How to get Relative Risk/ OR with 95% CI for PROC GENMOD with repeat observations

[sms1891]

Dear all,

I am trying to figure out how to get the RR/ OR with 95% CI for the same using repeated measure analysis using PROC GENMOD.

I have about 18 predictor variables, 15 of them are 0/1, with certain categories as control groups (please see my code below). Two variables have more than two levels and the last variable is a continuous variable. How do I get the RR/OR for all of the variables (categorical and continuous). I used lsmeans option but I can't get the continuous variable RR. Is there any other method to get the RR and ORs with 95% CI?

proc genmod data=Combined1; 
class  ID_No Race (ref="1") FebrileFlag (ref="1") Temp_HAI (ref="0") Temp_Grp (ref="1") Temp_Weekend (ref="0") Catheter_Use (ref="0") CentralLine_Use (ref="0") ELX_GRP_1 (ref="0") ELX_GRP_10 (ref="0") DM (ref="0")
ELX_GRP_15 (ref="0") ELX_GRP_14 (ref="0")  ELX_GRP_17 (ref="0") SIRS (ref="0") WBC_Grp (ref="0") Ward (ref="12") Disc_Status  (ref="1")/ param=glm;
model Relevant_BloodCult(event = '1') =Race FebrileFlag Temp_HAI Temp_Grp Temp_Weekend Catheter_Use CentralLine_Use ELX_GRP_1 ELX_GRP_10 DM ELX_GRP_15 
ELX_GRP_14  ELX_GRP_17 SIRS WBC_Grp Ward  Disc_Status LOS / dist=poisson link=log;
repeated subject = ID_No/ type=exch covb corrw;
lsmeans Race FebrileFlag Temp_HAI Temp_Grp Temp_Weekend Catheter_Use CentralLine_Use ELX_GRP_1 ELX_GRP_10 DM ELX_GRP_15 ELX_GRP_14  ELX_GRP_17 SIRS WBC_Grp Ward  Disc_Status  /diff exp cl;
run;

Please see the attached code. 

Any help with the SAS code would be greatly appreciated.

Thanks!

[sms1891]

Here are my other codes that I tried. param=ref in the first code. And estimate 1 -1/ exp in the second. But I am having issues with variable with more than 2 categories and continuous variable in the second set.

proc genmod data=Combined1; 
class  ID_No Race (ref="1") FebrileFlag (ref="1") Temp_HAI (ref="0") Temp_Grp (ref="1") Temp_Weekend (ref="0") Catheter_Use (ref="0") CentralLine_Use (ref="0") ELX_GRP_1 (ref="0") ELX_GRP_10 (ref="0") DM (ref="0")
ELX_GRP_15 (ref="0") ELX_GRP_14 (ref="0")  ELX_GRP_17 (ref="0") SIRS (ref="0") WBC_Grp (ref="0") Ward (ref="12") Disc_Status  (ref="1")/ param=ref;
model Relevant_BloodCult(event = '1') =Race FebrileFlag Temp_HAI Temp_Grp Temp_Weekend Catheter_Use CentralLine_Use ELX_GRP_1 ELX_GRP_10 DM ELX_GRP_15 
ELX_GRP_14  ELX_GRP_17 SIRS WBC_Grp Ward  Disc_Status LOS / dist=poisson link=log;
repeated subject = ID_No/ type=exch covb corrw;
lsmeans Race FebrileFlag Temp_HAI Temp_Grp Temp_Weekend Catheter_Use CentralLine_Use ELX_GRP_1 ELX_GRP_10 DM ELX_GRP_15 ELX_GRP_14  ELX_GRP_17 SIRS WBC_Grp Ward  Disc_Status  /diff exp cl;
run;

proc genmod data=Combined1; 
class  ID_No Race (ref="1") FebrileFlag (ref="1") Temp_HAI (ref="0") Temp_Grp (ref="1") Temp_Weekend (ref="0") Catheter_Use (ref="0") CentralLine_Use (ref="0") ELX_GRP_1 (ref="0") ELX_GRP_10 (ref="0") DM (ref="0")
ELX_GRP_15 (ref="0") ELX_GRP_14 (ref="0")  ELX_GRP_17 (ref="0") SIRS (ref="0") WBC_Grp (ref="0") Ward (ref="12") Disc_Status  (ref="1")/ param=ref ;
model Relevant_BloodCult(event = '1') =Race FebrileFlag Temp_HAI Temp_Grp Temp_Weekend Catheter_Use CentralLine_Use ELX_GRP_1 ELX_GRP_10 DM ELX_GRP_15 
ELX_GRP_14  ELX_GRP_17 SIRS WBC_Grp Ward  Disc_Status LOS / dist=poisson link=log;
repeated subject = ID_No/ type=exch covb corrw;
estimate "FebrileFlag" FebrileFlag 1 -1/ exp;
estimate "Temp_HAI" Temp_HAI 1 -1/ exp;
estimate "Temp_Grp" Temp_Grp 1 -1/ exp;

estimate "Temp_Weekend" Temp_Weekend 1 -1/ exp;
estimate "Catheter_Use" Catheter_Use 1 -1/ exp;
estimate "CentralLine_Use" CentralLine_Use 1 -1/ exp;

estimate "ELX_GRP_1" ELX_GRP_1 1 -1/ exp;
estimate "ELX_GRP_10" ELX_GRP_10 1 -1/ exp;
estimate "DM" DM 1 -1/ exp;
estimate "ELX_GRP_15" ELX_GRP_15 1 -1/ exp;
estimate "ELX_GRP_14" ELX_GRP_14 1 -1/ exp;
estimate "ELX_GRP_17" ELX_GRP_17 1 -1/ exp;
estimate "SIRS" SIRS 1 -1/ exp;
estimate "Ward" Ward 1 -1/ exp;
estimate "Disc_Status" Disc_Status 1 -1/ exp;
run;

[StatDave]

See this note about estimating relative risks and this note about estimating odds ratios in GENMOD.