Routinely available perioperative variables and EHR-captured postoperative adverse-event indicators after adult abdominal surgery: A retrospective cohort study using MOVER EPIC
Anqi Shi, Xinyue Cui, Hanling Zheng, Yifan LiBackground
Perioperative electronic health record (EHR) datasets are increasingly used to study postoperative morbidity, although database-recorded adverse-event indicators may not correspond to standardized complication endpoints. We characterized these indicators after adult abdominal surgery in MOVER EPIC and examined associations with routinely available perioperative variables.
Methods
This retrospective cohort study identified adult abdominal surgical encounters from 2017–2022 through procedure-name screening and case-level deduplication. The primary outcome was a broad composite of EHR-captured postoperative adverse-event indicators recorded within the indexed perioperative encounter. The secondary outcome was a narrower composite of severe EHR-captured adverse-event indicators. Multivariable logistic regression was performed in the primary complete-case cohort, adjusting for age, sex group, ASA physical status, procedure group, and operating-room duration. Sensitivity analyses included ASA-excluded models, alternative duration specifications, first-encounter restriction, penalized regression, patient-level cluster-robust inference, and inverse probability weighting (IPW).
Results
Among 5,679 encounters, the broad composite occurred in 166 (2.92%), the narrow severe composite in 95 (1.67%), and death in 23 (0.40%). In the primary complete-case cohort (n = 5,349; 147 broad events), ASA IV + was strongly associated with the broad composite compared with ASA III (adjusted OR 4.66, 95% CI 2.92–7.45; p < 0.001). Laparotomy/adhesiolysis was also associated with higher odds (OR 1.79, 95% CI 1.02–3.13; p = 0.041), whereas spline analyses suggested modest nonlinearity in the age association; operating-room duration was not independently associated. IPW estimates were materially similar, although the estimate for ASA I–II was less precise.
Conclusions
Clinician-assigned ASA physical status was the most consistent correlate of EHR-captured postoperative adverse-event indicators. ASA may reflect comorbidity burden and perceived perioperative vulnerability rather than directly measuring physiological reserve. These findings should not be interpreted as standardized 30-day postoperative complication risks.