Rheumatology Data Reflect Poor Adoption of ICD-10

Inflammatory arthritis codes increased 30-fold in the transition from the ninth to the 10th revision of the International Classification of Diseases (ICD-9 and -10), yet few were used in clinical practice, according to new research.

Most of the new codes for inflammatory arthritis in ICD-10 were rarely used, if at all, from 2015 to 2021.

"About 10-20 codes were comprising the majority of usage for inflammatory arthritis patients in ICD-10," first author Justin Zhu, a researcher and medical student at Yale University in New Haven, Connecticut, told Medscape Medical News. "The other 380 or 400 codes just weren't seeing a lot of use."

Justin Zhu

The findings show the difficulties of transitioning to a new system, he added, and emphasize the need for additional training to improve adoption of ICD-11. The new coding system launched globally in January 2022, but it is not clear when it will be implemented in the United States.

ICD-10 was launched in the United States in 2015, with the goal of enabling greater specificity in identifying health conditions. For example, the new coding system allowed users to include information on laterality and anatomic location for the first time. The total number of codes increased from 14,500 with ICD-9 to 70,000 with ICD-10, with the number of inflammatory arthritis diagnosis codes growing from 14 to 425.

To see how these ICD-10 codes were utilized compared with ICD-9, Zhu and colleagues used national multi-insurance administrative claims data to find inflammatory arthritis diagnostic codes for over 5.1 million patients. About half were coded in ICD-9, while the remaining half were coded in ICD-10. Zhu and colleagues defined "higher-usage codes" as those that were used more than 1% of the time.

The findings were published in a research letter in JAMA Network Open on April 18.

For ICD-9, four of the available 14 codes (28.6%) were higher-usage codes. In contrast, only nine of the 425 ICD-10 codes (2.1%) were frequently used. Though ICD-10 allowed for increased granularity in diagnosis, data showed that nonspecific codes were most popular. Of the 20 most used ICD-10 arthritis codes, 65% contained "unspecified or other specified" in its wording.

The researchers also found that there was no significant change in these higher-usage codes throughout the study period from 2015 to 2021, suggesting there was not a detectable learning curve in ICD-10 usage among physicians and coders. They also found that clinician specialty did not change code usage patterns.

"The percentage of codes used was not better for rheumatologists (who might be expected to be more refined users of such codes) than primary care clinicians," Zhu and colleagues wrote.

Moving to ICD-11 Brings Challenges as Well as Opportunities

Zhu noted that the study highlights the challenges of adopting new technological systems into daily practice, which can inform the eventual transition to ICD-11.

"There is this need to emphasize training as well as just invest more in improving adoption of ICD-11," he said.

Michael Pine, MD, MBA, of MJP Healthcare Innovations, LLC in Evanston, Illinois, added that ICD-11 needs to be more user-friendly to be useful in practice. While ICD-10 allowed for greater granularity in coding, it did not result in "usable granularity, in terms of the things doctors really want to communicate," he told Medscape Medical News.

Michael Pine, MD, MBA

And the transition to ICD-11 could pose greater challenges; rather than ICD-10's taxonomy system, ICD-11 is formatted as an ontology.

"Although ICD-11 retains some precoordinated codes that convey multifaceted compound concepts, its structure and syntax also provide for post-coordination, a new feature to the ICD that supports the customized combination of concepts and modifier codes to capture previously inaccessible clinical nuance," he wrote in a co-authored invited commentary.

This added clinical nuance, however, will potentially make coding more complex, he said. One solution is to automate coding, such that clinicians could input information in a natural clinical format that makes sense to them, which would then be translated into ICD-11 code by a program. (This would then be translated back to the user in the natural clinical format to ensure accuracy.)

This type of process would limit how much any one person would need to know about ICD-11 to code diagnoses effectively, while also taking full advantage of the increasing specificity of the new coding system, he said.

Such a program does not yet exist but could be possible with intensive investment in the transition to ICD-11.

The findings of this study serve as a cautionary tale for future transitions to new systems without considering the importance of user experience and usability, Pine noted. If the United States takes an approach for the adoption of ICD-11 that is similar to that used for ICD-10, it is likely to be "just another overhyped transition" that will make users unwilling to adopt any new system moving forward out of frustration.

But if the United States takes a different, innovative approach, the opposite could be true.

"In short, the US must decide whether it is time to invest considerable resources and effort into a 21st-century information system that could overcome such hindrances as asymmetric information for decision-making, faulty risk adjustment in performance evaluations and payment formulas, and burdens imposed by current coding and documentation practices," the commentary reads.

"It will allow us to make the best of what computers do and the best of what clinicians do," Pine added, "and get them to work together in ways which would not have been conceivable 50 years ago."

No information on study funding was provided. Zhu and Pine did not disclose any competing interests.