Need help with interpreting a logistic regression result with restricted cubic splines.

[J_Park]

I am trying to teach myself to run a logistic regression result with restricted cubic splines.  I found great information here and was able to build a SAS code.   I am using heart.csv from 

https://support.sas.com/documentation/onlinedoc/viya/examples.htm   I got the odds ratios but not sure how to interpret them.  I didn't set the reference, so I am not sure what the reference for these odds ratios was.   Is it possible to change the reference?

data heart1;

set heart;

if status = "Alive" then Death = 0;

else if status = "Dead" then Death = 1;

run;

proc logistic data=heart1; 

effect spl = spline(cholesterol / details naturalcubic basis=tpf(noint)                 

               knotmethod=percentilelist(10 50 90));  

model Death (event='1')= spl  / selection=none covb OUTROC=ROCDATA rocci;  

oddsratio cholesterol / at(cholesterol=174 223 285)

                  diff=ref cl=pl;

run;

[StatDave · Accepted solution]

The estimated odds ratios are the 1-unit changes in odds starting at each of the values you specified in the AT option. That is, the first estimate, 1.010, is the change in odds from 174 to 175. If what you want are the changes for more than 1 unit, then use the UNITS statement in addition to the ODDSRATIO statement. 

[J_Park]

Thank you so much!!!

[J_Park]

May I get a reference for this?  I need to explain this to my colleagues, and it would be great to have a reference.  Thank you!

[StatDave]

This is covered in "Odds Ratio Estimation" in the Details section of the PROC LOGISTIC documentation.