Discussion

Comprehensive manual IPC surveillance is “gold standard” for infections after CIED implantation and is considered a core component of many IPC programs ^{Reference Storr, Twyman and Zingg7} however it is human resource intensive and costly. When it is not possible to do this level of surveillance, administrative data may offer a novel alternative. A provincial IPC committee in Alberta will now adapt this and trial it as a mechanism of province wide surveillance demonstrating the successful operationalization of a policy change for IPC surveillance.

The American Heart Association has a national CIED infection initiative aimed at addressing gaps in awareness of detection and treatment of CIED infections. ⁵ This initiative includes a national CIED infection summit and a national education plan to provide guidelines, educational resources and care pathways. ⁵ Using administrative data as a mechanism for IPC surveillance can complement these other initiatives to reduce CIED infections through timely identification of trends in CIED infections allowing for investigation and intervention when necessary.

This policy change has several strengths. Firstly, this type of surveillance can largely be automated, thus reducing human resource requirements and cost. Exploring ways to adopt this type of policy change is necessary given that as health systems continue to enhance their use of automation and artificial intelligence, including in IPC, ^{Reference Wiemken and Carrico14} there will be increased need for collaboration and processes to support change. This policy will improve health equity as it can be available at all sites offering CIED implantation even if there are not the resources for wide scale manual CIED infection surveillance on an ongoing basis. This type of work can also be used as a blueprint for other types of infection surveillance (eg, neurosurgery, cardiac surgery, orthopedic surgery) where comprehensive traditional surveillance is not possible due to limited human resources. This may lead to enhanced collaboration between IPC and different surgical services and offer opportunities for patient safety reviews and improvements in patient outcomes.

This work is limited in that prior to this being adapted to other types of surgery there would need to be a comprehensive robust evaluation and validation work done on the administrative data to ensure accuracy in the codes at identifying infection. Additionally, validation work should be done in alternative regions outside of Alberta to ensure the reliability and generalizability. Unlike jurisdictions with diagnosis-linked reimbursement, Alberta hospitals are not paid per diagnosis code; rather, coded discharge abstracts (including diagnosis and procedure codes) are a mandatory component of provincial and national hospital reporting creating a structural incentive for complete and standardized coding independent of any direct reimbursement link. In jurisdictions where standardized diagnostic coding is not similarly mandated or systematically audited (eg, systems relying on optional or fee-for-service-only coding) the sensitivity and completeness of an administrative-data algorithm such as this one may be reduced, and local validation would be needed before adoption. This model relies on provincewide linkage of administrative data across a single, centralized health system covering >99% of residents, which is not universally available. In more fragmented healthcare systems (eg, those with multiple competing payers, non-interoperable electronic health records, or incomplete hospital-level reporting) the completeness of the denominator (CIED implantations) and numerator (infection-coded encounters) would need to be assessed.

Another potential limitation of this approach is the inherent latency of administrative data. CIED implantation and infection-related hospital encounters typically become available in the administrative data extracts only after a reporting lag of approximately 12 weeks. However, as traditional IPC surveillance is typically not reported more than quarterly this is not anticipated to create a substantial time delay. Finally, this type of surveillance is useful largely for monitoring trends as infection rates will be overestimated due to their uncommon occurrence. While the ICD-10 algorithm used in this work has excellent sensitivity (90.6%) and specificity (98.6%) against a chart-review gold standard, its positive predictive value is considerably lower (32.5%), reflecting the low baseline prevalence of CIED infection. ^{Reference Rennert-May, Leal and MacDonald9} In practical terms, this means that when the algorithm flags a case as a possible complex CIED infection, it will be a true infection roughly one-third of the time, and IPC teams reviewing flagged cases for confirmation should expect a corresponding chart-review workload of approximately two false positives for every confirmed true positive (given overall low infection rates this is not anticipated to be a substantial workload burden). This is an important operational consideration for sites adopting Policy Option 2, where flagged cases are intended to be confirmed by IPC chart review rather than acted upon directly, and is consistent with our stated position that trend-level monitoring (rather than case-level diagnostic certainty) is the appropriate use of this administrative data approach. Increases in trends would still suggest an overall increase in infection and trigger the necessary IPC steps to identify and reduce infection risk.

To move this policy forward, engagement between academic researchers, clinicians, and operational decision makers was necessary. Evidence around implementation of policy suggests that successful engagement and collaboration between academics and knowledge users/decision makers requires research that is relevant and has clear recommendations, as well as being aligned with strategic goals of the decision makers. ^{Reference Oliver, Innvar, Lorenc, Woodman and Thomas15} This supports policy adoption and knowledge translation. Additionally, researchers should be familiar with the policy making process and the knowledge users, and ensure availability for communication and questions. ^{Reference Oliver, Innvar, Lorenc, Woodman and Thomas15,Reference Otten, Dodson, Fleischhacker, Siddiqi and Quinn16} In the current work, these conditions were met by involving members of the research team who would also be some of the knowledge users of the policy change as well as including an IPC research scientist who was able to bridge the connection between the academic team and operational decision makers. While replicating these relationships may not be practical for every type of policy implementation it is critical to consider how to optimize these factors when trying to enact policy change. ^{Reference Jessani, Siddiqi, Babcock, Davey-Rothwell, Ho and Holtgrave17}

Administrative data may offer an alternative to comprehensive IPC surveillance for CIED infections, or a hybrid option that can be used to supplement traditional surveillance. This can allow for the tracking of infection rates and careful foundational work has led to an evidence-informed policy decision change to use this for the province of Alberta. Future studies should endeavor to apply this work to other types of surgical infections, and our group is currently exploring using administrative data to track CIED infections nationally across Canada which will demonstrate widespread generalizability of this novel mechanism for IPC surveillance.