Predicting deep surgical site infection in patients receiving open posterior instrumented thoracolumbar surgery--- A-DOUBLE-SSI risk score: a large retrospective multicenter cohort study in China.

Abstract

To develop a practical prediction model to predict the risk of deep surgical site infection (SSI) in patients receiving open posterior instrumented thoracolumbar surgery. Data of 3,419 patients in 4 hospitals from Jan 1, 2012 to Dec 30, 2021 were evaluated. We used clinical knowledge-driven, data-driven and decision tree model to identify predictive variables of deep SSI. Forty-three candidate variables were collected, including 5 demographics, 29 preoperative, 5 intraoperative, and 4 postoperative variables. According to model performance and clinical practicability, the best model was chosen to develop a risk score. Internal validation was performed by using bootstrapping methods. After open posterior instrumented thoracolumbar surgery, 158 patients (4.6%) developed deep SSI. Clinical knowledge-driven model yielded 12 predictors of SSI, while data-driven and decision tree model produced 11 and 6 predictors, respectively. Knowledge-driven model, which had the best C- statistics [0.81 (95% CI: 0.78-0.85)] and superior calibration, was chosen due to its favorable model performance and clinical practicality. Moreover, twelve variables were identified in clinical knowledge-driven model, including age, BMI, diabetes, steroid use, albumin, duration of operation, blood loss, instrumented segments, powdered vancomycin administration, duration of drainage, postoperative CSF leakage, and early postoperative activities. In bootstrap internal validation, the knowledge-driven model still showed optimal C-statistics (0.79, 95% CI: 0.75-0.83) and calibration. Based on these identified predictors, a risk score for SSI incidence was created: the A-DOUBLE-SSI (Age, D [Diabetes, Drainage], O [duration of Operation, vancOmycin], albUmin, B [BMI, Blood loss], CSF Leakage, Early activities, Steroid use, and Segmental Instrumentation) risk score. Based on the A-DOUBLE-SSI score system, the incidence of deep SSI increased in a graded fashion from 1.06% (A-DOUBLE-SSIs score ≤8) to 40.6% (A-DOUBLE-SSIs score>15). We developed a novel and practical model, A-DOUBLE-SSIs risk score, that integrated easily accessible demographics, preoperative, intraoperative, and postoperative variables and could be used to predict individual risk of deep SSI in patients receiving open posterior instrumented thoracolumbar surgery.