Health Care Technology Data Providers · Head-to-head
ONC Health IT Dashboard and Open Data vs The Cancer Imaging Archive (TCIA)
Which health care technology data providers data fits your job: ONC Health IT Dashboard and Open Data, or The Cancer Imaging Archive. API, files, or your warehouse. Daily, weekly, or hourly.
ONC Health IT Dashboard and Open Data
The Cancer Imaging Archive (TCIA)
Coverage, side by side
| ONC Health IT Dashboard and Open Data | The Cancer Imaging Archive | |
|---|---|---|
| Geographic | United States - national totals plus state/territory and county breakdowns depending on the series | Global contributor institutions, primarily US NCI-sponsored trials and international research cohorts |
| Temporal | Calendar-year series 2008 through 2026 releases; adoption briefs span 2008-2024, clinician file covers program years 2019-2024 | Collections added continuously since 2010; acquisition windows vary cohort by cohort |
What each contains
They tie on 1 attribute. Pick by fit, not by loyalty.
| ONC Health IT Dashboard and Open Data | The Cancer Imaging Archive | |
|---|---|---|
| Publisher | Office of the National Coordinator for Health IT (ONC/ASTP) | National Cancer Institute / University of Arkansas for Medical Sciences |
| Subject lens | US health IT adoption and policy: certified EHR uptake, interoperability, information blocking, patient portals | De-identified cancer imaging: DICOM radiology and digital pathology with outcomes, genomics and expert annotations |
| Scope | Four content types - Dashboards, Data Briefs, Quick Stats, raw Datasets - across eight topic filters; about 20 named machine-readable series plus clinician-file and Lantern archives | About 240 collections (211 Complete, 26 Ongoing) keyed by cancer type, modality and research focus |
| Geographic coverage | United States - national totals plus state/territory and county breakdowns depending on the series | Global contributor institutions, primarily US NCI-sponsored trials and international research cohorts |
| Temporal coverage | Calendar-year series 2008 through 2026 releases; adoption briefs span 2008-2024, clinician file covers program years 2019-2024 | Collections added continuously since 2010; acquisition windows vary cohort by cohort |
| Detail level | One percentage per geography-year-measure; provider-level rows in the Promoting Interoperability clinician file | Patient, then imaging study, then DICOM series, then image instance, with per-collection annotation tables |
| Formats | JSON, CSV, XML, HTML dashboards, XLSX | DICOM, ZIP, JSON, XML, CSV, SVS, TIFF |
| Documented fields | 10, marked verified | 10 down to individual DICOM tags, marked verified |
| Best for | Benchmarking adoption and interoperability by state, county and year | Assembling imaging training and validation cohorts with ground-truth labels |
What each does better
ONC Health IT Dashboard and Open Data
A yardstick someone already standardized. The measures arrive as published shares - share of non-federal acute care hospitals with certified EHR technology, share with a basic EHR system with notes, share sending, receiving, finding and integrating electronic health information - so benchmarking one state against another needs no modeling step.
Fifty-state coverage with county depth. aha.csv alone spans all 50 states plus territories across 15+ annual periods, and the series reach below the state where the underlying surveys allow - national totals, state and territory cuts, and county breakdowns depending on the dataset.
Policy breadth beyond the hospital ward. The same portal tracks office-based physicians (nehrs.csv), certified-health-IT developers by provider segment and program year (EHR-vendors-count-dataset.csv), Meaningful Use acceleration, Surescripts e-prescribing, REC KPI masterfiles, state privacy and consent laws, hospital TEFCA network participation and information blocking.
Provider-level linkage. Beneath the aggregates sits the Promoting Interoperability clinician file for program years 2019-2024, whose npi, chpl_id and cehrt_id_clean fields join individual clinicians to the certified products they reported - a bridge from policy percentages to named practices.
The Cancer Imaging Archive
Resolution no registry can match. Every collection decomposes patient, then imaging study, then DICOM series, then image instance, with Modality (CT, MR, PT, MG), BodyPartExamined and Manufacturer documented per series down to the DICOM tag. The policy record stops at percentages; this one hands over the pixels.
Clinical payload attached to the images. Where a cohort supplies them, tables join the imagery to patient outcomes, treatment details, genomics and expert analyses - segmentations and annotations. HANCOCK carries whole slide images and tissue microarrays for 763 head and neck subjects rather than radiology at all.
Cohort scale inside a single collection. NLST alone reaches 26,254 subjects; EAY131 holds 2,813 and UTSW-Glioma 625 glioma cases with demographic, diagnosis, treatment and molecular-test tables beside the MR imaging. Training-set assembly is a filter operation, not a collection project.
Machine-learning-ready labeling. Expert segmentations ship inside the collections - BraTS-PEDs pairs 457 pediatric glioma subjects with segmentation files - which is why computer-vision teams, not policy analysts, dominate this record's persona tagging.
Where they're equivalent
More than their subjects imply. Both publishers are federal health agencies speaking with statistical authority rather than vendor marketing - ASTP/ONC on the policy side, the NCI Cancer Imaging Program with Frederick National Laboratory involvement on the imaging side. Both document exactly ten fields apiece, both dictionaries verified during research, and both records score 9/10 on Datadory's rubric against a 7.81 catalog-wide average.
Both ship machine-readable formats - JSON, CSV and XML appear on both format lists, alongside HTML dashboards and XLSX for the policy record and DICOM, SVS and TIFF for the imaging record. Both update in place rather than being reissued as new products. And both serve the same strongest personas: data scientists and ML engineers tag each record near the top, followed by developers building products and academics publishing with citable federal data.
Where they part company is tense and object. ONC reports what institutions adopted; TCIA preserves what the disease did to people. One speaks in percentages of hospitals, the other in instances of imaging.
The verdict
Verdict: sample both, pick by fit - they answer different halves of the same health-technology question, and neither subsumes the other.
Take ONC Health IT Dashboard and Open Data if the unit of analysis is a place and the question is policy: benchmarking state EHR adoption, tracking interoperability and information-blocking trends, comparing certified-product vendor counts, sizing digital-health markets by county. That shape suits market researchers and competitive-intel teams mapping where adoption lags.
Take The Cancer Imaging Archive (TCIA) if the unit of analysis is a person and the question is clinical: assembling cancer-detection training cohorts, validating segmentation models against expert labels, pairing imagery with outcomes and genomics, building teaching files. That shape fits model builders and academic researchers first.
Rule of thumb: if the noun in your question is a state, a county or a year, start with the dashboard; if it is a tumor, an organ or a scan, start with the archive. Ranked shelf context lives in the best health care technology datasets rundown.
Sample both, pick by fit. See ONC Health IT Dashboard and Open Data · See The Cancer Imaging Archive
Or take both in one feed
Yes - stacked in sequence they cover both ends of a health-technology adoption curve. Read the policy record first for the diffusion picture: which states and counties moved on certified EHR technology, how interoperability measures trend, where physician adoption trails hospitals. Then read the archive for the clinical frontier: which cancer types, modalities and cohorts have imagery rich enough to support model work today.
Two cautions straight from the records. There is no join key - one row describes a place-year, the other a patient, study or series - so document the allocation assumption whenever a combined figure spans both. And respect the calendars: calendar-year measures on one side, per-cohort acquisition windows on the other, meaning joins happen on periods you define rather than timestamps you inherit. A strategist sizing an oncology-AI market can pair state adoption percentiles with the imaging cohorts actually available for research partnerships; see the data scientists use cases for the modeling side of that pairing.
Datadory ships either record alone or both merged onto one delivery calendar, normalized to their documented field dictionaries with sample rows attached for validation - delivered daily, weekly, or hourly, your call. Or take both in one feed.
API, files, or your warehouse. Daily, weekly, or hourly.
Fair questions
Is ONC Health IT Dashboard and Open Data better than The Cancer Imaging Archive (TCIA)?
Better at different jobs. The dashboard wins on policy reach: eight topic filters, about 20 named series, all 50 states plus territories across 15+ annual periods, and both hospital and physician adoption lines. TCIA wins on resolution: about 240 collections decomposing to the individual DICOM instance, with outcomes, genomics and expert segmentations attached.
What do the two datasets have in common?
Structure and standing. Each documents a ten-field dictionary verified during research, each scores 9/10 on Datadory's rubric against a 7.81 catalog average, each ships machine-readable formats beside human-readable reporting, and each comes from a federal health agency speaking with statistical authority rather than vendor marketing.
Which covers more geography, the ONC dashboard or The Cancer Imaging Archive?
Depends on the lens. The dashboard is exhaustive inside one country - national totals plus state, territory and county breakdowns, with aha.csv spanning all 50 states plus territories. TCIA's contributor base is global, but coverage is cohort-driven: whichever disease areas and institutions contributed collections, from dozens of subjects to 26,254 in NLST.
Whose field documentation runs deeper?
Depth for depth, TCIA's. Its ten documented fields descend into individual DICOM tags - PatientId at 0010,0020, Modality at 0008,0060, BodyPartExamined at 0018,0015 - so every column traces to a published standard. ONC's ten fields are verified and analysis-ready, but full measure definitions live in per-brief companion dictionaries rather than one master schema.
Can ONC adoption rates be joined to TCIA imaging collections?
Not directly - there is no shared key. One row describes a place-year, the other a patient, study or series. Teams bridge through geography and calendar: aggregate imaging activity to the state or county level, compare against adoption percentiles for matched periods, and document the allocation assumption behind any combined figure.
Can Datadory deliver both datasets together?
Yes. Either record arrives alone or merged onto one delivery calendar, normalized to its documented field dictionary with sample rows attached for validation - delivered daily, weekly, or hourly, your call. Name the states, years, cancer types and modalities when you request the sample and it lands pre-cut. Or take both in one feed.