The Data Trust Index: SuperTruth Publishes the First Formal Framework for Health Data Integrity Scoring

The Data Trust Index is now a published, citable academic framework.

Today SuperTruth released the first formal paper describing the DTI: an eight-dimensional scoring system that assigns a continuous integrity score from 0 to 100 to health data records before any AI model, clinical decision system, or downstream pipeline touches them.

The paper is available on Zenodo at doi.org/10.5281/zenodo.19601616 and the full source is at github.com/evil-robot/supertruth-dti.

What the paper covers

The DTI framework has been running in production since SuperTruth's engagement with imaware, where it standardized 105,000 diagnostic records, reduced data preparation time by 95%, and surfaced a customer segment accounting for 20% of revenue that had been invisible in fragmented data. Until now, the methodology existed only as a patent application and a production system.

This paper formalizes it:

DTI dimension weights: default scoring configuration

The eight dimensions and their default weights: Provenance (25%), Consent (20%), Recency (15%), Quality (10%), Concordance (10%), Validation (10%), Breadth (5%), Stability (5%)

Modality-specific recency decay functions — why a CBC panel ages differently than a continuous HRV stream

A concordance scoring methodology for corroborating signals across independent sources

A formal Algorithm 1 pseudocode for the full scoring procedure

Comparison to existing frameworks (Weiskopf and Weng, ONC, FDA SaMD, ICH E6) showing DTI is the only framework that produces a runtime record-level score

A prospective multi-site validation protocol as a pre-registered methodological framework

The four trust tiers

DTI scores map to four operational tiers:

Bronze (55-69): Exploratory. Useful for discovery. Remediation required before clinical use.

Silver (70-79): Operational-grade. Fit for analytics and population health.

Gold (80-89): Clinical-grade. Suitable for patient-facing AI and institutional models.

Platinum (90+): Regulatory-grade. Suitable for FDA submission and clinical decision support.

Why this matters

The FDA's 2021 AI/ML-Based Software as a Medical Device guidance explicitly identified training data quality as a gap in the framework for clinical AI. TEFCA is building national interoperability infrastructure. The 21st Century Cures Act opened patient data access. None of these initiatives established a standardized, record-level quality metric that AI systems can condition on before incorporating data.

DTI addresses that gap. The score travels with every record permanently. It does not disappear when data moves downstream.

Publishing this as a citable preprint means the framework is now available for other researchers, health system architects, and regulatory reviewers to build on, critique, and extend.

What comes next

The paper includes a pre-registered prospective validation protocol for a multi-site study comparing DTI-filtered model performance against unfiltered baselines on clinical prediction tasks. If your organization works with health data at scale and wants to participate in that validation, reach Louis Simeonidis at louis@supertruth.ai or (215) 918-4140.

Cite as: Snyder, J. A. (2026). The Data Trust Index: A Multidimensional Framework for Evaluating Health Data Integrity in AI Systems. Zenodo. https://doi.org/10.5281/zenodo.19601616