Same-day EHR-coded sevoflurane versus propofol exposure and long-term coded dementia in older surgical adults: a propensity-matched cohort study
Related Experiment Videos
Chih-Lang Lin1,2,3,4,5, Shyh-Chyi Chang6, Kuan-Chou Lin7,8
1Division of Gastroenterology and Hepatology, Department of Internal Medicine, Dali Branch, Jen-Ai Hospital, Taichung, Taiwan.
Background And Objectives:
Whether perioperative EHR records of sevoflurane and propofol identify clinically meaningful differences in long-term dementia risk is uncertain. We compared same-day EHR-coded sevoflurane versus propofol exposure in relation to subsequent coded dementia among older adults undergoing major inpatient surgery.
Methods:
We conducted a multinational propensity-score-matched cohort study using the TriNetX federated EHR network. Adults aged 60 years or older undergoing major inpatient surgery from January 1, 2010, to December 31, 2024, were classified by a same-calendar-day record of the index procedure and sevoflurane or propofol. The aggregate export did not consistently provide route, bolus-versus-infusion status, dose, duration, or anesthesia timestamps; therefore, exposure groups were interpreted as EHR-coded patterns rather than validated exclusive maintenance regimens. The primary outcome was incident coded dementia from index + 1 day. Death was treated as censoring for the cause-specific Cox estimand. Sensitivity analyses included 90- and 180-day landmarks, a higher-specificity dementia definition, stricter exposure coding, and a 5-year dementia-or-death composite.
Results:
After matching, 35,363 patients were included in each group. Matching retained 99.94% of the sevoflurane-coded cohort but 2.49% of the much larger propofol-coded cohort. Coded dementia occurred in 701 patients (1.98%) in the sevoflurane-coded group and 384 (1.09%) in the propofol-coded group (hazard ratio [HR], 1.84; 95% CI, 1.62-2.08; P < 0.001). The 5-year dementia-or-death composite occurred in 3,024 patients (8.55%) and 2,195 patients (6.21%), respectively (HR, 1.43; 95% CI, 1.33-1.52). Most measured covariates were balanced after matching, although residual imbalance remained for diastolic blood pressure and C-reactive protein.
Conclusions:
In a selected matched common-support EHR population, same-day sevoflurane-coded exposure was associated with a higher subsequent incidence of coded dementia than same-day propofol-coded exposure. Exposure misclassification, restricted overlap, residual confounding, surveillance differences, and competing mortality preclude causal inference or prescribing recommendations.