Mirvie

A model produces the result, and the measurement carries much of the weight. Encompass sequences cell-free RNA from a second trimester blood sample, and a regularized logistic regression over six genes plus blood pressure and body mass index turns the reading into a low or elevated risk call; without the model there is no result. The signal came from learning across thousands of transcriptomes, which is what makes this machine learning rather than a fixed formula. But the laboratory measurement and the company's sequenced cohort of about 11,000 pregnancies are the asset, and the model itself is small, one part of the product rather than the whole of it.

Oversight runs through the ordering clinician, and one part is undescribed. The test is ordered by a provider, and the ordering page states who qualifies: a singleton pregnancy, age 35 or over by the due date, and none of the listed high risk conditions (prior preeclampsia, diabetes, chronic hypertension, kidney disease, lupus, antiphospholipid syndrome). The result is a two-level call, low or elevated risk, and the site tells patients to discuss it and the action plan with their care team; the terms say Mirvie gives no medical advice.

What is missing is the behavior of the follow-up service. Patients with elevated results receive a home blood pressure kit and medication reminders, and SMS health check-ins run through a named provider, Memora Health, but nothing published says what those messages do when a patient reports a high reading or symptoms.

The model is published in full; the update practice is not. The validation paper names the classifier: an L1-regularized logistic regression over six genes (four placental: PAPPA2, KISS1, XAGE2 and ADAM12; two immune: CD163 and VSIG4) plus mean arterial pressure and body mass index, with corrections for gestational age and a gene-based estimate of fetal sex. Features, coefficients and decision boundary were locked before validation, and the training cohort (5,399 pregnancies from 11 sites, collected June 2020 to March 2023) is described with its demographics.

What is not stated is whether the test sold as Encompass runs that locked classifier unchanged, or how and when a retrained model would reach patients.

The model is Mirvie's own and runs in its laboratory, and the parties around the patient are named. The terms of service name three providers: Getlabs for at-home blood draws, 1health.io for delivering results to patients, and Memora Health for SMS check-ins and reminders, each under its own terms. The classifier is the company's, published in its validation paper, so no outside model sits between the sample and the result.

No subprocessor list is published, the hosting provider is not named, and no terms say what the three named providers may do with the results and health data they handle.

A prospective multicenter validation with a locked model is published, and it shows modest positive predictive value. The Miracle of Life study enrolled pregnancies at 11 US sites and through mobile phlebotomy; the classifier was locked on 5,399 and validated on a pre-specified, held-out set of 3,000 split by time within each site (Nature Communications, 2025). Among people without high risk factors, the area under the curve was 0.74 for the placental form of hypertensive disease, 0.83 when limited to delivery at or before 35 weeks, and 0.88 in those aged 35 and over, where the negative predictive value was 99.7 percent and the positive predictive value 8.9 percent. Test positive pregnancies spent more days in neonatal intensive care.

Mirvie funded the study and most authors hold equity. A companion cohort study (JAMA Network Open, 2025) found current guideline criteria label 89 percent of pregnancies as increased risk. No trial yet shows that testing changes outcomes; the aspirin adherence figure on the provider page comes from a 2026 preprint survey.

The documents say what the data may be used for, not what is kept. The Encompass terms reserve the right to use a patient's results and questionnaire answers to develop and improve the service and for research, with an email opt-out, and the privacy policy (last updated 30 April 2026) allows de-identified aggregate data to be used for any lawful purpose. Retention is stated only as long as necessary. Nothing is said about how long blood samples and RNA sequences are kept, whether they are banked, or whether they train future models, which matters because the cell-free RNA analyzed comes largely from the placenta and fetus as well as the mother.

The company publishes no safety engineering either, meaning guardrails, hallucination mitigation or how a safety event is handled.

A privacy document reaches the service, and no HIPAA position is taken. The Encompass privacy policy (last updated 30 April 2026) covers people who order the test and says the company receives their health information to provide it. As a laboratory reporting its own results, Mirvie would ordinarily be a covered entity in its own right, but no notice of privacy practices is published and neither the policy nor the terms mention HIPAA. The results platform and messaging provider operate under their own terms.

Controls are asserted with nothing independent behind them. The Encompass privacy policy says the company uses physical, organizational and technical measures and cannot guarantee security; no SOC 2 report, ISO/IEC 27001 certificate, HITRUST certification, penetration testing or trust portal appears on either site or in a search. The laboratory's CLIA certificate governs test quality, not information security.

No device claim is made, and the test is scoped as a laboratory service. Encompass is performed in a CLIA-certified laboratory as a laboratory developed test, and the site says so in its footer; no FDA clearance or approval is claimed. A federal court vacated FDA's rule on laboratory developed tests in March 2025, so the test's validation rests on the company's published studies rather than on agency review.

The site tells patients to discuss results and the action plan with their care team, and the terms say the company gives no medical advice.

Subgroup performance and a bias correction are published with their method. In the validation paper (Nature Communications, 2025), the locked classifier was compared across moderate risk groups in a held-out set of 3,000: it performed similarly in nulliparous (984), Black (476) and obese (801) participants to the whole group, and reached an area under the curve of 0.88 in those aged 35 and over (403), against 0.74 across everyone without high risk factors. During training the team found the model's sensitivity and specificity differed by fetal sex and corrected for it; they also report that race added no predictive information beyond the RNA, and race is not an input. The provider page's claim that the test performs equally across race and ethnicity rests on this analysis.

Two limits: the subgroup comparisons are shown in a figure rather than tabulated, and the headline 91 percent sensitivity is for people aged 35 and over without high risk conditions, the only group the test is sold for, not for the whole cohort.

Error rates are published with their method and denominators, and nothing stands behind them. The validation paper reports, for the held-out set, areas under the curve by group and, for people aged 35 and over, a negative predictive value of 99.7 percent and a positive predictive value of 8.9 percent for the placental form of hypertensive disease that drives preterm preeclampsia; the patient page repeats the 99.7 percent and says 9 of 10 eligible pregnancies that developed preterm preeclampsia were called elevated risk. A reader can see that most elevated results will not be followed by the severe disease.

The terms of service state that Mirvie gives no medical advice, disclaim warranties, limit liability and have the user indemnify the company, including for disputes with its results, messaging and blood draw providers. No guarantee or remediation attaches to a result.

Integration is offered with no system named. The ordering page lists a web form customized for the practice, direct EHR integration where available, and a print and fax requisition; patients see results through the 1health platform. No record system, interface standard or practice using an integration is named.

Because aspirin prophylaxis has to start early, how the result reaches the chart before the next visit is the question for a practice to settle.

A laboratory service in one place, with the data path implied. Blood is drawn at the provider's office or at home by Getlabs, and the test is run in Mirvie's CLIA-certified laboratory, which the billing guide places in San Francisco. Results reach patients through the 1health platform and messages through Memora Health, each under its own terms.

Where results, questionnaire answers and sequence data are hosted, and whether any processing leaves the United States, is not stated.

The price is published, with its tiers and how it is charged. The billing guide sets a flat self-pay price of $295 per test, applied automatically if insurance denies the claim, and a financial assistance schedule by household size and income that brings the cost to $0, $75 or $150. Encompass is billed to commercial insurance and Medicaid under the general code CPT 81599, is HSA and FSA eligible, and has interest-free payment plans; uninsured patients are told they can ask for a Good Faith Estimate. A patient can start through her own clinician with the standard requisition, with no sales conversation.

The list price billed to insurers is not published, and a practice joining the program is onboarded through the company.

Deliberately narrow, and sold only where it was validated. Encompass is offered for singleton pregnancies in people aged 35 or over by the due date without listed high risk conditions, from a blood draw between 17 weeks 4 days and 22 weeks, in US prenatal care. That is the group in which the held-out validation reported its best performance (403 people, area under the curve 0.88), and the ordering page excludes the high risk conditions the validation also excluded.

The study cohort was broader, so extension to younger patients has published data behind it but is not offered; preterm birth and fetal growth restriction are research programs, not products.