Abdominal wall reconstruction in ventral hernia repair: do current models predict surgical site risk?

Complications from ventral hernia repair (VHR) pose a significant healthcare burden. Risk assessment and stratification models are thus incentivized to improve cost-effectiveness and patient outcomes. The Ventral Hernia Risk Score (VHRS) and Ventral Hernia Work Group Classification (VHWG) are metrics that attempt to stratify and predict surgical site infection (SSI) and surgical site occurrence (SSO) risk based on patient characteristics. Our study aims to evaluate these models and assess external validity.

A retrospective review of all VHR procedures between October 2013 – August 2022 performed by the senior authors was conducted. Demographic, comorbidity, perioperative and outcome-related information was collected. Non-SSI and non-SSO cohorts were compared to SSI and SSO cohorts respectively to assess possible significant differences in patient demographics and operative characteristics. The VHRS and VHWG models were applied to each patient to predict risk. The Youden index of the respective Receiver Operating Characteristic (ROC) curves defined optimal score cutoffs for both models. Area under curve (AUC) was reported to assess model prediction quality.

A total of 1,414 patients who underwent VHR was identified, of which 175 (12.4%) experienced SSI and 367 (26.0%) SSO. Mean follow-up was 1.72 years [30 days, 13.65 years]. Patient demographics were similar between both non-SSI and SSI as well as non-SSO and SSO cohorts. However, comorbidities including prior non-VHR abdominal surgery (SSI: p < 0.001; SSO: p < 0.001), prior-VHR (SSI: p = 0.001; SSO: p-0.012), and prior mesh infection (p = 0.004) were significant between non-SSI and SSI cohorts as well as non-SSO and SSO cohorts. Operative characteristics including mesh plane (SSI: p = 0.008; SSO: p < 0.001) and adhesiolysis (SSI: p < 0.001; SSO: p < 0.001) were also significant in similar manner. Youden index of VHRS suggested a score of 7 as the optimal cutoff for increased SSI risk and 6 for SSO risk. The AUC was 0.609 for the VHRS-SSI model and 0.5882 for the VHRS-SSO model. VHWG grade of 3 was the optimal cutoff for both SSI and SSO. Model AUC was 0.616 for VHWG-SSI and 0.614 for VHWG-SSO.

Our study presents the largest external validation cohort for assessing the VHRS model. The VHRS was not superior toc the VHWG for SSI or SSO prediction. While the VHRS was designed for simplicity and basis in obvious patient or operative characteristics, it fails to appropriately weight pre-operative measures and more holistically evaluate clinical factors. Both models have limited predictability and generalizability in patients undergoing ventral hernia repair.