Hi,

I want to find a way to turn binary logistic regression model's coefficients/odds ratios to a scorecard for the occurrence of y. The problem is as:

"1000 patients were treated with a procedure j, and after 1-year follow-up we learned that 23 died in first quarter of the year, 6 died in the second quarter, 14 more by third quarter and 5 more by the last quarter. At the end of the year, 48 people out of 1000 have died. On the basis of age [continuous] (x1), gender [male=0, female=1] (x2), race [caucasian=1, African american=2, asian=3, others=4] (x3), BMI [underweight=0, normal=1, overweight=2, obese=3] (x4), access to care status [community clinic=0, urban health center=1, tertiary care hospital=2, academic hospital=3] (x5), a regression model is fit for the occurrence of overall mortality in first quarter (23 out of 1000) and 1-year mortality (48 out of 1000). The model provided us with coefficients for each value with a reference to "0" of each variable".

Question 1: How to convert the regression equation (where y=0 or 1) to a scorecard out of n score (e.g. out of 20)?
Question 2: How to make a score that can predict the % of occurrence of the event y? (e.g. Caucasian Obese Male Patient treated with procedure j with age 60 having access to academic hospital is 15% probability of death)
Question 3: How can I consider working on bootstrapping in this situation? How will it help me?
Question 4: After building a scorecard, what is the best way to validate it? Use another dataset to check or run through other similar data?

Help will be appreciated.

Thanks.