PROJECT SUMMARY/ABSTRACT Candidate: Andrew J. Read, MD, MS is a gastroenterologist and health services researcher at the University of Michigan with a research focus on improving diagnosis of gastrointestinal (GI) tract cancers. Dr. Read has prior training in biostatistics but not in mixed methods or implementation science. This NCI K08 award will train him to become a leader in the translational science of medical prediction, providing him with the skills to develop tailored implementation strategies for cancer prediction models and test these strategies in future clinical trials. Research Context: Iron deficiency anemia (IDA) is a common sign of many diseases, including GI tract cancers. Despite this important association, IDA is often under-recognized or under-investigated, resulting in delays in diagnosis. Fortunately, the electronic health record (EHR) contains potential diagnostic clues that can be leveraged to improve diagnosis of GI tract cancers. Specifically, algorithms can be developed to detect subtle changes in complete blood count (CBC) parameters to predict GI tract cancers. However, prediction models have rarely been implemented in clinical practice. Identifying the barriers and facilitators to implementing a model can allow for more customized implementation strategies to improve the chances of successful implementation. Research Aims: Dr. Read will (1) Refine a prediction model for detection of GI tract cancers using longitudinal laboratory data from the Veterans Health Administration (VA), the largest integrated healthcare system in the United States; (2) Identify barriers and facilitators to implementation of prediction models in clinical practice using mixed methods with an explanatory sequential design, incorporating a clinician survey followed by semi- structured clinician interviews; and (3) Develop and test components of a prediction model implementation strategy in a clinical setting, using Implementation Mapping. Training Aims: Dr. Read will develop expertise in: (1) Developing advanced longitudinal prediction models using a national dataset; (2) Using and applying mixed methods and implementation science frameworks to identify barriers and facilitators to successful implementation of a prediction model; (3) Applying Implementation Mapping to develop and test an implementation strategy for a novel clinical prediction model.