Bacterial Species-Specific Hospital Mortality Rate for Intra-Abdominal Infections

Patients and Methods

We conducted a retrospective review of IAI at our institution, a tertiary-care facility and regional Level I trauma center, over an eight-year period (June 1999–June 2007). The MetroHealth Medical Center Institutional Review Board approved the study.

The electronic medical records and hospital database were queried for patients' assigned International Classification of Disease (ICD)-9 codes for IAI, including intra-abdominal abscess and diffuse peritonitis. The original study population was identified from the administrative database utilizing ICD-9 coding to identify IAIs. Other ICD-9 diagnoses and other data were obtained utilizing the electronic database. All data included in the dataset were confirmed by chart review. Multiple organ dysfunction syndrome scores (MODS) were evaluated for this specific study. Infectious complications were identified prospectively by both surgical intensivists and infection control practitioners.

All patients met one of the following inclusion criteria for secondary peritonitis: (1) New IAI after any non-gynecologic intra-abdominal operation; (2) perforated stomach or duodenum without intervention for >24 h; (3) perforated small or large bowel; (4) histologically or intra-operatively confirmed perforated appendicitis; (5) perforated gallbladder; or (6) other cause of secondary bacterial peritonitis. Traumatic gastrointestinal tract perforations treated within 12 h were excluded. An IAI was defined as an intra-abdominal infection episode (IE) and was classified as appendiceal or non-appendiceal in origin to permit subgroup analysis. An IE involved complete resolution of the IAI.

The following information was collected for the study: patient age, gender, source of infection, type of initial intervention (percutaneous drainage vs. operation), Charlson Comorbidity Index score [3], and MODS at presentation and on post-operative day (POD) 7 [6]. Complications were categorized as infectious or non-infectious. Infectious complications included surgical site infection (SSI) urinary tract infection, pneumonia, CRBSI, and post-operative IAI, defined according to the National Healthcare Safety Network (NHSN) surveillance guidelines [9]. Non-infectious complications were wound dehiscence, a major cardiac event (dysrhythmia or myocardial infarction), and deep venous thrombosis or pulmonary embolism. Treatment failure was defined as the need for re-intervention (either reoperation or percutaneous drainage) after treatment of the initial infection. The outcomes were death, treatment failure, cardiac event, prolonged ventilator support (>72 h), pneumonia, CRBSI, urinary tract infection, SSI, and length of intensive care unit (ICU) stay.

Source control was attempted at the time of initial intervention. Intra-operative cultures were obtained at the discretion of the surgeon. Samples from all infections drained percutaneously were sent for culture. Specimens were cultured for aerobic and anaerobic organisms using the Becton, Dickinson and Company (BD) BBL™ Port-A-Cul™ product (Franklin Lakes, NJ). Fungal cultures were performed only when specifically requested; however, typically, such cultures are requested and so were available in most cases.

Antimicrobial treatment was not standardized but generally included initial empiric broad-spectrum antimicrobial agents targeted to the expected microbial flora and later tailored depending on the culture results. Treatment length was not evaluated in this study. However, a typical antimicrobial course lasted seven to ten days or until resolution of signs and symptoms.

The statistical analysis was performed using SPSS (SPSS Inc., Chicago, IL). Continuous and discrete variables were compared using the Student t-test and Fisher exact test where appropriate. Continuous data were reported as the mean±standard error of the mean. A p value<0.05 was considered significant. Backward stepwise logistic regression analysis was performed to assess for independent predictors of death, treatment failure, and complications. All variables with p<0.10 on bivariable analysis were included in the analysis, and a receiver operator characteristic was calculated. Using pathogen species as well as other clinical data, analyses were performed to evaluate the association independently with death, treatment failure, and complications. The microbial variables were limited to species that grew in 10 or more intra-abdominal cultures.

Results

Discussion

Intra-abdominal infections cause substantial morbidity and many deaths. In previous work at our institution, we demonstrated that age, major cardiac events, and CRBSI were independent predictors of death in patients with IAI [8]. Our current investigation demonstrated that the presence of specific organisms also is associated with a higher mortality rate. Appendiceal IAIs were excluded from this analysis because of the low morbidity and mortality rates, both in our study and in the literature [10]. We had no deaths in this series of patients. Our results demonstrated that Clostridium cultured from IAIs were an independent predictor of death, whereas Streptococcus was significantly predictive of survival. Additionally, cultures that grew Bacteroides species other than B. fragilis, Enterococcus (E. faecalis or E. faecium), or C. perfringens demonstrated a significant association with a higher mortality rate.

We are not aware of other studies investigating the relation between death and a specific infecting organism. However, there are publications addressing the diagnosis, management, and treatment of IAIs and intra-abdominal sepsis. Swenson et al. investigated patient characteristics associated with the growth of resistant pathogens from an intra-abdominal source. This study noted relations between the anatomic source of infection, patient characteristics, growth of resistant pathogens, and death [11]. The authors concluded that clinical characteristics may help identify patients potentially infected with resistant pathogens, specifically Candida and enterococci.

Montravers et al. investigated Candida peritonitis and noted that the infection was an independent predictor of death [12]. Fungal cultures were not performed routinely, and Candida was not associated with death. However, fungus seen on Gram stain was associated with treatment failure. Furthermore, Szeto et al. found a higher mortality rate from peritonitis in patients undergoing peritoneal dialysis who had cultures positive for Pseudomonas or Enterococcus [13]. Our study demonstrated a significant association of Pseudomonas with death, but logistic regression analysis failed to confirm Pseudomonas as an independent predictor of death. This fact likely is attributable to a Type II error and lack of power in our study.

There were other limitations to this study. It was retrospective and based on ICD-9 code assignments in the administrative database. Patients thus may have been excluded on the basis of coding errors, which may have contributed to the small sample, leading to a Type II error. Also, the low overall mortality rate limited our ability to identify factors associated with death. Additionally, specific populations such as patients with solid organ allografts were not evaluated. Our findings therefore should not be applied to a transplant population.

Source control and appropriate antimicrobial management are fundamental to the successful management of IAIs. However, this study did not examine specifically the adequacy of source control or the appropriateness of the antimicrobial treatment. Additionally, infectious complications were not differentiated by hospital-acquired or community-acquired status, yet these factors may contribute to death or treatment failure. We also did not use an Acute Physiology and Chronic Health Evaluation II score, which is associated with outcomes in many studies of infections. Lastly, the initial source of the post-operative infections was not identified.

Our previous work demonstrated that whether an infection was community-acquired or nosocomial had less impact on the patient mortality rate than did intrinsic patient characteristics [8]. The patient's physiologic and co-morbidity scores support the relation between co-morbidities, greater physiologic derangements, and death. Now we can further address the impact of the specific infecting organism on the likelihood of death. In our work, enterococcal species (E. faecalis and E. faecium) as well as Bacteroides other than B. fragilis and Clostridium were clear predictors of death. Our study demonstrates that IAIs associated with positive cultures of these specific organisms are associated with poorer outcomes. The specific cause and effect relation of these associations remains unclear. The growth of these organisms may occur in sicker patients and therefore herald overall host deterioration. Also, it is possible that these infections are more difficult to treat because of inadequate empiric antimicrobial coverage [14] or the selection of virulence factors resulting in resistant organisms [15]. As we look to improve and refine treatment regimens for IAIs, the ability to recognize organisms associated with death may alert the treating surgeon to a potentially poor outcome and trigger more aggressive treatment.