Hello, I am investigating the effect of a -0.25 unit decrease of my continuous variable has on survival. My model has the continuous variable (PRALBUM) and other dichotomous variables; no interaction terms. Since the HAZARDRATIO units option can only handle positive increments, I set units = 0.25. Next, I exponentiated the parameter estimate (0.97862) after multiplying by -1.0 to obtain the HR. To obtain the 95% CI, I multiplied the standard error (0.21116) by 1.0 and 1.96, then added/subtracted that value (0.41387) to/from the parameter estimate, and exponentiated those values. Final, result: HR = 2.66 (1.76, 4.02). I used the HAZARDRATIO statement to calculate the HR (95% CI) for my dichotomous variables. My question is: since I exponentiated the negative product of the parameter estimate of the continuous variable, do I need to do anything to the dichotamous variables to reflect the fact that I did that? Or, am I able to report the HR provided? I feel like I need to do something to those other estimates as they are only reflecting a 0.25 unit increase in the continuous variable, not the -0.25 unit decrease. Below is my code and Parameter Estimates: *Adjusted;
proc phreg data=c.nsqip;
class
ALBUM_NEW (PARAM = REF REF = "Normal")
PRSGOT_new (PARAM = REF REF = "<=40")
race_new (PARAM = REF REF = "White")
BMI_new (PARAM = REF REF = "Normal")
OPTIME_num (PARAM = REF REF = "0")
ASACLAS (PARAM = REF REF ="1-2")
PRHCT_new (PARAM = REF REF = ">=30")
PRPLATE_new (PARAM = REF REF = "<400K")
PRCREAT_new (PARAM = REF REF = "<=1.2")
WNDCLAS (PARAM = REF REF = "1-Clean")
cpt_group (PARAM = REF REF = "MIS")
DYSPNEA_num (PARAM = REF REF= "0");
model survival_time*failure(0) =
PRALBUM OTHBLEED_num cardiac_event_num Pulmonary_event_num infectious_event_num/ties=efron rl = pl details;
hazardratio PRALBUM/units=0.25 cl = pl;
hazardratio OTHBLEED_num/diff=pairwise cl = pl;
hazardratio cardiac_event_num/diff=pairwise cl = pl;
hazardratio Pulmonary_event_num/diff = pairwise cl = pl;
hazardratio infectious_event_num/diff=pairwise cl = pl;
run;
*Calculate HR and CI;
data albumin;
x = -1.0*(-0.97862);
y=1.96*1*0.21116;
HR = exp(x);
LB = exp(x-y);
UB = exp(x+y);
run;
proc print data= albumin;
run; Thank you
... View more
