Weiyi Peng
Research Mentor: Yun Jiang
Mentor Department: Not Available, Nursing
Author(s): Yun Jiang
Session: Session 5 (2:00 PM – 2:50 PM)
Presentation Type: Poster 55
Abstract
Adherence to adjuvant endocrine therapy (AET) is critical for improving breast cancer outcomes but remains suboptimal during long-term treatment. This study models adherence as a dynamic longitudinal process using linked electronic health records, pharmacy claims, and neighborhood socioeconomic data for 537 patients receiving tamoxifen or anastrozole. Recurrent neural network models were trained to predict future adherence risk. SHAP analysis identified prior adherence patterns as the strongest predictors, with demographic and clinical factors contributing smaller effects.
