Protective Effects of Enhanced Recovery After Surgery Components on Postoperative Paralytic Ileus: A Double Machine...

The evidence regarding which perioperative factors contribute to postoperative ileus (POI) is limited. Even within a multiple intervention enhanced recovery after surgery (ERAS) protocol, traditional statistical methods struggle to clarify these relationships. Advanced artificial intelligence-based approaches offer ways to overcome these limitations.

Consecutive data from patients undergoing elective colonic surgery were extracted from the ERAS Interactive Audit System database across six Swedish and one Norwegian centers between 2010 and 2024. Adjusted odds ratios were calculated to assess potential risk factors for POI. To estimate the causal effects of individual ERAS interventions on POI, we applied double machine learning.

A total of 7409 patients were included, of whom 351 (4.7%) developed POI. In multivariable logistic regression, several factors were significantly associated with increased risk of POI, including male sex, American Society of Anesthesiologists class ≥3, perioperative contamination, open surgical approach, right hemicolectomy, stoma closure, and prolonged operative time. Causal analysis using double machine learning was applied to evaluate the effect of individual ERAS interventions. Of the 11 preoperative and intraoperative ERAS components assessed, eight demonstrated a protective direction of effect against POI. However, only the avoidance of surgical drains achieved statistical significance. No intervention showed consistent evidence of increasing the risk of POI.

Avoidance of surgical drains was the only ERAS intervention with an association implying a significant protective effect against POI. Although the individual impact of other preoperative and intraoperative elements was limited, most showed a trend toward reduced risk, suggesting that they are unlikely to be harmful and may offer collective benefit.