Solutions
One clearinghouse. Eight ways in.
The same infrastructure underneath, described by what it changes for the person using it. Every record scored 0 to 100 before anyone acts on it. Nothing of yours moves. Pick the seat you sit in.
Outside health? The same proof runs in insurance, finance, media, the public sector and place data. See where it runs, sector by sector.
Health systems
Which parts of this chart can I act on right now?
The chart in front of you was assembled from systems that were never built to agree with each other. Some of it is current. Some was auto-populated from a prior record. Some has consent for the use you have in mind and some does not, and the EHR stores all of it identically.
DTI™ changes one thing: every element of the record is scored 0 to 100 for where it came from, whether consent covers this use, and how current it is, before a clinician or a model acts on it. You see which parts you can act on now, which need a second look, and why. Nothing is copied out of your EHR. The record stays; the score travels.
The rule that bites
HIPAA
Patients who consented to treatment under HIPAA did not consent to AI training. Consent is a scored dimension of every record, so a record authorized for care is told apart from one authorized for training, and a revocation reaches every downstream use.
Health plans
When a regulator asks how we knew, can we replay the answer?
A member calls a specialist listed as in network. The provider left four months ago and the directory was never updated. That is a recency problem, and no credentialing audit would have caught it. Risk models have the same problem one layer down.
DTI changes one thing: a record's currency, source and concordance are scored, not assumed. Records below your DTI floor never enter the risk model, and a stale directory entry is flagged before a member sees it. When a regulator asks how you knew, the answer is a replayable score, not a recollection.
DataSpine adds the member's place context: 804 attributes about every US place, each value with its source and vintage (DataSpine, 8 Sep 2026).
The rule that bites
No Surprises Act
Federal law puts the liability for an inaccurate directory on the plan when a member is billed out of network. The recency dimension flags records whose freshness has decayed below your directory threshold before they are published.
Pharmacy
Is the check I run at the counter running on the whole record?
Allergies, interactions and prior fills often live in systems the counter cannot see: another pharmacy, a discharge summary, a cash-pay script that never touched the claim stream. The interaction check runs against a partial list and reports a clean result.
DTI changes one thing: the complete record arrives at the moment of dispense with every element dated, sourced and scored, so the check you run is the check that matters. Consent is verified in real time rather than assumed from the existence of the prescription, and a revocation reaches the counter the instant it is made.
The rule that bites
State board of pharmacy audits
When a dispensing decision is questioned, you must show what was known at the time of the check. Every element carries its source, its date and its score, so the answer is a replayable record, not a recollection.
Research
Will a reviewer accept this cohort?
Five institutions pool records for one study. One runs Epic with a documented chain of custody; another keys data in by hand. Side by side in a federated dataset the records carry different weight, and nothing marks the difference.
DTI changes one thing: every record from every site carries the same 0 to 100 score, and each arrives with its provenance, its consent scope and a score you can replay in front of an IRB, a journal or a regulator. Federated queries run where the data already lives. Nothing moves, so nothing is de-identified twice or explained to a second privacy office. Clean Rooms handle the cross-institutional work.
DataSpine adds tract-level context alongside outcomes, each figure with its source and vintage. Alongside, never behind.
The rule that bites
Common Rule (45 CFR 46)
IRB approval covers the protocol, not every record in a multi-site dataset. Consent is tracked per record, and a cross-site query returns only the records whose consent covers the intended use.
Pharma and life sciences
Will this real-world evidence survive a regulatory reviewer?
Real-world evidence comes from EHRs, claims, labs, wearables and registries. Most of it has never been scored for provenance, consent coverage or recency. A lab result from a CLIA-certified facility and a self-reported entry from a wellness app enter most RWE pipelines as equivalent inputs.
DTI changes one thing: each record is scored before the model sees it, and the score travels with the record into the training run and into the submission. A reviewer can see what the evidence was worth, record by record, instead of taking the dataset's word for it.
The rule that bites
21 CFR Part 11
Electronic records in a submission must show authenticity and integrity. Every transformation in the audit trail carries a timestamp and an attribution, which is the documentation Part 11 asks for.
Government and VA
Can this record stand up to an inspector general?
A veteran sees a community provider under the MISSION Act. A Medicaid patient moves counties. The data exists, spread across VA systems, state HIEs, civilian EHRs and federal registries, and nothing scores it before a program acts on it.
DTI changes one thing: every record is scored against its source. When a VA record and a civilian record disagree, concordance flags the discordance with attribution. When a rural record has not been verified in years, recency says so before a case manager relies on it. Nothing moves out of federal systems; the score travels.
DataSpine gives a program the place behind the record at the level it reports: 804 attributes for 331,007 US geographies, from the nation to the block group, each value with its source and vintage (DataSpine, 8 Sep 2026).
The rule that bites
VA MISSION Act (2018)
Community care produces records in civilian systems that must be reconciled with the VA record. The Act created the access; it did not create a way to score the reconciliation. Concordance does.
Oncology
What are patients searching before they ever reach a clinic?
VIOLET finds what patients search before they see a doctor, and answers with the nearest enrolling trials, the evidence behind them and a number to call. It is in live pediatric clinical use today, and the map carries adult oncology too. Evidence is triangulated across six sources: ClinicalTrials.gov, PubMed, FDA openFDA, Google Scholar, Google News and patents.
The clearinghouse rule holds here as well. A gold mark means measured, from a named source, with a date. VIOLET's own derivations are said in words, and unknown is never rendered as zero.
The rule that bites
What your compliance officer will ask
There is no patient record in VIOLET. It reads public evidence and aggregate search interest. A sign-in is logged; what you search is not, and that no-retention rule is enforced in the database schema, not only promised.
Health AI developers
Where did this training record actually come from?
You have trained on books, code and the scraped internet. The one dataset that matters most for health AI, patient records, lab results, wearable streams and clinical notes, is the one where provenance, consent and recency are never verified before the model sees it. DTI scores every record 0 to 100 before your training run or your inference pipeline touches it.
The rule that bites
HIPAA
Protected health information used to train a model needs authorization beyond the treatment relationship. Consent is scored per record, so a training run can be held to records authorized for it.
Every record dated, sourced, consented and scored, before anyone acts on it.
Bring a sample from your own environment. We will score it in front of you.