amjepidajeAmerican Journal of Epidemiology1476-62560002-9262Oxford University Press10.1093/aje/kwq215ORIGINAL CONTRIBUTIONSPopulation-Level Impact of Osteoporotic Fractures on Mortality and Trends Over Time: A Nationwide Analysis of Vital Statistics for France, 1968–2004ZiadéNellyJouglaEricCosteJoël**Correspondence to Dr. Joël Coste, Biostatistics and Epidemiology Unit, Pavillon Saint-Jacques, Hôpital Cochin, 27 rue du Faubourg Saint-Jacques, 75674 Paris Cedex 14, France (e-mail: coste@cochin.univ-paris5.fr).1510201012820101728942951432010962010American Journal of Epidemiology © The Author 2010. Published by Oxford University Press on behalf of the Johns Hopkins Bloomberg School of Public Health. All rights reserved. For permissions, please e-mail: journals.permissions@oxfordjournals.org.2010Osteoporotic fractures are one of the leading causes of death in the elderly population, but mortality may have been reduced by the advances in management and prevention during recent decades. The authors analyzed the population-level impact of these fractures on mortality in France from 1968 to 2004. About 20 million death certificates registered in metropolitan France from 1968 to 2004 were analyzed. Osteoporotic fractures were identified by using a previously developed methodology. Age-specific and standardized mortality rates were calculated by site of fracture and sex, and time trends were evaluated. Associated causes of death were compared between the extreme periods of the study by the observed/expected pairs method; 440,890 (2.2%) death certificates reported an osteoporotic fracture. Osteoporotic fractures overall, particularly hip and skull fractures, declined by half during the study period, exceeding the decline in general mortality and resulting in fracture-deceased subjects being older. However, pelvis, vertebral, and rib fractures became more frequent. Associated causes of death increased with time, except for decubitus ulcers, indicating a change in the pattern of the death process. Despite a 50% decline, osteoporotic fractures still have a significant impact on mortality. The pattern of the death process has changed, with an increased role for comorbidities.cause of deathdeath certificatesfractures, boneInternational Classification of Diseasesmortalityosteoporosisvital statisticsOsteoporotic fractures are one of the leading causes of death in the elderly population (1) and a major contributor to the burden of disease (2). This burden is expected to increase as life expectancy increases (3), and it is a growing issue in all aging societies (4, 5).Treatments have been developed to prevent osteoporotic fractures and related mortality. Hip prosthesis, becoming widespread in the 1970s, reduced complications due to rapid mobilization and was further improved by effective anticoagulation and better anesthetic procedures (6–8). More recently, medical treatments have been widely used to reduce the incidence of osteoporotic fractures; they include postmenopausal hormonal replacement therapy (9) and bone-protective agents such as bisphosphonates (10).Despite the growing importance of osteoporotic fractures and the advances in their management, few studies have addressed the consequences for mortality in large populations or the changes in these consequences. Most available information comes from variously assembled cohorts, followed for short periods, and assessing mortality after a fracture event (11–15).Assessing the impact of osteoporotic fractures on the mortality of the entire population (i.e., how frequently osteoporotic fractures contribute to death) is valuable from a public health perspective. Nationwide vital registration systems provide consistent, relevant, and nationwide data, and these systems are available, exhaustive, low cost, and suitable for international comparisons. This approach requires identification and analysis of causes of death reported on death certificates and coded according to the International Classification of Diseases (ICD), following rules issued by the World Health Organization.We have developed a methodology (16) for estimating the contribution of osteoporotic fractures to mortality in the general population, based on the analysis of cross-classifications of essential components related to current definitions of osteoporotic fractures (age, sex, site, and mechanism of fracture). Here, we use this methodology to investigate the impact of osteoporotic fractures on mortality in France between 1968 and 2004. We also examined changes in the patterns of associated causes of death during the same period.MATERIALS AND METHODSData sourceMortality data were obtained from the Center of Epidemiology on Medical Causes of Death (“CépiDc”; INSERM), where all French death certificates are collected, checked for quality, and coded. All data from 1968 to 2004 for metropolitan France were analyzed. Variables including sex, age at death, and all coded causes of death were recorded.Death certification and codingIn France, death certification is mandatory and necessarily performed by a medical doctor. Death certificates are exhaustive, and the data since 1968 are available (17). The certificate is established according to World Health Organization recommendations, and the medical cause of death section consists of 2 parts. The first is used for reporting the sequence of events leading to death, with the underlying cause stated last. The second part reports the other contributing causes of death (“associated causes of death”). The medical information provided by the French death records is detailed; a mean of 3.1 causes was reported for each death in 2003 (17).Coding consists of attributing a digital code to each cause of death, according to the latest revision of the ICD, and then selecting the underlying cause according to World Health Organization rules. During the study period, 3 changes were made to the French coding system. First, the ICD was revised twice: The Eighth Revision was used from 1968 to 1978 (ICD-8), the Ninth Revision from 1979 to 1999 (ICD-9), and the Tenth Revision from year 2000 (ICD-10). Second, since the year 2000, 80% of death certificates have been coded by software that also attributes the underlying cause. The knowledge base included in the French software uses the Mortality Medical Data System decision tables of the National Center for Health Statistics, US Centers for Disease Control and Prevention. Third, the number of coded contributing causes was increased from 1 (1968–1978) to 2 (1979–1999) and then to be exhaustive (since 2000).Two types of ICD codes are used to describe injuries: one describes the nature of the injury or injury code (I-code, e.g., hip fracture), and the other describes the mechanism of the injury (E-code, e.g., fall). According to ICD rules, the mechanism of injury (E-code) is always selected as the underlying cause of death and is the one routinely published. If the certifying medical doctor omits an E-code, a nonspecific code is automatically assigned (X59 in ICD-10) and selected as the initial cause of death.Method for identification of an osteoporotic fracture-related deathWe previously developed a multiple-step method to estimate the contribution of osteoporotic fractures to mortality in the general population (16). Current definitions of osteoporotic fractures were identified, and their components were individualized and translated into an operational definition using age, sex, site, and mechanism of fracture. The main results were, first, the identification of an age threshold of 70 years, after which fractures are reported with increasing frequency. Second, the patterns of rates for men and women were similar. Third, a similar pattern of rates was found for all sites of fracture, including skull fractures, suggesting that all fractures, and not merely classical “osteoporotic” fractures, should be considered. Fourth, the mechanism of injury was useful for categorizing fractures as osteoporotic, by allowing the exclusion of high-energy fractures. According to current definitions of osteoporotic fractures, high-energy fractures should be excluded and only “fragility-related” fractures considered. However, certifying medical doctors may omit the mechanism of fracture; we adjusted for this underdeclaration of the mechanisms of fractures by calculating a corrective factor for each site of fracture, by sex. Fractures occurring between 20 and 30 years of age are obviously nonosteoporotic, so the sensitivity of the high-energy code for detecting high-energy fractures (or corrective factor) was calculated as the number of reports of the high-energy code divided by the total number of fractures in this age category. High-energy fractures were multiplied by this site- and sex-specific corrective factor to control for their underestimation. The estimated number of osteoporotic fractures was, thus, all fractures occurring after the age of 70 years minus the corrected number of high-energy fractures.The method was developed for ICD-10, so I-codes and E-codes in ICD-8 and ICD-9 were identified (Web Table 1; this information is described in the first of 2 supplementary tables referred to as ‘‘Web Table’’ in the text and posted on the Journal’s website (http://aje.oupjournals.org/).). Sites of fractures were individualized as follows: hip, vertebrae, pelvis, ribs, skull, multiple, and other. Bridge-coding tables allowed the estimation of the comparability ratios between ICD versions (18). We performed a detailed dissection of the codes and multiple simulations, comparing the numbers of mentions of the related causes of death in the years bracketing ICD modifications: 1974–1978, 1979–1983, 1995–1999, and 2000–2004. Some discrepancies were found. First, fractures of the pelvis were well individualized in ICD-8 and ICD-9 but were grouped with lumbar spine fractures in ICD-10, artificially decreasing mentions for pelvis fractures and increasing those for spine fractures over time. Second, an E-code for unspecified mechanism of fracture was included in the falls section in ICD-8 and ICD-9 (E887) but not in ICD-10 (X59: exposure to unspecified factor), artificially decreasing the reporting of falls over time. On the basis of this analysis, fractures of the lumbar spine and pelvis were grouped together because this minimized variation associated with the transition from ICD-9 to ICD-10, and the change of classification of the unspecified E-code E887/X59 was taken into account.Table 1.Standardized Mortality Rates per 100,000 Persons Through the Study Period, by Sex and Site of Osteoporotic Fracture, France, 1968–2004a1968–19721973–19781979–19841985–19891990–19941995–19992000–2004% ChangebMen    Hip153.24162.12139.54108.0986.2574.6870.37−54.08    Pelvis + vertebrae3.996.887.866.657.357.587.9599.24    Skull4.856.164.252.773.152.41.8−63.01    Ribs2.975.055.364.614.254.385.7295.52    Multiple1.411.621.611.120.930.771.720.83    Others27.2529.9729.4422.919.8918.8221.95−19.47    All193.72211.8188.05146.12121.82108.63109.49−43.48    External causesc372.5368.27387.6397.33362.08339.57268.34−27.96    General mortalityd8,888.58,274.637,852.757,873.937,033.156,6876,302.01−29.05Women    Hip294.79304.9240.54196.51142.8109.8102.38−65.27    Pelvis + vertebrae5.699.0110.39.529.389.779.9675.01    Skull3.483.453.123.022.821.931.79−48.63    Ribs1.482.682.682.692.382.453.74152.21    Multiple2.32.412.72.181.531.564.1982.34    Others45.0250.9750.3542.213634.342.04−6.62    All352.76373.43309.69256.13194.9159.8164.1−53.48    External causesc361.78368.03347.88341.84310.29209.46176.05−51.34    General mortalityd6,637.046,244.725,893.975,875.375,476.015,246.214,949.3−25.43aMortality rates are related to external causes of death, and general mortality rates are included for comparison.bComparison of mortality rates between 1968–1972 (reference) and 2000–2004. All changes were statistically significant (Pchi-square < 0.0001), except for multiple fractures in men (P = 0.10).cNumber of death certificates for subjects over 70 years of age reporting any external cause of death × 100,000/general population over 70 years of age.dTotal number of death certificates for subjects aged over 70 years × 100,000/general population over 70 years of age.We also verified that the previously identified thresholds were also valid for ICD-8 and ICD-9. Applying the same methodology, we found that the age threshold of 70 years can still be used. Because the high-energy threshold used to calculate the corrective factor allowing exclusion of high-energy fractures varied with time, it was calculated for 5-year periods for each sex and site of fracture (Web Figure 1) (http://aje.oupjournals.org/).Figure 1.Standardized mortality rates, by fracture site, France, 1968–2004. Continuous line for men and dashed lines for women. A, all fractures (right axis scale: all-cause mortality rate, represented in open squares for men and open circles for women); B, hip; C, pelvis and vertebrae; D, ribs; E, others; and F, skull.Death ratesCrude death rates were calculated annually by site, sex, and age, using demographic data provided by the National Institute for Statistics and Economic Studies for estimating the denominator. Age-adjusted mortality rates were calculated by using a direct method of standardization to account for differences in age composition and the French reference population of 1990 (a census year situated near the middle of the study period). Age-specific death rates were calculated for osteoporotic fractures by 5-year age group (70–74, 75–79, 80–84, 85–89, 90–94, and older than 94 years).Statistical analysisStandardized mortality rates were compared between the beginning (1968–1972) and the end (2000–2004) of the study period with a χ2 test (19). The trends of age-specific death rates were estimated by Poisson regression analysis (20), using the “85–89 years” age group as reference. The mean age at death was calculated yearly by using death certificates with osteoporotic fractures and compared with the mean age at death for the entire population after the age of 70 years; trends in the difference were studied by linear regression.Measurements of the association between osteoporotic fractures and other causes of death used the ratio method of Israel et al. (21): The ratio of the number of observed pairs of causes of death to the number of expected pairs of causes was calculated, based on the assumption of independence. An observed/expected ratio (O/E ratio) greater (less) than 1 indicates that more (fewer) deaths with paired causes were reported than would be expected by chance if the paired causes were independent. Chi-square statistics were used to test the null hypothesis H0: O/E ratio = 1. Two time periods were compared: 1968–1972 and 2000–2004, by site of fracture. For each O/E ratio, 95% confidence intervals were defined and, in the absence of an appropriate statistical test, the overlap of these intervals was used to assess statistical significance.RESULTSA total of 19,989,760 death certificates were registered in metropolitan France between 1968 and 2004. Of these, 440,890 (2.2%) reported an osteoporotic fracture, and 306,245 (69% of osteoporotic fractures and 1.5% of all death certificates) reported a hip fracture. During the period 2000–2004, 46,849 (1.78%) death certificates reported an osteoporotic fracture.Fewer men than women died with osteoporotic fractures, with a standardized mortality sex ratio (men/women) of 0.61 for the entire study period. This ratio was not stable across sites between 1968 and 2004; it varied from 0.52 to 0.69 for hip, from 0.70 to 0.80 for pelvis and vertebrae, from 2.06 to 1.52 for ribs, and from 1.41 to 1.02 for skull, thus reducing the gap between men and women.Hip fractures were the most frequently reported on death certificates followed by pelvis and vertebrae, skull, ribs, and multiple fractures. Standardized mortality rates changed through the study period and between the different sites (Table 1). There was a statistically significant halving of osteoporotic fractures overall (−43.5% in men and −53.5% in women) between 1968 and 2004, mainly due to decreasing reports of hip and skull fractures. The pattern was, however, similar for most sites: a small increase early in the study period and a subsequent continuous decline from the mid-1970s (Figure 1). This decline paralleled but exceeded both the 25%–29% decline in general mortality and the decline in external causes-related mortality (in France, mainly road traffic accidents).Although the decline in skull fractures was substantial, they constitute only a minority of all fractures (1.49% between 1968 and 2004), and their contribution to the general trend was small. Unlike hip and skull fractures, reports of pelvis and vertebral fracture increased gradually with time. Rib and “other” fractures increased initially and then stabilized in the 1980s (the sharp increase in the year 2000 is largely an artifact of the exhaustive coding of causes from that date).A statistically significant estimate was found for the third (cubic) power of year, indicating a polynomial—and not linear—trend over time. There were 2 inflexion points: an increase in mortality rates for all fractures at the beginning of the study period, followed by a rapid decrease that slowed slightly at the end.Age-specific, 5-year mortality rates associated with osteoporotic fractures were calculated (Figures 2 and 3). For all osteoporotic fractures, mortality rates decreased for all age categories, most markedly in younger age groups. Poisson regression estimates for decreasing trends with time were statistically significant for hip and “other” fractures in all age groups and for skull fractures in men younger than 85 years (Web Table 2). The increasing trends for pelvis, vertebrae, and rib fractures were apparently more pronounced in the older age groups (not statistically significant except for pelvis and vertebrae fractures in women).Table 2.Ratios of Actual to Expected Number of Pairs of Causes of Death Including Low-Energy Fractures, for ICD Chapters and Selected Causes, France, 1968–1972 and 2000–2004a1968–19722000–2004All Deaths, no.Observed, no.Expected, no.Observed/ Expected Ratio95% Confidence IntervalAll Deaths, no.Observed, no.Expected, no.Observed/ Expected Ratio95% Confidence IntervalICD chapters    Infectious and parasitic diseases23,551257837.000.310.27, 0.35140,1683,6035,139.170.700.68, 0.72    Neoplasms274,6441,0149,760.780.100.10, 0.11518,5574,73319,012.570.250.24, 0.26    Diseases of the blood6,11171217.180.330.26, 0.4144,0831,2231,616.280.760.72, 0.81    Endocrine diseases105,8882,0973,763.230.560.53, 0.58252,6307,4909,262.520.810.79, 0.83    Diseases of the nervous system90,5432,0053,217.880.620.06, 0.65313,77710,99111,504.440.960.94, 0.97    Diseases of the circulatory system848,8589,90730,168.210.330.32, 0.331,059,49126,33838,845.580.680.67, 0.69    Diseases of the respiratory system205,0962,7847,289.060.380.37, 0.40428,16211,17515,698.290.710.70, 0.73    Diseases of the digestive system87,0035183,092.070.170.15, 0.18169,0943,4266,199.730.550.53, 0.57    Diseases of the genitourinary system37,8754031,346.070.300.27, 0.33142,1553,6375,212.020.700.68, 0.72    Diseases of the musculoskeletal system11,040209392.360.530.46, 0.6140,8081,7661,496.021.181.13, 1.24    Diseases of skin and subcutaneous tissue12,4121,031441.122.342.20, 2.4854,5233,2811,999.051.641.59, 1.70    Others not classified elsewhere485,3089,36317,247.730.540.53, 0.55971,22928,00835,609.510.790.78, 0.80Selected causes of death (ranked by association with osteoporotic fractures in 2000–2004)    Decubitus ulcer10,6011,010376.762.682.52, 2.8544,6492,9901,637.031.831.76, 1.89    Pulmonary embolism14,675246521.550.470.41, 0.5357,1292,9582,094.601.411.36, 1.46    Gastrointestinal ulcer6,76041240.250.170.12, 0.239,013357330.461.080.97, 1.20    Dementia19,205484682.540.710.65, 0.77113,8444,4744,174.021.071.04, 1.11    Rheumatoid arthritis3,76157133.670.430.32, 0.545,363201196.631.020.89, 1.17    Ischemic cardiopathy141,2787345,020.990.150.14, 0.16157,6205,0435,779.040.870.85, 0.90    Pyelonephritis1,0381236.890.330.17, 0.544,546134166.680.800.67, 0.95    Hypertensive diseases25,256320897.590.360.32, 0.40173,1524,9576,348.510.780.76, 0.80    Pneumonia42,8211,0171,521.850.670.63, 0.71153,4414,0685,625.820.720.70, 0.75    Gastrointestinal hemorrhage1,574855.940.140.06, 0.2628,9546641,061.580.630.58, 0.67    Renal failure7,19146255.570.180.13, 0.24115,3142,6554,227.920.630.60, 0.65    Diabetes47,0561,0691,672.360.640.60, 0.6868,7081,4922,519.140.590.56, 0.62    Malignant neoplasms253,1938908,998.420.100.09, 0.11428,3313,57415,704.490.230.22, 0.24Total no. of times reported2,272,97580,7814,253,093155,937Total no. of death certificates1,630,36051,1221,930,42350,171Abbreviation: ICD, International Classification of Diseases.aNinety-five percent confidence intervals for the observed/expected ratio are given in a separate column. An observed/expected ratio of >1 indicates a positive association with low-energy fractures.Figure 2.Age-specific mortality rates for men on a log10 scale, France, 1968–2004. Diamonds, ages 70–74 years; squares, ages 75–79 years; triangles, ages 80–84 years; dashed line, ages 85–89 years (reference); crosses, ages 90–94 years; circles, ages >94 years. All interactions between time and age groups are significant (P < 0.05) except for pelvis and vertebrae, ribs, and skull fractures in the >90-year age groups. A, all fractures; B, hip; C, pelvis and vertebrae; D, ribs; E, others; and F, skull.Figure 3.Age-specific mortality rates for women on a log10 scale, France, 1968–2004. Diamonds, ages 70–74 years; squares, ages 75–79 years; triangles, ages 80–84 years; dashed line, ages 85–89 years (reference); crosses, ages 90–94 years; circles, ages >94 years. All interactions between time and age groups are significant (P < 0.05) except for the pelvis and vertebrae (90–94-year age group), ribs (>90-year age groups), and skull fractures. A, all fractures; B, hip; C, pelvis and vertebrae; D, ribs; E, others; and F, skull.The mean age at death increased continuously. The mean age at death associated with osteoporotic fractures increased more (P < 0.0001) than that for general mortality in both sexes (with a difference of 2.30 years for men and 1.77 years for women) (Figure 4).Figure 4.Mean age at death, France, 1968–2004. Osteoporotic fracture-related mortality is represented with continuous lines for men and dashed lines for women; the all-cause mortality rate is represented by open squares (men) and open circles (women).The number of causes of death for death certificates reporting osteoporotic fracture doubled from an average of 1.6 causes per certificate in 1968–1972 to 3.1 in 2000–2004 (contrasting with 1.4 in 1968–1972 and 2.2 in 2000–2004 for all certificates). The incidence of comorbidities in cases of fracture-related deaths thus increased substantially.Diseases of the circulatory system were the comorbidities most frequently associated with osteoporotic fractures (19.4% of death certificates in 1968–1972 and 53.2% in 2000–2004). The most common were ischemic cardiopathy and hypertensive diseases. Other frequent comorbidities were diabetes, pneumonia, and decubitus ulcer in 1968–1972 and pneumonia, renal failure, and dementia in 2000–2004 (Table 2). An O/E ratio greater than 1, indicating a positive association with osteoporotic fractures, was found for diseases of the musculoskeletal system, decubitus ulcer, pulmonary embolism, and dementia in the period 2000–2004. All associations (O/E ratios) increased with time, a trend also found for all sites studied individually (data not shown), with 2 exceptions: Decubitus ulcer, which was significant in 1968–1972, decreased significantly with time for all sites, and the diseases of the respiratory system (including pneumonia) decreased only for rib fracture-associated deaths.DISCUSSIONWe analyzed nearly 20 million death certificates covering 37 years and identified a significant contribution of osteoporotic fractures to nationwide mortality. Osteoporotic fractures were reported on 440,890 death certificates (2.2% during the entire period), supporting the established belief that they represent a major public health problem. By comparison, during the same period, diabetes, renal failure, and septicemia were mentioned on 430,967, 312,881, and 222,665 death certificates, respectively.Between 1968 and 2004, the mortality rates associated with osteoporotic fractures were reduced by about half, in both sexes and in all age groups, largely due to changes associated with hip and skull fractures. This decrease was apparent despite the increase in the number of coded causes of death per certificate, indicating a true significant decline. The decline exceeded that of general mortality, translating into a significant increase over time in the mean age at death from osteoporotic fractures. However, mortality rates associated with fractures at some sites increased. In some cases (rib and multiple fractures), this was a consequence of the change in coding rules with the transition to ICD-10 in 2000 and the increase of the total number of coded causes; in others (pelvis and vertebrae), it was not. These sites represented a minority of the total, such that these increases had little impact on osteoporotic fracture-associated mortality overall.The decline in the incidence of and mortality associated with hip fractures has been reported in large cohort studies (22, 23) and posthospitalization follow-up studies (15, 22). However, none of these studies fully explains the causes of the decrease in mortality. Our study allows various hypotheses concerning management and behavioral changes to be considered. During recent decades, temporally compatible with the observed falling mortality rates, surgical management and postsurgical care of hip fractures have clearly improved (7, 8). The use of antiosteoporotic treatments, initially hormone replacement therapy, increased steeply in France in the late 1980s and early 1990s (9), but this can explain only a small part of the change because the decrease started in the mid-1970s and affected both women and men (albeit less so). The decrease in hip fracture rates reported elsewhere was similarly much too large to be explained by antiosteoporotic medication (22, 24). Other possible causes include changes in smoking habits, improved general health, calcium and vitamin D supplementation, regular exercise, awareness of falls, and moderating alcohol intake (clearly documented in France) (25). The absence of age- and sex-specific differences in mortality trends is in favor of a more general cause, for example, improved postsurgical management and behavioral changes, rather than a medication-related effect. One potential bias should be addressed. Increased awareness of the severity of osteoporotic fractures and promotion of their medical management during the later years of the study period (10) may have increased the frequency of reporting of some fracture types (vertebral and pelvis fractures). The consequent increased reporting of less severe sites not traditionally associated with high risk of mortality may have decreased the observed general mortality rates associated with fractures. This bias was controlled for by analyzing each site of fracture separately. The observed trends in mortality rates were clearly site specific. Wrist fractures, clearly related to osteoporosis in all awareness campaigns, were seldom identified in our sample, indicating no artificial effect on the physicians’ certification habits. There was a 50% reduction of mortality rates associated with hip and skull fractures over the study period. Hip fracture may lead to death in the immediate postfracture period, so failure to report it on death certificates is less likely, even early during the study period.The skull site is not traditionally associated with osteoporotic fractures. However, our data suggest that some skull fractures, occurring at an older age and not associated with a high-energy mechanism, follow a pattern similar to that of all other osteoporotic fractures. Therefore, skull fractures were included in the analysis, but their contribution to the overall trends is small (1.49% of all osteoporotic fractures).Interestingly, we found an increase with time for all comorbidities associated with osteoporotic fractures at the moment of death, with 2 exceptions. For decubitus ulcer, the association weakened, reflecting a better mobility after surgery, and for diseases of the respiratory system, the association weakened in rib fracture deaths only. Similar increases in comorbidities have been observed by others (22), suggesting a change in mortality patterns, with comorbidities increasingly contributing to the death process, as fractured subjects at risk of death became older and frailer. Some causes of death were less associated with fractures than expected. The values are consistent throughout the study period, so medical certification errors are unlikely to be the cause. Some causes of death may “compete” at 2 levels. The first is true competition, leading to different mortality patterns (26). Patterns of mortality may differ according to age, sex, and time period, with one main pattern associating neoplasms and cardiovascular and digestive diseases in those aged 45–84 years and another pattern represented by injury and poisoning. The second level is an artificial competition in the mind of a certifying medical doctor. If death is associated with a “serious” condition, such as cancer, severe infection, or cardiovascular disease, all other causes, and not only fractures, are underdeclared (the cause is serious enough, and the certifying physician does not need to seek other associated causes).Our study has several strengths including the analysis of a very large set of death certificates exhaustively covering a large population over a long period (1968–2004). Nevertheless, it has some limitations. Diagnosis errors, unavailable medical records at the time of certification, and information missing from death certificates have all been documented (27). However, our study did not aim to estimate all osteoporotic fractures leading to death in France but, rather, it attempts to identify cases in which the certifying physician considered the fracture to have contributed to death. Moreover, underreporting is probably less frequent for injury-related causes of deaths such as fractures than for chronic medical conditions. Another limitation is that death certificate data represent only mortality rapidly following a fracture. Although the risk remains elevated up to 10 years following a hip fracture (12), the highest risk of mortality is immediately after the fracture event (4, 11). In a recent systematic review (11), excess mortality after a hip fracture ranged from 5.9% (among patients aged 50–74 years) to 50% during the first year after fracture and remained elevated for many years. One-quarter to one-third of the mortality occurred during the first month after fracture, around half within 3 months, and 70% within 6 months. The mechanism of fractures was poorly documented in our study. In particular, low-energy mechanisms were seldom reported. Work with the Swedish population-based register showed that low-energy trauma was responsible for 53% of fractures in those over 50 years of age and for 80% in those over 75 years of age between 1993 and 2004 (13). A better understanding of injury mechanisms is necessary for preventive interventions, and better reporting of mechanisms on the death certificate would be valuable. Coding the causes of death by using the ICD may lead to errors and inaccuracies. However, this is unlikely for fracture data, because there is no ambiguity in ICD codes for such injuries.In conclusion, we found a significant decline of reports of osteoporotic fractures on death certificates between 1968 and 2004, starting in the 1970s, in both genders and all age groups. This was driven mainly by the decline of hip and skull fractures and coincided with better surgical and postsurgical management and general behavioral changes. Nevertheless, the impact of osteoporotic fractures on mortality remains significant, and the trends through time were opposite for the sites pelvis, vertebrae, and ribs. Finally, the mortality pattern changed, suggesting an increasing role for comorbidities in the death process as subjects become older and frailer.AbbreviationsICDInternational Classification of DiseasesICD-8ICD-9, and ICD-10, International Classification of Diseases, Eighth, Ninth, and Tenth Revisions, respectivelyO/E ratioobserved/expected ratioAuthor affiliations: Research unit APEMAC, EA 4360, Nancy-Université, Université Paris-Descartes, Université Metz Paul Verlaine, Paris, France (Joël Coste, Nelly Ziadé); Centre d’épidémiologie sur les causes médicales de décès (CépiDc), National Institute for Health and Medical Research, Le Vésinet, France (Eric Jougla); and Faculté de Médecine, Université Saint-Joseph, Beirut, Lebanon (Nelly Ziadé).Conflict of interest: none declared.1.GenantHKCooperCPoorGInterim report and recommendations of the World Health Organization Task-Force for OsteoporosisOsteoporos Int19991042592642.JohnellOKanisJAJonssonBThe burden of hospitalised fractures in SwedenOsteoporos Int20051622222283.CummingsSRMeltonLJEpidemiology and outcomes of osteoporotic fracturesLancet20023599319176117674.JohnellOKanisJEpidemiology of osteoporotic fracturesOsteoporos Int200516suppl 2S3S75.OinumaTSakumaMEndoNSecular change of the incidence of four fracture types associated with senile osteoporosis in Sado, Japan: the results of a 3-year surveyJ Bone Miner Metab201028155596.ScheerlinckTHaentjensPFractures de l'extrémité supérieure du fémur chez l'adulte. (In French)2003Paris, FranceScientific and Medical Publishing Elsevier SAS7.HandollHHFarrarMJMcBirnieJHeparin, low molecular weight heparin and physical methods for preventing deep vein thrombosis and pulmonary embolism following surgery for hip fracturesCochrane Database Syst Rev.20002CD0003058.HandollHHSherringtonCMobilisation strategies after hip fracture surgery in adultsCochrane Database Syst Rev.20071CD0017049.RingaVJaussentIGuégenRTrends in the use of hormone replacement therapy in eastern France between 1986 and 1993Eur J Public Health19999430030510.StaffordRSDrielingRLHershALNational trends in osteoporosis visits and osteoporosis treatment, 1988–2003Arch Intern Med2004164141525153011.AbrahamsenBvan StaaTArielyRExcess mortality following hip fracture: a systematic epidemiological reviewOsteoporos Int200920101633165012.BliucDNguyenNDMilchVEMortality risk associated with low-trauma osteoporotic fracture and subsequent fracture in men and womenJAMA2009301551352113.BergströmUBjörnstigUStenlundHFracture mechanisms and fracture pattern in men and women aged 50 years and older: a study of a 12-year population-based injury register, UmeåSweden. Osteoporos Int20081991267127314.VestergaardPRejnmarkLMosekildeLHas mortality after a hip fracture increased?J Am Geriatr Soc.200755111720172615.RobertsSEGoldacreMJTime trends and demography of mortality after fractured neck of femur in an English population, 1968–98: database studyBMJ2003327741877177516.ZiadéNJouglaECosteJUsing vital statistics to estimate the population-level impact of osteoporotic fractures on mortality based on death certificates, with an application to France(2000–2004). BMC Public Health. 2009;9:344. (doi:10.1186/1471-2458-9-344)17.PavillonGLaurentFCertification et codification des causes médicales de décès. (In French)Bull Epidémiol Hebd200330/3113413818.PavillonGBoileauJRenaudGBridge coding ICD9-ICD10 and effects on French mortality data. Presented at the WHO Family of International Classifications Network MeetingReykjavik, Iceland, October 24–30, 200419.CarriereKCRoosLLComparing standardized rates of eventsAm J Epidemiol1994140547248220.RodríguezGPoisson models for count data [electronic monograph]2007Princeton, NJPrinceton University(http://data.princeton.edu/wws509/notes/c4.pdf). (Accessed November 2, 2009)21.IsraelRARosenbergHMCurtinLRAnalytical potential for multiple cause-of-death dataAm J Epidemiol1986124216117922.BrauerCACoca-PerraillonMCutlerDMIncidence and mortality of hip fractures in the United StatesJAMA2009302141573157923.LeslieWDO'DonnellSJeanSTrends in hip fracture rates in CanadaJAMA2009302888388924.AbrahamsenBVestergaardPDeclining incidence of hip fractures and the extent of use of anti-osteoporotic therapy in Denmark 1997–2006Osteoporos Int201021337338025.PyöräläETrends in alcohol consumption in Spain, Portugal, France and Italy from the 1950s until the 1980sBr J Addict199085446947726.CosteJBernardinEJouglaEPatterns of mortality and their changes in France (1968–99): insights into the structure of diseases leading to death and epidemiological transition in an industrialised countryJ Epidemiol Community Health2006601194595527.SwiftBWestKDeath certification: an audit of practice entering the 21st centuryJ Clin Pathol2002554275279