Background: This study examined the degree of accuracy of billing data in an academically affiliated family practice.
Methods: The progress notes from 1253 consecutive visits were independently reviewed by two family physicians, and the diagnoses, use of procedures, and level of service were determined for each visit. Discrepancies between the reviewers were resolved by consensus. These data were compared with the data on the corresponding billing form that had been completed by the care providers (ie, physicians on the faculty, physicians in training, family nurse practitioners, and nurses).
Results: There was poor agreement between the billing form and progress note on level of service and number of diagnoses (kappa = 0.37 and kappa = 0.28, respectively). The progress note usually indicated that a higher level of service should have been billed for a visit than actually was billed. Underreporting of the number of diagnoses was substantial; the billing forms listed only 69% of the diagnoses identified in the progress notes. In 60% of visits, each diagnosis on the billing form had a matching diagnosis in the progress note. This could be improved to 78% of visits if broad categories of disease were used. Residents were similar to faculty in the accuracy of reported level of service and types of diagnoses, but were more likely to underreport the number of diagnoses.
Conclusions: Ambulatory care data from computerized billing files may not be sufficiently accurate for proper reimbursement of physician services or for use in research.