on the GBS supplemented with clinical parameters; 9 % of patients had UGI endoscopy
in one hospital and their outcome was similar to that of patients who had presented
to another hospital where endoscopy was performed more liberally (in 74 % of patients)
[13]
-
In a mostly retrospective analysis, the policy of not admitting low-risk patients
and scheduling UGI endoscopy on the next working day was evaluated in 139 patients
over a 5-year period. It was found to be safe as none of the patients received endoscopic
therapy, blood transfusion or surgery for UGI bleeding. A GBS ≤ 2 supplemented with
clinical criteria was used to select low-risk bleeders [14].
One limitation of our study is the relatively small number of patients with a GBS
of 0 who were included, particularly in the second study phase (n = 11). However,
this proportion (12.5 %) lies in the range of previous studies (5 – 22 %) [12]. Although the population of patients with a GBS of 0 was small, the reduction in
duration of hospital stay and decreased costs were highly significant, which were
in line with our basic hypothesis. Another limitation is that our study is a single
center and single country study. Despite that, studies performed in other countries,
using a GBS of 0 as cut-off predicted a good and safe clinical evolution. Similar
results should be expected in reproducing this study elsewhere. In an observational
study, Le Jeune et al. have reported that patients with a GBS ≤ 2 could also be considered
for early discharge, potentially doubling the number of patients eligible for a “non-admission”
management [15]. In our study, 3 (5 %) of our 59 patients with a GBS ≤ 4 needed an intervention
and one patient with a GBS of 2 died at day 2 post-gastrectomy for a gastric cancer
discovered at UGI endoscopy performed upon admission to our hospital. On the other
hand, the strengths of our study include its prospective, interventional, design with
the inclusion of a high number of patients and a 30-day follow-up rate of 95 %.
We included prospectively all patients with UGI bleeding regardless of whether the
bleeding was variceal or not. It was important not to exclude cirrhotic patients because
these patients often present with non-variceal bleeding. As in other trials [8], the GBS showed its ability to discriminate severe from non-severe variceal bleeding
as well as non-variceal bleeding.
In conclusion, we showed that implementing a triage policy based on the GBS for selecting
out-of-hospital UGI bleeders who are eligible for outpatient management effectively
reduces duration of hospital stay and costs.
