Yirui (Iris) Hu, PhD

Research Interests

Dr. Hu has expertise in several applied statistical areas, such as predictive modeling using machine learning, causal inference in observational disease-based studies, advanced study designs, longitudinal data analysis, clustering and outlier detection. Her statistical expertise can obtain insights to inform internal decision-making using real world evidence from Electronic Health Records (EHR) data. Dr. Hu has collaborated with clinicians as co-Investigators to support grant preparation and manuscript submission. From 2017 to present, Dr. Hu has served as a member for Geisinger Scientific Review Committee to support protocol reviews. In 2019 spring, Dr. Hu has contributed a research methods course 'Introduction to Meta-Analyses in Medical Research'.

• Study Design: Support statistical analysis plan and sample size estimation for interventional studies, including cluster randomized trials, and stepped-wedge designs.
• Evidence Based Medicine: Contribute data-driven understandings from case report or Meta-Analyses to assist health care professionals towards investigating the safety and effectiveness between interventions; evaluating the association between risk factors and clinical outcomes.
• Predictive Analytics: Leverage traditional statistical regression and machine learning algorithms to support decision making. The standard predictive modeling pipeline includes Data preprocessing, missing data imputation, feature selection, model evaluation (Support Vector Machine, Decision Tree, Random Forest, eXtreme Gradient Boosting, Neural Network), and final model implementation.
• Group-Based Trajectory Modeling: Identify potential patient subgroups to enhance treatment effects, due to the heterogeneity of treatment effects in study population. To better classify individuals and identify unique attributes that are predictive of outcomes is an important objective in precision medicine.
• Real World Evidence: Obtain insights from large repositories of structured and unstructured patient EHR data in observational disease-based studies. Investigate on the integration of genomic information into EHR to guide decision making and improve patient care.