Model of Karkouti et al for Predicting the Need for Blood Transfusion Associated with a First-Time Elective Coronary Bypass Graft Operation

Populations: Elderly adults requiring first-time CABG procedures at Toronto General Hospital between January 1, 1997 and September 30, 1998. Patients who were excluded: patients with autologous blood, patients requiring re-exploration for bleeding, and patients who were transfused after the first post-operative day (page 1195).

Parameters from multivariate analysis:

(1) age in years

(2) sex (male = 1, female = 0)

(3) body weight in kilograms

(4) preoperative hemoglobin in g/dL

X =

= (-0.7064 * (points for sex)) + (0.0366 * (age in years)) – (0.0641 * (body weight in kilograms)) – (0.743 * (preoperative hemoglobin) + 12.3189

probability of requiring transfusion =

= 1 / (1 + EXP((-1) * X))

The authors selected a probability of 20% (0.20) as the point for predicting the need for transfusion (sensitivity 82%, specificity 71%).

Based on this decision, they created tables for men (Table 8) and women (Table 9) with the pre-operative hemoglobin associated with the need for transfusion using age and weight as axes.

However, if one reworks the equation (using probability = 0.2) , one can derive the following equation:

pre-op hemoglobin in g/dL below which transfusion is likely =

= (1.3863  – (0.7064 * (points for sex)) + (0.0366 * (age in years)) – (0.0641 * (body weight in kilograms)) + 12.3189) / 0.743