Clinical research
Pause-Health.ai's clinical foundations sit in three layers: peer-reviewed menopause-care guidelines we already cite in code (factual), a validation and bias-monitoring program designed against those guidelines (planned, gated on the first design-partner BAA), and a transparency commitment for how we will publish methodology and results regardless of direction.
What's referenced
Wired in code todayThe peer-reviewed menopause-care guidelines the Care Router policy and MSCP-pathway heuristics already cite. Each entry notes where in the open-source codebase the reference actually shows up, so a clinical reviewer can verify the claim without taking our word for it.
Position statements on hormone therapy, vasomotor symptoms, and the MSCP (Menopause Society Certified Practitioner) credential.
Where in code: lib/menopause-society.ts (deep-link referral path), lib/risk-band.ts (HRT suitability heuristic), /proposal/menopause-society
Clinical Practice Guidelines on management of menopausal symptoms; postmenopausal bleeding workup priorities.
Where in code: lib/risk-band.ts (postmenopausal bleeding red-flag check), lib/care-router.ts (urgent-gynecology pathway trigger)
Global consensus statements on menopause hormone therapy, cardiometabolic considerations, and quality of life endpoints.
Where in code: lib/risk-band.ts (CVD / BMI red-flag check for HRT deferral)
Clinical Practice Guideline on postmenopausal hormone therapy and treatment of menopausal symptoms.
Where in code: lib/care-router.ts (mscp-virtual-visit + mscp-in-person pathway thresholds)
What's planned · validation program
Gated on first design partnerThe validation and bias-monitoring program designed to run alongside the first design-partner deployments. Each block is pilled to distinguish what's a finalized study design from what's gated on cohort availability.
~10k women across 3-5 design-partner provider organizations, stratified by menopause stage (peri / post), symptom-cluster severity, and care-access setting. Recruitment begins after the first design-partner BAA is in force.
Care-Router pathway concordance vs. blinded clinician adjudication; symptom-burden trajectory (vasomotor / sleep / mood) at 30 / 90 days post-intake; time-to-MSCP-visit for high-burden cohorts.
Two-clinician adjudication of pathway routing on a stratified random sample, blinded to model output. Disagreements resolved by a third senior reviewer.
Pre-registered with the validation protocol; primary analysis is the pathway-concordance Cohen's kappa with 95% confidence intervals. Sensitivity analyses by age, ethnicity, and care-access setting.
Subgroup pathway-distribution + concordance reported per cohort dimension (age, race / ethnicity, insurance status, geography) at the same cadence as the validation report. Designed; cadence depends on cohort enrollment rate.
Validation studies submitted to the design-partner organization's IRB. The platform's prototype use of synthetic personas is not human-subjects research.
What we will publish · transparency commitment
Methodology, results, and bias-monitoring reports — published regardless of direction. The codebase is already public; the research artifacts will follow as the validation program runs.
Pause-Health.ai Care Router policy, grounding architecture, and validation protocol — preprint posted at the start of the validation phase so peers can critique the design before results land.
Primary and pre-registered secondary endpoints reported regardless of direction. Submission to a peer-reviewed clinical informatics or menopause-focused journal.
Public report on subgroup pathway distributions + concordance, refreshed at the cadence the validation cohort supports. Methodology and code published alongside.
The full Care Router policy, risk-band heuristic, and grounding integration are public at github.com/hucmaggie/pause-health.ai today — no opaque ML black box.
For clinical reviewers + research partners
The investor-brief deep-dive on hypotheses we're carrying into design-partner research is at /proposal/insights. Source code is public at github.com/hucmaggie/pause-health.ai. For clinical-advisory inquiries, validation collaborations, or pre-registration discussions, email research@pause-health.ai.
