Hi All,
I am working with a dataset that is looking to examine the effect of a policy change on the probability of transplantation among patients of different racial backgrounds. I used a competing risk regression analysis (Fine and Gray model) because there are other competing events like death or being removed from the waitlist due to becoming too sick. I've generated a lot interpretable data but I am struggling to figure out how to compare cumulative incidences of transplantation before and after the policy change for each race group. Here is some data.
1. Event - transplant (1); competing risk - waitlist removal (2); Pre-Policy period; unadjusted model
stset survival_time, failure(transplant_died_toosick==1)
stcrreg i.race if policy_period==0, compete(transplant_died_toosick==2)
SHR
2. Black | .9619023 .0974831 -0.38 0.702 .7886181 1.173263
Asian | .6681912 .0738626 -3.65 0.000 .5380318 .8298383
Hispa | .99846 .0616004 -0.02 0.980 .8847395 1.126798
Other | .744357 .1647499 -1.33 0.182 .4823733 1.148628
Interpretation: When compared to white patients, black patients have 4% lower probability of transplantation in the pre-policy era (SHR=0.96)
2. Event - transplant (1); competing risk - waitlist removal (2); Post-Policy period; unadjusted model
stset survival_time, failure(transplant_died_toosick==1)
stcrreg i.race if policy_period==0, compete(transplant_died_toosick==2)
SHR
2. Black | .9547059 .0580335 -0.76 0.446 .8474768 1.075502
Asian | .6172711 .0407567 -7.31 0.000 .5423423 .7025519
Hispa | .6792907 .0303931 -8.64 0.000 .6222586 .74155
Other | .7191012 .108336 -2.19 0.029 .5352448 .9661122
Interpretation: When compared to white patients, black patients have 5% lower probability of transplantation in the pre-policy era (SHR=0.95)
3. Now, I want to check the effect of the policy change, using an interaction term
stcrreg i.race##i.policy_period, compete(transplant_died_toosick==2)
SHR. p-val
2. Black | .955005 .0577595 -0.76 0.447 .8482508 1.075194
Asian | .6135333 .0403288 -7.43 0.000 .5393702 .6978937
Hispan | .6790761 .0301681 -8.71 0.000 .6224489 .740855
Other | .7139707 .1077011 -2.23 0.026 .5312248 .9595829
policy_period |
post-policy | .6519206 .0252123 -11.06 0.000 .6043317 .7032569
race#policy_period |
2. Black#post-policy | 1.007852 .1195527 0.07 0.947 .7987777 1.271649
Asian#post-policy | 1.089303 .1404977 0.66 0.507 .8459821 1.402608
Hispan#post-policy | 1.469301 .1120302 5.05 0.000 1.265345 1.706132
Other#post-policy | 1.040953 .2799749 0.15 0.881 .6144577 1.76348
4. Now, I am using lincom to compare SHR between groups. For example,
To compare cumulative incidence/probability of black patients to white patients in the pre-policy period.
lincom 2.race + 2.race#1.policy_period, shr
To compare cumulative incidence/probability of black patients to white patients in the post-policy period.
lincom 2.race + 2.race#1.policy_period, shr
To compare cumulative incidence/probability of transplant of black patients pre-policy to black-patients in post-policy.
lincom 2.race + 2.race #0.policy_period + 2.race#1.policy_period, shr
The challenge I have is that I want to generate interpretable cumulative incidence values at 1 year. SHR=0.96 does not tell me about an actual cumulative incidence, does it? Is there a way to print out survival tables so that I can compare 1-year cumulative incidence between pre-policy and post-policy of each race group?
Thank you!
Emile
I am working with a dataset that is looking to examine the effect of a policy change on the probability of transplantation among patients of different racial backgrounds. I used a competing risk regression analysis (Fine and Gray model) because there are other competing events like death or being removed from the waitlist due to becoming too sick. I've generated a lot interpretable data but I am struggling to figure out how to compare cumulative incidences of transplantation before and after the policy change for each race group. Here is some data.
1. Event - transplant (1); competing risk - waitlist removal (2); Pre-Policy period; unadjusted model
stset survival_time, failure(transplant_died_toosick==1)
stcrreg i.race if policy_period==0, compete(transplant_died_toosick==2)
SHR
2. Black | .9619023 .0974831 -0.38 0.702 .7886181 1.173263
Asian | .6681912 .0738626 -3.65 0.000 .5380318 .8298383
Hispa | .99846 .0616004 -0.02 0.980 .8847395 1.126798
Other | .744357 .1647499 -1.33 0.182 .4823733 1.148628
Interpretation: When compared to white patients, black patients have 4% lower probability of transplantation in the pre-policy era (SHR=0.96)
2. Event - transplant (1); competing risk - waitlist removal (2); Post-Policy period; unadjusted model
stset survival_time, failure(transplant_died_toosick==1)
stcrreg i.race if policy_period==0, compete(transplant_died_toosick==2)
SHR
2. Black | .9547059 .0580335 -0.76 0.446 .8474768 1.075502
Asian | .6172711 .0407567 -7.31 0.000 .5423423 .7025519
Hispa | .6792907 .0303931 -8.64 0.000 .6222586 .74155
Other | .7191012 .108336 -2.19 0.029 .5352448 .9661122
Interpretation: When compared to white patients, black patients have 5% lower probability of transplantation in the pre-policy era (SHR=0.95)
3. Now, I want to check the effect of the policy change, using an interaction term
stcrreg i.race##i.policy_period, compete(transplant_died_toosick==2)
SHR. p-val
2. Black | .955005 .0577595 -0.76 0.447 .8482508 1.075194
Asian | .6135333 .0403288 -7.43 0.000 .5393702 .6978937
Hispan | .6790761 .0301681 -8.71 0.000 .6224489 .740855
Other | .7139707 .1077011 -2.23 0.026 .5312248 .9595829
policy_period |
post-policy | .6519206 .0252123 -11.06 0.000 .6043317 .7032569
race#policy_period |
2. Black#post-policy | 1.007852 .1195527 0.07 0.947 .7987777 1.271649
Asian#post-policy | 1.089303 .1404977 0.66 0.507 .8459821 1.402608
Hispan#post-policy | 1.469301 .1120302 5.05 0.000 1.265345 1.706132
Other#post-policy | 1.040953 .2799749 0.15 0.881 .6144577 1.76348
4. Now, I am using lincom to compare SHR between groups. For example,
To compare cumulative incidence/probability of black patients to white patients in the pre-policy period.
lincom 2.race + 2.race#1.policy_period, shr
To compare cumulative incidence/probability of black patients to white patients in the post-policy period.
lincom 2.race + 2.race#1.policy_period, shr
To compare cumulative incidence/probability of transplant of black patients pre-policy to black-patients in post-policy.
lincom 2.race + 2.race #0.policy_period + 2.race#1.policy_period, shr
The challenge I have is that I want to generate interpretable cumulative incidence values at 1 year. SHR=0.96 does not tell me about an actual cumulative incidence, does it? Is there a way to print out survival tables so that I can compare 1-year cumulative incidence between pre-policy and post-policy of each race group?
Thank you!
Emile
Comment