Errors bars indicate 95% confidence intervals.
Separate regression lines are fitted to the period 1985 through 1995 (black) and the period 1996 through 2005 (blue). Solid lines indicate regression lines fitted to data points for the corresponding time period; dashed lines indicate portions of regression lines extrapolated over the remaining time period. Errors bars indicate 95% confidence intervals.
Leslie WD, O’Donnell S, Jean S, Lagacé C, Walsh P, Bancej C, Morin S, Hanley DA, Papaioannou A, Osteoporosis Surveillance Expert Working Group FT. Trends in Hip Fracture Rates in Canada. JAMA. 2009;302(8):883-889. doi:10.1001/jama.2009.1231
Author Affiliations: Department of Medicine, University of Manitoba, Winnipeg, Canada (Dr Leslie); Public Health Agency of Canada, Ottawa, Ontario, Canada (Mss O’Donnell and Lagacé, Mr Walsh, and Dr Bancej); Institut National de Santé Publique du Québec, Quebec City, Quebec, Canada (Ms Jean); McGill University, Montreal, Quebec, Canada (Dr Morin); University of Calgary, Calgary, Alberta, Canada (Dr Hanley); and McMaster University, Hamilton, Ontario, Canada (Dr Papaioannou).
Context Hip fractures are a public health concern because they are associated with significant morbidity, excess mortality, and the majority of the costs directly attributable to osteoporosis.
Objective To examine trends in hip fracture rates in Canada.
Design, Setting, and Patients Ecologic trend study using nationwide hospitalization data for 1985 to 2005 from a database at the Canadian Institute for Health Information. Data for all patients with a hospitalization for which the primary reason was a hip fracture (570 872 hospitalizations) were analyzed.
Main Outcome Measures Age-specific and age-standardized hip fracture rates.
Results There was a decrease in age-specific hip fracture rates (all P for trend <.001). Over the 21-year period of the study, age-adjusted hip fracture rates decreased by 31.8% in females (from 118.6 to 80.9 fractures per 100 000 person-years) and by 25.0% in males (from 68.2 to 51.1 fractures per 100 000 person-years). Joinpoint regression analysis identified a change in the linear slope around 1996. For the overall population, the average age-adjusted annual percentage decrease in hip fracture rates was 1.2% (95% confidence interval, 1.0%-1.3%) per year from 1985 to 1996 and 2.4% (95% confidence interval, 2.1%-2.6%) per year from 1996 to 2005 (P < .001 for difference in slopes). Similar changes were seen in both females and males with greater slope reductions after 1996 (P < .001 for difference in slopes for each sex).
Conclusions Age-standardized rates of hip fracture have steadily declined in Canada since 1985 and more rapidly during the later study period. The factors primarily responsible for the earlier reduction in hip fractures are unknown.
Osteoporosis is a common condition that predisposes individuals to skeletal fractures. Worldwide, the number of people who have suffered a prior osteoporotic fracture was estimated to be 56 million in 2000 with approximately 9 million new osteoporotic fractures each year.1 Because the prevalence of osteoporosis increases with age, the global burden of osteoporosis is projected to rise markedly over the next few decades as the number of elderly individuals increases.2
Osteoporosis increases the risk of fracture at most skeletal sites3; however, fractures of the proximal femur are of particular concern because they are associated with significant morbidity, loss of independence, excess mortality, and the majority of the costs directly attributable to osteoporosis.4- 7 The case-fatality rate for hip fractures can exceed 20% at 1 year.8,9
The incidence of hip fractures is an index of osteoporosis burden and the potential impact of preventive efforts in the population. The objective of this study was to examine the trends of hip fractures in Canada.
The Canadian Institute for Health Information (CIHI) collects and analyzes information on health and health care in Canada and makes this publicly available. Data from the Hospital Morbidity Database (HMDB), a database housed at the CIHI, were used in the current analysis. The HMDB includes administrative, clinical, and demographic information on hospital inpatient events and provides national discharge statistics from Canadian health care facilities by diagnoses and procedures. The HMDB includes data from the Discharge Abstract Database but also appends data for provinces/territories that do not participate in the Discharge Abstract Database in order to provide national geographic coverage. Through a data quality enhancement program, CIHI ensures a high quality of information in the HMDB and its other data holdings.10 The data were obtained through the Public Health Agency of Canada (PHAC). The study was reviewed and approved by the PHAC through the approvals process for peer-review publications.
We identified all hospitalizations over the period from January 1, 1985, to December 31, 2005, from the HMDB in which the most responsible diagnosis (the primary reason for hospitalization) was a proximal femoral fracture. The most responsible diagnosis excludes hospitalizations due to complications and revisions. In our data set, we found that approximately 11% of the proximal femoral fracture cases had a second hospital admission with the same diagnosis during the same calendar year, largely due to second fractures and interhospital transfers. It was not possible to distinguish a second fracture from interhospital transfer across all years of data, and therefore both hospitalizations were counted. Fracture diagnoses were coded according to the International Classification of Diseases (the Ninth Revision [ICD-9], the Ninth Revision, Clinical Modification [ICD-9-CM], and the 10th Revision, Canada [ICD-10-CA]). Prior to 1985, reporting of hospitalizations to CIHI was voluntary and hip fracture data were likely incomplete. Therefore, we used data from 1985 onwards when standardized coding methods were followed. The ICD-9 and ICD-9-CM were used from 1985 until 2000 throughout Canada with staggered introduction of ICD-10-CA starting in 2001. Not all provinces/territories transitioned between ICD systems in the same year.
We identified hip fractures using the following diagnosis codes: 820.x from the ICD-9-CM and S72.0-.2 from the ICD-10-CA. Procedure codes and physician claims were not used for defining hip fractures, but previous studies have shown a high degree of concordance with hospitalization data.11 The annual number of hip fractures was tabulated for the study period (1985 to 2005) and stratified by province, sex, and age groups (initially as 5-year intervals with aggregation into the following age groups: <55 years, 55-64 years, 65-74 years, 75-84 years, and ≥85 years). We stratified the denominator similarly using national census data with interpolated estimates for between-census years. Northern territories (Northwest Territories, Yukon, and Nunavut), encompassing 0.3% of the Canadian population, did not provide data across all years and were excluded from analysis.
Yearly changes in age-specific hip fracture subgroups (stratified by sex) were assessed using the Cochran-Armitage test for linear trend. This test is sensitive to the linearity between the response variable (number of fractures) and the ordinal category (calendar year). The trend test gives evidence of increasing linear trends, stable trends over time, or decreasing trends. Annual unadjusted (crude) fracture rates per 100 000 person-years with 95% confidence intervals (CIs) were calculated for the overall population, for each sex, and for previously defined age groups. Annual age-standardized hip fracture rates per 100 000 person-years were calculated for the entire Canadian population and stratified by sex. Rates were direct-adjusted to the 1991 age structure of the Canadian population to allow for comparison over time.
We used joinpoint regression analysis to identify points where a statistically significant change over time occurred in the linear slope of the trends in hip fracture rates (Joinpoint Regression Program, version 3.3, April 2008; Statistical Research and Applications Branch, National Cancer Institute, Bethesda, Maryland). In joinpoint analysis, the best-fitting points correspond to where the rate changes significantly (increases or decreases).12 The analysis starts with the minimum number of joinpoints and tests whether 1 or more joinpoints are statistically significant in the model and should be added. In the final model, each joinpoint indicates a statistically significant change in trend, and an annual percentage change is computed for each of those segments by means of generalized linear models assuming a Poisson distribution. A P value of less than .05 was taken to indicate a statistically significant effect.
During the 21 years of observation, we identified 570 872 hospitalizations for hip fractures in Canada. Slightly less than three-quarters of the hip fractures occurred in females (Table 1). The age-specific hip fracture rates decreased within each age group for females and males (all P for trend <.001). Among females (Table 2), the largest percentage decrease was in those aged 55 to 64 years (46.3%; 95% CI, 40.3%-53.3%), while the largest absolute decrease was in those aged 85 years and older (785.7 per 100 000 person-years; 95% CI, 687.8-897.6). For males (Table 3), the largest percentage decrease was again seen in those aged 55 to 64 years (32.5%; 95% CI, 24.9%-42.4%), with the largest absolute decrease in those aged 85 years and older (455.5 per 100 000 person-years; 95% CI, 350.7-591.6).
From 1985 to 2005, the number of hip fractures and the unadjusted (crude) hip fracture rate increased (Figure 1). To adjust for differences in the population structure over time, we calculated age-standardized rates of hip fractures for females, males, and the overall population. Figure 2 demonstrates that these rates have progressively declined from 1985 to 2005 in both males and females. In females, age-adjusted hip fracture rates decreased from 118.6 per 100 000 person-years (95% CI, 115.9-121.4) in 1985 to 80.9 (95% CI, 79.2-82.6) in 2005. Likewise, the age-adjusted hip fracture rates for males decreased from 68.2 per 100 000 person-years (95% CI, 65.6-70.8) to 51.1 (95% CI, 49.4-52.7). This represents a 31.8% decrease in females and a 25.0% decrease in males over 21 years.
To evaluate whether the change in age-adjusted hip fracture rates was constant across the years of observation, we performed joinpoint regression analysis for the rate of change from 1985 to 2005 (Table 4). Joinpoint regression analysis identified a change in the slope around 1996 for both males and females. When expressed as average percentage change from the 1985 baseline, the decrease in hip fracture rates was 1.2% (95% CI, 1.0%-1.3%) per year from 1985 to 1996 and 2.4% (95% CI, 2.1%-2.6%) per year from 1996 to 2005 (P < .001 for difference in slopes). Large changes were seen in both females and males with greater slope reductions after 1996 (P < .001 for difference in slopes). The rate of decrease was slightly greater in females than males both prior to 1996 (1.3% [95% CI, 1.1%-1.5%] per year vs 0.8% [95% CI, 0.5%-1.1%] per year) and after 1996 (2.4% [95% CI, 2.2%-2.6%] per year vs 2.0% [95% CI, 1.6%-2.4%] per year).
Our study identifies significant changes in the number and pattern of hip fractures occurring in Canada during 21 years of observation. Standardized rates of hip fracture have been decreasing with a more rapid decrease during the latter half of the study period, although the exact inflection point is difficult to determine precisely and our data are also consistent with a more gradual (nonlinear) change. For the age range 55 to 64 years, rates have decreased by almost one-half in females and by about one-third in males. The observation of declining age-adjusted fracture rates with a simultaneous increase in crude fracture rates is attributable to a changing age structure of the population, with a growing number of older people who are at greatest risk for hip fracture. The oldest segment of the population (aged ≥85 years) is growing faster than others, and although they showed the greatest absolute decrease in hip fracture rates, their relative decline was less than among those aged 55 to 64 years.
Similar trends have been reported in other countries, including the United States.13- 15 Gehlbach et al13 examined hospital discharges from the Nationwide Inpatient Sample from 1993 through 2003 among women and men aged 45 years and older. Over the 11-year study period, the age-adjusted rates for both women and men fell by about 20%. Melton et al16 reported that hip fracture rates appeared to start decreasing in the United States beginning in the 1950s in females and after 1980 in males. Ethnicity may be important because standardized hip fracture incidence in California (1983-2000) showed a decrease in non-Hispanic white females, no significant change among black or Asian females, and a significant increase among Hispanic females.17 There is no field for ethnicity in the data source used and therefore no way to explore this question in the Canadian population.
Not all countries have reported a decrease in hip fractures. Germany recently reported national hip fracture data covering the period 1995 through 2004 and did not see a consistent change, although the observation period may not have been long enough to identify smaller changes.18 The region around Geneva, Switzerland, observed a decrease in hip fractures between 1991 and 2000 in females (1.4% [95% CI, 0.1%-2.6%] per year) but not in males.15
An ecological correlation between an increase in bone mineral density testing and a decrease in osteoporotic fractures was recently reported from the province of Ontario, Canada, for the years 1996 to 2001.19 Our study extends the period of observation of fracture rates over a longer period and examines changes by sex. We found that trends toward decreasing hip fracture rates were evident before widespread availability of bone density testing or the modern era of pharmacotherapy. Further, the similar pattern observed in males and females argues against sex-specific interventions such as hormone replacement therapy or oral contraception. The greater decline in hip fracture rates in the latter period of the study could reflect a more widespread use of bone density testing and treatment, although this does not explain the reductions seen in males who have not been a target group for osteoporosis screening and treatment.20 Indeed, low treatment rates are seen in females with the highest risk, while male osteoporosis is even less frequently recognized and treated.21- 23
Our findings raise a compelling question: what other factors could contribute to the decline in hip fracture rates? There is no clear answer to this very important question.24 A secular increase in the average number of reproductive years and exposure to circulating endogenous hormones has been reported in females, although this would not account for changing hip fracture rates in males.25 There is little evidence to suggest that improvements in physical activity, calcium intake, vitamin D status, or falls prevention have affected hip fracture rates at the population level.26- 29 Declining smoking rates could be associated with reductions in hip fractures because smoking is a risk factor,30 but it is doubtful that this would be sufficient to account for the magnitude of the change in the rates of hip fractures.31 The possibility of a birth cohort effect resulting in a healthier aging population with improved functional ability and reduced risk of injurious falls has been proposed.14 Overweight and obesity are epidemic in modern societies and may contribute to reduced fracture rates. Greater body weight is associated with higher bone density and nonovarian aromatization of estrogen and provides padding over the trochanter.
Our study has obvious implications for health care administrators responsible for coordinating and prioritizing health care delivery. Estimates of the future burden of osteoporosis are predicated on the ability to accurately predict fracture incidence rates. Indeed, an examination of the burden of hip fractures for elderly French females concluded that inaccurate modeling could lead to a 70% overestimation in the number of hip fractures.32 A recent analysis from the United States projected that the incidence of osteoporotic fractures would increase by almost 50% between 2005 and 2025.5 However, the projection assumed static fracture rates (based on rates from 2001). The authors acknowledged that any projections would be sensitive to secular trends in fracture incidence rates. As the projected cost of osteoporotic fractures of $25.3 billion for 2025 was largely dependent on hip fractures, significant departures from these assumptions could translate into deviations from the estimates of several billion dollars.
A major strength of this study is the use of national population-based data spanning many years. Although administrative data reliably identify hip fractures, we do not have information on body weight, smoking history, diet or nutritional markers, or other lifestyle variables that could be involved in the changing risk of hip fractures. The national data source used also does not have medication records or ambulatory care data. As noted earlier, readmissions for hip fracture during the same calendar year will lead to some “double-counting” (approximately 11%), although a substantial number of these are likely attributable to a second fracture because 1-year risk of a subsequent fracture after a prior hip fracture is reported to be in the range 2.3% to 10%.33- 35 The change in hospital diagnostic coding from ICD-9-CM to ICD-10-CA occurred after 2000 and implementation dates differed among the provinces. However, this does not explain the decrease that was evident well before any change in coding classification.
In summary, this analysis of hip fracture rates at a national level demonstrates a significant reduction during the period 1985 through 2005. This decrease was evident in both females and males, with an onset that precedes large-scale use of diagnostic testing for osteoporosis or modern pharmacotherapy. The factors contributing to the earlier reduction in hip fractures are currently unknown. Despite the significant reduction in standardized hip fracture rates, the absolute number of hip fractures continues to increase. Hip fractures continue to exert major effects on the population, particularly the elderly, and on the health care system, related to the morbidity, costs, and mortality from these fractures. Therefore, the decreasing incidence rates are not grounds for complacency toward osteoporosis prevention and treatment.
Corresponding Author: William D. Leslie, MD, MSc, Department of Medicine (C5121), University of Manitoba, 409 Tache Ave, Winnipeg, MB R2H 2A6, Canada (email@example.com).
Author Contributions: Dr Leslie had full access to all of the data in the study and takes responsibility for the integrity of the data and the accuracy of the data analysis.
Study concept and design: Leslie, Morin, Hanley.
Acquisition of data: Leslie, O’Donnell, Walsh, Bancej.
Analysis and interpretation of data: Leslie, O’Donnell, Jean, Lagacé, Walsh, Bancej, Hanley, Papaioannou.
Drafting of the manuscript: Leslie.
Critical revision of the manuscript for important intellectual content: Leslie, O’Donnell, Jean, Lagacé, Walsh, Bancej, Morin, Hanley, Papaioannou.
Statistical analysis: Leslie, Jean, Walsh.
Administrative, technical, or material support: Leslie, O’Donnell, Lagacé, Bancej, Papaioannou.
Study supervision: Morin.
Financial Disclosures: Dr Leslie reported receiving honoraria for lectures from Merck Frosst Canada; research support from Merck Frosst Canada and Amgen Pharmaceuticals Canada; and unrestricted educational and research grants from the Alliance for Better Bone Health, Sanofi-Aventis, Procter & Gamble Pharmaceuticals Canada, Genzyme Canada, and Amgen Pharmaceuticals Canada. Dr Morin reported serving on advisory boards for Procter & Gamble, Sanofi-Aventis, Servier, Amgen, Eli Lilly, and Novartis; serving on the speakers' bureau for Procter & Gamble, Sanofi-Aventis, and Servier; and receiving an unrestricted research grant from Amgen. Dr Hanley reported serving on advisory boards for Amgen Canada, Eli Lilly Canada, Merck Frosst Canada, Novartis Canada, NPS Pharmaceuticals, Nycomed, Procter & Gamble Canada; working on clinical trials for Amgen, Aventis, Eli Lilly Canada, Merck Frosst, Novartis, NPS Pharmaceuticals, Pfizer, Procter & Gamble, Roche, Wyeth-Ayerst; and receiving speaking honoraria from Amgen, Aventis, Eli Lilly Canada, Merck Frosst, Novartis, NPS Pharmaceuticals, Nycomed, Procter & Gamble Canada, and Wyeth-Ayerst. Dr Papaioannou reported serving as a consultant or on the speakers' bureau for Amgen, Aventis Pharma, Eli Lilly, Merck Frosst Canada, Novartis, Procter & Gamble Pharmaceuticals, Servier, and Wyeth-Ayerst; working on clinical trials for Eli Lilly, Merck Frosst, Novartis, Procter & Gamble, and Sanofi-Aventis; receiving unrestricted grants from Amgen, Eli Lilly, Merck Frosst, Procter & Gamble, and Sanofi-Aventis; and receiving other support from the Ontario Ministry of Health and Long Term Care.
Osteoporosis Surveillance Expert Working Group: Jacques Brown, MD, Laval University, Quebec City, Quebec, Canada; Ann Cranney, MD, MSc, University of Ottawa, Ottawa, Ontario, Canada; David A. Hanley, MD, University of Calgary, Calgary, Alberta, Canada; Susan Jaglal, PhD, University of Toronto, Toronto, Ontario, Canada; Sonia Jean, MSc, Institut National de Santé Publique du Québec, Quebec City; Famida Jiwa, MHSC, DC, Osteoporosis Canada, Toronto; Stephanie Kaiser, MD, Dalhousie University, Halifax, Nova Scotia, Canada; David L. Kendler, MD, Prohealth Clinical Research Centre, Vancouver, British Columbia, Canada; William D. Leslie, MD, MSc, University of Manitoba, Winnipeg, Canada; Suzanne Morin, MD, MSc, McGill University, Montreal, Quebec, Canada; Alexandra Papaioannou, MD, MSc, McMaster University, Hamilton, Ontario, Canada; Kerry Siminoski, MD, University of Alberta, Edmonton.
Disclaimer: The analyses and conclusions in this report reflect the opinions of individual experts and not their affiliated organizations.