This page explains how Vertix's AI components work, what they are designed for, what they are not designed for, and your role as the licensed clinician.
1. What Vertix is
Vertix is a Clinical Decision Support (CDS) tool qualified as Non-Device CDS under 21st Century Cures Act § 3060 (codified 21 U.S.C. § 360j). The tool:
- Displays, analyzes, and communicates clinical information from peer-reviewed literature, clinical guidelines, and government clinical sources.
- Provides recommendations that are informational and supportive of, but not a substitute for, your independent clinical judgment.
- Allows you to independently review the basis for every recommendation through inline citations and evidence trails.
- Is not intended for time-critical, life-threatening decisions and is not for use with implants, life-support, or similar devices.
Because Vertix meets these four Cures Act § 3060 criteria, it does not require FDA 510(k) clearance.
2. What the AI components do
- Live Co-Pilot: surfaces relevant evidence and DDx options during your session in real time.
- Multi-Paradigm Reasoning: runs the same case through 34 paradigm-aware engines (CBT, psychodynamic, systemic, biomedical and 30 more) and presents lenses in parallel.
- Citational DNA: every clinical claim is linked to the exact span of a peer-reviewed source paper.
- Replication-Weighted Retrieval: papers that failed replication are deprioritized.
- Devil's Advocate Engine: argues against the current treatment plan with contrarian evidence.
- 60-Second Case Formulation: turns intake notes into multi-lens formulation, DDx, safety screen, and five cited papers in under a minute.
3. What the AI does NOT do
- It does not diagnose patients. Diagnosis is your professional responsibility.
- It does not prescribe medications. Prescription decisions are yours and must follow your scope of practice.
- It does not replace mandatory reporting obligations. State-specific reporting requirements remain your duty.
- It does not provide emergency medical care. In a crisis, contact 988 (Suicide & Crisis Lifeline) or 911.
- It is not certified for use in surgical, life-support, or implanted-device contexts.
4. Clinician review is required
Every recommendation produced by Vertix must be independently reviewed by you against:
- Your professional judgment and clinical experience.
- The patient's actual clinical presentation, history, and preferences.
- The standard of care in your jurisdiction.
- Your scope of practice under your license.
You retain full clinical authority and full responsibility for every decision. Recommendations are starting points, not endpoints.
5. Training data and patient privacy
Our reasoning engines are trained exclusively on peer-reviewed literature, clinical guidelines, and government clinical sources. Your patient data is never used to train any AI model — ours or anyone else's. Patient data is architecturally separated from the training pipeline at the database layer (Patient Privacy Firewall, ADR-036).
6. Known limits and bias
AI systems trained on published literature inherit the demographic biases of that literature. Most published mental health research overrepresents non-Hispanic white US samples. To address this, every recommendation in Vertix surfaces a Bias & Generalizability Auditor that calculates the demographic footprint of the underlying evidence and flags when generalizability to your patient's population is limited.
You should treat all AI outputs as one input among many, and weight them appropriately when the evidence base does not match your patient's demographic profile, cultural context, or clinical complexity.
7. Audit trail
Every query, source consulted, recommendation generated, and clinician action is logged in an immutable audit trail. The full chain is exportable as legal-quality PDF or structured JSON for court submission, licensing board review, or payer audit.
8. Updates and changes
The AI components are updated continuously. Material changes to capabilities, scope, or limits will be communicated by email and reflected in the changelog.
9. Questions
Questions about how the AI works in your specific use case: [email protected].
10. Diagnostic criteria database — data sourcing
AI you can stand behind. Every word from a real paper. Other AIs guess. Vertix cites.
We separate what the literature says from what the AI synthesizes. The Reference Library is the peer-reviewed mental-health corpus, ingested from OpenAlex, PubMed, CrossRef and Semantic Scholar under a strict domain scope filter (psychiatry, clinical psychology, clinical neuroscience, public health). The Decision Engine is the curated, dual-coded ICD-10-CM and DSM-5-TR, evidence-ranked subset that the AI actually uses to answer you — embedded for semantic retrieval, BM25-indexed for keyword retrieval, Tier A/B/C/D classified.
No open web. No phantom citations. No “the model says”. A chain of real papers from your screen to the DOI, ready when anyone asks — colleague, payer, audit, or court. The criteria-reconstruction methodology follows Feist Publications v. Rural Telephone Co., 499 U.S. 340 (1991): factual content is not copyrightable. Every criterion in our database has a transparent provenance trail to source papers.
Vertix uses ICD-10-CM (US public domain, NCHS/CMS) as the operational coding anchor. DSM-5-TR is referenced via tag scheme (chapter, disorder name, ICD-10-CM cross-walk) and via short fair-use quotations of ≤30 words with attribution per US fair-use doctrine. We do not reproduce DSM-5-TR criteria sets verbatim. PDM-2 is referenced bibliographically only. ICD-11 codes are used as cross-reference (codes are facts; ICD-11 textual descriptions remain under WHO CC BY-ND 3.0 IGO and are not derivative-modified). RDoC (NIMH public domain) and HiTOP (open academic) are research-grade dimensional companions.
This approach is governed by our internal Diagnostic Criteria Reconstruction Policy (PP-16 v1.1), which establishes per-manual operational rules, the extraction pipeline, audit-trail requirements, and the Postgres schema that make every criterion defensible in legal or regulatory review. ICD-10-CM annual fiscal-year cut-over per PP-16 §10 ensures Vertix stays aligned with the version mandated for US billing (currently FY 2026; cut-over to FY 2027 by Sept 29, 2026).