Public-tier visualizations

Data visualization and data art built only on the public tier of the NIH All of Us Research Program Data Browser. The public tier is aggregate-only: counts rounded to 20, counts under 20 suppressed, no identifiers, geography no finer than US state. Nothing here goes below that floor.

Data are from the All of Us Research Program public Data Browser. This is an independent project and is not endorsed by NIH or the All of Us Research Program.

Pieces

• RepresentativenessEnrollment to date and the sequenced participants of the current release in each state as marks, one per 20, each population travelling between the state the program reports them in and where they would sit if every state had enrolled, or been sequenced, in proportion to its Census population; the two never the same marks; territories listed off-map.Open the map
• Silent RecordsThe whole cohort as one field of marks, one per 20 participants, dimming to each data layer in turn, from the one that covers most of the cohort to the one that covers least, then relighting.Open the piece
• Four Kinds of GenomeIn this release, "genome" means four different measurements. A pyramid of marks by age and sex at birth, one per 20 participants, changing shape in place from one data type to the next; what shrinks is where the next type reaches fewer people. The structural-variant shrink is a subset, the one relation the program states.Open the page
• ThinningA corridor of the enrollment lifecycle: one column per status as the program reports it, one mark per 20 people, the marks the next status counts crossing to it column by column; what stays behind is the thinning between two counts.Walk the corridor
• Sixty-Eight Thousand WristsThe participants with Fitbit data as marks, one per 20, moving slowly between their place in the whole cohort and rows of N with one mark each: the scatter and the ratio are the same marks.Open the piece
• The StripA seismograph of enrollment, played back: a playhead sweeps how many were new each day since 2017, the day-to-day differences of the program's own cumulative counts, one column per day and one mark per 20 people, unsmoothed, with a version sized for a receipt printer.Play the strip
• The DictionaryEvery standard concept of an EHR domain as one typographic wall, type size following the square root of its participant count, the words travelling as marks when the domain changes. Medicine's vocabulary is enormous; the lived record is a small language.Read the wall
• Cohort AtlasHow the cohort fans out into survey modules, physical measurements, genomic data types, EHR domains, and the Fitbit slice, each drawn at its share of all participants; under it, participants by release from the catalog.Open the atlas

How the numbers are made

Every figure on this site is computed in code from a stored raw response of a program-published source. Each response is kept verbatim with its URL and fetch time, keyed by release, and is fetched once per release. Derived tables are rebuilt from those files, and an audit test ties every derived number back to the raw response it came from before a page counts as done. The pages carry the derived data inline, so what you see is what the audit checked. Methods walks through the sources, the definitions used on each page, the 20-count rule, and the program terms this project follows.