A clinical prediction model for predicting the surgical site infection after an open reduction and internal fixation procedure considering the NHSN/SIR risk model: a multicenter case-control study

Abstract

Introduction: Surgical site infection (SSI) is one of the most common surgical-related complications worldwide, particularly in developing countries. SSI is responsible for mortality, long hospitalization period, and a high economic burden.

Method: This hospital-based case-control study was conducted in six educational hospitals in Tehran, Iran. A total of 244 patients at the age of 18-85 years who had undergone open reduction and internal fixation (ORIF) surgery were included in this study. Among the 244 patients, 122 patients who developed SSIs were selected to be compared with 122 non-infected patients used as controls. At the second stage, all patients (n = 350) who underwent ORIF surgery in a hospital were selected for an estimation of the standardized infection ratio (SIR). A logistic regression model was used for predicting the most important factors associated with the occurrence of SSIs. Finally, the performance of the ORIF prediction model was evaluated using discrimination and calibration indices. Data were analyzed using R.3.6.2 and STATA.14 software.

Results: Klebsiella (14.75%) was the most frequently detected bacterium in SSIs following ORIF surgery. The results revealed that the most important factors associated with SSI following an ORIF procedure were found to be elder age, elective surgery, prolonged operation time, American Society of Anesthesiologists score of ≥2, class 3 and 4 wound, and preoperative blood glucose levels of >200 mg/dl; while preoperative higher hemoglobin level (g/dl) was found to be a protective factor. The evidence for the interaction effect between age and gender, body mass index and gender, and age and elective surgery were also observed. After assessing the internal validity of the model, the overall performance of the models was found to be good with an area under the curve of 95%. The SIR of SSI for ORIF surgery in the selected hospital was 0.66 among the patients aged 18-85 years old.

Conclusion: New risk prediction models can help in detecting high-risk patients and monitoring the infection rate in hospitals based on their infection prevention and control programs. Physicians using prediction models can identify high-risk patients with these factors prior to ORIF procedure.

Keywords: ORIF surgery; nosocomial infection; prediction model; standardized infection ratio; surgical site infection.

Figure 1

Nomogram used for the estimation of probability of SSIs following ORIF procedure based on predictors of ORIF model considering incidence outcome (SSI) of 8% in the population. Based on nomogram, each variable was given a 0–100 value that can be found to the “points” bar. The sum of these points corresponds to the “total points” bar. Total points for each patient refer to a predicted probability of SSI based on the ORIF model.

Figure 2

Box plot and discrimination slope of model in train (development) and test data (validation) using split-sample validation. The discrimination slopes are estimated as the difference in means of predictors for those with and without SSI following ORIF procedure. These plots show how a prediction model can discriminate those with and without the SSI.

Figure 3

Calibration plot of ORIF model in train (development) and test data (validation) using split-sample validation. Calibration plot shows agreement of the predicted probability using model with the observed proportion of SSI (nearing of observed SSI to the 45° line) and shows clinical usefulness using the observed number of predictors above or below thresholds (arrow).