Beyond the Report: Communication Governability in the Era of Artificial Intelligence

Executive Summary

🛡️ The Last Governance Boundary: Traditional AI governance asks "Was the model accurate?" CCG asks "Did the information survive the journey from inference to action?"
⚕️ Clinical Reality: The report is not the product. The clinical decision is the product. Outcomes rely on deterministic information transfer, not just data generation.
⚙️ Systems Engineering: Communication is no longer a "soft skill." It is a deterministically governable information layer requiring strict state management (S0-S5).
📊 Incident Taxonomy: Introduces C1-C6 failure modes (Context Loss, Semantic Distortion, Escalation Failure, Authority Diffusion, Acknowledgement Failure, Communication Overload) to formalize communication errors as governance breaches.
🏥 RATSe-Health Integration: Communication governance operates as a core module within the broader RATSe-Health architecture, ensuring continuous accountability.

Abstract

Radiology has traditionally been viewed as a diagnostic specialty whose primary output is the radiology report. However, as noted in earlier clinical paradigms, the report is not the product; the clinical decision is the product.

Historically, communication failures between radiologists and referring physicians have contributed to delayed diagnoses, missed critical findings, inappropriate treatment decisions, and patient harm. [2] [3]

The emergence of Artificial Intelligence (AI) introduces a new paradox. While AI improves reporting speed, workflow efficiency, and information extraction [4], it simultaneously risks increasing communication distance between radiologists and clinicians. More data generation does not equate to safer information transfer.

This paper argues that communication must shift from being viewed as an informal professional skill to a deterministic, governable information layer. [1] We introduce:

Clinical Communication Governability (CCG)

defined as:

The ability of a healthcare system to ensure that clinically relevant information is transferred, interpreted, acknowledged, and acted upon while preserving meaning, accountability, authority, traceability, and patient safety.

Integrating seamlessly into the existing RATSe-Health architecture, CCG establishes that communication is not complete when a report is transmitted; it is complete only when clinical meaning reaches the accountable decision-maker.

1. The Evolution from Generation to Governability

In 2022, before the widespread adoption of generative AI in healthcare, practical communication principles were proposed for radiology reporting. [5] These included:

understanding the referring physician's question
highlighting critical findings
attaching representative images
communicating uncertainty
discussing urgent findings directly
avoiding excessive radiology-specific terminology
structuring reports around clinical decision-making

These recommendations were developed from clinical practice rather than technology trends. The subsequent rise of AI has not reduced their importance; it has fundamentally altered their systemic role.

As radiology workflows become increasingly automated through ML inference and agentic summarization, communication itself becomes the critical safety boundary. The paradigm has shifted.

The Old Challenge

Can we generate reports faster?

The New Challenge

Can critical information reach the right clinician, at the right time, with the right meaning?

2. The Communication Paradox of AI

AI promises:

faster reporting
structured reporting
automated prioritization
automated summarization
automated alerts

Yet AI also creates:

information overload
alert fatigue
communication dilution
loss of context
reduced direct clinician interaction

The result is a paradox:
More information does not necessarily produce better communication.

In many healthcare environments:

More Data ≠ More Understanding

The communication problem therefore evolves from information generation to information governability.

3. Communication as a Safety-Critical System

Traditional Thinking

Image → Report → Clinician

Modern Reality

Image → AI Analysis → AI Report Draft → Radiologist → EHR → Alert Systems → Clinician → Clinical Action

Every transition introduces risk. Failures may occur through:

omission

misunderstanding

ambiguity

delay

alert fatigue

workflow interruption

semantic distortion

Communication therefore behaves like a clinical system. And clinical systems require governance.

4. The Last Governance Boundary

The core thesis of Clinical Communication Governability (CCG) is that communication acts as the final failsafe in the diagnostic continuum.

Traditional AI Governance Asks:

"Was the model accurate?"

CCG / RATSe-Health Asks:

"Did the information survive the journey from image to action?"

AI creates a distinct paradox in this journey. It promises structured reporting, automated prioritization, and faster turnaround. Yet, it concurrently generates information overload, semantic dilution, loss of context, and a reduction in direct clinician interaction.

Clinical Information Journey

🖼️ Image

→

⚙️ Inference

→

CCG Governed Segment

📄 Report

→

💬 Communication

→

⚖️ Decision

→

⚡ Action

→

✅ Outcome

5. The Communication State Machine

To treat communication as a governable layer, we must abandon the notion of communication as a singular event. Runtime governability requires us to model communication as a state machine. Failure can occur at any transition.

Generated

AI/Radiologist creates report

→

Delivered

Arrives in EHR/Inbox

→

Acknowledged

Clinician opens/reads

→

Understood

Meaning preserved

→

Actioned

Clinical decision made

→

Closed Loop

Traceable completion

Traditional systems focus only on getting from S0 to S1. True Clinical Communication Governability demands systemic verification through S5.

6. Communication Incident Taxonomy

If communication is an information governance layer, communication failures are formal governance incidents. Aligning with IRHAI Incident Taxonomy, we classify communication failures into five specific modes.

C1 — Context Loss

Examples: Missing clinical question, incomplete history, absent indication.

Result: Correct report, but wrong clinical interpretation.

C2 — Semantic Distortion

Examples: Excessive jargon, ambiguous terminology, AI-generated summarization errors.

Result: Information technically delivered, but clinical meaning lost.

C3 — Escalation Failure

Examples: Critical findings not directly communicated, alert fatigue, notification system failures.

Result: Diagnosis known by system, but action critically delayed.

C4 — Authority Diffusion

Examples: AI recommendation accepted without review, uncertainty regarding inter-departmental responsibility.

Result: Process executes, but with no accountable human actor.

C5 — Acknowledgement Failure

Examples: Report successfully delivered to inbox, but clinician never reviewed it.

Result: Communication technically successful (S1), but operationally failed (S2 missing).

C6 — Communication Overload

Examples: Excessive alerts, multiple AI notifications, dashboard fatigue, duplicated escalations.

Result: Critical information hidden within excessive information.

7. RATSe-Health: The Communication Module

Rather than existing as a separate protocol, Clinical Communication Governability acts as an implementation module directly within the established RATSe-Health architecture.

R — Responsible Communication

Was the communication clinically necessary and patient-centered?

Requires answering the clinical question, communicating relevant findings contextually, and avoiding unnecessary algorithmic complexity.

A — Accountable Communication

Who owns the communication state transition?

Requires an identified reporting radiologist, documented hand-offs, and escalation ownership.

Principle: Communication authority (C4) remains non-transferable.

T — Transparent Communication

Can clinical meaning be deterministically reconstructed?

Requires clear language free of semantic distortion (C2), structured impressions, explicit uncertainty disclosure, and image-supported evidence.

S — Safe Communication

Could a state transition failure cause patient harm?

Safety requires engineered escalation pathways to prevent C3 incidents (e.g., missed aortic dissection, delayed stroke notification).

E — Ethical & Equitable Communication

Did communication reach all stakeholders fairly?

Requires addressing language barriers, ensuring information accessibility across diverse patient populations, and confirming equitable distribution of critical notifications.

Legal Defensibility (RATSe-ML Integration)

A communication pathway should be deterministically reconstructable to ensure medico-legal defensibility and systemic auditability. Governance requires answering:

▶ Was the report delivered?
▶ Was it acknowledged?
▶ By whom?
▶ When?
▶ Was escalation triggered?
▶ Was the recommendation acted upon?

8. The CARE Operational Model

To operationalize CCG at the individual report level and prevent C1 (Context Loss), every communication must address CARE.

C — Context

Why was the examination performed?

A — Assessment

What is the imaging diagnosis?

R — Risk

Why does it matter?

E — Execution

What should happen next?

9. The Future Radiologist

Historically

Image Interpreter

→

Today

Diagnostic Consultant

→

Tomorrow

Clinical Communication Governor

The future radiologist will not be defined by image interpretation alone. AI will increasingly assist interpretation.

The differentiator will become:

contextual understanding
uncertainty management
communication
accountability
clinical integration

The radiologist becomes the bridge between artificial intelligence and clinical action.

Conclusion

As healthcare rapidly integrates ML and agentic systems, governance must expand beyond algorithms to encompass the pathways through which diagnostic information flows.

The future of radiology will not be determined solely by who interprets images best, nor will outcomes be secured merely by generating data faster. Outcomes depend deterministically on information surviving the journey from inference to action.

Communication is no longer a soft skill.

Communication is a Governable Information Layer.

References & Suggested Reading

Governing Artificial Intelligence in Radiology: A Systematic Review of Ethical, Legal, and Regulatory Frameworks Diagnostics (Basel), PMC. Analyzes the ethical, legal, and regulatory models for integrating AI in clinical imaging to ensure accountability and transparency.
Closing the loop on test results to reduce communication failures: a rapid review of evidence, practice and patient perspectives BMJ Quality & Safety, PMC. Synthesizes evidence regarding communication breakdowns of diagnostic test findings and practical interventions to reduce patient harm.
Communication errors in radiology - Pitfalls and how to avoid them Clinical Imaging, PubMed. A comprehensive framework describing common communication failures among providers and outlining solutions to optimize patient outcomes.
Opportunity and Opportunism in Artificial Intelligence–Powered Data Extraction: A Value-Centered Approach American Journal of Roentgenology (AJR). Explores the paradigm shift of AI data extraction in radiology and the associated risks of separating radiologic results from their clinical context.
How to Improve Communication in the Radiology Report Abdominal Imaging Blog, September 2022. Practical guidelines and clinical recommendations for enhancing the clarity and effectiveness of diagnostic communications.