This is the first nation-wide, population-based, epidemiologic study of hip fractures in Korea and one of the few studies worldwide reporting the incidence of hip fractures based on national health insurance claims data. The age-standardized annual incidence rate for the entire population of Korea over 50 years of age in 2003 was estimated as 104.06 per 100,000 inhabitants, which is about 27.8% lower than that estimated in an earlier study conducted on inhabitants residing in selected regions of Korea in 2001 (133 per 100,000) . The difference in the estimated incidence rates between the two studies can be explained by the difference in the dataset, methods used in each study to identify incidence cases, and the population characteristics in the study regions. In the earlier study, hip fracture cases admitted to selected hospitals in Gwangju City and Chonnam Province were identified on the basis of medical records and radiographs. Among the identified incidence cases in that study, 88% underwent surgery. By applying this operation rate to the incidence rate estimated by our study, the total number of incidence cases with and without operation was estimated as 118.3 per 100,000 (= 104.6/0.88), thereby reducing the gap in the incidence estimates between the two studies from 27.8% to 12.4%.
After standardization according to the age distribution of the US white population in 1990 , the incidence rates of hip fractures in Korea were 173 per 100,000 in women and 91 per 100,000 in men. Interestingly, the standardized incidence rates of both genders were substantially lower than those of most of the Asian countries including Taiwan (505 per 100,000 in women and 225 per 100,000 in men), Hong Kong (459 and 180 per 100,000), Singapore (442 and 164 per 100,000), and Thailand (269 and 114 per 100,000) .
The risk of 1-year mortality following hip fracture was 16.55%, which is about 2.85 times higher than the mortality rate of the general population in South Korea over 50 years of age. Despite the heterogeneity of the study population in terms of age and ethnicity, and the potential discrepancy in the quality of follow-up care after the fracture, the post-fracture mortality at 1 year observed in other countries is similar to that in Korea. A recent U.S. study has reported that about 11.5% of 950 patients over age 65 admitted to selected hospitals with a hip fracture between 1987 and 1997 died after 1 year . The 1-year mortality rate was 15.7% among 541 fracture patients over 60 years of age admitted to one university hospital in Greece between 1989 and 1992 . In a population-based, case-control study in Sweden, about 10.6% of 2,245 women aged 50-81 years admitted with a hip fracture died within 1 year . The comparison with other Asian countries also showed similar results. The 1-year mortality of patients with surgically treated hip fractures at a hospital in Taiwan between 1998 and 2006 was about 14.0% . A Japanese study by Muraki et al. reported that the 1-year mortality rate following hip fracture was 11.5% .
Osteoporosis is recognized as a problem predominantly in elderly women [24, 25]. Consequently, women have higher incidence rates of hip fracture than men. In Korea, the female to male ratio of the age-standardized incidence rates of hip fractures for those aged 50 and older is 2.37. This ratio is comparable to western countries such as Sweden (2.2, 1987-1991), Norway (2.3, 1983-1984), Australia (2.7, 1989-1990), and in U.S. whites (2.7, 1983-1984), and in some Asian countries such as Thailand (2.3, 1997-1998), Hong Kong (2.5, 1997-1998), and Malaysia (2.4, 1997-1998) . Hip fracture rates were higher in women of all age groups except for those under 55 years. This pattern is commonly observed in populations of other countries, although the starting age at which women show a higher hip fracture rate varies from country to county: 60 in Taiwan and Japan [4, 9], but mostly 55 or 50 in other countries such as Mexico, Argentina, Iran, and Morocco [1–3, 5].
Although women showed consistently higher incidence rates of hip fractures than men, the high incidence rates among older men is worthy of attention. After age 75, Korean men experienced hip fractures at a rate of 211.42 to 517.51 per 100,000 inhabitants. This rate is similar to that observed in other Asian countries, such as Japan (209.0 to 780 per 100,000), Hong Kong (404 to 1,639), Singapore (611), Malaysia (320), and Thailand (227 to 727) [4, 6]. The increased incidence with increasing age in both genders confirms the equal vulnerability of both genders to the aging process. Thus, preventive strategies are important for both genders.
It is interesting to observe higher post-fracture mortality in Korean males than in females. Gender differences in post-hip fracture mortality have been reported in earlier studies showing that men were about twice as likely as women to die during the first years after hip fracture [26–28]. The observed difference between men and women could be explained by under treatment of osteoporosis  and higher infection rates in men following hip fracture [26, 27].
Low socioeconomic status among the elderly population seems to be associated with an increased risk of fracture-caused mortality. The adjusted hazard ratio of post-fracture mortality was significantly higher among those enrolled in the MA program than among those in the HI program. Patients with low socioeconomic status, in general, have poorer health status and therefore tend to have poorer health outcomes from the same conditions as compared to those with higher socioeconomic status [29, 30]. Thus, more effective public health strategies to treat osteoporosis and to prevent fracture incidence should be implemented that target the indigent elderly population in Korea.
The regional variation in fracture-related mortality among Korean elderly is noteworthy. Those living in places other than the capital city consistently showed a higher risk of death during the first year after hip fracture, with a minimum of 21.4% or a maximum of 65.6% additional chance of death. This discrepancy may be due to the lack of access to health care providers or the poor quality of health care provided in non-city or small city areas. Further investigation is necessary to figure out the reasons for this regional discrepancy and therefore to reduce this discrepancy.
From our analysis, patients receiving medication to treat osteoporosis after fracture were 42.4% less likely to die within the year following fracture. The association between reduced mortality and post-fracture use of anti-osteoporosis drugs in elderly hip fracture patients has been addressed in earlier studies. A prospective analysis from Finland has reported a 43% reduction in deaths at 36 months following hip fracture in females who used prescribed calcium plus vitamin D supplementation concomitantly with anti-osteoporosis drugs . Also, an observational study from Canada has shown that mortality is significantly lower in the group treated with anti-osteoporosis drugs than in the untreated group following hip fracture . Because elderly people with prior fractures are at a higher risk for future fractures , clinical practice guidelines recommend the initiation of pharmacologic treatment of osteoporosis after the first fracture . However, under-treatment of osteoporosis [35, 36] and low adherence to oral anti-osteoporosis treatment following hip fracture  is commonly observed. In our study, only 20.9% of the fractured patients received anti-osteoporosis drugs following hip fracture. Thus, a more aggressive practice of diagnosing and providing medication therapy for osteoporosis seems to be an urgent need in Korea.
Interestingly, patients that underwent an operation for hip fracture from hospitals or clinics tend to have a better chance of survival than those from tertiary hospitals. Although our multivariate model tried to include as many patient characteristic variables as possible from the NHI data, we could not comprehensively control for hip fracture severity and patient health conditions due to the lack of clinical information available from the claims records. Thus, it cannot be concluded that the quality of care and patient outcomes after hip fracture treatment are better among hospitals and clinics than among tertiary hospitals in Korea. It might be possible that patients with more severe conditions underwent hip fracture repairs from larger and more advanced hospitals (i.e., tertiary or general hospitals) and therefore the prognosis after surgery tended to be poorer. A more thorough adjustment for patient baseline disease severity is necessary to make a concrete conclusion about this finding.
Since a hip fracture is a severe condition, most of the care episodes are initiated with hospital admission and most prior studies have relied on hospitalized hip fractures to identify incident cases [14, 17]. However, if we restrict the study sample to include hospitalized fractures only, we will underestimate incidence rates by excluding less severe cases that are initiated with office visits. To overcome this problem, the present study defined incident cases either by initial admission or a visit containing a diagnosis of hip fracture. However, while most hospital admissions coincide with an acute new episode of illness, it is difficult for outpatient visits to determine whether the case is a new episode or part of a post-episode course. In addition, the accuracy of diagnostic codes among the claims data for office visits is lower than that of hospital admission [38, 39]. Thus, to improve the accuracy of identifying incidence cases based on initial fracture visits, we excluded patients with a claims record with a hip fracture during the 6 months prior to the initial visit. Furthermore, to confirm the hip fracture cases, hip fracture-related operation codes were used in addition to the diagnostic codes.
Several methodological issues arise from the studies using administrative claims data. First, not all patients with fractures have access to hospitals. The incidence based on insurance claims records would be underestimated if many hip fractures in Korea were not diagnosed or treated in health care institutions. However, due to the emergent character of hip fractures, it is believed that virtually all hip fractures present to health services . Thus, the incidence of hip fractures identified from claims records is considered to be close to the actual incidence.
Second, the reliance on ICD-10 diagnostic codes to identify incident fractures may cause misclassification of incident hip fractures due to the inherent nature of claims data such as voluntary or non-voluntary miscoding behavior. However, a recent validation study for the diagnostic codes of the NHI claims database in Korea has alleviated this concern. It has been revealed that about 70% of primary, secondary, or tertiary diagnosis codes from NHI claims records coincide with those from medical records. In addition, the accuracy of diagnosis codes tended to be higher for claims from hospital admissions compared with office visits, and for claims of severe conditions compared with mild conditions [38, 39].