OBJECTIVES: Secular decreases in hip fracture rates have been reported in some countries. Whether this phenomenon applies to other fracture sites has not been studied. Our objective was to examine fracture patterns across twenty years in a population-based cohort to inform health care planners on trends in costs and resource needs related to fractures. METHODS: We used the Population Health Research Data Repository for the Province of Manitoba, Canada which is a comprehensive collection of databases including physician visits, hospitalizations and pharmaceutical prescriptions. Age and sex-adjusted fracture rates were calculated for men and women age fifty years and older from 1986 to 2006 according to fracture site (defined by ICD-9-CM codes) and mechanism (presence/absence of ICD-9-CM external injury codes). Generalized linear models with generalized estimating equations were used to derive adjusted annual rates and test the linear change overall, and for men and women separately. RESULTS: Osteoporotic fractures (non-traumatic fractures of the hip, forearm, spine and humerus) showed a significant linear decline (0.8% per-annum [95% CI 0.3-1.2%]), with a greater decline in women (1.0% [0.4-1.7%]) than in men (0.5% [0.0-1.3%], P<.05 for sex interaction). Similar trends were seen for all fractures sites: hip 0.9% (0.2-1.7%), forearm 0.8% (0.4-1.3%), humerus (0.7% [0.2-1.2%]) and spine (0.5% [0.0-1.0%])  A greater reduction in traumatic fractures was observed (1.8% per-annum [95% CI 1.0-2.6%]), with a greater decline in men (2.2% [1.3-3.1%]) than in women (1.3% [0.2-2.4%], P<.05 for sex interaction). Similar results were seen when testing the difference between the initial 5 years (1986-1991) and the final five years (2001-2006) of data. CONCLUSIONS: We observed a decrease in both non-traumatic (osteoporotic) and traumatic fracture rates over the study period. This decline was apparent in years prior to widespread osteoporosis testing or availability of modern pharmacotherapy.