• John W Peabody, Jeff Luck, Sharad Jain, Dan Bertenthal, and Peter Glassman.
• San Francisco Veterans Affairs Medical Center, San Francisco, California, USA. peabody@psg.ucsf.edu
• Med Care. 2004 Nov 1;42(11):1066-72.

BackgroundAdministrative data play a central role in health care. Inaccuracies in such data are costly to health systems, they obscure health research, and they affect the quality of patient care.ObjectivesWe sought to prospectively determine the accuracy of the primary and secondary diagnoses recorded in administrative data sets.Research DesignBetween March and July 2002, standardized patients (SPs) completed unannounced visits at 3 sites. We abstracted the 348 medical records from these visits to obtain the written diagnoses made by physicians. We also examined the patient files to identify the diagnoses recorded on the administrative encounter forms and extracted data from the computerized administrative databases. Because the correct diagnosis was defined by the SP visit, we could determine whether the final diagnosis in the administrative data set was correct and, if not, whether it was caused by physician diagnostic error, missing encounter forms, or incorrectly filled out forms.SubjectsGeneral internal medicine outpatient clinics at 2 Veterans Administration facilities and a large, private medical center participated in this study.MeasuresA total of 45 trained SPs presented to physicians with 4 common outpatient conditions.ResultsThe correct primary diagnosis was recorded for 57% of visits. Thirteen percent of errors were caused by physician diagnostic error, 8% to missing encounter forms, and 22% to incorrectly entered data. Findings varied by condition and site but not by level of training. Accuracy of secondary diagnosis data (27%) was even poorer.ConclusionsAlthough more research is needed to evaluate the cause of inaccuracies and the relative contributions of patient, provider, and system level effects, it appears that significant inaccuracies in administrative data are common. Interventions aimed at correcting these errors appear feasible.

26 keywords...

hide…