Abstract
BACKGROUND: Procedure-specific postoperative pain in otolaryngology varies substantially, yet real-world associations between time-sensitive analgesic processes and patient-centered outcomes remain incompletely characterized in Chinese practice settings. METHODS: This single-center prospective observational cohort (17 January 2021-31 March 2025) enrolled consecutive adults undergoing tonsillectomy, septoplasty, functional endoscopic sinus surgery (FESS), tympanoplasty/mastoidectomy, or microlaryngoscopy at a tertiary academic hospital in Guangzhou, China. Electronic health and anesthesia records captured 48-h worst pain [numeric rating scale (NRS) 0-10] and a composite process quality indicator [scheduled non-opioid ≤6 h plus timely post-anesthesia care unit (PACU) rescue ≤30 min for severe pain (NRS ≥ 7)]. Procedure category was the primary exposure. Worst pain was analyzed using ordinary least squares regression with restricted cubic splines and provider fixed effects; the process composite using logistic regression with heteroskedasticity-consistent standard errors. RESULTS: Among 1,632 patients (mean age 38.5 years; 55.2% male), severe PACU pain occurred in 31.4%, rescue opioids in 42.5%, and timely rescue in 81.5% of those rescued (34.7% overall). Patients receiving scheduled non-opioids at discharge (80.3%) had lower 48-h pain (median 4.0 vs. 4.6), reduced opioid exposure (8.7 vs. 11.6 mg oral morphine equivalent), and higher satisfaction. Procedure category dominated pain outcomes: tonsillectomy versus FESS β = 1.14 [95% confidence interval (CI) 0.91-1.38]; microlaryngoscopy β = -2.02 (-2.34 to -1.70). For the process composite, higher opioid exposure predicted lower attainment (odds ratio per 5 mg 0.92; 95% CI 0.90-0.94), non-steroidal anti-inflammatory drug use predicted higher attainment (1.25; 1.00-1.57). CONCLUSION: Postoperative pain in ear, nose, and throat surgery is strongly procedure-dependent. Non-opioid-first regimens and timely rescue represent actionable quality improvement targets, requiring multicenter validation before broader implementation.