Determinants of in-hospital mortality and length of stay for acute intestinal gangrene

The American Journal of Surgery 187 (2004) 482– 485

Brief report

Determinants of in-hospital mortality and length of stay for acute intestinal gangrene Karthikeshwar Kasirajan, M.D.a,c,*, Edward J. Mascha, M.S.b, Daithi Heffernan, M.D.a, Jody Sifuentes III, M.S.a a

Department of Surgery, University of New Mexico School of Medicine, Albuquerque, NM, USA b Department of Biostatistics, Cleveland Clinic Foundation, Cleveland, OH, USA c Department of Surgery, Emory University Hospital, Room H-122A, 1364 Clifton Rd., NE, Atlanta, GA 30322, USA Manuscript received June 16, 2003; revised manuscript November 15, 2003

Abstract Background: Intestinal gangrene carries high operative mortality and morbidity rates. This study was undertaken to identify predictors of in-hospital death and length of stay. Methods: Retrospective review of hospital data over a 6-year period identified 107 patients diagnosed with acute bowel gangrene. Results: Among the baseline factors that had a significant univariable association with mortality (51%) were age (P ⫽ 0.04), symptom duration (P ⫽ 0.01), preoperative and postoperative pH and lactic acid (P ⱕ0.02), history of hypertension (P ⫽ 0.001), and renal failure (P ⫽ 0.008). Symptom duration and history of hypertension were independent risk factors for mortality. Longer length of stay was univariably associated with symptom duration (P ⫽ 0.006), systemic acidosis (P ⱕ0.005), vascular etiology (P ⫽ 0.04), amount of resected bowel (P ⫽ 0.001), and need for second-look procedures (P ⬍0.001). Conclusions: The presence of multiple risk factors predictive of a high mortality rate may aid more realistic decision making for physicians, patients, and family members. © 2004 Excerpta Medica, Inc. All rights reserved. Keywords: Acute mesenteric ischemia; Mortality; Outcomes

Acute intestinal ischemia is being diagnosed with increasing frequency, and continues to be associated with high mortality rates of 60% to 80% [1–5]. These have largely remained unchanged since the 1960s. Long-term prognosis is equally dismal with reported 3-year 34% postdischarge survival rates. Early intervention is crucial, with the potential for 100% intestinal viability if symptoms are less than 12 hours, 56% if symptoms are 12 to 24 hours, but only 18% if symptoms are greater than 24 hours in duration before diagnosis [6]. However, very few manuscripts address the outcomes of patients once frank intestinal gangrene has been diagnosed intraoperatively. The aim of this study was to review our institutional experience of patients with gangrenous bowel and to evaluate baseline factors that deter-

* Corresponding author. Tel.: ⫹1-404-727-8407; fax: ⫹1-404-7273316. E-mail address: [email protected]

mine the perioperative mortality rate and length of stay (LOS).

Methods The records of all patients admitted to a single university-based tertiary referral center with a diagnosis of acute intestinal gangrene between 1991 and 2001 were reviewed and systematically analyzed. Only patients with an operative or pathological diagnosis of bowel gangrene were included in the study. Demographic data, baseline risk factors, duration of symptoms, etiology, vital signs on presentation, initial laboratory values, initial diagnoses, and diagnostic studies were recorded. Operative interventions, findings, length of bowel resected, and pathology reports were also noted. Etiologies were also assessed based on preoperative, intraoperative, and postoperative findings. Primary outcome events evaluated were intensive care unit (ICU)

0002-9610/04/$ – see front matter © 2004 Excerpta Medica, Inc. All rights reserved. doi:10.1016/j.amjsurg.2003.12.049

K. Kasirajan et al. / The American Journal of Surgery 187 (2004) 482– 485

LOS, total LOS, and in-hospital mortality. Relationship between the baseline factors and the outcome parameters were studied. The study was approved by the Institutional Review Board. Statistical analysis Association between baseline and postoperative factors and either ICU LOS or total LOS was assessed using Spearman’s correlation coefficient for continuous or ordinal predictors and using the Wilcoxon rank-sum test for nominal predictors. Nonparametric tests were used because the LOS variables were not normally distributed. Multivariable models were attempted using linear regression on the log-transformed of LOS⫹1 variables. Univariable and multivariable logistic regression analysis was used to assess for associations between with baseline and postoperative factors and mortality. Odds ratios (OR) and 95% confidence intervals (CI) are reported. In multivariable analysis we only considered variables with at least 80% of the data not missing. The significance level was 0.05 for each hypothesis. Results During the study period, 107 patients (55 female) with acute intestinal ischemia were identified. The median age was 67 years (range 42 to 90). Etiology for bowel gangrene was embolic occlusion in 20, in situ thrombosis in 55, nonocclusive mesenteric ischemia in 6, mesenteric venous thrombosis in 6, volvulus in 3, hernia in 8, and unknown etiology in 9 other patients. Total intensive care unit length of stay The median LOS in the ICU was 4 days (quartiles 1 to 9). Table 1 summarizes the risk factors and association with length of ICU stay. Only 9 factors were predictive of increased length of ICU stay: time from symptom onset to surgery, preoperative and postoperative serum albumin, postoperative pH, amount of resected bowel, male sex, vascular etiology, and need for second-look operation. Total length of hospital stay The median LOS was 10 days (quartiles 3 to 18). Factors predictive of prolonged length of hospital stay were time from symptom onset to surgery, preoperative and postoperative pH, postoperative white blood cell count, postoperative lactic acid, vascular etiology, amount of resected bowel, and need for second-look operation (Table 1). No multivariable models resulted for either LOS variable. Perioperative mortality Overall, 55 patients (51%) died during the perioperative period, the majority secondary to multisystem organ failure.

483

In univariable analysis the risk factors significantly associated with mortality included older age, time from symptom onset to surgery, preoperative serum albumin, preoperative and postoperative pH and lactic acid levels, postoperative base deficit, history of hyperlipdemia, and renal failure (Table 2). Time from symptom onset to surgery (OR 1.2; 95% CI 1.03 to 1.4) and history of hypertension (OR 4.6; 95% CI 1.8 to 12.1) were independent predictors of mortality in a multivariable model at the 0.05 significance level. This model has an estimated accuracy (c-statistic) of 73% on the data with which it was formed. Further, adding history of renal (P ⫽ 0.15) and pulmonary (P ⫽ 0.08) morbidity gave a model with 78% accuracy in predicting early mortality. For example, a patient presenting on day 5 and having all three risk factors has an estimated 87% chance (95% CI 67% to 96%) of perioperative death, whereas a patient presenting on day 0 with all three risk factors has an estimated 73% chance (95% CI 47% to 89%). A patient presenting on day 5 of symptoms onset with none of the three risk factors has an estimated 32% probability (95% CI 15% to 54%) of mortality, whereas a patient entering on day 0 with none of the risk factors has only an estimated 15% probability (95% CI 7% to 32%). A long delay before presentation—say, 20 days— yields a high estimated probability of early death even in the absence of the three risk factors (88%; 95% CI 31% to 99%).

Comments Although a significant number of the patients in our report had an underlying vascular etiology, bypass at time of bowel resection was performed in only 14 patients. However, vascular etiology was not by itself predictive of a higher mortality. The mortality rate was 54% when associated with a vascular etiology, compared with 37% when bowel gangrene was secondary to a nonvascular event (P ⫽ 0.18). Mortality associated with specific etiology is given in Table 3. The duration of the symptoms at presentation was of significance, so that the time to definitive treatment of acute ischemia appeared to play a significant role in patient outcomes and morbidity and mortality. In our patient population there was a discrepancy in the time from presentation to operative intervention with respect to survival, with median (quartiles) of 1 (0, 2) and 1 (0, 5) and range of 0 to 7 and 0 to 62 for nonsurvivors and survivors, respectively (P ⫽ 0.013). A proportion of these patients were transferred from another institution. That may also reflect the difficulty for some of our patients to reach medical attention. With the need for expedient intervention, the role of diagnostic investigations would need to be done in a timely fashion. Laboratory investigations have been notoriously unhelpful in predicting acute ischemia. In our study population the white blood cell count was elevated both before and after

484

K. Kasirajan et al. / The American Journal of Surgery 187 (2004) 482– 485

Table 1 Relationship of baseline and postoperative factors with intensive care unit (ICU) and total length of stay (LOS) n

ICU LOS

Total LOS

Correlation* Continuous variables Age Time to surgery Preoperative laboratory studies WBC Albumin Lactic acid pH CO2 Base deficit Postoperative laboratory studies WBC Albumin Lactic acid pH CO2 Base deficit Extent of gangrene Resected (cm) Categorical variables Male sex

Vascular etiology Smoking Hypertension Hyperlipidemia Cardiac Renal COPD Second-look surgery

Level F M N Y N Y N Y N Y N Y N Y N Y N Y

P value

Correlation*

P value

105 106

0.11 0.29

0.27 0.003

⫺0.03 0.27

0.76 0.006

107 77 33 77 76 71

0.12 ⫺0.27 0.06 0.22 0.02 ⫺0.15

0.21 0.002 0.73 0.06 0.87 0.21

0.12 ⫺0.19 ⫺0.19 0.22 0.005 ⫺0.18

0.24 0.09 0.28 0.05 0.97 0.13

88 43 17 71 71 70 100 95 n 55 52 19 87 33 71 37 63 44 18 49 57 72 33 67 39 72 34

0.07 ⫺0.32 ⫺0.30 0.27 0.07 ⫺0.19 ⫺0.05 0.23 Median (quartiles) 3 (1, 8) 4 (2, 10) 2 (1, 3) 4 (2, 10) 4 (1, 9) 3 (1, 9) 4 (1, 10) 3 (1, 9) 4 (2, 10) 3 (1, 10) 3 (2, 9) 4 (1, 10) 3 (1, 8) 4 (2, 10) 4 (2, 10) 3 (1, 9) 2 (1, 6) 10 (4, 13)

0.52 0.04 0.25 0.02 0.56 0.11 0.66 0.02†

0.26 ⫺0.26 ⫺0.77 0.33 0.20 ⫺0.01 0.004 0.33 Median (quartiles) 9 (4, 16) 13 (3, 20) 6 (2, 9) 12 (4, 19) 10 (5, 16) 9 (3, 18) 13 (7, 18) 9 (3, 18) 11 (5, 17) 10 (5, 18) 10 (4, 17) 10 (4, 19) 10 (3, 19) 10 (4, 17) 11 (6, 18) 7 (2, 18) 8 (2, 15) 17 (10, 28)

0.03 0.002 0.84 0.97 0.65 0.63 0.23 0.26 ⬍0.001

0.02 0.09 ⬍0.001 0.005 0.09 0.84 0.97 0.001†

0.22 0.039 0.74 0.24 0.8 0.82 0.95 0.12 ⬍0.001

* Spearman correlation coefficient. † Wilxocon rank-sum test. P ⬍0.05 was considered significant. WBC ⫽ white blood cell count; COPD ⫽ chronic obstructive pulmonary disease; F ⫽ female; M ⫽ male; Y ⫽ yes; N ⫽ no.

surgery and was not statistically related to the outcome parameters evaluated. Abnormal laboratory studies are often indicators of advanced necrosis. Lactate, although often cited as a specific marker of bowel gangrene, may often be a late finding. In our patient population, it reflected late disease with a high mortality rate. Persistent acidosis after bowel resection (low pH and high lactate) was a powerful predictor of death (P ⬍0.02), although these data were not available for many of the patients and were therefore not considered in multivariable modeling. We were able to develop a model for predicting mortality with about 78% accuracy using days from symptoms to presentation and history of hypertension, pulmonary, and renal morbidity. Age was not independently associated with

mortality once the above factors were accounted for. A long symptom history (more than 15 days) and history of hypertension and pulmonary and renal morbidity would have an almost certain fatal outcome (estimated probability 98% [CI 81% to 100%]). Gangrenous bowel continuous to be associated with a high mortality rate and a dismal long-term survival. For patients who left the hospital, the median survival time was only 2 years, and 21% of survivors continued to have symptoms of chronic mesenteric ischemia. In conclusion, early detection of bowel ischemia before progression to gangrene may be the only chance for improved early and long-term survival. An aggressive diagnostic approach (mesenteric angiograms, magnetic reso-

K. Kasirajan et al. / The American Journal of Surgery 187 (2004) 482– 485

485

Table 2 Association between baseline variables and perioperative mortality

Continuous variables Age (years) Time to surgery (days)* Preoperative laboratory studies WBC Albumin Lactic acid pH CO2 Base deficit Postoperative laboratory studies WBC Albumin Lactic acid pH CO2 Base deficit Extent of gangrene Resected (cm) Categorical variables Male Vascular etiology Smoking Hypertension Hyperlipidemia Cardiac Renal Second-look surgery

Alive (mean ⫾ SD)

Dead (mean ⫾ SD)

Odds ratio

95% CI

P value

59.6 ⫾ 18 1 (0, 2)

66.8 ⫾ 17 1 (0, 5)

1.02 1.20

1.00–1.05 1.03–1.39

0.04 0.01

15.3 ⫾ 8.8 3.2 ⫾ 0.9 2.5 ⫾ 2.5 7.4 ⫾ 0.1 35.4 ⫾ 8.4 ⫺1.3 ⫾ 4.5

13.9 ⫾ 7.8 2.8 ⫾ 0.9 7.3 ⫾ 5.1 7.3 ⫾ 0.1 36.1 ⫾ 12.5 ⫺1.1 ⫾ 8.2

0.98 0.60 1.78 0.00 1.01 1.00

0.93–1.03 0.35–1.02 1.14–2.78 0.00–0.39 0.96–1.05 0.94–1.08

0.36 0.05 0.01 0.02 0.78 0.90

12.2 ⫾ 6.5 2.3 ⫾ 0.7 1.4 ⫾ 0.4 7.4 ⫾ 0.1 39.9 ⫾ 8.5 0.6 ⫾ 3.3 1.8 ⫾ 1.2 76.8 ⫾ 104.7 No† 45% 37% 42% 32% 48% 45% 42% 50%

11.7 ⫾ 6.2 2.1 ⫾ 1 8.7 ⫾ 8.4 7.3 ⫾ 0.1 35.9 ⫾ 12.7 ⫺1.5 ⫾ 9.4 2.1 ⫾ 1.1 63.9 ⫾ 116.5 Yes‡ 58% 54% 54% 60% 44% 56% 70% 56%

0.99 0.67 3.59 0.00 0.96 0.96 1.32 1.00

0.92–1.06 0.30–1.50 0.74–17.43 0.00–0.24 0.92–1.01 0.89–1.03 0.92–1.90 1.0–1.0

0.74 0.32 0.02 0.01 0.12 0.20 0.13 0.57

1.64 0.87 1.56 3.17 0.88 1.57 3.22 1.27

0.76–3.51 0.71–1.07 0.68–3.60 1.35–7.43 0.29–2.64 0.73–3.39 1.34–7.75 0.56–2.87

0.21 0.18 0.29 0.007 0.81 0.25 0.008 0.57

* Median (quartiles). † Mortality if risk factor is absent. ‡ Mortality if risk factor is present. CI ⫽ confidence interval; WBC ⫽ white blood cell count.

nance angiograms, Duplex evaluation of vessels, or early diagnostic laprotomy) in elderly patients presenting with abdominal pain out of proportion to their physical findings is recommended, before progression to bowel gangrene [7]. Further studies are needed to evaluate the impact of an early and aggressive diagnostic and therapeutic approach to paTable 3 Relationship between etiology and perioperative mortality Etiology

Number

Mortality

Percent mortality

Embolism In situ thrombosis Nonocclusive mesenteric ischemia Venous thrombosis Volvulus Hernia Unknown

20 55 6 6 3 8 9

12 31 2 2 1 1 3

60% 56% 33% 33% 33% 13% 33%

tients presenting with symptoms suggestive of bowel ischemia. References [1] Klempnauer J, Grothues F, Bektas H, et al. Long-term results after surgery for acute mesenteric ischemia. Surgery 1997;121:239 – 43. [2] Ottinger LW, Austen WG. A study of 136 patients with mesenteric infarction. Surg Gynecol Obstet 1967;124:251– 61. [3] Hertzer NR, Beven EG, Humphries AW. Acute intestinal ischemia. Am Surg 1978;44:744 –9. [4] Sachs SM, Morton JH, Schwartz SI. Acute mesenteric ischemia. Surgery 1982;92:646 –53. [5] Slater H, Elliott PW. Primary mesenteric infarction. Am J Surg 1972; 123:309 –11. [6] Wadman M, Syk L, Elmstahl S. Survival after operation for ischemic bowel disease. Eur J Surg 2000;166:872–7. [7] Kasirajan K, O’Hara PJ, Gray BH, et al. Chronic mesenteric ischemia: open surgery percutaneous angioplasty and stenting. J Vasc Surg 2001;33:63–71.