Give your entire health system an AI advantage without exposing PHI.

No. Your health system chooses which tables and datasets go into the De-Identified Digital Twin. Most organizations start with the data one initiative needs, then add more as new use cases come up. Source PHI stays where it is, inside your environment.

The twin reflects your source data as of its most recent refresh. Your health system sets how often that refresh happens. Source PHI stays inside your environment the whole time. When analysis runs in Secure Runtime, it runs against your current source data, so final results reflect the latest records.

Subsalt is a privacy-preserving platform for health system data. It generates a De-Identified Digital Twin of your structured EHR data under the HIPAA Expert Determination method. Approved users explore the twin, then run analysis against source PHI in Secure Runtime. They get de-identified aggregate results, not row-level patient data, so your health system can support AI, research, and analytics without broadly exposing PHI.

Subsalt connects to your enterprise data warehouse, EHR exports, and registries through standard interfaces. It runs inside your existing on-prem or cloud environment. PHI stays within your health system’s security and compliance boundary. Subsalt adds a governed layer for de-identification and aggregate analytics, without creating a new data silo.

Subsalt uses the HIPAA Expert Determination method, evaluated by Datavant, rather than relying only on Safe Harbor. De-identification documentation accompanies the outputs covered by the determination. It documents the dataset, transformations, residual-risk assessment, and assumptions. That gives privacy, legal, and compliance teams defensible evidence that de-identification meets HIPAA’s “very small risk” standard for the intended use.

Traditional tools typically produce a single de-identified dataset and hand it to analysts. Subsalt provides a governed environment instead. Users explore the De-Identified Digital Twin through SQL or compatible AI chat, then run approved analysis against source PHI in Secure Runtime. The result is repeatable, privacy-preserving access to health system data, without distributing row-level records.

Access is role-based and project-based. Approved researchers, analysts, AI teams, and clinicians can explore the digital twin and submit analysis. Subsalt returns aggregate outputs such as counts, rates, summary tables, cohort metrics, and charts. These suit research, quality improvement, posters, and internal reports. Subsalt never returns raw patient data.

Subsalt gives AI teams a governed path to model-ready data. They can understand cohorts, features, and outcomes from the digital twin, then validate performance with aggregate metrics computed from source PHI in Secure Runtime. Models and external tools never see raw PHI, which helps health systems align AI initiatives with data governance and privacy requirements.

Yes. Subsalt shortens the path from research question to analysis-ready results by reusing a consistent de-identified environment for cohorts and queries. Investigators can explore data, refine hypotheses, and get aggregate outcomes more quickly. That supports academic medical centers, multi-site studies, registries, and real-world evidence programs.

Subsalt turns high-level governance policies into operational controls. Committees still decide which projects and roles are approved. Subsalt enforces de-identification, access control, and output checks consistently. That lets governance teams support more AI, research, and analytics use cases without multiplying manual data extracts and one-off privacy reviews.

Those platforms pool de-identified data from many organizations into a shared national asset. That is useful for benchmarking, but it is not your population, and it does not answer questions about your own patients. Subsalt is infrastructure for a single health system. It makes the clinical data you already own usable under your own governance, security, and partnership strategy. You never contribute it to a shared marketplace or cede control of your PHI.

Most organizations begin with one initiative currently blocked by data access. That might be a research program, an AI project, a strategic partnership, or a quality-improvement effort. Subsalt sets up a De-Identified Digital Twin and Secure Runtime for that use case, shows how the path from request to result changes, and then scales the same capability across other data initiatives.

No. Subsalt can generate synthetic data at query time, so you do not have to store a separate synthetic dataset. You also have the option to cache synthetic data behind the endpoint for faster retrieval, in milliseconds rather than minutes. Neither approach requires a lot of extra storage.

Subsalt connects to your data warehouse, such as Snowflake or Databricks, rather than directly to your EHR. It works with the patient data your health system already brings into that warehouse.