Validation of a Risk Model for Surgical Site Infections After Breast Reconstruction
Frederick Wang, M.D., M.A.S.1, Natalie Hall, B.S.2, Hani Sbitany, M.D.1.
1University of California, San Francisco, San Francisco, CA, USA, 2Alpert Medical School of Brown University, Providence, RI, USA.
PURPOSE:The Breast Reconstruction Risk Assessment (BRA) Score is a model developed to predict the risk of surgical site infections (SSI) within 30 days after breast reconstruction. This was derived from the National Surgical Quality Improvement Program (NSQIP) database, where the average risk of SSI was 3.75%. We aim to externally validate this model in our cohort of patients who have undergone total skin-sparing mastectomy (TSSM) with immediate breast reconstruction.
We reviewed all cases of TSSM with immediate breast reconstruction between 2005-2013 performed at our institution. SSIs in our cohort were defined as any infections requiring oral antibiotics, intravenous antibiotics, or procedures for resolution. The BRA Score variables included type of reconstruction, age, BMI, ASA class 3+, bleeding disorder, prior PCI or cardiac surgery, diabetes, active smoking, dyspnea, and hypertension. We compared the BRA Score model with models derived from our own dataset. We generated a new model using variables associated with complications after breast reconstruction, including age, BMI, diabetes, incision, lymph node dissection, prior radiation, and acellular dermal matrix.
We identified 746 patients who had undergone TSSM with immediate breast reconstruction, with an overall 30-day SSI risk of 5.2%. The 1-year SSI risk was 16.4% in the 684 patients who underwent prosthetic breast reconstruction. The BRA Score for 30-day SSI did not fit our data well (C-statistic 0.509). We generated a risk prediction model using specific variables associated with breast reconstruction, and ten-fold cross-validation yielded improved discrimination for overall 30-day risk of SSI (C-statistic 0.583) and 1-year SSI risk in prosthetic breast reconstruction (C-statistic 0.625).
Many risk prediction models have been published, but few are used clinically due to lack of validation or relevance to the population studied. The BRA Score model does not incorporate several important variables associated with breast reconstruction outcomes, such as lymph node dissection status, prior radiation, and the use of acellular dermal matrix. Our new model with these variables achieves better discrimination with our data but requires external validation. An improved SSI risk model could be developed using a larger cohort, with variables specific to breast reconstruction.
|Reconstruction||30-day SSI||1-year SSI|
|Prosthetic (684)||35 (5.1%)||112 (16.4%)|
|Pedicle TRAM (36)||1 (2.8%)||n/a|
|Free flap (22)||3 (12%)||n/a|
|Latissimus (1)||0 (0%)||n/a|
Back to 2017 Posters