The Natural History of the Systemic

Inflammatory Response Syndrome (SIRS)

A Prospective Study
M. Sigfrido Rangel-Frausto, MD, MSc;
Didier Pittet, MD; Michele Costigan, RN, BSN; Taekyu Hwang, MS;
Charles S. Davis, PhD; Richard P. Wenzel, MD, MSc

Objective.\p=m-\Definethe epidemiology of the four recently classified syndromes SEPSIS and septic shock are significant
causes of morbidity and mortality. In the
describing the biologic response to infection: systemic inflammatory response syn- United States, it has been estimated that
drome (SIRS), sepsis, severe sepsis, and septic shock.
there are approximately 500 000 new epi¬
Design.\p=m-\Prospectivecohort study with a follow-up of 28 days or until discharge sodes of sepsis each year with an asso¬
if earlier. ciated 35% crude mortality rate.13 Shock
Setting.\p=m-\Threeintensive care units and three general wards in a tertiary health is present in 40% of patients with sepsis
care institution. and adversely affects the prognosis.3
Methods.\p=m-\Patientswere included if they met at least two of the criteria for SIRS: Moreover, among hospitalized patients
fever or hypothermia, tachycardia, tachypnea, or abnormal white blood cell count. in noncoronary intensive care units
Main Outcomes Measures.\p=m-\Developmentof any stage of the biologic re- (ICUs), sepsis has been reported to be
sponse to infection: sepsis, severe sepsis, septic shock, end-organ dysfunction, the most common cause of death.4 Re¬
and death. cently, the US Vital Statistics Report lists
Results.\p=m-\Duringthe study period 3708 patients were admitted to the survey sepsis as the 13th leading cause of death.6
units, and 2527 (68%) met the criteria for SIRS. The incidence density rates for SIRS
in the surgical, medical, and cardiovascular intensive care units were 857,804, and For editorial comment see 155.
542 episodes per 1000 patient-days, respectively, and 671,495, and 320 per 1000
patient-days for the medical, cardiothoracic, and general surgery wards, respec- Sepsis and septic shock are commonly
tively. Among patients with SIRS, 649 (26%) developed sepsis, 467 (18%) devel- used terms. The hypothesis is that they
oped severe sepsis, and 110 (4%) developed septic shock. The median interval represent increasingly severe stages of
from SIRS to sepsis was inversely correlated with the number of SIRS criteria (two, the same disorder. These stages do not
three, or all four) that the patients met. As the population of patients progressed from necessarily imply increasing severity of
SIRS to septic shock, increasing proportions had adult respiratory distress infection, rather an increasing severity of
syndrome, disseminated intravascular coagulation, acute renal failure, and shock. the systemic response to infection.6 Al¬
Positive blood cultures were found in 17% of patients with sepsis, in 25% with se- though the term sepsis syndrome as origi¬
vere sepsis, and in 69% with septic shock. There were also stepwise increases in nally described by Bone and colleagues
identified a population of patients at risk
mortality rates in the hierarchy from SIRS, sepsis, severe sepsis, and septic shock: for adult respiratory distress syndrome
7%, 16%, 20%, and 46%, respectively. Of interest, we also observed equal num- (ARDS) and death,7,8 in common usage it
bers of patients who appeared to have sepsis, severe sepsis, and septic shock but now appears both confusing and ambigu¬
who had negative cultures. They had been prescribed empirical antibiotics for a ous. Newer categories and more precise
median of 3 days. The cause of the systemic inflammatory response in these definitions have evolved from discussions
culture-negative populations is unknown, but they had similar morbidity and mor- at a recent Consensus Conference.9 Spe¬
tality rates as the respective culture-positive populations. cifically, the term systemic inflamma¬
Conclusions.\p=m-\Thisprospective epidemiologic study of SIRS and related con- tory response syndrome (SIRS) was de¬
ditions provides, to our knowledge, the first evidence of a clinical progression from veloped to imply a clinical response aris¬
SIRS to sepsis to severe sepsis and septic shock. ing from a nonspecific insult and includes
(JAMA. 1995;273:117-123) two or more of the following: (1) tem¬
perature greater than 38°C or less than
36°C, (2) heart rate greater than 90 beats
From the Division of General Medicine, Clinical Epide-
per minute, (3) respiratory rate greater
Department of Internal Medicine, University Hospital, than 20 breaths per minute or a PC02 less
miology, and Health Services Research, Department of Geneva, Switzerland (Dr Pittet). Dr Wenzel has served as
Internal Medicine (Drs Rangel-Frausto and Wenzel and a consultant for Pfizer Roerig, New York, NY. than 32 mm Hg, or (4) white blood cell
Ms Costigan), and Division of Biostatistics, Department Reprint requests to Division of General Medicine, count greater than 12.0 X109/L or less than
of Preventive Medicine (Dr Davis and Mr Hwang), C-41 GH, University of Iowa Hospitals and Clinics, Iowa
University of Iowa College of Medicine, Iowa City; and City, IA 52242 (Dr Wenzel). 4.0xl09/L or the presence of more than
Infection Control Group, Division of Infectious Diseases, 0.10 immature neutrophils.

Downloaded From: http://jama.jamanetwork.com/ by a Northern Ontario School of Medicine User on 11/02/2013

 Sepsis is now defined as the presence
of SIRS associated with a confirmed in¬
fectious process.9 Discarding the term
sepsis syndrome, members of the Con¬
sensus Conference defined severe sepsis
as the presence of sepsis with either
hypotension or systemic manifestations
of hypoperfusion. Manifestations of hy-
poperfusion include lactic acidosis, oli-
guria, or altered mental status. Septic
shock was defined as sepsis with hypo¬
tension despite adequate fluid resusci¬
tation, associated with hypoperfusion ab¬
normalities that included, but were not
limited to, lactic acidosis, oliguria, or an
acute alteration in mental status.9
Infections are thought to be the most
common causes of SIRS,9 mediated via
macrophage-derived cytokines, which
target end-organ receptors.10·11 Because
mortality with sepsis is high,3 new thera¬
peutic agents have been developed tar¬
geting various intermediate steps within
the currently understood sepsis cas¬
cade.12 These agents include monoclonal
antibodies directed at endotoxin13·14 or
tumor necrosis factor1516 and antagonists
directed at interleukin-1 receptors.17 Al¬
though the new agents may be life spar¬
ing, they are potentially very expen¬
sive, and a major concern is how they
can be prescribed optimally. Unfortu¬
nately, epidemiologie data related to the
incidence, natural history, and outcome
ofwhat is called sepsis are not available.
Important questions arise: What pro¬
portion of hospitalized patients have
SIRS, sepsis, severe sepsis, or septic
shock? Are the distributions different in
critical care units compared with wards?
Do these terms define increasingly se¬
vere responses to the same disorder, and
do they develop sequentially? If they de¬
velop sequentially, what is the interval
for progressing from one stage to the
next? What proportions of patients with
each syndrome develop specific end-or¬
gan dysfunctions (ie, ARDS, disseminated
intravascular coagulation [DIC], acute re¬
nal failure [ARF], or shock)? What is the
mortality rate for each syndrome?
To date, no study has systematically
examined the rates and distribution of
any syndrome intended to define sepsis
in hospitalized patients. The study re¬
ported herein addresses the epidemiol¬
ogy and natural history of SIRS, sepsis,
severe sepsis, and septic shock.
METHODS
Patient Population
The University of Iowa Hospitals and
Clinics, Iowa City, is a 900-bed teaching
referral center where approximately
30 000 patients are admitted each year.
The hospital serves patients through¬
out Iowa and those in bordering states.
Downloaded From: http://jama.jamanetwork.com/ by a Northern Ontario School of Medicine User on 11/02/2013
Two hundred beds are designated for
intensive care, including those for criti¬
cal care and those for organ transplan¬
tation. Of the 700 non-critical care beds,
472 are designated for adults (excluding
pediatrics, psychiatry, and medicine-psy¬
chiatry). During the study period (Au¬
gust 1,1992, to April 30,1993), concur¬
rent incidence surveys for the four pu¬
tative stages of sepsis were done in three
critical care units and three wards.
Surveillance
A 1-month pilot study was performed
in April 1992 in the surgical ICU.18 The
purpose of the pilot study was to ex¬
amine the feasibility of performing in¬
cidence surveys in the larger study, gain
insight into specific problems, and gather
preliminary data to develop a plan for
analysis. Study nurses had the oppor¬
tunity to test the case report forms and
gain experience in the assessment of
end-organ dysfunctions.
Prospective surveillance during the
9-month study was performed in three
adult ICUs (medical ICU with 12 beds,
surgical ICU with 24 beds, and cardio¬
vascular ICU with 13 beds), and in three
general wards (a general surgery ward
with 26 beds, a cardiothoracic surgery
ward with 24 beds, and a medical on¬
cology ward with 22 beds), which his¬
torically have had the highest incidence
of nosocomial bloodstream infections.
Rates for nosocomial bloodstream in¬
fections in the ICUs have varied be¬
tween 9.2 and 19.6 per 100 admissions
per year, and those on the three study
wards selected have varied between 9.3
and 26.2 per 100 admissions per year.19,20
Surveillance was performed by three
experienced, specifically trained re¬
search nurses who visited all study units
each weekday. On Mondays, all data
were collected retrospectively from the
charts of patients admitted on the week¬
ends after a review of the daily census
of the respective units. Each weekday,
the study nurses reviewed the medical
charts and nursing notes and identified
all patients who were 16 years of age or
older meeting two or more of the crite¬
ria for SIRS. At the time of admission
to the study, a special case report form
was completed, recording demographic
and clinical data. Patients were excluded
if discharged from the ICU with less
than a 12-hour stay. All patients admit¬
ted to the study were followed up for 28
days or until discharge from the hospi¬
tal if it occurred before 28 days. End
points included the following: the de¬
velopment of sepsis, severe sepsis, sep¬
tic shock, and end-organ dysfunctions.
Subsequent evaluations of survival were
made by telephone or chart review at 3
months and 6 months.
Definitions of the Biologic
Response to Infection
The definitions we used for SIRS, sep¬
sis, severe sepsis, and septic shock were
those reported by the American College
of Chest Physicians-Society of Critical
Care Medicine Consensus Conference,9
as noted herein. We prospectively noted
whether patients who met criteria for
SIRS had two, three, or all four of the
criteria. A confirmed infectious process
was required for diagnosing sepsis. For
severe sepsis, the presence of hypoten¬
sion was defined as either one of the fol¬
lowing: a reduction of systolic pressure to
less than 90 mm Hg, a 50% reduction in
hypertensive patients, or a systemic mani¬
festation of peripheral hypoperfusion,
such as lactic acidosis, oliguria, or acute
alteration of mental status. Septic shock
was defined as a patient with hypoten¬
sion not responsive to a 500-mL intrave¬
nous fluid challenge plus manifestations
of peripheral hypoperfusion. Patients
without hypotension but receiving more
than 5 µg/kg per minute of dopamine or
any other vasopressor agent were also
considered to be in shock.
We also gathered data on patients who
appeared to have sepsis, severe sepsis,
or septic shock but who never had posi¬
tive cultures. All of the culture-nega¬
tive patients received empirical antibi¬
otics, defined as any antibiotic therapy
administered in the absence of culture-
proven infection to a patient with SIRS
who was not in a 48-hour window fol¬
lowing surgery (Table 1). Rates of oc¬
currence and outcome measures were
compared for patients with each syn¬
drome stratified by those with docu¬
mented infection (culture-positive) vs
those with undocumented but clinically
suspected infection (culture-negative).
Our hypothesis was that a large pro¬
portion of patients who appear septic
have no proven site of infection.
Definitions of End-organ
Dysfunctions
Adult respiratory distress syndrome
was defined as the presence of respira¬
tory insufficiency (Pa02/Fl02 [fraction of
inspired oxygen] ratio <175), with bilat¬
eral infiltrates consistent with pulmonary
edema, in the absence of heart failure (if
measured, a pulmonary capillary wedge
pressure <18 mm Hg) or primary pul¬
monary disease.21 Acute renal failure was
defined as an acute increase in the serum
creatinine concentration greater than 180
µ /L (2.0 mg/dL), a doubling in the
admission creatinine level in a patient
with chronic renal failure, or the require¬
ment for acute dialysis or ultrafiltration.
Oliguria was defined as a urinary output
less than 0.5 mL/kg per hour for at least
 1 hour or less than 30 mL for 2 hours.
Lactic acidosis was diagnosed if the
plasma lactate level was greater than 2.0
mmol/L. Disseminated intravascular co¬
agulation was defined as a decrease in the
platelet count of 25% or more from the
baseline with any increase in prothrom-
bin time. Mental status changes were
scored using the Glasgow Coma Scale.22
Specifically, we defined mental status
changes as a Glasgow score of 11. On the
general wards, where the Glasgow score
was not routinely recorded, the presence
or absence of mental status changes was
recorded as noted in the nurses' daily
notes. Severity-of-illness scores, using
Acute Physiology and Chronic Health
Evaluation II (APACHE II)23 and
McCabe and Jackson24 classifications,
were also recorded on admission to the
study. When patients underwent surgery,
the American Society of Anesthesiology
score was also recorded.25
Diagnostic Categories
and Infection Data
We stratified the underlyingdiagnoses
of our patient population using Inter¬
national Statistical Classification of
Diseases, Ninth Revision, definitions and
combined diagnoses into the following
general categories: neoplasias, cardio¬
vascular diseases, trauma, diabetes mel-
litus, pancreatitis, renal diseases, res¬
piratory diseases, and diseases of the
gastrointestinal tract.
When patients met criteria for infec¬
tion, we stratified by the site of infec¬
tion within each of the categories of sep¬
sis, severe sepsis, and septic shock. We
were especially interested in what pro¬
portion of patients in each category had
positive blood cultures.
Reliability, Validity,
and Generalizability
On three occasions concurrent surveys
were performed on 25 surgical ICU pa¬
tients to examine reliability: once before
the study and twice during the study at 1
and 3 months. Interobserver reliability
was measured as the ability of our study
nurses to view the same charts and
record the same data. Twenty-five vari¬
ables that were systematically recorded
on the case report forms were used for
this purpose. Concordance among observ¬
ers was calculated as values according
to the following formula: K=(observed
agreement-expected agreement)/(l -ex¬
pected agreement).26 Good reproducibil-
ity was considered whenthe values were
from 0.40 to 0.75; a value greater than 0.75
denoted excellent reliability. The validity
ofthe data recorded by the research nurses
was also estimated by comparing the data
recorded by each nurse with that of a
single combined evaluation by one of two
Downloaded From: http://jama.jamanetwork.com/ by a Northern Ontario School of Medicine User on 11/02/2013
Table 1.—Definitions
Consensus Conference definitions9:
Systemic inflammatory response syndrome (SIRS). Two or more of the following:
1. Temperature >38°C or <36°C
2. Heart rate >90 beats/min
3. Respiratory rate >20 breaths/min
4. White blood cell count >12.0x109/L, <4.0x109/L, or >0.10 immature forms (bands).
Sepsis. SIRS plus a documented infection (positive culture for organism).
Severe sepsis. Sepsis associated with organ dysfunction, hypoperfusion abnormalities, or hypotension.
Hypoperfusion abnormalities include, but are not limited to, lactic acidosis, oliguria, or an acute alteration
in mental status.
Septic shock. Sepsis-induced hypotension despite fluid resuscitation plus hypoperfusion abnormalities.
Culture-negative populations:
Culture-negative sepsis. SIRS plus empirical antibiotic treatment for a clinically suspected infection but in
whom all cultures were negative.
Culture-negative severe sepsis. SIRS associated with organ dysfunction, hypoperfusion, or hypotension. All
cultures were negative, yet empirical antibiotic treatment for a clinically suspected infection was
prescribed. Hypoperfusion abnormalities include, but are not limited to, lactic acidosis, oliguria, or an acute
alteration in mental status.
Culture-negative septic shock. SIRS associated with hypotension despite fluid resuscitation plus
hypoperfusion abnormalities. All cultures were negative, yet empirical antibiotic treatment for a clinically
suspected infection was prescribed.
physicians (D. P. or S.R.F.) considered to Statistical Analysis
be the gold standard. Sensitivity and speci¬ Rates of SIRS, sepsis, severe sepsis,
ficity were examined for all variables re¬ and septic shock were calculated as cases
corded on the case report forms.
per 100 patients and as cases per 1000
The interrater reliability data showed
patient-days. Based on the data collected
progressive improvement from the pre- in the pilot study,18 SIRS with three or
study values as the nurses gained further four of the criteria were predetermined
training and experience. Overall, sta¬
tistics were 0.56 (range, 0.20 to 1.0), 0.64
categories for our analysis, separate from
SIRS with only two of the criteria. Pa¬
(range, 0.30 to 1.0), and 0.91 (range, 0.60 tient characteristics and the outcomes of
to 1.0), respectively, for the three sequen¬ interest (different syndromes) and rela¬
tial evaluations. In the three validity stud¬ tive risk (RR) and their corresponding
ies, the three nurses had a sensitivity of 95% confidence intervals (CIs) were cal¬
80% (range, 57% to 100%), 83% (range, culated. For continuous variables, mean
50% to 100%), and 85% (range, 75% to values were compared using two sample
100%) and a specificity of 88% (range, 82% t tests for independent samples. Differ¬
to 94%), 80% (range, 78% to 90%), and ences in proportions were compared us¬
96% (range, 80% to 100%), respectively.
ing a 2 test or Fisher's Exact Test when
To estimate the generalizability of our
appropriate. Mean values are reported
data, 1-day prevalence studies of the ± 1 SD. All tests of significance were two
entire adult hospital population were tailed. Kaplan-Meier survival curves were
performed on two occasions. Our major calculated for three end points: time to
goal was to compare the rates of all re¬ sepsis (for patients with SIRS), time to
corded syndromes in the three routinely severe sepsis (for patients with sepsis),
surveyed general wards with those in and time to septic shock (for patients with
the 27 nonsurveyed general wards. severe sepsis). In these survival analy¬
SIRS: A Continuum to Shock ses, individuals who died of other causes
or did not progress to the outcome were
The hypothesis tested was that SIRS, censored at the time of death, at discharge
sepsis, severe sepsis, and septic shock from the hospital, or at 28 days, which¬
represent a hierarchical continuum of ever came first.27 All statistical analyses
an inflammatory response to infection. were carried out using UNI VARI ATE,
The hypothesis was examined in three
FREQ, and LIFE TEST procedures in
ways: (1) by determining the proportion SAS Version 6.™
of patients with a specific syndrome who
had been classified on at least 1 day in
a previous syndrome (eg, of all patients RESULTS
with severe sepsis, those who had pre¬ During the 9-month study period, 3708
viously met criteria for sepsis or SIRS); patients were admitted to the three gen¬
(2) by noting the interval in time and eral wards and the three critical care units
specifically the median interval between that were surveyed. Of these, 2527 (68%)
one syndrome and the next in the hier¬ met two or more criteria for SIRS and
archy for those who progressed clini¬ were followed up prospectively for 28 days
cally to more severe systemic inflam¬ or until discharge from the hospital. The
matory responses; and (3) by examining in-hospital follow-up period for the 2527
the rates of end-organ dysfunction and subjects represented 30126 patient-days.
mortality for populations of patients who This figure represents 89% of the 33 973
met criteria for each of the defined syn¬ patient-days for the entire population ad¬
dromes. mitted to the study areas. There were
 1528 males (60%) and 999 females (40%).
The mean (±SD) age for men was
54.7±17.2 years (range, 16 to 95 years)
and for women was 55.7±18.1 years
(range, 16 to 92 years). On average, the
APACHE II score on admission to the
critical care units was 18.5 ±9 (range, 2 to
71). Surgical procedures were performed
on 1179 patients (47%), with a mean
American Society of Anesthesiology score
of 2.79 (range, 1 to 5; median, 3). The
severity-of-illness score described by
McCabe and Jackson allowed us to clas¬
sify 421 (17%) of our study patients with
rapidly fatal diseases, 469 (19%) with ul¬
timately fatal diseases, and 1634 (65%)
with nonfatal conditions; three patients
(0.1%) were not classified.
During the 28-day follow-up period, 649
patients (26%) developed evidence of sep¬
sis, 467 (18%) developed severe sepsis,
and 110 (4%) developed septic shock
(Table 2). An additional 892 (35%)
thought to be clinically septic and
prescribed antibiotics
empirically for a
median of 3 days (range, 1 to 26 days), but
they had no site with a positive culture.
The causes of their systemic inflamma¬
tory responses are unknown.
Of the 2527 patients who met the cri¬
teria for SIRS, 1821 (72%) developed
three of the criteria for SIRS, and 975
(39%) met all four criteria for SIRS
some point in time. The incidence of the
syndromes within each of the study units
(Table 3) shows higher rates for patients
in the surgical ICU and medical ICU
compared with those in the cardiovas¬
cular ICU and medical or surgical wards.
In these critical care units, SIRS oc¬
were curred at a rate of 857 and 804 episodes
were per 1000 patient-days, respectively. At
the other end of the clinical spectrum,
culture-proven septic shock occurred at
a rate of 40 and 57 episodes per 1000
patient-days, respectively, in the surgi¬
cal and medical ICUs. Lower rates were
noted for patients in the general wards.
at Diagnostic Categories
and Infection Data
When one compares the underlying dis¬
eases categories in culture-positive and

Table 2.—Frequency of Syndromes*

culture-negative categories of systemic
inflammatory responses (Table 4), it is
No. of Patients Patients With apparent that cardiovascular diseases are
Category_Patients_Surveyed, %_Sepsis, %_ 1.5 to 2.0 times more common in the cul¬
Total
surveyed_3708_100_^_ ture-negative group. Such patients ap¬
SIRS_2527_68_100_ peared to have an inflammatory response
Sepsis to infection and received empirical anti¬
Culture-positive 649 17 26
892 35
biotics, but bacterial and yeast cultures
Culture-negative 24 were never positive. Patients with cul¬
Severe sepsis
Culture-positive 467 13 18 ture-negative severe sepsis also had a 1.5
Culture-negative 527 14 21 greater proportion with trauma who mim¬
icked the culture-positive syndrome of
Septic shock severe sepsis yet never met criteria for
Culture-positive 110 3 4
2
infection. In the sepsis category, there
Culture-negative 84 3
were 1.4 to 2.3 times as many patients
*Patients with positive cultures are those meeting Consensus Conference9 definitions; also, we Identified patients with gastrointestinal and respiratory dis¬
withsystemic Inflammatory response syndrome (SIRS) and clinically suspected infections (culture-negative) if they eases in the culture-negative group than
received empirical antimicrobial therapy. in the culture-positive group (Table 4).
Table 3.—Incidence Density of Systemic Inflammatory Response Syndrome (SIRS), Sepsis, Severe Sepsis, and Septic Shock: Episodes per 1000 Patient-Days
Sepsis Severe Sepsis Septic Shock
Site* SIRS Culture-Positive Culture-Negative Culture-Positive Culture-Negative Culture-Positive Culture-Negative
Surgical ICU 857 305
Medical ICU 368 126 263 95
Cardiovascular ICU 542 76 119 82 14
Medical oncology ward 671 295 38
Cardiothoracic surgery ward 495 46 81 13 34
General surgery ward 320 61 72 19

*ICU indicates intensive care unit.

Table 4.—Number of Patients in ICD-9 Diagnostic Categories With Culture-Positive and Culture-Negative Syndromes*
Sepsis Severe Sepsis Septic Shock
Culture-Positive Culture-Negative Culture-Positive Culture-Negative Culture-Positive Culture-Negative
Category (n=182)tt (n=366)tt (n=358)tt (n=457)tt (n=110)t (n=84)t
Neoplasias 49 83 42 55 16
Cardiovascular diseases 26 96 79 166t 32 43
Trauma 18 32 63 10
Diabetes mellitus 12
Pancreatitis
Renal diseases
Respiratory diseases 17» 15 28
Gastrointestinal diseases 21 61 28 41

Hematology diseases 10
Others 59 65 139 76 34 22

*
ICD-9 indicates International Statistical Classification of Diseases, Ninth Revision.
tNumber of patients in each category that never progressed to the next category. For example, the 182 patients in the group of culture-positive sepsis are those who never
developed (during the 28-day follow-up) severe sepsis or septic shock. This type of comparison allowed us to perform statistical analyses by category of underlying disease.
t-P<.05.

Downloaded From: http://jama.jamanetwork.com/ by a Northern Ontario School of Medicine User on 11/02/2013

 Table 5.—Prevalence of Systemic Inflammatory Response Syndrome (SIRS), Sepsis, Severe Sepsis, and
Septic Shock in Routinely Surveyed (n=3) and Nonsurveyed Wards (n=27)*
Prevalence Survey No. 1 Prevalence Survey No. 2
I -1
Routinely Routinely Routinely Routinely
Surveyed Wards Nonsurveyed Wards Surveyed Wards Nonsurveyed Wards
Syndrome (n=39t) (n=315t) (n=41t) (n=334f)
SIRS 64 (47-79) 30 (25-35) 61 (45-76) 32 (27-37)
Sepsis 8(2-21) 6(4-10) 20 (9-35) 8(5-11) "-SO.2
Severe sepsis 3(0.06-13) 2 (0.5-4) 3 (2-6)
Septic shock 0 O 2 4 6 8 10 12 14 16 18 20 22 24 26 28
Interval Between SIRS and Sepsis, d
*Number per 100 patients (95% confidence interval). The proportion of patients with culture-negative sepsis in
both surveys was 18% and 17% on the routinely surveyed wards and 7% and 8% on the routinely nonsurveyed wards. 1.0
The proportion for culture-negative severe sepsis In both surveys was 3% and 2% on the routinely surveyed wards Sepsis Culture-Negative
and 1 % and 2% in the routinely nonsurveyed wards. The proportion of culture-negative septic shock was 0% in both Sepsis Culture-Positive
surveys for routinely and nonroutlnely surveyed wards.
tThe number of patients (census) on the day of the survey. tio
¿•
S 0.6

Positive blood cultures were found in SIRS: A Continuum to Shock

io
S .S
S
c 0.4
^
16.5% of those with sepsis, 25.4% ofthose
with severe sepsis, and 69.1% of those To examine the hypothesis of a con¬ -1— -1— —r-
2 4 6 8 10 12 14 16 18 20 22 24 26 28
with septic shock. Pneumonia occurred tinuum for SIRS, we determined the Interval Between Sepsis
in 9%, 25%, and 26% of the patients with and Severe Sepsis, d
proportion of patients in each syndrome
sepsis, severe sepsis, and septic shock,
respectively. Urinary tract infections
were found in 26% of the patients with
who had been classified at least 1 day in
a previous syndrome. We also noted the
interval for progressing from one syn¬
rW
° œ0.6
sepsis, 38% of those with severe sepsis, drome to the next. Of those who pro¬ ¿.en
and 25% of those with septic shock. Sur¬ gressed to culture-proven septic shock =

a c
o
0.4
gical wound infections were found in 16% (n=110), 78 (71%) had been previously O
2
'w
of those with sepsis, 10% of those with classified as severe sepsis, sepsis, or o-l 0.2 Severe Sepsis Culture-Negative
Severe Sepsis Culture-Positive
severe sepsis, and 5% of those with sep¬ SIRS. The remaining 32 (29%) met cri¬
tic shock. teria for septic shock on the first day of 2 4 6 8 10 12 14 16 18 20 22 24 26 28
admission to the study. Of those who Interval Between Severe Sepsis
Generalizability of the Data met the Consensus Conference criteria9 and Septic Shock, d

Two, 1-day hospital-wide prevalence for severe sepsis (culture-proven;

surveys were performed in adult wards.
The hospital census in the adult non-
critical care wards was 354 patients (75%
occupancy) in the first survey and 375
(80% occupancy) in the second survey.
Distributions of the prevalence rates of
sepsis stages among patients in the three
routinely surveyed general wards were
compared with those of the 27 nonsur-
veyed general wards (Table 5). The
prevalence rates for SIRS, sepsis, se¬
vere sepsis, and septic shock were con¬
sistent in both surveys. However, there
was a somewhat higher prevalence of
sepsis among the three surveyed wards
in the second survey compared with the
first survey (20% vs 8%). Importantly,
the prevalence surveys demonstrated
that the three wards routinely studied
in the incidence surveys had twice the
rates of SIRS as the 27 routinely non-
surveyed wards. In general, an equiva¬
lent number of patients were noted with
culture-negative sepsis, culture-negative
severe sepsis, and culture-negative sep¬
tic shock vs those with positive cultures
in both surveys. From 2% to 3% of pa¬
tients on either routinely surveyed or
routinely nonsurveyed wards met cri¬
teria for severe sepsis in the prevalence
surveys. No patients on any of the gen¬
eral wards met criteria for septic shock
during the two prevalence surveys.
n=467), 271 (58%) had been classified
previously as sepsis or SIRS. The re¬
maining (42%) met the criteria for se¬
vere sepsis the first day of admission.
Among patients with sepsis (n=649), 285
(44%) had earlier met at least two cri¬
teria for SIRS. The remaining 56% met
criteria for sepsis on the first day of
admission to the study.
The progression of SIRS to sepsis
was also examined by stratifying the
SIRS population into three groups: those
who met only two of the four SIRS cri¬
teria, those who met three of the crite¬
ria, and those who met all four criteria.
Of interest, 50% of the patients with
two criteria for SIRS subsequently de¬
veloped a third criterion by day 7. Among
those with two SIRS criteria, 32% de¬
veloped sepsis by day 14, whereas for
those with three SIRS criteria, 36% de¬
veloped sepsis by day 14. Of those with
all four criteria, 45% developed sepsis
by day 14, and the median interval to
sepsis was 21 days (Figure 1, top). The
median interval from sepsis to severe
sepsis was 1 day; 64% developed severe
sepsis by day 14 (Figure 1, middle). The
median interval from severe sepsis to
septic shock was more than 28 days.
Nevertheless, by day 14,23% developed
septic shock (Figure 1, bottom).
The progression of patients from SIRS
Figure 1.—Top, Progression of patients who met
systemic inflammatory response
the criteria of the
syndrome (SIRS) to sepsis. SIRS2 indicates two of
the criteria were met; SIRS3, three of the criteria
were met; and SIRS4, four of the criteria were met.
Data include only patients with culture-proven in¬
fection meeting Consensus Conference criteria.9
The interval to sepsis was progressively shorter as
more criteria were met. Middle, Progression of sep¬
sis to severe sepsis. The time to sepsis was similar
among patients with culture-negative and culture-
positive sepsis. Bottom, Progression of severe
sepsis to septic shock. The time to septic shock was
similar among patients with culture-negative and
culture-positive severe sepsis.
to more severe inflammatory responses
but without a positive culture was also
examined. All ofthese patients received
empirical antibiotics for clinically sus¬
pected infection. Of patients with two
SIRS criteria, 37% developed culture-
negative sepsis by day 14. When three
of the criteria were fulfilled, 43% devel¬
oped sepsis, and with four criteria, 46%
developed sepsis by day 14. Patients
with culture-negative sepsis had a me¬
dian interval to severe sepsis of only 1
day, the same as was noted with the
culture-proven syndromes; by day 14,
61% had developed severe sepsis (Fig¬
ure 1, middle). Similarly, the median in¬
terval of culture-negative severe sepsis
to shock was more than 28 days, and by
day 14,16% had developed septic shock
(Figure 1, bottom).

Downloaded From: http://jama.jamanetwork.com/ by a Northern Ontario School of Medicine User on 11/02/2013

 Table 6.—Outcome of Systemic Inflammatory Response
Septic Shock: Attack Rates for End-organ Dysfunction
Adult Respiratory
Syndrome Distress Syndrome, %
SIRS
Two criteria
Three criteria
Four criteria
Sepsis
Culture-positive
Culture-negative
Severe sepsis
Culture-positive
Culture-negative
Septic shock
Culture-positive
Culture-negative 18
*P<.05.
End-organ Dysfunctions
Our hypothesis was that progressively
increased rates of end-organ dysfunc¬
tion would occur in each of the catego¬
ries of sepsis if there was a continuum
from SIRS to shock. We observed in¬
creased inflammatory responses to in¬
fection associated with higher propor¬
tions of end-organ dysfunction (Table
6). The attack rates of ARDS, DIC, ARF,
and shock increased directly as patients
met two, three, and four criteria for
SIRS. Rates of ARDS, DIC, ARF, and
shock were higher in the group of pa¬
tients with sepsis and a culture-proven
infection compared with those with a
negative culture. Among patients with
severe sepsis, rates of ARDS and DIC
were similar, regardless of the presence
or absence of a proven microbiological
focus of infection. However, ARF and
shock occurred more frequently in the
culture-proven stratum, meeting the
Consensus Conference criteria. No dif¬
ference was seen in the rates of ARDS,
DIC, or ARF in the group of patients
with shock and culture-proven or cul¬
ture-negative infection.
Mortality
We hypothesized that progressively in¬
creased mortality rates would occur in
each of the categories of sepsis if there
was a continuum from SIRS to shock. On
the first day of admission to the study,
1206 patients met the definition for SIRS
with only two of the criteria, and 69 (6%)
subsequently died. This rate is twice as
high as the crude mortality (3% [37/1191])
among patients who never met any of the
criteria for SIRS (P=.002; RR, 1.9; 95%
CI, 1.2 to 2.9). Of the 924 patients who had
three of the criteria for SIRS on the first
day, 84 (9%) subsequently died (P=.004;
RR, 1.6; 95% CI, 1.2 to 2.3), compared
with those with two criteria. Furthermore,
397 patients met SIRS with four of the
SIRS criteria the first day; of these, 71
Downloaded From: http://jama.jamanetwork.com/ by a Northern Ontario School of Medicine User on 11/02/2013
Syndrome (SIRS), Sepsis, Severe Sepsis, and
Disseminated
Intravascular Acute Renal
Coagulation, % Failure, %
8
15
19 19
16* 19*
20
18 23*
16
38
38 38
(18%) subsequently died (P<.001; RR, 2.2;
95% CI, 1.5 to 3.1), compared with those
with three criteria. Mortality was also ex¬
amined as a function ofthe individual syn¬
drome, regardless of when in the
follow-up period the criteria were met
(Figure 2). The mortality rate in the group
with culture-positive sepsis (16%) was
higher than that among those with SIRS
and suspected but undocumented infec¬
tion (10%). Patients with severe sepsis
who had a positive culture also had a some¬
what higher mortality rate (20%) com¬
pared with patients with severe sepsis
and negative cultures (16%). Patients who
developed shock had the same mortality
rate in both the documented and undocu¬
mented infection groups (46%). It should
be emphasized that all figures represent
an overall or crude mortality rate. The
mortality rate directly attributable to the
infections or systemic inflammatory re¬
sponses is unknown.
The crude mortality rate in the 28-day
follow-up was 9% (224 deaths). An addi¬
tional 111 patients (4%) died in the ensu¬
ing 3 months after discharge from the
hospital. Of these patients dying in the
3-month postdischarge period, 46 (41%)
were previously classified in the group
with SIRS only, 11 (10%) with sepsis, 49
(44%) with severe sepsis, and five (5%)
with septic shock. One hundred thirteen
(4%) died between 3 and 6 months after
discharge. Of those who died in the latter
interval after discharge, 31 (27%) were
previously classified with SIRS only, 28
(25%) with sepsis, 50 (44%) with severe
sepsis, and four (4%) with septic shock.
Thus, the crude mortality rate 6 months
after discharge was 17% for the entire
cohort.
COMMENT
Most prior studies of the epidemiol¬
ogy of sepsis have been retrospective.
Although such work has been important
for defining risk factors, retrospective
21
27
20*
28*
22
Figure 2.—Mortality in systemic inflammatory re¬
sponse syndromes (SIRS). A progressively higher
100 mortality was observed as more severe inflamma¬
100 tory response criteria were met. Only data for
patients meeting the Consensus Conference crite¬
ria9 are illustrated. The mortality rates in the group
with culture-negative syndromes were similar to
those with positive cultures: 10% for culture-
negative sepsis, 16% for culture-negative severe
sepsis, and 46% for culture-negative septic shock.
SIRS2 indicates two of the criteria were met for
SIRS; SIRS3, three of the criteria were met; and
SIRS4, four of the criteria were met.
28-day
studies are limited in their ability to
define the natural history of the syn¬
drome. We present here a prospective
epidemiologie study of SIRS and related
conditions using criteria proposed in a
recent Consensus Conference.9 These
data clearly illustrate that the Consen¬
sus Conference criteria are consistent
with the hypothesis that a clinical pro¬
gression from SIRS to septic shock de¬
fines the natural history of the inflam¬
matory response to infection.
Well-performed prospective studies
should demonstrate both high interrater
reliability for data collection and validity
of the data. An estimate of the general-
izability of the data is also necessary, and
there should be sufficient follow-up to
detect important end points of interest.
Each of these issues has been addressed
in this study. First, interrater reliability
was carefully assessed before and during
the study and found to be excellent. Sec¬
ond, the validity of the reporting team
was excellent: by the third study, the
overall sensitivity and specificity were
85% and 96%, respectively. Third, the
prevalence studies comparing the rates
on the general wards with those on the
three routinely surveyed wards showed
that the prevalence rates of SIRS and
sepsis on the nonsurveyed wards were
one half of the rates of the routinely sur¬
veyed wards. This is an important find¬
ing to gauge the applicability of the data
to other hospitals. Prevalence and inci¬
dence are related such that prevalence is
a direct function of both incidence and the
average duration of the condition under
study.29 Assuming that the average du¬
rations of SIRS and sepsis are similar on
surveyed and nonsurveyed wards, our
data on incidence of SIRS on the general
 wards may overestimate incidence for
most general wards by a factor of two.
The 28-day mortality rate was 9%,
and an additional 8% of patients died
within 6 months of discharge. Recently
conducted clinical trials of new immu¬
nothérapies for sepsis have used mor¬
tality at 28 or 30 days as the major end
point.13·14 Our data suggest that an
equivalent number of deaths occur in
the 6 months after hospital discharge
and that careful follow-up longer than 1
month might give an additional mea¬
sure of the impact of these new agents.
Overall, SIRS was found to occur fre¬
quently in the patients surveyed. Rates
were especially high in ICUs and in our
study wards with known high rates of
nosocomial bloodstream infection. Pa¬
tients with only two criteria of SIRS
had twice the rate of death (7%) as those
who never developed SIRS (3%). The
incorporation of more than two criteria
for SIRS was associated with an in¬
creased probability of identifying a sub¬
set of patients likely to develop sepsis in
a short time with higher rates of organ
References
1. Parker MM, Parrillo JE. Septic shock: hemody-
namics and pathogenesis. JAMA. 1983;250:3324\x=req-\
3327.
2. Wenzel RP. Anti-endotoxin monoclonal antibod-
ies: a second look. N Engl J Med. 1992;326:1151\x=req-\
1153.
3. Bone RC, Fisher CJ Jr, Clemmer TP, Slotman
GJ, Metz CA, Balk RA. A controlled clinical trial of
high-dose methylprednisolone in the treatment of
severe sepsis and septic shock. N Engl J Med.
1987;317:653-658.
4. Niederman MS, Fein AM. Sepsis syndrome, the
adult respiratory distress syndrome, and nosoco-
mial pneumonia: a common clinical sequence. Clin
Chest Med. 1990;11:633-656.
5. Centers for Disease Control and Prevention, Na-
tional Center for Health Statistics. Mortality Pat-
terns\p=m-\UnitedStates, 1990. Monthly Vital Stat Rep.
1993;41:5.
6. Bone RC. Sepsis syndrome: new insights into its
pathogenesis and treatment. Infect Dis Clin North
Am. 1991;5:793-805.
7. Bone RC, Fisher CJ Jr, Clemmer TP, Slotman
GJ, Metz CA, Balk RA. Sepsis syndrome: a valid
clinical entity. Crit Care Med. 1989;17:389-393.
8. Bone RC. Sepsis, the sepsis syndrome, multi-
organ failure: a plea for comparable definitions.
Ann Intern Med. 1991;114:332-333.
9. American College of Chest Physicians\p=m-\Society
of Critical Care Medicine Consensus Conference.
Definitions for sepsis and organ failure and guide-
lines for the use of innovative therapies in sepsis.
Crit Care Med. 1992;20:864-875.
10. Parrillo JE. Pathogenetic mechanisms of septic
shock. N Engl J Med. 1993;328:1471-1477.
11. Nathan CF. Secretory macrophages. J Clin In-
vest. 1987;79:319-326.
Downloaded From: http://jama.jamanetwork.com/ by a Northern Ontario School of Medicine User on 11/02/2013
dysfunctions and higher mortality (10%).
In our opinion, SIRS with only two cri¬
teria—as initially proposed9—is less
helpful in defining a subset of ICU and
ward patients who are at especially high
risk for development of sepsis than SIRS
with three or all four of the criteria. The
latter may be the cohort for whom new
interventions including effective immu¬
nothérapies should be directed.
In our study, half of the patients in¬
cluded in our definition of sepsis did not
have a documented infection; however,
they were prescribed antibiotics (me¬
dian, 3 days) by their physicians for sus¬
pected infection. Outcomes in the group
with culture-negative septic shock were
similar to those meeting the criteria of
the Consensus Conference of culture-
proven septic shock. However, patients
with culture-negative severe sepsis
tended to have lower rates of ARF and
shock than those with positive cultures.
We cannot conclude at this time that
culture-positive and culture-negative
syndromes have different pathogeneses.
However, our preliminary data suggest
12. Ziegler EJ, McCutchan JA, Fierer J, et al. Treat-
ment of gram-negative bacteremia and shock with
human antiserum to a mutant Escherichia coli.
N Engl J Med. 1982;307:1225-1230.
13. Ziegler EJ, Fisher CJ Jr, SprungCL, et al. Treat-
ment of gram-negative bacteremia and septic shock
with HA-1A human monoclonal antibody against en-
dotoxin: a randomized, double-blind, placebo-con-
trolled trial. N Engl J Med. 1991;324:429-436.
14. Greenman RL, Schein RMH, Martin MA, et al.
A controlled clinical trial of E5 murine monoclonal
IgM antibody to endotoxin in the treatment of gram\x=req-\
negative sepsis. JAMA. 1991;266:1097-1102.
15. Exley AR, Cohen J, Buurman W, et al. Mono-
clonal antibody to TNF in severe septic shock. Lan-
cet. 1990;335:1275-1277.
16. Wherry J, Wenzel R, Wunderink R, et al. Mono-
clonal antibody to human necrosis factor (TNF Mab):
multi-center efficacy and safety study in patients
with the sepsis syndrome. In: Program and ab-
stracts of the Interscience Conference on Antimi-
crobial Agents and Chemotherapy; October 17-20,
1993; New Orleans, La. Abstract 696:246.
17. Opal SM, Fisher CJ, Slotman GJ, et al. The
phase II interleukin-1 receptor antagonist (IL-1ra)
sepsis syndrome trial: analysis of clinical, cytokine,
and microbial features with outcome. In: Program
and abstracts of the Interscience Conference on
Antimicrobial Agents and Chemotherapy; October
11-14, 1992; Anaheim, Calif. Abstract 1570:372.
18. Pittet D, Rangel-Frausto MS, Tarara D, Co-
stigan M, Wenzel RP. SIRS, sepsis, and severe
sepsis: incidence, morbidities, and outcomes in SICU
patients. In: Program and abstracts of the Inter-
science Conference on Antimicrobial Agents and
Chemotherapy; October 17-20,1993; New Orleans,
La. Abstract 1301:358.
that underlying cardiovascular disease
and trauma are related to a systemic
inflammatory response in culture-nega¬
tive populations.
Most importantly, we provide strong
epidemiologie evidence for the hypoth¬
esis that SIRS and related conditions
represent a hierarchical continuum of
increased inflammatory response to in¬
fection. Forty-four percent to 71% of
patients in any category had progressed
from a previous state of biologic response
syndrome. Other patients either pro¬
gressed through more than two stages
within a 24-hour window or appeared to
have skipped a stage, for example, go¬
ing from sepsis to shock within a 24-
hour time period. Positive blood culture
rates, end-organ failure rates, and mor¬
tality increased with each subsequent
stage of systemic inflammatory response.
This study was supported in part by a grant from
Pfizer Roerig, New York, NY.
We are grateful to Rachel Happel, RN, Frosti
Chichelly, RN, Nancy Wagner, RN, and Debra
Tarara, RN, for data collection.
19. Broderick A, Mori M, Nettleman M, Streed S,
Wenzel RP. Nosocomial infections: validation of
surveillance and computer modeling to identify pa-
tients at risk. Am J Epidemiol. 1990;131:734-742.
20. Pittet D, Omahen J, Tarara D, Wenzel RP.
Current risk factors for nosocomial bloodstream
infection in a tertiary health care center. In: Pro-
gram and abstracts of the Interscience Conference
on Antimicrobial Agents and Chemotherapy; Sep-
tember 29-October 2, 1991; Chicago, Ill. Abstract
800:232.
21. Murray JF, Matthay MA, Luce JM, Flick MR. An
expanded definition of the adult respiratory distress
syndrome. Am Rev Respir Dis. 1988;138:720-723.
22. Teasdale G, Jennett B. Assessment of coma and
impaired consciousness: a practical scale. Lancet.
1974;2:81-84.
23. Knaus WA, Draper EA, Wagner DP, Zimmer-
man JE. APACHE II: a severity of disease clas-
sification system. Crit Care Med. 1985;13:818-829.
24. McCabe WR, Jackson GG. Gram-negative bac-
teremia, II: clinical, laboratory, and therapeutic ob-
servations. Arch Intern Med. 1962;110:856-864.
25. Owens WD, Felts JA, Spitznagel EL Jr. ASA
physical status classifications: a study of consis-
tency of ratings. Anesthesiology. 1978;49:239-243.
26. Woolson R. Statistical Methods for the Analy-
sis of Biomedical Data. New York, NY: John Wiley
& Sons Inc; 1987:252-260.
27. Kalbfleisch JD, Prentice RL. The Statistical
Analysis of Failure Time Data. New York, NY:
John Wiley & Sons Inc; 1980:10-20.
28. SAS Institute. SAS/STAT User's Guide: Ver-
sion 6.4th ed. Vol 2. Cary, NC: SAS Institute; 1990.
29. Freeman J, Hutchinson GB. Prevalence, inci-
dence, and duration. Am J Epidemiol. 1980;112:
707-723.