Issue:
A pharmaceutical client approached DMS to evaluate whether the use of a specific anti-thrombotic agent was related to secondary prevention of thromboembolic complications associated with symptomatic peripheral artery disease (PAD) or was used as primary prevention of thromboembolic complications after a stent procedure among patients that also had concomitant coronary artery disease (CAD) with PAD.
Solution:
- We constructed a Classification Tree (CT) model which is a structured machine learning approach, along with a logistic regression approach was used to classify key variables associated with initiation of the anti-thrombotic drug.
- The results of this analysis showed that PAD patients with comorbid CAD and a recent history of a stenting procedure were more likely to initiate the anti-thrombotic drug, suggesting that comorbid CAD and prior stenting could confound effectiveness analyses associated with the anti-thrombotic drug and should be carefully controlled for.
- This research was presented at the ISPOR 2015 annual meeting in the U.S. References available upon request.