INTRODUCTION
The prognostic significance of age, body surface area (BSA) burned, and the diagnosis of inhalation injury in burned patients has been well established (1-4 5-12 13, 14 14
Studies in trauma patients have suggested that organ dysfunction developing soon after trauma (i.e. during the first 48 h) rapidly resolves in a high proportion of patients and may reflect reversible derangement in organ function induced by the inciting event or incomplete resuscitation. Thus, it has been proposed that later assessment of organ function (i.e. 3 days after injury) is more adequate to define multiple organ failure in trauma patients and to evaluate the relationship between trauma-induced organ dysfunction and prognosis (15, 16
In the present study, we addressed the hypothesis that in burn patients, early or late organ dysfunction resulting as a consequence of injury may be related to outcome. We have assessed organ dysfunction by using the Sequential Organ Failure Assessment (SOFA) score. This score describes organ dysfunction in critically ill patients and has been shown to be related with mortality (17, 18 inhalation injury . We also studied the impact on outcome of the change in organ dysfunction during the first 4 days after injury.
PATIENTS AND METHODS
Patients
We have studied patients admitted to our intensive care burn unit (ICBU) meeting the following inclusion criteria: diagnosis of thermal injury, total BSA (TBSA) burned greater than 20%, delay from the time of injury to the time of ICBU admission less than 12 h, and length of ICBU stay greater than 3 complete days. Exclusion criteria were length of ICBU stay less than or equal to 3 complete days (due to either early death or brief admission for monitoring; 237 patients), burns due to nonthermal injury (such as chemical [5 patients] or electrical burns [97 patients]), associated nonburn trauma (9 patients), delay from time of injury to ICBU admission greater than or equal to 12 h (15 patients), admission due to non-burn-related diagnoses such as meningococcal sepsis or toxic epidermal necrolysis (41 patients), admission for postoperative care (6 patients), and records unavailable (24 cases). Thus, from 873 admissions, 439 patients comprised the study cohort. This study was approved by our Institutional Human Use Committee and informed consent was obtained.
Patients staying in the ICBU for less than 3 days were excluded because the objective was to analyze the impact of early and late organ dysfunction on outcome. Thus, patients not having the opportunity to develop burn-induced organ dysfunction because of early death or early discharge were not of interest for the purpose of the present study.
Data collection
We collected prospectively the following data: age, TBSA and full-thickness BSA (FTBSA) burned, diagnosis of inhalation injury , sex, mechanisms of injury, date and time of injury and of ICBU admission, ICBU and hospital mortality, and all information necessary to calculate the SOFA score (Table 1 ).
TABLE 1: The SOFA score
Patient management
Patients receive aggressive fluid resuscitation (initiated following the Parkland formula) according to hourly urine output and to the blood pressure (19 20
Measurements
The SOFA score (Table 1 ) was measured on admission (SOFA 0) based on data obtained at the time of ICBU admission, on day 1 (SOFA 1) based on data available at 0800 h the next morning (day 1), and on days 2 (SOFA 2), 3 (SOFA 3), and 4 (SOFA 4) based on data at 0800 h the next days. The difference between SOFA 4 and SOFA 0 (ΔSOFA 0−4) was calculated. The neurological component of the SOFA score was calculated according to the Glasgow coma score as assessed at the scene and after intensive care unit (ICU) admission. In sedated patients, the score was given based on the previous available assessment before sedation.
Definitions and calculations
Total BSA and FTBSA burned were calculated on admission by the plastic surgery fellow using the Lund and Browder diagram and reassessed over the next 3 days by the plastic surgery staff. Mortality represents hospital mortality.
Inhalation injury was diagnosed only if there were signs of inflammation in the lower airway. These included bronchoscopic evidence of inflammatory changes in the lower airway or the presence of soot in the tracheal aspirate. The presence of noncardiogenic pulmonary edema was not considered for the diagnosis of inhalation because extensive burn by itself, without inhalation injury , is a risk factor for noncardiogenic pulmonary edema. Facial edema or upper airway edema visualized by laryngoscopy was not considered for the diagnosis of inhalation because they are not necessarily associated with lower airway inflammation. The requirement for mechanical ventilation was not considered for the diagnosis of inhalation because patients may require ventilatory support or airway protection for indications unrelated to inhalation injury .
Selection of variables
Age, TBSA, and FTBSA burned were categorized to simplify and make the analysis easier to interpret. The cutoff points were chosen for each variable based on four criteria: observation of the value that best discriminates survivors form nonsurvivors by the area under the receiving operator characteristic curve, observation of the change in mortality in a histogram representing mortality for each decade change, clinical relevance, and categories used in previous studies. We therefore categorized age (>60 or ≤60 yrs), TBSA burned (>30% or ≤30%), and FTBSA burned (>20% or ≤20%). Categorization was performed in part to make it the interpretation of the risk of death estimates and the significant interaction terms easier. The reported association between organ dysfunction measurements (SOFA 1, SOFA 4, and ΔSOFA 0−4) and risk of death in the succeeding sentences did not change if the regression analysis was done using continuous, rather than categorical, variables for age and BSA burned.
We chose SOFA 1 for the analysis because in a multivariate model, including SOFAs 0 to 4, the score associated with mortality with the highest odds ratio (OR) for mortality was SOFA 1 (data not shown).
We chose SOFA 4 as a measure of late organ dysfunction because previous studies have considered that dysfunction developing soon after trauma (i.e. during the first 48 h) rapidly resolves in a high proportion of patients and may reflect reversible derangement in organ function induced by the inciting event or incomplete resuscitation, whereas later assessment of organ function (i.e. 3 days after injury) would be more adequate to define multiple organ failure and to evaluate the relationship between trauma-induced organ dysfunction and prognosis (15, 16
The difference between SOFA 0 and SOFA 4 was chosen as the most representative value reflecting the change in organ dysfunction from admission to day 4.
Estimation of the relationship between SOFA 1, SOFA 4, and ΔSOFA 0−4 with the severity of burn trauma and with mortality
We analyzed the association between SOFA 1, SOFA 4, and ΔSOFA 0−4 with mortality using bivariate and multivariate analyses.
In the bivariate analysis, we compared the SOFA 1, SOFA 4, and ΔSOFA 0−4 scores between survivors and nonsurvivors. For the bivariate analysis, continuous variables were compared by the Student t test, and categorical variables were compared by the chi-square test.
Because it was anticipated that older patients and those with more extensive burns or with inhalation injury would have higher SOFA scores, a multivariate analysis was necessary to study the independent association between the SOFA scores and mortality after adjusting for other mortality risk factors.
Thus, we performed multivariate logistic regression analyses to determine in estimative models the relationship between SOFA 1, SOFA 4, and ΔSOFA 0−4 and mortality.
Multivariate logistic regression analyses (backward technique) were performed. In this modeling technique, each variable is excluded from the maximal model regardless of its P value if the change in the OR corresponding to the variable of interest (SOFA 1, SOFA 4, or ΔSOFA 0−4, depending on the analysis) is less than 10%. Interaction variables remain in the model if their P value was less than 0.05. If an interaction remains in the model, each of its components also does, regardless of their P value. The diagnosis of collinearity is performed.
The maximal model included in all cases age, TBSA and FTBSA burned, diagnosis of inhalation injury , and sex either because they were significantly associated with the dependent variable (mortality) in bivariate analysis or because they were clinically relevant for their association with mortality. The associations between SOFA 4 and mortality and between ΔSOFA 0−4 and mortality were also adjusted for SOFA 1. In addition, clinically relevant interactions were included in the models. The diagnosis of collinearity was performed.
If, after this process of elimination, the variable of interest (SOFA 1, SOFA 4, or ΔSOFA 0−4) remained significantly associated with mortality, it was concluded that there exists a significant relationship with mortality after having adjusted for all other variables included in the maximal model.
This strategy differs from that used in predictive models because the objective of this study was not to predict mortality but rather to estimate an association between measurements of organ dysfunction and mortality. In addition, inclusion of early deaths would have been appropriate if the objective would have been mortality prediction.
Values are mean ± SD and percentages. We used the statistical package SPSS 12.1.
RESULTS
Relationship between mortality and other variables
Demographic characteristics of study patients are shown in Tables 2 and 3 . Causes of burns included flame (84%), scald (10%), and other causes (6%). Mortality was higher in older patients, those who had more extensive burns or suffered inhalation injury , and in female patients (Table 3 ).
TABLE 2: Patient characteristics (n = 439)
TABLE 3: Mortality according to different categorical variables
Of the 439 patients, 195 had combined burn and inhalation injury , 244 had only burns, and 3 had isolated inhalation injury . Given the small number, the results presented here did not change after excluding from the analysis patients with isolated inhalation.
On day 1, patients with inhalation injury had, as compared with those without inhalation, higher scores for the cardiovascular, respiratory, hematological, and neurological components of the SOFA score, and a greater degree of impairment in organ function over time (data no shown). In addition, a higher percentage of patients with inhalation had respiratory SOFAs 3 to 4 (PaO2 /FiO2 ≤ 200) on day 1 as compared with those without inhalation (21% vs. 7%; P < 0.001). Thus, although gas exchange abnormalities were associated with inhalation, most patients with inhalation (79%) had PaO2 /FiO2 greater than 200 on day 1.
Almost all patients in our study were apparently well resuscitated because only 17 of 439 (3.9% of all patients) had a urine output less than 0.5 mL kg−1 h−1 on day 0, and 5 (1.1% of all patients) on day 1, and all patients were treated to aim for a normal blood pressure. In fact, the urine output on day 1 was consistent with overresuscitation in this population of patients (1.61 ± 0.04 and 1.63 ± 0.09 mL kg−1 h−1 for survivors and nonsurvivors, respectively; P = 0.57). Sequential Organ Failure Assessment scores and mortality were not related with having a low or a normal urine output on days 0 and 1 (data not shown). Mortality rates of patients without or with oliguria (mean urine output <0.5 mL kg−1 h−1 ) on day 0 (79/422 [19%] vs. 2/17 [12%], respectively; P = 0.75) or on day 1 (80/434 [18%] vs. 1/5 [20%], respectively; P = 0.75) were similar. Thus, changes in organ function and the reported relationship between organ dysfunction and mortality cannot be attributed in these patients to resuscitation failure.
Estimation of the relationship between SOFA 1 and SOFA 4 with mortality
Sequential Organ Failure Assessment scores from admission to day 4 (SOFA 0 to SOFA 4) were higher in nonsurvivors than in survivors (Table 4 ). Sequential Organ Failure Assessment 1 score was higher in patients older than 60 years and in those with TBSA burned greater than 30%, FTBSA burned greater than 20%, or inhalation injury (Table 5 ).
TABLE 4: Relationship between SOFA score and ΔSOFA 0−4 and its components with mortality
TABLE 5: Relationship between SOFA 1 and categorical variables
In an estimative model by multivariate logistic regression analysis, SOFA 1 score was significantly associated with mortality after adjusting in the maximal model for age, TBSA and FTBSA burned, diagnosis of inhalation injury , sex, and for their interactions with SOFA 1. The effect of SOFA 1 on outcome was only significantly influenced by FTBSA burned, as indicated by the significant interaction SOFA 1 × FTBSA burned greater than 20%. Thus, there are two OR values for the relationship between SOFA 1 and mortality according to the different levels of the interaction variable (Table 6 ).
TABLE 6: Mutivariate regression analysis to estimate the relationship between SOFA 1 and outcome, adjusting for other variables
A similar analysis was performed to test the association between SOFA 4 and mortality. Sequential Organ Failure Assessment 4 was independently related to outcome, and only SOFA 1 remained in the final estimative model (Table 7 ). Increasing SOFA 1 or SOFA 4 scores were associated with progressively higher mortality rates (Fig. 1 ).
TABLE 7: Mutivariate regression analysis to estimate the relationship between SOFA 4 and outcome, adjusting for other variables
Fig. 1: Mortality rate in relation to SOFA 1, SOFA 4, and ΔSOFA 0−4 .
Estimation of the relationship between ΔSOFA 0−4 and mortality
The difference between SOFA 0 and SOFA 4 was higher in nonsurvivors than in survivors (Table 4 ). In addition, patients with TBSA burned greater than 30% and FTBSA burned greater than 20% and inhalation injury had higher ΔSOFA 0−4 (Table 8 ).
TABLE 8: Relationship between ΔSOFA 0−4 and categorical variables
Analyzing the difference between survivors and nonsurvivors in the several components of the ΔSOFA 0−4 score over time, there were significant differences in the respiratory, cardiovascular, and hematological components of the SOFA score but not in the renal, hepatic, and neurological components (Table 4 ).
Increasing ΔSOFA 0−4 was associated with progressively higher mortality rates (Fig. 1 ): mortality rates of patients showing a decreased (ΔSOFA 0−4 < 0; n = 105), unchanged (ΔSOFA 0−4 = 0; n = 75), or increased (ΔSOFA 0−4 > 0; n = 259) scores were 5.7%, 9.3%, and 26.3%, respectively (P < 0.0001). In multivariate logistic regression analysis, ΔSOFA 0−4 was significantly associated with mortality (Table 9 ).
TABLE 9: Mutivariate regression analysis to estimate the relationship between ΔSOFA 0−4 and outcome, adjusting for other variables
Organ dysfunction and mode of death
According to the inclusion criteria, all patients survived at least 3 full natural days after admission. Nonsurvivors died a median of 15 (interquartile range [IQR], 8-32 days). Patients without and with inhalation injury died a median of 16 (IQR, 7-29) and 15 (IQR, 9-34) days, respectively (P = 0.07).
Patients dying from hypoxemia had higher respiratory SOFA 4 (3.1 ± 1.0 vs. 2.3 ± 0.9, respectively; P = 0.001) and respiratory ΔSOFA 0−4 scores (1.5 ± 1.6 vs. 0.7 ± 1.4, respectively; P = 0.04) than those dying from other causes. Respiratory SOFAs 0 and 1 did not differ in the two groups.
Similarly, patients dying from shock had higher cardiovascular SOFA 4 than those dying from other causes (2.8 ± 1.5 vs. 2.0 ± 1.6, respectively; P = 0.036). The hematological component at day 1 also differed in patients dying form shock versus other causes (1.1 ± 0.9 vs. 0.6 ± 0.8, respectively; P = 0.02).
Scores at other times or other components did not differ according to the mode of death.
DISCUSSION
The main findings of this study are that organ failure on day 1 after ICBU admission, as measured by the SOFA score, is significantly associated with mortality of burned patients after adjusting for other risk factors. Furthermore, dynamic assessment of organ dysfunction from admission to day 4 is also independently associated with mortality.
After the recognition that mortality in burned patients is strongly associated with age, BSA burned, and diagnosis of inhalation injury (1 5-8 8-12 5-12 15-18 13, 14
The Thermal Injury Organ Failure score was developed specifically for burn patients by Saffle et al. (13 inhalation injury , and the presence of sepsis, and performed very well to predict survival after injury.
Using a score validated in general critically ill patients (the Multiple Organ Dysfunction [MOD] score) (21 14
Assessment of early organ dysfunction
Using the SOFA score, we have quantified organ dysfunction on admission and after the resuscitation phase in a large cohort of 439 burned patients who had BSA burned greater than 20% or inhalation injury . The SOFA score allows for the assessment of organ dysfunction not as an all-or-none phenomenon but rather as a continuum of organ function alteration, permits the consideration of the time course of the degree of dysfunction, and is based on simple parameters. The SOFA score has been validated to assess organ dysfunction in patients with sepsis and general critically ill patients and has been shown to be related to mortality (17, 18
The SOFA 1 score was related to mortality, age, BSA burned, and the presence of inhalation injury (Table 4 ). Furthermore, the SOFA score on day 1 was independently associated with outcome of these patients (Table 5 ). To our knowledge, this is the first study that shows an independent (i.e. after adjusting for other mortality risk factors) association between some measure of organ dysfunction and mortality in burn patients. Higher SOFA 1 scores were associated with more elevated mortality rates, from less than 1% for a SOFA score from 0 to 1 to rates ranging from 80% to 100% for SOFA scores greater than 9 (Fig. 1 ). Our finding, in keeping with the results of other studies (12-14 inhalation injury ) with mortality.
Assessment of late organ dysfunction
We measured the SOFA score on day 4 because it has been suggested that organ dysfunction early after injury may represent reversible organ dysfunction more related primarily to the insult itself and probably to insufficient resuscitation, rather than to the host response to injury (15, 16
The assessment of organ dysfunction on the fourth day after trauma was associated with mortality both in bivariate analysis (Table 4 ) and after adjusting for other mortality risk factors, including the SOFA 1 score (Table 7 ). Increasing SOFA 4 scores were associated with higher mortality rates from approximately 1% for a SOFA score from 0 to 1 to rates ranging from 69% to 78% for SOFA scores greater than 9 (Fig. 1 ). Thus, organ dysfunction assessed after the resuscitation phase is related to mortality, even considering the degree of organ dysfunction immediately after injury.
Assessment of the change in organ dysfunction
Not only a static measurement but also a dynamic assessment of organ dysfunction such as the ΔSOFA 0−4 was related to other risk factors (such as BSA burned or the diagnosis of inhalation injury ; Figure and Table 8 ). There were increasing mortality rates for patients with improved, unchanged, or worsened organ function as assessed by the ΔSOFA 0−4.
This finding is in keeping with results in general ICU patients showing that the change in organ dysfunction over time is related to outcome (22-24 23 23 24
This study adds to our knowledge of burn-induced organ failure. First, our results suggest that other factors, besides those classically considered as determinants of burn mortality (inhalation, age, and BSA burned) and captured by the SOFA score, may play a role in the development of postinjury organ dysfunction and mortality (i.e. different genetic background, different ability to mount an inflammatory response, or different comorbidity).
Second, we observed that not only the absolute value but also the change in the degree of organ dysfunction during the resuscitation phase differed between survivors and nonsurvivors. The components of the SOFA score that differed between survivors and nonsurvivors were the cardiovascular, respiratory, and hematological ones. It can be speculated that this change in organ dysfunction, particularly the change in cardiovascular and respiratory function, could be amenable to some therapeutic intervention inasmuch as it happens after ICU admission.
Third, the validation of the SOFA score in patients in general critically ill patients has been a step forward to standardize a descriptive measure of organ function in critically ill patients, assess in an objective manner changes in organ function (as a result of either their natural course or of some intervention), and compare different populations of patients with different conditions. Thus, the validation of this score in burn patients seems appropriate.
Limitations of the study
It can be argued that some of our patients could have had a deficient resuscitation that might have worsened organ dysfunction and determined a poor clinical course after the first 3 to 4 postburn days. This concern applies to the validation of this score in other population of critically ill patients (10-16, 21-24
We excluded from the present analysis patients with late arrival to our center because they may have developed organ dysfunction before admission to our center. In addition, exclusion of patients with a length of stay in our unit of 3 days or less (because of early death or discharge) was considered appropriate. The exclusion of these patients precludes the possibility of mortality prediction in those patients dying early after injury. However, the objective of this study was not outcome prediction but, rather, to analyze the association between changes in organ function over the first days after injury and mortality, and this can only be studied in patients that remained admitted after the resuscitation phase.
Finally, most of our cases had flame injury (84%), and only 10% had scald burns. Thus, we cannot conclude firmly as to the performance of the SOFA score in nonflame injuries.
SUMMARY
Critically ill burn patients respond differently to burn injury for a same given magnitude of insult (as measured by age, BSA burned, and inhalation injury ), as shown by a different degree of organ dysfunction induced by insults of comparable magnitude. This different response can be assessed by the SOFA score and is related to survival. Furthermore, the time course of organ dysfunction during the first few days after trauma differs in nonsurvivors from survivors, suggesting that interventions before organ dysfunction deteriorates can prevent worsening organ dysfunction and death.
REFERENCES
1. Bull JP, Squire JR: A study of mortality in a burn unit.
Ann Surg 130:160-173, 1949.
2. Smith DL, Cairns BA, Ramadan F, Dalston JS, Fakhry SM, Rutledge R, Meyer AA, Peterson HD: Effect of
inhalation injury , burn size, and age on mortality: a study of 1447 consecutive burn patients.
J Trauma 37:655-659, 1994.
3. O'Keefe GE, Hunt JL, Purdue GF: An evaluation of risk factors for mortality after burn trauma and the identification of gender-dependent differences in outcomes.
J Am Coll Surg 192:153-160, 2001.
4. Ryan CM, Schoenfeld DA, Thorpe WP, Sheridan RL, Cassem EH, Tompkins RG: Objective estimates of the probability of death from burn injuries.
N Engl J Med 338:362-366, 1998.
5. Hamit RF: Factors associated with death of burned patients in a community hospital.
J Trauma 18:405-418, 1978.
6. Tweed A, Ross JF: A review of the mortality in burn units of the Victoria General Hospital and the Izaak Walton Killam Hospital, January 1967 to April 1977.
Ann Plast Surg 2:491-498, 1979.
7. Aikawa N, Shinozawa Y, Ishibiki K, Abe O, Yamamoto S, Motegi M, Yoshii H, Sudoh M: Clinical analysis of multiple organ failure in burned patients.
Burns Incl Therm Inj 13:103-109, 1987.
8. Yao YM, Sheng ZY, Tiam HM, Yu Y, Wang HM, Guo ZR, Gao WY: The association of circulating endotoxinemia with the development of multiple organ failure in burned patients.
Burns 21:255-258, 1995.
9. Sheridan RL, Ryan CM, Yin LM, Hurley J, Tompkins RG: Death in the burn unit: sterile multiple organ failure.
Burns 24:307-311, 1998.
10. Huang YS, Yang ZC, Liu XS, Chen FM, Li A, Crowther RS: Serial experimental and clinical studies on the pathogenesis of multiple organ dysfunction syndrome (MODS) in severe burns.
Burns 24:708-716, 1998.
11. Huang YS, Yang ZC, Chen FM, Crowther RS, Li A: Effects of early eschar excision en masse at one operation for prevention and treatment of organ dysfunction in severely burned patients.
World J Surg 23:1272-1278, 1999.
12. Marshall WG, Dimick AR: The natural history of major burns with multiple subsystem failure.
J Trauma 23:102-105, 1983.
13. Saffle JR, Sullivan JJ, Tuohig GM, Larson CM: Multiple organ failure in patients with thermal injury.
Crit Care Med 21:1673-1683, 1993.
14. Cummings J, Purdue GF, Hunt JL, O'Keefe GE: Objective estimates of the incidence and consequences of multiple organ dysfunction and sepsis after burn trauma.
J Trauma 50:510-515, 2001.
15. Moore FA, Sauaia A, Moore EE, Haenel JB, Burch JM, Lezotte DC: Postinjury multiple organ failure: a bimodal phenomenon.
J Trauma 40:501-512, 1996.
16. Sauaia A, Moore FA, Moore EE, Haenel JB, Read RA, Lezotte DC: Early predictors of post-injury multiple organ failure.
Arch Surg 129:39-45, 1994.
17. Vincent JL, de Mendonca A, Cantraine F, Moreno R, Takala J, Suter PM, Sprung CL, Colardyn F, Blecher S: Use of the SOFA score to assess the incidence of organ dysfunction/failure in intensive care units: results of a multicenter, prospective study.
Crit Care Med 26:1793-1800, 1998.
18. Vincent JL, Moreno R, Takala J, Willatts S, De Mondiça A, Bruining H, Reinhart CK, Suter PM, Thijs LG: The SOFA (Sepsis-related Organ Failure Assessment) score to describe organ dysfunction/failure.
Intensive Care Med 22:707-710, 1996.
19. Lorente JA, Ezpeleta A, Esteban A, Gordo F, de la Cal MA, Díaz C, Arévalo JM, Tejedor C, Pascual T: Systemic hemodynamics, gastric intramucosal PCO2 changes, and outcome in critically ill burn patients.
Crit Care Med 28:1728-1735, 2000.
20. de La Cal MA, Cerda E, Garcia-Hierro P, van Saene HK, Gomez-Santos D, Negro E, Lorente JA: Survival benefit in critically ill burned patients receiving selective decontamination of the digestive tract: a randomized, placebo-controlled, double-blind trial.
Ann Surg 241:424-430, 2005.
21. Marshall JC, Cook DJ, Christou NV, Bernard GR, Sprung CL, Sibbald WJ: Multiple organ dysfunction score: a reliable descriptor of a complex clinical outcome.
Crit Care Med 23:1638-1652, 1995.
22. Ferreira FL, Bota DP, Bross A, Mélot C, Vincent JL: Serial evaluation of the SOFA score to predict outcome in critically ill patients.
JAMA 286:1754-1758, 2001.
23. Russell JA, Singer J, Bernard GR, Wheeler A, Fulkerson W, Hudson L, Schein R, Summer W, Wright P, Walley KR: Changing pattern of organ dysfunction in early human sepsis is related to mortality.
Crit Care Med 28:3405-3411, 2000.
24. Levy MM, Macias WL, Vincent JL, Russell JA, Silva E, Trzaskoma B, Williams MD: Early changes in organ function predict eventual survival in severe sepsis.
Crit Care Med 33:2194-2202, 2005.
