Lucem Health

A learned model is the engine of each program, and the program around it carries its own value. The Reveal page says every program is powered by a validated clinical AI model that finds patterns in record data people cannot and identifies patients at higher risk of life threatening or chronic disease, deployed on a platform with analytics on program impact and provider behavior and a continuous improvement process. The models are partners' work (Medial EarlySign for three programs, an iRhythm relationship for arrhythmia), but a learned component still sits at the center of every program.

Without the models there is no list of patients to reach; with them, the company sells the deployment, outreach design and measurement around the list, which it calls operationalizing clinical AI.

A model produces a ranked list of patients who may warrant a test they have not had, and humans do everything after that: clinicians review, outreach staff contact the patient, and the diagnostic decision belongs to the ordering physician. Nothing is diagnosed and no order is placed automatically. The design is deliberately additive to existing population health processes rather than replacing them, which the company states as a selling point.

The oversight gap is structural rather than clinical and belongs to the arrangement rather than the software: a model surfacing patients and a separate organization acting on them means responsibility for a missed patient is split across at least three parties, and no published material assigns it.

The transparency here is inherited, and inherited transparency has a gap in it. Lucem points to partner algorithms validated in multiple health systems with findings published in leading journals, which is true and is a real strength of the underlying models, but that literature belongs to the developer and describes the model in general rather than the instance a given customer runs.

What is not published by anyone is the part Lucem controls: which model version is deployed at a site, whether it is recalibrated against local data, what thresholds are set and by whom, and how those choices change who appears on the list. A published performance figure for a model is not a published performance figure for a deployment of that model, and the difference lives exactly where this company sits.

The parties are identifiable and the flow between them is not, which is the characteristic shape of an orchestration business. Partner algorithm developers are named through the published literature the company points to, so whose models are running is a matter of record. What is absent is the map.

Patient data originates in a health system, moves through this platform, and is scored by a model belonging to a third company, in one case headquartered in another country, and nothing published describes which party holds identifiable data at which step, whether the model developer receives records or only returns scores, whether anything is retained for model improvement at any hop, or which of the three parties a patient would approach to exercise a right.

That last question is the one with no obvious answer: a patient asking who holds their data would be told by each party that another one does. Any assurance about the flow can only come from one of the three.

Real programs running at real health systems, with the evidence pointing at the wrong thing. Everything cited belongs to the model developers: peer reviewed validations of the underlying algorithms across multiple health systems, and third party analyst recognition of Medial EarlySign. Those establish that the models detect what they claim to detect. They say nothing about the layer this company sells.

Lucem's entire proposition is that clinical artificial intelligence fails on implementation rather than on accuracy, so the numbers that would prove its value are implementation numbers: what share of surfaced patients were contacted, what share completed the test, how that compares to the site's baseline, and how much of the effect survives after the launch period. None is published, and a claim of finding up to eight times the cancers in colorectal screening is stated without the denominator or the comparison it implies.

The path patient information takes through a program is undocumented. The Reveal page says programs run on record data a health system already holds without extensive data or systems integration; the privacy policy (last updated 17 April 2026) covers website visitors and the terms cover the website, and neither mentions patient records. No retention period, minimization statement, training use or deidentification step is published for the records the models score.

The gap is widest at the chain of custody. Patient data originates in a health system, moves through this company's platform, and is scored by a model belonging to another company, in one case headquartered outside the United States; nothing says which party holds identifiable data at which step, whether the model developer receives records or only returns scores, or whether anything is kept for model improvement.

No statement of status and no privacy document that reaches the programs. The privacy policy (last updated 17 April 2026) covers visitors to the website and mentions HIPAA nowhere, and the terms (last updated December 2021) govern use of the website; no HIPAA statement, business associate status or agreement is published. A website notice does not reach a product that scores patient records.

The subcontractor question raises the stakes: an agreement between a health system and Lucem does not by itself cover a model developer acting further down the chain, and HIPAA requires those obligations to flow down to every subcontractor that handles protected health information. Nothing published names which entities sit in the chain or shows that each is bound.

Nothing independent is published about security. No SOC 2, HITRUST or ISO 27001 attestation, trust center, penetration testing statement or vulnerability disclosure policy appears on the home page, Reveal page, privacy policy or terms; the privacy policy's security language covers the website and says no transmission or storage can be guaranteed secure.

The company's investors and early customers include large health systems, which normally require an attestation as a condition of contracting, so a report may exist privately; a buyer cannot establish it before a sales process. The supplier chain adds a second party, the model developer, whose security posture matters wherever patient data reaches it.

Disease detection is marketed with no regulatory position stated, on a basis a clinician cannot review. The Reveal page says each program "is powered by a validated clinical AI model able to discover patterns in data that humans cannot" and "identifies patients that appear to be at a higher risk for life threatening or chronic diseases, even in the absence of symptoms, screening results, or other traditional indicators"; the home page sells "AI-powered disease detection", and the colorectal program promises to "find up to 8x the cancers". What the ranking predicts is an undiagnosed disease, and what it asks is that a patient be tested: clinical on both counts. The company says clinicians keep practicing medicine as they always have, which assigns responsibility but states no position, and the patterns the model uses are ones the company says people cannot see, so the clinician receiving a flag cannot reconstruct why it fired.

No clearance, intended purpose statement or regulatory position appears on the home page, Reveal page, privacy policy or terms. The arrhythmia program involves a cleared cardiac monitor, and that clearance covers the monitor, not the model that decides who should wear one.

The company publishes nothing on how the models are governed or tested at a site. The home page describes practical, responsible AI and the Reveal page a continuous improvement process and analytics on program impact and provider behavior; no governance framework, validation before go live, subgroup performance, model card or drift monitoring appears in its published material.

The structure concentrates every governance question at the point where nothing is published. The company runs models it did not build, over data it does not own, to reach patients it never sees: who validates a model against a site's own population, who watches for drift, who notices if performance degrades for one group of patients, and which party answers for it are all unstated. The company running the deployment would see a problem first, the company that built the model understands it, and they are different companies. Early detection models that miss more cases in groups with thinner records would widen the gaps the programs exist to close.

Nothing is published on what happens when a program's model is wrong. Surfaced patients go to clinicians and outreach staff who decide whether to test, which lets a person decline a flag, but the characteristic failure of an early detection list is the patient who is never on it, and nothing published lets anyone see or challenge that: no audit of who was scored and excluded, no statement of the operating threshold or who sets it, and no route for a patient or clinician to question a score. The site's terms cover the website and disclaim all warranties; no customer terms are public.

The gap sits exactly where this company operates. A published performance figure for a partner's model is not a figure for a deployment of that model, and threshold setting alone can move a program from useful to unusable without touching the model, since the same algorithm at two operating points produces different alert volumes and different missed case rates. No warranty, indemnity or remediation commitment appears on the home page, Reveal page, privacy policy or terms.

A named record system is verifiable in its marketplace, and the programs are built to run on data a health system already holds. The NextGen marketplace carries a vendor listing for Lucem Health Reveal naming the NextGen Enterprise versions supported and describing standards based integration with CRMs, patient engagement and outreach platforms; the Reveal page says programs run on existing record data without new instrumentation, workflow changes or extensive integration, and sit alongside existing population health tools.

The listing names one record system and read access. What the programs write back into the record, which other record systems are supported, and which interface standards carry the data are not described.

Nothing is published about where the programs run or where patient data rests. The Reveal page says the programs work on existing record data without installing new software, which implies a hosted service, and the NextGen marketplace listing describes standards based integration; no hosting provider, region, tenancy model or residency commitment appears on the home page, Reveal page, privacy policy or terms.

The question is not routine here because of the supplier chain. The models behind several programs come from a company headquartered outside the United States, so whether scoring happens inside the health system's environment, inside Lucem's, or inside the developer's decides whether patient data crosses a border at all, and nothing published distinguishes the three.

How the programs are sold is discoverable; the price is not. Lucem sells Reveal early detection programs (arrhythmias, colorectal cancer, liver disease, lung cancer, type 1 diabetes) to health systems, framed as raising the clinical and financial yield of scarce care resources without adding staff, changing workflows or installing new software, and sells patient identification to life sciences companies whose products those patients might benefit from. Programs run on one platform and scale with patient volumes and priorities; the route in is a conversation.

The unit of charge (per program, per patient scored, per patient reached) is not stated, and neither is whether the model developer is paid separately. The last announced financing was a $7.7 million round led by Mayo Clinic, Granger Management and Mercy, and a program depends on two suppliers, the company and the model developer, so it can fail at either end.

Broad for a company of this size, and the breadth is a direct consequence of the business model. Named programs cover lung cancer, colorectal cancer, liver disease, arrhythmias, type 1 diabetes, progression from prediabetes to diabetes, and lower gastrointestinal disorders, which is seven disease areas from a small team.

Because each program is a partner model wrapped in a common operating layer, adding a condition is a partnership and a configuration rather than a research program, so coverage scales in a way that a company building its own models cannot match. The corresponding limit is that depth in any one area belongs to whoever built that model. Setting is United States health systems, spanning ambulatory and primary care wherever the record reaches.