BACKGROUND: Tableau is a powerful tool for displaying data in a visually appealing, interactive manner. However, it is primarily designed for straightforward aggregation and analysis of count data rather than dealing with complex survey designs. Public Health Seattle & King County (PHSKC) is adopting Tableau for displaying >200 community indicators. The data for these indicators largely come from population-based surveys or data sources with confidential information that cannot be shared at the individual level. This presentation will feature a live demonstration of the approaches PHSKC is taking to overcome these limitations and a discussion of lessons for other public health jurisdictions.
METHODS: We developed two strategies that address confidentiality and survey weighting issues: 1) pre-aggregating data using existing statistical software but developing scripts to automate the output into a format suitable for Tableau, and 2) using raw survey data but calling R via Tableau’s built-in integration functionality to apply survey weights and produce summary data. For display of spatial data in Tableau, we have generated custom polygon paths that allow Tableau to draw PHSKC’s health reporting areas and display relevant health data.
RESULTS: Each approach has strengths and limitations. Pre-aggregating data requires analysts to anticipate all desired data permutations and conduct the analyses accordingly. However, confidentiality can be assured, which in turn allows data to be displayed on the free-to-use Tableau Public. Calling R from Tableau using raw data offers much more flexibility to the end user and requires much less analyst time for data preparation. However, to protect the data, a paid Tableau Server solution is required. In addition, there are limitations to the types of values that can be returned from R (usually single numbers), which makes simultaneously displaying multiple slices of data a challenge.
Both approaches have allowed PHSKC to develop sets of templates, by data source, with custom geographies displayed on a map. We have adopted standardized color and naming schemes that provide a consistent feel across the indicators. PHSKC has begun publishing initial versions of the indicators on Tableau Public and embedding them on its website.
CONCLUSIONS: Tableau will enhance community groups’ ability to obtain the data they need. It will also greatly improve the efficiency of producing public-facing health-indicator data. The two approaches shared show that it is possible to display survey or confidential data in aggregate form, even on publicly accessible websites.