Keighley et al reported a model for predicting mortality for a patient with candidemia. This can help to identify a patient who may require more aggressive management. The authors are from multiple institutions in Australia.
Patient selection: candidemia
Parameters:
(1) age in years
(2) location at the time of diagnosis of Candidemia
(3) history of prior surgery
(4) any hematological malignancy
(5) chronic organ (renal, heart, lung) dysfunction
(6) attributable source
(7) duration of antibiotic therapy in days
|
Parameter
|
Finding
|
Points
|
|
age in years
|
<= 65 years
|
0
|
|
|
> 65 years
|
1.5
|
|
location when diagnosed
|
ICU
|
1
|
|
|
other
|
0
|
|
history of prior surgery
|
no
|
0
|
|
|
yes
|
1
|
|
hematological malignancy
|
no
|
0
|
|
|
yes
|
1
|
|
chronic organ dysfunction
|
no
|
0
|
|
|
yes
|
1
|
|
attributable source
|
GI or unidentified
|
0
|
|
|
other
|
1
|
|
antibiotic therapy
|
< 10 days
|
0
|
|
|
>= 10 days
|
1
|
total source =
= SUM(points for all of the parameters)
Interpretation:
• minimum score: 0
• maximum score: 7.5
• A score >= 2 has a sensitivity of 95% and specificity of 48% (Youden 0.43).
• A score >= 3 has a sensitivity of 88% and specificity of 76% (Youden 0.64).
• A score > 2 had a mortality rate >= 20% versus < 20% if score <= 2.
Performance:
• The area under the ROC curve is 0.74.
