Administrative database concerns: accuracy of International Classification of Diseases, Ninth Revision coding is poor for preoperative anemia in patients undergoing spinal fusion

Cross-sectional study.

To objectively evaluate the ability of International Classification of Diseases, Ninth Revision (ICD-9) codes, which are used as the foundation for administratively coded national databases, to identify preoperative anemia in patients undergoing spinal fusion.

National database research in spine surgery continues to rise. However, the validity of studies based on administratively coded data, such as the Nationwide Inpatient Sample, are dependent on the accuracy of ICD-9 coding. Such coding has previously been found to have poor sensitivity to conditions such as obesity and infection.

A cross-sectional study was performed at an academic medical center. Hospital-reported anemia ICD-9 codes (those used for administratively coded databases) were directly compared with the chart-documented preoperative hematocrits (true laboratory values). A patient was deemed to have preoperative anemia if the preoperative hematocrit was less than the lower end of the normal range (36.0% for females and 41.0% for males).

The study included 260 patients. Of these, 37 patients (14.2%) were anemic; however, only 10 patients (3.8%) received an "anemia" ICD-9 code. Of the 10 patients coded as anemic, 7 were anemic by definition, whereas 3 were not, and thus were miscoded. This equates to an ICD-9 code sensitivity of 0.19, with a specificity of 0.99, and positive and negative predictive values of 0.70 and 0.88, respectively.

This study uses preoperative anemia to demonstrate the potential inaccuracies of ICD-9 coding. These results have implications for publications using databases that are compiled from ICD-9 coding data. Furthermore, the findings of the current investigation raise concerns regarding the accuracy of additional comorbidities. Although administrative databases are powerful resources that provide large sample sizes, it is crucial that we further consider the quality of the data source relative to its intended purpose.