Using multiple correlation nod linear regression approaches, we investigate
d the association between the amount of mediastinal drainage for the first
24 postoperative hotels and clinical variables as well as multiple haematol
ogical tests performed at three time points: before anaesthesia induction,
10 minutes after protamine administration and just after skin closure, on 4
6 patients undergoing primary coronary artery bypass grafting. Three models
from the three times were then developed to predict mediastinal drainage.
The number of internal mammary grafts, the total number of grafts and plasm
a fibrinogen concentration were useful predictors of mediastinal drainage a
t all three times. The platelet count taken only after skin closure was fou
nd to provide additional predictive information. Each regression model expl
ained approximately 60% of the variation in postoperative mediastinal drain
age. The information obtained front these predictive models is useful in de
fining high-risk populations.