"Competing risk" which increases probability of outcome

Hi, I have been looking up posts regarding competing risks in regression modelling and am hoping for some advice.

Using a dataset of patients with cardiovascular risks at baseline, I am aiming to model a composite outcome of 3 variables - 1) atrial fibrillation incidence (diagnosis), 2) hospitalisation for rhythm-related disturbances, and 3) cardiovascular death, at 2 years follow-up.

To my understanding, a competing risk would be an event which would modify the probability of experiencing / precludes the occurrence of one of the 3 events in the composite outcome. Therefore, I intend to include non-cardiovascular deaths as a competing risk.

It was brought to my attention that there are patients in the cohort who have implanted cardiac devices - pacemakers - during the follow-up period. It is known that patients with implanted devices are more commonly diagnosed with atrial fibrillation because the device enables continuous monitoring as compared to patients without devices who need to rely on awareness of symptoms.

So, I now have a variable (implanted device: Yes/No) which modifies the probability of experiencing one of the events in the composite outcome, but instead of precluding the occurrence, it inflates the probability of the event.

I wonder if anyone has thoughts on how to handle this variable? Would it be appropriate to consider this a “competing risk”? Or perhaps an adjustment variable in the regression model?

Many thanks in advance!

2 Likes

If the objective of the study was the effect of the implanted device on atrial fibrillation, the exposure would be a time-dependent variable. I have also asked myself several times how to analyze that and it has not been clear to me. Let’s see if some expert can clarify this for us clinicians. It is a model that represents common situations.

This may be a place for a state transition model. A nice resource is from Therneau.

1 Like

Yes, it seems that Therneau explains it very well there. The truth is that this man writes very nice and useful tutorials. It is very grateful that besides explaining the mathematical properties he teaches us how to use these models.

1 Like