Assessing readmission databases: how reliable is the information?

Databases are being used to shape health care policy. However, the reliability of coding information entered into the databases may be difficult to validate. In this study, we assess readmission data from an institutional database that identified 1515 readmissions (708 patients) after total hip or total knee arthroplasty during a 5-year interval. After exclusions, 223 readmissions (190 patients) underwent medical record review. Bleeding, wound-related, and arthroplasty-related complications constituted most (62.8%) of readmissions. Bleeding and wound complications were nearly 6 times more frequently associated with readmission than venous thromboembolism events. On secondary review, there was discordance between the diagnosis obtained by a surgeon reviewer and coding for diagnoses consistent with periprosthetic infection (996.66, 77, 78, and 998.59) in 70% of cases. The findings of our study raise questions regarding the validity of accepting information obtained from larger databases without closer scrutiny.