Description

Ruiz-Gonzalez et al developed a bedside predictive model and scoring system to evaluate patients with community-acquired pneumonia. The goal is to identify the probable etiology for the pneumonia, with selection of initial empiric therapy. The authors are from the University Hospital of Lleida in Catalonia, Spain.


Diagnostic groups:

(1) virus-like: viral, Chlamydia pneumoniae, Mycoplasma pneumoniae, Coxiella burnetii

(2) bacterial: Streptococcus pneumoniae, Hemophilus influenzae, Streptococcus viridans, Streptococcus pyogenes, Enterococcus faecium

(3) unknown etiology (? everything else)

 

Parameters:

(1) age and/or comorbidity

(2) onset

(3) WBC count

 

Co-morbidities in patients:

(1) chronic bronchitis

(2) HIV infection

(3) chronic heart failure

(4) malignancy

(5) chronic liver disease

(6) diabetes mellitus

(7) stroke

(8) dementia

 

where:

• Other comorbid conditions (immunodeficiency, immunosuppression, splenectomy, etc.) were not listed.

 

Parameter

Odds Ratio

95% CI

age or comorbidity

6.9

2 – 23

abrupt onset

31

6 – 150

abnormal WBC count

2

0.6 – 7.0

 

 

Parameter

Finding

Points

age or comorbidity

age > 65 years, OR >= 1 comorbidities present

3

 

age <= 65 AND no comorbidities

0

onset

abrupt

5

 

not abrupt

0

WBC count

leukocytosis (>= 11,000 per µL)

2

 

normal WBC count (4,001 – 10,999 per µL)

0

 

leukopenia (<= 4,000 per µL)

2

 

bedside score =

= SUM(points for all 3 parameters)

 

Interpretation:

• minimum score: 0

• maximum score: 10

• The higher the score, the more likely that the patient has a bacterial pneumonia.

 

Different cut-off points for different goals:

(1) maximal sensitivity for bacterial pneumonia: >= 5

(2) maximal specificity for bacterial pneumonia: = 10

(3) maximal specificity for virus-like pneumonia: <= 3

 

Diagnostic Goal

Sensitivity (95% CI)

Specificity (95% CI)

maximal sensitivity for bacterial pneumonia

89% (78 - 96%)

63% (54 – 81%

maximal specificity for bacterial pneumonia

57% (40 – 74%)

94% (86 – 100%)

virus-like pneumonia

63% (48 – 77%)

89% (78 – 100%)

 

The area under the ROC curve was 0.84.

 

Limitations:

• This may apply to patients from the general population. I would not expect it to work as well in patients with highly specialized needs or risk factors (transplants, etc.).

• The differential is between bacterial and virus-like pneumonias, while fungal pneumonias are not represented.


To read more or access our algorithms and calculators, please log in or register.