2442. Detection of Prosthetic Hip and Knee Joint Infections Using Administrative Databases – A Validation Study

BACKGROUND: Forming large cohorts to study prosthetic joint infections (PJIs) is a challenge without an existing surgical registry, as is the case in Canada. Administrative databases are an option, yet PJI diagnostic codes are insensitive. There is a need to improve the detection of PJIs from within administrative databases. METHODS: Individuals who had a primary arthroplasty at four hospitals in Toronto, Canada from 2010 to 2016 were identified using Canadian Classification of Health Intervention codes (based on the International Classification of Disease, Tenth Revision). Each re-admission to the same hospital until December 31, 2016 was reviewed for the presence of a PJI. The performance characteristics (sensitivity, specificity, positive and negative predictive values) of combinations of diagnostic and procedure codes when compared with the gold standard of chart review were calculated. The primary outcome was the algorithm that maximized sensitivity and positive predictive value. RESULTS: 27,843 primary arthroplasties were performed with 8595 readmissions, of which 572 involved a PJI. Median follow-up was 1258 days (interquartile range (IQR) 614–1891 days), with median time to first re-admission of 352 days (IQR range 166–725 days). PJI codes exhibited a sensitivity of 0.86 (95% confidence interval (95% CI) 0.83–0.89) and positive predictive value (PPV) of 0.89 (95% CI 0.86–0.92). The best performing algorithm is a combination of a PJI code or joint spacer insertion procedure code or insertion of a peripherally inserted central catheter along with an arthroplasty code (sensitivity 0.90, 95% CI 0.88–0.93 and PPV 0.89, 95% CI 0.86–0.91). Using timing from primary arthroplasty, spacer insertion codes and presence of a subsequent arthroplasty procedure code identified 68% (71/105) of first stage and 74% (108/146) of debridement with joint retention procedures during the first re-admission for a PJI. CONCLUSION: Combinations of diagnosis and procedure codes can reliably identify PJIs from administrative databases. Individual orthopaedic procedure codes and timing from primary arthroplasty can inform the surgical procedure performed. This PJI detection algorithm could be used for PJI surveillance and research. DISCLOSURES: All authors: No reported disclosures.