Nest Genomics

A learned layer is claimed on a workflow and guideline product whose value stands without it. Most of what Nest delivers is workflow inside the record: risk questionnaires, education and consent, lab agnostic ordering, structured results storage and care plans that follow published guidelines. Nest's AI philosophy page says its models are trained on genomic datasets to identify patterns and suggest workflows that generic language models miss, and its home page describes AI powered algorithms that surface patients for testing.

Neither page says which function a trained model performs, and an earlier GenomeWeb account described the care plans as rule based. Read together, a learned component is claimed for patient identification and workflow suggestion, beside a core that works on rules.

Ask which outputs come from a trained model and which from guideline rules, and what the model was trained on.

Human oversight is asserted without a described mechanism. Nest's privacy policy says its analysis alone is never used for automatic decision making and only augments customers' human decisions, and the AI philosophy page refers to human in the loop verification and development with genetic counselors and key opinion leaders. Clinicians order tests and act on care plans.

What is not published is how a flagged patient or a suggested workflow is reviewed, who can override a care plan, or what happens when a suggestion conflicts with the clinician's judgment.

Ask how suggestions appear in the clinician's workflow, what must be accepted individually, and how overrides are recorded.

The approach is described in general terms with nothing identified. Nest's AI philosophy page says its models are trained on genomic datasets to identify patterns and suggest workflows, and that care plans follow guidelines with custom workflows for complex results. No model, model type, training data source, supplier or version practice is named.

Ask which models Nest runs, what they were trained on, whether any is a third party language model, and how updates are introduced.

The architecture is described and no model provider is named. Nest names AWS as its host and Google Analytics, PostHog and a self hosted OpenReplay as analytics tools in its privacy policy. Its AI philosophy page contrasts its models with generic language models without saying whether any outside model service is used, and no subprocessor list for patient data is published.

Ask whether any feature sends patient data to an outside model provider, and for the list of subprocessors that receive patient data.

A named customer with vendor reported figures and no method. Nest's October 2024 athenahealth announcement reports that at Ms. Medicine, a primary care network for women, 80 percent of patients completed the risk assessment, 52 percent were identified as high risk and 38 percent met genetic testing criteria, and that Nest was in use at more than 10 health systems and 50 clinics. Rady Children's Hospital San Diego, Baptist Health, Geisinger, Dana-Farber Cancer Institute and Mount Sinai are named as users.

No denominator, period, comparison or downstream result, such as tests completed or cancers found, is given, and no study of the platform is published. Scale of use is not evidence of benefit.

Ask for testing and follow up rates before and after deployment at a named site, with patient numbers.

General assurances, with the questions the models raise left open. Nest's privacy policy says provider customers control patient data, that records are kept at least six months after service and account data until closure, and that its analysis supports rather than replaces human decisions. Its security page says data is encrypted in transit and at rest on HIPAA eligible AWS services.

Nothing published says whether customer patient data trains the models the AI philosophy page describes, which de identification method applies, or what reaches any outside model service. Safety engineering for the models, meaning guardrails and how an error would be detected and handled, is not published.

Ask whether patient data from your system trains Nest's models, and whether any outside model service receives it.

Business associate status is stated and supported, and the agreement is not published. Nest Health Inc.'s privacy policy, updated August 2026, covers its website, mobile app and web platform for genetic testing decision support, says Nest acts as a business associate to healthcare providers and processes patient information under the provider's business associate agreement or Nest's standard one, and says patient information is governed by the provider's privacy policies and HIPAA. Its security page states HIPAA compliance.

Nest's standard business associate agreement is referred to and not published, and the policy does not say whether it binds subcontractors.

Ask for Nest's standard business associate agreement and its subcontractor terms.

A recognized certification is named without the artifact. Nest displays SOC 2 Type II certification and describes encryption, access controls, monitoring, secure development and vulnerability testing on its security page, with a security contact for questions and vulnerability reports. The SOC 2 report, its date and its scope are not published.

Ask for the current SOC 2 Type II report and its period, and the cadence of vulnerability testing.

No device claim is made, and no regulatory position is published. Nest's care plans follow published guidelines and are shown to clinicians, who act on them, so each recommendation has a basis a clinician can check; Nest says its analysis is never used for automatic decisions. It has not said how it classifies the patient identification and care plan features, for example under the clinical decision support provisions that exempt some software from device regulation.

Ask Nest for its regulatory analysis of its patient identification and care plan features.

A development process is described and no evaluation is published. Nest says its platform was developed with genetic counselors and key opinion leaders and that recommendations are guideline driven with human verification. No accuracy evaluation of the models, no results by ancestry, age or sex, and no change control for models or care plan content are published, which matters because family history based screening misses some groups more than others.

Ask how Nest evaluates its patient identification across patient groups and who approves changes to models and care plans.

Clinician review is the mechanism, and nothing stands behind the output. Nest's suggestions and care plans reach clinicians, who decide, and Nest says its analysis is never used for automatic decisions, so a wrong suggestion can be caught before it affects care.

No error rate for patient identification, no warranty, indemnity or remediation commitment for care plan content, and no route for a patient affected by a missed or wrong recommendation is published.

Ask how often patient identification misses people who meet testing criteria, and what Nest commits to when a care plan is wrong.

Named record systems with ordering and results inside them, and a marketplace listing. Nest names Epic, athenahealth, eClinicalWorks and AdvancedMD, with Redox among its integration routes, and announced in October 2024 that the Nest Genetics Copilot is available through athenahealth's Marketplace program, with Ms. Medicine named as a customer. It describes ordering from any laboratory and structured ingestion of results that stay current in the record.

No integration documentation or listing detail shows which data is written back into each record system, and no Epic site is named as live on Nest's pages.

Ask for a live reference on your record system and which data, such as family history, orders and results, is written back as discrete fields.

A single hosted service with location implied. Nest's security page says the platform is built on HIPAA eligible AWS services with encryption in transit and at rest, and its privacy policy limits the products to users in the United States and Australia. No AWS region, data location commitment for either country, or tenant isolation is described.

Ask which AWS regions hold US and Australian data and how each health system's data is separated.

No price is published and how the platform is sold can be established. Nest sells to health systems and clinics, runs inside their record systems including through the athenahealth Marketplace, and routes orders to any partner laboratory, so the buyer is the provider organization rather than a laboratory. No subscription fee, per patient or per order charge, or implementation fee is published.

Ask whether Nest charges the health system, the ordering laboratory or both, and the unit of charge.

Coverage is stated with nothing validating it yet. Nest serves health systems, primary care and women's health networks, cancer centers and a children's hospital, across hereditary cancer, cardiovascular genetics and secondary findings from broader sequencing, and is offered in the United States and Australia. No evaluation is published for any setting or specialty.

Ask which conditions and patient groups the care plans and patient identification cover, and results from a deployment like yours.