I guess you are talking about classifying new observations. In proc logistic, use the outmodel option. This saves the fitted model information in a data set. For example,
proc logistic data=training_data outmodel=mylib.Fitted_Model;
....
....
Use the above data Fitted_Model to model your new dataset,
proc logistic noprint inmodel=mylib.Fitted_Model;
score data=New_Data out=Scored_Data;
....
Scored_Data will have the data and the predicted values.
Romakanta