Morbidity and Mortality of Bloodstream Infections in Patients With Severe Burn Injury

The study was conducted in the burn unit of Ghent University Hospital, Ghent, Belgium. The unit serves a geographic area of about 2.6 million inhabitants. Approximately 60 to 80 patients are admitted to the unit each year. Among the total population of burn patients, the median total burned surface area (TBSA) is 13%, median age is 30 years, approximately 12% have inhalation injury, the median length of stay is 12 days, and mortality is 7%.¹ 

The unit has 6 separate isolation rooms; each room is equipped with a shower and bath. Intensivists from the surgical intensive care unit and plastic surgeons are responsible, respectively, for intensive care and wound care.

A mixed crystalloid-colloid scheme is used for fluid replacement. In the 1980s, human albumin was used as the colloid; since the 1990s, semisynthetic colloids such as starches and gelatin solution have been used. Replacement is started with an hourly dose of 2 mL/kg per 1% burned surface area. On the first day, one-half of the calculated fluid requirement is administered within the first 8 hours after the burn; the other one-half is administered during the next 16 hours. In the first 24 hours after the burn, one-third of the total fluid volume consists of colloids; the other two-thirds, hypertonic saline (1 L of 0.9% sodium chloride + 50 mEq sodium bicarbonate). During the second 24 hours, colloids account for two-thirds of the total fluid volume. Volume replacement is guided by a patient’s urine output and hemodynamic status.

Early enteral nutrition has been the practice in the unit since 1998. During the period 1985 to 2001, patients were showered daily with chlorhexidine solution. During the last 2 years of the study, a povodoneiodine solution was used. From 1985 to 1998, partial-thickness burns were covered with silver sulfadiazine; thereafter hydrocolloid dressings have been used. Full-thickness burns are covered with cerium nitrate–silver sulfadiazine. Early excision of burn wounds has never been the practice in the burn unit. Use of polarized light to stimulate wound healing in partial-thickness burns was introduced in the unit for scientific purposes in 1996. Since 1998, all partial-thickness burns are treated with polarized light, limiting the need for surgery for full-thickness burns and difficulty in healing deep dermal burn wounds. Twice weekly microbiological monitoring indicates each patient’s colonization status. No antibiotic prophylaxis is used.

A retrospective, pairwise-matched (matching ratio 1:2 or, if not feasible, 1:1), risk-adjusted cohort study was performed with “exposed” patients admitted to the burn unit between 1992 and 2006 in whom microbiological evidence indicated BSI. Nonexposed patients were selected from a database that included all burn patients admitted to the unit. The study was approved by the appropriate ethics committee.

A prospective, case- and laboratory-based surveillance program of BSIs by the infection control team was used for the retrospective search for all burn patients with BSI. Patients were presumed to have a BSI if microorganisms were detected in blood cultures. Subsequent ad hoc determination of clinical significance, and presumed or definite initial focus of infection were established by mutual agreement among the attending intensivist, infectiologist, and microbiologist. In patients who had multiple episodes, only the first BSI episode was considered.

The purpose of matched cohort studies is to achieve reliable estimates of attributable mortality through accurate adjustment for confounding covariates. Hence, strict matching on prognostic factors is crucial.^13,16–21  Therefore, exposed patients were matched with nonexposed patients according to identical Belgian Outcome in Burn Injury (BOBI) scores.²²  For the BOBI 10-point classification, 3 major risk factors—mortality, age, and TBSA—and inhalation injury are considered.^22,23  Age is divided into 4 risk categories (0–3 points); TBSA, into 5 (0–4 points). The presence of inhalation injury is scored as 3 additional points. This matching procedure resulted in an equal a priori expected mortality and allows assessment of the impact of a subsequent complication. Nonexposed patients were selected within a 5-year time frame before or after the admission year of the index BSI patient. Nonexposed patients were required to have a time to discharge at least equal to the time to event in the corresponding exposed patient.¹⁷  Selection of non-exposed patients was made without knowledge of outcome. If data on more than 2 potential nonexposed patients were available, matching was based on the admission date nearest to that of the index exposed patient.

Definitions of BSI, determination of BSI sources, methods for antimicrobial susceptibility testing, and appropriate antimicrobial therapy have been reported elsewhere.^24–26  Antimicrobial resistance was defined^24,27,28  as resistance to fluconazole for Candida species, as resistance to methicillin for staphylococci, as resistance to vancomycin for enterococci, as resistance to ampicillin for streptococci, as production of extended-spectrum β-lactamases for Enterobacter species, and as resistance to 1 of the following agents for gram-negative nonfermenting bacteria: ceftazidime, piperacillin, ciprofloxacin, imipenem, or meropenem.

Blood samples for cultures were obtained routinely if a patient’s temperature increased to more than 38.4°C or if bacteremia was suspected because of unstable hemodynamic status, chills, or new organ failure. Samples were processed according to the BacT/Alert (Organon Teknika Corp, Durnham, North Carolina) procedure. Susceptibility testing was done in accordance with the latest guidelines recommended by the National Committee on Clinical Laboratory Standards or the Clinical and Laboratory Standards Institute at any time during the study period.

Evaluation of clinical outcome was based on the need for organ support (eg, mechanical ventilation, renal replacement therapy, vasopressor or inotropic support) and in-hospital mortality. Excess length of hospitalization and ventilator dependency were used as major indicators of added morbidity. Excess length of hospitalization was calculated by subtracting the median length of hospitalization of the nonexposed group from the median length of hospitalization of the exposed group. Because exposed patients were matched on the basis of exposure time (time to event in exposed patients vs time to discharge in nonexposed patients), the difference in length of stay indicates the added proportion of hospitalization due to the infectious complication.¹⁷ 

Mann-Whitney U and χ²  tests were used as appropriate. Relationships with mortality were assessed by using logistic regression analysis. The following variables were entered in the regression model and were stepwise removed if P > .10: age, sex, TBSA, inhalation injury, duration of hospitalization, BSI, acute kidney injury, and need for vasopressor support. Odds ratios and 95% confidence intervals are reported. Covariates with a plausible relationship with mortality or P < .10 in univariate analysis were included in the model. Because BSI was the variable of interest, it was kept in the model regardless of the associated P value. All tests were 2 tailed.

During the 15-year period, 1125 patients were admitted to the burn unit. In total, 178 episodes of BSI occurred (prevalence, 15.8 episodes per 100 admissions) in 76 patients. Among the 76 patients with BSI, 43 had multiple episodes (mean, 3.4 episodes). Table 1 summarizes the causative pathogens involved. In 28% of the BSI episodes, multiple bacteria were detected. Median time between admission and onset of BSI was 11 days (IQR, 5–20). A total of 39 episodes were primary BSIs (39/7651%); of these, 23 were due to contaminated catheters (23/76, 30%). The burn wound was the main source of secondary BSIs (36/76, 47%). In 58 patients, appropriate antibiotic therapy was initiated within 48 hours. In-hospital mortality was 12% (n=9) among the 76 patients who had BSI and 9% (n = 5) among the 55 patients whose first episode was due to a single pathogen. Of the 21 patients with a polymicrobial first episode, 19% died (n = 4).

Matching was successful for all patients, but for 42 exposed patients, only 1 suitable nonexposed patient was found. Thus, the matching procedure resulted in a matched cohort study with 76 exposed and 103 nonexposed subjects. Table 2 summarizes the characteristics of the exposed and nonexposed groups. Compared with nonexposed patients, patients with BSI had a higher need for vasopressor or inotropic support. Need for ventilatory assistance and renal replacement therapy did not differ significantly between the 2 groups. Hospital mortality also did not differ significantly between the groups. However, duration of hospital stay was 25 days longer for patients with BSI (61 days) than for nonexposed patients (36 days) and duration of mechanical ventilation was 11 days longer (21 vs 10 days). Logistic regression analysis confirmed that BSI did not affect survival (Table 3).

In this cohort of burn patients with BSI, the infections did not contribute to mortality but did cause additional morbidity; compared with the nonexposed patients, patients with BSI had a greater need for vasopressors or inotropic agents and longer durations of ventilator dependency and hospitalization. The absence of mortality attributable to BSI in the exposed group was confirmed by multivariate regression analysis. Instead, older age, greater TBSA, renal replacement therapy, and need for vasopressors or inotropic agents were risk factors for mortality. Strangely enough, inhalation injury, a factor that strongly compromised prognosis in other studies,^1,29–31  was not associated with death. Probably, inhalation injury contributes more to early deaths, whereas our BSI cohort generally included survivors of the acute phase of injury who were prone to late-onset infectious complications.

In contrast to previous reports,^9,10,14,15  our findings indicate that once adjusted for prognostic covariates, BSI does not necessarily impede survival in burn victims. No excess mortality^{24,25,32–35}  and dramatic mortality rates attributable to infection have been reported before in bacteremic critically ill patients.^15,19,36,37  The likelihood of survival depends on unchangeable characteristics such as severity of disease, causative pathogen, antimicrobial resistance, and the patient’s age.^{12,13,38–40}  Optimizing the odds of survival requires prompt initiation of appropriate therapy^3,41–43 ; in our study, 58 of the 76 patients (76%) received appropriate antibiotic therapy. However, mortality in patients who did not receive appropriate therapy within the critical time frame of 48 hours (12%) was not worse than mortality in patients who did receive appropriate therapy (11%). Possibly, most BSIs treated inappropriately were caused by coagulase-negative staphylococci (methicillin-resistant), known as low-virulence pathogens. Anyhow, insufficient power in the study hampered this particular analysis.

Because matched cohort studies are prone to selection and survival bias,^13,17,44  reliability of the nonexposed group is crucial. To avoid survival bias, we required that nonexposed patients have a time to discharge at least as long as the time to event (ie, onset of BSI) of the exposed patients. Additionally, we matched patients on the basis of the BOBI score because this classification summarizes the most powerful prognostic indicators.^22,23  An advantage of this matching procedure is that the validity of the nonexposed group can be judged by comparing the observed and expected mortality rates. In our study, the observed mortality of the nonexposed group (17.5%) was within the 95% confidence interval of the expected mortality (14.7%–30.9%), indicating a reliable nonexposed group with an outcome in line with the expectations. The percentages of TBSA differed statistically between the 2 groups, but the clinical relevance of the observed difference (30% vs 40%) was minor because mortality associated with the percentage of TBSA only starts to increase substantially from more than 40%, as indicated by Ryan et al.⁴⁵  Also in the BOBI score, TBSAs from 21% to 40% correspond to 1 point. Therefore, exposed and nonexposed patients might have different percentages of TBSA but still have identical BOBI scores and hence have similar outcome predictions. Consequently, our finding that TBSA was significantly lower among the nonexposed patients is overruled by the equal BOBI scores.

A disadvantage of this study is the single-center design. Nevertheless, the homogenous standard of care for both exposed and nonexposed patients is an advantage of this approach. Another potential weakness is that exposed patients were recruited during a long period in which favorable evolutions in survival occurred. We tried to counter this problem by selecting nonexposed patients on the basis of the year of admission, but a 5-year frame was necessary to match all exposed patients. Selecting controls from a 5-year period may be borderline acceptable, because survival improved during the 5-year period (odds ratio, 0.73; 95% confidence interval, 0.56–0.94).¹ 

In conclusion, in this cohort of burn patients, BSI did not adversely affect survival. BSI was, however, associated with a significant increase in the durations of mechanical ventilation and hospitalization, thereby representing a substantial economic burden. These data underscore the need for vigorous application of evidence-based, cost-effective preventive measures.

These results were presented in part at the Congress of the European Burn Association, Lausanne, Switzerland, September 2–5, 2009. We thank Mr J. Deschuijmer and Prof G. Verschraegen of the hospital hygiene team at Ghent University Hospital for their continual efforts in the surveillance of nosocomial bloodstream infections.