Semantic Health

The models are the product. Semantic Health describes custom clinical AI and NLP, trained on millions of clinical records and tuned to each customer's patient population, that reads the chart, suggests diagnosis and procedure codes with the supporting documentation (Semantic Coder) and audits coded charts before billing to find unsupported or missing codes (Semantic Auditor). A coder confirming a suggestion is deciding, not doing the reading; without the models there is no product.

The oversight structure is described with the threshold missing. Nothing is submitted autonomously: Semantic Coder suggests codes before coding and Semantic Auditor surfaces findings before billing, each with references to the documentation, for coders and auditors to validate with minimal workflow change. What is not published is the confidence threshold for a suggestion, what the auditor does not check, or how a rejected suggestion feeds back.

The approach is named without version or accuracy discipline. Semantic Health describes custom clinical AI and NLP trained on millions of clinical records, tuned to each customer's patient population, using machine learning to follow clinical context and changing coding guidelines, and contrasts it with rules based computer assisted coding. No model, version, accuracy figure or update practice is published.

A structurally short chain. Semantic Health describes custom clinical AI of its own and offers on premises deployment, where charts need not leave the hospital. For the cloud option, the host and any subprocessor are not named and no subprocessor list is published, and its parent, AAPC, is a further party whose access is not described.

Named customers without outcome figures. Semantic Health names The Hospital for Sick Children and Humber River Hospital in Toronto, and Boston Children's Hospital and Cedars-Sinai in the United States. The site describes pre bill audit coverage of 100 percent of coded data, which is scope rather than outcome, and publishes no precision, acceptance, coder productivity or revenue figure.

General assurances that do not answer the questions AI raises. Semantic Health's page for US hospitals offers deployment on premises or in its SOC 2 and HIPAA compliant cloud environment, and says its models are trained on millions of clinical records and tuned to each customer's patient population. Nothing published states retention, de identification, or whether customer charts contribute to the shared models. Canadian hospital records processed by a company owned by a United States parent raise a residency question provincial health information law treats seriously. Safety engineering is not published.

Compliance is claimed without the document behind it. Semantic Health describes a HIPAA compliant cloud environment; no business associate statement, agreement or privacy document reaching the service is published. For Canadian customers the equivalent obligations run under provincial health information law, which the site does not address.

A recognized certification is named without the artifact. Semantic Health's page for US hospitals describes its cloud environment as SOC 2 and HIPAA compliant. The statement attaches to the environment, so it is unclear whether the report is Semantic Health's own or its host's, and neither its type nor its period is given. There is no trust center or penetration testing statement.

No device pathway applies and none is claimed: suggesting billing codes and auditing coded data are administrative determinations. The regulatory weight sits in program integrity: a finding that further specifies an inpatient diagnosis can change the payment grouping, which is closely examined in federal audits. Semantic Health's design answer is an evidence trail to the documentation on every suggestion and validation by the customer's coders before billing. Canadian inpatient coding runs on a different classification system under provincial oversight, and nothing published describes how the models are validated against each regime.

One design commitment and no evaluation. Semantic Health states every opportunity is fully explainable with references to the clinical documentation, so a reviewer can reject a finding on the evidence. It also prioritizes the cases with the highest impact, and its parent described the platform at acquisition as finding cases with the highest revenue impact; ranking by revenue directs auditors to charts where money is available rather than charts most likely to be wrong. No acceptance rate, breakdown by specialty or physician, bias testing or external audit is published.

Mechanisms exist and nothing stands behind them. Semantic Health's page for US hospitals states every opportunity is fully explainable with references to the clinical documentation and can be validated by the customer's team before billing, and Semantic Auditor checks whether assigned codes are supported. No precision, false positive or acceptance rate, warranty or indemnity is published; the page's line that compliance is 'ensured by AI audits' is a claim, not a commitment. The coder or auditor validates every suggestion, so the provider owns every submitted claim.

Named systems without documentation of the connection. Semantic Health's page for US hospitals names Epic, Cerner and Meditech and says it integrates with all major EHR and billing systems at any level of digitization. No integration method, standard, listing or statement of what writes back is published.

Options are stated and residency is not. Semantic Health's page for US hospitals says customers can deploy on premises or in its SOC 2 and HIPAA compliant cloud environment. The on premises option keeps charts in the hospital; for the cloud option no provider, region or tenancy is published, and nothing says whether Canadian hospital data is processed in Canada.

Every commercial term waits for a sales conversation. Semantic Health publishes no price, unit of charge, pricing basis, pilot terms or implementation timeline; the only commercial structure visible is the choice between on premises and cloud deployment.

Coverage is named with validation behind part of it: inpatient hospital coding and pre bill auditing in the United States, and health information management coding in Canada, supported by named hospitals in both countries (The Hospital for Sick Children, Humber River Hospital, Boston Children's Hospital, Cedars-Sinai). Coverage is not enumerated by specialty, and two of the four references are children's hospitals.