Abstract: Frequent sequential patterns of antidepressant medication changes were obtained through the use of electronic health record data from Weill Cornell Medicine (WCM) for 35,846 patients. SPADE algorithm was applied to mine sequential patterns at drug class and generic name levels. Antecedent- consequent rule Induction pairs were formed and digraphs made for highest supported changes. The method effectively identifies temporal relationships between medications and has diverse applications for patient clustering and elucidating adverse drug reactions.

Learning Objective 1: To identify various frequent temporal associations between changes in antidepressants at an outpateint care facility

Learning Objective 2 (Optional): To be able to apply sequential pattern mining algorithms on Electronic Health Records

Learning Objective 3 (Optional): Understand the advantages of using machine learning pattern mining techniques in the health field for visualizing treatment regimes and to inform clinical prescribing patterns.


Kartikey Grover, Weill Cornell Medicine
Joseph DeFerio, Weill Cornell Medicine
Min-hyung Kim, Weill Cornell Medicine
Amit Sheth, Wright State Univeristy
Samprit Banerjee, Weill Cornell Medicine
Jyotishman Pathak (Presenter)
Weill Cornell Medicine

Presentation Materials: