3 days ago
How AI Can Shape The Future Of Clinical Documentation Improvement
Sameer Ather, MD, PhD, Co-Founder & CEO of XpertDox, leveraging AI to enhance medical coding efficiency and healthcare productivity.
Clinical documentation is the written record of a patient-physician encounter.
Before the advent of electronic health records (EHRs), these encounters were often inadequately documented. While this allowed for faster visits and more personal interaction between doctor and patient, it created several issues.
Without a written record, follow-up visits lacked context, making continuity of care more difficult. In cases requiring multidisciplinary input, providers had no shared documentation to review, limiting collaboration and informed decision-making. The absence of structured records also made it difficult for healthcare systems to monitor or assess care quality.
Insurance companies and hospital administrators, too, lacked insight into the care being provided. This lack of transparency made assessing performance, ensuring compliance or justifying reimbursement challenging. Documentation, or the lack thereof, became a silent barrier to systemic efficiency and accountability.
Today, comprehensive and timely clinical documentation is not just preferred, it's essential. It is a cornerstone for care coordination, patient safety, quality reporting and reimbursement. As expectations grow, so too does the pressure on clinicians to document thoroughly, accurately and efficiently.
Clinical documentation has evolved immensely since EHRs were introduced in the 1990s, and the next major evolution is the introduction of AI. Let's look at how AI will impact how clinical interactions are recorded.
The Role Of Clinical Documentation Improvement (CDI)
Clinical documentation improvement (CDI) ensures that medical records are complete, accurate and clinically meaningful. It captures the complete picture of a patient's condition, the care delivered and the outcomes achieved.
The key benefits of CDI include:
• Better Patient Care: Comprehensive records enable clinicians to make informed decisions based on a clear understanding of patient history and treatment.
• Regulatory Compliance: Accurate, rule-aligned documentation supports billing and legal standards compliance.
• Fair Reimbursement: When documentation reflects the complexity of care, providers are more likely to be compensated appropriately.
• Improved Data Use: High-quality documentation supports analytics, planning and population health management.
Current State Of CDI
Today, CDI is essentially a manual and retrospective process. Specialists review a subset of medical records after the visit has occurred. If documentation is incomplete or ambiguous, they send queries to the provider, often days or weeks later, when recall may be limited. Due to staffing constraints, only a fraction of cases are reviewed, leaving many gaps unaddressed.
Although widely recognized as essential, CDI is not yet fully integrated into clinical workflows. Rather than easing the administrative burden, it often adds to it. For CDI to achieve its potential, it must become real-time and seamlessly embedded into the care process.
Without CDI, for example, a provider might document only "diabetes." This leads to a generic code that doesn't reflect all aspects of the patient's condition, including any complications from diabetes. With CDI support, the physician is prompted to include "diabetic nephropathy," allowing accurate assignment of a Hierarchical Condition Category (HCC) code and ensuring appropriate reimbursement and clinical clarity.
The Historical Evolution Of AI In Clinical Documentation
Until the early 2010s, most hospitals relied on paper records. With the adoption of EHRs, the shift to digital documentation opened the door to innovation. Early efforts focused on rule-based systems that searched for predefined terms in provider notes. However, these notes are often unstructured and lengthy, limiting the scalability of such systems.
The emergence of machine learning and natural language processing (NLP) changed the landscape. Models, including gradient boosted decision trees and large language models (LLMs), have made it possible to interpret and analyze vast amounts of clinical text. This enabled the development of more innovative CDI tools to improve accuracy across diverse clinical scenarios.
According to the 2024 Association of Clinical Documentation Integrity Specialists (ACDIS) survey, 56.58% of respondents (download required) reported that technology has helped identify straightforward documentation gaps, freeing up CDI teams to focus on more complex cases—an increase from 49.37% in 2023.
AI's Potential Role In Improving CDI
AI is particularly well-suited for reading and interpreting complex clinical narratives. It can process unstructured data at scale and offer real-time support to providers. For instance, AI-powered tools can:
• Identify missing or ambiguous information during or immediately after a clinical encounter.
• Transcribe and organize spoken interactions between patients and clinicians.
• Recommend accurate diagnoses and codes based on current and historical data.
• Simplify technical language to support better patient understanding.
Looking ahead, AI tools will not only flag deficiencies—they will proactively surface likely conditions, suggest evidence-based coding and highlight potential compliance risks. Over time, these tools will be able to adapt to individual documentation styles to offer personalized support.
Challenges With The AI Model
Despite their promise, AI models—particularly LLMs—pose notable challenges in clinical settings. Their probabilistic nature means they may generate different outputs for the same input, which is incompatible with the deterministic standards of clinical care. Accuracy, reproducibility and traceability are essential when patient safety is at stake.
One persistent issue is hallucinations, where the AI fabricates information that may seem plausible but is factually incorrect. While cleaning training data and applying rule-based constraints have helped reduce these errors, they do not eliminate them. Incorporating human oversight can further mitigate risk but introduces additional workflow complexity.
LLMs also lack transparency. Their black-box architecture makes auditing claims or explaining specific outputs difficult, raising concerns during clinical reviews or regulatory scrutiny. Although models like gradient boosting machines (GBMs) offer more interpretability and consistency, they struggle to capture nuanced relationships in unstructured medical text.
Conclusion
AI has the potential to transform clinical documentation from a tedious task into a strategic enabler of quality care. By synthesizing complex narratives and offering real-time, intelligent support, technologies like LLMs and GBMs can help clinicians document more accurately and efficiently.
However, this transformation must be approached with care. Ensuring clinical accuracy, protecting patient privacy and maintaining provider trust are essential.
With thoughtful integration, rigorous validation and ongoing human oversight, AI-powered clinical documentation improvement can enhance—not replace—clinical judgment, paving the way for a future where technology supports care without compromising its integrity.
Forbes Technology Council is an invitation-only community for world-class CIOs, CTOs and technology executives. Do I qualify?
