Background: Older HNC survivors often experience post-treatment pain, and they have concerningly high rates of unsafe opioid prescribing per CDC guidelines (e.g., high doses, concurrent benzodiazepines). Because interventions to improve opioid safety should target providers who prescribe opioids to HNC survivors, we aimed to identify post-treatment opioid prescriber specialty. Methods: Using the national linked Surveillance, Epidemiology, and End Results (SEER) registry and Medicare claims data, we identified adults age >65 diagnosed 2014-2017 with stage I-III HNC and who had ≥6 months treatment-free follow-up prior to censoring at further treatment, new cancer, hospice, death, or end 2019. Each opioid fill was assigned to a prescriber specialty: oncology (medical, radiation), surgery, primary care (internal medicine, family practice, geriatrics), pain, or other. Prescriber patterns (number of fills, duration of fill, annual opioid receipt) were summarized across follow-up and stratified by year. Multinomial logistic regression models captured trends in prescribing, accounting for clinical and sociodemographic characteristics. Results: Among 5547 HNC survivors, 2951 (53%) had ≥1 opioid fill (median 2.1 years follow-up). Primary care providers (PCPs) prescribed 49% of all fills (46%-53% each year). PCPs prescribed opioids to 49% of survivors with ≥1 opioid fill, a greater share than other specialties. PCPs prescribed longer supplies of opioids (median 30 days/fill and median 45 days/year) than oncologists or surgeons. The likelihood of an opioid being prescribed by an oncologist or surgeon was ≥5 times lower than by a PCP. Over follow-up, rates of opioid prescribing by oncologists decreased relative to rates of prescribing by PCPs (Table). Conclusions: Primary care involvement in opioid prescribing remains high over post-treatment HNC survivorship. Interventions to address unsafe opioid prescribing for HNC survivors should target the primary care setting, as is typical for opioid-reduction efforts in the non-cancer population.