ImportancePostoperative opioid overprescribing leads to persistent opioid use and excess pills at risk for misuse and diversion. A learning health system paradigm using risk-stratified pancreatectomy clinical pathways (RSPCPs) may lead to reduction in inpatient and discharge opioid volume.ObjectiveTo analyze the outcomes of 2 iterative RSPCP updates on inpatient and discharge opioid volumes.Design, Setting, and ParticipantsThis cohort study included 832 consecutive adult patients at an urban comprehensive cancer center who underwent pancreatic resection between October 2016 and April 2022, comprising 3 sequential pathway cohorts (version [V] 1, October 1, 2016, to January 31, 2019 [n = 363]; V2, February 1, 2019, to October 31, 2020 [n = 229]; V3, November 1, 2020, to April 30, 2022 [n = 240]).ExposuresAfter V1 of the pathway established a baseline and reduced length of stay (n = 363), V2 (n = 229) updated patient and surgeon education handouts, limited intravenous opioids, suggested a 3-drug (acetaminophen, celecoxib, methocarbamol) nonopioid bundle, and implemented the 5×-multiplier (last 24-hour oral morphine equivalents [OME] multiplied by 5) to calculate discharge volume. Pathway version 3 (n = 240) required the nonopioid bundle as default in the recovery room and scheduled conversion to oral medications on postoperative day 1.Main Outcomes and MeasuresInpatient and discharge opioid volume in OME across the 3 RSPCPs were compared using nonparametric testing and trend analyses.ResultsA total of 832 consecutive patients (median [IQR] age, 65 [56-72] years; 410 female [49.3%] and 422 male [50.7%]) underwent 541 pancreatoduodenectomies, 285 distal pancreatectomies, and 6 other pancreatectomies. Early nonopioid bundle administration increased from V1 (acetaminophen, 320 patients [88.2%]; celecoxib or anti-inflammatory, 98 patients [27.0%]; methocarbamol, 267 patients [73.6%]) to V3 (236 patients [98.3%], 163 patients [67.9%], and 238 patients [99.2%], respectively; P < .001). Total inpatient OME decreased from a median 290 mg (IQR, 157-468 mg) in V1 to 184 mg (IQR, 103-311 mg) in V2 to 129 mg (IQR, 75-206 mg) in V3 (P < .001). Discharge OME decreased from a median 150 mg (IQR, 100-225 mg) in V1 to 25 mg (IQR, 0-100 mg) in V2 to 0 mg (IQR, 0-50 mg) in V3 (P < .001). The percentage of patients discharged opioid free increased from 7.2% (26 of 363) in V1 to 52.5% (126 of 240) in V3 (P < .001), with 187 of 240 (77.9%) in V3 discharged with 50 mg OME or less. Median pain scores remained 3 or lower in all cohorts, with no differences in postdischarge refill requests. A subgroup analysis separating open and minimally invasive surgical cases showed similar results in both groups.Conclusions and RelevanceIn this cohort study, the median total inpatient OME was halved and median discharge OME reduced to zero in association with a learning health system model of iterative opioid reduction that is freely adaptable by other hospitals. These findings suggest that opioid-free discharge after pancreatectomy and other major cancer operations is realistic and feasible with this no-cost blueprint.