CAP Discusses Distributed Genomics, Digital Pathology

 

AI on CAP on DIGITAL PATHOLOGY and REMOTE BIOINFORMATICS

 

https://chatgpt.com/g/g-p-699fbe1fc2c8819189cf64e8233f3a55-danaher-dig-pathol/c/6a8f3daa-126c-83e8-ae47-b699f7b1c832

 

https://chatgpt.com/share/6ab35108-06e8-83e8-b0f2-4c09492ebb45

 

Yes. The CAP record is unusually clear—and it directly conflicts with CMS’s categorical premise that these computational-pathology analyses “do not require” a CLIA laboratory and may be performed by “any non-regulated entity.”

Strongest CAP authority

The most useful document is CAP’s February 23, 2026 response to the HHS Health Sector AI RFI. It is recent, formal, addressed to HHS, and explicit:

• CAP identifies itself as a CLIA-deemed accrediting organization that accredits more than 8,300 laboratories.
• CAP states that its accreditation checklists require laboratory-director validation or verification before patient testing, “including AI and machine learning algorithms.”
• CAP describes FDA regulation and CLIA oversight as “complementary and necessary” for clinical AI.
• CAP says laboratories must validate or verify local AI performance under CLIA before implementation.
• Later in the letter, CAP states directly that CLIA requires pathologists, acting as laboratory directors, to assess AI systems before local implementation.
• CAP recommends using its accreditation and proficiency-testing infrastructure to evaluate, monitor, and safely deploy AI tools.

This is not an inference from general laboratory principles. CAP expressly places clinical AI under CLIA laboratory-director oversight and CAP accreditation requirements.

CAP’s actual inspection checklist

CAP’s 2025 accreditation checklist edition is described in detail in the February 2026 CAP TODAY article, “Digital path practices reflected in latest checklist changes”.

The Laboratory General Checklist now contains a section expressly titled:

Digital Pathology Including Remote Data Assessment

CAP explains that it revised this section specifically to encompass developments in digital algorithms and artificial intelligence. Its definition expressly includes pathologists’ review and diagnostic interpretation of tissue using digital pathology and whole-slide images. It also ties the activity to the laboratory’s CLIA address and explains when a remote interpreting location becomes a separate referral laboratory.

Page two of the CAP article identifies particularly useful checklist requirements:

• Inspectors sample digital-pathology policies and procedures.
• Inspectors sample reports generated through digital pathology.
• Inspectors review digital-pathology validation and verification records.
• GEN.50630 requires the laboratory to validate or verify digital-pathology systems used for clinical diagnostic purposes, with approval by the laboratory director or a qualified designee.
• GEN.52860 requires digital-pathology services to be included in the laboratory’s quality-management system.
• CAP inspectors may examine scan-failure rates, image-quality criteria, risk-mitigation procedures, discordance reconciliation, and other ongoing performance measures.

That is probably the best operational evidence for your letter. CAP is not merely saying that a pathologist should use good judgment when purchasing software. CAP has placed clinical digital pathology inside the inspection checklist used to determine continuing laboratory accreditation.

CAP’s specific AI-validation guidance

CAP’s educational guidance, “How to Validate AI Algorithms in Anatomic Pathology”, is also remarkably direct:

• Both CLIA and CAP require validation of a new test, device, or diagnostic aid before reporting patient results.
• CAP applies its whole-slide-imaging validation principles to image-analysis algorithms.
• CAP states that any image-analysis or image-recognition system—FDA-authorized or not—must be validated before clinical use.
• Validation occurs in the laboratory, using cases representative of the laboratory’s intended clinical use and patient population.
• The medical director establishes acceptance criteria and approves implementation.
• CAP even supplies suggested report language for a non-FDA-authorized AI system treated as a laboratory-developed test.

Although this is educational guidance rather than the checklist itself, it demonstrates how CAP interprets and operationalizes CLIA and CAP requirements for H&E-based AI.

Second formal CAP policy letter

CAP made essentially the same point in its December 1, 2025 comments to FDA on measuring and evaluating AI performance.

CAP states that:

• CAP is a federally deemed CLIA laboratory accrediting organization.
• CAP checklist validation and verification requirements expressly include AI and machine-learning algorithms.
• AI represents a new component of pathologists’ responsibilities as CLIA laboratory directors and section directors.
• CLIA and CAP impose quality requirements beyond the manufacturer’s FDA-authorized operating instructions.
• Laboratories should establish controls, performance metrics, corrective actions, and procedures addressing AI performance drift.

Again, CAP treats AI applied in clinical pathology as part of the laboratory test system—not as an unregulated computer transaction detached from the laboratory.

CMS’s own CLIA guidance

There is also an apparent inconsistency within CMS itself. In its May 11, 2023 CLIA Post-PHE Guidance, CMS treated the remote review of digital laboratory data, results, and images as testing performed under a primary laboratory’s CLIA certificate:

• The remote work must fall within the specialties or subspecialties on the primary laboratory’s certificate.
• The laboratory director remains responsible.
• Records are retained by the primary laboratory.
• Survey findings and enforcement actions attach to that laboratory’s CLIA certificate.
• Digital proficiency-testing review is treated as occurring under the primary site’s CLIA number.

This does not decide every possible algorithm-only service, but it refutes any general theory that converting tissue information into digital images automatically removes subsequent clinical interpretation from the CLIA framework.

The precise conflict with the OPPS proposal

CMS states in the CY 2027 OPPS proposed rule that stand-alone algorithmic analyses are separate from a CLIA laboratory’s examination of human material. CMS then concludes that they:

• do not require CLIA-regulated laboratory services or entities;
• can be performed by any “non-regulated entity” possessing the software;
• are “other diagnostic tests,” rather than clinical diagnostic laboratory tests; and
• should therefore be removed from the CLFS and paid as SaMS under OPPS.

For a purely secondary calculation performed on a finalized laboratory result, CMS’s distinction may sometimes be defensible. It is much less persuasive for computational pathology applied to an H&E whole-slide image. The WSI is a direct digital representation of the patient’s tissue specimen, and the algorithm produces a patient-specific diagnostic, prognostic, or predictive result. CAP expressly subjects that clinical process to laboratory validation, laboratory-director oversight, quality management, reporting requirements, and inspection.

Payment classification and regulatory classification need not always be identical. But CMS’s proposal does not merely select a different payment system. It makes the broader factual and regulatory assertion that no CLIA entity is needed. That is where the collision becomes unavoidable.

Suggested comment-letter language

CMS’s proposed treatment of computational pathology as Software as a Medical Service creates a direct conflict with the existing federal laboratory-quality framework. CMS states that these algorithmic analyses do not require a CLIA-certified laboratory and may be performed by an otherwise unregulated entity. In contrast, the College of American Pathologists—the nation’s largest federally deemed CLIA accrediting organization—expressly applies its laboratory accreditation requirements to artificial intelligence and machine-learning algorithms used for patient testing. CAP requires laboratory-director validation or verification before implementation, and its 2025 Laboratory General Checklist includes a dedicated section entitled “Digital Pathology Including Remote Data Assessment.” CAP inspectors review digital-pathology policies, patient reports, validation records, and quality-management activities; checklist requirements GEN.50630 and GEN.52860 specifically address laboratory validation, laboratory-director approval, and continuing quality management of digital-pathology systems.

This conflict is especially pronounced for computational-pathology tests that apply an algorithm to an H&E whole-slide image to produce a patient-specific diagnostic, prognostic, or predictive result. Such a service is not simply an arithmetic calculation performed on a previously reported laboratory value. It is the clinical examination and interpretation of a digital representation of human tissue. CAP’s current accreditation framework treats that activity as patient testing conducted under the responsibility of the CLIA laboratory director.

CMS should therefore withdraw or qualify its categorical conclusion that SaMS analyses performed on laboratory-test data require no CLIA-regulated entity. At minimum, CMS should clarify that classification and payment of a service under OPPS does not displace otherwise applicable CLIA requirements; distinguish purely secondary analysis of finalized data from computational examination of whole-slide tissue images; and consult with CMS’s CLIA program and federally deemed accrediting organizations before finalizing the regulatory status of computational-pathology services.

One terminology refinement: CAP is not universally “CLIA’s inspection body.” The technically exact phrase is “a CMS-approved, federally deemed CLIA accrediting organization.” CAP states that, for CAP-accredited laboratories, a CAP inspection occurs in lieu of a CMS inspection. That wording will be harder for CMS to evade.