Abstract
Background:
Gastrointestinal hemorrhage is a common cause of hospital admission. However, there are little data to inform practice around blood count monitoring – a cornerstone of management. We hypothesize that more frequent testing leads to increased resource utilization without improvement in patient outcomes.
Methods:
This retrospective observational cohort study examined all patients admitted to a large academic medical institution primarily for gastrointestinal bleeding between July 10, 2014 and January 1, 2018. We identified 1150 patients admitted for gastrointestinal hemorrhage. Patients under 18, who developed bleeding while hospitalized, or who were transferred were excluded. The primary outcome was the number of complete blood counts collected in the first 48 hours of admission. Propensity matched analysis was performed to assess blood transfusion, units of blood transfused, time-to-endoscopy, mortality, and 30-day readmission rate.
Results:
On average, 5.6 complete blood counts were collected in the first 48 hours. 67% of the cohort was transfused (average of 2.6 units of packed red blood cells). When matched for comorbidity, anticoagulant use, location (ward vs. intensive care unit), vital signs, hemoglobin level, and INR, patients having more frequent monitoring had similar hospital length of stay and mortality rates, but were more likely to receive a blood transfusion (0.93 vs. 0.76, p<0.05), and if transfused, receive more blood (4 vs. 2 units, p<0.05).
Conclusion:
Blood count monitoring occurs more frequently than is likely necessary, is associated with a higher likelihood of blood transfusion, and does not affect patient outcomes, suggesting patient care might be improved by less frequent monitoring.
Keywords: gastrointestinal bleeding
Introduction
Gastrointestinal bleeding is a common cause of hospital admission with widely agreed upon approaches to management. The incidence is estimated to be anywhere from 48–160 patients per 100,000 population1,2. Patients are typically triaged based on the severity of bleeding and the presence or absence of comorbidities. Those judged to have more severe bleeding are frequently admitted to intensive care units while those with milder bleeding are admitted to standard hospital floor beds. In both settings, the overarching goal of management is maintaining patient stability until the bleeding resolves and/or an intervention is performed3. Monitoring the patient is typically accomplished by following hemodynamics and trending blood count levels. Hemodynamic instability and rapidly decreasing or low blood counts typically necessitate intervention - in the form of transfusion, endoscopy, or other invasive procedures.
Practitioners typically rely on monitoring of blood counts to judge whether a patient should have resuscitation with blood products. While this approach is commonly used, there is no agreed upon frequency with which blood counts should be monitored. Frequency is often determined by the perceived severity of bleeding or the comfort level and experience of the practitioner in caring for acute gastrointestinal bleeding. To our knowledge, there is no evidence addressing the appropriate frequency of blood count monitoring in patients with gastrointestinal bleeding. In our experience, blood counts are often collected more frequently than likely necessary, leading to unnecessary harm and increased cost while further reducing blood counts.
In this study, we hypothesized that while blood count testing in critical situations is an important cornerstone of patient management, it is performed at unreasonably frequent intervals. Therefore, we performed a detailed investigation of the frequency with which blood counts are monitored in patients with gastrointestinal bleeding.
Methods
Study population
This was a retrospective observational cohort study of all patients admitted to our institution with acute gastrointestinal bleeding between July 10, 2014 and January 01, 2018. We excluded patients below the age of 18 years, those that developed a gastrointestinal bleed at any time other than prior to their admission and those who were transferred to our institution from another hospital for acute gastrointestinal bleeding. The institution’s Clinical Data Warehouse was used to identify all patients admitted with upper and lower gastrointestinal bleeding using ICD-9 and ICD-10 codes. This study was approved by the institutional review board at the Medical University of South Carolina.
Data collection and definitions
Clinical variables were abstracted at the time of admission and included baseline demographics, medical and surgical history, Charlson Comorbidity Index (CCI), outpatient use of antiplatelet and anticoagulation therapy, vital signs and laboratory data. Hospital variables collected included number of blood draws in the first 48 hours of admission, time to endoscopic procedure, blood transfusions, hospital length of stay, hospital disposition, and mortality. Medical records were then reviewed (IJ and RS) to determine the endoscopic findings and the specific source of bleeding. Upper GI bleeding was defined as hematemesis, melenemesis, or melena or a documented upper GI source of bleeding identified by endoscopy. Lower GI bleeding was defined as hematochezia or melena with no source of bleeding identified by upper endoscopy, or a documented lower GI source of bleeding identified by colonoscopy.
The principal aim of this study was to determine the frequency of complete blood count draws during the first 48 hours of hospitalization in patients presenting with gastrointestinal bleeding. Secondary outcomes included number of transfusions received during the admission, time to endoscopic intervention, hospital length of stay, and hospital death.
Statistics
Baseline and outcome variables were presented as mean ± SD for continuous variables, and frequencies and percentages for categorical variables. Comparison of variables (baseline and outcome variables) between patients who received 1–3, 4–6 and ≥7 complete blood counts during their first 48 hours of hospital stay was carried out using the Chi- square or fisher test and one-way anova or Kruskal-Wallis test as indicated.
We then conducted a propensity score matched analysis using multinomial logistic regression to match patient in 3 groups (1–3, 4–6, ≥7 complete blood count blood draws during their first 48 hours of hospital stay) for their baseline characteristics, which included initial hemoglobin, vital signs on presentation (blood pressure and heart rate), admission to the ICU, diagnosis of cirrhosis, Charlson Comorbidity Index score, and use of anticoagulation or antiplatelet therapy4. Significance was considered at a p-value less than or equal to 0.05 and statistical analyses were performed using SPSS (version 24).
Results
Of 860 patients meeting inclusion criteria, 45% were female with an average age of 62 (Table 1). The average Charlson Comorbidity Index was 2 and 16% of patients were on either antiplatelet therapy or anticoagulation (Table 1). The mean admission hemoglobin was 8.8 g/dL. The majority of the population (71%) was admitted to the general medicine ward while 29% required ICU admission.
Table 1.
Clinical features
| Variables | N (%) or Mean (SD) |
|---|---|
| N=860 | |
| Age (years) | 62 (16) |
| Gender (F) | 385 (45%) |
| Race | |
| White | 502 (58%) |
| Black | 325 (38%) |
| Other | 34 (4%) |
| CCI score | 2 (2) |
| Cirrhosis | 136 (16%) |
| Antiplatelet therapy | 80 (9%) |
| Anticoagulation therapy | 63 (7%) |
| Required ICU admission | 249 (29%) |
| Systolic BP | 119 (26) |
| Diastolic BP | 69 (19) |
| HR | 93 (19) |
| Hemoglobin | 8.8 (1.9) |
| Hematocrit | 44 (26) |
| Platelets | 186 (80) |
| INR | 1.3 (0.6) |
Nearly all patients (96%) underwent endoscopy; most received an EGD (56%) compared to colonoscopy (24%), capsule endoscopy (7%), and enteroscopy (7%) (Table 2). Of the patients who had a presumed source of their bleeding identified (71%), 49% were classified as upper GI hemorrhages and 22% were determined to be lower GI bleeds. The most common etiologies of upper GI bleeding were peptic ulcer disease (31%), esophageal varices (18%), esophagitis (9%), and arteriovenous malformations (9%). The most common identified sources of lower GI bleeding were hemorrhoids (8%) and diverticulosis (8%) (Table 2).
Table 2.
Endoscopic findings*.
| Upper Endoscopy | 483 (56%) |
| Peptic ulcer disease | 150 (31%) |
| Esophageal varices | 87 (18%) |
| AVM | 45 (9%) |
| Esophagitis | 45 (9%) |
| Portal hypertensive gastropathy | 18 (4%) |
| Gastritis | 17 (4%) |
| Mass | 8 (2%) |
| Other | 40 (8%) |
| No lesion identified | 73 (15%) |
| Lower Endoscopy | 209 (24%) |
| Hemorrhoids | 73 (35%) |
| Diverticulosis | 71 (34%) |
| Other | 22 (11%) |
| No lesion identified | 41 (20%) |
| Enteroscopy | 61 (7%) |
| AVM | 25 (41%) |
| No lesion identified | 36 (59%) |
| Capsule Endoscopy | 61 (7%) |
| AVM | 15 (25%) |
| Other | 5 (8%) |
| No lesion identified | 41 (67%) |
| No Endoscopy | 46 (6%) |
The lesion at endoscopy that demonstrated the putative culprit bleeding lesion is shown. No patient had two bleeding lesions identified.
Patients were divided into three groups based on the number of complete blood counts they had collected over the first 48 hours of their admission: 3 or fewer (least frequent), 4–6 (intermediately frequent), and ≥7 (most frequent). In the first 48 hours of hospitalization, 48% of patients had 3 or fewer complete blood counts collected, while 29% had 4–6 complete blood counts and 24% had 7 or more complete blood counts obtained (Table 3). The maximum number of complete blood counts collected in the first 48 hours was 14 while the minimum number was one. In the baseline analysis, clinical characteristics at presentation were generally similar among the different groups with a few exceptions; they had similar heart rates and diastolic blood pressure, although the most frequent blood draw group had lower systolic blood pressure. The mean hemoglobin was 9.1 mg/dL in the least frequent blood draw group and 8.3 mg/dL in the most frequent blood draw group. All three groups had similar demographic features, average Charlson Comorbidity Index score of 2 and similar rates of cirrhosis and use of antiplatelet or anticoagulation therapy. Patients who underwent more frequent blood monitoring were more likely to receive a transfusion (55% v 69% v 89%) and, if transfused, received more units of blood (2 units v 3 units). Interestingly, in the unmatched analysis, mortality was greatest in patients who had fewer complete blood counts drawn.
Table 3.
Complete blood count monitoring, patient characteristics, and outcomes
| Variables | Number of CBCs drawn in the first 48 hours of admission for GI bleed | ||
|---|---|---|---|
| 1–3 (N=409) | 4–6 (N=248) | 7 (N=203) | |
| Age (years) | 62 (16) | 61 (16) | 62 (15) |
| Gender (F) | 173 (42%) | 114 (46%) | 98 (48%) |
| White | 254 (62%) | 132 (53%) | 116 (57%) |
| Black | 135 (33%) | 110 (44%) | 80 (39%) |
| Other | 20 (5%) | 6 (2%) | 7 (4%) |
| CCI score | 2 (2) | 2 (2) | 2 (2) |
| Cirrhosis | 62 (15%) | 35 (14%) | 39 (19%) |
| Antiplatelet therapy | 54 (13%) | 13 (5%) | 13 (6%) |
| Anticoagulation therapy | 46 (11%) | 13 (5%) | 4 (2%) |
| ICU admission | 101 (25%) | 67 (27%) | 81 (40%) |
| Systolic BP (mmHg) | 126 (26) | 117 (28) | 110 (23) |
| Diastolic BP (mmHg) | 69 (19) | 71 (19) | 68 (17) |
| HR | 91 (18) | 94 (20) | 97 (21) |
| Hemoglobin (g/dL) | 9.1 (2.0) | 8.5 (1.7) | 8.3 (1.6) |
| Platelets/μl | 196 (72) | 185 (86) | 170 (86) |
| INR | 1.3 (0.6) | 1.3 (0.5) | 1.4 (0.7) |
| APTT (s) | 34 (13) | 32 (12) | 34 (14) |
| BUN (mg/dL) | 27 (20) | 33 (23) | 36 (27) |
| Hospital LOS (days) | 4.7 (3.3) | 4.6 (2.6) | 5.2 (2.9) |
| RBC transfusion | 225 (55%) | 172 (69%) | 180 (89%) |
| RBC units | 2 (2) | 2 (2) | 3 (2) |
| Death | 31 (8%) | 8 (3%) | 6 (3%) |
| 30-day readmission | 139 (34%) | 89 (36%) | 76 (37%) |
CCI – Charlson Comorbidity Index
ICU – intensive care unit
BP- blood pressure
HR – heart rate
RBC – red blood cell
LOS – length of stay
Analysis with propensity score matching revealed no significant differences in baseline clinical characteristics. In particular, ICU admission was similar among the groups, and there were no significant differences in hemodynamic features (systolic blood pressure or heart rate) among the groups (Table 4). There were also no significant differences between groups in admission hemoglobin (9 g/dL vs. 8 g/dL vs. 8 g/dL). However, there were statistically significant differences in blood transfusion among groups. The group of patients having fewer complete blood counts drawn were transfused approximately three-quarters of the time, while those with ≥7 CBCs drawn were almost universally transfused. Additionally, patients undergoing more frequent monitoring received more total units of blood (4 vs. 2). There was no significant difference in hospital LOS, mortality, or readmission rate based on frequency of monitoring.
Table 4.
Propensity score matched outcomes
| Variables | Number of complete blood counts drawn in the first 48 hours of admission for GI bleed | P-value | ||
|---|---|---|---|---|
| 1–3 (N=106) | 4–6 (N=106) | 7 (N=106) | ||
| CCI score | 3 (1) | 2 (2) | 3 (2) | NS |
| Cirrhosis | 26 (25%) | 13 (12%) | 22 (21%) | NS |
| Antiplatelet therapy | 2 (2%) | 5 (5%) | 4 (4%) | NS |
| Anticoagulation therapy | 2 (2%) | 6 (6%) | 1 (1%) | NS |
| ICU admission | 44 (42%) | 33 (31%) | 41 (39%) | NS |
| Systolic BP (mmHg) | 106 (24) | 104 (25) | 104 (19) | NS |
| HR | 97 (18) | 98 (21) | 99 (22) | NS |
| Hemoglobin (g/dL) | 8.1 (1.4) | 8.5 (1.8) | 8.2 (1.2) | NS |
| Platelets/μl | 174 (64) | 191 (83) | 169 (52) | NS |
| RBC transfusion | 80 (76%) | 80 (76%) | 98 (93%) | <0.05 |
| RBC units | 2 (1) | 2 (1) | 4 (2) | <0.05 |
| Time to endoscopy (hrs) | 42 | 41 | 39 | NS |
| Hospital LOS (days) | 5.2 (3.4) | 5.1 (2.6) | 5.4 (3.2) | NS |
| Death | 11 (10%) | 5 (5%) | 6 (6%) | NS |
| 30-day readmission | 39 (37%) | 40 (38%) | 41 (39%) | NS |
CCI – Charlson Comorbidity Index
ICU – intensive care unit
BP- blood pressure
HR – heart rate
RBC – red blood cell
LOS – length of stay
Discussion
Here, we have shown that more frequent monitoring is associated with a higher rate of blood transfusion. Furthermore, patients who received blood transfusion were given a larger total amount of blood. It was also notable that more frequent testing did not affect outcomes, such as time to endoscopic intervention, hospital length of stay or readmission.
We did not identify a statistically significant difference in mortality between the propensity score matched groups. However, analysis of the entire cohort demonstrated that there were more patient deaths in patients having less frequent monitoring. This is unlikely to be a result of less severe disease in this group since this was a propensity match analysis. Rather, it raises the real possibility that a higher mortality rate could be an untoward result of more frequent blood count monitoring (due to any of a number of possible downstream events such as more frequent blood transfusion, infection, anemia, etc…). However, we urge caution in interpreting this result, and propose that further investigation to address this issue is warranted.
In our study, the demographics of our population and etiologies of hemorrhage were consistent with those reported in previously published studies examining gastrointestinal bleeding5,6,7,8,9,10. Admission hemodynamics and ICU admission were used as markers of bleeding severity and were not significantly different between propensity matched groups. This suggests that the increased rate of transfusion in patients undergoing more frequent blood count monitoring was not due to their degree of hemorrhage. Lower admission hemoglobin may prompt practitioners to transfuse a patient, but our results suggest initial blood counts do not affect this theoretical tendency. Finally, we did not identify significant differences in comorbidities, as measured by Charlson Comorbidity Index score, use of antiplatelet or anticoagulant use, or diagnosis of cirrhosis, that may encourage more liberal blood transfusion.
Limiting blood transfusion is beneficial from multiple standpoints – including by reducing the risk of transfusion reactions, and decreasing the financial, physical, and psychological costs associated with blood transfusion. There has been significant investigation in the area of transfusion thresholds, leading to well defined limits at which patients should receive blood11,12. Guidelines published by the American Association of Blood Banks in 2016 recommend transfusion for hemoglobin 7–8g/dL as opposed to 9–10g/dL13. Reducing transfusion preserves a finite resource while also decreasing the likelihood of adverse events caused by blood transfusion, such as acute hemolytic reactions, transfusion associated circulatory overload, transfusion related acute lung injury, and transmission of infection.
Additionally, limiting unnecessary laboratory testing is vital in the current age of maximizing quality while reducing spending. A recent study examined inappropriate lab testing and found a 7.4% rate of unnecessary ordering of repeat lab tests, such as serial blood count monitoring14. The costs associated with collecting and running a complete blood count include blood collection supplies, nursing time, and use of lab equipment. If blood count monitoring is being performed too frequently, as our study suggests, this is a prime opportunity to reduce these costs.
In addition to the financial cost of laboratory testing, there is the psychological effect of frequent phlebotomy. For one, blood draws are often painful, and thus cause physical and emotional discomfort for many patients. Additionally, blood draws are often performed during sleeping hours, and it is well appreciated that lab draws disrupt sleep, which is oftentimes necessary in the critically ill patient. Sleep deprivation and disruption of circadian rhythm is connected with increased susceptibility to infection, risk of hospital delirium, and disruptions in endocrine homeostasis15.
We recognize limitations of this study. First, given that this was a retrospective observational study, ascertainment bias is always a concern. However, we performed a rigorous propensity matched analysis, which controlled for baseline clinical characteristics, and in this group, vital signs and the need for ICU admission were closely matched. We intentionally did not stratify by etiology of bleeding, which raises the possibility that specific causes of hemorrhage may affect transfusion outcomes differently. For example, patients with peptic ulcer disease may be more aggressively transfused than patients with diverticular bleeding. However, endoscopic stigmata were similar in the different groups (as in Tables 3 and 4), suggesting that this was not a confounding variable. Additionally, since this study was performed at a single center, our results may not be generalizable to other institutions. A further point of caution is that while we included patients with severe hemorrhage in this study, including those with variceal bleeding or bleeding from a visible vessel, many patients in our study had less severe bleeding. As a result, the findings of this study are likely more applicable to patients with less aggressive bleeding and may not be applicable in cases of extremely aggressive bleeding; clinical judgement should be applied on a case by case basis in these patients. Despite these limitations, our findings provide sorely needed evidence to guide blood count monitoring in acute gastrointestinal bleeding.
Conclusion
We conclude that blood count monitoring is performed more often than is likely necessary in hospitalized patients with gastrointestinal bleeding. More frequent monitoring was associated with a higher likelihood of blood transfusion, while not affecting patient outcomes. This relationship persisted independent of patient comorbidities, severity of bleeding, and admission blood levels. Our results suggest that given the number of patients admitted with gastrointestinal bleeding and the physical and financial costs of laboratory testing, more frequent laboratory testing may harm patients and waste resources. We conclude that there are opportunities to develop better, data-driven approaches to blood count monitoring in patients with gastrointestinal bleeding and that patient care might be improved by less frequent blood count monitoring.
Figure 1. Patient cohort.
A consort diagram of the patients included in the study is shown.
Clinical Significance.
There are little data to inform practice around blood count monitoring – a cornerstone of management of patients with gastrointestinal bleeding.
More frequent blood count monitoring in patients with gastrointestinal bleeding was associated with a higher likelihood of blood transfusion, while not affecting patient outcomes. This relationship persisted independent of patient comorbidities, severity of bleeding, and admission blood levels.
Blood count monitoring occurred more frequently than is likely necessary.
Acknowledgments
Funding Source: DCR was supported in part by the National Institutes of Health, P30 DK123704.
Footnotes
Conflicts of Interest: The authors have no conflicts of interest relevant to this work to disclose.
Publisher's Disclaimer: This is a PDF file of an unedited manuscript that has been accepted for publication. As a service to our customers we are providing this early version of the manuscript. The manuscript will undergo copyediting, typesetting, and review of the resulting proof before it is published in its final form. Please note that during the production process errors may be discovered which could affect the content, and all legal disclaimers that apply to the journal pertain.
References
- 1.Wuerth BA, & Rockey DC Changing epidemiology of upper gastrointestinal hemorrhage in the last decade: A nationwide analysis. Digestive Diseases and Sciences. 2017;63(5):1286–1293. doi: 10.1007/s10620-017-4882-6 [DOI] [PubMed] [Google Scholar]
- 2.El-Tawil AM Trends on gastrointestinal bleeding and mortality: Where are we standing? World Journal of Gastroenterology. 2012;18(11):1154. doi: 10.3748/wjg.v18.i11.1154 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 3.Barkun A, Bardou M, Kuipers E, Sung J, Hunt R, Martel M, et al. International consensus recommendations on the management of patients with nonvariceal upper gastrointestinal bleeding. Annals of Internal Medicine. 2010;152(2):101. doi: 10.7326/0003-4819-152-2-201001190-00009 [DOI] [PubMed] [Google Scholar]
- 4.McCaffrey DF, Griffin BA, Almirall D, Slaughter M, Ramchang R, Burgette L A tutorial on propensity score estimation for multiple treatments using generalized boosted models. Stat Med 2013;32(19):3388–414. doi: 10.1002/sim.5753 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 5.Hreinsson J, Gumundsson S, Kalaitzakis E, Bjornsson E Lower gastrointestinal bleeding. European Journal of Gastroenterology & Hepatology. 2013;25(1):37–43. doi: 10.1097/meg.0b013e32835948e3 [DOI] [PubMed] [Google Scholar]
- 6.Leerdam MV Epidemiology of acute upper gastrointestinal bleeding. Best Practice & Research Clinical Gastroenterology. 2008;22(2):209–224. doi: 10.1016/j.bpg.2007.10.011 [DOI] [PubMed] [Google Scholar]
- 7.Hearnshaw S, Logan R, Lowe D, Travis S, Murphy M, Palmer K Acute upper gastrointestinal bleeding in the UK: Patient characteristics, diagnoses and outcomes in the 2007 UK audit. Gut. 2011.60(10):1327–1335. doi: 10.1136/gut.2010.228437 [DOI] [PubMed] [Google Scholar]
- 8.Gralnek IM, Neeman Z, Strate LL. Acute lower gastrointestinal bleeding. New England Journal of Medicine. 2017;376(11):1054–1063. doi: 10.1056/nejmcp1603455. [DOI] [PubMed] [Google Scholar]
- 9.Kamboj AK, Hoversten P, Leggett CL. Upper Gastrointestinal Bleeding: Etiologies and Management. Mayo Clinic Proceedings. 2019;94(4):697–703. doi: 10.1016/j.mayocp.2019.01.022. [DOI] [PubMed] [Google Scholar]
- 10.Hreinsson JP, Kalaitzakis E, Gudmundsson S, Björnsson ES. Upper gastrointestinal bleeding: incidence, etiology and outcomes in a population-based setting. Scandinavian Journal of Gastroenterology. 2013;48(4):439–447. doi: 10.3109/00365521.2012.763174. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 11.Villanueva C, Colomo A, Bosch A, Concepcion M, Hernandez-Gea V, Aracil C, et al. Transfusion strategies for acute upper gastrointestinal bleeding. New England Journal of Medicine. 2013;368(24)2341–2341. doi: 10.1056/nejmx130015 [DOI] [PubMed] [Google Scholar]
- 12.Odutayo A, Desborough M, Trivella M, Stanley A, Doree C, Collins G, et al. Restrictive versus liberal blood transfusion for gastrointestinal bleeding: A systematic review and meta-analysis of randomised controlled trials. The Lancet Gastroenterology & Hepatology. 2013;2(5):354–360. doi: 10.1016/s2468-1253(17)30054-7 [DOI] [PubMed] [Google Scholar]
- 13.Carson J, Guyatt G, Heddle N, Grossman B, Cohn C, Fung M, et al. Clinical practice guidelines from the AABB. JAMA, 2016;316(19):2025. doi: 10.1001/jama.2016.9185 [DOI] [PubMed] [Google Scholar]
- 14.Zhi M, Ding E, Theisen-Toupal J, Whelan J, Arnaout R The landscape of inappropriate laboratory testing: A 15-year meta-analysis. PLoS ONE. 2013;8(11). doi: 10.1371/journal.pone.0078962 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 15.Pisani M, Friese R, Gehlbach B, Schwab R, Weinhouse G, Jones S Sleep in the intensive care unit. American Journal of Respiratory and Critical Care Medicine. 2015;191(7):731–738. doi: 10.1164/rccm.201411-2099ci [DOI] [PMC free article] [PubMed] [Google Scholar]