Bacterial infections are diagnosed in 30% to 50% of all hospital admissions for cirrhotic decompensation and are directly responsible for up to 50% of deaths in cirrhosis [[1]x[1]Wong, F., Bernardi, M., Balk, R., Christman, B., Moreau, R., Garcia-Tsao, G. et al. Sepsis in cirrhosis: report on the 7th meeting of the International Ascites Club. Gut. 2005; 54: 718–725

CrossRef | PubMed | Scopus (200)See all References, [2]x[2]Navasa, M., Fernández, J., and Rodés, J. Bacterial infections in liver cirrhosis. Ital J Gastroenterol Hepatol. 1999; 31: 616–625

PubMedSee all References]. These patients are twice as likely to die of sepsis than individuals without cirrhosis [3x[3]Foreman, M.G., Mannino, D.M., and Moss, M. Cirrhosis as a risk factor for sepsis and death: analysis of the National Hospital Discharge Survey. Chest. 2003; 124: 1016–1020

CrossRef | PubMed | Scopus (132)See all References][3]. Furthermore, bacterial infections carry significant morbidity, precipitating cirrhosis decompensation (hepatic encephalopathy, ascites, variceal bleeding, and/or jaundice) and acute kidney injury (AKI) development [4x[4]D'Amico, G., Garcia-Tsao, G., and Pagliaro, L. Natural history and prognostic indicators of survival in cirrhosis: a systematic review of 118 studies. J Hepatol. 2006; 44: 217–231

Abstract | Full Text | Full Text PDF | PubMed | Scopus (698)See all References][4]. In fact, renal failure occurs in 33% of patients with cirrhosis who have spontaneous bacterial peritonitis (SBP) and in 27% of those with sepsis unrelated to SBP, with a mortality rate of 41% to 66% [[5]x[5]Sort, P., Navasa, M., Arroyo, V., Aldeguer, X., Planas, R., Ruiz-del-Arbol, L. et al. Effect of intravenous albumin on renal impairment and mortality in patients with cirrhosis and spontaneous bacterial peritonitis. N Engl J Med. 1999; 341: 403–409

CrossRef | PubMed | Scopus (899)See all References, [6]x[6]Terra, C., Guevara, M., Torre, A., Gilabert, R., Fernández, J., Martín-Llahí, M. et al. Renal failure in patients with cirrhosis and sepsis unrelated to spontaneous bacterial peritonitis: value of MELD score. Gastroenterology. 2005; 129: 1944–1953

Abstract | Full Text | Full Text PDF | PubMed | Scopus (152)See all References]. These severe complications frequently result in admission to the emergency department (ED) [[2]x[2]Navasa, M., Fernández, J., and Rodés, J. Bacterial infections in liver cirrhosis. Ital J Gastroenterol Hepatol. 1999; 31: 616–625

PubMedSee all References, [7]x[7]Loo, N.M., Souza, F.F., and Garcia-Tsao, G. Non-hemorrhagic acute complications associated with cirrhosis and portal hypertension. Best Pract Res Clin Gastroenterol. 2013; 27: 665–678

Abstract | Full Text | Full Text PDF | PubMed | Scopus (3)See all References].

Patients with cirrhosis require frequent visits to the ED. In a large population study performed in England, 47.3% of patients had a first diagnosis of cirrhosis during an emergency admission. The remaining 52.7% were diagnosed in an outpatient setting, but 37.8% of them had a subsequent emergency hospital admission over a 5-year period [8x[8]Ratib, S., Fleming, K.M., Crooks, C.J., Aithal, G.P., and West, J. 1 and 5 year survival estimates for people with cirrhosis of the liver in England, 1998-2009: a large population study. J Hepatol. 2014; 60: 282–289

Abstract | Full Text | Full Text PDF | PubMed | Scopus (16)See all References][8].

Early recognition and treatment of severe complications are generally related to better clinical outcomes. In patients with SBP, the most common bacterial infection in cirrhosis, treatment with suitable antibiotics and large volume albumin infusion in the first 6 hours from diagnosis have reduced mortality rates from 80% to 20% in the last 3 decades, in addition to reducing renal failure development in almost 70% [[5]x[5]Sort, P., Navasa, M., Arroyo, V., Aldeguer, X., Planas, R., Ruiz-del-Arbol, L. et al. Effect of intravenous albumin on renal impairment and mortality in patients with cirrhosis and spontaneous bacterial peritonitis. N Engl J Med. 1999; 341: 403–409

CrossRef | PubMed | Scopus (899)See all References, [9]x[9]Garcia-Tsao, G. Bacterial infections in cirrhosis: treatment and prophylaxis. J Hepatol. 2005; 42: S85–S92

Abstract | Full Text | Full Text PDF | PubMed | Scopus (67)See all References]. However, the benefit of such treatment may be restricted to high-risk patients (serum creatinine >1 mg/dL and/or serum bilirubin >4 mg/dL), who should be recognized as early as possible [[10]x[10]Terg, R., Gadano, A., Cartier, M., Casciato, P., Lucero, R., Muñoz, A. et al. Serum creatinine and bilirubin predict renal failure and mortality in patients with spontaneous bacterial peritonitis: a retrospective study. Liver Int. 2009; 29: 415–419

CrossRef | PubMed | Scopus (67)See all References, [11]x[11]Poca, M., Concepción, M., Casas, M., Alvarez-Urturi, C., Gordillo, J., Hernández-Gea, V. et al. Role of albumin treatment in patients with spontaneous bacterial peritonitis. Clin Gastroenterol Hepatol. 2012; 10: 309–315

Abstract | Full Text | Full Text PDF | PubMed | Scopus (21)See all References, [12]x[12]Sigal, S.H., Stanca, C.M., Fernandez, J., Arroyo, V., and Navasa, M. Restricted use of albumin for spontaneous bacterial peritonitis. Gut. 2007; 56: 597–599

CrossRef | PubMed | Scopus (67)See all References].

Despite the great importance of identifying high-risk patients among cirrhotic decompensated individuals, there are no current criteria developed or validated in the emergency setting, although this is the place where most patients with decompensated cirrhosis will be evaluated and would benefit from early detection and treatment of bacterial infections and other severe complications.

In this article, we evaluated patients with decompensated cirrhosis who visited the ED and were admitted to the hospital. Our aim was to evaluate the early predictors, on the arrival to the ED, of the presence of bacterial infection and inhospital mortality.

This was a retrospective single-center study performed in the ED of a Brazilian university hospital. The Hospital das Clínicas, which belongs to the University of São Paulo's Medical School (HCFMUSP), is a tertiary university hospital with more than 180000 annual ED visits, of which 14000 result in hospital admission. In this hospital, the hepatology service includes a 24-hour endoscopy unit, interventionist radiology, a hepatology intensive care unit, and a liver transplantation service. The study was approved by the ethical review board of HCFMUSP (CAPPesq).

From January 1, 2009, to June 30, 2011, consecutive patients seen in the ED of HCFMUSP with decompensated cirrhosis were identified in the hospital database. Medical records of patients with ED diagnosis of cirrhosis (International Statistical Classification of Diseases, 10th Revision [ICD-10], code K74) or correlated conditions (viral hepatitis [ICD-10 code B18], alcoholic liver disease [ICD-10 code K70], autoimmune hepatitis [ICD-10 code K75.4], hemocromatosis [ICD-10 code E83.1], Wilson disease [ICD-10 code E83.0], other liver diseases [ICD-10 code K76], ascites [ICD-10 code R18], esophageal varices [ICD-10 code I85], and peritonitis [ICD-10 code K65]) were screened for inclusion by 2 independent researchers. For patients with more than 1 hospital admission during the study period, only data from the first admission were considered.

Inclusion criteria were (a) patients aged 18 years or older; (b) cirrhosis diagnosis by liver biopsy or a combination of clinical, laboratory, radiological, and endoscopic data; and (c) admission to the ED related to cirrhosis decompensation (hepatic encephalopathy, ascites, variceal bleeding, and/or jaundice).

Exclusion criteria were (a) patients with terminal illness (metastatic neoplasia or severe cardiopulmonary or neurologic disease); (b) hospital discharge in less than 24 hours after admission; and (c) patients with previous solid organ transplantation.

The medical record review was performed by 3 fifth year medical students and 2 members of the medical staff (1 emergency physician and 1 hepatologist). Medical students were trained on the first 20 charts and could then contact the staff at any time to clarify doubts.

Data from ED admission were inserted into standardized electronic spreadsheets (software Excel, Microsoft Office 2007), which included age, sex, cirrhosis etiology, cirrhosis complications (ascites, encephalopathy, variceal bleeding, hepatorenal syndrome, and spontaneous bacterial peritonitis), comorbidities, clinical variables (heart rate, blood pressure, temperature, and Glasgow Coma Scale), laboratory results (C-reactive protein [CRP], leukocytes, neutrophils, lymphocytes, neutrophil/lymphocyte relationship, creatinine, urea, sodium, international normalized ratio [INR], bilirubin, and albumin), culture results, Child-Pugh classification, Model for End-Stage Liver Disease (MELD) score, MELD-Na score, site of infection, and infection classification according to International Sepsis Definitions Conference (ISDC) 2001 (absent, possible, probable, or definite and sepsis, severe sepsis, or septic shock) [13x[13]Levy, M.M., Fink, M.P., Marshall, J.C., Abraham, E., Angus, D., Cook, D. et al. 2001 SCCM/ESICM/ACCP/ATS/SIS International Sepsis Definitions Conference. Crit Care Med. 2003; 31: 1250–1256

CrossRef | PubMed | Scopus (2552)See all References][13]. Standardized electronic spreadsheets also included all creatinine values measured during the first 48 hours from hospital admission, length of hospital stay, and cause of death in nonsurvivors.

The presence of AKI was defined according to Acute Kidney Injury Network criteria [14x[14]Mehta, R.L., Kellum, J.A., Shah, S.V., Molitoris, B.A., Ronco, C., Warnock, D.G. et al. Acute Kidney Injury Network: report of an initiative to improve outcomes in acute kidney injury. Crit Care. 2007; 11: R31

CrossRef | PubMed | Scopus (2019)See all References][14]. For AKI determination, we considered serum creatinine measures during the first 48 hours from hospital admission. Acute kidney injury was defined as an absolute increase in serum creatinine of more than or equal to 0.3 mg/dL or a percentage increase in serum creatinine of more than or equal to 50% (1.5-fold from baseline). In this study, data on urine output were not available [15x[15]de Carvalho, J.R., Villela-Nogueira, C.A., Luiz, R.R., Guzzo, P.L., da Silva Rosa, J.M., Rocha, E. et al. Acute kidney injury network criteria as a predictor of hospital mortality in cirrhotic patients with ascites. J Clin Gastroenterol. 2012; 46: e21–e26

CrossRef | PubMed | Scopus (27)See all References][15].

All collected data were assessed by another investigator (AQF) for the identification of disagreements and obvious problems.

Emergency physicians managed patients with cirrhosis at their discretion. At HCFMUSP, most physicians follow American Association for Study of Liver Disease or European Association for Study of the Liver guidelines for cirrhosis complications management. Physicians were not aware of this study.

Patients' infection status was defined according to the ISDC 2001 [13x[13]Levy, M.M., Fink, M.P., Marshall, J.C., Abraham, E., Angus, D., Cook, D. et al. 2001 SCCM/ESICM/ACCP/ATS/SIS International Sepsis Definitions Conference. Crit Care Med. 2003; 31: 1250–1256

CrossRef | PubMed | Scopus (2552)See all References][13]. Patients classified with definite, probable, or possible infection were considered to be patients with bacterial infection.

The severity of their infection was then classified in sepsis, severe sepsis, or septic shock, according to standard definitions [13x[13]Levy, M.M., Fink, M.P., Marshall, J.C., Abraham, E., Angus, D., Cook, D. et al. 2001 SCCM/ESICM/ACCP/ATS/SIS International Sepsis Definitions Conference. Crit Care Med. 2003; 31: 1250–1256

CrossRef | PubMed | Scopus (2552)See all References][13].

Mortality was determined during hospital stay. Patients who were discharged from hospital were considered survivors. There was no follow-up after hospital discharge.

Statistical analysis was performed with the software R 3.0.1 (The R Foundation for Statistical Computing, 2013). Discrete variables were reported as percentages. Continuous variables were presented as medians with interquartile ranges.

Using area under the receiver operating characteristic (ROC) curve (AUROC) and maximum Youden Index, the continuous variables' effectiveness and best cutoff values to identify patients with bacterial infection and to identify survivors were determined. Continuous variables were categorized according to these cutoff values.

Simple binary logistic regression models were built, and variables with P ≤ .1 were inserted in multiple binary logistic regression models to determine variables independently associated with bacterial infection and to determine variables independently associated with mortality according to the Akaike Information Criterion.

Scores using variables associated with bacterial infection and mortality were built from the resized regression coefficients giving a range of scores from 0 to 100 points. Receiver operating characteristic curve analysis was performed to evaluate the scores' discrimination effectiveness using AUROC and maximum Youden Index. Sensitivity, specificity, positive predictive value (PPV), negative predictive value (NPV), positive likelihood ratio (PLR), and negative likelihood ratio (NLR) were calculated at maximum Youden Index cutoffs.

Bootstrap analysis was used for scores validation (1000 random bootstrap samples of patients drawn with replacement from the original sample) to calculate the optimism on the effectiveness measures.

Sampling unit was disregarded when there was at least 1 variable involved in the analysis with missing data.

During the study period, there were 277 ED admissions from 159 patients with cirrhosis. Ten patients were excluded: 7 for hospital discharge in less than 24 hours after admission, 2 with terminal illness (metastatic cancer), and 1 with previous liver transplantation. The remaining 149 patients were considered in our analysis. Baseline characteristics of eligible patients are shown in Table 1Table 1.

The cirrhosis decompensations that caused ED admission were ascites in 113 (75.8%) patients, hepatic encephalopathy in 88 (59.1%), jaundice in 43 (29.0%), and variceal bleeding in 12 (10.1%).

Acute kidney injury according to Acute Kidney Injury Network criteria occurred in 61 patients (40.9%) during the first 48 hours after admission. Median length of hospital stay was 5 days (2-9 days).

Seventy-two patients (48.3%) were diagnosed with bacterial infection: 27 (37.5%) with spontaneous bacterial peritonitis, 20 (27.8%) with urinary tract infection, 11 (15.3%) with pneumonia, 9 (12.5%) with skin and soft tissue infection, and 5 (6.9%) with other sites of infection. The classification of severity of infection was sepsis in 9 patients (12.5%), severe sepsis in 14 (19.4%), and septic shock in 4 (5.6%). The other 45 patients (62.5%) did not meet criteria for sepsis.

Ninety-nine cultures were performed, of which 30 (30.3%) were positive. Isolated microorganisms were Escherichia coli in 12 (40.0%) cultures, Klebsiella pneumoniae in 10 (33.3%), Staphylococcus aureus in 3 (10.0%), other gram negatives in 3 (10.0%), and other gram positives in 2 (6.7%).

During hospital stay, 36 patients (24.2%) died. Cause of death was sepsis in 23 patients (63.9%), variceal bleeding in 5 (13.9%), AKI in 3 (8.3%), and other causes in 5 (13.9%). Among patients with AKI, 31.1% died, compared to 17.6% among those without. Mortality rate in patients with bacterial infection was 31.9% compared to 18.1% in those without.

In the simple binary logistic regression model, variables associated with bacterial infection that were subsequently inserted into the multiple binary logistic regression model are shown in Table 2Table 2.

After the multiple binary logistic regression, variables independently associated with bacterial infection were lymphocytes less than or equal to 900/mm³ (odds ratio [OR], 3.85 [95% confidence interval [CI], 1.47-10]; P = .006) and CRP greater than 59.4 mg/L (OR, 5.05 [95% CI, 1.93-13.2]; P = .001). In the ROC curve analysis, lymphocytes had an AUROC of 0.60 (0.516-0.701); and CRP, an AUROC of 0.72 (0.622-0.817).

In the simple binary logistic regression model, variables associated with mortality that were subsequently inserted into the multiple binary logistic regression model are shown in Table 3Table 3.

After the multiple binary logistic regression, variables independently associated with mortality were creatinine greater than 1.5 mg/dL (OR, 4.35 [95% CI, 1.87-10.1]; P = .001) and INR greater than 1.65 (OR, 3.71 [95% CI, 1.6-8.61]; P = .002). In ROC curve analysis, creatinine had an AUROC of 0.756 (0.657-0.855) and INR an AUROC of 0.701 (0.598-0.803).

The score designed to predict bacterial infection (Bacterial Infection in Cirrhosis Score [BIC Score]) can be calculated using the following formula:

Receiver operating characteristic curve analysis showed an AUROC of 0.82 (0.734-0.905), with the best cutoff value of 33.55. Patients with a BIC Score above this value had an 80% chance of bacterial infection, compared to a 28% chance in patients with a BIC Score less than 33.55. Sensitivity, specificity, PPV, NPV, PLR, and NLR are shown in Table 4Table 4 as well as sensitivity, specificity, PPV, and NPV calculated after correction for optimism in the bootstrap validation setting.

The score designed to predict mortality (Mortality in Cirrhosis Emergency Department Score [MED Score]) can be calculated using the following formula:

Receiver operating characteristic curve analysis showed an AUROC of 0.801 (0.714-0.889), with the best cutoff value of 27.37. Patients with a MED Score above this value had a mortality rate of 60%, compared to a mortality rate of 11% in patients with a MED Score less than 27.37. Sensitivity, specificity, PPV, NPV, PLR, and NLR are shown in Table 5Table 5 as well as sensitivity, specificity, PPV, and NPV calculated after correction for optimism in the bootstrap validation setting.

The AUROCs for MELD and MELD Na scores were 0.766 and 0.779, respectively. The FigureFigure shows the ROC curve of MED Score.