Diagnosis of infection after cardiovascular surgery (DICS): a study protocol for developing and validating a prediction model in prospective observational study.

Abstract

Postoperative infection (PI) is one of the main severe complications after cardiovascular surgery. Therefore, antibiotics are routinely used during the first 48 hours after cardiovascular surgery. However, there is no effective method for early diagnosis of infection after cardiovascular surgery, particularly, to determine whether postoperative patients need to prolong the use of antibiotics after the first 48 hours. In this study, we aim to develop and validate a diagnostic model to help identify whether a patient has been infected after surgery and guide the appropriate use of antibiotics. In this prospective study, we will develop and validate a diagnostic model to determine whether the patient has a bacterial infection within 48 hours after cardiovascular surgery. Baseline data will be collected through the electronic medical record system. A total of 2700 participants will be recruited (n=2000 for development, n=700 for validation). The primary outcome of the study is the newly PI during the first 48 hours after cardiovascular surgery. Logistic regression penalised with elastic net regularisation will be used for model development and bootstrap and k-fold cross-validation aggregation will be performed for internal validation. The derived model will be also externally validated in patients who are continuously included in another time period (N=700). We will evaluate the calibration and differentiation performance of the model by Hosmer-Lemeshow good of fit test and the area under the curve, respectively. We will report sensitivity, specificity, positive predictive value and negative predictive value in the validation data-set, with a target of 80% sensitivity. Ethical approval was obtained from Medical Ethics Committee of Affiliated Nanjing Drum Tower Hospital, Nanjing University Medical College (2020-249-01). Chinese Clinical Trial Register (www.chictr.org.cn, ChiCTR2000038762); Pre-results.