ABSTRACT: OBJECTIVES:The aim of this study was to estimate the comparative safety of initiating commonly used opioids among older, long-stay United States nursing home residents with fracture hospitalizations. METHODS:We conducted a new-user retrospective cohort study of nursing home residents initiating short-acting oxycodone, hydrocodone, or tramadol by merging the 2011-2013 Minimum Data Set 3.0 to Medicare hospitalization and pharmacy claims. Residents (??65 years, no cancer or hospice use) contributed treatment episodes (>?120 days with no prior opioid claims) and were followed for 180 days until incident fracture hospitalization (hip, femur, humerus, pelvis, radius/ulna), death (competing risk), treatment changes (e.g., discontinuation), or administrative censoring. Competing risks models with inverse probability of treatment weighting were used to estimate subdistribution hazard ratios (HRSD) and 95% confidence intervals (CI). RESULTS:Overall, 110,862 residents contributed 134,432 treatment episodes: 14,373 oxycodone; 69,182 hydrocodone; and 50,877 tramadol initiators. The incidences of fracture hospitalizations per 100 person-years were 9.4 (95% CI 7.5-11.7) for oxycodone, 7.9 (95% CI 7.1-8.8) for hydrocodone, and 5.0 (95% CI 4.3-5.7) for tramadol initiators. In weighted models, oxycodone initiators had a similar rate of fractures to hydrocodone initiators (HRSD 1.08, 95% CI 0.79-1.48). Tramadol initiators had lower fracture rates than hydrocodone initiators (HRSD 0.67, 95% CI 0.56-0.80). CONCLUSIONS:The lower rate of fractures that we documented among tramadol initiators compared with hydrocodone initiators is consistent, albeit attenuated compared with prior studies among community-dwelling older adults. However, overall fracture rates were lower than in community settings, potentially due to the limited risk of falling in this population with limited mobility.
