Harnessing AI for Clinical Coding: Promise and Challenges Ahead

Hello and welcome to today’s discussion on an exciting evolution happening at the intersection of healthcare and technology: Automated Clinical Coding. Before we dive in, let’s first understand what exactly clinical coding is and why it’s so essential.

Clinical coding is the process of translating complex medical information from a patient’s health records into structured codes. These codes are incredibly vital as they allow for statistical analysis and tracking of diseases and treatments. But it’s also a cognitive, labor-intensive, and time-consuming task that demands high levels of consistency.

The question is, could we leverage technology to make this process more efficient and accurate? Hang Dong, Matúš Falis, William Whiteley, and their team believe we can. Drawing from their project experience over two and a half years and their extensive discussions with clinical coding experts across Scotland and the UK, they put forward their perspectives on the subject. So, let’s unwrap what they have to say!

A Shift Towards Automated Clinical Coding

The idea that underlies their proposition is automating clinical coding. They suggest that this cognitive process could potentially be supported by an automated system driven by Artificial Intelligence (AI) and Natural Language Processing (NLP) technologies. This could significantly enhance the speed and accuracy of clinical coding.

It sounds simple, right? Just let the machines do the work! But, as you may have guessed, it isn’t quite as straightforward as it seems.

Challenges: Deep Learning, Explainability, and Consistency

When you’re dealing with medical data, there are several factors at play that make automation a challenging task. One of the major hurdles that the team identifies is the gap between the current deep learning-based approach to clinical coding and the requirement for explainability and consistency in real-world practice.

For instance, the deep learning algorithms currently in use are like “black boxes,” and their decision-making process is often inscrutable. This lack of transparency is a significant problem when you’re dealing with something as crucial and sensitive as patient data.

That’s why the team suggests that knowledge-based methods, which can represent and reason the standard process of a task, may need to be integrated with deep learning methods for clinical coding. This would ensure not just efficiency but also the much-needed explainability.

The Road Ahead: An Exciting Journey

Despite the technical and organizational hurdles, automated clinical coding is a promising prospect. It’s important to note, though, that coders should still play an integral role in the development process. After all, their expertise and experience can guide the creation of more efficient, accurate, and transparent systems.

Looking ahead, there’s a lot to achieve. Developing and deploying an AI-based automated system to support coding is a task for the next five years and beyond. And while the journey may be challenging, the potential benefits are too significant to ignore.

Automated clinical coding could revolutionize healthcare analytics, improve patient care, and make the whole system more efficient. It’s a fascinating space to watch, and we can’t wait to see where the road leads!

Let’s remember, the future of healthcare lies in the synergy between technology and the human touch. So, stay tuned as we continue to explore more about this exciting frontier in the upcoming posts.


Dong, H., Falis, M., Whiteley, W. et al. Automated clinical coding: what, why, and where we are?. npj Digit. Med. 5, 159 (2022). https://doi.org/10.1038/s41746-022-00705-7

Related Articles


This site uses Akismet to reduce spam. Learn how your comment data is processed.