Long-term mortality of patients admitted to the intensive care unit for gastrointestinal Bleeding.

Objectives: Long-term mortality data for gastrointestinal (GI) bleeders is scarce in the literature. The aim of this prospective study was to determine the long-term mortality of patients admitted to two intensive care units with a primary diagnosis of GI bleeding.

Methods: The charts of patients admitted to the medical intensive care unit (MICU) with GI bleeding were reviewed and the data of the patients' first day in the MICU was used to calculate APACHE III and Charlson scores. A GI bleeding score was computed by combining endoscopic findings and units of blood transfused during patients' MICU stay. Mortality data was obtained from the Vital Statistics Department of Montgomery County, Dayton, OH. Survival data and predictability of mortality based on these scores were assessed.

Results: Mean age of the 66 patient cohort was 58.6 years. Twenty-six of 51 patients with upper GI bleeding, five of seven patients with lower GI bleeding, and four of eight patients with unknown site of bleeding died within 7 years. Charlson score correlated significantly with the mortality prediction, whereas the APACHE III and bleeding scores did not.

Conclusions: All-cause and GI bleeding-related 7-year mortality for patients admitted to the MICU with GI bleeding was lower than the rates cited in the literature. The Charlson score was helpful in predicting mortality.

Key Words: APACHE III, Charlson Comorbidity Index, gastrointestinal bleeding, long-term mortality

**********

Long-term mortality data for gastrointestinal (GI) bleeders is scarce in the literature. (1) The primary aim of this prospective study was to evaluate long-term "all-cause" and GI bleeding-related mortality rates for patients admitted to the medical intensive care unit (MICU) with a primary diagnosis of GI bleeding. A secondary aim was to assess the value of APACHE III (Table 1), GI bleeding (Table 2), and Charlson (Table 3) scores in predicting long-term mortality.

Materials and Methods

This study protocol was approved by the institutional review boards of Wright State University and the two participating hospitals. Medical charts were reviewed for 66 patients with a primary diagnosis of GI bleeding who were admitted to MICUs in two major Midwest university teaching hospitals during 1995. In addition to demographics, data on the patients' first day of admission to the MICU were collected prospectively to calculate an APACHE III score (2) and a Charlson Comorbidity Index score. (3) The number of units of transfused blood that patients received during their stay in the MICU, and results of GI endoscopies (esophagogastroduodenoscopy, colonoscopy) were used to compile a bleeding score.

Mortality data were collected at periodic intervals from the Vital Statistics Division of Montgomery County, OH, through June 30, 2002. The cause of death (both primary and contributory) was determined from death certificates. If GI bleeding was given as the primary cause of death, the cause of death was considered as directly related to GI bleeding. If the primary cause of death was not related to bleeding, but GI bleeding was stated as the contributory cause, it was considered as an indirect cause of death. The survival period was estimated from the date of admission to the date of death, or through June 30, 2002.

Survival data and predictability of mortality risks were analyzed by the Kaplan-Meier method and Cox proportional hazards regression.

Results

Of 66 patients, 18 were women. The mean age of our subjects was 58 [+ or -] 16.1 years (range, 20-94 years). Fifty-four patients were admitted directly to the MICU, and 12 patients were transferred to the MICU from hospital wards. Of 54 direct admissions to the MICU, 25 died during the period studied, five as a direct result of GI bleeding, while in nine others GI bleeding was a contributory cause. Of 12 transfers to the MICU, 10 died during the study period. GI bleeding was the primary cause of death in 1 of these 10 patients, and in 3 others GI bleeding was a contributory cause of death.

Altogether there were 51 patients with upper GI (UGI) bleed: bleeding ulcer, 24; variceal bleeding, 12; and other bleeding lesions (eg, gastroesophageal reflux disease, erosions, Mallory-Weiss), 15. There were seven patients with lower GI (LGI) bleeding, and eight with bleeding from an unknown site. Forty-eight patients received two or more units of blood transfusion, and 18 received either one unit of blood or no transfusion. In total, 35 patients had died and 31 were alive as of June 30, 2002 (Fig. 1).

GI bleeding was the primary cause of death in six members of the study group, three of whom died (in less than 1 month) in the hospital. Twenty-nine patients died as a result of non-GI causes, and in 12 of these, GI bleeding was a contributory cause of death. Seven patients died in the hospital during their admission for GI bleeding, and in all of these GI bleeding either directly caused (four patients) or contributed to (three patients) their deaths. Of the five patients with variceal bleeding who died, GI bleeding was either directly or indirectly the cause in every case. The overall mortality rate for UGI bleeding was 51%, versus 71% for LGI bleeding, and 50% for patients with unknown site of bleeding. Seventy-four percent of deaths in patients with UGI bleeding and 33% of deaths in patients with LGI or unknown site of bleeding were either directly or indirectly related to the bleeding. There was no statistically significant difference between patients who died and patients who survived regarding 1) the source of bleeding (UGI, LGI, or unknown site), 2) having received blood transfusions, 3) the sex of the patients, or 4) whether the patients were direct admissions to the MICU or transfers from other areas of the hospital. As determined by the APACHE III, bleeding index, and Charlson scores, increased age and a higher Charlson score both indicated a significantly higher risk of mortality (Table 4).

[FIGURE 1 OMITTED]

Discussion

Mortality for UGI bleeding has been reported (4,5) to have been constant at around 8 to 10% over the past 40 years. Most of the mortality data in available studies are, however, related either to in-hospital or 30-day mortality. Mortality figures for GI bleeding that reach beyond the 1-month mark are scarce in the literature.

Higher mortality rates in patients with GI bleeding are attributed to associated comorbid conditions. Therefore, we used APACHE III and Charlson scores to evaluate the role of comorbid conditions in increasing mortality rates. The stigmata of recent hemorrhage has been a standard assessment in UGI bleeding for more than 20 years. In recent years, the same criteria are being applied to LGI bleeding as well. These criteria are very helpful to the clinician in deciding the source and severity of bleeding and in predicting rebleeding rates and short-term mortality.

The number of units of blood transfused has been used traditionally for referral to surgery and also for assessing the severity of GI bleeding. The severity of GI bleeding is accepted to be mild if the hemoglobin drop is less than 2 grams, massive if the continued need for blood transfusion is 3 units a day or more, and moderate for the cases in between these extremes. Most of ours were cases either of moderate or massive bleeding.

Endoscopic findings of stigmata of recent hemorrhage (especially for UGI bleeding) and the number of units of transfused blood required have independently been shown to correlate well with the severity of GI bleeding; thus, we combined both parameters to compute a bleeding score.

Aggarwal et al, (6) in a study of 582 patients with cirrhosis who were admitted to the MICU, found the in-hospital all-cause mortality rate of these patients to be 49%. An APACHE III score of greater than 90 was one of the three significant predictors of mortality, in addition to the use of mechanical ventilation and use of pressors. We could not determine the exact mortality for patients with variceal bleeding in this study. We were surprised to find that the bleeding score and APACHE III scores in our study were not helpful in predicting long-term mortality. We cannot explain why the APACHE III score did not turn out to be a good predictor of mortality even though it is an objective score indicating the severity of comorbid conditions. As most of our patients had APACHE III scores less than 90, and only seven patients died in the hospital, we surmise that APACHE III scores were not as helpful as the Charlson score. The Charlson score is easier to compute compared with the APACHE III score, and was a good predictor of mortality. Recently, the Rockall score (5) has been found to be useful in predicting mortality and also re-bleeding rates, but its value in predicting long-term mortality was not evaluated.

Hudson et al (1), in a prospective study of 487 patients with UGI bleeding, found a 50%, 6-year mortality, and in less than 2% of patients GI bleeding was the cause of mortality. Our study is in agreement with these findings.

In a national audit of 4,185 patients with UGI bleeding in the United Kindom, Rockall et al (5) found an overall mortality rate of 14%, and 33% mortality for inpatients who were emergent admissions with UGI bleeding. Our data show much lower in-hospital mortality rates for similar patients.

In a prospective study of 85 patients with cirrhosis who were admitted with GI bleeding, Afessa and Kubilis (7) found that the mortality rate was 21%. In the current study, 5 of 12 patients with variceal bleeding died over the period of 7 years. This mortality rate is lower than the 30% in-hospital and 60% 1-year mortality reported in the literature for variceal bleeders.

Mortality for LGI bleeding is generally found to be lower than that for UGI bleeding. (8,9) Our patients with LGI bleeding had a higher mortality (71%) than the ones with UGI bleeding (51%) or the ones with unknown site of bleeding (50%). This may be related to the small number of patients with LGI bleeding in our study.

Forty-eight (73%) of 66 patients in our study required two or more units of blood by transfusion, and were representative of severe GI bleeders. We were surprised to find the low all-cause long-term mortality rates for patients with severe GI bleeding in the present study. GI bleeding-related mortality was much lower than expected. Among the GI bleeding-related deaths, in most GI bleeding was a contributory cause only, and in only a small fraction was bleeding the primary cause of death.

The strengths of our study included: 1) inclusion of sicker patients (MICU admissions needing blood transfusion), 2) availability of 7-year mortality data, 3) having obtained copies of death certificates to estimate the cause of death, and 4) inclusion of patients with different sites of GI bleeding. Some limitations of our study are as follows: 1) there was a smaller number of patients; 2) the cause of death obtained from death certificates may not be very accurate, although severe GI bleeding leading to death or contributing to the cause of death is unlikely to be wrong; and 3) the mortality rate may have been an underestimate, as some of our patients might have died outside of Montgomery County, OH.

Conclusion

In our study, the long-term (7-year) mortality for patients admitted to the MICU with GI bleeding was lower than that reported in the literature. (1,4,5,10) GI bleeding caused death in only a small percentage of patients, although GI bleeding was a contributory cause of death in many patients.

Life is pleasant. Death is peaceful. It's the transition that's
troublesome.
--Isaac Asimov

Table 1. APACHE III Score

Factors used to compute the score
  1. MICU admission route
  2. Age
  3. Chronic health evaluation (comorbid conditions)
  4. Neurologic (Glasgow Coma Scale Score)
  5. Laboratory tests: arterial blood gases, renal function tests, liver
  function tests
  6. Coefficient for GI bleeding

Table 2. GI Bleeding Score

A. Number of blood transfusions required
      within 24 hours of admission
  More than 4 U in the first 24 h         3
  3-4 U in 24 h                           2
  2 U in 24 h                             1
  0-1 U in 24 h                           0
B. Endoscopy criteria
  GI bleeding
    Spurting                              3
    Oozing or adherent clot               2
    Blood at site                         1
    Clean                                 0

Table 3. Charlson Comorbidity Index

Condition                                                        Score

Myocardial infarct                                                 1
Congestive heart failure                                           1
Peripheral vascular disease                                        1
Cerebrovascular disease                                            1
Dementia                                                           1
Chronic pulmonary disease                                          1
Connective tissue disease                                          1
Ulcer disease                                                      1
Mild liver disease                                                 1
Diabetes mellitus without end-organ damage                         1
Hemiplegia                                                         2
Moderate to severe renal disease (serum creatinine > 2.0 mg/dL)    2
Diabetes mellitus with end-organ damage                            2
Any tumor                                                          2
Leukemia                                                           2
Lymphoma                                                           2
Moderate to severe liver disease                                   3
Metastatic solid tumor                                             6
Acquired immunodeficiency syndrome                                 6

Table 4. APACHE III, Bleeding, Charlson scores and relative risk of
death among GI bleeders admitted to the MICU (Cox proportional hazards
regression) (a)

Variable        Coefficient  Standard error    Z      P     Relative
                                                              risk

APACHE III        0.01421       0.01386       1.49  0.1361    1.01
Bleeding score   -0.15322       0.10553      -1.45  0.1465    0.86
Charlson score    0.35439       0.11885       2.98  0.0029    1.43

(a) MICU, medical intensive care unit.

Acknowledgments

The authors thank APACHE III Medical Systems for providing us coefficients for GI bleeding; Timothy Sorg, MD, and Virginia Wood, MD, for helping us in obtaining institutional review board consent; and Kim Hagler, Diana Ramsey, and Sharon Wilkinson for secretarial support.

Accepted December 2, 2003.

Please see Ronald A. Leo's editorial on page 922 of this issue.

References

1. Hudson N, Faulkner G, Smith MJS, et al. Late mortality in elderly patients surviving acute peptic ulcer bleeding. Gut 1997;37:177-181.

2. Knaus WA, Wagmer DP, Draper EA, et al. The APACHE III prognostic system: risk prediction of hospital mortality for critically ill hospitalized adults. Chest 1991;100:1619-1636.

3. Pompei P, Charlson ME, Douglas RG Jr. Clinical assessments as predictors of one year survival after hospitalization: implications for prognostic stratification. J Clin Epidemiol 1988;41:275-284.

4. Siverstein FE, Gilbert DA, Tedesco FJ, et al. The National ASGE survey on upper gastrointestinal bleeding. Gastrointest Endosc 1981;27:80-93.

5. Rockall TA, Logan RFA, Devlin HB, et al. Incidence of and mortality from acute upper gastrointestinal hemorrhage in the United Kingdom. BMJ 1995;311:222-226.

6. Aggarwal A, Ong JP, Younossi ZM, et al. Predictors of mortality and resource utilization in cirrhotic patients admitted to the medical ICU. Chest 2001;119:1489-1497.

7. Afessa B, Kubilis PS. Upper gastrointestinal bleeding in patients with hepatic cirrhosis: clinical course and mortality prediction. Am J Gastroenterol 2000;95:484-489.

8. Longstreth GF. Epidemiology and outcome of patients hospitalized with acute lower gastrointestinal hemorrhage: a population-based study. Am J Gastroenterol 1997;92:419-424.

9. Zuckerman GR, Prakash C. Acute lower intestinal bleeding: etiology, therapy and outcomes. Gastrointest Endosc 1999;49:228-238.

10. Rockall TA, Logan RFA, Devlin HB, et al. Risk assessment after acute upper gastrointestinal hemorrhage. Gut 1996;38:316-321.

RELATED ARTICLE: Key Points

* Long-term mortality (more than 30 days) of gastrointestinal bleeders is reported only scarcely in the literature.

* Seven-year data for a subset of patients with severe GI bleeding indicated that all-cause and GI bleeding-related long-term mortality were low.

* The Charlson score was helpful in predicting the long-term mortality of gastrointestinal bleeders.

Narasimh Gopalswamy, MD, Vikas Malhotra, MD, Niranjan Reddy, MD, Brij M. Singh, MD, Ronald J. Markert, PHD, Satya Sangal, PHD, and Roy Jordan, BA

From the Department of Medicine, VA Medical Center, and Department of Medicine, Wright State University, Dayton, OH.

The authors have no financial disclosure, as this study was not funded, and none of the authors have any proprietary interest.

Reprint requests to Narasimh Gopalswamy, MD, VA Medical Center, 4100 West Third Street, 111, Dayton, OH 45428. Email: [e-mail]