Modeling and prediction of allergic trajectories in children - PROJECT SUMMARY/ABSTRACT I seek to become an independent, NIH-funded academic allergist-immunologist and clinical research informatician invested in improving outcomes in pediatric allergy. To this end, I will utilize electronic health record (EHR) data to study developmental patterns of pediatric allergic diseases and harness this knowledge to develop clinically relevant allergy prediction models using artificial intelligence approaches. I am in my final year of fellowship in pediatric allergy-immunology at the Children’s Hospital of Philadelphia (CHOP) having completed MD-PhD and pediatric residency training. I possess a strong background in experimental and clinical immunology and have pivoted to clinical research informatics for my post-doctoral research work, as it is aligned with my long-term career goals. Despite my development of multidisciplinary collaborations that have led to several publications, I require dedicated training in bioinformatics, statistics, and epidemiology in order to successfully conduct research independently in this space. The K08 Mentored Clinical Scientist Development Award is the optimal means by which I can obtain this additional experience and training. My training plan includes formal coursework through certificate programs at Penn. I will complement this with hands-on training with an exceptional team of mentors (Dr. David Hill, primary mentor; Dr. Robert Grundmeier and Dr. Rich Tsui, co-mentors) and advisors (Dr. Alex Fiks, Dr. Jonathan Spergel, and Dr. John Holmes). My training will also greatly benefit from the first-rate clinical and research environments at CHOP/Penn. To complement my training plan, I have developed an integrated research proposal that will allow me to model the development of allergic diseases in children. There is considerable variability from child to child in the sequence and number of allergic conditions that they develop, an emerging concept in the allergy field termed allergic trajectories. The basis for these heterogeneous allergic trajectories is incompletely understood. To date, allergy epidemiology has been measured largely via patient questionnaires, which are susceptible to reporting and self-selection biases. EHR databases contain demographic and longitudinal clinical data and thus represent an alternative approach for the study of allergic disease risk factors; however, they have been scarcely used for the study of their trajectories and for the development of disease risk prediction models. Building on my published and preliminary work, I will use the multi-institutional Comparative Effectiveness Research through Collaborative Electronic Reporting (CER2) and CHOP primary care network EHR databases to test the hypothesis that heterogenous allergic trajectories exist that are differentially associated with specific clinical, demographic, and/or environmental factors. The aims of my study are (1) to determine the major pediatric allergic trajectories and their biologic and non-biologic risk modifiers using multivariable logistic regression and Cox proportional hazards testing and (2) to develop and test EHR-based allergy risk prediction models that consider clinical and demographic information using machine learning and exploratory natural language processing. Study results will be externally validated in the independent, multi-institutional PEDSnet database. This work will yield clinically relevant tools that are expected to inform personalized medicine efforts focused on preventing and predicting pediatric allergy. By completing the proposed research and career development plans, I will be well positioned to function as an independent physician-scientist investigator at the intersection of pediatric allergy-immunology and clinical research informatics.
