Li et al reported a model for predicting immune thrombocytopenia. The authors are from multiple institutions across Canada.
Patient selection: thrombocytopenia
Definition for ITP: platelet count < 100 * 10^9/L with response to either high-dose corticosteroids or IV immune globulin (to count above 50 * 10^9/L and >= 2 times baseline)
Parameters:
(1) lowest platelet count in 10^9/L
(2) platelet count variability
(3) maximum platelet volume in fL
(4) history of major bleed based on ITP bleeding scale
The platelet count variability is derived in Supplementary Table 4, It is defined as:
PCV =
= LOG10((sum square of platelet count) + (sum square of changes in platelet counts) + 0.001)
|
Parameter
|
Finding
|
Points
|
|
lowest platelet count
|
|
-0.034 * (count)
|
|
platelet count variability
|
|
0.151 * (variability)
|
|
max platelet volume
|
|
0.222 * (volume)
|
|
history of major bleed
|
no
|
0
|
|
|
yes
|
0.874
|
value of X =
= SUM(points for all of the parameters) - 5.18
probability of ITP =
= 1 / (1 + EXP((-1) * X))
Performance:
• The area under the ROC curve is 0.83.
