b r a z j i n f e c t d i s .

2 0 1 3;1 7(2):179–183

The Brazilian Journal of

INFECTIOUS DISEASES
www.elsevier.com/locate/bjid

Original article

Pneumonia severity index compared to CURB-65 in

predicting the outcome of community acquired pneumonia
among patients referred to an Iranian emergency
department: a prospective survey

Mostafa Alavi-Moghaddam a , Hooman Bakhshi a , Bareza Rezaei a , Patricia Khashayar b,∗

a Imam Hossein Hospital, Shahid Beheshti University of Medical Sciences, Tehran, Iran
b Endocrinology and Metabolism Research Center, Tehran University of Medical Sciences, Tehran, Iran

a r t i c l e i n f o a b s t r a c t

Article history: Objective: To compare the prognostic value of the pneumonia severity index and the sever-
Received 15 November 2011 ity score for community-acquired pneumonia (CURB-65) in predicting mortality and the
Accepted 3 October 2012 need for ICU admission of patients with community-acquired pneumonia referred to our
Available online 28 February 2013 emergency department.
Materials and methods: This prospective study was performed on patients with community-
Keywords: acquired pneumonia admitted to the emergency department of Imam Hossein Medical
CURB-65 Center, Tehran, Iran. A questionnaire with demographic information, clinical signs and
PSI symptoms, laboratory and radiographic findings was completed for each patient. The infor-
Community-acquired pneumonia mation required for calculating the pneumonia severity index and CURB-65 were extracted
from the medical records. The patients’ clinical outcome was also recorded within a month
after admission.
Results: We studied 200 patients with community-acquired pneumonia (122 men, 78
women). The sensitivity and specificity of CURB-65 in predicting mortality were 100% and
82.3%, respectively. As for pneumonia severity index, the rates were 100% and 75%, respec-
tively. The sensitivity and specificity rates of CURB-65 and pneumonia severity index in
predicting mortality and need for ICU admission were 96.7% and 89.3%, and 90% and 78.7%,
respectively.
Conclusion: CURB-65 seems to be the preferred method to predict mortality and need for
ICU admission in patients with community-acquired pneumonia. Despite their comparable
specificity and sensitivity, CURB-65 is much easier to implement.
© 2013 Elsevier Editora Ltda. All rights reserved.

Corresponding author at: Endocrinology and Metabolism Research Center, Shariati Hospital, Kargar St, 14114 Tehran, Iran.
∗

E-mail address: [email protected] (P. Khashayar).

1413-8670/$ – see front matter © 2013 Elsevier Editora Ltda. All rights reserved.
http://dx.doi.org/10.1016/j.bjid.2012.10.012
180 b r a z j i n f e c t d i s . 2 0 1 3;1 7(2):179–183

ical board committee of Shahid Beheshti University of Medical

Introduction Sciences, Tehran, Iran, approved this study.

Community acquired pneumonia is a common disease with

Study population
a lifetime prevalence of 20–30% in developing and 3–4% in
developed countries.1 The mortality rate among hospital-
Patients aged 18 years or more diagnosed with community
ized patients diagnosed with the condition has been reported
acquired pneumonia on the grounds of their clinical and par-
to range from 4% to 21% in different settings.1 The rate is
aclinical findings by the emergency and/or infectious disease
reported to be higher, even as high as 50%, among patients
residents and/or specialists were enrolled in the study. Those
admitted to ICU.2
whose diagnosis changed during the course of treatment were
The condition imposes a heavy burden on the healthcare
excluded.
system in terms of its high cost both for diagnosing and treat-
ing the condition as well as for the hospital and ICU stay.3
Study protocol
This heavy cost points out the importance of predicting the
need for hospitalization (ICU or ward) as well as the outcome
A questionnaire including the demographic data, clinical, par-
of these patients.2
aclinical and imaging findings of the patients was completed
One of the earliest studies of outcome in community
for each patient. PSI and CURB-65 scores were calculated
acquired pneumonia was conducted in 1982 by the British
for each patient. The CURB-65 is a 5-item index while PSI
Thoracic Society. In this study 453 adult patients were evalu-
uses 20 items to predict the patient’s outcome. An emergency
ated in 25 medical centers. Patients who had two of the three
medicine resident was responsible for filling out the question-
risk factors were described as being 21 times more likely to die
naires.
because of their condition. Based on these findings, the British
The need for ICU stay as well as the risk of dying was com-
Thoracic Society devised the BTS1 scoring system which has
pared according to the calculated PSI and CURB-65 CPRs. The
shown a high predictive value compared to other indices in
outcome of patients was also recorded within a month after
this regard.4,5
their admission at the hospital.
In 1997, Fine et al. devised the pneumonia severity index
(PSI) method based on their study of more than 50,000
Data analysis
outpatients and also hospitalized patients diagnosed with
community acquired pneumonia. This scoring system, which
Data were analyzed using SPSS ver 13. Results were expressed
divided the patients into five main categories based on their
as mean ± SD. Student’s t-test was used for statistical compar-
30-day prognosis prediction, also showed acceptable predic-
isons. Sensitivity, specificity, and relative risks were calculated
tive values.6
for each study outcome using standard formulas.
The main concern while designing PSI was the manage-
ment of outpatients at high risk of dying from the disease.
The predictive value of PSI in these patients was later con- Results
firmed by several other studies.5–7 The main shortcoming of
PSI was the fact that age was not a predictive component in Two hundred patients with community-acquired pneumonia
this index. Later on, PSI scoring system was recommended by were enrolled (122 males, 78 females). Their mean age was
the Infectious Diseases Society of America as a predicting tool 68 ± 18 years, ranging from 18 to 68 years.
for patients with community acquired pneumonia.6 Among these patients 148 (74%) were hospitalized in dif-
The severity score for the community acquired (CURB-65) ferent hospital wards and 52 (26%) were admitted to ICU. The
and the PSI are two of the most prominent methods in this most common cause of the condition in males under the age
regard.2,3,8–12 The efficacy of these two clinical prediction rules of 50 was injection drug abuse and high blood glucose (Fig. 1).
(CPRs), however, had never been compared in Iran or in any In females of the same age, however, viral diseases (influenza)
other regional countries. and high blood glucose were the prevailing causes. The most
Considering the high prevalence of community acquired
pneumonia in Iran, the present study was designed to com- The most prevalent underlying conditions
pare the prognostic value of these two CPRs in our emergency
Urea (65) 25
department (ED) as a model representing EDs in developing 38%

countries. Respiratory rate (106) 21

19.70%
Systolic blood pressur (113) 26
23.00%

Patients and methods PH (10)

90%
9

Cercbrovascular diseas (63) 12

17.40%
Study design Heart failur (73) 20
28.80%
Confusion (76) 34
This observational comparative study was conducted on 44.80%
Plural effusion (20) 6
patients diagnosed with community acquired pneumonia, 30%
referred to the ED of Imam Hossein Medical Center in 2009.
Imam Hossein hospital is a teaching general hospital where Fig. 1 – The most prevalent underlying conditions in the
more than 70,000 patients are seen in its ED annually. The eth- studied population.
 b r a z j i n f e c t d i s . 2 0 1 3;1 7(2):179–183 181

Table 1 – Distribution of patients’ outcome based on PSI class.

Variable PSI class

I II III IV V

Death 0 (0) 0 (0) 0 (0) 0 (0) 36 (100)

Cure 4 (2.4) 3 (1.8) 13 (7.9) 103 (62.8) 41 (25)

Total 4 (2) 3 (1.5) 13 (6.5) 103 (51.5) 77 (38.5)

Table 2 – Distribution of patients’ outcome based on CURB-65 class.

Variable CURB-65 class

I II III IV V

Death 0 (0) 0 (0) 0 (0) 0 (0) 36 (100)

Cure 4 (2.4) 3 (1.8) 13 (7.9) 103 (62.8) 41 (25)

Total 4 (2) 3 (1.5) 13 (6.5) 103 (51.5) 77 (38.5)

Table 3 – Sensitivity, specificity, positive predictive value, negative predictive value, and diagnostic accuracy of PSI and
CURB 65 methods in forecasting mortality.
Class Sensitivity Specificity PPV NPV DA

PSI
II 100 (90.4, 100) 2.4 (0.9, 6.1) 18.4 (13.6, 24.4) 100 (51.0, 100) 20 (15.1, 26.1)
III 100 (90.4, 100) 4.3 (2.1, 8.5) 18.7 (13.8, 24.7) 100 (64.6, 100) 21.5 (16.4, 27.7)
IV 100 (90.4, 100) 12.2 (8.0, 18.1) 20 (14.8, 26.4) 100 (83.4, 100) 28.0 (22.2, 34.6)
V 100 (90.4, 100) 75 (67.9, 81.0) 46.8 (36.0, 57.8) 100 (98.0, 100) 79.5 (73.4, 84.5)

CURB-65
I 100 (90.4, 100) 0.6 (0.1, 3.4) 18.1 (13.4, 24.0) 100 (20.6, 100) 18.5 (13.7, 24.5)
II 100 (90.4, 100) 5.5 (2.9, 10.1) 18.9 (13.9, 25.0) 100 (70.1, 100) 22.5 (17.3, 28.8)
III 100 (90.4, 100) 82.3 (75.8, 87.4) 55.4 (43.4, 66.8) 100 (97.2, 100) 85.5 (80.0, 89.7)
IV 75 (58.9, 86.3) 97.0 (93.1, 98.7) 84.4 (68.3, 93.4) 94.6 (90.1, 97.2) 93.0 (88.6, 95.8)
V 11.1 (4.4, 25.3) 99.4 (96.6, 99.9) 80 (37.6, 96.4) 83.6 (77.8, 88.1) 83.5 (77.7, 88)

common underlying condition in the whole population was Six of the 14 patients who were admitted in a ward, instead of
heart failure. ICU, died.
Overall, 36 patients died (18%) during the study period; of Based on PSI scoring system, the majority of discordances
those, six had been hospitalized in different wards and the were reported in patients categorized as classes 3–5, with
remaining 30 were admitted to ICUs. Thirty patients died in the highest occurring with patients of class 5. Among the 30
hospital, whereas six died in the 30-day follow-up period after patients who should have been admitted to ICU but were hos-
discharge. pitalized in a ward, there were six deaths during the study
The average hospital length of stay for patients with period.
community-acquired pneumonia categorized as classes 1 and Sensitivity, specificity, positive and negative predictive
2 based on the PSI scoring system was two days, three days values, and diagnostic accuracy of CURB-65 in predicting mor-
for class 3, five days for class 4, and 10.5 days for class 5. For tality were 100%, 82.3%, 55.4%, 100% and 85.5%, respectively.
CURB-65, these values were two days for classes 1 and 2, 14.5 The same rates for PSI were 100%, 75%, 46.8%, 100% and 79.5%,
days for class 3, and nine days for higher classes. respectively (Table 3). Table 4 shows the efficacy of the two
Heart failure and age were both significantly associated tools in forecasting ICU admission (Table 4).
with mortality (p < 0.05). Among clinical and paraclinical Receiver operating characteristic (ROC) curves for the pre-
findings, having cardiovascular disease as the underlying con- diction of mortality and need for ICU admission using PSI and
dition, low blood pH and high urea levels, and decreased CURB-65 are shown in Figs. 2 and 3, respectively. CURB-65
consciousness level were statistically correlated with mortal- showed a better predictive value in foreseeing both the need
ity (p < 0.05). for ICU admission and mortality than PSI.
The distribution of patient’s outcome based on their PSI
and CURB-65 scores is shown in Tables 1 and 2.
All 36 deaths occurred in patients with PSI levels of 5 and Discussion
higher. There was 12% discordance between decisions on the
need for hospitalization or the type of ward the patient was Although there are strong similarities between these two
to be admitted when using CURB-65. As for PSI, the rate was methods at first glance, important differences make them
high as 27.5%. The highest discordance rate was reported in unique. PSI uses a long list of predicting factors and its imple-
patients categorized as class three and higher using CURB-65. mentation needs various clinical and paraclinical information
182 b r a z j i n f e c t d i s . 2 0 1 3;1 7(2):179–183

Table 4 – Sensitivity, specificity, positive value, negative predictive value and diagnostic accuracy and their relative 95% CI
of PSI and CURB-65 methods in forecasting ICU admission.
Class Sensitivity Specificity PPV NPV DA

PSI
II 100 (92.9, 100) 2.7 (1.0, 6.7) 25.5 (19.9, 32.0) 100 (51.0, 100) 27.0 (21.3, 33.5)
III 100 (92.9, 100) 4.7 (2.3, 9.3) 25.9 (20.2, 32.5) 100 (64.6, 100) 28.5 (22.7, 35.1)
IV 100 (92.9, 100) 13.3 (8.8, 19.7) 27.8 (21.8, 34.7) 100 (83.9, 100) 35 (28.7, 41.8)
V 90 (78.6, 95.7) 78.7 (71.4, 84.5) 58.4 (47.3, 68.8) 95.9 (90.8, 98.3) 81.5 (75.5, 86.3)

CURB-65
I 100 (88.7, 100) 0.7 (0.1, 3.7) 16.8 (12.0, 22.9) 100 (20.7, 100) 17.2 (12.4, 23.4)
II 100 (88.7, 100) 6 (3.2, 11.0) 17.5 (12.6, 23.9) 100 (70.1, 100) 21.7 (16.3, 28.4)
III 96.7 (83.3, 99.4) 89.3 (83.4, 93.3) 64.4 (49.8, 76.8) 99.3 (95.3, 99.9) 90.6 (85.4, 94.0)
IV 30 (16.7, 47.9) 98.0 (94.3, 99.3) 75.0 (46.8, 91.1) 87.5 (81.7, 91.7) 86.7 (80.9, 90.9)
V 16.7 (7.3, 33.6) 100 (97.5, 100) 100 (56.6, 100) 85.7 (79.8, 90.1) 86.1 (80.3, 90.4)

ROC curve while CURB-65 is designed to be as simple as possible using a

limited set of information.
Source of the curve
Based on the nature of these two tools, their predictive
PSI class Reference line
CURB-65 value largely depends on the environment in which they are
implemented. In a hospital setting in developing countries like
1.0
Iran with scarce resources, simple methods such as CURB-65
0.8 are preferred as they put less pressure on the country.
In line with previous studies, both PSI and CURB-65
Sensitivity

0.6 showed high negative predictive value and low positive pre-
0.4
dictive value in predicting mortality and the need for ICU
admission.6,7 In our results, however, CURB-65 had better
0.2 accuracy in predicting mortality and the need for ICU admis-
sion among patients with community-acquired pneumonia.
0.0
0.0 0.2 0.4 0.6 0.8 1.0
While CURB-65 had a high sensitivity in predicting mortality
and need for ICU admission, PSI was shown to have a high
1 - specificity
specificity in this regard. Both indices can therefore be used
Diagonal segments are produced by ties.
even in low risk patients as a guide for early discharge.
Fig. 2 – ROC curves for mortality prediction using PSI and According to other studies, the mortality risk and the need
CURB-65. for ICU admission were higher as the scores increased in both
PSI and CURB-65.4,7 Our study, similarly, revealed that the
mortality increased with age, presence of underlying heart
failure, high blood levels of urea, pH lower than 7.35, and
decreased consciousness level. The most common underlying
ROC curve
condition in this study was heart failure, which had a statis-
Source of the curve tically significant relation with mortality, whereas decreased
PSI class Reference line consciousness level was associated with higher mortality.
CURB-65
Musher et al. in a study on 170 patients with community-
1.0 acquired pneumonia found heart conditions namely CHF in
33 (19.7%) of the patients.13 Corroborating our results, Lich-
0.8 man et al. reported that 6.8% of their patients had severe heart
diseases.14
Sensitivity

0.6 In a study by Man et al. on 1016 patients with community-

acquired pneumonia, authors compared the outcome of
0.4
patients categorized as level 4 and 5 using CURB-65 and PSI
methods. Their results indicated that being level 5 by CURB-65
0.2
had the best predictive value for patients with community-
0.0 acquired pneumonia admitted to the ED.15 In line with our
0.0 0.2 0.4 0.6 0.8 1.0 results, Shah et al. reported both PSI and CURB-65 to have
1 - specificity equal sensitivity to predict death from community-acquired
Diagonal segments are produced by ties. pneumonia, adding that PSI was more sensitive in predicting
ICU admission than CURB-65.1 This may be because CURB-65
Fig. 3 – ROC curves for predicting ICU admission using PSI model does not consider decompensated co-morbidity due to
and CURB-65. community-acquired pneumonia and results in limited appli-
cation in the elderly.16 In another study PSI was reported to
 b r a z j i n f e c t d i s . 2 0 1 3;1 7(2):179–183 183

have the highest sensitivity followed by CURB-65 in predict- 4. Anonymous. Community-acquired pneumonia in adults in
ing mortality. For predicting ICU admission, however, other British hospitals in 1982–1983: a survey of a etiology,
indices such as modified ATS, SMART-COP and IDSA/ATS were mortality, prognostic factors and outcome. Br Thoracic Soc
Public Health Lab Serv Q J Med. 1987;62:195–220.
reported to perform better than PSI and CURB-65,17 as these
5. Aujesky D, Fine MJ. The pneumonia severity index: a decade
indices were originally designed to assess ICU admission after the initial derivation and validation. Clin Infect Dis.
rather than mortality. Therefore, a poor performance could 2008;47 Suppl. 3:S133–9.
be found if applied in predicting mortality. 6. Fine MJ, Auble TE, Yealy DM, et al. A prediction rule to identify
low-risk patients with community-acquired pneumonia. N
Limitation Engl J Med. 1997;336:243–50.
7. Capelastegui A, Espana PP, Quintana JM, Areitio I, Gorordo I,
Egurrola M. Validation of a predictive rule for the
Considering the limited number of ICU beds in our hospital,
management of community-acquired pneumonia. Eur Respir
it is possible that certain patients were admitted to different J. 2006;27:151–7.
wards due to unavailability of ICU beds. Not having an avail- 8. American Thoracic Society, Infectious Diseases Society of
able ICU bed may have affected the physician’s decision on America. Guidelines for the management of adults with
whether the patient needed ICU admission. hospital-acquired, ventilator-associated, and
healthcare-associated pneumonia. Am J Respir Crit Care Med.
2005;171:388–400.
Conclusion 9. Mandell LA, Wunderink RG, Anzueto A. management of
community-acquired pneumonia in adults. Clin Infect Dis.
The severity score for community-acquired pneumonia seems 2007;44:27–40.
10. Charles PG, Wolfe R, Whitby M, Fine MJ. SMART-COP: a tool for
to be the preferred method to predict the need for ICU admis-
predicting the need for intensive respiratory or vasopressor
sion and the prognosis of patients seen at EDs. Despite having support in community-acquired pneumonia. Clin Infect Dis.
comparable specificity and sensitivity with PSI, CURB-65 is 2008;47:375–84.
much easier to be implemented. 11. Yu KT, Wyer PC. Evidence behind the 4-hour rule for initiation
of antibiotic therapy in community-acquired pneumonia.
Ann Emerg Med. 2008;51:651–60.
Conflict of interest 12. Waterer GW, Kessler LA, Wunderink RG. Delayed
administration of antibiotics and atypical presentation in
All authors declare to have no conflict of interest. community-acquired pneumonia. Chest. 2006;130:11–5.
13. Musher DM, Rueda AM, Kaka AS. The association between
pneumococcal pneumonia and acute cardiac events. Clin
references
Infect Dis. 2007;45:158–65.
14. Lichtman JH, Spertus JA, Reid KJ. Acute non-cardiac
conditions and in-hospital mortality in patients with acute
1. Shah BA, Ahmed W, Dhobi GN, Shah NN, Khursheed SQ, Haq myocardial infarction. Circulation. 2007;116:1925–30.
I. Validity of pneumonia severity index and CURB-65 severity 15. Man SY, Lee N, Antonio GE, Chauss M. Prospective
scoring systems in community acquired pneumonia in an comparison of three predictive rules for assessing severity
Indian setting. Indian J Chest Dis Allied Sci. 2010;52:9–17. of community-acquired pneumonia. Eur Respir J.
2. Lim WS. Defining community acquired pneumonia severity 2006;27:30–5.
on presentation to hospital: an international derivation and 16. Niederman MS. Making sense of scoring systems in
validation study. Thorax. 2003;58:377–82. community acquired pneumonia. Respirology (Carlton, Vic).
3. Moran GJ, Talan D. Pneumonia. In: Marx JA, Hockberger RS, 2009;14:327–35.
Walls RM, et al., editors. Rosen’s emergency medicine: 17. Fang WF, Yang KY, Wu CL, et al. Application and comparison
concepts and clinical practice. 7th ed. Philadelphia: Elsevier; of scoring indices to predict outcomes in patients with
2009. p. 627–938. healthcare associated pneumonia. Crit Care. 2011;15:R32.