Variations in postoperative patient-reported outcomes (PROs) attributed to patients, surgeons, or surgical centers.

Abstract

e23277 Background: Postoperative PROs reflect patients’ subjective experiences and are influenced by numerous factors, including disease severity, surgery type and complexity, duration of hospitalization, and perioperative support. By measuring variation in postoperative PROs after cancer surgery, we sought to gain insight into potentially modifiable factors at the patient, surgeon and surgical center levels that might inform interventions to improve outcomes. Methods: We studied patients who had colectomy, hysterectomy, lobectomy, or pancreaticoduodenectomy at 6 centers participating in the SIMPRO study who completed ≥1 ePRO survey. For each surgical procedure, we fit a multilevel mixed-effects regression model with random effects for patient, surgeon, and center to partition the variance in total symptom scores. The total symptom score is the summation of 12 frequently reported postoperative symptoms (range 0-34), with higher scores indicating greater symptom burden. The fixed-effects covariates in the model included patient age, sex, race, ethnicity, comorbidities, primary language, poverty, insurance, marital status and time since surgical discharge. Coefficients were estimated for fixed effects. Intraclass correlation coefficients (ICCs) were calculated from the estimated variance components to quantify the proportion of variance attributable to patients, surgeons and centers (Table). Results: A total of 3808 patients had surgery (1316 hysterectomy, 1072 colectomy, 590 lobectomy, 198 pancreaticoduodenectomy) performed by 35, 52, 23, and 13 surgeons respectively across 6 centers. Most variation in PROs was attributable to patients (75% to 78%). The proportion of variation attributable to surgeons and centers was below 3% for hysterectomy, colectomy and lobectomy. Unexplained variation was less than 26% across procedures. Conclusions: Across surgeries for four common cancers, most variation in postoperative symptoms occurred at the patient level. The unexplained (residual) variation could be due to cancer stage, postoperative complications, supportive care programs, or unmeasured confounders. Identifying and incorporating potentially modifiable factors may enable targeted interventions to reduce the postoperative symptom burden. Clinical trial information: NCT03850912 . Decomposition of variance in postoperative PROs: intraclass correlation coefficients attributable to patients, surgeons, and centers ( ICC, %) . Hysterectomy Colectomy Lobectomy Pancreaticoduodenectomy Patients 75.00% 77.67% 76.10% 77.75% Surgeons 1.90% 0.55% 2.40% Center 1.10% 1.21% 2.00% Unexplained 22.00% 20.57% 19.50% 22.25%