Dimer Health

AiME does substantive clinical work: real time patient engagement, symptom interpretation against the individual's history, medications, and comorbidities, and continuous risk signal detection. Held back from A because what a health system buys is a staffed transitional care service in which a licensed Transitionist owns the episode; the AI is the monitoring and triage layer in front of a physician led practice rather than the deliverable itself.

The escalation target is named and accountable, which is what distinguishes a real oversight model from an assurance. AiME monitors and detects, then escalates to a dedicated Transitionist, a licensed clinician who owns the episode of care from discharge through recovery. Assigning ownership of the episode to a named human role, rather than routing alerts into a queue, is the correct design for the post discharge window where deterioration is the failure mode.

The system is named and its clinical grounding is asserted. The technology is not described.

What is public: a named clinical system, described as developed by clinicians and grounded in physician validated protocols, which engages patients conversationally, interprets symptoms against the individual's history, medications and comorbidities, and detects emerging risk. That is a clear functional account and the protocol grounding claim is meaningful, because it implies a defined clinical logic rather than an unconstrained generative system.

What is absent: whether the underlying model is the company's own or licensed, what the protocols are and who validated them, how symptom interpretation is performed, how risk detection works, what evaluation exists, how versions are managed, and how a protocol change is tested before it reaches patients.

The protocol claim is the one most worth pressing, precisely because it is the strongest thing said. Physician validated is a process claim rather than a performance claim: it says clinicians approved the logic, not that the logic was measured against outcomes. For a system operating in the highest risk window after discharge, a buyer needs to know how often the protocols produce the right escalation, not only that clinicians endorsed them at design time.

Ask which models underlie the conversational layer, whether any third party service processes patient conversations, what the protocols cover and do not, and what validation exists beyond clinician endorsement.

Nothing identifies any party in the chain: no model or model family, no foundation model provider for the conversational layer, no hosting arrangement and no sub processor list was located in two passes, and no retention or training position was found. The content this service generates is unusual in kind and deserves describing precisely, because it is not the sort of thing a retention policy written for clinical records would cover.

A patient recovering at home describes symptoms, worries, medication confusion and daily circumstances to a conversational system at any hour, often through a caregiver. That produces a record of illness experience that exists nowhere in the hospital's own documentation, is generated outside a clinical encounter, and is far more revealing of a person's circumstances than a discharge summary.

It is also created during a period when the patient is least able to evaluate what they are agreeing to. Two onward flows follow. Information written back to the medical record becomes permanent and visible to every future clinician, which is appropriate for clinical findings and less obviously so for everything a patient says at two in the morning.

And the referring health system and any payor are commissioning parties with an interest in utilisation, while a patient speaking to what feels like a supportive companion may not understand that. Ask what is retained and for how long after the episode closes, what trains models, what reaches the hospital and payor, and what the patient is told.

The company is credited with large reductions in avoidable readmissions, which is the right endpoint for transitional care and a measurable one. However no figure, cohort, study design, or named health system was retrieved, so the claim cannot be sized or verified. Readmission reduction is also heavily confounded by patient selection, making an unstated methodology a material gap. Founded 2023 with a Series A in March 2026.

No retention position, training use statement or de identification posture was located.

The data this service generates is unusual in kind. A patient recovering at home describes symptoms, worries, medication confusion and daily circumstances to a conversational system at any hour, often through a caregiver. That produces a record of illness experience that exists nowhere in the hospital's own documentation, is generated outside a clinical encounter, and is far more revealing of a person's circumstances than a discharge summary. It is also created during a period when the patient is least able to evaluate what they are agreeing to.

Three questions follow and none is addressed. What is retained from those conversations, and for how long after the episode of care closes. Whether conversation content develops the models, which matters because a system described as improving through clinician validated protocols has an obvious appetite for real interaction data. And what is shared with the referring health system or payor, since both are commissioning parties with an interest in utilisation, and a patient speaking to what feels like a supportive companion may not understand that.

The write back to the medical record is a fourth. Information entering the chart becomes permanent and visible to every future clinician, which is appropriate for clinical findings and less obviously so for everything a patient says.

Ask what is retained, what trains models, what reaches the hospital and payor, and what the patient is told.

No business associate statement, availability, scope or contracting entity was located from the company's own materials, and one claim in circulation needs correcting rather than crediting.

A development contractor's published case study states that it helped the company through a compliance process and obtain a seal of good faith for the health privacy rule. That should not be read as an attestation. There is no certification for the health privacy rule. The federal department responsible for it does not endorse, accredit or recognise any certification, and seals of this kind are commercial products sold by compliance services firms rather than findings by an independent auditor against a defined standard. A buyer seeing such a seal has learned that the company engaged a consultancy, which is worth something operationally and nothing evidentially.

The substantive position needs establishing separately and is more complicated than for a software vendor. The company delivers care through licensed clinicians, bills Medicare, Medicaid and commercial insurance, and operates across several states, which points toward it being a provider and therefore a covered entity in its own right rather than only a business associate of the discharging hospital. Both relationships probably exist at once: covered entity for the care it delivers, business associate for data it receives from a partner health system.

Ask which entity is the treating provider, which signs with a hospital partner, and whether a notice of privacy practices is published.

No SOC 2, HITRUST, ISO 27001 or equivalent attestation was located, and no trust centre or security page was found.

One claim in circulation should not be mistaken for one. A development contractor's case study describes helping the company obtain a seal of good faith relating to the health privacy rule and implementing compliant infrastructure. A seal of that kind is a commercial product from a compliance services firm, not an independent examination of controls against a defined standard by an accredited assessor. It is not an attestation and should not be counted as one.

Context is worth stating in fairness. This is a company founded in 2023 with around twenty million dollars raised in total, and a full attestation programme is a material cost at that stage. This index has recorded elsewhere that certification frameworks can exclude early entrants who may nonetheless be well built. The grade reflects what a counterparty can verify, not a judgement that controls are absent.

What raises the stakes is the customer set rather than the company's size. The service operates across eight states with hospital partners and a regional payor, holds conversational health information, and writes into hospital record systems. Write access into a partner's medical record system is a security question for that partner as much as for the vendor.

Ask what independent examination exists or is planned, how write access to partner systems is authenticated and logged, and what the incident notification commitment is.

No regulatory pathway is claimed, none was located, and no rationale is published. On this record the boundary is closer than for most care management vendors and the reason is specific.

The company describes its system as engaging patients in real time and interpreting their symptoms in the context of that individual's medical history, medications and comorbidities, detecting emerging risk and escalating when intervention is needed. Read plainly, that is patient specific assessment rather than general health information, and the assessment is delivered to the patient rather than to a clinician.

Both halves matter. The exemption that keeps most decision support outside device regulation depends on a health care professional being able to review the basis independently rather than relying on the output; a person recovering at home is not a health care professional and has no way to review anything. The alternative ground, that software provides general information rather than individualised assessment, is unavailable when the product's stated distinction is that it reasons against the patient's own history and medication list.

The setting raises the stakes. This operates in the fortnight after discharge, when a large share of readmissions occur, and the characteristic failure is reassurance: a patient told their symptom is expected who does not then seek care.

The company's mitigations are real. Clinician review of care plans before release is stated, escalation to a named clinician is the design, and the practice is physician led. Ask for the published regulatory rationale, the intended use statement, and what the system will not answer.

No governance framework, evaluation methodology, performance data or applicability statement was located.

The domain question here is escalation, and it is sharper than in most monitoring products because of when it operates. In the fortnight after discharge the cost of a missed escalation is a readmission or worse, and the cost of an unnecessary one is a clinician's time. Those are not symmetric, and a system tuned to reduce low acuity burden on providers, which the company names as a benefit, is being tuned in the direction where the expensive error lives. Nothing published states the escalation threshold, the false negative rate, or how the balance was chosen.

A second question concerns who the system serves well. It engages patients by conversation, in the days after a hospital stay, when people are unwell, medicated, tired and often frightened. Comprehension, language, health literacy and cognitive state after an acute illness all vary enormously, and a patient who cannot describe symptoms precisely to a conversational system is exactly the patient most likely to deteriorate unnoticed. Caregivers are named as users too, which helps for some patients and not for those without one.

The company states that every recommendation is reviewed by its clinical team, which is a real control and is credited on the autonomy axis. It does not substitute for knowing how the system performs across the people it talks to.

Ask for escalation sensitivity, outcomes by patient language and living situation, and what happens when a patient stops responding.

The strongest claim on this record is a process claim standing in for a performance claim, and separating the two is the whole assessment. The system is described as grounded in physician validated protocols, which is meaningful and implies defined clinical logic rather than an unconstrained generative system. Physician validated says clinicians approved the logic at design time.

It does not say the logic was measured against outcomes, and those are different things: a protocol endorsed by experts can still escalate too rarely in practice, because endorsement tests whether the rules look right and deployment tests whether they fire correctly against real presentations. For a system operating in the highest risk window there is, the days after discharge, a buyer needs the escalation accuracy rather than the design time endorsement, and nothing publishes it.

No accuracy figure, no evaluation methodology, no false reassurance rate and no warranty, indemnity or remediation commitment was located. The setting sharpens the consequence. A patient at home describes symptoms to a conversational system at any hour, often through a caregiver, and a missed escalation there means someone stays home who should have been seen, with no clinician in the loop to notice.

Ask what validation exists beyond clinician endorsement, how often the protocols produce the correct escalation, what the system does when a presentation falls outside them, and what a patient is told about its limits.

Described specifically and in both directions, which is more than most vendors of this size manage.

A published account of the platform's construction sets out concrete behaviour rather than a general integration claim. Patient entry with telephone verification checks whether a record already exists in the electronic medical record system, and on submission patient health information is transferred into that system, with the application automatically creating or updating the patient profile. Integration with call centre software is described alongside it.

That is a read and a write, and the write is the harder half. A transitional care service that creates or updates a patient record in the discharging institution's system is putting information where the patient's own clinicians will see it, which is what makes the service continuous with hospital care rather than parallel to it. For a post discharge product that is the whole proposition: the value is lost if what happens at home stays in the vendor's system.

Held at B rather than A because no interoperability standard, named record platform or interface documentation was located, the description comes from a development contractor's case study rather than the company's own materials, and nothing states what is written back beyond profile level information. Whether clinical findings, escalations and outcomes reach the record as structured data is the question that determines usefulness.

Ask which systems are integrated in production, what standards are used, and what specifically is written back.

No hosting location, region, tenancy model, retention schedule or subprocessor list was located.

The architecture is partly inferable and more distributed than a single platform. There is a patient facing application, a conversational layer that may or may not depend on a third party model service, integration writing into partner electronic medical record systems, call centre software in the path, and a clinical team operating across eight states. Each is a distinct component with its own data handling, and none is described.

Two questions are specific rather than generic. If the conversational system depends on an externally hosted model, patient described symptoms are leaving the company's environment, and whether that provider retains prompts, for how long, and under what terms is the first thing to establish for any product of this shape. And the multi partner structure means encounter data from several hospital systems and a payor accumulates in one place, so separation between partners is a question those partners should ask on their patients' behalf.

A third follows from the corporate shape. Care is delivered by licensed clinicians and billed to insurance, so a clinical record exists and someone holds it. Where it sits, and what happens to it if a hospital partnership ends, should be settled before enrolment rather than after.

Ask where the platform runs, whether any third party model service processes patient conversations and on what terms, how partners are separated, and who holds the clinical record.

No public pricing. Delivered as a clinical service line billed through health system and payor partnerships and aligned with value based care arrangements. Buyers should establish whether the economics are per episode, per member, or shared savings against readmission performance, since a service with clinical staffing costs behaves very differently from software under each.

Tightly and deliberately scoped to the transition from hospital discharge through recovery, a defined and clinically high risk window rather than general chronic care. Narrow scope precisely stated is what this axis rewards.