BACKGROUND - Hospital databases contain vital demographic patient information, which is increasingly being used as a basis to dictate care. It is hypothesized that the validity of data administratively generated from such sources is suboptimal, especially for rare subspecialties. The authors examined three databases to determine their concordance in an academic orthopaedic oncology subspecialty practice.
METHODS - A 2-year retrospective review was performed on three databases searching for seven fundamental variables: additions/deletions; identification number; birthdate; procedure date; admit/discharge date; procedure code; and diagnostic code. Two university-maintained hospital databases (medical records and physician billing) were compared to the surgeon's personal handwritten daily log, which served as the "gold standard."
RESULTS - All seven variables were in agreement with the physician's log in only 60% of the medical records and 61% of the physician billing patient entries (n = 564). On more detailed statistical analysis using chi(2), cross tabulations, and the K statistic for interobserver agreement, it was determined that poor concordance exists among the databases.
CONCLUSION - Surgeons delivering quartenary care should maintain his or her own database because the hospital's information often differs on one or more important variables. Further investigation into the accuracy of hospital databases regarding commonly practiced medical disciplines appears warranted.