Why automated clinical coding is the key to saving the NHS valuable time and money
The demand for NHS services is at an all time high. This means clinicians are submitting more patient data than ever before into electronic health record systems. As a result of the high volume of data, many NHS Trusts are now hiring a team clinical coder at an extremely high cost.
Clinical coders record information about every patient who visits hospital, investigating all aspects of their journey from start to finish. For instance, they will record why a patient was admitted to hospital, how long they stayed, what treatment they were given and recovery time. Clinical Coders use alphanumeric code to record all of this information on the computer system. These records can then be understood throughout the NHS and used to plan for future patient care.
The current international standard for medical coding is ICD-10 (the tenth version of International Classification of Disease codes), from the World Health Organisation. ICD‑10 has over 14,000 codes for diagnoses and the next iteration, ICD-11 will contain over 55,000 diagnostic codes.
For decades, medical coders have relied on “code books” to look up the right code for classifying a disease or treatment, which slows the process down dramatically. And it is not just a matter of finding the right code. There are interpretation issues. With ICD-10 and prior versions of the classification scheme, there is often more than one way to code a diagnosis or treatment, and the medical coder has to decide on the most appropriate choices.
To add to this, fluctuations in patient admissions also puts pressure on Trust’s and at peak times, often they are forced to turn to agency staff paying a basic rate of up to £35 per hour for a single coder. So the current process is neither efficient or cost effective. But what if this process could be automated?
By automating clinical coding, NHS Trusts could have a clear view on how many coders are needed based on the number of admissions. When coding tools are part of the clinical documentation process, physicians could be offered appropriate coders based on the patient’s history and current documentation.
Currently, there are no such cases of clinical coding having been automated in the UK. However, over in America the story is different. For a number of years, they have been working on incorporating state-of-the-art machine learning methods and other aspects of artificial intelligence to enhance the system’s ability to analyse clinical documentation, charts and notes to determine which codes are relevant to a particular case. Some medical coders are now working hand-in-hand with AI-enhanced computer-assisted coding systems to identify and validate the correct codes (click the link at the bottom of the blog to find out more).
So how does it work?
As we’ve already determined, manual clinical coding is expensive, time consuming and prone to error. Using an automated system would apply a more consistent and transparent understanding to clinical records and would allow clinicians to gain easier access to historical information.
Traditionally, clinical coding using the international classification of diseases codes are derived from free text notes. This means you could produce automated clinical coding using text mining on the free text notes which would allow for keywords to be selected from the text. This would be done using natural language processing (NPL) keyword extraction. These keywords can then be classified into the corresponding codes.
Electronic health record use a structured clinical vocabulary, known as SNOMED CT, to ensure information is recorded in a clear, consistent, and comprehensive manner. The use of SNOMED CT indicates that free text notes should be written in a similar way. Therefore, it may be possible to use the SNOMED CT as an addition to any training data to enhance the model.
It’s important to consider that automated coding is made especially challenging by the idiosyncrasies of clinical text, the large number of disease codes and their unbalanced distribution, so automating the entire process may not be possible yet. Having said that, even using AI-assisted automation still has a great potential to save valuable resources and real-time availability of codes which would improve oversight of patient care and would accelerate research, making it a very worthwhile investment for NHS Trusts.
To read the thoughts and opinions of the clinical coders in the US who are already using AI-assisted clinical coding systems, click here.