Effectiveness of disseminating consensus management recommendations for ulcer bleeding: a cluster randomized trial

Alan N. Barkun; Mamatha Bhat; David Armstrong; Martin Dawes; Allan Donner; Robert Enns; Janet Martin; Paul Moayyedi; Joseph Romagnuolo; Larry Stitt

doi:10.1503/cmaj.120095

Article Figures & Tables

Figures

Tables

Download figure
Open in new tab
Download powerpoint
Figure 1:
Selection of hospitals and patients for inclusion in the trial.
Download figure
Open in new tab
Download powerpoint
Figure 2:
Timetable of randomization, intervention, follow-up and data collection. Chart numbers shown do not include an additional 10% duplicate independent data entry for validation purposes. Data collection timelines were tabulated based on conservative estimates (based on pilot testing) of case volumes of 4–5 eligible patients weekly in the period of care and 4–5 charts abstracted weekly in the data collection period. The timelines allowed additional time for data collection (3–5 weeks if a 2-week turn-around was allowed for the preparation of feedback sheets).

Tables

Figures

View popup

Table 1:

Patient-level analysis of compliance to guidelines for the management of nonvariceal upper gastrointestinal bleeding

Guideline	Before intervention, % (no. of patients)		After intervention, % (no. of patients)		Intracluster correlation coefficient	Percentage difference (95% CI)

	Experimental group	Control group	Experimental group	Control group
Primary outcome

Endoscopic hemostasis followed by high-dose IV PPI	13.2 (20/152)	7.1 (11/155)	9.8 (14/143)	4.8 (6/124)	0.04	−5.0 (−12.0 to 2.0)

Secondary outcomes

Injection and/or thermal coagulation without clips and bolus IV PPI followed by infusion for 72 h for high-risk ulcers	8.6 (13/152)	5.2 (8/155)	6.3 (9/143)	4.8 (6/124)	0.07	−1.5 (−8.1 to 5.2)

Injection and/or thermal coagulation with clips and bolus IV PPI followed by 8 mg/h infusion for 72 h and no oral PPI. If the patient received a pre-endoscopy infusion, the absence of bolus IV PPI was ignored	69.1 (105/152)	58.7 (91/155)	57.3 (82/143)	62.9 (78/124)	0.18	5.6 (−12.7 to 23.8)

Injection and/or thermal coagulation with clips and pre-endoscopy IV bolus PPI plus 80-mg bolus IV PPI followed by 8 mg/h infusion for 72 h after endoscopy and no oral PPI	1.3 (2/152)	1.9 (3/155)	1.4 (2/143)	0.8 (1/124)	−0.05	−0.6 (−2.5 to 1.3)

Injection and/or thermal coagulation upon endoscopy for high-risk ulcers with clips alone	81.6 (124/152)	74.8 (116/155)	72.7 (104/143)	72.6 (90/124)	0.35	−0.1 (−20.4 to 20.1)

Injection and/or thermal coagulation without clips for high-risk ulcers alone	49.3 (75/152)	56.1 (87/155)	50.4 (72/143)	46.8 (58/124)	0.33	−3.6 (−25.4 to 18.2)

80-mg IV bolus PPI followed by 8 mg/h PPI infusion for 72 h after endoscopy and no oral PPI	15.3 (27/177)	10.0 (17/170)	13.6 (21/155)	15.9 (24/151)	0.27	2.3 (−12.1 to 16.8)

80-mg IV bolus PPI followed by 8 mg/h infusion for 72 h after endoscopy and no oral PPI. If patient received a pre-endoscopy infusion, the absence of bolus IV PPI afterward was ignored	82.5 (146/177)	80.0 (136/170)	81.3 (126/155)	89.4 (135/151)	0.06	8.1 (−1.4 to 17.6)

Pre-endoscopy bolus of PPI plus 80-mg bolus IV PPI followed by 8 mg/h infusion for 72 h after endoscopy and no oral PPI	1.1 (2/177)	1.8 (3/170)	1.9 (3/155)	9.3 (14/151)	0.35	7.3 (−2.9 to 17.5)

Use of validated scoring system based on a combination of clinical and endoscopic characteristics to stratify patients into low- and high-risk categories	0.0 (0/402)	0.2 (1/424)	1.9 (7/361)	0.0 (0/389)	0.19	−1.9 (−5.0 to 1.1)

Early endoscopy (in the first 24 h) with risk classification by clinical/endoscopic criteria	62.9 (253/402)	65.6 (278/424)	57.3 (207/361)	63.1 (245/388)	0.09	5.6 (−5.6 to 16.9)

No hemostasis for low risk lesion	88.9 (200/225)	93.7 (223/238)	94.2 (196/208)	89.0 (218/245)	0.13	−5.2 (−13.2 to 2.8)

Endoscopic hemostasis for adherent clot	87.9 (29/33)	88.0 (22/25)	80.7 (25/31)	80.8 (21/26)	0.09	0.1 (−22.9 to 23.1)

Endoscopic hemostasis for other active bleeding or visible vessel	85.9 (152/177)	83.3 (155/186)	93.5 (143/153)	86.1 (124/144)	0.06	−7.4 (−15.8 to 1.1)

Low-risk patients who received treatment	11.1 (25/225)	6.3 (15/238)	5.8 (12/208)	11.0 (27/245)	0.13	5.2 (−2.8 to 13.2)

Low-risk patients not given IV PPI bolus or infusion after endoscopy	61.8 (139/225)	59.2 (141/238)	56.3 (117/208)	64.1 (157/245)	0.05	8.0 (−2.9 to 18.9)

Note: CI = confidence interval, IV = intravenous, PPI = proton pump inhibitor.

View popup

Table 2:

Characteristics of hospitals included in the study

Characteristic	No. (%)
Characteristic	Experimental group n = 21	Control group n = 22
University hospital	13 (61.9)	18 (81.8)
Availability of a nurse on call	17 (81.0)	17 (77.3)
Residents assist in care of patients with upper gastrointestinal bleeding	16 (76.2)	16 (72.7)
Teaching/education/administrative activities occurred at site that could alter patient management	0 (0.0)	1 (4.6)
More than 400 beds	14 (66.7)	15 (68.2)
Geographic location
East^*	10 (47.6)	7 (31.8)
Ontario and West^†	11 (52.4)	15 (68.2)

↵* Quebec, Newfoundland, Nova Scotia, New Brunswick and Prince Edward Island.
↵† British Columbia, Alberta, Saskatchewan and Manitoba.

View popup

Table 3:

Characteristics of patients with available outcome data, before and after intervention

Characteristic	Before intervention, % (no. of patients)^*		After intervention, % (no. of patients)^*

	Experimental group n = 402	Control group n = 424	Experimental group n = 361	Control group n = 389
Age, yr, mean ± SD	68.6 ± 15.3	66.3 ± 15.7	68.0 ± 14.8	67.8 ± 16.4

Sex, male	240 (59.7)	274 (64.6)	218 (60.4)	220 (56.6)

No. of comorbid illnesses, median (range)	2.0 (0–10)	2.0 (0–7)	3.0 (0–8)	2.0 (0–9)

Inpatient onset of bleeding	106 (26.4)	58 (13.7)	71 (19.7)	69 (17.7)

Rockall score recorded	0 (0.0)	0 (0.0)	7 (1.9)	0 (0.0)

Blatchford score^†	0 (0.0)	1 (0.2)	0 (0.0)	0 (0.0)

Presence of initial hemodynamic instability	110 (27.4)	129 (30.4)	134 (37.1)	113 (29.1)

Bleeding site: stomach	211 (52.5)	246 (58.0)	191 (52.9)	222 (57.1)

Bleeding stigmata identified at endoscopy

High-risk lesion	177 (44.0)	186 (43.9)	153 (42.4)	144 (37.0)

Active bleeding	75 (18.7)	83 (19.6)	65 (18.0)	63 (16.2)

Adherent clot	48 (11.9)	46 (10.9)	42 (11.6)	37 (9.5)

Nonbleeding visible vessel	54 (13.4)	57 (13.4)	46 (12.7)	44 (11.3)

Low-risk lesion	225 (56.0)	238 (56.1)	208 (57.6)	245 (63.0)

Clean base ulcer	176 (43.8)	199 (46.9)	182 (50.4)	208 (53.5)

Nonprotuberant pigmented dot	49 (12.2)	39 (9.2)	26 (7.2)	37 (9.5)

Note: SD = standard deviation.
↵* Unless otherwise indicated.
↵† The Blatchford score is a risk stratification scale used in the pre-endoscopic assessment of patients with upper gastrointestinal bleeding.

View popup

Table 4:

Clinical outcomes before and after intervention

Characteristic	Before intervention, % (no. of patients)^*		After intervention, % (no. of patients)^*		Between-group comparison, p value

	Experimental group	Control group	Experimental group	Control group	Unadjusted^†	Adjusted^‡
High-risk patients who received endoscopic therapy	67.8 (152/177)	83.3 (155/186)	94.7 (143/151)	86.1 (124/144)	0.09	0.05

Index endoscopy within 24 h	62.8 (252/401)	65.6 (278/424)	57.5 (207/360)	63.1 (245/388)	0.3	0.1

Time to index endoscopy, h, mean ± SD^†	38.0 ± 69.8 n = 401	28.1 ± 40.1 n = 424	33.5 ±51.6 n = 361	28.2 ± 37.2 n = 388	0.8	0.6

Targeted irrigation of adherent clot performed	68.8 (33/48)	54.4 (25/46)	73.8 (31/42)	70.3 (26/37)	0.8	–

Endoscopic treatment performed	44.0 (177/402)	40.1 (170/424)	42.9 (155/361)	38.8 (151/389)	0.5	0.9

Injection	37.3 (150/402)	31.4 (133/424)	36.0 (130/361)	29.3 (114/389)	0.2	0.9

Epinephrine/saline alone	30.4 (122/402)	30.4 (129/424)	28.5 (103/361)	27.0 (105/389)	0.8	0.9

Other	7.0 (28/402)	0.9 (4/424)	7.5 (27/361)	2.3 (9/389)	0.2	–

Thermal	18.4 (74/402)	21.9 (93/424)	19.4 (70/361)	17.7 (69/389)	0.7	0.4

Clips	16.9 (68/402)	9.7 (41/424)	14.4 (52/361)	13.6 (53/389)	0.8	0.3

Other (ligation)	0.0 (0/402)	0.7 (3/424)	1.1 (4/361)	0.3 (1/389)	0.1	–

Pre initial endoscopy, total daily dose of PPI, mg, mean ± SD

Oral	66.4 ± 46.4 n = 45	68.3 ± 69.4 n = 48	80.0 ± 46.4 n = 37	68.6 ± 49.5 n = 49	0.4	0.3

Intravenous bolus	80.0 ± 24.3 n = 277	81.4 ± 25.8 n = 288	81.4 ± 23.1 n = 254	81.5 ± 21.5 n = 288	0.95	0.9

Intravenous constant infusion	145.0 ± 150.1 n = 273	124.1 ± 129.3 n = 289	124.2 ± 126.6 n = 254	120.1 ± 113.8 n = 288	0.7	0.7

Post initial endoscopy, total daily dose of PPI, mg, mean ± SD

Oral	91.7 ± 64.1 n = 196	98.8 ± 77.7 n = 246	116.8 ± 74.9 n = 206	116.1 ± 69.0 n = 181	0.95	0.3

Intravenous bolus	102.1 ± 53.6 n = 76	97.4 ± 68.6 n = 62	104.5 ± 80.8 n = 62	100.4 ± 57.3 n = 51	0.7	0.7

Intravenous constant infusion	380.4 ± 245.9 n = 285	380.5 ±233.3 n = 306	430.7 ± 258.8 n = 252	402.6 ± 264.5 n = 288	0.3	0.6

Intensive care unit admission	15.7 (63/402)	14.9 (63/424)	15.5 (56/361)	13.1 (51/389)	0.6	0.5

Intensive care unit duration of stay, d, mean ± SD^†	6.0 ± 8.1 n = 63	4.4 ± 5.0 n = 63	4.7 ± 3.7 n = 56	6.5 ± 8.4 n = 51	0.4	0.4

Re-bleeding, need for surgery or angiography to stop bleeding	9.2 (37/402)	9.0 (38/424)	11.9 (43/361)	8.2 (32/389)	0.1	–

Re-bleeding	8.0 (32/402)	8.0 (34/424)	11.1 (40/361)	7.2 (28/389)	0.1	–

Need for surgery to stop bleeding	2.5 (10/402)	1.7 (7/424)	1.9 (7/361)	2.1 (8/389)	0.9	–

Need for angiographic procedure to stop bleeding	2.5 (7/280)	2.0 (6/300)	2.8 (10/361)	2.1 (8/389)	0.5	–

Deaths	6.0 (24/402)	6.4 (27/424)	6.4 (23/361)	5.7 (22/389)	0.7	–

Bleeding-related death	1.5 (6/402)	0.7 (3/424)	2.2 (8/361)	2.3 (9/389)	0.9	–

Length of stay, d, mean ± SD^§	10.6 ± 15.4 n = 362	9.6 ± 15.7 n = 381	9.1 ± 10.1 n = 331	9.4 ± 12.1 n = 351	0.8	0.7

Hospital stay less than 3 d^¶	20.0 (77/386)	26.7 (109/408)	20.1 (71/354)	22.3 (83/373)	0.7	0.9

High-risk patients with hospital stay less than 3 d	6.9 (12/174)	14.6 (27/185)	8.5 (13/153)	11.8 (17/144)	0.4	0.2

Note: PPI = proton pump inhibitors, SD = standard deviation.
↵* Unless otherwise indicated.
↵† χ² for categorical variables or t-tests for continuous variables.
↵‡ Adjusted for baseline, hospital size (≤ 400 v. > 400 beds), Ontario (yes v. no) and interaction of hospital size and Ontario location. Adjustment for site was made for p values using the generalized estimating equations for dichotomous outcomes and mixed models for continuous outcomes.
↵§ Time to index endoscopy, ICU length of stay and hospital length of stay were log-transformed to improve normality.
↵¶ Unless patient died during hospital stay.