Characterizing infections in prosthetic breast reconstruction: A validity assessment of national health databases

Introduction: Current guidelines in the United States require reporting only the 30-day postoperative outcomes to standardized databases, including the National Surgical Quality Improvement Program (NSQIP). Thus, many breast implant-related complications go unreported in standard databases. We sought to characterize late periprosthetic infections following implant-based breast reconstruction. Methods: We conducted a retrospective analysis of all women who underwent expander/implant reconstruction from 2005 to 2014 at two institutions. All periprosthetic infections were identified and divided into early and late cohorts (<= 30 days or > 30 days). Infection was defined as any episode where antibiotics were initiated or a prosthetic device was explanted because of clinical evidence of the infection. Results: In the 1820 patients (2980 breasts) identified, 421 periprosthetic infections occurred (14%). Of these, 173 (41%) were early and 248 (59%) were late (mean time to infection = 66.4 +/- 101.9 days). Patients with late infections were more likely to be current smokers or have diabetes than patients with early infections (p < 0.034 for both). Infections caused by gram-negative bacteria and antimicrobial-resistant strains of Staphylococcus were more common in the early infection group (p < 0.001 for both). Implant loss due to infection was more common in the late infection group (p = 0.037). Discussion: Late periprosthetic infections following implant-based breast reconstruction are underestimated in national outcome databases and have unique risk factors and microbiology compared to early infections. A system-level change in reevaluating and redefining a timeline for tracking and treating implant infections is necessary, given the substantial morbidity associated with, and frequency of, late periprosthetic infections. (C) 2017 British Association of Plastic, Reconstructive and Aesthetic Surgeons. Published by Elsevier Ltd. All rights reserved.