Increasingly, experts and interested laypeople are turning to the explosion of online data to form and explore hypotheses about relationships between public health intervention strategies and their possible impacts. We have engaged in a multi-year collaboration to use and design semantic techniques and tools to support the current and next generation of these explorations. We introduce a tool, qb.js, to enable access to multidimensional statistical data in ways that allow non-specialists to explore and create specific visualizations of that data. We focus on explorations of health data - in particular aimed at helping to support the formation and analysis of hypotheses about public health intervention strategies and their correlation with health-related behavior changes. We used qb.js to formulate and explore the hypothesis that youth tobacco access laws have consistent, measurable impacts on the rate of change in cigarette smoking among high school students over time. While focused in this instance on one particular intervention strategy (i.e., limiting youth access to tobacco), this analytics platform may be used for a wide range of correlational analyses. To address this hypothesis, we converted population science data on tobacco-related policy and behavior from ImpacTeen to a Resource Description framework (RDF) representation that was annotated with the RDF Data Cube vocabulary. A Semantic Data Dictionary enabled mapping between the original datasets and the RDF representation. This allowed for the creation and publication of data visualizations using qb.js. The RDF Data Cube representation made it possible to discover a significant downward effect from the introduction of nine youth tobacco access laws on the rate of change in smoking prevalence among high school-aged youth.

We demonstrate an early version of a semantic web tool, Global Health Explorer (GHX), that can be used to conduct public health surveillance using Twitter. Our infrastructure can use any controlled vocabulary to extract term uses in Twitter and supports hypothesis formation and exploration of data sets using visual analysis. The resulting data, gathered in RDF, makes it possible to analyze term usage through both temporal and spatial dimensions. GHX uses the qb.js framework to visualize and explore these data across dimensions, intially time and location. This allows users of GHX to monitor terms from pre-existing ontologies to conduct public health surveillance. We have prototyped the use of GHX to monitor terms from the NCI Thesaurus related to in uenza-like illnesses.