Material and Method: A retrospective cross-sectional study was conducted among patients hospitalized in six general medicine wards at Siriraj Hospital. ICD-10 code data was identified and retrieved directly from the hospital administrative database. Patient comorbidities were captured using the ICD-10 coding algorithm for the Charlson comorbidity index. Infectious conditions were captured using the groups of ICD-10 diagnostic codes that were carefully prepared by two independent infectious disease specialists. Accuracy of ICD-10 codes combined with microbiological data for diagnosis of urinary tract infection (UTI) and bloodstream infection (BSI) was evaluated. Clinical data gathered from chart review was considered the gold standard in this study.

Results: Between February 1 and May 31, 2013, a chart review of 546 hospitalization records was conducted. The mean age of hospitalized patients was 62.8±17.8 years and 65.9% of patients were female. Median length of stay [range] was 10.0 [1.0-353.0] days and hospital mortality was 21.8%. Conditions with ICD-10 codes that had good sensitivity (90% or higher) were diabetes mellitus and HIV infection. Conditions with ICD-10 codes that had good specificity (90% or higher) were cerebrovascular disease, chronic lung disease, diabetes mellitus, cancer, HIV infection, and all infectious conditions. By combining ICD-10 codes with microbiological results, sensitivity increased from 49.5 to 66% for UTI and from 78.3 to 92.8% for BSI.

Conclusion: The ICD-10 coding algorithm is reliable only in some selected conditions, including underlying diabetes mellitus and HIV infection. Combining microbiological results with ICD-10 codes increased sensitivity of ICD-10 codes for identifying BSI. Future research is needed to improve the accuracy of hospital administrative coding system in Thailand.

Keywords: Comorbidities, ICD-10, Infectious conditions