Clinical Risk Characteristics of Upper Gastrointestinal Hemorrhage Severity: A Multivariable Risk Analysis

Background Upper gastrointestinal hemorrhage (UGIH) is one of the common clinical manifestations encountered in most emergency departments. Patient characteristics indicating UGIH severity in developing countries may be different from those in developed countries. The present study was designed to explore clinical prognostic indicators for UGIH severity. Methods A retrospective cohort study was conducted in a university affiliated tertiary hospital in Kamphaeng Phet, Thailand. Medical folders of patients with UGIH were reviewed. Patients were grouped into 3 severity levels, based on criteria proposed by The American College of Surgeon. Pre-defined prognostic indicators were compared. The prognostic indicators for UGIH severity were analyzed by a multivariable continuation ratio ordinal logistic regression and presented with odds ratios. Results From 1,043 eligible medical folders, 984 (94.3%) complete folders were used in analysis. There were 241, 631 and 112 patients in the mild, moderate and severe UGIH groups. Six independent indicators of severe UGIH were, hemoglobin < 100 g/dL (OR = 13.82, 95% CI = 9.40 to 20.33, P < 0.001), systolic blood pressure < 100 mmHg (OR = 11.01, 95% CI = 7.41 to 16.36, P < 0.001), presence of hepatic failure (OR = 5.50, 95% CI = 1.14 to26.64, P = 0.037), presence of cirrhosis (OR = 2.03, 95% CI = 1.32 to 3.11, P = 0.001), blood urea nitrogen ≥ 35 mmol/L (OR = 1.73, 95% CI = 1.25 to 2.40, P = 0.001), and pulse rate ≥ 100 per minute (OR = 1.72, 95% CI = 1.21 to 2.45, P = 0.003). Conclusions Pulse rate ≥ 100 per minute, systolic blood pressure < 100 mmHg, hemoglobin < 10 g/dL, blood urea nitrogen ≥ 35 mmol/L, presence of cirrhosis and presence of hepatic failure are prognostic indicators for an increase in UGIH severity levels. They are potentially useful in UGIH risk stratification.


Introduction
Upper gastrointestinal hemorrhage (UGIH) is one among the common clinical manifestations encountered in most emergency departments. The annual incidence varies from 50 to 150 episodes per 100,000 populations [1], with 11-14% case fatality [2,3]. Mortalities from UGIH increase in elderly patients and in patients with co-morbidities [4,5]. Advances in medical technology, particularly endoscopic instrumentations, in the past 10 years do not seem to decrease case fatality from UGIH [2,3,5]. The costs of patient care are still very high in the United States, approximately 750 million US dollars per year [6].
Usual managements of patients with UGIH began with patients screening, clinical assessment and evaluation, resuscitation, endoscopy [7] and surgical intervention when indicated.
Endoscopy is essential and is the key measures to evaluate the risks of re-bleeding and mortality. At the same time, endoscopy is valuable for the pathological and anatomical diagnosis, patient risk stratification and hemostasis [8]. Endoscopy is usually recommended within 24 hours after the patient vital signs are stabilized. Although endoscopy is very accurate in patient risk stratification [9], it may not be performed within the first 24 hours in all patients with UGIH, especially in areas with limitations in health resources and/ or medical personnel. Studies over the past 10 years were focused on how to triage UGIH cases into categories such as emergent, semi-elective or as out-patients, in order to define appropriate time for endoscopy, but the results are still inconclusive [8].
Various clinical risk characteristics were adopted in the process of patient assessment. These characteristics included Manuscript accepted for publication August 15, 2012 a the patient age, presence of shock, professional diagnosis, hemoglobin level measured on arrival at emergency rooms, presenting symptoms, size of ulcers, stigmata of hemorrhage and blood transfusion [10][11][12][13][14][15][16][17][18][19][20]. Those clinical risks were further applied into a prognostic scoring algorithm to evaluate UGIH patients risk of death and/or re-bleeding [7], and also to screen for and to select high risk patients into intervention within an appropriate time.
The most well known prognostic scores are probably "The Rockall Score" [21] and "The Blatchford Score" [22]. The Rockall Score was developed in the UK in 1995 [2], published in 1996 [3] and was validated in the following year [23]. The objective of the score was to predict patient poor clinical outcomes. It was also validated in many other settings, but with diverse conclusions. Some authors reported good prediction for re-bleeding, but poor prediction for death [24,25]. The others reported the opposite directions [26,27]. However, all authors agreed that a decrease in scores results in an over-estimation and that an increase in scores results in a loss of discrimination [7]. The Blatchford Score was proposed in 2000 [22] to be used in patients evaluation before endoscopy, and to search for patients requiring intervention, such as blood transfusion, hemostasis, either with endoscopy and/or surgery [22]. The score comprised both clinical examinations and laboratory tests [2,22]. Validation studies of the score concluded that it can be used in screening UGIH patients on admission into those with high risk who needed blood transfusion, endoscopy or intervention, and patients with low risk who do not need such interventions [28][29][30]. External validation of the Blatchford score reported a high sensitivity of 99-100% in high risk patients [29,30], but a low specificity of 13% [30]. Other prognostic scores such as "The Baylor College Score" [31] and "The Cedars-Sinai Score" [32] were not worldwide, partly because they required early endoscopy and that re-bleeding was unacceptably under-estimated [7].
The clinical risk characteristics reported in developed countries may be different from developing countries. The present study was initiated to explore and investigate for the different clinical risk characteristics associated with the severity of UGIH in a tertiary hospital in a developing country.

Patients
Kamphaeng Phet Hospital is a university affiliated tertiary hospital located in the lower northern part of Thailand. It represents a common tertiary hospital of the country. Medical folders of patients admitted to the hospital with UGIH during 2009 -2010 were searched from the hospital computer database system, using ICD-10, K-920-hematemesis, K-921-melena, and K-922-gastro intestinal hemorrhage un-

Definitions of severity
The severity of UGIH was operationally defined by the following criteria: 1) Severe; patients who hemostasis was done under surgical intervention, patients with grade 3 and 4 shock ( Table 1) [33], or patients who died of UGIH; 2) Moderate; patients who hemostasis was done under endotherapy (endoscopy), patients with re-bleeding, patients with grade 1 and 2 shock [33], or patients who required blood transfusion; 3) Mild; patient without shock, patients who did not require any intervention under endoscopy, or patients who required no blood transfusion.

Data analysis
Patient parameters were compared across the 3 severity categories by chi-squared tests for trend and a two-way analysis of variance by rank (Friedmann ANOVA). To maximize the statistical power of analysis of the 3 ordinal clinical outcomes (severe, moderate and mild), we chose a multivariable ordinal continuation ratio logistic regression to analyze the prognostic indicators for UGIH severity.
The study protocol was approved by The Kamphaeng Phet Hospital Ethical Committee for Clinical Research. Patient consent forms were not required in this retrospective data review. Patient names and other traceable identifications were confidential and omitted in all process of data management.
The 2 authors received no outside grants and reported no conflicts of interests.
Patients were categorized into 3 levels of severity; mild (n = 241), moderate (n = 631) and severe (n = 112) following the aforementioned criteria. The three severity groups were similar in gender, presentation with hematemesis, and hematochezia ( Table 2). The characteristics that were different among the three groups were; age, presentation with coffee ground vomiting, melena, syncope, pulse rate, systolic blood pressure, hemoglobin level, blood urea nitrogen, presence of cirrhosis, hepatic failure, cardiac failure, and renal failure.

Discussion
The prognostic indicators for UGIH severity detected in our study agreed with findings in the previous studies of prognostic score [13,19], such as The Rockall Score (age, systolic blood pressure, pulse rate, presence of hepatic failure, cardiac failure, and renal failure) [2,3,23], and The Blatchford Score (blood urea nitrogen, hemoglobin systolic blood pressure, pulse rate, presentation with melena, syncope, presence of liver failure, and cardiac failure) [22,34,35]. These indicators are straightforward and need no further explanations. However, the age parameter was different from the Blatchford's and the other studies, proposing that the patient age [22,34,35] was not associated with a decision for intervention.
These similarities and discrepancies show that in developing countries, some clinical indicators detected in developed countries are similarly related to the severity of UGIH. It is worth noticing that presence of cirrhosis, the parameter not mentioned earlier in other studies, also increased the UGIH severity by 2.03 folds in our study. This may indicate that, in Thai patients, as well as in other developing countries, cirrhosis may also play an additional role in the severity of UGIH.
Patient medical files with incomplete information (5.7%) were excluded from analysis and may leave some doubts on representativeness of the study. However, we believe that the missing data were likely to be at random and that this small percentage of missing was unlikely to bias the overall findings.
The clinically and statistically important indicators detected from this study may be used in future investigations, such as to develop a prognostic scoring algorithm to screen for severity levels of patients presenting with UGIH.

Conclusions
The pulse rate ≥ 100 per minute, systolic blood pressure < 100 mmHg, hemoglobin < 10 g/dL, blood urea nitrogen ≥ 35 mmol/L, presence of cirrhosis and hepatic failure are the risk characteristics that increase UGIH severity levels. They may guide clinicians to pay some attentions to patients with these risks. For future implications, these risk characteristics may