Documente Academic
Documente Profesional
Documente Cultură
Paul C. Hebert, M.D.;t Alanaj Drummond, Gordon R. Bernard, M.D., FC.C.P; andJames R.N.;Joel
A.
system organ failure (MSOF) score may of patients who have sepsis syndrome. Using an MSOF scoring system, we prospectively determined the presence or absence of respiratory, cardiovasculai; renal, hepatic, gastrointestinal, hematologic, and neurologic organ failure on day 1 of sepsis syndrome in 154 consecutive patients who had sepsis syndrome in the ICU of a tertiary care, teaching hospital. We used 30-day hospital mortality as the primary outcome variable. Overall 30-day mortality was 34 percent. There was a strong linear association between number of organ system failures and 30-day mortality (p<O.000l). Mortality was 20 percent in patients who had less than 3 organ system failures (n 111) and 70 percent in patients who had 3 or more organ system failures (n43). Survival was assessed using the Cox proportional hazards model and was found to be significandy different (pzz0.01) between the two groups defined by the aforementioned dichotomy after adjustment for age and sex using time to death as the primary outcome. The
A simple
predict
multiple
increase
organ interval,
mortality
odds
were
risk of death associated with 3 or more failures was 2.77 (95 percent confidence 2.74 to 2.83). Usinglogistic regression, the adjusted ratios (OR) for covariates most predictive of mortality in relative
system
hematologic
(OR
cardiovascular
6.2),
neurologic
(OR
4.4),
hepatic
and age (1.05 per year). The logistic model using the seven organ system failures and age as covariates accurately predicted outcome 75 percent of the time with a sensitivity of 51 percent and specificity of 87 percent. In conclusion, a simple scoring system tabulating the number of organ system failures present on day 1 of sepsis syndrome predicts the mortality of patients who have sepsis syndrome with reasonable
(OR=3.4), (0R2.6),
accuracy.
(Chest
1993;
104:230-35)
respiratory MSOFmultiple
distress system
syndrome; organ
CIconfldence
failure;
ORodds
epsis
syndrome and of
to dysis a sepsis
who
have
AIDS
requiring
ICU
admission.9
MSOF
developed of the MSOF predictor score
defined
Multisystem cause
as three in 55 percent
system and
survivors. In critically score on the day ofICU of mortality (an AIDS scoring than was
is a complication
of other
illness7 as well as of sepsis syndrome. FurMSOF scoring has been used to predict mortality of patients who have a variety of critical illnesses.t7 However, there are few studies that have determined whether a MSOF scoring system can be used at onset of sepsis syndrome who a MSOF have scoring to predict rigorously system mortality defined originally in a group of patients sepsis syndrome 17 We have used utility adapted
system21),
MSOF score could be useful in outcome ofpatients who have sepsis syndrome. MSOF oforgan predicts an ICU asked scoring system 30-day who whether have the syndrome system, failures hospital sepsis develwas we system survivors Finally, from we the
1 of sepsis of patients
syndrome in we day on
by Bihari and colleagues7 and have assessed its in outcome prediction in patients who had adult distress syndrome (ARDS)8 and in patients
1 of sepsis
respiratory
mortality. Furthermore, which specific organ in discriminating syndrome. developed model sepsis
useful
*Fmrn
the
Division
of Critical
Care
Medicine,
St.
Pauls
Hospital
(Drs. Hebert and Russell, and Ms. Drummond), the Department ofHealth Care and Epidemiology, University ofBritish Columbia, Vancouver, British Columbia, Canada (Dr. Singer), and Division of Pulmonary and Critical Care, Vanderbilt University, Nashville, Tenn (Dr. Bernard). tRecipient of the 1991-1992 Merck-Frosst Critical Care Medicine Research Fellowship for Western Canada. Supported in part by grant 1 R01HL43167-02 CLTR from the National Institute of Health, Bethesda, Md. Reprint requests: Dr. Russell, St. Paulr Hospital, 1081 Burrard Street, Vancouver, BC, Canada V5Z 1Y6
predetermined accurately
Patients Patients
and
Study
in the
Design
ICU of St. Pauls Hospital, Vancouver, a tertiary
230
A Simple
MSOF
Scoring
System (Hebertetal)
Table
1-Diagnostic Sepsis
Who
Had
Table
2-Multiple Ibtients
system: Hg
System
of
At temperature: rate: rate >20 38.3#{176}C (101#{176}F) or 35.5#{176}C (96#{176}F) (in the breaths/mm absence ofbeta blockers) ventilation >10 LI 90 beats/mm
1. 2. 3.
an A-a
Respiratory Bt
or minute
4. system: mm adequate or urine system: If the meet bilateral must diffuse patient over output, PaO,, primary established infiltrates catheter as the two points sole significant 70 infused are 12, solution system: 333. hypotension, than 90 mm 1 h given artery Hg that wedge or a fall the filling pressure 500 1 h then and the ml of 6. in 5. BP >40 Hg for greater
phosphatase tube by
U/L
nasogastric accompanied 2 U of packed criteria: count, and or a melena cells white rectum
(pulmonary
is unresponsive 1 h) 0.5 mm diagnosis ARDS on chest failing
red blood
2. 3.
rnl/kgfh
Pulmonary
Hg (room criteria
air) or a PaOJFIo2 (PaO,fFIo2<200, PAWP for the pulmonary in mental Coma of infection. one status Scale)
blood abnormal
count >2
40 x 10#{176}/L; or
fibrmn degradation nervous in the system
coagulation
disseminated
mm
7. Central
decrease *Each absent organ
at least was
3 ifprimary determined
4.
Central
(a decrease
failure on day
or value
1 ofsepsis
must meet
a suspected
They group B
to determine
presence
or absence
of each
at least
to fuffill
patients
immediately
prior
to sedation
or paralysis.
tResting conditions. jThe failing organ organ underlying care ment Sepsis system failure disease
the new
primary and
Statistical The whether organ Haenszel organ was with patients, organ as three using between effects We time goal
Analysis
of the primary mortality failures. failures
secondary
study with we
was
to determine the number a Mantelnumber there and of three defined 1 and A Cox for the type of differences 2 of of
were
prospectively Effects
screened of Ibuprofen
for enrollin Human 1990 for St. eligible a known vital dysfunction if they were virus. in the syndrome ICU or 154 or proven sepsis Pauls for or sign to
used the
x#{176} statistic
April
syndrome Hospital inclusion suspected
1992,
study in
trial. During the period from December 250 patients with at least group A criteria
1) were screening the current of infection in group B (Fable expected or were as the present later identified log. Patients study at least accompanied A plus one were 6 h), had 1). Patients to survive positive criteria MSOF and entered were if there by all new into was organ the considered of the
mortality. whether
(Table
between
we patients comparison as the 1 and patients failures
patients
defined having having less
without
two MSOF groups than
question, 1) and
respectively, organ
as follows: system for model the was 30 days of groups risk of death
alterations listed moribund brain Day ICU could could of the study. The deaths more, syndrome citing time who 1 was have have 250
2).
Kaplan-Meier
for human
calculated
day of onset
in Table time the met at the during
syndrome is, sepsis to the criteria hospital 30 days. therapies outcome calculated nurse system 2. The to failures difficulty Coma Scale (A.
as defined
by the
determined 100
associated
As a result,
patients
the
We considered
end-point
relevant
because
60
clinical
30-day within
40
primary
care ofthe study.
J.
D.) 1 a
20
or absent
Number
of Organ
Failures
individual
coding
patients addressed
determination values
Ficua 1. Mortality per number oforgan system failures of patients who had sepsis syndrome. There was a significant linear increase in mortality (percent) with each additional organ system failure
MH,I<0.#{176}#{176}#{176}1). Mortality
paralyzed
difficulty
assigning
system
failures
to 100 percent
with system
CHEST
I 104 I 1 I JULY,
1993
231
Table
3-Clinical
Nonsurvivora
Characteristics ofSepsis
in Survivors Syndrome
Survivors
and
RESULTS
Nonsurvivors (n)
The clinical characteristics nonsurvivors of sepsis syndrome 3. Cardiac disease (medical and
of
Clinical
Age, <20 yr
Characteristics
(n)
postoperative),
5 21
1 4 18 30 35/18 6 2 4 1 4 8 2 6 4 2 7 6 1
20-39 40-59
60 Sex, M/F underlying failure diagnosis
33
42 70/31 12 6 13 8 (excluding disease disease disease pneumonia) 3 12 9 3 6 4 patients 14 7 4
monia, and trauma were the most common primary underlying diagnoses. The overall 30-day mortality of our patients who had sepsis syndrome was 34 percent. Using the Mantel-Haenszel x2 we found a strong linear association (p(O.OOOl) between the number oforgan system ranged system failures and 30-day mortality from 10 percent for patients failure to 100 percent organ system Kaplan-Meier patients failures. survival 1 and (Fig 1). Mortality who had no organ who were had
Primary
in patients curves
constructed in patients
in groups
2. Mortality
Malignancies
Renal disease Postoperative Cardiac Noncardiac Other
less than three organ system percent (n = 111) compared in patients failures,
comparing
with
MSOF
(three
organ using
system 30-day
failure. mortality
A logistic as the model, intervals was also based Model individual the
was
constructed system and system of likely with was used 95 parameters. (OR) organ
oforgan
as model ratios CI) for each outcome outcome were sensitivities, All statistical version
adjusted for the effects of age and sex was highly significant (p 0.00502, Fig 2). The increase in relative risk of death associated with three or more organ system failures was 2.77 (95 percent CI, 2.74 to 2.83). The logistic regression model confirmed the prediction based on the Mantel-Haenszel organ system failures predicted also enabled a quantification
mortality
x2 that
30-day of the system
individual on the
observed predictive to establish were version done 2.2
patients
of the
who had
model. and one
survival
with
individual
organ
established
on which
significant increase in the risk of death (l#{224}ble 4) was noted in patients who had hematologic (OR = 6.3 with 95 percent
a of 0.05
significance on a Sun
100 80
CI from 1.3 to 29.8), neurologic (OR=4.4 CI from 1.9 to 10.3), hepatic (OR = 3.4 CI from 1.06 to 10.7), and cardiovas(OR = 2.6 with 95 percent CI from 1.2 to 5.8) system failure. Age was also used as a model for The probability each additional age of the of death increased by year of life (p<0.001) sample of patients who had 5
60 40
Cl)
20
0 0 6 12 18 24 30
Odds Ratios ofOrgan System Failures Associated With 30-Day Mortality of Who Had Sepsia Syndrome
Odds Confidence Intervals 0.69-4.00 1.18-5.77 0.76-9.81 1.06-10.70 0.22-5.41 1.29-29.77 1.86-10.32 1.02-1.08 logistic regression. p Values 0.2581 0.0179 0.1240 0.0394 0.9129 0.0228 0.0007
Covariates Respiratory
Ratios5 1.66 2.61 2.73 3.37 1.09 6.20 4.39 1.05 calculated using
Days
FIGURE 2. Group survival over 30 days of patients who had sepsis syndrome. Patients were divided into two groups. Group 1 (n = 111) had less than three organ systems that failed and group 2 (n =43) had three or more organ systems that failed. Kaplan-Meier curves for each group were computed using time to death as the outcome. Using a Cox proportional hazards model adjusted for age and sex, the group 1 survival curve was signfficantly different from the group 2 survival curve (p<0.00502).
Cardiovascular
Renal
Hepatic Gastrointestinal Hematologic Neurologic
Age
#{216}Jds ratios
0.0010
232
A Simple
MSOF
Scoring
System
(Hebert
et a!)
Table
5-Logistic
Observed
Dead
Model*
3). The
respiratory
(n
74),
neurologic
(n =
65),
and
Alive
(n = 63) organ systems were the most systems to fail in patients who had Hepatic (n = 23), renal (n = 17), he12) organ in our paindividual
77
Predicted
Dead 27 26 (17.5%) (16.8%) 53 *Sensitjvity cent; predictive 13 88 (8.4%) (57.1%) 101 percent; value specificity 40 114 154 88/101 percent;
matologic (n = 13), and gastrointestinal (n system failures were much less frequent tients.
organ
Alive
The system
30-day
associated as follows:
with
failures
hematologic,
=
positive
27/53
= 50.9 =
88/114
=
67.5
= =
87.1
per-
predictive value
27/40 percent;
negative 115/154
= was
77.2
accuracy
50 percent;
respiratory,
74.7 percent.
A logistic
model
also
used
to
calculate
DIscussIoN
probability of death threshold probability our model, patients nonsurvivors outcome specificity to identify ity = 50.9 101 patients (Table enabled of the only percent) who The
patients. Using a determined from as survivors or with the observed and able
of this
study
was
a strong
linear
association between a simple MSOF ber of organ system failures on syndrome, patients finding three or and 30-day hospital syndrome who had failures)
.
score, the numday 1 of sepsis mortality Another MSOF in ICU as synimportant (defined of sepsis
(sensitiv88 of the
=
at onset
(specificity alive
=
drome had a significantly greater day period than did patients who The overall mortality in the while mortality was 70 percent compared with 20 percent for have MSOF. The Kaplan-Meier both groups clearly illustrate
mortality over a 30did not have MSOF. 34 percent with MSOF not of
or
74.7
in 115 of 154 patients had sepsis syndrome. the sum type of the of organ
tabulated
of organ system
system failures
failure (Fig
is much
30
Survivors
(I) C
*i
20
:;
Nonsurvivors
I-
0
w
10
z
0
I I iiL1
i1
iti
1 2
#{149}
I:
5 6
Total
Distribution syndrome. Survivors are pattern. There are seven organ. From left to right, bar 3, neurologic failure hematologic failure (n = proportion of nonsurvivors
FIGURE
Number
of organ
failures
3.
of patients and represented by bars per total bar 1 represents (n =45); bar 4, 1 1) and bar 7, increases with
individual organ system failures in patients who had sepsis the solid black portion of the bar and nonsurvivors by the cross number of organ system failures each representing a different respiratory failure (n = 57); bar 2, cardiovascular failure (n 49); hepatic failure (n = 13); bar 5, renal failure (n = 13); and bar 6, gastrointestinal failure (n 12). The plot demonstrates that the the number of organ failures.
CHEST
I 104
I 1 I JULY,
1993
233
in
patients
with
MSOF
at
onset
of
sepsis
than in patients without MSOF. Using regression, we were able to determine the risk of3O-day mortality associated with each failure. associated hepatic with an A signfficant increase in the risk with neurologic, hematologic, failure. Age was also signfficantly increased risk of death in our
used for definition of individual in this study were chosen there are no standard accepted organ we have definitions who had system found
failures that could that a simple MSOF was associated and with of patients
these
of patients
ARDS8
who
patients who had sepsis syndrome. We believe that our sample of patients who had sepsis syndrome is similar to other studies of patients who have sepsis syndrome because we used a very similar definition Bone and prospectively and of sepsis syndrome, modified from because we identified patients
The
absence review
and who were admitted to our ICU. values establishing the presence or failure were determined and by clinical judgment; that hematologic and with hepatic significantly the criteria (OR
=
by are
however,
6.20), in-
calculated
the
MSOF
score
at onset
(OR= were
failure
of sepsis syndrome, and because we observed a 30day mortality in our patients (34 percent) that was similar to the 30 percent mortality reported by Bone and colleagues.4 We believe used that this simple to predict that MSOF mortality Several the other number score early has not
(OR
3.37)
associated
creased risks of death of patients who had sepsis syndrome. There was overlap between the definitions of cardiovascular and neurologic criteria of sepsis syndrome (Table 1) and of organ failure (Table 2). In contrast, there was no overlap of hematologic or hepatic failure because they were not part ofthe sepsis syndrome definition. Thus, to have either hematologic or hepatic
organ
previously
in the
to mortality
failure,
the
patient
had
to have
ill adultse,1a,l3,ll7m, despite using quite definitions for individual organ failures and A major advantage of this MSOF scoring its simplicity. systems were individual Organ classified system system failures on a dichotomous failure was
is
dysfunction as defined by sepsis syndrome. might have contributed to the seriousness presence of hematologic and hepatic failure. Our study suggests that may be useful in predicting patients who have not documented sensitive is certainly treatment trials. We accurate in sepsis the ability individual found a prime groups that the MSOF outcomes
of seven scale.
organ
scored
or absent using simple criteria on day 1 of sepsis syndrome. that we minimized used and the masking potential observer potential patients Therefore, score
sepsis syndrome. that the MSOF known candidate at baseline the MSOF
to interventions
to compare placebo and in randomized control scoring system patient model, was not
patient. There are two First, we studied disciplinary whether this ICU.
is
not 30-day
established mortality
in prediction syndrome.
MSOF
predicts
of patients who have and in other settings ment and patient number of patients failures; therefore,
sepsis syndrome in other ICUs such as the Emergency DepartSecondly, there were had five and six organ regarding a small system in
wards. who
model developed in this study including age and the presence system failure were accurate correctly cent of (specificity) negative percent, There
conclusions
patients
these groups may not be generalizable even though a statistically significant linear trend was observed. The small number of patients in these groups may be explained partly by the screening process and by the entry criteria of the Cardiopulmonary Effects of Ibuprofen in Sepsis clinical trial. We excluded patients who were moribund, patients with suspected or proven brain immunodeficiency be more syndrome study.
234
predicted mortality (sensitivity) in 51 perpatients and correctly predicted survival in 87 percent ofpatients. The positive and predictive values were 68 percent and 77 respectively. are several reasons why predictive models limitations in individual First, predictive scales patient outcome are derived from
death, and patients positive for human virus. These excluded patients may to have MSOF on the patients included day 1 of sepsis in this MSOF
likely than
groups of patients groups may have patients. Secondly, tions render such
and therefore trends observed in little or no meaning for individual wide CIs around individual predicpredictions essentially meaningless.
A
System
(Hebertet&)
Thirdly,
scales be valid
derived when
from applied
one
sample to another
of
DE, failure:
RL,
Polk
HC.
patients sample
115:136-40 CJ, organ LW. organ E, Baue MM, JAMA PM, 1990; Marshall of Maier 121:196-208 pus Multiple 23:775-87 multisystem failure. Clin organ Chest S. reverse organ Multiple 11 12 13 14 Norton multiple Faist failure Descamp syndrome.
a simple MSOF scoring system mortality of patients who had sepsis simple MSOF scoring system may in have predicting sepsis mortality syndrome. of For to of concould who
J Surg
H,
several
applications
J Trauma
Posttraumatic
failure.
Dorinsky Med 15 Villar Multiple
Multiple
organ
This MSOF
J,
Care RC,
JJ,
6:75-80
MA,
Qumntana
J,
organ
in acute et al.
also be used for continuous quality improvement to compare groups ofpatients who have sepsis syndrome and who were admitted at different periods to an ICU. However, predictive mortality ther
larger
J Crit
16 Bell
JJ,
Jiskala failure.
Smith
failure
Intern 17 18 Ruokonen multiple Russell
in adult 99:293-97
J,
and Am
Kari
shock
Crit Care
Lockhat than Rev PM, with Chest EA, consumption Respir
JJ,
studies other
to validate this MSOF scoring samples of patients who have sepsis settings and from other centers.
respiratory
from
syndrome.
JJ,
failure
Russell predicts
J,
of
REFERENCES 1 Bone and 2 Bone Care 3 Bone clinical ofsevere 58 4 Bone drome: 5 Young 3rd ed. RC, Fisher clinical eds. Draper system M, prostacyclmn NV, Smithies ill patients. CJ, Clemmer Sepsis Crit sepsis. entity. Principles EA, Wagner failure. Gimson on N Engl organ failure. Ann oxygen TP, Care Slotman Group. GL, Med GJ, and Sepsis Douglas diseases. JE. 202:685-93 effects and JL. 1988; The of vasodiuptake microbiin Prognosis 25 The synRG 23 24 Methylprednisolone a valid LS. New WA, D, organ Gram-negative York: Churchill Severe Study In: Mandell 1990; Zimmerman 1985; delivery 1987; R, Arch Surg Meakins RC. RC. Med RC, trial Gram-negative Chest agree CJ, Lets sepsis; on terminology: Clemmer Sepsis shock. iT, background, definition Slotaman Group. clinical features 20 of sepsis. GJ, in the and Crit The
et al. Multiple
respiratory
102:1823-28 DP, Zimmerman system. prognostic syndrome.
secondary
APACHE Care system Med for
to AIDS-related Knaus 1985; 21 Justice 320:1388-93 22 Freidman Fundamental Book, libshirani model. Knaus M, 1991; Schafer prediction do they 1111-16 et hospital Clin WA, al. 1985; LM, WA, II: a severity 13:818-29 AC,
intervention.
1991; 100802-08
Wagner
1991;
Fisher
19:973-76
Severe Study N Engl A controlled treatment 317:6531987;
the acquired
immunodeficiency Furberg of clinical 191-211 R. A plain Invest Wagner APACHE Med DP, III for mans 5:63-8 Draper prognostic critically CD, trials.
Med
1989;
In:
of high-dose septic
methylprednisolone
Med
DL.
proportional
1989; 17:389-93
ofinfectious 611-36
of
C, Care
et al Med
with
intensive
unit: 18:
individual
317:397-403 123:309-15
JC, Christou
ofmultiple
Horn
CHEST
I 104
I 1 1 JULY.
1993
235