Evaluation of Definitions of Sepsis

Abstract

To the EditorWe appreciate the comments of Dr. Pilz and his colleagues in regard to the new definition of SIRS and its comparison with the traditional definition of sepsis syndrome. First, we wish to emphasize that both SIRS and sepsis syndrome are syndromes, that is, constellations of specific signs and symptoms. They are not alternative diagnoses for sepsis and are to be used only with a specific diagnosis.1Bone RC Balk RA Cerra FB Dellinger RP Fein AM Knaus WA et al.Definitions for sepsis and organ failure and guidelines for the use of innovative therapies in sepsis.Chest. 1992; 101: 1644-1655Abstract Full Text Full Text PDF PubMed Scopus (7440) Google Scholar For example, sepsis syndrome has always been defined as a set of physiologic findings associated with infection (ie, a diagnosis compatible with infection).2Bone RC Fisher CJ Clemmer TP Slotman GJ Metz CA Balk RA. Sepsis syndrome: a valid clinical entity.Crit Care Med. 1989; 17: 389-393Crossref PubMed Scopus (592) Google Scholar When this was done in a clinical trial situation, use of the rubric “sepsis svndrome” identified a group of patients with infection with a mean hospital mortality rate of approximately 40 percent.3Bone RC. Toward an epidemiology and natural history of SIRS.JAMA. 1992; 268: 3452-3455Crossref PubMed Scopus (425) Google Scholar Use of the sepsis syndrome definition improved the precision and uniformity of many recent clinical trials involving critically ill patients with sepsis.Like all syndrome definitions, however, the sepsis syndrome definition uses categorical cutoffs (eg, temperature > 38° or < 36°C) to identify patients. Many patients may fail to meet these exact cutoffs and still may be quite similar to included patients. As our ability to measure and estimate patient risk improved, it was also recognized that use of the sepsis syndrome definition may have identified groups of patients with a relatively uniform group mortality rate but with a very wide distribution of individual patient risk.4Knaus WA Sun X Nystrom PO Wagner DP. Evaluation of definitions for sepsis.Chest. 1992; 101: 1656-1662Abstract Full Text Full Text PDF PubMed Scopus (120) Google Scholar These concepts led to the suggestion of a less restrictive syndrome definition, such as that of SIRS, that could be combined with more precise risk prediction techniques and diagnostic labels.3Bone RC. Toward an epidemiology and natural history of SIRS.JAMA. 1992; 268: 3452-3455Crossref PubMed Scopus (425) Google Scholar This could expand the number of patients eligible for clinical investigation while increasing the precision of description for each individual patient through use of either a severity score or risk prediction. These would be measured in a continuous, rather than a categorical, manner.In regard to patients admitted to ICUs with sepsis, the use of the SIRS definition, provided it is combined with a severity measure, appears to accomplish these objectives well. Pilz and colleagues ask how it applies to other conditions. To answer this, we applied SIRS criteria to all 17,440 ICU admissions in the original APACHE III database (Tables 1, 2). Because SIRS was designed to be applied only to patients with diagnoses compatible with infection or inflammation, this application is artificial. Nevertheless, the results in Table 1 show that of patients with specific diagnoses of sepsis, 92 percent met the criteria for SIRS. Of patients at high risk of infection (eg, those with ARDS), 77 percent met the definition. Likewise, 71 percent of patients with inflammation (eg, due to trauma or pancreatitis) also met the criteria for SIRS. Even in the category of patients at low risk of infection, 57 percent met the SIRS criteria. In all of these categories, patients who met the criteria for SIRS had higher hospital mortality rates and a greater incidence of infections as measured by discharge diagnoses.Table 1Profile of Infections in 17,440 ICU Admissions From APACHE III DatabasePresence of Infection, % †Documented infections in ICD-9 coding system.Patient Subgroup (%)*Definitions for each subgroup are shown in Table 2. Values in parentheses are percentage of subgroup or subsubgroup. SIRS = met criteria for diagnosis of SIRS; Not SIRS = did not meet criteria for diagnosis of SIRS.Mortality Rate,SepticemiaOther InfectionsInfection (10)32Admission sepsis (31)48SIRS (92)4847Not SIRS (8)Other infections (69)25SIRS (77)2714Not SIRS (23)1810High risk of infection (22)24SIRS (71)28828Not SIRS (29)16319Low risk of infection (59)14SIRS (57)18422Not SIRS (43)8215Inflammation (10)10SIRS (71)11215Not SIRS (29)60.810* Definitions for each subgroup are shown in Table 2. Values in parentheses are percentage of subgroup or subsubgroup. SIRS = met criteria for diagnosis of SIRS; Not SIRS = did not meet criteria for diagnosis of SIRS.† Documented infections in ICD-9 coding system. Open table in a new tab Table 2Classification of ICU Admisssion Diagnoses From APACHE III Database*GI = gastrointestinal; S/P = status post.InfectionSepsisBacterial pneumoniaBacterial meningitisCholangitis/cholecystitisRenal abscessPeritonitis, GI perforation/ruptureCellulitisHigh risk of infectionViral/fungal/parasitic infectionsPulmonary congestionShockS/P resuscitationComaHematologic disordersS/P emergency surgeryS/P GI/propharynx surgeryGI obstruction/vascular insufficiencyS/P vascular surgery with graftS/P transplantationViral/fungal/parasitic pneumoniaARDS, COPD, asthma, aspirationHemorrhagic/cardiogenic shockS/P cardiac/respiratory arrestNontraumatic comaNeutropenia, coagulopathyS/P emergency abdominal aortic aneurysmectomyS/P GI neoplasmGI paralysisFemoral-popliteal bypassS/P renal transplantationAscitesLow risk of infectionElective nonabdominal surgeryCardiovascular diseaseNonoperative neoplasmsNeurologic diseaseMetabolic/endocrinologic diseaseElective thoracic surgeryAcute myocardial infarctionCongestive heart failureRhythm disturbanceThyroid neoplasmMyasthenia gravis seizuresDiabetic ketoacidosis myxedemaGI bleeding drug overdoseInflammationTraumaFracturePancreatitisGI inflammatory disease* GI = gastrointestinal; S/P = status post. Open table in a new tab A similar message is illustrated in Table 3, which shows the proportion of patients with one type of infection (pneumonia), one inflammatory condition (head trauma), and a diagnosis associated with low risk of infection (congestive heart failure) who met SIRS criteria. The pneumonia and head trauma categories contain the most patients with SIRS, but even in the congestive heart failure subgroup, half of the patients met the SIRS criteria.Table 3SIRS Classification Status in Subgroups of ICU Admissions*SIRS = met criteria for diagnosis of SIRS; Not SIRS = did not meet criteria for diagnosis of SIRS.SubgroupProportion of Subgroup, %Mortality Rate, %PneumoniaSIRS8341Not SIRS1730Head traumaSIRS7120Not SIRS297Congestive heart failureSIRS5329Not SIRS4716* SIRS = met criteria for diagnosis of SIRS; Not SIRS = did not meet criteria for diagnosis of SIRS. Open table in a new tab We believe these results suggest that the SIRS definition is meeting its goal well, provided it is used as intended. Use of the SIRS definition identifies the majority of patients with clinical diagnoses of infection or inflammation (more than would qualify using the sepsis syndrome criteria) while still identifying patients at a mortality risk higher than average. If combined with a patient-specific risk score or prediction, the SIRS definition may then be able to meet another of its goals— increasing the number of patients eligible for clinical evaluation without sacrificing precision. Because many patients without infection or inflammation also meet criteria for SIRS, however, it must only be used with a precise clinical diagnosis. To the EditorWe appreciate the comments of Dr. Pilz and his colleagues in regard to the new definition of SIRS and its comparison with the traditional definition of sepsis syndrome. First, we wish to emphasize that both SIRS and sepsis syndrome are syndromes, that is, constellations of specific signs and symptoms. They are not alternative diagnoses for sepsis and are to be used only with a specific diagnosis.1Bone RC Balk RA Cerra FB Dellinger RP Fein AM Knaus WA et al.Definitions for sepsis and organ failure and guidelines for the use of innovative therapies in sepsis.Chest. 1992; 101: 1644-1655Abstract Full Text Full Text PDF PubMed Scopus (7440) Google Scholar For example, sepsis syndrome has always been defined as a set of physiologic findings associated with infection (ie, a diagnosis compatible with infection).2Bone RC Fisher CJ Clemmer TP Slotman GJ Metz CA Balk RA. Sepsis syndrome: a valid clinical entity.Crit Care Med. 1989; 17: 389-393Crossref PubMed Scopus (592) Google Scholar When this was done in a clinical trial situation, use of the rubric “sepsis svndrome” identified a group of patients with infection with a mean hospital mortality rate of approximately 40 percent.3Bone RC. Toward an epidemiology and natural history of SIRS.JAMA. 1992; 268: 3452-3455Crossref PubMed Scopus (425) Google Scholar Use of the sepsis syndrome definition improved the precision and uniformity of many recent clinical trials involving critically ill patients with sepsis.Like all syndrome definitions, however, the sepsis syndrome definition uses categorical cutoffs (eg, temperature > 38° or < 36°C) to identify patients. Many patients may fail to meet these exact cutoffs and still may be quite similar to included patients. As our ability to measure and estimate patient risk improved, it was also recognized that use of the sepsis syndrome definition may have identified groups of patients with a relatively uniform group mortality rate but with a very wide distribution of individual patient risk.4Knaus WA Sun X Nystrom PO Wagner DP. Evaluation of definitions for sepsis.Chest. 1992; 101: 1656-1662Abstract Full Text Full Text PDF PubMed Scopus (120) Google Scholar These concepts led to the suggestion of a less restrictive syndrome definition, such as that of SIRS, that could be combined with more precise risk prediction techniques and diagnostic labels.3Bone RC. Toward an epidemiology and natural history of SIRS.JAMA. 1992; 268: 3452-3455Crossref PubMed Scopus (425) Google Scholar This could expand the number of patients eligible for clinical investigation while increasing the precision of description for each individual patient through use of either a severity score or risk prediction. These would be measured in a continuous, rather than a categorical, manner.In regard to patients admitted to ICUs with sepsis, the use of the SIRS definition, provided it is combined with a severity measure, appears to accomplish these objectives well. Pilz and colleagues ask how it applies to other conditions. To answer this, we applied SIRS criteria to all 17,440 ICU admissions in the original APACHE III database (Tables 1, 2). Because SIRS was designed to be applied only to patients with diagnoses compatible with infection or inflammation, this application is artificial. Nevertheless, the results in Table 1 show that of patients with specific diagnoses of sepsis, 92 percent met the criteria for SIRS. Of patients at high risk of infection (eg, those with ARDS), 77 percent met the definition. Likewise, 71 percent of patients with inflammation (eg, due to trauma or pancreatitis) also met the criteria for SIRS. Even in the category of patients at low risk of infection, 57 percent met the SIRS criteria. In all of these categories, patients who met the criteria for SIRS had higher hospital mortality rates and a greater incidence of infections as measured by discharge diagnoses.Table 1Profile of Infections in 17,440 ICU Admissions From APACHE III DatabasePresence of Infection, % †Documented infections in ICD-9 coding system.Patient Subgroup (%)*Definitions for each subgroup are shown in Table 2. Values in parentheses are percentage of subgroup or subsubgroup. SIRS = met criteria for diagnosis of SIRS; Not SIRS = did not meet criteria for diagnosis of SIRS.Mortality Rate,SepticemiaOther InfectionsInfection (10)32Admission sepsis (31)48SIRS (92)4847Not SIRS (8)Other infections (69)25SIRS (77)2714Not SIRS (23)1810High risk of infection (22)24SIRS (71)28828Not SIRS (29)16319Low risk of infection (59)14SIRS (57)18422Not SIRS (43)8215Inflammation (10)10SIRS (71)11215Not SIRS (29)60.810* Definitions for each subgroup are shown in Table 2. Values in parentheses are percentage of subgroup or subsubgroup. SIRS = met criteria for diagnosis of SIRS; Not SIRS = did not meet criteria for diagnosis of SIRS.† Documented infections in ICD-9 coding system. Open table in a new tab Table 2Classification of ICU Admisssion Diagnoses From APACHE III Database*GI = gastrointestinal; S/P = status post.InfectionSepsisBacterial pneumoniaBacterial meningitisCholangitis/cholecystitisRenal abscessPeritonitis, GI perforation/ruptureCellulitisHigh risk of infectionViral/fungal/parasitic infectionsPulmonary congestionShockS/P resuscitationComaHematologic disordersS/P emergency surgeryS/P GI/propharynx surgeryGI obstruction/vascular insufficiencyS/P vascular surgery with graftS/P transplantationViral/fungal/parasitic pneumoniaARDS, COPD, asthma, aspirationHemorrhagic/cardiogenic shockS/P cardiac/respiratory arrestNontraumatic comaNeutropenia, coagulopathyS/P emergency abdominal aortic aneurysmectomyS/P GI neoplasmGI paralysisFemoral-popliteal bypassS/P renal transplantationAscitesLow risk of infectionElective nonabdominal surgeryCardiovascular diseaseNonoperative neoplasmsNeurologic diseaseMetabolic/endocrinologic diseaseElective thoracic surgeryAcute myocardial infarctionCongestive heart failureRhythm disturbanceThyroid neoplasmMyasthenia gravis seizuresDiabetic ketoacidosis myxedemaGI bleeding drug overdoseInflammationTraumaFracturePancreatitisGI inflammatory disease* GI = gastrointestinal; S/P = status post. Open table in a new tab A similar message is illustrated in Table 3, which shows the proportion of patients with one type of infection (pneumonia), one inflammatory condition (head trauma), and a diagnosis associated with low risk of infection (congestive heart failure) who met SIRS criteria. The pneumonia and head trauma categories contain the most patients with SIRS, but even in the congestive heart failure subgroup, half of the patients met the SIRS criteria.Table 3SIRS Classification Status in Subgroups of ICU Admissions*SIRS = met criteria for diagnosis of SIRS; Not SIRS = did not meet criteria for diagnosis of SIRS.SubgroupProportion of Subgroup, %Mortality Rate, %PneumoniaSIRS8341Not SIRS1730Head traumaSIRS7120Not SIRS297Congestive heart failureSIRS5329Not SIRS4716* SIRS = met criteria for diagnosis of SIRS; Not SIRS = did not meet criteria for diagnosis of SIRS. Open table in a new tab We believe these results suggest that the SIRS definition is meeting its goal well, provided it is used as intended. Use of the SIRS definition identifies the majority of patients with clinical diagnoses of infection or inflammation (more than would qualify using the sepsis syndrome criteria) while still identifying patients at a mortality risk higher than average. If combined with a patient-specific risk score or prediction, the SIRS definition may then be able to meet another of its goals— increasing the number of patients eligible for clinical evaluation without sacrificing precision. Because many patients without infection or inflammation also meet criteria for SIRS, however, it must only be used with a precise clinical diagnosis. We appreciate the comments of Dr. Pilz and his colleagues in regard to the new definition of SIRS and its comparison with the traditional definition of sepsis syndrome. First, we wish to emphasize that both SIRS and sepsis syndrome are syndromes, that is, constellations of specific signs and symptoms. They are not alternative diagnoses for sepsis and are to be used only with a specific diagnosis.1Bone RC Balk RA Cerra FB Dellinger RP Fein AM Knaus WA et al.Definitions for sepsis and organ failure and guidelines for the use of innovative therapies in sepsis.Chest. 1992; 101: 1644-1655Abstract Full Text Full Text PDF PubMed Scopus (7440) Google Scholar For example, sepsis syndrome has always been defined as a set of physiologic findings associated with infection (ie, a diagnosis compatible with infection).2Bone RC Fisher CJ Clemmer TP Slotman GJ Metz CA Balk RA. Sepsis syndrome: a valid clinical entity.Crit Care Med. 1989; 17: 389-393Crossref PubMed Scopus (592) Google Scholar When this was done in a clinical trial situation, use of the rubric “sepsis svndrome” identified a group of patients with infection with a mean hospital mortality rate of approximately 40 percent.3Bone RC. Toward an epidemiology and natural history of SIRS.JAMA. 1992; 268: 3452-3455Crossref PubMed Scopus (425) Google Scholar Use of the sepsis syndrome definition improved the precision and uniformity of many recent clinical trials involving critically ill patients with sepsis. Like all syndrome definitions, however, the sepsis syndrome definition uses categorical cutoffs (eg, temperature > 38° or < 36°C) to identify patients. Many patients may fail to meet these exact cutoffs and still may be quite similar to included patients. As our ability to measure and estimate patient risk improved, it was also recognized that use of the sepsis syndrome definition may have identified groups of patients with a relatively uniform group mortality rate but with a very wide distribution of individual patient risk.4Knaus WA Sun X Nystrom PO Wagner DP. Evaluation of definitions for sepsis.Chest. 1992; 101: 1656-1662Abstract Full Text Full Text PDF PubMed Scopus (120) Google Scholar These concepts led to the suggestion of a less restrictive syndrome definition, such as that of SIRS, that could be combined with more precise risk prediction techniques and diagnostic labels.3Bone RC. Toward an epidemiology and natural history of SIRS.JAMA. 1992; 268: 3452-3455Crossref PubMed Scopus (425) Google Scholar This could expand the number of patients eligible for clinical investigation while increasing the precision of description for each individual patient through use of either a severity score or risk prediction. These would be measured in a continuous, rather than a categorical, manner. In regard to patients admitted to ICUs with sepsis, the use of the SIRS definition, provided it is combined with a severity measure, appears to accomplish these objectives well. Pilz and colleagues ask how it applies to other conditions. To answer this, we applied SIRS criteria to all 17,440 ICU admissions in the original APACHE III database (Tables 1, 2). Because SIRS was designed to be applied only to patients with diagnoses compatible with infection or inflammation, this application is artificial. Nevertheless, the results in Table 1 show that of patients with specific diagnoses of sepsis, 92 percent met the criteria for SIRS. Of patients at high risk of infection (eg, those with ARDS), 77 percent met the definition. Likewise, 71 percent of patients with inflammation (eg, due to trauma or pancreatitis) also met the criteria for SIRS. Even in the category of patients at low risk of infection, 57 percent met the SIRS criteria. In all of these categories, patients who met the criteria for SIRS had higher hospital mortality rates and a greater incidence of infections as measured by discharge diagnoses. A similar message is illustrated in Table 3, which shows the proportion of patients with one type of infection (pneumonia), one inflammatory condition (head trauma), and a diagnosis associated with low risk of infection (congestive heart failure) who met SIRS criteria. The pneumonia and head trauma categories contain the most patients with SIRS, but even in the congestive heart failure subgroup, half of the patients met the SIRS criteria. We believe these results suggest that the SIRS definition is meeting its goal well, provided it is used as intended. Use of the SIRS definition identifies the majority of patients with clinical diagnoses of infection or inflammation (more than would qualify using the sepsis syndrome criteria) while still identifying patients at a mortality risk higher than average. If combined with a patient-specific risk score or prediction, the SIRS definition may then be able to meet another of its goals— increasing the number of patients eligible for clinical evaluation without sacrificing precision. Because many patients without infection or inflammation also meet criteria for SIRS, however, it must only be used with a precise clinical diagnosis. Evaluation of Definitions of SepsisCHESTVol. 105Issue 3PreviewWith great interest we read in the June 1992 issue of Chest the new definitions of sepsis and systemic inflammatory response syndrome (SIRS)1 and, in another journal, a comparison of SIRS with “sepsis syndrome.”2 Evaluating these definitions, Knaus et al3 showed that “SIRS” was superior to “sepsis syndrome” regarding its sensitivity for identifying the 519 patients with a clinically defined primary diagnosis of sepsis out of the APACHE III patient database4 (sensitivity, 97 percent vs 59 percent). Full-Text PDF