Background: As more and more research data become available, we are able to build on this information and to achieve better research results more quickly. This presentation will show several bioinformatics related databases and programs supported by the National Institutes of Health help researchers share and learn from each other. We will also describe how the research results are beneficial to public health. The NIH's Bioinformatics Resource Center for viral families will be described in this presentation as well as the Immunology database and analysis portal (ImmPort) system. In addition, we will show specific examples of data integration, analytics and visualization of genomic and other information can be used to inform public health. Purpose: This presentation will help bridge the gap between the clinical research world and practical applications for public health leaders, managers and researchers. Methods: Use of data curation, standardization, and analytics to understand clinical research information. Results: We will show examples of bioinformatics databases that are being used to further health research, and will show how analytics of clinical and genomic information can result in valuable information that can improve the health of the public. Conclusions: Effective use and reuse of clinical and genomic research information can inform and improve public health.

background: Large data sets exist for both clinical/healthcare and community/public health populations—often gathering distinct or unique perspectives on various subgroups. However, merging information together from disparate data can be difficult without linking primary keys. Statistical matching, and propensity score matching, provide approaches to align data and enable effective examination of the superset of information, and help to make useful comparisons across variables. purpose: This presentation provides an example of current statistical matching and propensity score matching approaches for linking datasets, and an examination of the results of the analyses when linking a large public health dataset (Behavioral Risk Factor Surveillance System dataset) with a large healthcare dataset that do not share any primary key variables. methods: Using the R statistical tool and various packages (including StatMatch and MatchIt), subjects from the two disparate datasets were matched using a set of demographic, health, and other variables. Various strategies were used (e.g., ranked nearest neighbor, random hot deck) to examine matching success on indicator variables and covariates. results: Using a subset of variables linked through statistical matching or propensity score matching, both public health and healthcare variables were examined using generalized regression analyses and other statistical learning methods. A complement of both public health and healthcare variables were found to be significant predictors of the chronic health conditions examined. Lasso regression documented the relative importance of the complement of factors. conclusions: Statistical matching and propensity score matching can be performed on partnered data sets and, if important assumptions are kept in mind during interpretation, can aid in exploring relationships within the joined information in the data superset.

background: Large populations generate big data, and issues of data quality exist when defining, collecting, storing, and using big data. Data scientists estimate that 60% to 80% of their time is consumed with cleaning and preparing data for analysis. Data quality not only impacts the processing of data, but the correct interpretation of findings and shaping policy decisions. purpose: This presentation documents and demonstrates a comprehensive approach and set of methods we incorporate to assure high data quality for work with federal clinical big data sets. methods: Using a Data Quality Assessment Framework, the authors examine and identify data errors and their impact on specific data-driven decisions. Using data quality dimensions of completeness, timeliness, consistency, integrity, and validity, relevant methods are used for 1)data profiling of fields, values, and relationships (missing; skewed; outlier); 2)application of predefined data standards (HIPAA; HL7 Clinical Standards; AHRQ Implementation Guides); 3)detailed quantitative indicators (scagnostics), and more. results: After examination of a large healthcare data set, we identified major data quality issues that can impact the interpretation, use, and application of the information derived from basic/advanced analytics. Awareness of data quality issues/metrics and appropriate corrective actions improved payment modernization approaches for healthcare. We discuss these results. conclusions: A comprehensive, structured approach to establishing, maintaining, assessing, and using data quality principles and practices is integral to assuring that data can be useful and provide appropriate impact.