Phenotyping Support Core

PROJECT SUMMARY – SUPPORT CORE The recent widespread adoption of electronic health records (EHRs) provides the opportunity to leverage real- world data generated as a byproduct of healthcare for clinical discovery. EHRs contain a wealth of valuable information, particularly for pregnant women, women who just gave birth, and for children, as these individuals have scheduled health care visits on a frequent basis which provide a source of longitudinal data. Further, these populations have been previously under-represented in research efforts. However, the data in EHRs are not always easily accessible for research. Pregnancy and childhood have not been focal points for development of phenotyping methods in EHRs. Algorithms to assess specific drug exposures, important covariates (which may change rapidly), and outcomes (which may be population-specific) have not been validated for obstetric or pediatric populations. The Phenotyping Support Core will provide full technical support for phenotyping within the proposed Center and other sites of the MRPRINT Hub. In Aim 1, the Core will support the Project 1 and Project 2 by assisting with medication exposure, therapeutic response, and adverse event outcomes. We will develop and validate natural language processing pipelines and machine learning approaches to extract drug outcomes. We will define and evaluate algorithms for detecting neonatal opioid withdrawal syndrome (NOWS) and related maternal and infant outcomes. We will implement the algorithms and organize datasets and results into common data models. In Aim 2, we will support the work of the MPRINT-Hub by assisting with validation of findings at other sites using real-world, EHR data, including Vanderbilt’s BioVU biobank resource. In Aim 3, we will develop and disseminate tools for EHR-based therapeutics research for the scientific community. We will assemble a suite of existing and new analytical tools for MPRINT research. All the tools and algorithms generated through the core will be published and shared on PheKB and with MPRINT Hub platforms, as appropriate. The tools generated from this project will significantly enhance our efficiency when using EHRs and offer a potential approach to identify new phenotypes and to generate novel research hypotheses for the benefit of pediatric and maternal populations.