Benchmarking variation in coding accuracy across the United States.

The objective of this study was to measure the consistency of coded medical data through information managers' reports of the overall coding error level in patients' medical records. Using a cross-sectional design, we examined the reported percent of records containing coding errors significant enough to change a diagnostic related group (DRG). Results indicate about 87 percent, 9 percent, and 5 percent of respondents reported that significant coding errors existed in less than 5 percent, 6-10 percent, and greater than 10 percent of the medical records in their institutions, respectively. Significant variation was found in the accuracy and consistency of coding practice and associated data quality across key demographic and organizational variables. Significantly large error rates in coded data exist in some organizations. Given variations across key demographic characteristics, providers may tend to distrust all coded data, when aggregated. As the United States moves toward an evidence-based medicine environment, the use of current patient data classification methods may be of limited value without increased attention to coding practices.