(171) Factors affecting readmission in patients with surgical site infection: A graphical and prediction model-based approach

Publication Number: 506

Background: Antimicrobial therapy is becoming ineffective as the microorganisms gain resistance. Surgical site infections (SSI) can cause a major change in the infection control policy during the patient’s hospital stay.

Objectives: To develop a prediction model for readmission after being diagnosed with SSI and to identify the major microbial isolates, with their trend of resistance.

Methods: This was a single-centred retrospective study carried out in a tertiary healthcare setting in India. 549 patients were diagnosed with SSI from January 1, 2016, to August 25, 2021, in ortho surgery (n=373), general surgery(n=135), and neurosurgery(n=41) departments. Patients’ demographics and microbial isolate data were collected using a case record form. Purposeful selection of covariates (P≤0.25) was used to identify the predictors of readmission. The model fit was validated using the omnibus test. Collinearity statistic was also used to eliminate non-significant variables. A receiver operating characteristic curve with the area under the curve (AUC) was utilised to assess the model discrimination. A cutoff plot at 0.50 was used to represent the sensitivity and specificity of the model. The resistance trend of microbial isolates was graphically represented.

Results: Of the 549 patients diagnosed with SSI, 137(24.9%) were readmitted due to repeated infections in the previously treated sites. A mean span of 152 ± 32 days from the surgery date to the recent readmission was estimated. Univariate logistic regression showed 40 significant variables (P≤ 0.25). The multivariate logistic regression eliminated three variables (surgical procedure time, diaminopyramidines used for SSI treatment, and Unknown characteristics for hospital admissions) due to insufficient comparator levels. Collinearity statistics excluded two variables based on VIF > 10 (peripheral surgical site and reconstruction surgeries). The model showed an AUC of 0.772 and an accurate prediction of 77.8% (AIC:568; BIC:722). 15 types of microorganisms were isolated from 75.4% of the readmitted patients. Methicillin-resistant Staphylococcus aureus (MRSA) (23.8%) was the primary isolate that showed resistance towards cloxacillin and ciprofloxacin/ofloxacin (25.69%), equally followed by erythromycin (18.4%) and gentamycin (6.25%).

Conclusions: MRSA-associated infections had a more significant influence on the readmissions in the included patients. The microbial data was not included in the prediction model, as more significant factors could give better predictability. This study can aid clinicians in detecting patients at risk for readmission with surgical site infections.