Background: Opioids are often used in nursing homes to manage non-malignant pain, but little is known about their long-term use, initiation, and comparative safety. Methods: We used the Minimum Data Set 3.0 from 2011-2013 merged to Medicare and facility characteristics data to study opioid use and safety among older, long-stay residents. The specific aims were to examine the 1) prevalence of long-term opioid use; 2) geographic variation in the initiation of commonly used opioids (oxycodone, hydrocodone, tramadol); and 3) comparative safety of commonly used opioids and fracture hospitalizations. Results: One in seven long-stay residents were prescribed opioids long-term. There was extensive geographic variation in the initiation of commonly used opioids, with oxycodone (9.4%) initiated less frequently than hydrocodone (56.2%) or tramadol (34.5%) but varying most extensively across the United States, with the majority of variation in prescribing explained by state of residence. Compared to hydrocodone initiators (7.9 fracture hospitalizations per 100-person years), those initiating tramadol had lower rates of fracture hospitalizations (subdistribution hazard ratio [HRSD] = 0.67, 95% Confidence Interval [CI]: 0.56-0.80), whereas oxycodone initiators had similar rates of fracture hospitalizations (HRSD=1.08, 95% CI: 0.79-1.48). Conclusion: The prevalence of long-term opioid use was twice as common in nursing homes as community settings, with initiation patterns varying extensively by region and being strongly driven by state of residence. Although initiating tramadol was associated with lower rates of fractures than hydrocodone, questions on opioid risks and benefits remain and are especially pertinent given the high mortality rates in this population.