Every prescription dispensed under the paid drug registers over seven years: who prescribed it, which drug, where, and what it cost.
2019 to 2025Data packages
The harmonised datasets behind the Data Atlas, delivered as files with a molecule crosswalk, a vintage manifest and an integrity report, for use in your own tools. Take one family, several, or the whole Atlas. Trade shipment records and global bilateral flows come as separate packages.
What this data holds
Every figure below was counted from the data itself, as received, before any processing. Each names the years it covers, and each sits inside one of the packages further down.
Every prescription dispensed under the paid drug registers over seven years: who prescribed it, which drug, where, and what it cost.
2019 to 2025Spending on medicines across the paid registers, tracked drug by drug, plan by plan and prescriber by prescriber.
2019 to 2025Every prescriber and healthcare organisation enumerated on the US national provider registry, with specialty, credentials and location.
2005 to 2026Every drug approval on record, from the 1930s to today, brand and generic, with application type and approval date.
1939 to 2026The patent and exclusivity listings that determine when a medicine loses protection and generics can enter.
current listingRegistered clinical studies with sponsor, phase, condition, intervention and site.
1999 to 2026Registered drug manufacturing sites worldwide, mapped and linked to what they make.
current registrationRegulatory inspections of those plants, each with its outcome and date.
2008 to 2026Individual problems cited during those inspections: the detail behind a plant's compliance record.
2008 to 2026Drug plans whose medicine coverage is held, across commercial and Part D books.
2024 to 2026Individual rulings on whether a plan covers a medicine: its tier, and whether prior authorisation, step therapy or quantity limits apply.
2024 to 2026Measured 2026-08-13. Each figure was counted directly from the held data, by record counts over the original files or read-only aggregations. Nothing is estimated. Held as 1,160 original files · 323.9mn records · 93 GB · 177 datasets covering 1939 to 2028, read exactly as received, and delivered harmonised onto one molecule key.
What every package includes
A package is the same resolved data the lenses read, without the lenses. It arrives as files, with the key that joins them and the paperwork that says where each record came from.
Data Atlas. Every dataset behind the platform with its record count, source, vintage and integrity check. Each package is one or more of its families, exported as files.
The packages
Dataset and record counts are those reported on the Data Atlas at build time. Each package lists the families it contains and the grain of the records, so you know what one record is before you load it.
Drug economics, 16 datasets and 143.3mn records: prescriber-by-drug utilisation, spending by drug and by manufacturer, pricing. Utilisation, 7 datasets and 68.4mn records: state drug utilisation and physician-administered drugs.
Formulary coverage cells, plan registries and restrictions: prior authorisation, step therapy and quantity limits, by plan and by drug. The largest family on the Atlas by dataset count.
Drug regulatory, 20 datasets and 0.7mn records: approvals, patents and exclusivity, product listings, shortages, recalls, drug master files, import alerts. Reference, 8 datasets and 6.1mn records: codes and crosswalks.
Quality, 24 datasets and 1.1mn records: inspections, findings, warning letters, quality measures. Facilities, 5 datasets and 2.0mn records: registered drug manufacturing sites, hospitals, care sites, cost reports. Devices, 6 datasets and 5.7mn records: device registries.
Providers, 4 datasets and 14.0mn records: the national provider registry and taxonomy. Industry payments, 4 datasets and 35.3mn records: payments to clinicians, general and research, with the paying company named.
Research, 1 dataset and 19.7mn records: the clinical trial registry with sponsor, phase, condition and site. Population, 1 dataset and 37k records: population counts to put prevalence and demand on a denominator.
Export and import shipment records, one record per shipment: product, exporter or importer, counterparty, destination or origin country, quantity, value and date. About 1.56mn export and 71k import records.
Bilateral trade flows for 105 reporter countries, by product and month, January 2024 to March 2026. Reporter, partner, product, direction, value and quantity in one table.
All twelve Atlas families in one delivery, with the molecule crosswalk that joins them, one manifest and one integrity report covering the whole set. The two trade packages are taken separately.
Trade packages are counted from the shipment and flow records and are not part of the 177-dataset Atlas count. Record counts refresh as each register is released.
How a package is made
Each register is read as released, resolved to one canonical molecule key, stamped with its vintage and checked before it is written out. The files you receive are the same files the platform reads.
Data or platform?
The data is the same either way. The difference is whether you bring your own tools to it, or want the lenses and Ask already built on top.
Use cases
The same records answer very different questions depending on who loads them. These are the ones they are asked most, with the packages each one reads.
Every one of these is answered from the same files, on the same molecule key. If the question you need answered is not on this list, tell us what it is and we will name the packages it needs.
Questions we are asked
Parquet and CSV, one file per table, with a manifest listing every file. A single DuckDB file on request, so the whole package opens in one step. Column names and types are the same across formats.
Each register is re-read when its next release is published. You receive the new files, a new vintage manifest naming the release, and a new integrity report with the updated counts. Earlier vintages are kept, not overwritten.
Yes. Every record carries the canonical molecule key, and the reference package supplies the codes and crosswalks that map your product, provider and plan identifiers onto the same keys. Join on one column, not on names.
Yes. The six Atlas packages above are each one or more families, and a single family can be taken on its own. The two trade packages are separate. The full Atlas is for teams that want everything on one key.
Name the package or the families you need, and what you will load them into. We confirm the source terms, the current vintage and the delivery format, and reply with what the package contains.