Birth Model

The models are the engine of the prediction and risk modules, and the platform carries value beside them. Birth Model says two patented AI models, built on more than 150,000 patient charts, predict delivery timing, and that risk intelligence flags postpartum hemorrhage, preeclampsia and shoulder dystocia; it also generates handoff summaries and delivery notes with coding.

Around them sit functions that stand without a model: the Motherboard view of every patient on the unit populated from the record, electronic booking for inductions and cesareans, census and staffing dashboards, quality reporting, and, for freestanding birth centers, Birth Center Hub, a record system of its own.

The home page describes risk intelligence as 'Real-time alerts for PPH, preeclampsia, and shoulder dystocia. Never miss a critical indicator.' The birth center page says the platform watches 'every patient, every minute, so no one slips through' and that 'Nothing gets missed', and the clinicians page says the patented algorithms 'accurately predict delivery times for all types of obstetric patients'.

Those are claims that the system does not miss what matters, and a clinician reviewing alerts cannot see what was never flagged, so they grade here. Beyond them, the stated posture is advisory: predictions and alerts inform staff and clinicians, handoff summaries are generated for review before a one-click sync to the record, and staffing alerts fire when census passes thresholds. No alert threshold, confidence limit or behavior on a wrong prediction is published. Delivery forecasts drive staffing, so an optimistic forecast shows up as thin coverage during a complication rather than as a visibly wrong prediction.

Granted patents and a training scale name the approach, without version discipline. The company page gives United States patents 11,664,100 and 12,133,741 under a virtual patent marking notice, the home page says two patented AI models predict delivery timing and the platform is 'Built on 150,000+ Patient Charts', and delivery ETAs are described as based on inputs such as the Bishop score.

Performance figures are stated without method: '±12 min prediction accuracy' and '99.1% Pushing Phase Accuracy'. A patent describes a claimed method, not necessarily the deployed model. The company does not say what drafts the handoff summaries and delivery notes or which model sits behind the risk alerts, and it publishes no discrimination or calibration for each risk model and no versioning or update practice.

No model provider is named, although the architecture is described. The delivery prediction models are Birth Model's own patented models, built by its engineering and machine learning team, but the platform also generates handoff summaries and delivery notes, and nothing says what produces them or whether any outside model receives chart data.

The privacy policy (January 2026) says information is shared with business partners and affiliates 'to provide integrated services' and with service providers under confidentiality agreements, naming none. A subprocessor list, if there is one, may sit in the trust center.

The headline claims are specific: delivery predicted within 12 minutes, 99.1 percent pushing phase accuracy, 30 minutes saved per handoff, and, in the founder's words on the company page, lowered primary cesarean rates, saved staffing costs and increased labor and delivery revenue. None carries a site, denominator, period or method.

One customer is named: the Birth Center of New Jersey, whose medical director gives the home page testimonial and sits on Birth Model's advisory board; Birth Center Hub is said to be live there. No hospital is named, and no study or outside evaluation is published.

Retention is stated and the questions a model raises are not. The privacy policy (January 2026) covers the website, mobile app and services, lists health data and race and ethnicity among what may be collected with consent, and says personal information is kept 'for a maximum of twenty-four (24) months past account termination'.

It lists evaluating and improving the services among the uses and says nothing about whether chart data trains or tunes the models, or whether anything is de-identified. The home page and footer claim HIPAA compliance, and the footer links a trust center. Safety engineering, meaning how a missed or false risk alert is investigated and handled as a safety event, is not published. Obstetric data includes pregnancy loss and termination, which carry legal consequences in some states.

HIPAA compliance is claimed and business associate status is not stated. The footer shows 'HIPAA Compliant' beside 'SOC 2 Type II', and mobile notifications are described as HIPAA-compliant. The privacy policy (January 2026) reaches the services and lists health data among what may be collected, but says nothing about business associate status, an agreement, or protected health information from customer record systems.

The site names a recognized certification with its type and says little about the report behind it. The home page and footer state 'SOC 2 Type II', and the footer links a trust center hosted on Vanta. The report's period and scope are not stated, nor whether it can be opened without a request, and the trust center itself may answer both.

No clearance is claimed and no regulatory position is stated. The delivery timing forecasts drive staffing and raise no device question, but the risk intelligence is patient-specific clinical prediction: 'Real-time alerts for PPH, preeclampsia, and shoulder dystocia' and 'Early warning alerts for clinical teams'. The terms say the services 'do not constitute medical advice'.

Alerts appear with the patient's chart data and highlighted risk factors, which gives a clinician something to review, while the model's basis for each risk score is not described. Held here with the tension named.

Accuracy is asserted and no evaluation by group is published. The site states prediction accuracy figures and describes a quality dashboard that lets a hospital filter its own outcomes, such as cesarean rates, by race; that analyzes the hospital's care, not the model's performance.

No calibration or discrimination for the risk models is published by race, ethnicity, age, parity, body mass or payer, although obstetric warning signs are known to be recognized unevenly across patient groups and a risk model trained on past records can carry that forward into lower predicted risk.

Mechanisms exist that let someone check an output, and nothing stands behind it. Alerts appear with the patient's chart data and highlighted risk factors, handoff summaries are generated for review before they sync to the record, and the hospitals page markets 'comprehensive documentation trails'. The terms (January 2026) provide the services 'AS-IS', disclaim that results will meet requirements, exclude liability for errors and inaccuracies, and send disputes to arbitration in New Jersey. The accuracy figures on the home page carry no method or denominator, so they are not an error rate a buyer can test, and no route is described for a patient harmed by a missed alert.

Named record systems with integration in both directions claimed, and no named production site on them. The home page says Birth Model 'connects bidirectionally with Epic, Oracle Health, and athenahealth using HL7 FHIR and ADT protocols', launches by SMART on FHIR with no extra clicks, syncs handoffs and documentation to Epic through the Epic Connection Hub, and links an athenahealth marketplace listing and an Epic Showroom search; the birth center page adds eClinicalWorks among the systems patients transfer to and from. No hospital running the integration is named, so production on a named system cannot be checked.

A single hosted service with location implied rather than committed. Birth Model runs as a web application with iOS and Android apps, and its implementation guide covers enterprise iOS deployment through Apple Business Manager and mobile device management (Intune, Jamf, Kandji, Workspace ONE, Mosyle), managed configuration and a shared device mode. No hosting provider, region, tenancy model or on premise option is stated.

No price is published, and the shape of the sale can be established: a demo for hospitals and birth centers, an ROI calculator, card payments for paid services under the terms, and enterprise app licensing through Apple Business Manager. The calculator models the benefit side while the cost stays behind the demo.

Coverage is stated clearly with nothing validating it yet: labor and delivery units in hospitals, and freestanding birth centers through Birth Center Hub, with clinicians reached on mobile. One site is named, the Birth Center of New Jersey, through a testimonial from its medical director, who also advises the company; no hospital is named and no results are published by unit size, acuity or staffing model. A high volume tertiary unit and a small birth center are very different settings for a delivery timing model, and nothing shows which have been served.