International statistical classification of diseases and related health problems coding underestimates the incidence and prevalence of acute kidney injury and chronic kidney disease in general medical patients.

BACKGROUND: The international classification of diseases (ICD) code is frequently used to identify renal impairment in epidemiological research. However, Australian studies examining accuracy of this administrative data in coding kidney injury are lacking. AIMS: To compare the ICD 10 coding with the kidney disease: improving global outcomes (KDIGO) criteria in diagnosing acute kidney injury (AKI) and/or chronic kidney disease (CKD).
METHODS: A retrospective study of 325 patients admitted to general medicine during January 2012 was performed. Sensitivity and specificity of ICD 10 in identifying AKI and CKD were calculated using KDIGO as gold standard.
RESULTS: The sensitivities of ICD 10 in identifying AKI and CKD were low for both (59.5% and 54.1%), but the specificities were high (86.2% and 90.2%). Using KDIGO criteria, we identified 72 AKI (22%), 56 CKD (17%), 64 AKI on CKD (19%) and 133 controls (40%). Compared to the control group, patients with AKI and AKI on CKD had longer length of stay (3.2 vs 4.9 days and 3.2 vs 4.8 days, P = 0.20). Renal impairment groups had increased in-hospital mortality rate (5% control, 6% AKI, 10% CKD, 9% AKI on CKD) and re-admission rate within 30 days (13% control, 20% AKI, 25% CKD, 26% AKI on CKD). After adjusting for age, gender and comorbidities, the difference in outcomes was not statistically significant.
CONCLUSION: This study shows that ICD 10 fails to identify almost half of the patients with AKI (40.5%) and CKD (45.9%) in our cohort. A total of 60% had evidence of renal impairment as defined by KDIGO.