To et al reported a model and data for predicting perioperative transfusion in elective hip and knee arthroplasty. One set of data was for patients who received a knee arthroplasty. The authors are from multiple institutions in Australia.
Patient selection: knee arthroplasty
Parameters:
(1) preoperative hemoglobin in g/dL
(2) sex
(3) age in years (less than 65 years; greater than or equal to 65 years)
(4) presence of both female sex and age >= 65 years
Category
Value of X Based on Figure 3
female, age LT 65 years
(-0.7042 * (hemoglobin)) + 6.41
female, age GTE 65 years
(-0.6727 * (hemoglobin)) + 7.308
male, age LT 65 years
(-0.731 * (hemoglobin)) + 7.1
male, age GTE 65 years
(-0.7041 * (hemoglobin)) + 6.895
probability of blood transfusion =
= 1 / (1 + EXP((-1) * X))
The general equation is shown on page 318 but the coefficients and intercept are not provided. Based on an analysis of the various models the following guesses at the values are:
Factor
Odds Ratio
Best Guess
a
7
b1
0.51
-0.68
b2
3.56
1.06
b3
-0.145
b4
-0.6
b5
3.37
1.3
Performance:
• The area under the ROC curve is 0.84.
To read more or access our algorithms and calculators, please log in or register.