>88% clinician concordance. Guideline-concordant. Audit-tagged.
A guideline-concordant recommendation on every referral. A cited primary clinical guideline on every recommendation. An audit-tagged recommendation record on every call.
Enter the requested specialty and clinical indication. The engine returns a guideline-concordant recommendation in under ten seconds.
Well-controlled essential hypertension on monotherapy meets ACC/AHA criteria for primary care management. Consider specialist consultation if BP becomes resistant (≥3 medications) or evidence of secondary causes emerges.
39 non-overridable escalations fire before any other engine logic. Cannot be bypassed by any user — including administrators. Categories: Urgent, Escalate, Emergent, Specialist Required. The boundary comes from clinical society guidelines (ACC/AHA, GOLD, ADA) — not from the tool.
PCP judgment always supersedes the engine. Override available for any result except the 39 hardcoded safety escalations. Every override requires a documented reason. The tool never acts alone.
Every recommendation timestamped, audit-tagged, and bound to engine version and guideline source. Immutable record per referral. Closes failure-to-diagnose documentation gaps (Studdert et al., NEJM 2006).
Hardcoded. Fire before all other engine logic. Cannot be bypassed by any user.
Routes to CMO. Outpatient vs. ED decision.
Routes to CMO. May need inpatient.
Immediate transfer. No delay.
Approved — specialist referral required.
Every fire is timestamped and audit-tagged (AU_URGENT / AE_ESCALATE / ET_EMERGENT / AAS). Immutable record per referral.
Not a black box. Every claim below maps to a verifiable piece of the production system, available for code review under NDA.
Across 40 specialty matrices spanning cardiology, endocrinology, GI, neurology, dermatology, orthopedics, urology, ENT, pulmonology, and twenty more. Each condition has guideline-concordant routing logic and a cited primary source.
Coordinators type how they think — "diabetis," "atrial fib," "bad back," "fluttering heart," "high BP that won't come down." The engine handles the messy reality of real referral text.
Cauda equina, ACS, hypertensive emergency, suspected stroke. These rules fire before any other engine logic and cannot be bypassed by any user — including administrators. Patient safety is hardcoded, not configurable.
Every recommendation is logged with engine version, guideline version, cited source, timestamp, and recommendation ID. Append-only via Postgres triggers. Field-level diff tracking. Exportable for plan audits and CMS data validation. The Referral Review Engine is the entry point to closed-loop accountability — every recommendation logged, every CM-qualifying referral routed onward.
Optional integration with your network configuration. Engine reports both appropriate-referral rate and out-of-network leakage rate. Routing logic respects in-network preferences while preserving guideline concordance.
Beyond the 39 hardcoded safety rules, board-certified clinicians review flagged recommendations and edge cases on a structured cadence. The engine surfaces; clinicians decide. The clinical voice in the loop is real, not theoretical.
Surgical checklists reduced mortality 47%. Not by replacing surgeons — by catching what cognitive load causes them to miss. (Haynes et al., NEJM 2009.)
The gap is cognitive load, not negligence. A PCP managing 2,000+ patients cannot hold 10,000 guideline recommendations in memory. ~45% of recommended care goes undelivered. (McGlynn et al., NEJM 2003.)
The engine provides the starting point. Clinical reasoning provides the endpoint. Complementary — not competing.
A clinician matching the engine's logic by hand needs UpToDate, Lexicomp, five society guideline databases (ACC/AHA, ATS/CHEST, ACR, ADA, USPSTF), and the EHR — across three applications and seven screens. The engine collapses that into a single recommendation surface with the guideline cited inline.
The engine isn't an added step. It replaces the steps clinicians already do — badly, inconsistently, or not at all. ~45% of recommended care doesn't get delivered when clinicians work from memory alone (McGlynn et al., NEJM 2003). The "cumbersome" tool prevents the errors that happen when clinicians wing it.
of recommended care not delivered to patients
of referrals lack adequate clinical information
of non-adherence from lack of awareness
of guideline failures from access barriers
The "cumbersome" tool prevents the errors that happen when clinicians wing it.
Every recommendation the engine produces feeds five operational metrics, available per PCP, per coordinator, per site, per specialty. Surfaces clinical and operational outliers — variance across PCPs, coordinator training gaps, specialty utilization anomalies, and aggregated community-health signals.
Recommendation mix per PCP, per site, per specialty. Identifies high-leakage specialties and PCPs whose patterns differ materially from peers.
Captured before the referral leaves the building. Drives the cost-avoidance line in the financial case.
Override exceeds 15% on any condition → pathway is revised. The system adapts to clinicians, not the other way around. >88% clinician concordance.
Zero missed safety escalations observed in production to date. Independent review of the safety-rule layer is the cleanest item to validate first.
40% total-avoided rate (validated production data) × $1,200 (all-in) per specialist episode (peer-reviewed eConsult anchor) yields ~$19,200 per PCP per month modeled. The full decomposed methodology — including imaging and ancillary categories — is on the For Plans & Providers page.
Aggregated to a community-health signal
Aggregated referral patterns indicate which conditions and specialties are most utilized in a given geography — a community-health signal usable by public health departments and county health systems to identify specialty-access gaps and emerging condition burden.
A clinical reference tool. Recommends. Does not prescribe, order, or refer.
The PCP's independent clinical judgment remains the standard of care.
The audit trail documents the recommendation and the clinician's decision — creating a defensible record.
Guideline-concordant documentation strengthens legal defensibility (Studdert et al., NEJM 2006).
We recommend general counsel review as part of pilot planning. Every client has.
A 30-minute demo. We'll walk through the recommendation flow, the audit trail, the override workflow, and the safety escalations — using a sample of your referral pipeline.
Request a demo →