Data Trust Index (DTI)
The Data Trust Index (DTI) is an eight-dimensional scoring framework, developed by SuperTruth Inc., that scores any record on a 0 to 100 scale; the published weights were calibrated in health. It is the authorization step: per record, in the moment, the score that decides whether data can be trusted for a given use. The score travels permanently with every record, so any model, clinician, or auditor downstream can see how much the record deserves to be trusted.
DTI scores the record, not the patient. Provenance, consent, recency, and concordance are properties of a data artifact, not attributes of a person. Nothing in the score judges, rates, or follows a human being; it tells the person acting on a record how much the record itself deserves to be trusted.
The eight dimensions and their default weights are Provenance (25%), Consent (20%), Recency (15%), Quality (10%), Concordance (10%), Validation (10%), Breadth (5%), and Stability (5%). Institutions can configure the weights for a given study or compliance context; the defaults are the patented weights.
Scores map to four trust tiers (Bronze, Silver, Gold, Platinum) plus a fail band below 55: Bronze (55 to 69) is exploratory, Silver (70 to 79) is operational-grade, Gold (80 to 89) is clinical-grade, and Platinum (90 and above) is regulatory-grade. A record below 55 is held for remediation.
DTI scoring is context-dependent. The same record may score differently depending on the study, the use case, and the governing compliance body, and every score is replayable: the inputs and weights that produced it are part of the record's audit trail.
The closest analogy is Visa, which decides in the moment whether a transaction can be trusted without ever holding your account. DTI is that authorization decision, made per record and per use.
Data trust
Data trust is the decision to act on a record, made on evidence instead of assumption. The evidence is five documented answers: where the record came from, whether consent covers this use, how current it is, whether independent sources agree, and whether outcomes validate it. SuperTruth measures it one record at a time with the Data Trust Index (DTI), a patented score of 0 to 100 across eight weighted dimensions, with published weights calibrated in health.
Trust is a decision, not a property of the warehouse. Cloud platforms secure the perimeter and call that trust; the data inside is exactly as reliable as it was when it arrived. Data trust is scored on the record itself, before it reaches any warehouse or model, and the score travels with the record wherever it goes.
Most organizations measure quality because quality is measurable inside the record itself. Trust needs the context the record does not carry on its own: chain of custody, consent scope, corroboration, and clinical linkage. A record can be perfectly clean and completely untrustworthy.
Clearinghouse (the Visa sense)
SuperTruth is the clearinghouse for health data in the Visa sense, not the claims sense. Visa does not lend you money and never holds your account; it decides in the moment whether a transaction can be trusted, and it keeps the record. DTI is the authorization step, the Ledger is the clearing, and Truth Units are the settlement.
In healthcare, "clearinghouse" usually means the pipes that route claims between providers and payers. Those clearinghouses hold the data, which is exactly why a breach of one is catastrophic. SuperTruth never takes custody of the record: zero-copy architecture means the data stays where it lives and only the verified signal travels.
The mapping is one-to-one. Authorization: DTI plus the trust-tier gate decides per event, in the moment, whether data can be used. Clearing: the Ledger records what moved, when, and under whose authority. Settlement: value flows back to the institution and the contributor. Network rules: the eight dimensions and the Bronze through Platinum tiers are the standard members agree to. Dispute: consent revocation propagates to every downstream use, which a credit-score model has no analog for.
The neutrality matters as much as the mechanics. Visa does not compete with its member banks, and SuperTruth does not compete with the institutions on its network: it does not sell their data, run their models, or hold their records.
Data trust vs. data quality
Data quality asks whether a record is complete, correctly formatted, and free of errors. Data trust asks a harder set of questions: where did this data come from, was consent obtained properly, how current is it, does it corroborate with independent sources, and has it been validated against outcomes. A record can be perfectly clean and completely untrustworthy.
Most organizations measure quality because quality is measurable inside the record itself. Trust requires context the record does not carry on its own: chain of custody, consent scope, corroboration, and clinical linkage.
This is also the difference between infrastructure trust and data trust. Cloud platforms secure the warehouse and call that trust; the data inside is exactly as reliable as it was when it arrived. SuperTruth scores the data itself, before it reaches any warehouse or model.
Data provenance
Data provenance is the documented origin and chain of custody of a record: which source produced it, under what conditions, and every hand it passed through on the way to the point of use. It is stronger than attribution and different from lineage. Lineage maps how a value moved inside your systems; provenance says whether the starting point deserves belief. In the Data Trust Index (DTI), SuperTruth's patented 0 to 100 score for any record, provenance is the heaviest of the eight dimensions at 25 percent, because a value that cannot say where it came from cannot be defended in front of a clinician, a regulator, or a court.
Provenance is stronger than attribution. A CLIA-certified lab result with device attestation and an intact chain of custody carries different weight than the same number retyped into a spreadsheet, even though the values are identical. A record can have immaculate lineage and worthless provenance: a number retyped from a fax, then tracked perfectly through nine pipelines, is still a retyped fax.
SuperTruth's rule for every surface it builds: a figure may not claim a period or a place its source does not cover, and unknown provenance is recorded as unknown, never invented.
Consent revocation
Consent revocation is a patient's withdrawal of permission for their health data to be used, and the hard engineering problem is propagation: every downstream system that holds or derives from the record has to honor the withdrawal, immediately and provably. In SuperTruth's consent step, revocation propagates downstream the moment it is made, and every consent event lands in an immutable, hash-chained audit log.
Static consent, collected once and assumed forever, is one of the five structural barriers to trusted health data. Consent has scope, duration, and hygiene, and records are re-scored automatically when consent changes.
The practical test is the pharmacy counter: a revocation made this morning has to reach the dispensing check this afternoon, not the next quarterly sync.
Zero-copy architecture
Zero-copy architecture verifies and scores data where it already lives instead of copying it into another vendor's warehouse. The record stays inside the institution's environment; only the verified signal, the score, the provenance, and the answer, travels.
Copying data multiplies risk: every copy is a new breach surface, a new consent question, and a new divergence from the source of truth. Zero-copy removes the copy instead of trying to secure it.
In practice this means no duplication, no external storage, no loss of provenance, and SQL-level traceability, and it is what lets federated research run across institutions without any records changing hands.
Temporal drift
Temporal drift is the decay of a data record's accuracy over time. A medication list from 2019, an address from a prior enrollment, or last quarter's risk factor may be dangerously wrong today while still looking perfectly clean. Data does not announce that it has expired.
Drift is modality-specific: a CBC result ages differently than a continuous heart-rate stream, and an allergy list ages differently than a genomic record. The Data Trust Index scores Recency with time-decay calibrated per modality, so the score reflects the record's currency at the moment of inference, not the moment of capture.
Models that ignore temporal decay produce inferences that are technically valid and clinically misleading, which is why drift is one of the five barriers the clearinghouse exists to remove.
Schema drift
Schema drift is a data source changing its field names, types, codes, or units without telling the systems downstream. The record still arrives and still looks clean; it just no longer means what the pipeline assumes it means. A renamed column, a lab that switches from mg/dL to mmol/L, or a code set that moves versions all count.
The Data Trust Index catches the break in two places. Quality scores resolution, completeness, missingness, and noise floor, so a field that suddenly empties or changes shape loses points at once. Stability scores test-retest reliability, so a source whose measurements stop agreeing with their own history is marked down before a model ever reads them.
Schema drift is the mechanical cousin of temporal drift: nothing about the record decayed, the container changed underneath it.
Semantic drift
Semantic drift is the same field coming to mean something different over time. A lab changes its assay method, a diagnosis is recoded under a new guideline, a wearable updates the algorithm behind a metric, and the values keep flowing under the old name with a new meaning.
Because the schema is unchanged, quality checks pass. The Data Trust Index catches semantic drift through Concordance, corroboration across independent sources (does the lab result still agree with the wearable, does the EHR still match the claim), and through Validation, the clinical linkage of a data type to outcomes and peer-reviewed evidence. When a field stops corroborating and its clinical meaning is no longer established, its score falls.
This is the failure that makes clean data dangerous: the record is complete, well formed, and wrong in a way no format check can see.
Provenance gap
A provenance gap is a record whose origin or chain of custody cannot be shown: who produced it, when, from what source, and through which hands it passed before it reached you. The record may be accurate. Nobody can prove it.
Provenance is the heaviest of the eight DTI dimensions, 25 percent of the score under the default weights, so a gap marks a record down harder than any other single failure. A record with a complete chain of custody can be replayed and audited; a record with a gap can only be believed.
Zero-copy architecture is the structural answer: when data is verified where it lives instead of copied into another warehouse, the chain of custody never breaks in transit, because there is no transit.
Federated query
A federated query runs a research question across multiple institutions' data without moving any of the data. The query travels to where each record lives, and what returns is the answer with its provenance, never the records themselves. Clean Rooms handle the cross-institutional work so no participant sees another's raw data.
Because nothing moves, nothing has to be de-identified twice or explained to a second privacy office, and every record in a resulting cohort arrives with its provenance, its consent scope, and a replayable trust score, which is what an IRB, a journal, or a regulator actually asks for.
Researchers can set DTI floors, minimum trust tiers a record must clear to enter the cohort, so the network's answers are only as inclusive as each question can defend.
AI data governance
AI data governance is the set of controls that decides what data an AI system is allowed to learn from and act on: where each record came from, whether consent covers the use, how current the value is, and whether the answer can be defended after the fact. SuperTruth implements it as an authorization step: every record is scored 0 to 100 before anything downstream touches it, and records below the floor never enter the model.
Most governance programs govern access: who may open the table. Governing AI means governing fitness: whether the record itself deserves to drive a decision. A perfectly permissioned record can still be stale, unconsented, or unprovenanced.
The practical mechanism is a trust floor. Set the minimum DTI a record must clear for a given use, and the pipeline enforces it the way a payment network enforces authorization: in the moment, per record, with a replayable reason.
Data lineage
Data lineage is the map of how a value moved and changed inside your systems: which pipelines transformed it, which tables it landed in, which reports consumed it. Provenance is the stronger claim underneath it: which real-world source produced the value in the first place, and under what conditions. Lineage tells you the path; provenance tells you whether the starting point deserves belief.
A record can have immaculate lineage and worthless provenance: a number retyped from a fax, then tracked perfectly through nine pipelines, is still a retyped fax.
SuperTruth scores both. Provenance is the heaviest-weighted DTI dimension at 25%, and the lineage of every transformation travels with the record so an auditor can replay how the score was earned.
Consent management
Consent management is the discipline of knowing, for every health record, what the patient agreed to, for which purposes, for how long, and honoring changes immediately. In SuperTruth's consent step, consent is verified in real time at the moment of use rather than assumed from a signature collected once, and consent is a scored dimension of every record: 20% of its Data Trust Index.
Consent has scope, duration, and hygiene. A record consented for care is not consented for research; a consent collected five years ago against a different privacy policy is not the same consent; and a revocation has to reach every downstream system provably, not eventually.
The engineering consequence: consent is data about data, and it travels with the record. When it changes, the record re-scores automatically everywhere the record is used.
DataSpine
DataSpine is the sourced record of every place in the United States: 804 attributes in nine categories, economics to food, from 102 live public sources, joined on one code across 331,007 geographies from the nation to the block group, with the source and the vintage kept on every value. Run by SuperTruth.
The nine categories and their counts are economics (266), health (161), demographics (108), environment (73), housing (71), education (41), psychographics (32), food (29) and technology (23). Health is one of the nine. The sources include Census, CDC, County Health Rankings, HUD, BLS, BEA, IRS, FDIC, FCC, Ookla, NCHS, NOAA, EPA, HRSA, USDA, Opportunity Insights, MIT Election Lab, US Religion Census, Zillow, Redfin, EIA, IMLS, CMS and NCES; 108 are cataloged and 102 are live. Together they hold 11,220,406 sourced data points, a count taken 8 Sep 2026 that grows continuously.
Every value carries the source that published it and the vintage it was published for, and every vintage ingested stays on file, so a figure used last year is retrievable with ?vintage=YYYY. Where a source holds no figure for a place the answer is "not on file", never a zero. Gaps can be filled from the nearest published geography, or on request with synthetic households generated from the place's own published distributions; a filled or synthetic value is always labeled and never counted among the sourced figures. No person is in it: every value describes a place.
Where this runs: the Gauntlet, the Truth Agent, VIOLET, Live or Die Here.
DTI score
A DTI score is the number the Data Trust Index produces for one record: 0 to 100, computed across eight dimensions (Provenance, Consent, Recency, Quality, Concordance, Validation, Breadth and Stability). It is the one name for that number wherever it appears: on a record, in a Gauntlet lane, in the Record Remix, in an audit trail. DTI scores any record; the published weights were calibrated in health.
The score maps to four trust tiers (Bronze, Silver, Gold, Platinum) plus a fail band below 55. Bronze (55 to 69) is exploratory, Silver (70 to 79) is operational-grade, Gold (80 to 89) is clinical-grade and Platinum (90 and above) is regulatory-grade. A record scoring below 55 is held for remediation and does not enter a tiered use.
The score describes the record, not the patient. It travels with the record, it is replayable from the inputs and weights that produced it, and it can differ for the same record under a different study, use or governing body.
Data truth
Data truth is the state of a record that has been proved before anything acts on it: its source is documented, the consent that covers this use is verified, it is current for its purpose, and it agrees with independent sources. It is not the same as data quality. Data quality asks whether a record is complete and correctly formatted; a record can be perfectly clean and completely untrustworthy. SuperTruth, a data truth company founded in 2018 in Philadelphia, measures data truth one record at a time with the Data Trust Index (DTI): a patented score of 0 to 100 across eight dimensions, on any record, with published weights calibrated in health.
Data truth is a property of the record, not of the system that stores it or the person it describes. A record has data truth when four questions have documented answers: where did it come from, was consent obtained for this use, how current is it, and does it agree with other sources. When those answers travel with the record, a model, a clinician or an auditor downstream can see how much the record deserves to be trusted without redoing the work.
Data truth is measured, not asserted. The measurement is a score on the record, replayable from its inputs and weights and sealed with the record so it cannot be quietly edited, and the score decides the use: a record below the floor set for a purpose does not enter it.
AI truth
AI truth is whether an answer from a model or an agent can be traced to a record that was true when it was used, and whether the agent behaved while producing it. It has two parts. The record part is data truth: the record was sourced, consented, current and scored before the model saw it, which SuperTruth measures with the Data Trust Index (DTI), 0 to 100 on any record. The agent part is behavior: VIGIL, run by SuperTruth, scores every agent decision on a Behavioral Integrity Index from 0.0 to 1.0, lands it in one of four gates (pass, hold, alert, collapse) before it propagates, and seals it in a hash-chained ledger nobody can quietly edit. SuperTruth's line "Data truth is AI truth" means a model or agent is only as true as the records it acts on.
AI truth has two failure modes, and they need two different instruments. The first is an answer that looks true and has no row behind it: a figure produced from memory, with no source and no year. The second is an agent that reports well on itself. The record test catches the first. The agent test catches the second.
AI truth is not AI detection. Tools that guess whether a text was written by a machine answer a different question. AI truth asks whether what the model or agent did can be traced and whether the agent behaved, and it is proved with a score and a receipt, not a guess about authorship.
