Inputs
Pregnancy complications, reproductive events, diagnosed conditions, symptoms, blood pressure, glucose, kidney function, and other cardiometabolic markers from visit summaries or, later, the EHR.
Products
HerArc
Astravia Analytics’ first product. A longitudinal early-detection engine for primary care that reconstructs maternal and reproductive history, connects it to downstream cardiovascular, metabolic, and oncological risk, and delivers a clinician-ready report in which every multiplier is traceable to peer-reviewed evidence.
What HerArc makes
A clinician-ready report, not a black-box diagnosis.
The product extracts pregnancy complications, reproductive events, diagnosed conditions, symptoms, and physiological changes—including blood pressure, glucose, kidney function, and other cardiometabolic markers—then maps them to long-term risk with an inspectable evidence trail.
Reconstruct the history
Relevant events and labs are extracted from visit summaries and health reports. The prototype accepts PDF uploads; later versions reconstruct history automatically through SMART on FHIR.
Surface missed patterns
Machine learning models identify patterns associated with frequently missed conditions such as PCOS, endometriosis, and thyroid dysfunction.
Apply a transparent evidence layer
Detected and reported conditions are combined with findings from peer-reviewed studies. Every risk multiplier is traceable to a specific source and assigned an evidence grade.
Place it in the visit
The product generates an explanation of risk factors, the evidence behind them, and screenings or evaluations to discuss with a physician. The system does not make the diagnosis.
Longitudinal summary
Maternal history connected to cardiometabolic follow-up
Hypertensive disorder of pregnancy
2× CVD
Grade A+
Gestational diabetes
7× T2D
Grade A+
Severe preeclampsia → kidney disease
9.38× CKD
Grade A
PCOS → coronary artery disease
2.27× CAD
Grade A
Illustrative multipliers from published literature. The physician can inspect underlying patient data and cited sources before acting. HerArc does not diagnose.
Inside HerArc
Extract. Detect. Cite. Deliver.
HerArc is built so a physician can see both the patient signal and the literature that supports the alert—without leaving the logic of a primary care visit.
Detection
Models flag patterns associated with frequently missed conditions, including PCOS, endometriosis, and thyroid dysfunction, then unify them into a phenotypic profile.
Outputs
A one-page, clinician-ready report: risk factors, cited evidence, evidence grade, and screenings or evaluations to discuss. No automated diagnosis.
How it runs on the Astravia platform
Pattern recognition, then deterministic clinical logic.
HerArc uses the company cascade hybrid: specialized agents for reproductive history, then a rules layer that applies cited hazard ratios so the physician remains the decision-maker.
Stage 1
Reproductive-history agents
Sub-models process cycle regularity, gestational blood-pressure spikes, insulin-resistance indicators, and related notes into a unified phenotypic profile.
Stage 2
Cited multipliers
Outputs become guideline-backed risk multipliers drawn from peer-reviewed literature, graded for evidence strength, and shown in the report rather than hidden in a composite score.
From prototype to workflow
Build and validate before hospital-system access is required.
The current HerArc prototype structures uploaded visit summaries and prior health reports using published clinical evidence. Partnerships and EHR integration strengthen the product later; they are not prerequisites for testing the core engine.
Now
Users upload PDFs of doctor visit summaries and previous health reports. The engine extracts history and generates cited reports for prototype evaluation.
Next
Testing with women in the target population and clinician review of whether summaries and recommendations are useful in real care conversations.
Scale
SMART on FHIR integration so longitudinal history is reconstructed automatically and findings surface inside the physician’s existing workflow, including Epic.
On the roadmap
HerArc first. Additional modules next.
Later products can reuse the same Astravia extraction, evidence grading, and EHR delivery spine for other historical risk domains. HerArc is the product we are designing and validating now.