Breaking ICD Codes: Identifying Ambiguous Respiratory Infection Codes via Regional Diagnosis Heterogeneity.

Purpose: We aimed to analyze regional variations in the assignment of International Classification of Diseases, 10th Revision (ICD-10) codes to acute respiratory infections, seeking to identify notable anomalies that suggest diverse diagnoses of the same condition.

Methods: We analyzed national weekly diagnosis data for acute respiratory infections (ICD-10 codes J00-J22) in Poland from 2010 to 2019, covering all 380 county-equivalent administrative regions and encompassing 292 million consultations. Data were aggregated into age brackets. We calculated the Kendall tau correlations between shares of particular diagnoses.

Results: We found staggering differences across regions in applied diagnoses that persisted even after disaggregating the data into age groups. The differences did not seem to stem from different levels of health care use, as there was no consistent pattern suggesting variability in milder diagnoses. Instead, there were numerous pairs of strongly negatively correlated codes implying classification ambiguity, with the most problematic diagnosis being J06 (acute upper respiratory infections of multiple and unspecified sites), which was used almost interchangeably with a diverse range of others, especially J00 (common cold) and J20 (bronchitis).

Conclusions: To the best of our knowledge, this is the first study using observable anomalies to analyze regional coding variability for the same respiratory infection. Although some of these discrepancies may raise concerns about misdiagnosis, the majority of cases involving interchangeably used codes did not seem to substantially impact treatment or prognosis. This suggests that ICD codes may have clinical ambiguities and could face challenges not only in fulfilling their intended purpose of generating internationally comparable health data but also in their use for comprehensive government health planning.