HealthVision
HealthVision is Health Data Analytics Institute's platform for predictive risk and chart summarization, built on longitudinal data covering more than 140 million patients and 500 billion encounters over 26 years. It combines EHR and claims data into a concise patient profile with risk estimates for mortality, readmission, complications and diagnoses, the factors behind each and links to the source records, opened inside an EHR or through a web application. Risk rosters help care teams prioritize follow up after discharge, heart failure care and serious illness conversations, perioperative teams get procedure specific risk estimates, and accountable care organizations use claims based rosters for outreach and network planning. Diagnostic Discovery ranks likely missing diagnoses by revenue impact and clinical significance for documentation and coding teams, with payment tied to net new revenue and a claimed average net revenue impact of more than $14 million per system.
Houston Methodist used the platform across eight hospitals to schedule follow up within seven days of discharge, and its readmission rate moved from 18.1% to 16.2%. At Dana-Farber Cancer Institute, the BRIDGE-SIC pilot used predictive selection and AI summaries to show care teams patients' documented preferences, and 902 of 936 summary points in a 58 patient pilot matched the underlying records. Houston Methodist Coordinated Care, HarmonyCares and Buena Vida y Salud appear among its partners.
Nassib Chamoun founded the company, based in Dedham, Massachusetts.
Capability Axes
An AI Health Index grade measures what a buyer can verify from public sources on the date shown. It is not a rating of how good the product is. A vendor can build an excellent system and grade low on an axis because it publishes nothing an outsider can check. How grades read
The patient profile combines chart summaries with predictions about readmission, mortality, complications and possible conditions. The predictions change as new information reaches the platform and the summaries link back to medical records. An acute care team can use the same engine to build a list of patients to contact after discharge.
The population health module ranks patients for outreach or care management, and Diagnostic Discovery uses predictions to help a hospital find possible undocumented conditions. Those tools are built around learned risk estimates and chart extraction. HealthVision embeds them in an existing EHR workflow or shows them through a web application.
Care managers at Houston Methodist get lists ordered by predicted risk and decide how to arrange follow up, and its ACO enrolls the top risk quartile for unplanned admission in an intensive program with physician visits and case review. A user can open the source material behind a summary or follow the contributing factors of a prediction, and Diagnostic Discovery sends likely conditions to clinical and coding review before anything reaches a chart or claim. People decide on every output in these deployments. No common escalation threshold or override path is set across the individual risk models and suggested diagnoses.
The predictive models draw on longitudinal data covering more than 140 million patients and tie each prediction to contributing information in claims and charts, with targeted models for hundreds of clinical endpoints and a separate language model workflow for summarizing notes. Each application uses a prediction its own way, a readmission score around discharge or a mortality score in serious illness care, and the predictive and generative tasks stay distinct. The foundation model behind the summaries is not named, and no model version or change schedule is given for any predictive endpoint.
Health Data Analytics Institute builds its own outcome specific predictive models, can host a customer's institutional risk models on the platform, and runs HealthVision on AWS in separate customer environments. Generative summaries use a language model, and the Epic listing refers broadly to LLM and GPT capabilities without naming who supplies the active models or where they run. No processor list is offered, and no terms say whether a model operator may keep notes, prompts or generated output.
Dana-Farber's 58 patient BRIDGE-SIC pilot found that 902 of 936 generated summary points accurately reflected prior documentation, with errors including overstated preferences and misattribution; the preliminary results appeared in a 2026 oncology meeting abstract, and a larger trial is enrolling.
At Houston Methodist, readmissions fell from 18.1% to 16.2% after risk based follow up across eight hospitals, a figure HDAI reports from a series that does not separate the software from the change in follow up practice, and the hospital confirms its use of the software. Anonymous academic medical center results cite a 2% average rise in case mix index and about $300 per admission, with no period or method. Earlier immunotherapy work did not test the current platform.
HealthVision synthesizes patient information from charts and claims and runs inference in an AWS environment that separates customers, with access controls, encryption and audit logging. It ships models already trained, which run on a hospital's live data without training a model on that hospital's records, though whether customer data feeds later versions of the shared models is not stated.
The general privacy policy covers contact information and service usage, and how chart content is handled inside the summarizer, how long those inputs are kept, and how an error in a clinical summary is handled are not addressed.
HealthVision reads and summarizes charts, and HDAI calls it HIPAA compliant and describes controls for protected patient data, including encryption and separate customer environments. The privacy policy covers website use and information HDAI obtains in offering services, with its detail on contact and usage information and nothing on the obligations HDAI takes on when it reads a patient chart. No business associate agreement, tier or subcontractor terms are offered, either for use inside an EHR or for the standalone application.
HDAI announced HITRUST r2 certification for its core information systems in February 2026, with the added AI Security Assessment and Certification, and the HITRUST directory confirms certification at the organization level. The homepage also carries an AICPA SOC mark, and HDAI describes encryption, access controls, network isolation and audit logging for its hosted environments.
The current report is not public, and the announcement describes the assessed systems only in general terms, so whether the scope reaches the language model providers and every version of the clinical integration is open.
Diagnostic Discovery ranks likely inpatient diagnostic and documentation omissions, including conditions not yet diagnosed or entered in the chart, and shows likelihood estimates with supporting clinical data for review. The acute care and perioperative versions forecast mortality, readmission and complications and help staff decide who needs attention first, and the models are sold as clinically validated.
No device clearance, regulatory basis or device exclusion is stated anywhere for these diagnostic and prognostic functions. The chart links give a reviewer context for a suggested diagnosis, but the risk forecasts that order clinical attention rest on model inference a clinician cannot open.
HealthVision shows chart sources and risk factors whenever a clinician reviews an output. HDAI holds a HITRUST AI Security Assessment and Certification for its core information systems and uses matched comparisons of similar patients to look for disparities in care, and its marketplace materials say race and social drivers are left out of the predictive inputs.
Its white paper reports that hospitals in the top fifth for diagnostic capture had about 20% lower mortality than their matched digital twins, a comparison between hospitals rather than patient groups. No performance is broken out by race, sex or clinical subgroup for the model that flags undocumented conditions or for the language model that summarizes charts.
Diagnostic Discovery is sold as guaranteed to add revenue to all but the very top performers, with payments contingent on reaching net new revenue targets and a retrospective claims audit offered without upfront cost. That commitment attaches to the module's revenue result rather than to the accuracy of a suggested diagnosis, and the target and contract terms behind it are not given. Reviewers can open the chart sources behind a suggested condition or risk summary before acting. The website terms cover the site, and a patient whose risks or preferences were summarized wrongly has no route to a remedy, nor is any accuracy guarantee given for clinical output.
Epic's Connection Hub lists Health Vision BEC Limited v3 for access inside the record system, and the platform reads clinical data from the chart to build summaries and risk profiles. Houston Methodist uses it in care delivery, with care managers planning scheduling and follow up from its rosters, and some deployments pull CMS or payer claims to refresh population rosters. The homepage says only that HealthVision integrates with the customer's EHR, and Epic is the one system named. Which actions it can write back into Epic, and whether any order or appointment is written automatically, are not shown.
Clinicians open HealthVision inside an EHR or through its own web application, and the analysis runs on AWS with customer environments kept apart. The privacy policy says information collected through its services is transferred to and processed in the United States and allows transfers through subcontractors, but its data categories are account, contact and usage information. No storage location is committed for patient charts or backups, and the view inside the EHR is a window onto the same hosted system rather than a local deployment option.
HDAI sells HealthVision to health systems and accountable care organizations through a demonstration and deployment process and promises results in weeks rather than months. Diagnostic Discovery starts with a retrospective claims audit at no upfront cost, and its payments are contingent on reaching agreed net new revenue targets, which ties the charge to a measured result for that module.
The acute care, perioperative and population health modules share the same predictive and chart platform. The minimum target, the share of revenue, a starting subscription price, charges for the other modules and installation or data mapping fees are not given.
An eight hospital system uses HealthVision to prioritize follow up for patients leaving acute care, Houston Methodist Coordinated Care applies risk estimates to a Medicare population for outreach and planning, and Dana-Farber's oncology pilot uses a 90 day mortality estimate to pick participants and summarizes prior serious illness conversations.
The platform also offers perioperative estimates across surgical service lines and claims based work for accountable care groups such as HarmonyCares. Published results cover cancer care and select hospital workflows, with no separate validation for each procedural specialty, each risk model or every care setting it is sold into.
Alternatives to HealthVision
Other products that do the same kind of work as HealthVision, and what each one does.
Nest Genomics, the trading name of Nest Health Inc. and a Y Combinator company, sells a genomic care platform that runs inside the electronic health record.
Renvio sells electronic records, billing and analytics for dialysis organizations, from independent centers and acute dialysis providers to skilled nursing facilities and home therapy programs.
HealthLeap screens every adult hospital inpatient every day for malnutrition risk.
Bayesian Health, a Johns Hopkins spinout founded by Suchi Saria, sells a platform of machine learning models that read the EHR continuously and flag inpatients at rising risk, with workflows that prepare order sets and documentation for clinicians to act on.
AKASA, formerly Alpha Health, sells generative AI for the provider side of the revenue cycle to hospitals and health systems.
Etiometry, based in Dedham, Massachusetts, sells a clinical intelligence platform for critical care that combines vendor neutral ICU data aggregation with four FDA cleared physiologic risk indices.
Pricing
Vendor-published figures are labeled as such. Figures labeled “Estimated” are derived from third-party sources and have not been confirmed by the vendor.
| Entry Price | Pricing Basis | BAA Tier | Implementation | Source |
|---|---|---|---|---|
|
Quote based
|
Diagnostic Discovery payments depend on meeting net new revenue targets. The wider platform is sold to hospitals and accountable care organizations through a demo and deployment process. | No business associate agreement is offered. | The Diagnostic Discovery retrospective audit is offered without upfront cost; no other implementation fee is shown. | Vendor Published |
Diagnostic Discovery is sold as guaranteed to add revenue for all but the top performers, with payment contingent on net new revenue targets. The target, revenue share, rate card and software price are not shown, and the acute care, perioperative and population health modules have no listed fees.