Harrell's C indexes in competing risk model

[haoduonge]

Hi,

Hi all,

Can we calculate the Harrell's C indexes in competing risk model as below?

Thank you!

Hao

proc phreg data=out_withclass;

class X1(REF='0');

model  TIME_YRS *OUTCOME(0)=X1 X2/eventcode=1 rl;

run;

[StatDave]

Use the CONCORDANCE option in the PROC PHREG statement. You can also use the PLOTS=ROC, PLOTS=AUC, and ROCOPTIONS to draw and control ROC curves. See the PHREG documentation for discussion of these options and an example. 

[haoduonge]

I think it only works with Cox model, not with competing risk model (F & G), right?

Hao

[StatDave]

Sorry... that's correct. The concordance statistics (and ROC analysis) are not available with that model.

[OsoGris]

You can try using the  CONCORDANCE option on the PROC PHREG statement.  I don't know if that's enabled, necessarily, for the competing risks analysis.