sign in sign up
<<
Home , Sepsis

Epidemiology and risk factors of sepsis after multiple trauma: An analysis of 29,829 patients from the Trauma Registry of the German Society for Trauma *

Arasch Wafaisade, PhD; Rolf Lefering, MD; Bertil Bouillon, MD; Samir G. Sakka, MD; Oliver C. Thamm, MD; Thomas Paffrath, MD; Edmund Neugebauer, PhD; Marc Maegele, MD; Trauma Registry of the German Society for Trauma Surgery

Objectives: The objectives of this study were 1) to assess spective subperiods (16.9%, 16.0%, 13.7%, and 11.9%; p < potential changes in the incidence and outcome of sepsis after .0001). For the subgroup of patients with sepsis, the mortality -respec ,(054. ؍ multiple trauma in Germany between 1993 and 2008 and 2) to rates were 16.2%, 21.5%, 22.0%, and 18.2% (p evaluate independent risk factors for posttraumatic sepsis. tively. The following independent risk factors for posttraumatic Design: Retrospective analysis of a nationwide, population- sepsis were calculated from a multivariate logistic regression based prospective database, the Trauma Registry of the German analysis: male gender, age, preexisting medical condition, Glas- Society for Trauma Surgery. gow Scale score of <8 at scene, Injury Severity Score, > Setting: A total of 166 voluntarily participating trauma centers Abbreviated Injury ScaleTHORAX score of 3, number of injuries, (levels I–III). number of red cell units transfused, number of operative Patients: Patients registered in the Trauma Registry of the procedures, and laparotomy. German Society for Trauma Surgery between 1993 and 2008 with Conclusions: The of sepsis decreased significantly complete data sets who presented with a relevant trauma load over the study period; however, in this decade the incidence (Injury Severity Score of >9) and were admitted to an intensive remained unchanged. Although overall mortality from multiple trauma has declined significantly since 1993, there has been no .(29,829 ؍ care unit (n Interventions: None. significant decrease of mortality in the subgroup of septic trauma Measurements and Main Results: Over the 16-yr study period, patients. Thus, sepsis has remained a challenging 10.2% (3,042 of 29,829) of multiply injured patients developed after trauma during the past 2 decades. Recognition of the iden- sepsis during their hospital course. Annual data were summarized tified risk factors may guide early diagnostic workup and help to into four subperiods: 1993–1996, 1997–2000, 2001–2004, and reduce septic complications after multiple trauma. (Crit Care Med 2005–2008. The incidences of sepsis for the four subperiods were 2011; 39:621–628) 14.8%, 12.5%, 9.4%, and 9.7% (p < .0001), respectively. In- KEY WORDS: multiple trauma; sepsis; risk factors; incidence; hospital mortality for all trauma patients decreased for the re- outcome; mortality; multivariate analysis

everal factors have been recog- survive initial trauma since drome and (6–11). In a large nized to substantially contrib- it represents the leading cause of death in trauma study, Osborn et al (12) identified ute to impaired outcomes after noncardiac intensive care units (ICUs) the Injury Severity Score (ISS), the Re- severe injury, but sepsis is one (1, 2). However, while overall mortality vised Trauma Score, the Glasgow Coma Sof the most significant in patients who from multiple trauma has gradually de- Scale (GCS) score, comorbidities, and creased in western countries over the male gender as predictors for posttrau- past few decades (3–5), there is only matic sepsis but were not able to adjust *See also p. 876. limited information on the develop- for transfusion or other therapy-related From the Departments of Trauma and Orthopedic ment of sepsis in injured patients over variables. Furthermore, the predictive Surgery (AW, BB, TP, MM) and and recent years. Epidemiologic data from value of specific injury patterns and the Intensive Care (SGS) and Institute for Re- the have shown a general role of distinct anatomical regions have search in Operative Medicine (AW, RL, OCT, EN, MM), decrease in the incidence of sepsis (2); not been elucidated. Altogether, indepen- University of Witten/Herdecke, Cologne-Merheim Med- ical Center, Cologne; and Committee on Emergency however, comparable European investi- dent risk factors for postinjury sepsis are Medicine, Intensive and Trauma Care (Sektion NIS), gations do not exist. still poorly defined, although they might German Society for Trauma Surgery, Berlin, Germany. In addition, independent predictors be useful adjuncts to identify high-risk The authors have not disclosed any potential con- for sepsis during trauma-associated hos- patients at an early stage. flicts of interest. For information regarding this article, E-mail: pitalization are debated controversially. Therefore, using a large multicenter [email protected] Several authors have reported that the population-based trauma registry, the Copyright © 2011 by the Society of Critical Care transfusion of allogenic blood during aims of the present study were 1) to in- Medicine and Lippincott Williams & Wilkins trauma resuscitation is associated with vestigate the incidence and outcome of DOI: 10.1097/CCM.0b013e318206d3df the systemic inflammatory response syn- sepsis after multiple trauma and potential

Crit Care Med 2011 Vol. 39, No. 4 621 temporal changes since the establish- Table 1. Basic demographic/clinical characteristics upon admission and outcomes of multiple trauma ment of the database in 1993 and 2) to patients with and without sepsis (1993–2008) identify independent risk factors for sep- sis after multiple trauma. Sepsis, No Sepsis, Demographic/Clinical Characteristic, Outcome 3,042 (10.2%) 26,787 (89.8%) p

PATIENTS AND METHODS Age (yrs; mean Ϯ SD)44Ϯ 19 42 Ϯ 21 Ͻ.0001 Male (%) 80.9 72.1 Ͻ.0001 The Trauma Registry of the Deutsche Ge- Blunt/penetrating trauma (%) 95.9/4.1 95.5/4.5 .356 sellschaft fu¨ r Unfallchirurgie. The Trauma (%) 12.7 9.0 Ͻ.0001 Registry of the Deutsche Gesellschaft fu¨ r Un- Injury Severity Score (points; mean Ϯ SD)33Ϯ 13 25 Ϯ 13 Ͻ.0001 Ն Ͻ AISHEAD score of 3 (%) 54.6 50.5 .0001 fallchirurgie (German Society for Trauma Ն Ͻ AISTHORAX score of 3 (%) 67.8 49.5 .0001 Surgery) (TR-DGU) (13) was founded in 1993 Ն Ͻ AISABDOMEN score of 3 (%) 29.8 18.0 .0001 as an initiative for prospective, standardized, Ն Ͻ AISEXTREMITIES score of 3 (%) 49.0 37.5 .0001 and anonymous documentation of severely in- SBP at the scene (mm Hg; mean Ϯ SD) 112 Ϯ 33 120 Ϯ 32 Ͻ.0001 jured patients. At present, 166 trauma centers SBP at the scene of Յ90 mm Hg (%) 26.5 17.4 Ͻ.0001 of all levels are participating voluntarily at the at the scene (beats per minute; mean Ϯ SD)98Ϯ 25 93 Ϯ 23 Ͻ.0001 TR-DGU, mainly from Germany, but also 17 score at the scene 10.0 Ϯ 4.8 11.1 Ϯ 4.6 Ͻ.0001 hospitals from Austria, Switzerland, The Neth- (points; mean Ϯ SD) erlands, Belgium, and Slovenia (http:// Glasgow Coma Scale score at the scene of Յ8 (%) 39.4 29.3 Ͻ.0001 www.traumaregister.de). Recently, Ͼ6,000 pa- Intravenous fluids, prehospital (mL; mean Ϯ SD) 1637 Ϯ 1241 1294 Ϯ 1055 Ͻ.0001 Ϯ Ϯ Ϯ Ͻ tients have been registered annually, and it is SBP at the ER (mm Hg; mean SD) 117 30 124 29 .0001 SBP at the ER of Յ90 mm Hg (%) 20.9 11.2 Ͻ.0001 estimated that the TR-DGU covers about 25% Heart rate at the ER (beats per minute; mean Ϯ SD)95Ϯ 23 90 Ϯ 21 Ͻ.0001 of all severe trauma cases in Germany (5). The (g/dL; mean Ϯ SD) 10.7 Ϯ 2.8 11.8 Ϯ 2.7 Ͻ.0001 Ϫ data are collected prospectively and are struc- (nL 1; mean Ϯ SD) 187 Ϯ 82 206 Ϯ 81 Ͻ.0001 tured into four consecutive time phases: 1) (Quick%; mean Ϯ SD)72Ϯ 23 80 Ϯ 22 Ͻ.0001 prehospital phase, 2) emergency room and (mmol/L; mean Ϯ SD) Ϫ4.2 Ϯ 5.1 Ϫ2.9 Ϯ 4.6 Ͻ.0001 initial treatment until ICU admission, 3) ICU, Intravenous fluids, ER to 3560 Ϯ 3641 2457 Ϯ 2547 Ͻ.0001 and 4) discharge and list of injuries and inter- (mL; mean Ϯ SD) ventions. The registry contains detailed infor- Preexisting medical condition (%) 35.4 28.3 Ͻ.0001 a mation on the demographics, injury severity pRBC transfusion (%) 45.4 24.9 Ͻ.0001 a Ն Ͻ and pattern, prehospital and in-hospital man- Massive transfusion ( 10 pRBC units) (%) 15.2 5.8 .0001 pRBC unitsa (n; mean Ϯ SD) 4.3 Ϯ 8.1 1.8 Ϯ 5.1 Ͻ.0001 agement, time course, and outcome of each unitsa (n; mean Ϯ SD) 2.3 Ϯ 6.3 0.9 Ϯ 3.5 Ͻ.0001 patient. In 2002, data collection moved from Operative procedures (n; mean Ϯ SD) 5.9 Ϯ 6.5 3.0 Ϯ 3.4 Ͻ.0001 paper form documentation to a central online Laparotomy (%) 27.3 15.0 Ͻ.0001 registration system via the Internet. The doc- failure (%) 82.9 34.8 Ͻ.0001 umentation includes standardized scoring sys- Cardiovascular (%) 63.5 20.1 Ͻ.0001 tems, e.g., the GCS and the Abbreviated Injury Respiratory (%) 63.5 17.3 Ͻ.0001 Scale (AIS) (14, 15). (%) 35.4 18.4 Ͻ.0001 This study was approved by the ethical Hepatic (%) 21.2 2.4 Ͻ.0001 Ͻ board of our institution and is in accordance Renal (%) 20.9 2.6 .0001 Hematologic (%) 18.5 5.2 Ͻ.0001 with the standards of the Helsinki Declaration Multiple-organ failure (%) 63.2 17.3 Ͻ.0001 of 1975. Intensive care unit length of stay (days; mean Ϯ SD)28Ϯ 22 10 Ϯ 12 Ͻ.0001 Study Population. Until 2008, 42,248 pa- In-hospital length of stay (days; mean Ϯ SD)45Ϯ 37 26 Ϯ 28 Ͻ.0001 tients have been registered in the TR-DGU Ventilator days (mean Ϯ SD)22Ϯ 19 6 Ϯ 9 Ͻ.0001 database. For the present study, patients were 30-day mortality (%) 15.7 12.1 Ͻ.0001 excluded 1) if any data on demographics, in- In-hospital mortality overall (%) 19.5 12.5 Ͻ.0001 jury severity, or prehospital and in-hospital course were missing (n ϭ 7466), 2) if patients AIS, Abbreviated Injury Scale; ER, emergency room; pRBCs, ; SBP, systolic presented with mild injury defined as ISS Ͻ9 . The small p values are based on the large sample size, and interpretation has to (n ϭ 3068), and 3) if they were not admitted to consider the clinical importance of the observed difference. a the ICU (n ϭ 1885). Therefore, a total of Blood products transfused between ER arrival and intensive care unit admission. 29,829 patients remained for further analysis. To evaluate temporal changes over the using the revised multiple-organ failure comparison of more than two groups (four study period, annual data were summarized (MOF) score of Goris et al (17). Since the subperiods of trauma years; Tables 2 and 3). into four subperiods as described previously introduction of the online version of the reg- Data are presented as mean Ϯ SD for contin- (2), for the complete study population (see istry in 2002, the Sequential Organ Failure uous variables and as percentages for inci- Table 2) and for the subgroup of patients with Assessment score was used to determine organ dence rates. sepsis (see Table 3). function status (18). MOF was defined as or- Statistical Analysis: Multivariate. A multi- Definitions of Sepsis and Organ Failure. In gan failure for at least two of the listed organs/ variate analysis was performed by stepwise lo- the data report form, the occurrences of sepsis systems. gistic regression with sepsis as the dependent and organ failure (cardiovascular, , cen- Statistical Analysis: Univariate and Strat- variable to identify independent risk factors tral nervous system, kidney, liver, coagula- ification. For categorical variables, data were for the development of posttraumatic sepsis. tion) during the hospital course are docu- compared between groups using the chi- The variables that were analyzed as potential mented dichotomously for each as either “yes” square test. For continuous variables, the predictors for sepsis were selected from the or “no.” Sepsis was coded according to the Mann-Whitney U test was applied for compar- literature and from the clinical experience of criteria of Bone (16) during the complete ison of two groups (sepsis vs. nonsepsis pa- our research group. These variables consisted study period. Organ failure was assessed by tients; Table 1) and the Kruskal-Wallis test for of patient characteristics (age, gender, preex-

622 Crit Care Med 2011 Vol. 39, No. 4 Table 2. Basic characteristics, injury patterns, and outcomes of the complete study population, stratified according to their respective trauma years (n ϭ 29,829)

Characteristic, Injury Pattern, Outcome 1993–1996 1997–2000 1997–2000 2005–2008 p n 1,543 3,771 8,403 16,112 Participating hospitals 19 72 104 166 Age (yrs; mean Ϯ SD)39Ϯ 19 39 Ϯ 19 42 Ϯ 20 44 Ϯ 21 Ͻ.0001 Male (%) 71.4 72.5 73.2 73.2 .414 Injury Severity Score (points; mean Ϯ SD)24Ϯ 12 27 Ϯ 14 25 Ϯ 13 27 Ϯ 13 Ͻ.0001 Ն Ͻ AISHEAD score of 3 (%) 43.3 49.9 50.4 52.2 .0001 Ն Ͻ AISTHORAX score of 3 (%) 48.9 46.1 48.9 54.1 .0001 Ն Ͻ AISABDOMEN score of 3 (%) 20.9 21.3 20.2 18.0 .0001 Ն Ͻ AISEXTREMITIES score of 3 (%) 46.1 44.0 38.1 37.0 .0001 Glasgow Coma Scale score at the scene of Յ8 (%) 30.4 34.9 30.6 29.0 Ͻ.0001 Preexisting medical condition (%) 1.5 17.4 29.4 34.8 Ͻ.0001 Packed transfusiona (%) 41.3 37.4 25.7 23.9 Ͻ.0001 Operative procedures (n; mean Ϯ SD) 3.5 Ϯ 3.3 3.6 Ϯ 3.6 3.1 Ϯ 3.8 3.3 Ϯ 4.1 Ͻ.0001 Laparotomy (%) 19.8 18.7 16.5 15.2 Ͻ.0001 Sepsis (%) 14.8 12.5 9.4 9.7 Ͻ.0001 Organ failure (%) 37.0 34.6 38.7 45.6 Ͻ.0001 Multiple-organ failure (%) 22.3 18.3 20.5 27.0 Ͻ.0001 Intensive care unit length of stay (days; mean Ϯ SD)13Ϯ 16 13 Ϯ 15 12 Ϯ 15 11 Ϯ 13 Ͻ.0001 In-hospital length of stay (days; mean Ϯ SD)31Ϯ 31 30 Ϯ 31 29 Ϯ 33 26 Ϯ 27 Ͻ.0001 30-day mortality (%) 16.3 15.1 12.8 11.3 Ͻ.0001 In-hospital mortality overall (%) 16.9 16.0 13.7 11.9 Ͻ0.0001

AIS, Abbreviated Injury Scale. aPacked red blood cells transfused between emergency room arrival and intensive care unit admission.

Table 3. Basic characteristics, injury patterns, and outcomes of the subgroup of patients who developed sepsis during the hospital course, stratified according to their respective trauma years (n ϭ 3042)

Characteristic, Injury Pattern, Outcome 1993–1996 1997–2000 2001–2004 2005–2008 p n 228 470 786 1558 Age (yrs; mean Ϯ SD)38Ϯ 16 41 Ϯ 17 43 Ϯ 19 45 Ϯ 20 Ͻ.0001 Male (%) 75.7 80.4 81.7 81.5 .195 Injury Severity Score (points; mean Ϯ SD)28Ϯ 11 33 Ϯ 13 31 Ϯ 13 34 Ϯ 13 Ͻ.0001 Ն Ͻ AISHEAD score of 3 (%) 43.9 52.1 48.2 60.1 .0001 Ն Ͻ AISTHORAX score of 3 (%) 60.1 63.0 64.6 71.9 .0001 Ն AISABDOMEN score of 3 (%) 32.0 33.4 29.4 28.5 .189 Ն Ͻ AISEXTREMITIES score of 3 (%) 61.0 54.3 50.1 45.1 .0001 Glasgow Coma Scale score at scene of Յ8 (%) 33.3 43.4 39.0 39.2 .13 Preexisting medical condition (%) 1.3 21.7 37.8 44.3 Ͻ.0001 Packed red blood cell transfusiona (%) 57.5 52.3 41.5 43.6 Ͻ.0001 Operative procedures (n; meanϮ SD) 5.7 Ϯ 4.4 6.0 Ϯ 5.5 5.6 Ϯ 7.2 6.0 Ϯ 6.7 .001 Laparotomy (%) 31.6 31.1 24.4 26.9 .031 No organ failure (%) 17.7 22.2 17.2 13.1 Ͻ.0001 Single-organ failure (%) 19.9 22.7 20.4 16.7 .014 Multiple-organ failure (%) 62.4 55.1 62.4 70.2 Ͻ.0001 Intensive care unit length of stay (days; mean Ϯ SD)26Ϯ 23 30 Ϯ 24 29 Ϯ 24 27 Ϯ 20 .012 In-hospital length of stay (days; mean Ϯ SD)51Ϯ 42 45 Ϯ 33 47 Ϯ 40 43 Ϯ 35 .14 30-day mortality (%) 15.4 16.6 17.6 14.5 .258 In-hospital mortality overall (%) 16.2 21.5 22.0 18.2 .054

AIS, Abbreviated Injury Scale. aPacked red blood cells transfused between emergency room arrival and intensive care unit admission. isting medical condition), type and severity of ered significant. Statistical analysis was 42 (Ϯ20) yrs and were injured substan- injury (ISS, body regions, open vs. closed frac- performed using standard statistical software tially (mean ISS ϭ 26 Ϯ 13). Detailed ture, etc.), early physiology (GCS, blood pres- (SPSS Version 15.0, SPSS, Chicago, IL). characteristics for patients with vs. pa- sure), and early treatment (operative proce- tients without sepsis are summarized in dures, laparotomy, transfusion). Additional Table 1. adjustment was performed for the trauma RESULTS year. Measures of discrimination (e.g., In both groups, patients were predom- Nagelkerke’s r2, area under the receiver oper- Univariate Analysis. From the total of inantly male with significantly more ating characteristic curve) were calculated. 29,829 eligible multiple trauma patients males among septic trauma patients Where appropriate, mortality and other in- during the entire study period, 3,042 (80.9% vs. 72.1%; p Ͻ .0001). Patients cidence rates are presented with 95% confi- (10.2%) developed sepsis during their with sepsis were injured more severely as dence intervals. A p value of Ͻ.05 was consid- hospital stay. Patients had a mean age of indicated by a higher ISS (33 vs. 25; p Ͻ

Crit Care Med 2011 Vol. 39, No. 4 623 .0001). Figure 1 depicts the association between the incidence of sepsis and trauma severity as reflected by stratifica- tion for ISS, in both male and female patients. Furthermore, the occurrence of sepsis during posttraumatic hospitaliza- tion was associated with more deranged physiologic and laboratory values during the initial trauma resuscitation. Patients developing trauma-associated sepsis had received significantly higher rates of transfusion of packed red blood cell units between emergency room arrival and ICU admission than nonseptic patients; fur- thermore, they underwent surgery more frequently. Figure 1. Incidence of sepsis in multiple trauma patients, grouped for male and female gender and Sepsis after trauma was significantly stratified for Injury Severity Score subgroups (n ϭ 29,829; 1993–2008); p Ͻ .0001. Error bars associated with prolonged ICU/hospital represent 95% confidence intervals. stay and with increased rates of single- organ failure (82.9% vs. 34.8%) and MOF (63.2% vs. 17.3%; both p Ͻ .0001). The organs that failed most frequently in pa- tients with sepsis were the cardiovascular system and the lungs (both 63.5%; Ta- ble 1). The overall hospital mortality to- taled 19.5% in the sepsis group vs. 12.5% in patients without sepsis (p Ͻ .0001). Stratification for Trauma Years. Table 2 displays the basic characteristics for the complete study population stratified into four subperiods according to the trauma year. Most changes observed from 1993 through 2008 were statistically signifi- Figure 2. Incidence of sepsis after multiple trauma from 1993 to 2008 (n ϭ 29,829); p Ͻ .0001. Linear cant, e.g., a decreasing rate of packed red regression analysis showed a decrease of Ϫ0.46% per year for sepsis incidence. Error bars represent blood cell transfusions and fewer laparot- 95% confidence intervals. omies. The incidence of sepsis declined until the third subperiod but then re- mained on a comparable level during the ble 4. The following independent risk zen plasma transfusion vs. no fresh fro- two latest subperiods (Fig. 2). Overall factors for the development of sepsis after zen plasma transfusion. mortality in all multiple trauma patients multiple trauma were calculated: male (Fig. 3, light gray columns) declined gender, age, preexisting medical condi- DISCUSSION from 1993 through 2008. Յ Table 3 lists the relevant data on the tion, GCS score of 8 at the scene, ISS, Incidence. To our knowledge, the cur- Ն subgroup of septic trauma patients cor- AISTHORAX score of 3, number of inju- rent analysis is the largest study on sepsis respondingly stratified into the four ries, number of red blood cell units trans- in severely injured patients. The only trauma year subperiods. As for the entire fused, number of operative procedures, comparable study, by Osborn et al (12), 2 ϭ study population, the incidence of MOF and laparotomy (r 0.124; area under included 30,303 trauma patients from significantly increased from 1993 until the receiver operating characteristic Pennsylvania over a 2-yr period; however, 2008, and over time, progressively fewer curve of 0.727; 95% confidence interval their study population was rather heter- sepsis patients presented without organ of 0.717–0.738). For the model, the per- ogeneous since the majority of the pa- failure. The development of mortality centage of correctly predicted cases was tients were injured only mildly (mean rates in septic trauma patients is also 89.4% and Cohen’s ␬ coefficient was 0.20 ISS ϭ 13 Ϯ 11) and therefore only 40% demonstrated in Figure 3 (dark gray col- if p(sepsis) of Ն20% was used as the were admitted to the ICU. Consequently, umns), showing no significant changes cutoff. The following variables were ex- the authors reported an incidence of regarding the complete study period. cluded by the model: systolic blood pres- posttraumatic sepsis of only 2%, which However, when the two latest subperiods sure at the scene of Յ90 mm Hg vs. Ͼ90 probably underestimates septic complica- were compared, the mortality rate de- mm Hg; systolic blood pressure at emer- tions in populations of critically injured creased significantly. gency room admission of Յ90 mm Hg vs. patients. Since in the present study only Ͼ Ն Predictors of Sepsis from Multivariate 90 mm Hg; AISHEAD score of 3; AIS- patients with a relevant trauma load in- Ն Analysis. A total of 19 variables were en- score of 3; AISEXTREMITIES dicated by an ISS of Ն9 were considered, tered into the stepwise logistic regres- score of Ն3; blunt vs. penetrating mech- the mean ISS was 26 Ϯ 13, with the sion. The final model is presented in Ta- anism; open vs. closed fracture; fresh fro- overall incidence of sepsis totaling 10.2%.

624 Crit Care Med 2011 Vol. 39, No. 4 over time; however, regarding this decade (2001–2008), the incidence rates re- mained constant. In contrast to our find- ings, a large epidemiologic study from the United States indicated that the inci- dence of sepsis increased progressively from 1979 until 2000 (2). The declining incidence of sepsis over the 16-yr study period in our analysis can partially be attributed to various nonspecific im- provements in over the past 2 decades, e.g., with respect to (21, 22) and weaning (23, 24). However, the impacts of these potential advances were neither investigated nor substantiated by our analysis since the respective data are not documented in the TR-DGU. With respect to packed red blood cell Figure 3. Mortality in all trauma patients (light gray columns;nϭ 29,829; p Ͻ .0001) and in the transfusion and operative procedures, we subgroup of trauma patients who developed sepsis (dark gray columns;nϭ 3042; p ϭ .054) from 1993 observed substantial changes regarding until 2008. In patients with sepsis, differences between two subsequent periods did not reach statistical these sepsis predictors from 1993 to ϭ significance except for the two latest subperiods, as depicted (p .026). Linear regression analyses 2008, particularly a dramatic reduction showed a slope of Ϫ0.433% per year for overall mortality and a slope of ϩ0.155% per year for sepsis in the packed red blood cell transfusion mortality. Error bars represent 95% confidence intervals. rate. Therefore, this observation repre- sents another possible reason for the de- Table 4. Independent risk factors for the development of sepsis after multiple trauma, derived from cline in sepsis incidence, as discussed a multivariate analysis using a stepwise logistic regression model with forward variable selection previously (25). Regression Odds Ratio (e␤; 95% Outcome. Previous reports have sug- Variable Entered Coefficient ␤ Confidence Interval) p gested a decrease in mortality after severe injury over recent years (3–5). However, Male gender (vs. female gender) 0.59 1.81 (1.61–2.03) Ͻ.0001 similar findings in our present study were Preexisting medical condition (vs. none) 0.48 1.62 (1.45–1.80) Ͻ.0001 not accompanied by improved outcomes Injury Severity Score (points) 0.01 1.01 (1.01–1.02) Ͻ.0001 Number of operative procedures in those patients with sepsis. Although 0 overall mortality from multiple injury de- 1 0.52 1.69 (1.37–2.07) Ͻ.0001 clined significantly from 1993 until 2008, Ն2 0.86 2.37 (1.98–2.83) Ͻ.0001 a we could not determine a significant de- Number of packed red blood cell units crease of mortality in the subgroup of 0 1–9 0.35 1.41 (1.27–1.58) Ͻ.0001 septic trauma patients regarding the en- Ն10 0.47 1.60 (1.37–1.88) Ͻ.0001 tire study period. Similarly, there was no Abbreviated Injury Scale (thorax) score of Ն3 (vs. Ͻ3) 0.26 1.30 (1.17–1.45) Ͻ.0001 significant change in the length of hos- Glasgow Coma Scale score of Յ8 at the scene (vs. Ͼ8) 0.28 1.32 (1.19–1.46) Ͻ.0001 pital stay among sepsis patients although Age (yrs) 0.01 1.01 (1.00–1.01) Ͻ.0001 Laparotomy (vs. no laparotomy) 0.16 1.18 (1.05–1.32) .007 demonstrated for the complete study Number of injuries 0.09 1.09 (1.07–1.11) Ͻ.0001 population. Interestingly, the occurrence Trauma years of MOF increased significantly in both 1993–1996 the entire cohort and the subgroup with Ϫ 1997–2000 0.28 0.76 (0.62–0.93) .007 sepsis, which has been reported by other 2001–2004 Ϫ0.50 0.61 (0.50–0.74) Ͻ.0001 2005–2008 Ϫ0.55 0.58 (0.48–0.70) Ͻ.0001 authors as well (2). These temporal Constant Ϫ4.63 changes with respect to organ failure, which is an indicator of the severity of Backward variable selection yielded the same results. The logistic regression model was started illness, are likely to have contributed to with 19 variables (see text). The standard category is included in parentheses behind the variable name the relatively static mortality rates and does not receive a coefficient in the model. Odds ratios have to be interpreted in relation to this among sepsis patients over the study pe- standard category. riod. Another potential explanation aPacked red blood cells transfused between emergency room arrival and intensive care unit admission. might be associated with the emergence of fungal pathogens and increasing mi- crobial resistance (2, 26). However, infor- This result is consistent with results from gation that found a sepsis rate of 12.4% mation on causative or a South African study on 450 severely among 454 surgical and medical ICUs the site of infection is not documented at injured patients (mean ISS ϭ 18 Ϯ 8) across Germany (20). all in the TR-DGU database, which repre- where 14.4% developed sepsis (19) or In the current study, the incidence of sents one major weakness of the current from a prospective, observational investi- postinjury sepsis decreased significantly investigation.

Crit Care Med 2011 Vol. 39, No. 4 625 Osborn et al (12) found a mortality of tween men and women (1, 2), which has mission laboratory values are docu- 23.1% in trauma patients with sepsis, been attributed to genetic (33), endocrine mented; further values that might be of which resembles the overall mortality (34), and socioeconomic (1, 2) aspects. Cor- interest (, C-reactive pro- rate of 19.5% in septic patients of the responding to prior studies, both age and tein) are not documented in our registry. current study considering the higher age preexisting medical conditions were also During the initial years, comorbidities as compared to that of our study popula- identified as independent predictors of sep- were registered very inconsistently, and tion (49 Ϯ 21 yrs vs. 44 Ϯ 19 yrs). In sis in our analysis (1, 7, 12). only their presence or absence was doc- 1995, Angus et al (1) calculated an annual Still, activation of the inflammatory umented. Thus, further information with mortality of 28.6% in a nontrauma pop- response is determined not only by these respect to specific diseases was not avail- ulation; however, only patients with se- individual host factors but also by the able. Accordingly, the occurrences of sep- vere sepsis were studied, and the mean magnitude of the initial trauma load (35), sis and MOF during trauma-associated age was 64 yrs. preferably quantified using the ISS as a hospitalization are documented only di- While general long-term data on sep- surrogate (15). Our present analysis sup- chotomously as either “yes” or “no” with- sis are available from the United States ports earlier findings that the ISS repre- out information on the respective day of showing a declining mortality in sepsis sents a potent predictor of infection and onset or the duration. Therefore, it can- patients (27.8% during the period of sepsis (7, 11, 12). However, only a few not be abstracted from the TR-DGU data 1979–1984 to 17.9% during 1995–2000) authors have investigated the influence of set whether the operative procedures oc- (2), further comparison remains difficult the injury pattern on the development of curred before or after the onset of sepsis. as no other trauma study exists assessing sepsis. First, we could demonstrate that In the interpretation of incidence data, it epidemiology of sepsis over a long study the incidence of sepsis significantly cor- should be considered that the sepsis def- period. We found a trend toward increas- relates with the number of injuries. Ac- inition according to Bone (16) has been ing mortality in sepsis patients in each cording to AIS scores, the anatomical re- applied throughout the complete study subsequent period, but interestingly, a gion with the highest predictive value for period. Just recently, the TR-DGU data significant decrease was observed when postinjury sepsis was the thorax. Chest set has been revised; thus, in the future the two latest subperiods were compared injuries have frequently been reported to sepsis is going to be defined according to (2001–2004 vs. 2005–2008). Certainly, contribute significantly to increased rates the recommended American College of the interpretation of outcome data must of acute respiratory distress syndrome, Chest Physicians/Society of Critical Care take into account that new insights from sepsis, and MOF (7, 36, 37). Accordingly, Medicine consensus criteria (43). As clinical trials have enhanced sepsis ther- the highest rate of organ failure in our mentioned above, the definition for organ apy over recent years (27–30). Accord- study was observed for the was updated for the TR-DGU par- ingly, as a key effort of the Surviving system (63.5%). ticipants in 2002, which might bias sta- Sepsis Campaign, evidence-based clinical Several authors have recognized allo- tistics related to the incidence of organ practice guidelines were developed and genic blood transfusion as an indepen- failure. The following aspects might be published in 2004 (31), and first analyses dent predictor of posttraumatic infection further potential sources for bias: The demonstrated that increasing compliance (7, 11) and systemic inflammatory re- mode of data entry changed in 2002 from with the guidelines was associated with sponse syndrome (6, 8). We observed sim- paper form to online registration via the improved outcomes (32). However, re- ilar results and a dose-dependent rela- Internet. A total of 7,466 patients with garding the role of these initiatives for tionship regarding posttraumatic sepsis, missing data were excluded from the the current study, the grade of adherence with massive transfusion (Ն10 red blood analysis. Participation at the TR-DGU is to sepsis guidelines among the participat- cell units) approaching a doubling of the voluntary; as estimated, only about 25% ing centers cannot be abstracted from risk. Potential underlying mechanisms of German multiple trauma patients are our database. However, it remains to be are associated with contaminating leuko- registered. There is a significant imbal- seen whether future advances in critical cytes and inflammatory mediators within ance in the patient numbers among the care will be accompanied by temporal packed red blood cell units (9, 10) and four subperiods, which reflects the pro- changes observable in our and other with the widespread depression of the im- gressively increasing number of trauma trauma registries. mune system through hemorrhagic centers participating in the registry since Sepsis Risk Factors. As the second ob- (6, 11, 37–39). 1993. The resulting unequal sample sizes jective, the present study was conducted As discussed above, we could assess a of the groups and subgroups studied to determine independent risk factors for decreasing number of operative proce- might bias the results. Also, death was sepsis after severe injury. In the investi- dures and fewer laparotomies from 1993 attributed to sepsis just on the basis of gation by Osborn et al (12), the following through 2008. This reflects the changing the mortality rates in patients diagnosed predictors of sepsis were identified: ISS, practice regarding splenic and hepatic in- with sepsis, since the direct cause of Revised Trauma Score, GCS score, co- juries with a more frequent use of non- death is not documented in the TR-DGU. morbidities, and male gender. However, operative treatment strategies, which It must be considered that information information on transfusion requirement, have been associated with lower rates of on “do not resuscitate” decisions is not operative procedures, or organ failure abdominal infection and sepsis in previ- registered as well. Finally, we conducted was not documented in their database. ous reports as well (7, 40, 41). Multiple a retrospective analysis, and only associ- Regarding the multivariate analysis of surgical procedures have been described ations are reported that do not necessar- the current study, we detected male gender to affect the as “second ily represent causal relationships. as a significant predictor for sepsis as well, hits” (7, 11, 38, 42). Despite these shortcomings, the pres- and large epidemiologic studies have al- Several limitations must be consid- ent analysis remains the largest study on ready demonstrated the same disparity be- ered regarding this study. Only basic ad- sepsis epidemiology in critically injured

626 Crit Care Med 2011 Vol. 39, No. 4 patients and is derived from a represen- sion in critically injured patients: A prospec- 21. Amato MB, Barbas CS, Medeiros DM, et al: tative database with respect to the gen- tive study. Injury 2006; 37:455–465 Effect of a protective-ventilation strategy on eral German trauma population. Addi- 7. Croce MA, Fabian TC, Waddle-Smith L, et al: mortality in the acute respiratory distress tionally, early recognition of the Identification of early predictors for post- syndrome. N Engl J Med 1998; 338:347–354 identified risk factors for sepsis may allow traumatic . Am Surg 2001; 67: 22. Drakulovic MB, Torres A, Bauer TT, et al: 105–110 Supine body position as a for prognostic considerations, guide early di- 8. Dunne JR, Malone DL, Tracy JK, et al: Allo- nosocomial pneumonia in mechanically ven- agnostic workup, and help to reduce sep- genic blood transfusion in the first 24 hours tilated patients: A randomised trial. Lancet tic complications after multiple trauma. after trauma is associated with increased sys- 1999; 354:1851–1858 temic inflammatory response syndrome 23. Ely EW, Baker AM, Dunagan DP, et al: Effect CONCLUSIONS (SIRS) and death. Surg Infect (Larchmt) on the duration of mechanical ventilation of 2004; 5:395–404 identifying patients capable of breathing Although the incidence of posttrau- 9. Flohe´ S, Kobbe P, Nast-Kolb D: Immunolog- spontaneously. N Engl J Med 1996; 335: matic sepsis decreased from 1993 until ical reactions secondary to blood transfusion. 1864–1869 2008, the general outcome improvement Injury 2007; 38:1405–1408 24. Kress JP, Pohlman AS, O’Connor MF, et al: in multiply injured patients seen over re- 10. He´bert PC, Fergusson D, Blajchman MA, et Daily interruption of infusions in cent years has not been accompanied by al: Clinical outcomes following institution of critically ill patients undergoing mechanical significant changes in the subgroup of the Canadian universal leukoreduction pro- ventilation. N Engl J Med 2000; 342: patients with sepsis. Therefore, sepsis af- gram for red blood cell transfusions. JAMA 1471–1477 2003; 289:1941–1949 25. Maegele M, Lefering R, Paffrath T, et al: ter severe trauma remains a substantial 11. Papia G, McLellan BA, El-Helou P, et al: Changes in transfusion practice in multiple challenge, and further efforts are neces- Infection in hospitalized trauma patients: In- injury between 1993 and 2006: A retrospec- sary to reduce morbidity and mortality. cidence, risk factors, and complications. tive analysis on 5389 patients from the Ger- Predictors for sepsis after severe trauma J Trauma 1999; 47:923–927 man Trauma Registry. Transfus Med 2009; derived from multivariate analysis were 12. Osborn TM, Tracy JK, Dunne JR, et al: Epi- 19:117–124 male gender, age, preexisting medical demiology of sepsis in patients with trau- 26. Climo MW, Sepkowitz KA, Zuccotti G, et al: condition, GCS score of Յ8 at the scene, matic injury. Crit Care Med 2004; 32: The effect of daily bathing with chlorhexidine Ն 2234–2240 on the acquisition of methicillin-resistant ISS, AISTHORAX score of 3, number of injuries, number of red blood cell units 13. Scoring study committee of the German So- aureus, -resis- transfused, number of operative proce- ciety of Trauma Surgery: Trauma register of tant Enterococcus, and healthcare-associ- dures, and laparotomy. Identification of the German Society of Trauma Surgery. Un- ated bloodstream : Results of a fallchirurgie 1994; 97:230–237 quasi-experimental multicenter trial. Crit injured patients at risk for sepsis at an 14. Association for the Advancement of Automo- Care Med 2009; 37:1858–1865 early stage may help to decrease this tive Medicine: The Abbreviated Injury Scale 27. Annane D, Se´bille V, Charpentier C, et al: deadly complication in trauma popula- (AIS). 1990 Revision. Barrington, IL, Associ- Effect of treatment with low doses of hydro- tions. ation for the Advancement of Automotive cortisone and fludrocortisone on mortality in Medicine, 1990 patients with . JAMA 2002; 288: ACKNOWLEDGMENTS 15. Baker SP, O’Neill B, Haddon W Jr, et al: The 862–871 injury severity score: A method for describ- 28. Bernard GR, Vincent JL, Laterre PF, et al: We thank Dr. Thomas Ostermann for ing patients with multiple injuries and eval- Efficacy and safety of recombinant human statistical review. uating emergency care. J Trauma 1974; 14: activated for severe sepsis. N Engl 187–196 J Med 2001; 344:699–709 REFERENCES 16. Bone RC: Let’s agree on terminology: Defi- 29. Elke G, Scha¨dler D, Engel C, et al: Current nitions of sepsis. Crit Care Med 1991; 19: practice in nutritional support and its asso- 1. Angus DC, Linde-Zwirble WT, Lidicker J, et 973–976 ciation with mortality in septic patients— al: Epidemiology of severe sepsis in the 17. Goris RJ, te Boekhorst TP, Nuytinck JK, et al: Results from a national, prospective, multi- United States: Analysis of incidence, out- Multiple-organ failure. Generalized autode- center study. Crit Care Med 2008; 36: come, and associated costs of care. Crit Care structive inflammation? Arch Surg 1985; 1762–1767 Med 2001; 29:1303–1310 120:1109–1115 30. Rivers E, Nguyen B, Havstad S, et al: Early 2. Martin GS, Mannino DM, Eaton S, et al: The 18. Vincent JL, de Mendonc¸a A, Cantraine F, et goal-directed therapy in the treatment of se- epidemiology of sepsis in the United States al: Use of the SOFA score to assess the inci- vere sepsis and septic shock. N Engl J Med from 1979 through 2000. N Engl J Med 2003; dence of /failure in inten- 2001; 345:1368–1377 348:1546–1554 sive care units: Results of a multicenter, pro- 31. Dellinger RP, Carlet JM, Masur H, et al: Sur- 3. Davenport RA, Tai N, West A, et al: A major spective study. Working Group on “Sepsis- viving Sepsis Campaign guidelines for man- trauma centre is a specialty hospital not a Related Problems” of the European Society agement of severe sepsis and septic shock. hospital of specialties. Br J Surg 2010; 97: of Intensive Care Medicine. Crit Care Med Crit Care Med 2004; 32:858–873 109–117 1998; 26:1793–1800 32. Levy MM, Dellinger RP, Townsend SR, et al: 4. Papa L, Langland-Orban B, Kallenborn C, et 19. Muckart DJ, Bhagwanjee S: American Col- The Surviving Sepsis Campaign: Results of al: Assessing effectiveness of a mature lege of Chest Physicians/Society of Critical an international guideline-based perfor- trauma system: Association of trauma center Care Medicine Consensus Conference defini- mance improvement program targeting se- presence with lower injury mortality rate. tions of the systemic inflammatory response vere sepsis. Crit Care Med 2010; 38:367–374 J Trauma 2006; 61:261–266; discussion syndrome and allied disorders in relation to 33. Hubacek JA, Stu¨ ber F, Fro¨hlich D, et al: Gene 266–267 critically injured patients. Crit Care Med variants of the bactericidal/permeability in- 5. Ruchholtz S, Lefering R, Paffrath T, et al: 1997; 25:1789–1795 creasing protein and Reduction in mortality of severely injured 20. Engel C, Brunkhorst FM, Bone HG, et al: binding protein in sepsis patients: Gender- patients in Germany. Dtsch Arztebl Int 2008; Epidemiology of sepsis in Germany: Results specific genetic predisposition to sepsis. Crit 105:225–231 from a national prospective multicenter Care Med 2001; 29:557–561 6. Beale E, Zhu J, Chan L, et al: Blood transfu- study. Intensive Care Med 2007; 33:606–618 34. Kno¨ferl MW, Angele MK, Diodato MD, et al:

Crit Care Med 2011 Vol. 39, No. 4 627 Female sex hormones regulate and hemorrhagic shock. Crit Care Med 1998; Blunt hepatic injury: A paradigm shift from function after trauma-hemorrhage and pre- 26:1372–1378 operative to nonoperative management vent increased death rate from subsequent 38. de Haan JJ, Lubbers T, Hadfoune M, et al: in the 1990s. Ann Surg 2000; 231:804– sepsis. Ann Surg 2002; 235:105–112 Postshock intervention with high-lipid en- 813 35. Sauaia A, Moore FA, Moore EE, et al: Early teral nutrition reduces inflammation and tis- 42. Pape HC, Giannoudis P, Krettek C: The risk factors for postinjury multiple organ sue damage. Ann Surg 2008; 248:842–848 timing of fracture treatment in failure. World J Surg 1996; 20:392–400 39. Napolitano LM, Kurek S, Luchette FA, et al: patients: Relevance of damage control or- 36. Moore FA, Moore EE, Read RA: Postinjury Clinical practice guideline: Red blood cell thopedic surgery. Am J Surg 2002; 183: multiple organ failure: Role of extratho- transfusion in adult trauma and critical care. 622–629 racic injury and sepsis in adult respiratory Crit Care Med 2009; 37:3124–3157 43. American College of Chest Physicians/ distress syndrome. New Horiz 1993; 40. Gauer JM, Gerber-Paulet S, Seiler C, et al: Society of Critical Care Medicine Consensus 1:538–549 Twenty years of splenic preservation in trau- Conference: Definitions for sepsis and organ 37. Wichmann MW, Ayala A, Chaudry IH: Severe ma: Lower early infection rate than in sple- failure and guidelines for the use of innova- depression of host immune functions follow- nectomy. World J Surg 2008; 32:2730–2735 tive therapies in sepsis. Crit Care Med 1992; ing closed-, soft-tissue trauma, 41. Malhotra AK, Fabian TC, Croce MA, et al: 20:864–874

628 Crit Care Med 2011 Vol. 39, No. 4