Chase et al developed 3 models for predicting bacteremia in patients in the Emergency Department. One model predicted the probability of bacteremia without specifying the cause. The authors are from Beth Israel Deaconess Medical Center in Boston.
Patient selection: patient in the Emergency Department with suspected bacteremia
Parameters:
(1) suspected endocarditis
(2) suspected line infection
(3) suspected urinary source
(4) indwelling venous catheter
(5) vasopressors in the ED
(6) temperature
(7) respiratory failure
(8) platelet count
(9) percent neutrophils in the differential count
(10) bandemia
Parameters
Finding
Points
suspected endocarditis
no
0
yes
2.41
suspected line infection
no
0
yes
1.55
suspected urinary source
no
0
yes
0.83
indwelling venous catheter
no
0
yes
0.54
vasopressors in the ED
no
0
yes
0.66
temperature
normal
0
abnormal (< 96.8 or > 100.4°F)
0.50
respiratory failure
no
0
yes
0.32
platelet count
>= 150,000 per µL
0
< 150,000 per µL
0.66
percent neutrophils
<= 80%
0
> 80%
0.56
bandemia
no
0
yes (> 5%)
0.88
where:
• Points are beta-coefficients for the model.
X =
= SUM(points for the 10 parameters) - 3.12
probability of bacteremia =
= 1 / (1 + EXP((-1) *X))
Performance:
• The area under the ROC curve was 0.71.
To read more or access our algorithms and calculators, please log in or register.
