Frequently Asked Questions

Even well-documented appeals can hit a wall — because clinical criteria define eligibility, not how payers actually enforce it. Same clinical scenario, same criteria, different outcomes. Experience beats correctness in this system — and that’s the gap we close.

Not at all. It’s patterned but undocumented. Look closely at large claim volumes and you’ll see consistent enforcement trends — they’re just never formally published.

We zero in on medical necessity disputes — the cases where payer interpretation varies most and human judgment tips the outcome.

Industry data shows 10–30% of denials require human expertise, and 80.7% of appealed denials are overturned, yet only 11.5% ever get appealed. The gap isn’t automation capability — it’s knowing which denials to contest and what evidence sequences work.

BH medical necessity criteria are inherently more subjective. “Clinically appropriate” for PHP or IOP requires judgment calls about symptom severity, functional impairment, and treatment history — and those calls vary by payer and reviewer.

A structured record of what actually worked — which evidence combinations, in what sequence, moved outcomes for a specific payer and denial type. It’s institutional memory, captured and reusable. See how it works →

A cluster is a targeting scope — a specific combination of payer + denial pattern + level of care (e.g., “United medical necessity denials for residential BH” = 1 cluster). A Precedent Object is a deliverable produced within a cluster — an individual evidence kit, narrative template, or reimbursement argument built for a specific denial scenario. A cluster defines where you’re working; precedent objects are what you produce there. One cluster typically yields multiple precedent objects.

We capture how your team solves complex medical necessity cases and turn that reasoning into reusable, structured intelligence: outcome-labeled precedent objects, expert-validated playbooks, explainable reasoning chains, and versioned intelligence.

All the major frameworks your team already knows: ASAM, MCG, and InterQual — captured at the enforcement level, not just the requirement level.

Yes — fully. All data handling follows HIPAA requirements with BAAs in place. Precedent Objects are de-identified by design: payer ID + denial code + clinical scenario + evidence elements + outcome. No PHI.

No — and that’s intentional. RCM platforms automate routine tasks. We focus on the 10–30% of complex cases where human judgment determines outcomes — and make that expertise compound over time.

Every cohort partner completes a Development Partner Sprint — a fixed-scope engagement scoped to your denial landscape. Lite Sprints target 1 cluster over 30–45 days. Full Sprints cover 3–5 clusters over 45–60 days. Pricing is discussed during the application process.

Most partners see measurable improvement within 60–90 days.

Fast. Sprint kickoff is 1–2 weeks, and you’ll see first precedent captures within 30–60 days.

Sprints work alongside your current workflow — no integration required. The Platform embeds guidance where your team already works.

No. We charge for expertise capture and intelligence access — never contingency fees.

AI RCM automates high-volume, low-complexity claims. We operate on the complex 30% that AI can’t template. Explainability shows what the model decided. Governance lets your team fix it when it’s wrong.

MCG and InterQual define what should be required. We capture how payers actually enforce those requirements. Requirements vs. enforcement — we operate on the gap between published criteria and real-world behavior.

The RCM industry optimized for speed and deferred governance. AI tools retrain every 2–6 months. Payer enforcement shifts every few weeks. That gap — governance debt — compounds against you. The longer you run on stale models, the more revenue you leak. Stratum closes the gap with 1–2 week drift response and human-editable precedent.

1–2 weeks. AI model retraining takes 2–6 months. Manual biller adaptation takes 30–90 days. Stratum detects drift through structured precedent monitoring — when a payer shifts how they enforce, your billers flag it and the precedent updates that same week.

Three things. Structural: Precedent Objects are human-governable — readable, editable, version-controlled. No black box. Compounding: Solving one payer-denial cluster gives adjacent clusters a 60–70% head start. Each sprint makes the next one faster. Speed: 1–2 week drift response vs. months. A competitor can copy the schema but not the library, the speed, or the compounding.

Yes — our Research track is built for academic health systems through ARPA-H co-development partnerships. Partners contribute in an advisory capacity or funded evaluation role, scoped to validation methodology and study design. First-mover access to precedent infrastructure, co-authorship on research, and openings for sub-awardee on ARPA-H submittal.