Aidin

Identifiable machine learning doing narrow work, inside a platform that is mostly directory, workflow and document handling.

Two components plausibly involve models. Document intake classifies material arriving by fax, email and a dedicated print path and assigns it to the correct patient and request, which is the same class of problem another vendor in this index is graded on and is genuine work, since post acute referral traffic still moves on paper and the classification is what removes manual sorting. Provider quality analytics compute comparative measures across a national directory, which is analysis even if it is not necessarily learned.

What surrounds it is not inference. A national care partner directory, referral routing, authorization tracking, document signature handling, task deadline flags and secure messaging are all software rather than intelligence, and a hospital adopting Aidin is buying a case management workflow first.

The company describes an artificial intelligence powered platform and no model, method or performance figure supports the description anywhere in public material. This index has already recorded a rejection where intelligent routing turned out to be configured rules, and the distinction matters: without any published detail, a buyer cannot tell whether document classification is learned or pattern matched, or whether at risk flags are model derived or threshold based.

Graded C because the document and analytics work is real and the platform around it plainly is not, and because the positioning outruns anything published. Ask which components are model driven and what they achieve over rules.

Human decision making is preserved by design, and one of the design choices puts the decision unusually far from the vendor.

The platform recommends nothing autonomously. Case managers manage referrals, assign tasks and work deadlines; the system surfaces flags for incomplete flows, at risk cases and overdue assignments; a person acts. Authorization tracking monitors requests to prevent denials rather than submitting or deciding them.

The patient choice model is the notable feature on this axis and it goes further than clinician oversight. Rather than an algorithm selecting a post acute provider, the platform presents quality, satisfaction and outcome data to the patient and the patient chooses. That places the consequential decision with the person affected by it, which is a stronger position than clinician review of a machine recommendation, and it aligns with the federal requirement that discharge planning offer genuine choice supported by quality data.

What is unexamined is the framing effect. Whoever controls which providers appear, in what order, and which quality measures are displayed exercises substantial influence over a choice that is nominally free, and nothing published describes how the list is ordered, how many options are shown, whether the hospital's preferred provider list biases presentation, or how quality measures are weighted. Presentation is where an open market either delivers choice or simulates it.

No escalation behaviour is described for a referral that goes unanswered or a deadline that passes.

Ask how the provider list is ordered and weighted, and what happens when no provider accepts.

The platform is described in detail and the technology inside it is not described at all.

Functional disclosure is actually good. Document intake paths are enumerated as fax, email and a dedicated print route, task flag categories are named, the referral, authorization and document workflows are laid out, and the tier structure specifies which capabilities appear at which level. A buyer can picture the product.

Beneath that there is nothing. No model is named, no method is described, no architecture is given, and no accuracy, precision or error figure exists for any component. Document classification, which is the clearest candidate for machine learning in the product, carries no accuracy rate, which matters because a misfiled referral document is a clinical record attached to the wrong patient. At risk flagging is not explained, so whether it is a learned prediction or a deadline rule is unknown, and those are entirely different products.

The quality analytics layer is the least specified and the most consequential. Which measures are drawn on, from which sources, over what period, and how they are combined into whatever a patient sees, is unpublished, and this is the component that shapes a patient's choice of where to receive care.

Graded D because the artificial intelligence claim in the company's positioning has no technical content behind it anywhere in public material.

Ask for document classification accuracy, whether at risk flags are learned or rule based, and the full methodology behind the provider quality presentation.

A dedicated pass located no cloud or infrastructure provider, no model or component vendor, no document processing or optical recognition supplier, no sub processor register and no statement on whether platform data contributes to model development.

One component is close to certain and entirely unnamed. Ingesting documents from fax and scanned print and attaching them to the correct patient requires optical character recognition and classification, and that capability is very rarely built in house at this scale, which means a third party service almost certainly processes clinical referral documents. Whether that is the case, and if so who, is undisclosed, so a hospital cannot determine which parties handle its patients' records.

The quality data has a supply chain of its own that is equally undisclosed. Provider quality, satisfaction and outcome measures originate somewhere, most plausibly in federal quality reporting programmes and survey instruments, and nothing states which sources are used, how they are licensed, how current they are, or whether anything is derived from platform activity. For a component that shapes patient decisions, provenance is not a technical footnote.

The training question is unaddressed and specific here. A two sided network accumulates a detailed record of which providers accept which patients and how they perform afterwards, which is exactly the corpus a placement recommendation model would need, and nothing states whether it is used that way.

Ask who performs document recognition, the sources and refresh cadence behind the quality measures, and whether referral outcomes train anything.

Named institutional customers with stated objectives, and no published figure against any of them.

The customer list is real and checkable, which is more than several vendors in this index offer. UCLA Health, five hospitals of a large academic medical centre, St. Luke's, Elmhurst Edward Hospital and Western Maryland Health System appear in company case material, spanning academic, community and regional systems. Each case describes a concrete operational problem: emergency department diversion driven by capacity constraint, average length of stay, referral management across inpatient and outpatient settings, and readiness for value based reimbursement.

Stating the problem a customer brought is useful and it is not an outcome. No quantified result was located for any named site: no length of stay reduction in days, no readmission rate change, no capacity figure, no referral turnaround time. The measures the platform targets are among the most cleanly measurable in hospital operations, since length of stay and readmission are tracked continuously by every customer for other reasons, which makes the absence of published numbers harder to explain than it would be elsewhere.

No peer reviewed publication, independent evaluation or third party analyst assessment was located.

One industry level caution belongs here. Published claims about artificial intelligence in post acute coordination circulate widely without attribution to any specific vendor, and figures encountered in that literature should not be attached to this record.

Ask for length of stay and readmission change at a named site, with the measurement period and comparator.

A dedicated pass located no encryption statement, no retention schedule, no access control model, no audit logging description, no deletion process and no position on whether platform data contributes to model development.

The data profile deserves the attention. A post acute referral packet is among the more complete clinical documents that moves between organisations, typically carrying diagnosis, functional status, medication list, insurance information and clinical notes, and this platform assembles and transmits those at volume. The document intake layer additionally ingests unstructured material from fax and email, which means content the sender did not curate arrives in the system.

The broadcast pattern raises the specific stewardship question. An open market referral sends patient information to multiple providers so they can bid to accept, which multiplies the copies of a clinical record in proportion to how many providers are approached. Nothing published describes how many receive a given referral, what each sees, what is redacted before acceptance, or what happens to the record held by every provider that did not take the patient.

That last point is the one worth asking about first, because a declined referral leaves patient data resident at an organisation with no ongoing treatment relationship and no obvious retention obligation.

One pre emptive note: a general privacy policy cannot move this grade. Only a published retention schedule, a statement of what recipients see and keep, and a deletion process for declined referrals will.

Ask what a referral packet contains, how many providers receive it, and what is deleted after decline.

Compliance is asserted at the product level with specificity about where it applies, and no contractual position is published.

What is stated is more particular than a badge. The platform is described as a health privacy compliant communication hub that consolidates fax, email and digital channels, and the secure messaging layer is described as compliant for sharing patient details across a care community. Naming the channels matters, because consolidating fax and email traffic into a governed system is a privacy improvement over the status quo it replaces, where patient records move by unencrypted fax between organisations with no audit trail.

The multi party structure is what makes the missing contractual detail consequential. Patient information crosses from a hospital to multiple competing post acute providers during a single referral, and often to providers that decline the patient and retain whatever they were sent. Whether receiving providers on the free tier execute agreements, what they may retain after declining, and how the platform limits onward disclosure are all unaddressed, and this is the central privacy question of a referral marketplace rather than an incidental one.

No business associate agreement template, execution requirement, negotiation stance or subcontractor position was located for either side.

Graded C because the compliance claim is specific to a real workflow improvement, and rather than higher because the distinctive multi party risk is unaddressed.

Ask what a declining provider retains, whether free tier participants sign agreements, and what limits onward disclosure.

A dedicated pass located no security page, no external attestation, no trust centre, no penetration testing statement, no vulnerability disclosure policy and no documentation offered under agreement. The only security relevant statement anywhere is that communication is health privacy compliant.

The exposure is broader than a single organisation's data. This platform sits between hospitals and a national network of post acute providers, holding referral packets that combine clinical, functional and insurance information for patients at their most vulnerable transition. A compromise would reach across many organisations at once rather than one, which is the characteristic risk of a network positioned product and the reason buyers ordinarily require an attestation before joining one.

The participant mix sharpens it further. Large academic health systems run mature security programmes; small skilled nursing facilities and home health agencies on a free tier frequently do not, and they are credentialed users of the same platform. How access is scoped for the least resourced participants, and what a compromised free tier account can reach, is unaddressed.

A component installed on hospital networks to capture printed documents adds a further surface with no published security description.

One pre emptive note: a restated privacy compliance claim cannot move this grade. Only an external attestation, or security documentation available under agreement, will.

Ask whether any external assessment exists, how free tier provider accounts are scoped and reviewed, and how the print capture component is secured.

The regulation that actually governs this product is named and built against, which is rare enough in this index to carry the grade.

Device regulation is not the relevant regime and the record should say so rather than treat its absence as a gap. Referral routing, document management and authorization tracking make no clinical claim, diagnose nothing and recommend no treatment, so no clearance is required and none is claimed. That is a clean position rather than an undeclared one.

What governs discharge planning is federal post acute care law, and the company addresses it directly. The platform is described as built to keep customers compliant with the federal standards for post acute patient discharge lists, which oblige hospitals to supply patients with quality data on available providers and to support a genuine choice among them. The open market design is a direct implementation of that obligation rather than a feature positioned beside it, and the quality, satisfaction and outcome data presented to patients is precisely what the requirement contemplates.

That alignment is the strongest regulatory posture recorded in this session, because most vendors here are silent about the regime that actually applies to them while making claims about one that does not.

What would lift it further is evidence rather than alignment. No attestation, audit or third party assessment of that compliance was located, and nothing describes how the quality data presented is sourced, refreshed or validated, which is what a regulator examining the requirement would test.

Ask for the source and refresh cadence of the quality measures, and any external assessment of discharge list compliance.

A dedicated pass located no governance framework, no fairness testing, no subgroup analysis, no methodology and no external audit.

The bias risk here is not the usual one and is arguably more consequential, because it operates on institutions rather than on patients directly and then reaches patients through them. The platform ranks and presents post acute providers using quality, satisfaction and outcome data, and provider quality measures are well known to correlate with the population a facility serves. Facilities caring for poorer, sicker and more socially complex patients score worse on readmission and outcome measures for reasons that are substantially attributable to their case mix rather than their care. A ranking that does not adjust for that will systematically steer patients away from safety net post acute providers and toward those serving easier populations, which concentrates disadvantage precisely where it already sits.

Nothing published describes which measures are used, whether any risk adjustment is applied, how measures are weighted, or how recency is handled.

A second and quieter mechanism sits in the same place. Provider participation is tiered, with a free level and paid levels above it, and nothing states whether tier affects visibility, ordering or presentation in the patient facing list. If it does, commercial standing influences a clinical choice presented as objective.

Document classification carries a more conventional exposure, since referral packets vary in format by originating institution.

Ask which quality measures are used and how they are risk adjusted, and whether paid tier affects visibility to patients.

Nothing allocates responsibility, and the two sided structure makes the question harder rather than simpler.

A dedicated pass located no service level agreement, no accuracy warranty, no uptime commitment, no indemnity and no remediation position for either the hospital side or the provider side.

Availability is the omission that matters most for this product. Discharge planning runs to a daily clock, beds are held while placements are arranged, and a platform outage during business hours stalls transitions across every customer simultaneously, extending length of stay for the exact reason the product was bought to reduce. No uptime commitment or historical availability figure was located.

The document layer carries a distinct and underappreciated risk. Classification attaches incoming clinical documents to patients and requests, and a misattribution places one patient's clinical record in another patient's referral packet, which is both a clinical safety event and a privacy breach in a single action. No accuracy figure exists and no correction or notification process is described.

The quality data creates a third exposure that no comparable vendor in this index carries. The platform presents provider quality information on which patients base a care decision, and nothing states what is warranted about that information's accuracy or currency, or what recourse exists for a provider misrepresented by it.

One pre emptive note: further customer names cannot move this grade. Only contractual terms, or published availability and classification accuracy, will.

Ask for the availability commitment, what is warranted about the quality data, and the correction process for a misfiled document.

Two integration stories exist, one named as a paid feature and the other genuinely unusual.

Record system integration is stated but not specified. It appears as a capability unlocked at paid tiers, which tells a buyer it exists and is chargeable, and no electronic health record is named, no interface standard is described and no marketplace or validated listing was located. For a discharge planning product that must live in the case manager's daily system, naming zero record systems is a real gap.

The second story is the interesting one and deserves credit rather than dismissal. The platform ingests from fax, email and a dedicated print path, which is interoperability with the actual state of post acute care rather than with an idealised one. Skilled nursing facilities, home health agencies and rehabilitation providers frequently have no interoperable record system at all, and a platform that meets them on fax is solving the harder half of the problem. Standards based integration is worth more when both sides have systems; here one side often does not.

What is missing on that side is any statement of what the receiving provider gets, whether the free tier includes any integration, and whether acceptance flows back into the hospital's record as a discrete event.

Graded C because unnamed record system integration and a genuine unstructured intake capability roughly offset. Ask which record systems are integrated in production, through what standard, and whether referral outcomes post to the chart.

Cloud delivery is stated and nothing else about the deployment is described.

What is published is that the platform is cloud based and accessed by hospitals and post acute providers through web interfaces, with a dedicated print path installed at customer sites to capture documents. That establishes a hosted multi tenant shape by inference rather than by statement.

Everything specific is absent. No cloud or hosting provider is named, no region is stated, no residency commitment is made, no tenancy or segregation model is described, and no backup, continuity or recovery position was located.

The multi tenant question carries particular weight for this product and is not a generic omission. Competing hospitals and competing post acute providers are all tenants of the same platform, and the referral data those tenants generate is commercially sensitive to one another: which facilities accept which patients, at what speed, and on what terms is exactly the information a competitor would value. How tenant data is segregated, and whether aggregate directory analytics are derived from it, is unaddressed in either direction.

The print path deserves a question of its own, since software installed inside a hospital to capture printed documents sits on the customer's network and nothing describes what it transmits or how it is secured.

Ask where the platform is hosted, how tenant data is segregated between competing organisations, and what the print capture component transmits.

The commercial model is disclosed in structure even though no price is published, which is a meaningful distinction on this axis.

What is stated is the shape of a two sided market. Post acute providers receive referrals under a free access plan, and paid tiers are named as Verified, Community and Enterprise, with what each unlocks described: record system integration, custom reporting and team workflow access sit above the free tier. Hospitals on the other side buy the case management and referral platform. Naming the tiers and their contents lets a post acute provider work out which level they need before contacting sales, and it reveals that integration is a premium feature rather than a baseline one, which is exactly the detail that surprises buyers late.

The two sided design is itself commercially informative. Free access on the receiving side builds directory density, and directory density is what makes the open market proposition work for hospitals, so a buyer can reason about why the pricing is arranged this way.

No figure appears anywhere for any tier or for the hospital side, no unit of charge is stated for the hospital platform, and no implementation fee, minimum commitment or contract term was located. Nothing indicates whether hospitals pay per case manager, per bed, per referral or per facility.

No return proxy accompanies it. Length of stay and capacity are the argument and no cost per referral or avoided bed day figure is supplied.

Ask for the hospital side unit of charge, tier pricing, and whether referral volume affects cost.

Both sides of the transition are covered, and depth within either is not evidenced.

The span is coherent. On the acute side, hospitals and health systems use it for discharge planning and case management across inpatient and outpatient referral flows, with named customers ranging from a large academic system to a regional community one, which suggests the product works at more than one scale. On the receiving side, post acute providers across the national directory participate, and the platform is described as managing referrals across the care spectrum rather than a single destination type.

What is missing is the composition. Post acute care is not one setting: skilled nursing, inpatient rehabilitation, long term acute care, home health and hospice differ in eligibility rules, authorization requirements and clinical criteria, and nothing states which the platform handles well or how directory density varies across them. Home health and skilled nursing referrals are different transactions and a hospital serving a rural catchment cares which one has depth in its region.

Geographic coverage is likewise unstated beyond a national directory claim. Provider density by market is the single factor determining whether the open market model produces real choice or a short list, and no figure describes it anywhere.

Nothing addresses specialty populations such as paediatric post acute care or behavioural health placement, both of which have distinct and scarce provider networks.

Ask for directory density by provider type and by region, and which post acute settings the authorization workflow supports.