Yale researchers now have the ability to access millions of EHR records from patients across the country using the platform Truveta, linked through the Yale DataMed website.
Thirty health systems across the country update their electronic health record data daily through Truveta, a secure, cloud-based environment where it is checked for quality and translated into the standard medical terminologies used in research (SNOMED CT, RxNorm, LOINC), reducing the need for extensive manual cleaning before use in research. The platform also includes closed claims data linked across commercial payers, Medicare, and Medicaid, along with longitudinal patient records spanning multiple care settings. Yale researchers can access the data by logging into the Yale DataMed website, which acts as a gateway to Truveta Studio, the platform which supports building and querying research cohorts directly against row-level, de-identified patient data, including a natural-language query assistant for researchers who prefer not to write code-based queries.
Truveta can be useful for early-stage exploratory work, cohort feasibility assessments, and studies that require statistical power beyond what a single-site dataset can provide — without the lead time of a prospective study. Its value comes from pairing that reach with detail: the records capture what was documented during a patient’s care, including diagnoses, procedures, vital signs, lab results, and medications, and some member systems also contribute clinicians’ notes and medical images. The linked insurance claims and mortality data fill in two gaps that health records alone leave open — what happens to patients after they seek care elsewhere, and whether researchers can reliably learn how their health outcomes turned out. That design reflects how Truveta began: health systems formed the consortium early in the COVID-19 pandemic, when no comprehensive, current national picture of patient care existed, and built it for research from the start rather than adapting records kept for billing.