Flexible on data, rigorous on evidence.
Whatever data fits the question, with peer-reviewed methods behind every result.
The right data for the question.
Engagements run on whichever data fits the question. There is no commercial incentive on our side to push clients toward our data. The incentive is to put the right data into the platform and produce the right answer.
Our data
RespondHealth's Harris-partnered asset, when the question fits the therapeutic areas and cohorts we cover.
Your data
The client's own records, when they hold the relevant data and bring us in for the analytics layer.
Linked data
Privacy-preserving tokenization to identify, license, and link third-party datasets when the right view of a question lives across multiple sources.
Multi-EHR coverage through Harris Computers.
The principal real-world data asset is provided through our partnership with Harris Computers: a multi-EHR database covering primary care and specialty practices. It pairs full structured tables, including procedures, diagnoses, labs, medications, and demographics, with extensive de-identified free-text clinical notes.
AmazingCharts
Primary care and small-practice settings.
CareTracker
Multi-specialty and ambulatory practices.
GEMMS
Cardiology and specialty medical groups.
DigiChart
Women's health and OB/GYN practices.
De-identified under Expert Determination.
The asset is de-identified to the standard set out in 45 CFR 164.514(b)(1), the Expert Determination method. This is the standard appropriate for research uses where structured fields plus extensive narrative notes need to be retained without re-identification risk.
55-year-old presenting for type 2 diabetes follow-up. A1c 7.8%, down from 8.4%. Continues metformin 1000 mg BID, tolerating well. Discussed adding a GLP-1 receptor agonist.
Plan: recheck A1c in three months. Referral placed to Dr. Albert Chen, endocrinology.
Peer-reviewed and externally validated.
The platform's extraction and reasoning layer has been published and evaluated across multiple cohorts and EHR systems. Statistical guarantees on extraction accuracy, longitudinal cohort assembly, and population-scale knowledge-graph search.
- Kim, Foty, Thakor, Skyler, Robinson, Cairns, Seyfert-Margolis (2026). Computable longitudinal patient journeys derived from structured and unstructured EHR data: development, validation, and clinical application in GLP-1 receptor agonist therapy. Under review. The platform end-to-end on a worked clinical case. Read paper
- Kim, Foty, Shrestha, Seyfert-Margolis (2025). Conformal Prediction and Verification of Large Language Model Extractions in EHR Data. AAAI Fall Symposium Series (FSS-25). Evaluated on 10,000 clinical visits across 898 practices and three EHR systems. Read paper
- Kim, Shrestha, Foty, DeLay, Seyfert-Margolis (2024). Structured Extraction of Real World Medical Knowledge using LLMs for Summarization and Search. arXiv:2412.15256. Demonstrated on 33.6M patients, with a Dravet syndrome case study. Read paper