Introduction
The International Classification of Diseases (ICD) codes are frequently used in healthcare for various purposes, including identifying specific conditions like acute optic neuritis (aON). However, the reliability of these codes, particularly when used in conjunction with diagnostic imaging such as MRI, remains a critical question. This study delves into the accuracy of ICD codes for aON, especially when combined with MRI findings within a two-month timeframe of the initial diagnosis code. Understanding the precision of these diagnostic tools is crucial for accurate patient identification and research.
Methods
A retrospective chart review was conducted utilizing a comprehensive medical record repository from a major healthcare institution, spanning from 1998 to 2019. The study included patients identified by ICD-9 and ICD-10 codes for optic neuritis who also underwent an MRI of the brain and/or orbits within two months of their earliest ICD code. The MRI results were categorized as either positive or negative for aON, based on the presence of gadolinium-contrast enhancement noted in the radiology report. Furthermore, the clinical diagnosis at the time of the initial ICD code was classified into several categories: aON, prior ON, considered ON, alternative diagnosis, or unknown. This classification was based on a thorough review of physician notes documented within seven days of the initial code. This rigorous methodology allowed for a detailed assessment of diagnostic accuracy.
Results
The study cohort consisted of 251 patients who had MRI results available within two months of receiving their initial optic neuritis ICD code. Among these, 49 patients had positive MRI findings, while 202 had negative MRI results. Intriguingly, even within the negative MRI group, a significant portion (16%) were clinically diagnosed with aON. Additionally, 20% had a history of prior ON, and 9% were considered to have ON as a possible diagnosis. A substantial 46% were ultimately diagnosed with alternative conditions, and in 9% of cases, the diagnosis remained undetermined at the time of coding. Overall, when considering all subjects, the accuracy of ICD codes alone for identifying acute ON was found to be only 25%. The accuracy improved slightly to 41% when considering both acute and prior ON, and further to 48% when including cases where ON was merely considered as a diagnosis. The study also identified factors associated with increased accuracy for clinically confirmed aON diagnosis, including a positive MRI, a higher number of ON ICD codes, codes assigned by ophthalmologists or neurologists within 2 months, and a neurology encounter within the same timeframe.
Conclusion
The findings of this study highlight the limited accuracy of ICD codes for optic neuritis, particularly for acute cases, even when MRI is conducted within two months of coding. While accuracy is somewhat better when considering a broader definition of ON (including prior or considered cases), it remains suboptimal for precisely identifying acute ON using ICD codes alone. Notably, the accuracy is significantly higher in cases with positive MRI findings, suggesting that a positive MRI in conjunction with ICD codes enhances diagnostic precision. The study strongly indicates that relying solely on ICD and Current Procedural Terminology codes to identify aON cases in research, especially in claims-based studies, may introduce a considerable risk of misclassification bias. Therefore, caution is advised when using these codes in isolation for research or clinical decision-making related to acute optic neuritis.
