LimbusHealth Assess · Our approach
A draft adult ADHD assessment report, prepared for doctor review.
The patient completes a guided history, screening questionnaires, and an adaptive text interview. Limbus links the evidence to its sources. The doctor reviews, corrects, speaks with the patient, and decides.
The designed pathway
The pathway, end to end.
The intended sequence is intake and consent, structured screening, an adaptive interview, and doctor review. A case can pause or route out at safety, scope, or evidence-readiness gates.
- 01
Stage 01
Guided intake
- 02
Stage 02
Structured screening
- 03
Stage 03
Adaptive interview
- 04
The human acts here
Doctor review
Safety and evidence-readiness gates run throughout. A completed session is not automatically diagnosis-ready, and no flagged condition is auto-resolved.
Human oversight
Nothing is finalized without the doctor.
LimbusHealth Assess is designed so that the software can prepare evidence, but cannot finalize a clinical conclusion. The reviewing doctor sees the sources, questions or corrects the draft, resolves uncertainty, and signs the decision. That authority boundary is part of the workflow, not a disclaimer added afterward.
Safety checks run before diagnostic processing. Risk indicators, contradictory information, and presentations outside the adult-ADHD assessment scope can pause or route the case for human attention rather than being resolved automatically.
Where the evidence is incomplete, Assess is built to preserve the gap. Ending an interview does not make a case diagnosis-ready.
Methodology
Criteria first. Probabilities do not overrule them.
The binding path is a structured adult-ADHD criteria record. Findings are mapped to the A–E criteria, functional impairment, childhood onset, and source passages. A criterion that is unsupported stays unsupported; a completed interview cannot turn it into a pass.
A Bayesian pipeline may run alongside that record as advisory evidence during development. It cannot rescue a failed criterion, erase a contradiction, or finalize a conclusion. Mimics, comorbid signals, and route-outs stay visible as questions for review, not stand-alone diagnoses produced by the system.
Interview design
Adaptive, text-first, and bounded.
The current development candidate is text-first. It uses an adaptive conversation to ask follow-up questions, but the interview remains inside a controlled adult-ADHD assessment design. Optional voice or avatar delivery is not part of the active candidate.
AI assists the conversation and evidence extraction. Deterministic services hold the criteria, safety, readiness, and release gates. The result is meant to give the reviewing doctor a traceable starting point, not a finished answer.
Validation · planned
How we'll validate.
Current automated tests and synthetic rehearsals are engineering evidence. They are not a diagnostic-accuracy result, clinical validation, or permission to use the system with patients.
The next evidence stages require locked challenge cases and full role-play interviews, independent review, and predefined acceptance criteria. Prospective shadow comparison can begin only after the applicable clinical, regulatory, privacy, security, operational, and release gates are signed.
Any future diagnostic-performance claim will need its own protocol, reference standard, analysis plan, and independent review. Limbus does not make that claim today.
Software verification, clinical validation, and authorization are separate evidence states.
Regulatory status · as of August 2026
LimbusHealth Assess is software as a medical device in development for adult ADHD assessment. Its Canadian classification and authorization pathway remains unresolved, and it is not authorized for clinical use (as of August 2026).
LimbusHealth Assess is treated internally as software as a medical device because it is being developed for a medical purpose. Its exact Canadian classification and authorization pathway remain unresolved; no Health Canada authorization has been granted and no clinical use is authorized.
LimbusHealth Treat is a separate product with its own intended use, evidence, regulatory analysis, and release decision. Assess status does not authorize Treat, and neither product is currently available for patient use.
We do not make claims about diagnostic accuracy or clinical efficacy. Those will be established through formal study and regulatory review before we share any results.
Assess waitlist
Join the list for future Assess access.
Choose whether you are joining as a patient, doctor, or clinical organization. We will contact this list about possible patient access, clinician pilots, and other opportunities to use Assess. It is not available for clinical use today, and joining is not a request for an assessment or a promise of access.
- For
- Patients, doctors & clinics
- Purpose
- Future access
- Status
- In development
ASSESS INTEREST RECORD
DOUBLE OPT-IN
Request recorded · confirmation sent
Check your inbox to confirm your Assess waitlist request.
Your request is not confirmed until you click the link. It expires in 24 hours.
This is not a request for clinical care or an assessment, and it does not promise product access.