Fungal wound infection (not colonization) is independently associated with mortality in burn patients

Abstract

Objective: To analyze the occurrence of fungal wound infection (FWI) after thermal injury and its relationship to mortality.

Background: FWI is an uncommon but potentially lethal complication of severe thermal injury.

Methods: The records of patients with thermal burns admitted to a single burn center (1991-2002) were reviewed. Analyses accounted for total burn size (TBS, percentage body surface area), full-thickness burn size (FTBS, percentage body surface area), age, inhalation injury, sex, and fungal-status category. Fungal colonization and infection were determined histopathologically.

Results: Criteria for inclusion were met by 2651 patients. Each patient's fungal-status category was defined according to the deepest level of fungal involvement observed during the hospital course: no fungus (2476 patients), fungal wound colonization (FWC, 121 patients), or fungal wound infection (FWI, 54 patients). Median TBS (9%, 47%, 64%, respectively) and mortality (5%, 27%, 76%, respectively) varied significantly among fungal-status groups. Logistic regression was used to detect significant independent associations. FWI was associated with higher TBS. Mortality was associated with TBS, FTBS, inhalation injury, FWI, and age. Unlike FWI, FWC was not independently related to mortality, the greater observed mortality in FWC being explained by other variables such as TBS. The odds ratio for FWI (8.16) suggested about the same mortality impact as augmenting TBS by 33%. A midrange TBS of 30% to 60% was required for most of the detectable association of FWI with mortality.

Conclusions: FWI accompanies larger burns and is associated with mortality in burn patients, particularly in those with TBS 30% to 60%. This association is independent of burn size, inhalation injury, and age.

FIGURE 1. A, Correlation of burn size with histopathologicallycategorized deepest level of fungus during the hospital course. Vertical lines represent the interquartile range (IQR) for TBS, connecting the 25th and 75th percentiles. TBS, total burn size (percentage body surface area [BSA]); FTBS (% BSA); FWC, fungal wound colonization; FWI, fungal wound infection. TBS: P < 0.001 no fungus versus FWC, no fungus versus FWI, and FWC versus FWI. FTBS: P < 0.001 no fungus versus FWC, no fungus versus FWI; and P < 0.01 FWC versus FWI. B, TBS distribution according to the fungal category of each patient.

FIGURE 2. A, Higher observed mortality with deeper fungal involvement. FWC, fungal wound colonization; FWI, fungal wound infection (invasion). Each category has significantly different (P < 0.001) mortality from each of the others. B, Observed mortality in each fungal-status category within each burn-size group. TBS, total burn size (percentage body surface area).

FIGURE 3. Observed mortality in relation to total burn size (TBS, percentage body surface area [BSA]) and fungal wound infection (FWI). FWI-No includes FWC and no fungus. TBS groups are represented in the following way: the first (left-most) symbol for FWI-Yes and for FWI-No represents TBS ≤30%, the second symbol TBS >30%–60%, and the third (right-most) symbol TBS >60%. For the second (middle) point in each curve, mortality is different between FWI-Yes and FWI-No (P < 0.001), although not significantly different (FWI-Yes vs. FWI-No) at the first points or at the third points.

FIGURE 4. Predicted mortality probability from the logistic model of Equation 3 developed with inclusion of fungal information as FWI. The curve has FWI set to “no,” TBS set to 45%, FTBS set to 30%, inhalation injury set to absent, and age points varying across the abscissa. The variables other than age must be held constant over the full range of age, to see only the effect of age in the predicted curve. Entries below the plot represent actual patient data in age groups chosen to represent various parts of the curve. Obs, observed deaths (mortality). Pred, predicted deaths (mortality); pred deaths as the sum of the probabilities calculated with Equation 3 (and actual patient data). Obs and pred mortality are in good agreement across age groups. Age is in years. TBS and FTBS are in percentage body surface area.

FIGURE 5. Receiver operator characteristic (ROC) curves comparing the ability of the mortality model including FWI as an independent variable (Equation 3) and the model without fungal information (Equation 2) to discriminate between observed survival and nonsurvival. For this comparison and these curves, the 2 models (each developed in all the patients) were applied only to the 298 patients with TBS >30%–60%. The area under the curve (AUC) from application of Equation 3 (AUC = 0.867) was significantly greater (P = 0.005) than that from application of Equation 2 (AUC =0.826). The straight diagonal line would indicate no discrimination (AUC = 0.5).