Understanding Public Health Systems through Data Integration and Visualization

Editor's Notes

• #3: Health providers, public health agencies, and social service systems all collect data Health – diagnoses, procedures Public health – WIC, immunizations, mental health Social service organizations – collect information on families while enrolling them in Medicaid Many programs serve the same clients with considerable overlap in data and databases maintained
• #4: HIV, diabetes, antibiotic resistance, access to healthcare, mental health, well-being
• #6: Public health and social services issues span multiple sectors
• #7: Data governance is “the specification of decision rights and an accountability framework to ensure appropriate behavior in the valuation, creation, storage, use, archiving, and deletion of information. It includes the processes, roles and policies, standards, and metrics that ensure the effective and efficient use of information in enabling an organization to achieve its goals.” Functional and regulatory requirements of multiple programs are challenging for a single system to support Every effective database needs a carefully designed schema that keeps data clean, avoids conflicts, serves its users’ varied needs, and accommodates future extensions. But, a data collection application that perfectly serves the governance needs of one entity may look very different from another data system that collections similar data, but for a different purpose. Take, for example, the different data systems that a social services organization might use to keep track of its operations and caseloads.
• #9: Application portfolios reflect the evolution and development of programs and systems they were built to support Many programs serve the same clients with considerable overlap in data and databases maintained – but the databases do not talk to each other
• #10: Data are siloed in different systems. Though they may support operational, fiscal, and regulatory requirements, data systems can fall short in supporting day-to-day informational needs . People are pretty crafty and want to get things done – they will always develop workarounds, but maybe not always the ones you’d wish they had developed. Paper logs No source/version control/fidelity
• #11: Program-centric emphasis of data systems makes it difficult to gain insights for policy-level decision-making. Again, a lot of data is being collected, but generally for very specific operational, governance, or reporting purposes – the data is not necessarily designed to be readily distilled for use in day-to-day or policy-level decision-making But it doesn’t have to be that way! Today’s technology allows for sophisticated linkage of databases – and particularly, the move toward cloud computing and the explosion of those technologies has exciting implications for the provenance and application of public health data The rest of this presentation will describe a data integration scenario that we face in our work at HSRI with social services data for the intellectual and developmental disability population, and the solution that we have come with up with to not only integrate disparate data sources, but democratize the data to make it available to social service agencies and policy makers
• #17: Integration = enable a central view across the system Data sourcing – what data is the most accurate and timely? Where does it come from? What problems does that data have? Transformation standards/ETL. Provide consistent codes and descriptions, even fix bad data. Physical structure – transformed and cleaned data needs to be reorganized and restructured – database normalization – in this case, turning data from program-centric data sources so that it is represented at the person- and system level. Subject oriented Present the information consistently Metadata – information about data in the data warehouse. Documents business definitions of data, enforces valid values, security characteristics, ownership, timeliness, quality, type, length, etc.. Maintain data history – even if the source systems do not Security and privacy – must resolve the important security, privacy, and governance requirements of the data elements to be integrated within the data warehouse Usability – Restructured so it makes sense for end users to understand. Decision-making queries are easier to write
• #23: John Tukey was not an artist – he was a statistician! Tukey lambda distribution – had his own fast fourier transform algorithm – also coined the term bit when working with von Neumann on computer designs - and also, relevant to this conversation, the boxplot! One of my favorite presentations was in Canada when I made a joke about boxplots to parliamentary staff – and they actually laughed! It actually wasn’t funny, had something to do with him winning the National Medal of Science but NOT for the boxplot. Believed that in statistics, too much emphasis was spent on confirmatory analysis as opposed to exploratory analysis. I really like boxplots because it displays a lot of information in a condensed picture. It can tell you about your outliers and what their values are. It can also tell you if your data is symmetrical, how tightly your data is grouped, and if and how your data is skewed. Take up less space than a histogram or a density plot. Can COMPARE groups of distributions
• #25: So getting back to public health – if you’re in public health you know all about the cholera outbreak of 1854, and how he became the father of modern epidemiology, in part because of his work tracing the source of cholera. Miasma theory – didn’t have germ theory – but he didn’t think the evidence was consistent with foul air. Used statistics, and also this dot map to illustrate a cluster of cases around the water pump.