Skip to main content
Alma Health AI
Home / Plans & Providers
For Healthcare Plans & Risk-Bearing Provider Groups

Three cost-avoidance levers. One platform. Audit-tagged on every recommendation.

Inappropriate-referral prevention, operational care continuity, and decision defensibility — built for healthcare plans and risk-bearing provider groups. AI surfaces guideline-concordant recommendations clinicians can act on immediately, reducing patient wait times and protecting provider licensure through documented, audit-tagged decision support. Every recommendation logged with source, version, and timestamp.

Downstream risk reduction — including avoided ER visits and preventable hospitalizations — is not yet quantified but directionally supported by operating data.

Each module deploys independently and integrates with your existing tech stack — or with Alma's other modules. Adopt the Referral Review Engine for immediate cost-avoidance impact, or all six modules for closed-loop accountability across the patient journey.

EHR integration
eClinicalWorks athenahealth NextGen Allscripts Greenway Veradigm Epic Cerner / Oracle Health Meditech
Drop-in integration for ambulatory EHRs.
Where the economics improve

Three operational levers, directly measurable.

01

Inappropriate-referral prevention

The Referral Review Engine identifies primary-care-manageable conditions and routes them to PCP-LED management with cited guideline rationale. Reduces specialty referrals 40%, appropriately — with >88% clinician concordance. Hardcoded safety escalations ensure conditions genuinely requiring specialty input are never deferred. With optional preferred-network integration, the engine also reports out-of-network leakage rate alongside appropriate-referral rate — bundled metrics for plan dashboards.
02

Operational care continuity

The Care Management Platform handles TCM/CCM enrollment with cumulative-time tracking and provides EHR-documented follow-through for high-risk patients identified across the platform. Care Management notes are required to be listed at the next encounter — making closed-loop accountability a clinical commitment, not just a data flow.
03

Decision defensibility

Every recommendation across the platform is logged with cited guideline source, engine version, guideline version, and timestamp. Append-only audit trail via Postgres triggers; field-level diffs preserved. Exportable for plan audits and CMS data validation. The override workflow requires a documented reason that becomes part of the recommendation record.
Why plans and providers care

"Specialty referrals are the largest controllable spend after inpatient. Defensible inappropriate-referral prevention is what every plan is trying to underwrite — and what most platforms can't actually prove."

The economics · 30–40% of specialty referrals are PCP-manageable per industry reporting · $1,200 (all-in: visit + imaging + workup + leakage) per inappropriate referral · referral rates and medical loss ratio are universally tracked across risk-bearing organizations
What we're not

Decision support — not autonomous decisioning.

There is a category of products that promise to remove the clinician from the loop. We are not that. The distinction matters at audit, at MA plan diligence, and most of all to the patient.

What we don't do

Autonomous AI decisioning

  • AI makes referral decisions that providers cannot override
  • Clinical judgment is replaced rather than supported
  • Safety decisions are AI-configurable, not hardcoded
  • Audit records are summary-level, not recommendation-level
  • Coverage and prior auth-style gatekeeping
What we are

Augmentative decision support

  • AI surfaces a recommendation; clinician decides
  • Override authority preserved on every non-safety decision
  • 39 hardcoded safety escalations are non-overridable by anyone
  • Immutable per-recommendation audit record with cited source
  • Clinical workflow tool — not a payer gatekeeping system
What a pilot looks like

30 days. Live in days. Customizable. Compliant.

Pilot structure

  • 01Day 1: Standalone platform available immediately. Referral teams can begin processing referrals through Alma's interface as requests come in — no integration required to start.
  • 02Week 1: EHR integration complete for most ambulatory and major hospital-grade systems. Our team handles the technical lift end-to-end.
  • 03Week 2: Joint clinical review. Alma physicians and your clinicians align on guideline application, specialty matrix calibration, and edge-case handling. Workflow adjustment list captured.
  • 04Weeks 3–4: Workflow customization complete. Automation tuned to Client's preferred level — from full clinician review of every recommendation through to audit-tagged auto-routing — while maintaining compliance and audit standards required for health plans.

Robust enough for health plan deployment. Flexible enough to meet each Client where they are.

Success criteria

  • A≥30% reduction in total referrals, appropriately, at pilot sites
  • BOverride rate <15% — clinicians find recommendations defensible
  • CZero missed safety escalations in independent review
  • DProvider satisfaction (NPS-style) ≥ +20 after 30 days
  • EEngine uptime ≥99% across the pilot window
The financial case

Modeled cost avoided — decomposed by category.

Two tables. The first is the conservative specialist-only floor. The second is the full external-orders opportunity — specialist referrals, imaging, and ancillary orders, all reviewed by the engine via workup-verification gates.

Specialist referral baseline.

Specialist referrals only. 40% total-avoided rate × $1,200 (all-in: visit + imaging + workup + leakage) per episode.

Panel size
Referrals/month
Net avoided (40%)
Annual savings
5 PCPs (~10K patients)
200
80
$1.16M
10 PCPs (~20K patients)
400
160
$2.31M
25 PCPs (~50K patients)
1,000
400
$5.76M
50 PCPs (~100K patients)
2,000
800
$11.6M
100 PCPs (~200K patients)
4,000
1,600
$23.1M

40 specialty referrals/PCP/month (320 visits × 12.5% specialty referral rate) × 40% total-avoided rate × $1,200 (all-in) cost per episode = $19,200/PCP/month. Cost-per-episode anchor consistent with peer-reviewed eConsult cost-savings literature: Anderson et al., AJMC 2018 ($655 cost reduction per avoided face-to-face cardiology consult); Anderson et al., Health Affairs 2019 ($82/patient/month specialty episode reduction).

Total external orders — specialist + imaging + ancillary.

All three categories reviewed by the engine via workup-verification gates. Category-specific cost anchors.

Panel size
Orders/month
Avoided (40%)
Annual savings
5 PCPs (~10K patients)
350
140
$1.43M
10 PCPs (~20K patients)
700
280
$2.86M
25 PCPs (~50K patients)
1,750
700
$7.14M
50 PCPs (~100K patients)
3,500
1,400
$14.3M
100 PCPs (~200K patients)
7,000
2,800
$28.6M

Decomposed at 40% total-avoided rate (validated production data): specialist refs (40 × 40% × $1,200 all-in) = $19,200/PCP/month + imaging (20 × 40% × $500 direct cost only, client-supplied) = $4,000/PCP/month + ancillary (10 × 40% × $150) = $600/PCP/month = $23,800/PCP/month combined. Imaging and ancillary cost anchors via Medicare PFS reference.

Note on what this model excludes

This model excludes the internal operational savings the engine generates — coordinator time recovered from chasing specialist notes, eliminated rework on incomplete referrals, and the HEDIS gap-closure opportunities recovered when the patient stays in primary care rather than being routed to a specialist. These are real and material; we exclude them to keep the financial case anchored to externally observable cost categories. Internal-savings modeling is available under NDA.

Anchored evidence base: Anderson et al., American Journal of Managed Care, 2018 (PMID 29350511). Anderson et al., Health Affairs, 2019. Forrest, Reasons for Outpatient Referrals from Generalists to Specialists (~30% of referrals possibly appropriate or inappropriate). Albini et al. (37% of referral forms classified inappropriate). Kaul et al., AJMC, 2016 (52% of US PCPs reported making unnecessary referrals on patient request).

Payer-ready data

Auditable, guideline-concordant referral data — for CMS, MA plans, and quality reviews.

The same engine outputs that drive operational analytics also produce the documentation payers and reviewers ask for. Three clients, one recommendation record per referral.

Medicare Advantage

STAR ratings and HEDIS measures.

  • — PCP-first management supports HEDIS measures
  • — Referral appropriateness for STAR ratings
  • — Reduced specialist utilization → lower MLR
  • — Audit trail demonstrates VBC delivery
MSSP & ACO

Auditable referral patterns.

  • — Referral patterns — now auditable
  • — Guideline adherence per encounter
  • — Shared savings via reduced per-episode cost
  • — Quality metrics tied to recommendation trail
Quality reviews

CMS audit and accreditation.

  • — Every recommendation timestamped and guideline-cited
  • — Audit tags per recommendation (CM, safety, best practice)
  • — Engine + guideline source version tracked per record
  • — CSV export — ready for CMS audit or accreditation
Regulatory tailwind

Aligned with the way healthcare is paying.

The Referral Review engine supports the full spectrum of value-based care. That includes CMS Innovation Center priorities — the Advancing Chronic Care with Effective, Scalable Solutions (ACCESS) Model, the Long-term Enhanced ACO Design (LEAD) Model, and the Medicare Shared Savings Program — and extends to Medicaid managed care, commercial risk arrangements, and employer direct contracting. Any organization measured on quality, total cost, and patient outcomes can use Referral Review as part of its clinical and operational foundation.

The market is moving.

CMS is tying reimbursement to documented guideline adherence. Medicare Advantage plans are requiring referral appropriateness data for STAR ratings. Value-based contracts reward reduced specialist utilization with auditable documentation.

Primary care groups that can prove guideline-concordant referral management will win contracts. Those that cannot will lose them.

The question is not whether you will be asked for this data. It is whether you can produce it when asked.

See it on a sample of your referral data.

A 30-minute working session. We'll walk through the recommendation flow, the audit record, the override workflow, and the safety escalations — using a sample of referrals from your network.

Request a working session →