Billing codes in big data often misidentify diagnoses

Use of billing codes in big data sets to find diagnoses can result in up to two-thirds of cases being mistakenly identified, new UCLA-led research finds.

Databases frequently used for medical research such as those for the Centers for Medicare & Medicaid Services or the National Inpatient Survey typically rely on ambulatory billing codes to identify diseases or medical procedures, but their accuracy is rarely verified in publications that rely on this data, the researchers write in a report published in the peer-reviewed journal British Journal of Surgery.

Though the findings in this paper focused on hernia diagnoses, reliance on billing codes in research reports can lead to similar discrepancies with other diseases or conditions, said Dr. Edward Livingston, health sciences professor of surgery at the David Geffen School of Medicine at UCLA and the research letter's senior author.

Researchers often assume that if a code appears for a certain diagnosis in one of these big data sets that the disease truly is present. Our research demonstrates that is not the case in many instances. Research relying on these codes to identify diseases may lead to false conclusions because of this problem."

Dr. Edward Livingston, health sciences professor of surgery, David Geffen School of Medicine, UCLA

The researchers examined records for 1.36 million patients, of whom 41,700 were diagnosed with hernias based on the coding-12,800 (45%) with diaphragmatic hernias, 7000 (24%) with ventral hernias, and 8,800 (31%) with inguinal hernias.

But the researchers had corresponding images for 28,600 of code-based diagnosed patients. Of those, the images verified that 10,234 (36%) actually had hernias; 4,325 (34%) diaphragmatic hernias, 3,069 (44%) were ventral hernias, and 2,840 (32%) were inguinal hernias.

The researchers suspect that the discrepancy stems from physicians basing their coding on the clinical problem for which they initially examined the patient and not on what they ultimately found. For instance, a patient visit for a possible hernia will be coded as a hernia in the record and remains that way even if that initial diagnosis is ruled out during subsequent examination.

"These findings highlight a fundamental weakness in using administrative data for disease identification," the researchers write. "Encounter coding occurs because a diagnosis is considered, and not necessarily proven. We found that reliance on billing codes for hernia identification could result in 2/3 of cases being erroneously identified. This issue extends beyond hernia, highlighting a serious limitation in using administrative data for clinical research. Validation of coding accuracy against actual disease presence is essential before assuming diagnosis validity."

Hila Zilicha, Dr. Douglas Bell, and Dr. Yijun Chen co-authored the paper.

The UCLA Department of Surgery Research Funds and the National Center for Advancing Translational Science (NCATS) of the National Institutes of Health under the UCLA Clinical and Translational Science Institute (UL1TR001881) funded the research.