Early prediction of 30-day mortality in patients with surgical wound infections following cardiothoracic surgery: Development and validation of the SWICS-30 score utilizing conventional logistic regression and artificial neural network

Abstract

Introduction

We aimed to create and validate the 30-day prognostic score for mortality in patients with surgical wound infection (SWICS-30) after cardiothoracic surgery.

Methods

This retrospective study enrolled patients with surgical wound infection following cardiothoracic surgery admitted to a Cardiologic Reference Center Hospital between January 2006 and January 2023. Clinical data and commonly used blood tests were analyzed at the time of diagnosis. An independent scoring system was developed through logistic regression analysis and validated using Artificial intelligence.

Results

From 1713 patients evaluated (mean age of 60 years (18–89), 55 % female), 143 (8.4 %) experienced 30-day mortality. The SWICS-30 logistic regression score comprised the following variables: age over 65 years, undergoing valve heart surgery, combined coronary and valve heart surgery, heart transplantation, time from surgery to infection diagnosis exceeding 21 days, leukocyte count over 13,000/mm3, lymphocyte count below 1000/mm3, platelet count below 150,000/mm3, and creatinine level exceeding 1.5 mg/dL. These patients were stratified into low (2.7 %), moderate (14.2 %), and high (47.1 %) in-hospital mortality risk categories. Artificial intelligence confirmed accuracy at 90 %.