Clarify Health

The models do real work here, inside an offer the company now sells as a service rather than software. Meridian is described as an outcomes partnership combining market data, technology and a field team, and its prioritization step, which ranks where a health system can gain most within its capacity and contract constraints, is described as AI driven. The company states more than 20 billion AI predictions in production on the Atlas data layer.

Remove the models and a buyer would still hold the data asset, the benchmarking built on it and a field team working playbooks, which is why this is not A. The prioritization itself does not exist without them, which is why it is not C.

The product now acts rather than only reporting, and oversight is described for one half of it. On the Meridian side, AI driven prioritization produces playbooks that the company's field teams and the health system's own staff carry out, so people make the network, affiliation and rate decisions. On the Arrival side, the product reaches patients directly: outreach to patients with documented need who are not seeking care, a digital front door that ends in a booked appointment, and live scheduling. Nothing on the home or about pages says what triggers those contacts, what a patient is told, or when a person steps in.

No accuracy or confidence figure is published for the rankings that decide which specialists receive redirected referrals.

Scale and data sources are quantified; method and performance are not.

What is disclosed is informative about shape. The dataset is quantified at more than 300 million mapped lives with more than 20 billion predictions generated, and source categories are named rather than left vague, covering health plan claims, pharmacy data, government sources and negotiated pricing rates. Techniques are identified as predictive analytics, machine learning and natural language processing, and the architecture is described as modular stackable building blocks on a common platform. Leadership has also explained the approach by analogy to the statistical methods popularized in professional sports, which makes the intent clear.

What is absent is anything measurable. No accuracy, discrimination, calibration or error figure is published for any model, no model card exists, no validation methodology is described, and no operating point is given for predictions that drive network and referral decisions.

One phrase recurs and carries no content. Precise methodologies is used repeatedly to describe the analytic approach, and precision is a claim that ordinarily arrives with a number. None accompanies it.

The prediction count is a volume measure rather than a quality one. Twenty billion predictions says how much the platform produces and nothing about how often it is right.

The architecture is described and no model provider is named. One data layer, the Atlas platform, sits under both products and combines claims across Medicare, commercial and managed Medicaid, prescription records, price transparency rates and social factors, joined to each customer's encounter data and contracts. The digital front door is supplied by Loyal Health, now owned by Clarify. No data supplier, claims aggregator, hosting provider, model provider or subprocessor is named on the home, about or privacy pages.

For a company whose main asset is licensed data, the data suppliers are the part of the chain that matters most, and whether their licenses permit the use made of them is something a buyer cannot check.

Vendor reported figures, credited to customers, with no method behind them. The home page states $3.2 billion of documented customer ROI and a typical $65 to $100 million of annual opportunity per health system, and says results are measured against a baseline agreed before work starts. No method, denominator, period or customer is given for any of it, and the main site pages name no customers at all.

The recognitions listed in the resources library, Best in KLAS 2024 for data analytics platforms for payers among them, measure customer satisfaction with an earlier payer analytics offer, not whether the current referral and activation work improves anything. No peer reviewed publication or model accuracy figure is published.

The public reports page lists annual reports each November since 2020, which it describes as measuring patient safety indicators across Medicare, Medicaid and commercially insured populations; they report on providers rather than on the product's effect.

General assurances that do not answer the questions artificial intelligence raises. The privacy policy, last updated 2 August 2024, says protected health information is used and disclosed only to provide the services and as HIPAA and the provider's agreement permit, is protected under the HIPAA rules, and reaches third parties for research only once de identified. It also permits use of personal information for developing new products and keeps it for as long as its purposes require. It does not say whether customer data trains models or joins the shared data layer.

The larger question sits outside that document. The Atlas layer links claims, prescription records, price data and social factors for more than 300 million patients and joins them to each health system's own encounter data. Nothing on the home, about or privacy pages describes the linkage method, the identifiers used, re identification safeguards or retention for that dataset.

Status is stated and supported by a substantive privacy document; the agreement and its scope are not published. The privacy policy, last updated 2 August 2024, says Clarify uses and discloses protected health information only to provide its services and as HIPAA and its agreement with the provider permit, that HIPAA obligations and provider contracts govern wherever they conflict with the policy, and that protected health information is safeguarded under the HIPAA rules.

Scope is the question. The policy is written for a provider and patient application, with patient registration and linked accounts, and says nothing about the Atlas data layer of more than 300 million patient journeys, the legal basis on which it is assembled, or how a health system's encounter data is handled once joined to it. No business associate agreement, execution requirement or subcontractor position is published.

The vendor's own material names a recognized certification, and nothing says whether it has been renewed. A HITRUST risk based certification was announced on its news pages as Clarify's own. No date survives for the announcement, and a risk based HITRUST certification runs for two years. The site carries no security page, trust center or attestation, and the footer links only the privacy policy, terms of use and acceptable use policy.

The holder is the vendor rather than a host. Scope and currency are the open questions for a platform holding linked records on more than 300 million patients. Loyal Health, which Clarify acquired, states its own SOC 2 Type 2; that covers the Loyal engine behind Arrival's digital front door and nothing wider.

No device claim is made and the product is scoped as network, access and financial work for health systems. Meridian recommends where referrals, affiliations, sites and rates should move, and Arrival contacts and books patients; neither is described as diagnosing or deciding treatment, and the clinical decision stays with the physician at the visit.

Outreach selects patients with documented need who are not seeking care, from claims and records their clinicians did not review for that purpose. The output is an invitation to book rather than a clinical finding, and the selection criteria are still a fair buyer question.

The regimes that do apply concern data. The public reports page carries annual quality reports labeled QE, the abbreviation CMS uses for its Qualified Entity program, under which Medicare data is released on condition of public reporting; the page does not spell out the designation.

Nothing published on how model behavior is governed or tested. The home page, about page, privacy policy and resources library carry no fairness testing, subgroup performance, evaluation method or governance framework. The leadership page names a Chief AI and Privacy Officer, Erik Talvola, and nothing describes what that role reviews or approves.

The mechanism at risk is specific. Meridian steers referrals toward specialists who score higher on quality and efficiency. Efficiency built from cost and utilization history carries the problem Obermeyer and colleagues documented in Science in 2019, where cost used as a proxy for need understated the needs of Black patients because less had been spent on their care at the same level of illness. A specialist whose patients arrive sicker, or have had less spent on them, can score as less efficient for reasons unrelated to the quality of care, and steering volume away from that specialist changes access for the people they serve.

For Clarify Meridian, the who we help page describes a single outcomes based fee, tied to improvement in value per patient per service line against a baseline set with the customer before work starts, and calls the fee structure 100 percent outcomes based rather than a software subscription. The measure and baseline are set per contract. That is a published commitment that can be tested, and it puts the whole fee at risk: when the improvement does not arrive, Clarify is not paid.

It answers for the partnership's result, not for what any one recommendation said. Results against a baseline move for many reasons, including the health system's own execution and payer behavior, and Meridian includes Clarify's field team, so the fee is a risk share on the whole engagement. Nothing runs to a specialist whose referrals were redirected or a patient steered on a flawed score, and no accuracy figure, warranty or remediation position is published. The home page describes Arrival as measured against an agreed baseline too, counted in booked in network appointments, without stating that its fee depends on it.

The headline figure of $3.2 billion in documented customer ROI carries no method, period or breakdown.

Integration is claimed and no system is named. The home page says each Meridian answer arrives as a playbook delivered into the system the customer's team already works in, that Arrival follows a referral until the patient is booked and seen, fills capacity through live scheduling and returns closed loop confirmation to the referring physician, and that Clarify's data is joined to the health system's own encounter data and contracts. Each of those needs a connection to scheduling, referral or record systems. None is named on the home or about pages, and no standard, listing or interface documentation is given.

The digital front door comes from Loyal Health, which Clarify acquired, and its integrations are a separate question for the buyer.

Nothing published about where the system runs or where data rests. The home page, about page and privacy policy name no hosting provider, region, residency commitment or tenancy model.

Tenancy is the question that matters. One data layer serves more than 60 health systems and joins each one's encounter data and contracts to the shared claims and price data, and health systems in the same market compete for the same referrals. Nothing read says how one customer's data and derived analysis are kept from another's.

The pricing model is discoverable even though no price is published. The who we help page describes Meridian's fee as a single outcomes based fee tied to improvement in value per patient per service line against a baseline set with the customer, 100 percent outcomes based and not a software subscription. The home page presents Meridian and Arrival as outcomes partnerships with Clarify's economics tied entirely to measured results, and frames its pitch around one test for any analytics vendor: what happens to its fees if nothing moves. So the cost driver is public before any conversation.

What cannot be sized is the share. No percentage of the measured improvement, fee floor, term, or treatment of results that fall short is published, and the home page's typical opportunity of $65 to $100 million a year per health system comes without the fee side.

Coverage is stated clearly with nothing validating it. The company now names one buyer, health systems, and says it works with more than 60 of them; the earlier spread across health plans, physician groups, surgical centers and life sciences companies no longer appears on the home, about or resources pages. Use cases are named: service line and site decisions, affiliations and employment, rate positioning, performance across fee for service, Medicare Advantage, MSSP, TEAM and commercial contracts, referral steering, and patient activation from referral to booked visit.

No customer is named, and nothing states in which settings or markets the recommendations have been shown to work. The data layer spans Medicare, commercial and managed Medicaid claims, so payer mix is broad; which health system types, sizes and regions make up the 60 is not said.