SCRIMED Atlas Intelligence Core v1
A continuously validated healthcare intelligence operating system.
Faith-neutral, compliance-centered operating model for governance, trust, ROI, interoperability, and agentic workflows. Atlas now exposes document intelligence, evidence-backed reasoning, Trust Cards, sandbox runtime, validation metrics, AI governance, and reimbursement readiness while live clinical execution remains gated.
Enterprise boundary
Enterprise Atlas surfaces must remain procurement-ready, compliance-focused, and institutionally neutral.
SCRIMED Atlas Intelligence Core v1 is a synthetic pilot and enterprise assessment operating layer. Medical-record, claims, payer, wearable, and telehealth workflows remain gated until tenant approval, BAA readiness, live connector controls, durable audit, and human review are approved.
Structural Intelligence Engine
Parse layout before LLM extraction across healthcare documents.
synthetic-ready
forms
Parse layout, labels, tables, and visual grouping before LLM extraction.
Business and synthetic healthcare forms only until live document controls are approved.
- field grouping
- checkbox state
- signature blocks
- required field detection
synthetic-ready
tables
Preserve table structure and source cell references before summarization.
No autonomous financial, billing, clinical, or coverage decision from extracted tables.
- row-column mapping
- merged cells
- header inference
- numeric unit detection
foundation-online
contracts
Extract clauses with page/section source references and governance review status.
Contract intelligence supports legal and operational review; it is not legal advice.
- section hierarchy
- defined terms
- obligations
- dates
- exceptions
synthetic-ready
referrals
Structure referral context into review queues with missing-information detection.
No autonomous referral acceptance, clinical triage replacement, or patient outreach.
- referral source
- requested specialty
- missing attachments
- routing cues
synthetic-ready
claims
Structure claims context into denial-risk and reviewer queues.
No final coding, billing, appeal, or reimbursement action without qualified review.
- claim lines
- denial codes
- payer reason text
- supporting documentation references
synthetic-ready
prior-authorizations
Map prior authorization packets to policy evidence, missing support, and reviewer checkpoints.
No payer submission, coverage guarantee, or medical-necessity determination.
- payer policy fields
- requested service
- evidence checklist
- authorization status
production-gated
medical-records
Medical-record parsing requires approved tenant controls and must preserve provenance before LLM extraction.
No live medical-record ingestion in current synthetic pilot mode.
- note sections
- medication tables
- lab tables
- problem list structure
- source provenance
Atlas Evidence Layer
Answers require citations, confidence, source attribution, and validation timestamps.
answer
Short operational answer or recommendation framed for human review.
citations
Source IDs, titles, URLs, and relevant evidence snippets or section references.
confidenceScore
Numerical confidence score with explanation of uncertainty and missing evidence.
sourceAttribution
Named source owner, version, validation timestamp, and scope boundary.
validationTimestamp
Timestamp for when source freshness and TrustQA checks were last evaluated.
humanReviewRequirement
Reviewer role and checkpoint required before any external, clinical, payer, or patient-facing action.
Operating subsystems
Atlas combines document intelligence, evidence, governance, reimbursement awareness, and validation.
synthetic-ready
Structural Intelligence Engine
Document layout understanding before LLM extraction across forms, tables, contracts, referrals, claims, prior authorizations, and gated medical records.
/atlas#structural-intelligence
- layout-first parsing
- source provenance
- field-level review
- no live PHI in synthetic pilots
foundation-online
Atlas Evidence Layer
Evidence retrieval from guidelines, protocols, publications, policies, standards, and buyer knowledge sources.
/atlas#evidence
- citations
- confidence score
- source attribution
- validation timestamp
synthetic-ready
Trust Card System
Attach provenance, confidence, source version, validation state, and human-review requirement to recommendations.
/trust#trust-cards
- source IDs
- confidence
- guideline version
- last updated
- review gate
synthetic-ready
Agent Sandbox Runtime
Isolated agent environments with memory, files, tools, audit logs, and workflow-specific boundaries.
/workflows#sandbox-runtime
- per-task sandbox
- scoped memory
- tool allowlist
- audit log
- blocked live execution
foundation-online
Continuous Validation Engine
Measure workflow outcomes instead of generic benchmark scores.
/observability#continuous-validation
- denial reduction
- time saved
- revenue impact
- escalation rate
- override rate
- trust metrics
foundation-online
Governance Layer
AI Asset Registry, shadow AI detection, model/prompt/source inventory, connector tracking, and audit trails.
/audit#governance
- asset registry
- shadow AI signals
- complete audit trail
- exception review
production-gated
Reimbursement Layer
CMS ACCESS-aware chronic care monitoring, telehealth, wearable integration, and outcome reporting readiness.
/atlas#reimbursement
- policy evidence
- reviewer approval
- consent model
- no reimbursement guarantee
Reimbursement Layer
CMS ACCESS-aware posture for chronic care monitoring, telehealth, wearables, and outcomes.
This layer supports readiness and reporting design; it does not guarantee reimbursement or perform billing actions.
production-gated
ACCESS-aligned chronic care monitoring readiness
care-plan adherence, remote monitoring review, escalation queue, outcome reporting
Supports readiness assessment and outcome-reporting design only; no reimbursement claim or care-management billing guarantee.
- CMS program policy
- tenant clinical protocol
- reviewer disposition
- patient consent model
synthetic-ready
Telehealth workflow intelligence
visit modality, follow-up need, access bottleneck, documentation completeness
No autonomous scheduling, patient instruction, billing, or clinical triage.
- tenant telehealth protocol
- state/regional policy review
- documentation policy
production-gated
Wearable integration planning
device signal availability, review thresholds, escalation policy, outcome trend
No live device monitoring or emergency-response automation in v1.
- device data agreement
- consent model
- clinical protocol
- alert fatigue policy
synthetic-ready
Outcome reporting layer
time saved, denial risk surfaced, access bottleneck, override rate, trust completeness
Reports operational pilot outcomes; clinical and financial claims require buyer validation.
- baseline method
- buyer-approved metric definitions
- audit trace
- validation timestamp