Gastric Surgical Site Infection Risk Prediction Model

Abstract

Background: Surgical site infections (SSIs) remain a common complication after an operation. Although most SSI can be treated by antibiotics, yet it has been shown to decrease health-related quality of life, increase the risk of readmission, and increase the costs of health care as well. Thus, we need to seek to wipe out or maintain its incidence rate as low as possible. As the strategies to make it happens, understanding the epidemiology and providing surgeons with appropriate risk factors become necessary.

Aim: Developed risk prediction model defining the patient with the high risk of pathogens infection in patients undergoing gastric surgery.

Methods: 4290 participants who underwent gastric surgery from July 2007 to December 2009 that successfully recorded and registered in KONIS, Korean Nosocomial Infections Surveillance System, were analyzed using lasso method to predict the emersion of SSI. Cross validation were applied in order to get tuning parameter value used in lasso process.

Results: Age, sex, NNIS Risk Index, multiple procedures in the same operation, re-operation at the same site, emergency, BMI, diabetes, as well as current smoking status were statistically significant factors for SSI after gastric surgery. Among them, re-operation was a factor that gave the largest contribution on the emergence of SSI with the probability alone about 0.14 or 8.8 times higher risk compared to non re-operation

followed by multiple procedures with probability 0.034. If high risk is defined as the probability larger than or equal to 0.33, thus when these both criteria were met, the risk would increase with probability about 0.20 which made the presence of re-operation and multiple procedures at once a kind of high-risk warning of getting infected after surgery. Moreover, when other additional factors were combined, the resulting risk would be even higher, with value in the range of 0.20 - 0.81. In terms of BMI, patients with BMI < 18.5 or 25 ≤ BMI < 30 showed no significant difference risk but patient with BMI ≥ 30 had higher risk as much as 25% compared to them who under/overweight.

Conclusion: Model building based on lasso problem is better than stepwise logistic regression and can produce a good and well calibrated risk prediction model on gastric SSI. This study shows that the emersion of gastric SSI is more affected by environmental/treatment factors, especially re-operation and multiple procedures, rather than host factors. Therefore, the surgeons are expected to be more careful in patient selection, preparation and medical care provision.