Overview of Clinical Coding in Saudi Arabia

Clinical coding transforms the descriptions of diagnoses, injuries, and clinical interventions documented by healthcare professionals into standardized alphanumeric codes. These codes are internationally recognized for organizing, analyzing, and reporting healthcare information. Within Saudi Arabia, the healthcare sector relies on several key systems:

• ICD-10-AM (International Classification of Diseases, 10th Revision, Australian Modification, 10th Edition): for classifying diseases and health conditions.
• ACHI (Australian Classification of Health Interventions): for coding clinical procedures.
• SBS (Saudi Billing System): for procedural coding specific to local billing requirements.
• GTIN (Global Trade Item Number): to catalog and track medications.
• GMDN (Global Medical Device Nomenclature): for identification and regulation of medical devices.
• DRG (Diagnosis-Related Grouper): for classification and reimbursement of inpatient cases.

Clinical Coding Process

• Clinical Documentation: Physicians and nurses record patient encounters, providing details on symptoms, diagnoses, treatments, and procedures.
• Coder Review: Trained health information professionals (clinical coders) meticulously review the patient record for completeness and accuracy.
• Code Assignment: Coders map the clinical concepts to standardized code sets using codebooks and electronic tools. For example, “acute myocardial infarction” is translated to its appropriate ICD-10-AM code, not just described in words.
• Data Entry: The assigned codes are entered into the health information systems for downstream applications, including billing, analytics, and reporting.

Inclusion and Exclusion Notes in Coding Standards

• Inclusion Notes: Define the conditions, synonyms, or subclassifications considered part of a code—even if not explicitly named in the code title. These facilitate the coding of related or equivalent terms under a single code.
• Exclusion Notes: Specify conditions not included within a code. These guide coders to refer to alternate codes, preventing misclassification and double coding.

Importance of Inclusion/Exclusion Notes

• Accuracy: Ensures the coded data reflects the true clinical picture.
• Consistency: Promotes standardized coding, allowing comparability across institutions and jurisdictions.
• Clarity and Compliance: Minimizes ambiguity, supporting correct billing and transparent reporting.

Key Coding Standards in Healthcare

Coding standards provide the syntax and grammar for converting clinical records into structured data for diverse stakeholders (healthcare providers, payers, regulatory bodies, and researchers). Commonly implemented standards include:

• ICD-10 AM: Diagnoses
• ACHI/SBS: Procedures/interventions
• GTIN: Medications and Herbals
• GMDN: Medical devices

Examples of coding standards for specific coding scenarios:

• 0001: Principal Diagnosis
• 0002: Additional Diagnoses
• 0003: Supplementary Codes for Chronic Conditions
• 0005: Syndromes
• 0016: General Procedure Guidelines
• 0019: Intervention Abandoned/Not Completed
• 0020: Bilateral/Multiple Procedures

Why Clinical Coding Is Critical

• Reimbursement and Billing: Accurate coding is the foundation for correct insurance claims and hospital reimbursement.
• Analytics and Reporting: Codified data enable meaningful health statistics and inform public health policy and research.
• Quality and Safety: Supports audits, benchmarking, and initiatives to enhance patient care and outcomes.

---

Professional Risks of Using General-Purpose Generative AI (e.g., ChatGPT) for Clinical Coding

Despite the advances of large language models (LLMs), their deployment in clinical coding is widely viewed as problematic for several reasons:

1. Lack of Accuracy and Hallucination Risk: LLMs are prone to generating incorrect, non-existent, or fabricated codes (“hallucinations”), particularly when confronted with complex or nuanced cases. Studies document a deterioration in coding accuracy with increasing case complexity.
2. Deficient Contextual and Clinical Judgment: Clinical coding requires interpretation of clinical nuance and regulatory considerations. General-purpose LLMs lack the capacity to consistently comprehend the intent and specificity embedded in clinical documentation.
3. Data Privacy and Security Limitations: Employing ChatGPT or similar LLMs generally involves transmitting sensitive patient data through cloud services and outside the country. This raises substantive concerns regarding compliance with Saudi data protection regulations-SDAIA PDPL, data confidentiality, and risk of breaches.
4. Regulatory, Ethical, and Legal Risks: Coding inaccuracies derived from general AI tools can precipitate substantial legal and financial liabilities for healthcare providers. Current regulations demand transparency, and accountability.
5. Incompatibility with Clinical Coding Rule-Based Decision Making: Clinical coding is inherently rules-driven. LLMs built for flexible text generation are not aligned with the rule-based and protocol-driven nature of coding, emphasizing the need for specialized engines integrated with coding logic.
6. Dependence on Outdated or Non-Specific Data: General-purpose large language models (LLMs) are trained on broadly available internet content and do not have access to proprietary or regularly updated resources such as ICD-10-AM codes. As a result, they are prone to inaccuracies and may generate incorrect or fabricated codes (“hallucinations”), particularly when precise and up-to-date coding information is required.

Conclusion

Given the complexity, regulatory requirements, and high stakes involved in clinical coding, relying on general-purpose generative AI models like ChatGPT is inappropriate at this time. Best practice remains the use of rigorously trained clinical coders, supported by dedicated coding tools and thoroughly validated decision support systems, designed to meet clinical, operational, and compliance needs.