308. Identification of Prosthetic Hip and Knee Joint Infections in Administrative Databases
Session: Poster Abstract Session: Bone and Joint Infections
Thursday, October 4, 2018
Room: S Poster Hall
Background: Canada lacks a prosthetic hip and knee joint infection (PJI) registry, leaving active surveillance to be orchestrated by individual hospitals, which is limited by cost and narrow scope. Administrative databases are potentially an ideal instrument for infection surveillance, but detection algorithms relying solely on PJI diagnostic codes alone have been hampered by low specificity. There is a need to develop improved strategies to efficiently and accurately identify PJIs using health administrative databases.

Methods: Combinations of International Classification of Disease, Tenth Revision, diagnostic and procedure codes were used to create testing cohorts among individuals treated at two institutions in Toronto, Ontario from April 1, 2015 until March 31, 2016. These cohorts were compared to a reference standard of PJIs, which were identified by chart reviews of every individual who underwent a hip or knee revision operation at these institutions during the study period. The primary outcomes were the performance characteristics of each algorithm.

Results: Over the one year study period there were 471 revision operations for 405 patients, of which 155 (33%) were performed for the treatment of a PJI. Of the 405 individuals, 108 (27%) had a PJI as the surgical indication; there were 57 (53%) two-stage procedures, 9 (8%) single-stage procedures, 34 (31%) incision and drainage procedures with implant retention, and 8 (7%) excisional arthroplasties. The combination of a revision operation code plus a PJI diagnosis code was the most robust detection method: sensitivity 0.86 (95% confidence interval, 0.79-0.91) and specificity 0.99 (0.98-1.00). Coupling codes for a revision operation and insertion of a peripherally inserted central catheter yielded a sensitivity of 0.45 (0.37-0.53) and specificity of 1.00 (0.98-1.00). PJI codes alone had a sensitivity of 1.00 (0.86-1.00) and specificity 0.50 (0.23-0.77).

Conclusion: The combination of a revision operation procedure code and a PJI diagnosis code is sensitive and specific for the detection of a PJI in administrative databases. This is a promising avenue for national PJI surveillance and has the potential to facilitate future research in the prevention and management of PJIs.

Christopher Kandel, MD, Department of Infectious Diseases, University of Toronto, Toronto, ON, Canada, J. Roderick Davey, MD, Department of Surgery, University of Toronto, Toronto, ON, Canada, Nick Daneman, MD, MSc, Division of Infectious Diseases & Clinical Epidemiology, University of Toronto, Toronto, ON, Canada and Allison McGeer, MD, MSc, Laboratory Medicine and Pathobiology, University of Toronto, Toronto, ON, Canada


C. Kandel, None

J. R. Davey, None

N. Daneman, None

A. McGeer, None

Findings in the abstracts are embargoed until 12:01 a.m. PDT, Wednesday Oct. 3rd with the exception of research findings presented at the IDWeek press conferences.