Plastic Surgery Research Council
Members Only  |  Contact  |  PSRC on Facebook
PSRC 60th Annual Meeting

Back to Annual Meeting Program


The Bra Score: Creating A General Risk Calculator For Breast Reconstruction Outcomes
Nima Khavanin, BS1, John Y.S. Kim, MD1, Armando A. Davila, MD2, Jon P. Ver Halen, MD3, Alexei S. Mlodinow, BA1, Kevin P. Bethke, MD1, Seema A. Khan, MD1, Jacqueline S. Jeruss, MD1, Nora M. Hansen, MD1, Karl Y. Bilimoria, MD1, Albert Losken, MD4, Neil A. Fine, MD1.
1Northwestern University Feinberg School of Medicine, Chicago, IL, USA, 2Loma Linda University, Loma Linda, CA, USA, 3Baptist Cancer Center / Vanderbilt Ingram Cancer Center, Memphis, TN, USA, 4Emory University Hospital, Atlanta, GA, USA.

Purpose: With over 90,000 prosthetic and autologous reconstructions each year, there have been many studies aimed at identifying risk factors associated with a perioperative surgical site infection. None, however, are sufficiently powered to provide an objective measure of an individual patient’s pre-operative risk for infection following the various reconstruction procedures. With data from over 300 surgical centers across the United States, the American College of Surgeon’s National Surgical Quality Improvement Program (NSQIP) registry provides high powered, validated data ideal for modeling a patient’s risk for perioperative complication. We aimed to develop a validated Breast Reconstruction Risk Assessment (BRA score) calculator and to assess a patient’s risk for post-operative infection across the various reconstruction modalities following mastectomy with immediate reconstruction.
Methods: Patients undergoing mastectomy with immediate prosthetic (n=12,612) or autologous (n=3,457) reconstruction between 2005 and 2011 were identified from the NSQIP database. Forward stepwise multiple logistic regression identified preoperative variables for inclusion in the model. Hosmer-Lemeshow and concordance statistics were computed to assess model calibration and discrimination. Overall model performance was evaluated with the Brier score. Bootstrap analysis was used to validate the model. Chi-squared tests were used to determine the influence of infection on 30-day readmission and reoperation. The validated model was used to develop an interactive risk calculator that accepts patient information and returns an estimated surgical site infection probability based on the logistic regression model. Predicted probabilities were calculated from the logistic function: Probability=1/(1 + e), where β is the summation of the model constant and the relevant covariates for a given patient. The web based risk calculator is available at www.brascore.org.
Results: The overall infection rate was 3.75 (603 out of 16,469 patients). The infection rate was greatest within the pedicled TRAM cohort (5.97%), followed by the free flap cohort (5.52%), prosthetic cohort (3.44%), and finally the latissimus cohort (2.80%). In addition to reconstructive modality, 5 predictors of infection were selected for inclusion within the model: BMI, age, ASA class, bleeding disorder, and history of percutaneous cardiac intervention or cardiac surgery. The model c-statistic was 0.682 and the optimism-corrected c-statistic 0.678. The model was well calibrated (HL p-value = 0.371) and the brier score was 0.036. Across all reconstructive modalities, patients with an infection experienced higher rates of reoperation (range of 38.1%-45.9% vs. 4.8%-13.5%) and readmission (50.9%-61.1% vs. 2.8%-5.3%) (all p < 0.001).
Conclusions: In this era of increasing emphasis on evidence-based decision making, there has been a reliance on population-based assessments of risk. Through this analysis of a large, validated multi-institutional database, we developed a modifiable risk calculator (the BRA score) to determine individual risk of outcomes following breast reconstruction. Applying the BRA score to surgical site infections, we found a gradient of risk among the common forms of reconstruction. While the precise manifestation of this BRA score can vary by database and measured outcome, it is a potentially useful educational construct to better manage physician and patient expectations.


Back to Annual Meeting Program