PROC Causalmed- Natural Indirect Effect value inverted

G_Scheiffele — Thu, 05 Dec 2024 16:42:28 GMT

Hi all!

I am utilizing PROC CAUSALMED to assess the role of vascular access modality (Central venous catheter against all other access types) on survival of one-year post dialysis. I am using SAS 9.4 on desktop

Here are the levels which I've used in PROC Causalmed:

Binary categorical exposure (Predialysis nephrology care (Y/N))

Binary categorical Mediator (Central venous catheter vs all else)

Binary categorical Outcome (Survival of one year post-dialysis (Y/N))

The model suggested is predialysis care -> access type -> odds of one-year survival

The individual associations have significant associations and directionality, as used within PROC LOGISTIC, to prove the individual associations.

Predialysis Care (Y) is associated with better odds of survival compared to no predialysis care.

Predialysis care (Y) is associated with lower central venous catheter use compared to no predialysis care.

Central venous catheter is associated with lower odds of survival compared to all other vascular access modalities.

Here is the sample code I am using to produce this output:

PROC CAUSALMED DATA = DF order = freq rorder = internal;

CLASS one_yr_mort vasbin nephcarebin;

MODEL one_yr_mort = nephcarebin vasbin nephcarebin*vasbin;

MEDIATOR vasbin = nephcarebin;

run;

From this, I get a general output such as this:

Odds Ratio Total Effect : 1.7928

Odds Ratio Natural Direct Effect: 1.3560

Odds Ratio Natural Indirect Effect: 1.3221

...

% Mediated: 55.0955

To my understanding,

This output is suggesting that compared to no nephrology care, individuals with predialysis nephrology care have a 79% greater odds of survival through one year. Directly, there is ~35% greater odds, and indirectly, by the disparate influence of all other vascular access types compared to Central venous catheter, nephrology care has a survival advantage of 32%. This leads to the product total effect, of which 55% of the variation is mediated by all other vascular modalities compared to CVC.

I am interested in describing to what extent central venous catheter, compared to all other modalities, is detrimental to survival outcome for this population.

What should I change to be able to be able to elucidate a potential detrimental effect of central venous catheter on survival in the context of the exposure being predialysis care?

Sincere thanks,

Grant

topic PROC Causalmed- Natural Indirect Effect value inverted in Statistical Procedures

PROC Causalmed- Natural Indirect Effect value inverted