Diagnosis of osteoporosis is based on bone mineral density (BMD) measurement, which is site dependent and commonly discordant between measurement sites. We aimed to determine the prevalence of osteoporosis diagnosed based on BMD T-scores measured by dual-energy x-ray absorptiometry (DXA) at different sites: the lumbar spine (LS) alone, femoral neck (FN) alone, or both. A total of 1712 women and 2028 men with LS and FN BMD measurements were enrolled. Over 50% discordance was found between osteoporosis classifications based on T-scores measured at the LS and FN. Use of the lowest T-scores measured at both the LS and right and left FN (rather than one site) significantly increased the prevalence of osteoporosis from 4.03 to 10.75% in postmenopausal women and 1.82 to 4.29% in men aged ≧50 years (p < 0.001). The trends of overall and age-adjusted prevalence of osteoporosis were similar in women and men. Osteoporosis was diagnosed at a higher rate if the USA reference rather than the Asia reference was used to calculate the T-score (26.64% vs. 10.75%). In conclusion, diagnosis based on the lowest T-score from multiple site BMD measurement can increase the prevalence of osteoporosis, demonstrating the higher sensitivity of the multiple site measurement strategy.