doi: 10.5858/arpa.2026-0016-CP.
_____
CAP Error Reduction: What Pathologists, Labs, and Genomics Stakeholders Need to Know
TLDR
CAP has released an important 2026 update on Interpretive Diagnostic Error Reduction in anatomic pathology. The headline is simple: pathology groups should have structured, documented, timely case-review processes to detect disagreements, reduce interpretive errors, and improve patient care. But the article is more interesting than a generic “peer review is good” statement. CAP is effectively saying that diagnostic accuracy in pathology is no longer just an individual professional virtue; it is a system property that must be designed, monitored, and improved.
The guideline update is grounded in a systematic review of the literature since the 2016 CAP guideline. It focuses mainly on surgical pathology and cytology, especially secondary review, targeted review, reviewer expertise, agreement, disagreement, and the emerging role of digital pathology and AI. CAP lands on two strong recommendations: first, pathologists should develop procedures to review cases for disagreements and potential interpretive errors; second, these reviews should be performed timely enough to affect patient care. CAP also recommends documented procedures, periodic monitoring, corrective steps when poor agreement is found, and use of clinically relevant grading systems, especially simpler systems with meaningful clinical cut points.
Strategically, this is a major move for CAP. It positions pathology not merely as a craft practiced by experts at microscopes, but as a modern diagnostic discipline with quality systems, measurable variation, AI-adjacent workflows, and accountability for patient-facing diagnostic outcomes. The implications extend beyond anatomic pathology. Genomics, molecular pathology, tumor profiling, and AI-enabled diagnostics all face analogous problems: complex interpretation, variable reporting, uncertain agreement, high clinical stakes, and the need for structured review before results become treatment decisions.
__________________________________________
What CAP Is Actually Addressing
The article is not about every kind of laboratory error. It is specifically about interpretive diagnostic error in anatomic pathology: the problem that arises when the final diagnosis based on tissue, cytology, or autopsy material is inaccurate, imprecise, delayed, or otherwise clinically misleading.
That distinction matters. A specimen can be accessioned correctly, processed correctly, stained correctly, and still generate an interpretive error. Conversely, what we casually call “pathologist error” may really reflect a chain of vulnerabilities: inadequate clinical history, specimen limitations, fatigue, cognitive bias, workload, ambiguous terminology, weak review processes, poor communication with clinicians, or absence of subspecialty expertise.
CAP’s framing is therefore more sophisticated than “pathologists make mistakes.” It treats diagnostic error as a diagnostic-process problem. The pathologist is central, but not alone. The patient, specimen pathway, clinical question, institutional setting, reviewer expertise, and communication loops all matter.
Scope of the Guideline
The guideline update asks a practical question: What are the most effective ways to reduce diagnostic errors in anatomic pathology?
The evidence review focuses on targeted review of surgical pathology and cytology cases, including review during the analytic or post-analytic phase. In plainer English, CAP is asking whether cases should be reviewed before or after sign-out, how cases should be selected, who should review them, whether expert or subspecialty review improves accuracy, whether AI helps, and whether ancillary studies reduce diagnostic error in specific settings.
The key point is that CAP does not mandate one universal review model. A large academic center with subspecialty sign-out, tumor boards, digital slides, and deep benches of experts will not look like a small community practice. CAP’s position is that each practice needs procedures that are documented, relevant to its setting, monitored, and capable of improvement.
That makes the article more flexible, but also more demanding. “We do peer review somehow” is not quite enough. CAP is nudging pathology groups toward a more formal quality-management model.
The Two Strong Recommendations
CAP issues two strong recommendations.
First, anatomic pathologists should develop procedures for review of pathology cases to detect disagreements and potential interpretive errors and improve patient care. This is the central recommendation. CAP is saying that case review should be built into practice, not treated as an informal favor in the hallway.
Second, anatomic pathologists should perform case reviews in a timely manner to have a positive impact on patient care. This is more consequential than it may sound. A retrospective review six months later may be useful for quality assurance, education, and risk management, but it does not help the patient if surgery, chemotherapy, or surveillance decisions have already been made. CAP’s preferred ideal is prospective review before finalization when feasible, while recognizing that timely retrospective review may still have a role.
The concept of timeliness is context-dependent. A difficult low-acuity case might tolerate an external consultation delay if the clinician knows what is happening. A high-acuity cancer case may not. CAP even draws an analogy to molecular testing in stage IV lung cancer: when the result drives near-term treatment, review and finalization must move with clinical urgency.
The Four Good Practice Statements
CAP also offers four good practice statements.
First, practices should have documented case-review procedures relevant to their own setting. This means the review process should fit the practice’s personnel, case mix, volume, complexity, and available expertise.
Second, practices should periodically monitor and document the results of case review. This includes tracking review rates, mandatory review categories, amended reports, addenda, frozen-section correlations, cytology-histology correlations, and other mechanisms.
Third, if reviews show poor agreement in a defined area, pathologists should take steps to improve agreement. This can include consensus conferences, calibration slide sets, terminology alignment, focused education, external consultation, or adoption of standardized criteria.
Fourth, when grading systems show poor agreement, pathologists should use established clinically relevant morphologic grading criteria, often with fewer tiers and clearer clinical cut points. This is one of the more practical insights in the guideline. Pathology grading systems often become more reproducible when they stop trying to divide borderline biology into too many intermediate categories and instead focus on distinctions that matter for treatment.
What the Data Show
One striking feature of the article is that diagnostic discrepancies remain common enough to matter. Across studies, discrepancy rates varied substantially depending on case type, review method, and setting. CAP reports that diagnostic error rates ranged from very low levels to more than 10% depending on review method and case type, and the table of studies reports median discrepancy and major discrepancy rates that are not trivial.
This does not mean that every discrepancy is a catastrophic error. Some disagreements are minor, some reflect terminology differences, and some occur in inherently difficult borderline diagnoses. But the numbers are large enough to support CAP’s core thesis: interpretive variation is not a rare curiosity. It is a recurring feature of pathology practice that deserves formal systems.
The article also reinforces a predictable but important finding: expertise matters. More experienced pathologists and subspecialty pathologists often show better agreement. Expert review at referral institutions can detect more discrepancies, although referral bias complicates interpretation. Team-based review and direct collaboration with clinicians may also improve accuracy.
The Approach: Not Random Policing, but Targeted Quality Design
The best reading of the CAP article is not that every case should be second-read by another pathologist. That would be impractical and probably wasteful. The better reading is that laboratories should build targeted review systems around known risk points.
Examples include:
High-risk diagnoses where an error changes therapy.
Low-volume areas where local expertise may be thin.
Borderline lesions known to have poor interobserver agreement.
New malignancy diagnoses before major treatment.
Cases headed to tumor board or multidisciplinary care.
Discordance between cytology and histology, frozen and permanent sections, or biopsy and resection.
Amended reports and addenda, which may reveal recurring process problems.
Clinician-requested reviews and “comfort reviews” for diagnostically uncertain cases.
This is a quality-system mindset. The goal is not to create a punitive culture or a bureaucratic overlay. The goal is to identify where review adds the most value and then monitor whether the process is working.
Why Timeliness Is Central
The article’s discussion of timeliness may be the most clinically important section. A second opinion is not equally valuable at all times. A correction before final sign-out can prevent harm. A correction after treatment has begun may still be educational and legally relevant, but the patient-facing benefit is diminished.
This issue is amplified in the post-Cures Act environment, where patients may see pathology results quickly through electronic portals. A premature or incorrect diagnosis can propagate through the medical record, alarm the patient, mislead clinicians, and generate later confusion even if amended.
Thus, CAP’s “timely review” recommendation is not just about turn-around time. It is about where in the clinical decision chain the review occurs. The closer the review is to initial sign-out, the more likely it is to prevent downstream harm.
AI and Digital Pathology: Carefully Introduced, Not Oversold
The article includes AI and digital pathology, but in a measured way. CAP notes that digital technologies and AI were among the reasons for updating the 2016 guideline, and the review includes a small body of AI-related literature.
The message is not “AI will fix pathology.” Rather, AI may help in certain types of review: detecting missed lesions, improving agreement, triaging cases, assisting with grading, or supporting image analysis. CAP also notes that AI tools may increase agreement between pathologists in some settings.
But CAP is careful. The article recognizes that AI still needs regulatory, workflow, validation, and implementation frameworks. AI is not substituted for professional judgment. Instead, it is placed inside a broader quality structure: review procedures, documentation, monitoring, and patient-centered decision-making.
This is strategically important. CAP is creating conceptual space for AI in pathology, but on CAP’s terms: as part of quality, safety, and diagnostic reliability, not simply as a productivity gadget.
Strategic Importance for CAP
This guideline matters strategically for CAP for several reasons.
First, it reinforces CAP’s role as the professional body that defines quality practice in pathology. In an era when outside forces increasingly shape diagnostics — FDA, CMS, private payers, AI companies, hospital systems, venture-backed platforms — CAP is asserting that pathologists themselves should define how diagnostic reliability is maintained.
Second, it aligns pathology with the broader patient-safety movement. Diagnostic error has been a major theme in medicine for more than a decade, but pathology can be oddly invisible in general patient-safety discussions. CAP is placing pathology squarely inside the diagnostic safety conversation.
Third, it creates a bridge between traditional pathology and digital pathology. CAP can support AI and whole-slide imaging without appearing to chase technology for its own sake. The argument becomes: digital tools are important because they can support error reduction, agreement, triage, documentation, and scalable review.
Fourth, it helps defend pathology’s professional value. If diagnostic quality depends on expertise, subspecialty judgment, second review, clinical correlation, and quality systems, then pathologists are not interchangeable commodity readers. They are central actors in a high-stakes diagnostic system.
Fifth, it may shape accreditation, institutional expectations, and payer thinking over time. Even if the guideline is voluntary, it may become part of what hospitals, pathology groups, malpractice reviewers, and quality committees regard as reasonable practice.
Why This Matters Beyond Anatomic Pathology
Although the guideline is formally about anatomic pathology, its logic extends naturally to genomics and molecular diagnostics.
Modern genomics is full of interpretive complexity. A tumor profile is not just a sequence file. It is an interpretation of variants, copy-number changes, fusions, biomarkers, tumor type, therapy rules, resistance mechanisms, guidelines, trials, regulatory labels, and payer constraints. The same is true, in different ways, for germline testing, pharmacogenomics, MRD, hereditary cancer panels, and emerging multi-omic diagnostics.
Genomics therefore faces many of the same issues CAP identifies in anatomic pathology:
Disagreement among experts.
Terminology drift across reporting systems.
Borderline interpretations where evidence is incomplete.
Dependence on clinical context.
Need for timely review before treatment decisions.
Role of tumor boards and molecular consensus conferences.
Potential value of AI tools, but only within validated workflows.
Need to monitor amendments, addenda, and discordant interpretations.
In genomics, the analogy to case review may include variant review boards, molecular tumor boards, expert curation, repeat analysis, knowledge-base updates, report amendment tracking, orthogonal confirmation in selected cases, and review of clinically actionable calls before release.
This is especially relevant as genomic tests become more comprehensive and more automated. The more a report claims to guide treatment, the more important it becomes to ask: Who reviewed the interpretation? Was there a structured review process? How are disagreements handled? How are report changes tracked? How are knowledge-base updates incorporated? When does AI assist, and when does a human expert decide?
The CAP error-reduction framework is therefore not just an anatomic pathology story. It is part of a larger movement toward diagnostic governance.
The Take-Home Message
CAP’s updated guideline is not flashy, but it is important. It says that diagnostic error reduction in pathology requires structure: case review, timely review, documentation, monitoring, agreement improvement, and clinically meaningful grading systems.
The article also subtly redefines what high-quality pathology means. It is not only the brilliance of an individual diagnostic pathologist. It is the ability of a practice to build systems that detect disagreement, learn from discrepancies, use expertise wisely, calibrate terminology, and intervene before errors affect patient care.
For pathology, that is a professional-quality agenda. For digital pathology and AI, it is a workflow and validation agenda. For genomics, it is a warning and an opportunity: complex diagnostic interpretation needs not only better algorithms, but better systems of review, accountability, and clinical integration.
Sidebar: Five Surprising Aspects of the CAP Article
1. CAP makes strong recommendations despite imperfect evidence.
The evidence base is real, but much of it is observational, heterogeneous, and not always tied directly to patient outcomes. CAP still concludes that the balance of evidence and clinical logic supports strong recommendations for structured and timely review.
2. The discrepancy rates are still substantial.
The update does not suggest that pathology has “solved” diagnostic variation since 2016. Discrepancy and major discrepancy rates remain meaningful, especially in selected, difficult, malignant, external-review, or single-organ settings.
3. Timeliness is treated as a patient-care concept, not just a lab metric.
A review is not simply “done” or “not done.” Its value depends heavily on whether it occurs before clinical management is initiated.
4. CAP gives AI a role, but not the starring role.
AI appears as a promising tool for missed-lesion detection, grading support, triage, and agreement improvement. But the foundation remains human judgment, structured review, documentation, and patient-centered practice.
5. Simpler grading systems may be safer than elaborate ones.
One of the most practical messages is that fewer grading tiers, tied to clinically meaningful cut points, can improve agreement. In pathology, more granularity is not always more truth.
The strongest citation anchors for this draft are: the article’s objective/design/results/conclusion and error definition; the two strong recommendations and four good practice statements; the systematic-review methods and 101 extracted articles; the discrepancy-rate table; the timeliness discussion; and the AI/future-considerations section.
