Preoperative risk prediction of surgical site infection requiring hospitalization or reoperation in patients undergoing vascular surgery

Machine-Learning Models for Predicting Surgical Site Infections using Patient Pre-Operative Risk and Surgical Procedure Factors

BACKGROUND

Surgical site infections (SSIs) are a significant healthcare problem as they can cause increased medical costs and increased morbidity and mortality. Assessing a patient's pre-operative risk factors can improve risk stratification and help guide the surgical decision-making process. Previous efforts to use pre-operative risk factors to predict the occurrence of SSIs have relied upon traditional statistical modeling approaches. The aim of this paper is to develop and validate, using state-of-the-art machine learning (ML) approaches, classification models for the occurrence of SSI to improve upon previous models.

METHODS

In this work, using the American College of Surgeons' National Surgical Quality Improvement Program (ACS NSQIP) database, the performances (e.g., prediction accuracy) of seven different ML approaches (Logistic Regression (LR), Naïve Bayesian (NB), Random Forest (RF), Decision Tree (DT), Support Vector Machine (SVM), Artificial Neural Network (ANN), and Deep Neural Network (DNN)) were compared. The performance of these models was evaluated using the area under the curve, accuracy, precision, sensitivity, and F1-score metrics.

RESULTS

Overall, 2,882,526 surgical procedures were identified in the study for the SSI predictive models' development. The results indicate that the DNN model offers the best predictive performance with 10-fold compared to the other six approaches considered (area under the curve = 0.8518, accuracy = 0.8518, precision = 0.8517, sensitivity = 0.8527, F1-score = 0.8518). Emergency case surgeries, American Society of Anesthesiologists (ASA) Index of 4 (ASA_4), BMI, Vascular surgeries, and general surgeries were most significant influencing features towards developing an SSI.

CONCLUSION

Equally important is that the commonly used LR approach for SSI prediction displayed mediocre performance. The results are encouraging as they suggest that the prediction performance for SSIs can be improved using modern ML approaches.

Groin Wound Infection after Vascular Exposure (GIVE) Risk Prediction Models: Development, Internal Validation, and Comparison with Existing Risk Prediction Models Identified in a Systematic Literature Review

OBJECTIVE

This study aimed to develop and internally validate risk prediction models for predicting groin wound surgical site infections (SSIs) following arterial intervention and to evaluate the utility of existing risk prediction models for this outcome.

METHODS

Data from the Groin wound Infection after Vascular Exposure (GIVE) multicentre cohort study were used. The GIVE study prospectively enrolled 1 039 consecutive patients undergoing an arterial procedure through 1 339 groin incisions. An overall SSI rate of 8.6% per groin incision, and a deep/organ space SSI rate of 3.8%, were reported. Eight independent predictors of all SSIs, and four independent predictors of deep/organ space SSIs were included in the development and internal validation of two risk prediction models. A systematic search of the literature was conducted to identify relevant risk prediction models for their evaluation.

RESULTS

The "GIVE SSI risk prediction model" ("GIVE SSI model") and the "GIVE deep/organ space SSI risk prediction model" ("deep SSI model") had adequate discrimination (C statistic 0.735 and 0.720, respectively). Three other groin incision SSI risk prediction models were identified; both GIVE risk prediction models significantly outperformed these other risk models in this cohort (C statistic 0.618 - 0.629; p < .050 for inferior discrimination in all cases).

CONCLUSION

Two models were created and internally validated that performed acceptably in predicting "all" and "deep" groin SSIs, outperforming current existing risk prediction models in this cohort. Future studies should aim to externally validate the GIVE models.

The association between preoperative length of stay and surgical site infection after lower extremity bypass for chronic limb-threatening ischemia

OBJECTIVE

Surgical site infection (SSI) is an important complication of lower extremity bypass (LEB) and the rate of SSI after LEB varies widely in the existing literature, ranging from 4% to 31%. Prolonged length of stay (LOS) has been implicated in the occurrence of SSI across multiple surgical disciplines. The impact of preoperative LOS in patients with chronic limb-threatening ischemia (CLTI) undergoing LEB is unknown. We examined the association of preoperative LOS on SSI after LEB.

METHODS

A retrospective analysis of the Society for Vascular Surgery Vascular Quality Initiative Infrainguinal Bypass Registry identified patients undergoing elective LEB for chronic limb-threatening ischemia from 2003 to 2019. Patients undergoing LEB for acute limb ischemia, urgent/emergent procedures, aneurysm, or who had concomitant suprainguinal bypass were excluded. The primary outcome measure was postoperative SSI. Multivariable forward stepwise logistic regression was then performed including all variables with a P value of less than .10 in both matched and unmatched cohorts to evaluate for demographic and perioperative predictors of SSI. Propensity score matching was used to create matched cohorts of patients for each LOS group.

RESULTS

A total of 17,883 LEB procedures were selected for inclusion: 0 days (12,362 LEB), 1 to 2 days (1737 LEB), and 3 to 14 days (3784 LEB). Patients with the greatest preoperative LOS were more likely to have vein mapping (0 days preoperative LOS, 66.3%; 1-2 days, 65.2%; 3-14 days, 73.2%; P < .01) or computed tomography angiography/magnetic resonance angiography (0 days, 32.1%; 1-2 days, 34.4%; 3-14 days, 38.4%; P < .01). Patients with 3 or more days of preoperative LOS had longer procedure lengths (0 days, 244 minutes; 1-2 days, 243 minutes; 3-14 days, 255 minutes; P < .01) and were more likely to have completion angiogram (0 days, 27.1%; 1-2 days, 29.5%; 3-14 days, 31.6%; P = .02). Multivariable logistic regression demonstrated that preoperative LOS of 3 to 14 days was associated with increased rate of SSI (odds ratio [OR], 1.92; 95% confidence interval [CI], 1.20-3.07; P = .01). Transfusion of 3 or more units (OR, 2.87; 95% CI, 1.89-4.36; P < .01) and prolonged procedure length (>220 minutes; OR, 1.86; 95% CI, 1.26-2.73; P < .01) were also significantly associated with postoperative SSIs.

CONCLUSIONS

Many factors including preoperative comorbidities and operative complexity covary with preoperative LOS as risk factors for SSI. However, when patients are matched based on comorbidities and factors that would predict overall clinical complexity, preoperative LOS remains important in predicting SSI.

Study protocol for the groin wound infection after vascular exposure (GIVE) audit and multicentre cohort study

INTRODUCTION

Surgical site infections (SSIs) following groin incision for arterial exposure are commonplace and a significant cause of morbidity and mortality following major arterial surgery. Published incidence varies considerably. The primary aim of GIVE will be to compare individual units' practice with established guidelines from The National Institute for Health and Care Excellence (NICE). Secondary aims will be to describe the contemporary rate of SSI in patients undergoing groin incision for arterial exposure, to identify risk factors for groin wound infection, to examine the value of published tools in the prediction of SSI, to identify areas of equipoise which could be examined in future efficacy/effectiveness trials and to compare UK SSI rates with international centres.

METHODS AND ANALYSIS

This international, multicentre, prospective observational study will be delivered via the Vascular and Endovascular Research Network (VERN). Participating centres will identify all patients undergoing clean emergency or elective groin incision(s) for arterial intervention during a consecutive 3-month period. Follow up data will be captured at 90 days after surgery. SSIs will be defined according to the Centres for Disease Control and Prevention (CDC) criteria. Data will be gathered centrally using an anonymised electronic data collection tool or secure email transfer.

ETHICS AND DISSEMINATION

This study will be registered as a clinical audit at all participating UK centres; research ethics approval is not required. National leads will oversee the appropriate registration and approvals in countries outside the UK as required. Site specific reports of SSI rates will be provided to each participating centre. Study results will be disseminated locally at each site, publicised on social media and submitted for peer-reviewed publication.

Scoring System to Predict the Risk of Surgical Site Infection in Patients with Esophageal Cancer after Esophagectomy with Cervical Anastomosis

BACKGROUND

Surgical site infection (SSI) surveillance has become increasingly important during the peri-operative period of esophagectomy with cervical anastomosis (McKeown esophagectomy). This study sought to clarify the risk factors for SSI and to develop a stratification scoring system to predict SSI after esophagectomy with cervical anastomosis.

PATIENTS AND METHODS

All patients who underwent elective esophagectomy with cervical anastomosis were studied between January 2010 and December 2016 in the Chinese Academy of Medical Sciences Cancer Hospital (CAMS). Univariable analysis and multivariable logistic regression were used to screen the independent risk factors. A risk stratification scoring system was developed based on multivariable logistic regression parameters. The model derivation set involved 711 consecutive cases, and the validation set involved 168 consecutive cases.

RESULTS

In the model derivation set, there were 711 patients, of whom 146 were found to have SSI and the incidence rate was 20.53%. Multivariable analysis found that SSI was associated independently with the following adverse risk factors: peripheral vascular disease, prior chest surgery, no pre-operative surgical antibiotic prophylaxis (SAP) administration within 120 minutes prior to incision, low serum albumin, and low pre-albumin at post-operative day zero to three, respectively. Each of these factors contributed one point to the risk score and a risk stratification scoring system was established. The SSI rates were increased gradually in the low, intermediate, high, and extremely high-risk groups (p < 0.001). The area under the receiver operating characteristic (AUROC) curve was 0.706 for the logistic regression model and 0.704 for the scoring system. In the validation set, the model performed equivalently (AUC = 0.824).

CONCLUSIONS

The validated stratification scoring system could predict accurately the risk of SSI after esophagectomy with cervical anastomosis. This could be helpful in the selection of high-risk patients requiring frequent monitoring and more aggressive interventions to decrease the incidence of SSI.

Innovative Approaches to Improve Anti-Infective Vaccine Efficacy

Safe and efficacious vaccines are arguably the most successful medical interventions of all time. Yet the ongoing discovery of new pathogens, along with emergence of antibiotic-resistant pathogens and a burgeoning population at risk of such infections, imposes unprecedented public health challenges. To meet these challenges, innovative strategies to discover and develop new or improved anti-infective vaccines are necessary. These approaches must intersect the most meaningful insights into protective immunity and advanced technologies with capabilities to deliver immunogens for optimal immune protection. This goal is considered through several recent advances in host-pathogen relationships, conceptual strides in vaccinology, and emerging technologies. Given a clear and growing risk of pandemic disease should the threat of infection go unmet, developing vaccines that optimize protective immunity against high-priority and antibiotic-resistant pathogens represents an urgent and unifying imperative.