Wong et al developed a model for identifying clinically significant bleeding in an adult with dengue fever. This can help to identify a patient who may benefit from more aggressive management. The authors are from Tan Tock Seng Hospital, Nanyang Technological University, and the National University of Singapore in Singapore.
Patient selection: adult with dengue fever
Outcome: clinically significant bleeding
Parameters on admission:
(1) gender
(2) temperature in °C
(3) nausea and/or vomiting
(4) absolute neutrophil count (ANC) in 10^9/L
(5) absolute lymphocyte count (ALC) in 10^9/L
(6) hematocrit in percent from 0 to 100
(7) platelet count in 10^9/L
Parameters
|
Finding
|
Points
|
gender
|
female
|
0
|
|
male
|
-1.05
|
temperature
|
<= 38°C
|
0
|
|
> 38°C
|
0.59
|
nausea and vomiting
|
absent
|
0
|
|
present
|
0.3
|
Parameter
|
Points
|
absolute neutrophil count (ANC)
|
0.27 * (ANC)
|
absolute lymphocyte count (ALC)
|
0.9 * (ALC)
|
hematocrit
|
-0.04 * (hematocrit)
|
platelet count
|
-0.007*(platelet count)
|
where:
• Female gender had an adjusted odds ratio of 2.85. The LN(female) was 1.05.
• The absolute lymphocyte count had an adjusted odds ratio of 0.4. The LN(ALC) was -0.9.
X = log odds =
= SUM(points for all 7 parameters) - 0.04
probability of clinically significant bleeding =
= 1 / (1 + EXP((-1) * X))
Interpretation:
• A log odds > -3.919 was used to identify a patient at risk for significant bleeding. This corresponds to a risk of 2%.
Performance:
• The area under the ROC curve was 0.76.
• The sensitivity was 0.87 and the specificity 0.38. The negative predictive value was 0.98.