Beruflich Dokumente
Kultur Dokumente
0.609, 0.641, and 0.778 for the ESI, MEWS, and MEDS
score respectively. The AUCs concerning prognostic accuracy were 0.617, 0.642, and 0.871 for ESI, MEWS, and
MEDS score respectively.
By using the MEDS score systematically, critically ill
patients with sepsis could be detected in the ED. Finally,
the MEDS score provides the basis for a risk adjusted
disposition management that follows objective criteria.
Keywords sepsis emergency department severity
illness scoring systems early identification prognosis
508 Severity illness scoring systems for early identification and prediction of in-hospital mortality
13
original article
Introduction
Sepsis is a complex, systemic inflammatory host
immune response to infection [1]. Severe sepsis is associated with a hospital mortality of 55.2% [2]. With 59,344
deaths per year, severe sepsis is the third most common
cause of death in Germany [3]. Furthermore, Martin
et al. [4] provide evidence about significant increases
in the incidence of sepsis in the USA between 1997 and
2000 (164,072 vs. 659,935 newly diagnosed sepsis cases
per year). High mortality rates can be explained by the
presence of a systemic excessive immune response and
the wide variety of causative organisms, both in type and
in pathogenicity [5]. By identifying patients with sepsis
at an early stage and treating it according to the guidelines, the damaging consequences of pathophysiological changes can be limited [6]. However, this constitutes
a great challenge in clinical practice. Due to knowledge
gaps and/or uncertainties concerning the diagnosis of
sepsis, the highly threatened sepsis patient is at risk to
be identified too late. The use of severity illness scoring systems could provide some assistance with regard
to the early and reliable identification of patients with
severe sepsis. For example, the five-level triage system
emergency severity index (ESI) was developed in
order to assess illness severity of patients in emergency
department and according to that, set treatment priorities [7]. In addition the National Institute for Health and
Clinical Excellence in the United Kingdom argues for
using early warning scores [8]. Early warning scores
indicate critical conditions if vital signs are out of normal range.
Our aim is to evaluate the following severity illness
scoring systems concerning their diagnostic accuracy
to detect patients with severe sepsis and septic shock
(SSSS) under patients presenting with suspected sepsis
to the emergency department (ED): emergency severity
index (ESI), modified early warning score (MEWS), mortality in emergency department sepsis (MEDS) score.
Furthermore, our secondary aim is to determine their
prognostic accuracy in predicting the in-hospital mortality of patients with suspected sepsis in ED. In addition,
we calculate the prognostic value of the Charlson comorbidity index (CCI).
13
Outcome measures
The primary endpoint was the diagnostic accuracy of
the ESI, MEWS, and MEDS score in detecting critically
ill sepsis patients under those with suspected sepsis. We
also studied the chief complaints of the patients with suspected sepsis. The prognostic accuracy of the mentioned
risk stratification tools and the CCI was our secondary
aim.
Since 2009 we are using the five-level-algorithm ESI
for the assessment of the medical treatment urgency
of emergency patients. ESI level 1 is given to acute lifethreatened ill patients or to patients who require immediate initiation of diagnostics and therapy. Patients in
a high-risk situation are classified as ESI 2, initiation of
diagnostics and therapy has to start within 10min following the initial triage assessment. The higher the ESI level,
the lower the medical treatment urgency [9].
The MEWS is a simple and disease-unspecific scoring
system which can be used bedside. Its scoring system
is based on physiologic parameters of vital functions.
Severity illness scoring systems for early identification and prediction of in-hospital mortality
509
original article
Especially in highly frequented care units its use is supposed to identify patients at high risk for rapid deterioration. Its scale ranges between 0 and maximum 14
points [10].
The MEDS score was developed in order to predict
the 28-day mortality of patients presenting with infection to the ED. Its maximum value is 27. The MEDS score
is being calculated by summarizing nine weighted elements: age >65 years, nursing home resident, rapidly
terminal comorbid illness, lower respiratory infection,
bands >5%, tachypnea or hypoxemia, septic shock,
platelet count <150,000/mm3, and altered mental status
[11]. Due to the fact that we do not measure bands routinely in ED, this element was not included in the MEDS
score calculation.
The chief complaints were extracted from the ED
admission document and classified according to the
Canadian Emergency Department Information System
(CEDIS) presenting complaint list [12].
The CCI reflects type and number of comorbid chronic
diseases. If the CCI is high, the 1-year mortality in longterm studies is rather caused by comorbid diseases than
by the studied disease. If the CCI is 5 points, the probability that the patient has died due to his comorbid disease is 85%. The maximum value of the CCI is 37 [13].
Statistical analysis
Statistical analyses were carried out with the software
package SPSS Statistics Version 20 (IBM, Munich).
Results are presented as means, standard deviations with
95% confidence intervals and numbers or percentages
respectively. Using the ShapiroWilk test we tested the
variables on their normal distribution. More than two
independent samples were compared using the H-Test
according to Kruskall and Wallis, in case of normal distributed and ordinal variables. The means of nominal
data were compared by means of the Pearsons Chisquared test. The p-values were not corrected according
to Bonferroni. All hypotheses testing were two-tailed,
p-values less than 0.05 indicate statistical significance.
The diagnostic and prognostic values of the respective
risk stratification tools were calculated by means of the
receiver operating characteristic (ROC) curve.
Results
In the course of the observation period we registered 151
patients that entered the ED with suspected sepsis. In
this study, 45% (n=72) of the studied patients received
the adjusted final diagnosis of SSSS; 33.1% (n=53) of
the studied patients had a so-called uncomplicated sepsis, a sepsis without organ dysfunction. Finally 21.9% of
patients (n=26) had no sepsis, but SIRS or locally confined infection.
The mean age of patients with suspected sepsis was
68.3 (18). Fifty-three patients (35.1%) were older than
70 years and 37 patients (24.5%) were older than 80 years;
54.3% of the studied patients were men. Characteristics
of vital signs, laboratory parameters, chronic comorbidities, and sources of infection are shown in Table 1. Inhospital mortality of all registered patients was 14.6%, of
patients with SSSS 27.8%.
An appropriate ESI triage level of 2 was given to
70.8% of patients with SSSS and 45.6% of patients without SSSS were assessed with an ESI level >2. In case of an
ESI level 2, 54.3% of patients actually had a SSSS. In case
of an ESI level >2, 63.2% actually had no severe sepsis.
Table2 shows the detailed characteristics of diagnostic and prognostic accuracy criteria for the ESI, MEWS,
MEDS score, and CCI. Furthermore Fig. 1 pictures the
diagnostic accuracy of the mentioned risk stratification
tools by means of a ROC-curve. With an area under curve
(AUC) of 0.778 (95% CI: 0.7040.853) the MEDS score has
the highest diagnostic accuracy among the considered
instruments. In contrast, the ESI has little diagnostic
accuracy (AUC 0.609; 95% CI: 0.5180.699), similar to the
MEWS (AUC 0.641; 95% CI: 0.5520.730). Mean MEWS
value of patients with SSSS is 3.9 points. Patients with
an uncomplicated sepsis (and patients without sepsis
respectively) show a mean MEWS value of 3.28 (and 1.81
respectively). For further information, the study results
of Subbe et al. indicate a critical status of ED patients
with a MEWS value 5 [10].
Chief complaints of study patients are heterogeneous.
Table 3 lists the most documented chief complaints of
patients who were admitted to the ED with suspected
sepsis. These are fever (21.2%), dyspnea (15.2%), and
unspecified minor complaints (11.9%) which are defined
as a deterioration of the general state of health. There is
no significant difference among the three groups examined concerning the frequency of their chief complaints,
with the exception of the chief complaint altered level
of consciousness (p=0.029), which was more often in
patients with SSSS.
Figure 2 pictures the prognostic accuracy of the ESI,
MEWS, MEDS score, and CCI in predicting the in-hospital mortality of patients with suspected sepsis in the
ED. With an AUC value of 0.871 (95% CI: 0.7960.945) the
MEDS score has the highest prognostic accuracy.
The distribution of the MEDS score calculated risk
groups in patients with suspected sepsis, is as follows:
60.3% of patients have a low risk of 28-day mortality
(MEDS score = 07 points), 25.8% have a moderate risk to
510 Severity illness scoring systems for early identification and prediction of in-hospital mortality
13
original article
Table 1 Demographic and baseline characteristics of patients with suspected sepsis, in accordance to their final adjusted
disease
Variables
No sepsis (n=26)
SSSS (n=72)
Demography
Age*
68.318
58.817.9
64.520.6
74.613.3
<0.001
Men, n (%)
82 (54.3)
16 (61.5)
28 (52.8)
38 (52.8)
0.718
Vital signs
Respiration rate, 1/min*
207
175
206
218
0.015
Oxygen saturation, %*
945
972
954
935
<0.001
10121
9019
10518
10123
0.009
12828
13028
13821
11931
<0.001
Temperature, C*
381.1
37.70.8
38.50.8
37.81.3
0.001
Laboratory parameters
Leucocytes, 103/L*
14.310.5
11.48
13.14.9
16.313.6
0.005
Thrombocytes, /nl
233.6115.9
268.5130
228.893.7
224.6124.2
0.169
Lactate, mmol/L*
2.31.5
1.60.6
1.60.5
2.91.8
<0.001
1.15
1.92.6
0.92.2
2.95.8
<0.001
Creatinine, mg/dl*
1.71.5
1.51.6
1.31.2
2.11.7
0.001
CRP, mg/dl
12.610.7
9.27.5
1110.2
1511.7
0.051
Procalcitonin, ng/ml*
9.329.5
0.81.7
3.19.3
14.137.3
0.020
52 (34.4)
7 (26.9)
13 (24.5)
32 (44.4)
0.046
42 (27.8)
6 (23.1)
17 (32.1)
19 (26.4)
0.656
40 (26.5)
2 (7.7)
16 (30.2)
22 (30.6)
0.058
Dementia, n (%)*
40 (26.5)
4 (15.4)
8 (15.1)
28 (38.9)
0.004
40 (26.5)
4 (15.4)
13 (24.5)
23 (31.9)
0.240
Cerebrovascular
disease, n (%)
28 (18.5)
2 (7.7)
7 (13.2)
19 (26.4)
0.051
Pulmonary disease,
n (%)
22 (14.6)
1 (3.8)
9 (17)
12 (16.7)
0.234
Others, n (%)
78 (51.7)
Charlson comorbidity
index
43 [34]
33 [24]
33 [24]
43 [35]
0.052
57 (37.7)
5 (19.2)
14 (26.4)
38 (52.8)
0.001
Pneumonia, n (%)*
51 (33.8)
1 (3.8)
23 (43.4)
27 (37.5)
0.001
Acute abdominal
infection, n (%)
13 (8.6)
1 (3.8)
5 (9.4)
7 (9.7)
0.635
Tissue/bone, n (%)
10 (6.6)
1 (3.8)
6 (11.3)
3 (4.2)
0.232
Others, n (%)
9 (6)
Source of infection
*A p-value 0.05 indicates statistical significance among the groups no sepsis, uncomplicated sepsis, and SSSS
13
Discussion
This is the first study evaluating the diagnostic and prognostic accuracy of the ESI triage system, the MEWS,
MEDS score, and CCI in patients with suspected sepsis
presenting to a German-speaking ED. The main results
Severity illness scoring systems for early identification and prediction of in-hospital mortality
511
original article
MEWS
MEDS
score
CCI
Sensitivity
0.708
0.366
0.592
Specificity
0.456
0.798
0.785
0.543
0.619
0.712
0.632
0.583
0.681
1.301
1.808
2.749
0.64
0.795
0.520
Sensitivity
0.727
0.429
0.857
0.818
Specificity
0.395
0.744
0.682
0.465
0.170
0.214
0.305
0.207
0.895
0.889
0.967
0.938
1.203
1.675
2.697
1.530
0.690
0.768
0.209
0.391
CCI Charlson comorbidity index, ESI emergency severity index, MEDS mortality in emergency department sepsis, MEWS modified early warning score
are the following: (1) while the MEDS score shows a significant diagnostic accuracy in identifying patients with
SSSS, ESI and MEWS do not. (2) The MEDS score is a valid
tool for prediction of in-hospital mortality in patients
with SSSS. The ESI, MEWS, and CCI show low prognostic
accuracy. (3) Patients with suspected sepsis enter the ED
with heterogeneous chief complaints like fever, dyspnea,
or deterioration of general health status. (4) About half of
patients with a high risk of 28-day mortality are admitted
to a unit of higher care like ICU or IMC.
In contrast to the MEDS score, ESI and MEWS do not
show high values of diagnostic accuracy in identifying
the critically ill sepsis patient. There are several reasons
for the low diagnostic accuracy. For being identified as
urgent emergency patient in the ESI triage assessment,
either a life-saving procedure or a high-risk situation
in terms of a potentially life-threatening symptom
is required. If the patient is confused, lethargic, disorientated, or suffers from pain, an ESI level of 2 is
indicated too. However, not every critically ill patient
with SSSS exhibits this features. Let us look at the
example of a 73-year-old man that enters the ED with
cough, yellowish secretion, and a body temperature of
39.2C. The patient actually has an organ dysfunction
of his kidney, but this is not evident at the time point
of triage assessment. According to the decision rules
of the ESI system, the patient is correctly assessed
with an ESI level of 3. Valuable minutes or even hours
can pass till physicians are able to assess the severity
of patients health status by themselves and therefore
initiate the necessary procedures. Our results indicate
that the ESI triage system has weaknesses with regard
to the early identification of critically ill patients with
severe sepsis. Patients with acute, sepsis-associated
encephalopathy or septic shock are identified reliably
by the ESI system and its criteria; sepsis patients with
acute kidney failure or strongly increased lactate levels
probably not.
The MEWS, which shows no diagnostic accuracy as
well, seldom achieves high values in our patient cohort.
Only 36.6% of patients with SSSS show a MEWS value
5 points. The average MEWS score of the critically ill
sepsis patient is 3.9, and therefore much lower than the
proposed cut-off value of 5 from Subbe et al. [10]. The
MEWS relevant values like systolic blood pressure, heart
frequency, respiratory frequency, body temperature, and
status of consciousness may be easily collected in the ED
setting but they do not seem to be out of normal range
in every case of SSSS. Given the fact that only strongly
increased or decreased values result in a high MEWS
value, many cases of SSSS remain undetected. In order
to detect critically ill patients reliably, the weighting of
small deviations could be increased.
The MEDS score calculation is based on parameters
that take greater account of organ dysfunctions. Thereby,
an organ dysfunction is operationalized through tachypnea or hypoxemia, presence of septic shock, a platelet
count lesser than 150,000/mm3, and an altered mental
status. In addition to these clinical criteria, the MEDS
512 Severity illness scoring systems for early identification and prediction of in-hospital mortality
13
original article
Table 3 The most documented chief complaints of patients with suspected sepsis at emergency department admission, following the classification of the Canadian Emergency Department Information System (CEDIS) presenting complaint list
Chief complaint at time point of admission
No sepsis (n=26)
SSSS (n=72)
Fever, n (%)
32 (21.2)
6 (23.1)
15 (28.3)
11 (15.3)
0.205
23 (15.2)
2 (7.7)
7 (13.2)
14 (19.4)
0.316
18 (11.9)
3 (11.5)
4 (7.5)
11 (15.3)
0.419
11 (7.3)
2 (7.7)
0 (0)
9 (12.5)
0.029
7 (4.6)
0 (0)
1 (1.9)
6 (8.3)
**
7 (4.6)
2 (7.7)
2 (3.8)
3 (4.2)
**
6 (4)
1 (3.8)
2 (3.8)
3 (4.2)
**
5 (3.3)
3 (11.5)
2 (3.8)
0 (0)
**
Others, n (%)
42 (27.8)
13
Severity illness scoring systems for early identification and prediction of in-hospital mortality
513
original article
Fig. 3 By means of mortality in
emergency department sepsis
(MEDS) score calculated 28day mortality risk of patients
with suspected sepsis in the
emergency department. The
stacked bar chart shows the
respective disposition pattern
from ED for each mortality risk
group (low risk = MEDS score
07, moderate risk = MEDS
score 812, high risk = MEDS
score 13)
514 Severity illness scoring systems for early identification and prediction of in-hospital mortality
13
original article
Conclusion
This work shows that the MEDS score is able to detect critically ill sepsis patients under those with suspected sepsis in ED. Furthermore, in-hospital mortality of patients
with suspected sepsis can be predicted by means of the
MEDS score. The application of the MEDS score as risk
stratification tool in ED could therefore facilitate early
risk detection and contribute to a risk adjusted disposition management following objective criteria.
Acknowledgments
The authors thank Rafaela Kljaic for her expert technical
assistance.
Conflict of interest
The authors declare that they have no conflict of interest.
References
1. Hagel S, Brunkhorst F. Sepsis. Intensivmed. 2011;48:5773.
2. Engel C, Brunkhorst FM, Bone HG, et al. Epidemiology
of sepsis in Germany: results from a national prospective
multicenter study. Intensive Care Med. 2007;33:60618.
3. Bundesministerium fr Bildung und Forschung. Gesundheitsforschung: Forschung fr den Menschen. In: Newsletter No. 17. 2004. http://www.gesundheitsforschung-bmbf.
de/_media/NL_17.pdf. Accessed 29 Jan 2013.
4. Martin GS, Mannino DM, Eaton S, et al. The epidemiology
of sepsis in the United States from 1979 through 2000. N
Engl J Med. 2003;348:154654.
5. Center for Sepsis Control and Care. Was ist Sepsis? 2012.
http://www.cscc.uniklinikum-jena.de/Sepsis.html.
Accessed 13 Jan 2013.
6. Kumar A, Roberts D, Wood KE, et al. Duration of hypotension before initiation of effective antimicrobial therapy is
the critical determinant of survival in human septic shock.
Crit Care Med. 2006;34:158996.
7. Christ M, Grossmann F, Winter D, et al. Triage in der Notaufnahme. Moderne, evidenzbasierte Ersteinschtzung der
Behandlungsdringlichkeit. Dtsch Arztebl. 2010;107:8928.
13
Severity illness scoring systems for early identification and prediction of in-hospital mortality
515