Different patient case mix by applying the 2003 SCCM/ESICM/ACCP/ATS/SIS sepsis definitions instead of the 1992 ACCP/SCCM sepsis definitions in surgical patients: a retrospective observational study

Background Revised consensus sepsis definitions have been published in 2003. The present study was performed to compare the prevalence of different stages of sepsis and ICU mortality rates and find out the case mix within the same collective of postoperative/posttraumatic patients applying either the original 1992 ACCP/SCCM or the revised 2003 SCCM/ESICM/ACCP/ATS/SIS sepsis definitions. Methods Retrospective observational single-centre study in surgical critically ill patients admitted to an University adult ICU. From 01/2007 to 12/2007, 742 patients were surveyed daily computer-assisted with respect to different stages of sepsis. Results Within the same patient collective, applying the 2003 definitions instead of the 1992 definitions, prevalence of severe sepsis (61 vs. 56) and septic shock (205 vs. 162) was higher (p < 0.001). In patients with septic shock according to the 2003 definitions, mortality rate of 22% was lower than that of 27%, when the 1992 definitions were used. Risk of death was increased for those patients classified to be in septic shock with any of the definitions (OR 6.5, p = 0.001). Sensitivity to predict deaths was slightly higher with the 2003 definitions (92%) than with the 1992 definitions (88%), and specificity was lower (31% vs. 49%). Conclusion The prevalence and mortality rates of various sepsis severity stages differ if defined by the 1992 or the 2003 definitions. Thus, transferring conclusions drawn from data sets regarding severity of sepsis generated with the 1992 definitions to the same population applying the 2003 definitions may be misleading. The 1992 definitions may under-classify patients with severe sepsis.


Background
Prior to 1987, none of the sepsis studies utilized standards for inclusion criteria [1]. Since the 1992 American College of Chest Physicians/Society of Critical Care Medicine (ACCP/SCCM) consensus conference on sepsis definitions [2], predefined sepsis criteria for patient enrolment in clinical trials and markers of organ dysfunction have been increasingly applied [1]. Comparing 176 trials from 1993 to 2001 after the ACCP/SCCM consensus conference with 57 trials from 1976 to 1992, revealed increasing use of standards for inclusion criteria in 65% vs. 11%, respectively, in most cases (69%) the ACCP/SCCM consensus conference definitions [1]. Uniformity of inclusion criteria is a prerequisite for comparability of sepsis studies. In 2001, the original 1992 ACCP/SCCM sepsis definitions [2] were modified to reflect the extended understanding of the pathophysiology of the systemic inflammatory response syndrome (SIRS) and sepsis syndrome at that time. The result was the 2003 revised Society of Critical Care Medicine/European Society of Critical Care Medicine/American College of Chest Physicians/American Thoracic Society, Surgical Infection Society (SCCM/ ESICM/ACCP/ATS/SIS) sepsis definition [3,4]. However, due to its complexity, the latter definition is not widely used, and papers and current studies still apply to the 1992 definitions.
In the 2003 sepsis definitions, it was stated that the diagnostic criteria for infection or systemic inflammation should be broadly useful for both clinicians caring for patients at the bedside and researchers designing observational studies and clinical trials to improve the understanding of sepsis and its treatment. An extended list of possible signs of the systemic response was included in the 2003 definitions to better reflect the reality at the bedside, especially, of how physicians diagnose sepsis in practice, regarding general, inflammatory, hemodynamic, organ dysfunction, and tissue perfusion variables. In Table 1, the diagnostic criteria for sepsis and organ dysfunction variables in the 1992 and 2003 definitions are specified and compared. Cut-off values changed from the 1992 definitions to the 2003 definitions, such as fever (from > 38.0°C to > 38.3°C) and tachypnea (from > 20 to > 30 breaths/min). Furthermore, in the 1990s, it was recommended to define organ failure by applying the Sequential Organ Failure Assessment (SOFA) score with greater than two points in one organ system [5]. In contrast to the 1992 definitions, the 2003 definitions provide clear cut-off values for organ dysfunctions. The use of the word "some" of the diagnostic criteria in the 2003 definitions instead of "two or more out of four" (temperature, heart rate, respiratory rate, and white blood cell count (WBC)) in the 1992 definitions, was introduced to facilitate a clinical diagnosis at the bedside. Once severe sepsis and septic shock are diagnosed, this may lead to far-reach-ing consequences since extensive management and treatment guidelines have been published for this issue [6,7].
However, it remains unclear how the changes in the definitions influence the prevalence and mortality rate of SIRS and sepsis in critically ill surgical patients. The 2003 sepsis definitions were updated for clinicians and researchers to facilitate a clinical diagnosis and to improve the understanding, diagnosis and treatment of sepsis. Therefore, the paper will focus on patients with sepsis. The present study was performed to determine how the prevalence of various stages of sepsis and mortality rates differs in the same patient population, if defined by either the original or the revised sepsis criteria. Moreover, it should be addressed whether a switch to the 2003 definitions might increase the number of patients who are expected to benefit from earlier and more appropriate critical care management.

Patients and data collection
The study is in compliance with the Helsinki declaration and was approved by the ethics committee of the University Hospital Ulm. The ethics committee waived informed consent due to the fact that for this study no blood was drawn and no intervention was performed. Thus, no informed consent was obtained from the patients. Patients were admitted to the Anaesthesiolgical ICU of the University Hospital Ulm after major trauma, great vessel, lung, brain or abdominal surgery. All surgical patients admitted to this ICU are routinely computer-assisted surveyed for sepsis, severity of disease (Simplified Acute Physiology Score (SAPS) II) [8] and organ dysfunctions (Sequential Organ Failure Assessment (SOFA) score) [5] on a daily basis by the ICU residents and staff physicians. To improve accuracy and to minimize inconsistencies in medical chart reviews, all residents and staff physicians were trained in the charts before initiation of the study. In addition, cases were selected, variables were defined, and data were entered on a daily base in standardized electronic case report forms leading through the different organ systems and relevant infection parameters. After data entry of different organ systems and infection parameters, the physicians directly received the results of the actual scores, the sepsis/SIRS classification, and, also, a longitudinal overview regarding the whole ICU course. Then, the residents and staff physicians checked, corrected, ascertained and re-ascertained the data and classifications. Monitoring and meetings were performed daily and inter-rater agreement tested. Charts were checked and corrected by the staff physicians before demission of the patients from the ICU and before final evaluation. Data of postoperative/posttraumatic patients admitted to our ICU during a one year period from 01-JAN-2007 until 31-DEC-2007 were analysed.

Definitions
Sepsis was defined using the original 1992 ACCP/SCCM [2] and the revised 2003 SCCM/ESICM/ACCP/ATS/SIS sepsis definitions [3,4]. In the present study, due to the 1992 definitions, a systemic inflammatory response syndrome (SIRS) was manifested in patients by two or more of the four conditions: temperature, heart rate, respiratory rate, and white blood cell count (WBC) ( Table 1). If SIRS was due to a documented infection, patients were classified as sepsis patients. In the present study, applying the 2003 sepsis definitions, SIRS and sepsis was classified as manifested if two or more of eight conditions given in the broadened diagnostic criteria list of possible signs of systemic inflammation were present. These criteria include general variables such as temperature, heart rate, respira-tory rate, altered mental status (not applied in the present study), significant edema or positive fluid balance, hyperglycemia, as well as inflammatory variables such as white blood cell count (WBC), plasma C-reactive protein and plasma procalcitonin (not applied in the present study). Severe sepsis was defined as sepsis plus organ dysfunction. Septic shock was defined as severe sepsis plus shock. Severity of sepsis is proposed to increase firstly by association with organ dsyfunctions and secondly by additional shock. As recommended [5], in the present study, applying the 1992 sepsis definitions, organ failure was regarded to be present if patients had lactic acidosis or oliguria, or reached greater than two points in one organ system (lung, coagulation, liver, kidney) using the SOFA score. Greater than two points are reached in the SOFA score for Sensitivity of predicting deaths in patients with documented infections was slightly higher using the 2003 definitions than the 1992 definitions (92% vs. 88%) while specificity was lower (31% vs. 49%) ( Table 3).

Patients with a high risk of death
As shown in Table 2, patients with shock were at a high risk of death.

Discussion
The present study mainly proposes that conclusions drawn from data where severity of sepsis is analysed by the 1992 or the 2003 sets of definitions may vary and have   To assess potential clinical consequences for surgical patients, at least three aspects (diagnosis, therapy and clinical studies) should be taken into account.
In the revised 2003 definitions, facilitating a bedside diagnosis of sepsis had primacy over research entry criteria [3,4]. The revised 2003 SCCM/ESICM/ACCP/ATS/SIS sepsis definitions were constructed to detect more cases of SIRS/sepsis. As shown in the present paper, this aim has been achieved, since out of the 742 surgical patients, the numbers of no SIRS/no sepsis patients have been decreased by 119 patients with the 2003 definitions.
Using the revised 2003 definitions, the increase in prevalence of sepsis was most pronounced in the more severe stages, i.e. in severe sepsis and septic shock (Table 2). Thereby, within the same population, about 27% more surgical patients were allocated to septic shock than they would have been using the 1992 definitions. Thus, applying the 2003 definitions instead of the 1992 definitions will create a case mix with more patients with sepsis, especially classified as more severely ill, i.e. with severe sepsis and septic shock. When comparing classification before vs. after the ACCP/SCCM conference, 16% (n = 57 studies) vs. 3% (n = 119 studies) explicitly required blood culture positivity as an inclusion criterion, respectively [1]. Only 49% vs. 68% of the studies before vs. after the ACCP/SCCM conference incorporated markers of organ dysfunctions into the inclusion criteria [1]. This indicates a trend to prefer criteria of host response (and organ dysfunction) vs. infection diagnostics for the definition of sepsis. Using the 1992 definitions, a threshold of greater than two points in one organ system in the SOFA score has been recommeded to assign organ dysfunction [5].
The higher prevalence of more severely ill classified patients with the 2003 definitions in the present study is due to a lower threshold to define organ dysfunction (in comparison equal to or less than two points in the SOFA score) with the 2003 definitions (Table 1).
Regarding therapeutic approaches, more surgical patients with septic shock are expected to be recommended for potential immunomodulatory treatment using the 2003 definitions. In 2004 and 2008, the "Surviving Sepsis Campaign" guidelines for the management of severe sepsis and septic shock were published to improve outcome [6,7]. Based on the increased prevalence of severe sepsis and septic shock in the present study, more patients may be applicable to and may benefit from the "Surviving Sepsis Campaign" management guidelines. On the other hand, this may lead to more undesirable consequences, such as harm to patients, more burden on staff and patients, and greater costs. One of the evidence based recommendations suggests the application of recombinant human activated protein C (rhAPC) in patients at high risk of death (Acute Physiology and Chronic Health Evaluation II, APACHE II >= 25 and septic shock, as defined by the 1992 definitions), since rhAPC has been shown to reduce mortality in these patients by 13% [6,7]. Applying the 2003 definitions, in our ICU, within one year, 43 more surgical patients, i.e. about 21% more patients, would have been potential candidates for this expensive and potentially harmful (risk of bleeding) therapy. Moreover, their mortality rate of 22%, in any case, would have been markedly lower than with the 1992 definitions (Table 2). Thus, it may be necessary to retrospectively or prospectively revaluate whether treatment recommendations based on the 1992 definitions hold true, especially, if the same population is classified with the 2003 definitions.
With respect to inclusion criteria for sepsis trials, the present study revealed that more surgical patients with septic shock, having a lower mortality rate, would have been enrolled using the 2003 definitions (Table 2). Thus, a properly powered study with the 2003 definitions would have to be larger, in order to demonstrate an equivalent mortality reduction (by e.g. 10%) compared with the same population classified by the 1992 definitions in which the mortality rate is higher.
Since sepsis definitions lack specificity and sensitivity, efforts have been made [3,4,9] and will be mandatory in the future to improve diagnosis and valid classification of patients to reduce variability within treatment groups, and to increase the probability of generating convincing results. In the present study in critically ill surgical patients, prediction of death by the 2003 definitions occurred with slightly higher sensitivity and lower specificity than by the 1992 definitions.  [3,4]. Moreover, to increase uniformity in patient enrolment, comprehensive demographic data including severity scores, such as the SAPS II [8] or SAPS 3 [11] to assess severity of illness and predict vital status at hospital discharge, will be required.
Regarding divergent classifications of patients, the present study shows that 309/742 patients were under-classified with the 1992 definitions, but 21, only, with the 2003 definitions (Table 4). Also, in the patients with severe SIRS and sepsis, and shock, markedly more patients were under-classified with the 1992 definitions as compared to the 2003 definitions. Especially these patients with high risk of death under-classified by the 1992 definitions are those who might benefit from earlier and more appropriate critical care management, if, in addition, detected by the 2003 definitions.

Conclusion
The present study shows that the definition of sepsis criteria strongly influences prevalence and mortality rates of sepsis. Use of the 1992 definitions may lead to under-classification of patients with severe sepsis at high risk of death with the consequence of delayed or lack of timely and appropriate critical care management. Thus, the present study underlines the need for a widespread use of common sepsis definitions to facilitate comparability, diagnosis, treatment recommendations, and enrolment strategies for clinical trials of surgical patients with sepsis. Due to the fact that only surgical patients have been enrolled in the present study, it has to be clarified whether the results are applicable to medical cases, also.

Key messages
Applying the 2003 SCCM/ESICM/ACCP/ATS/SIS sepsis definitions instead of the 1992 ACCP/SCCM sepsis definitions in surgical patients resulted in different patient case mix with under-classification with the 1992 definitions.
Within the same surgical patient collective, applying the 2003 definitions instead of the 1992 definitions, prevalence of severe sepsis and septic shock was higher.
In surgical patients with septic shock, mortality rate was lower using the 2003 definitions than with the 1992 definitions.
Risk of death was increased for those surgical patients classified to be in septic shock in both definitions compared to those classified to be without septic shock in both definitions (OR 6.5, p = 0.001).
The 2003 definitions predicted fatal outcome in surgical patients with slightly higher sensitivity and lower specificity than the 1992 definitions.