Osteoporosis International

, Volume 22, Issue 2, pp 685–692

Epidemiology of hip fractures in Austria: evidence for a change in the secular trend

Authors

    • Department of Internal Medicine, Division of Endocrinology and Nuclear MedicineMedical University of Graz
  • A. Svedbom
    • I3 Innovus
    • Medical Management CentreKarolinska Institutet
  • A. Fahrleitner-Pammer
    • Department of Internal Medicine, Division of Endocrinology and Nuclear MedicineMedical University of Graz
  • T. Pieber
    • Department of Internal Medicine, Division of Endocrinology and Nuclear MedicineMedical University of Graz
  • H. Resch
    • Medical Department IISt. Vincent Hospital
  • E. Zwettler
    • Ludwig Boltzmann Institute of OsteologyHanusch Hospital Vienna
  • M. Chandran
    • Department of EndocrinologySingapore General Hospital
  • F. Borgström
    • I3 Innovus
    • Medical Management CentreKarolinska Institutet
Original Article

DOI: 10.1007/s00198-010-1271-9

Cite this article as:
Dimai, H.P., Svedbom, A., Fahrleitner-Pammer, A. et al. Osteoporos Int (2011) 22: 685. doi:10.1007/s00198-010-1271-9

Abstract

Summary

Hip fracture incidence rates in Austria over a period of 20 years (1989–2008) were assessed. Age-standardized incidence rates increased until 2005 but decreased thereafter. This change in the secular trend was primarily driven by a decrease in hip fracture incidence in women.

Introduction

The aim of the prevailing study was to assess the incidence rates of hip fractures including changes in trend in the Austrian population over a period of 20 years (1989–2008).

Methods

The number of hip fractures was obtained from the Austrian Hospital Discharge Register for the entire population ≥50 years of age. A correction factor for multiple registrations of the same diagnosis was determined. Incidence rates (cases per 100,000) of hip fracture were calculated in 5-year age intervals. To analyze the overall change in hip fracture for the period, average annual change expressed as incidence rate ratios (IRRs) was calculated.

Results

The age-standardized incidence in women increased until 2005, from 493 to 642, and decreased thereafter. In men, it increased at a measured pace until 2006, from 192 to 280, and decreased thereafter with a slight rebound in 2008. The age-standardized incidence in the entire population increased until 2005, from 376 to 496, and decreased thereafter. The IRR for the last 3 years (2006–2008) was significantly below the IRR for the first 17 years (0.94, ρ < 0.01), driven by a lower IRR in women (0.91, ρ < 0.01) and to a lesser extent by a lower IRR (not statistically significant) in men (0.96, ρ < 0.15).

Conclusion

The present study indicates that since 2006, age-standardized incidence of hip fractures has been declining in the Austrian population aged 50 years and above. This reversal in the secular trend has primarily been driven by a decrease in hip fracture incidence in women.

Keywords

Age-adjusted incidenceCrude incidenceHip fractureOsteoporosisTrend analyses

Introduction

According to a recommendation of a consensus conference held in 1991, osteoporosis is defined as a “systemic skeletal disease characterized by low bone mass and microarchitectural deterioration of bone tissue with a consequent increase in bone fragility and susceptibility to fracture” [1]. Fractures are the most significant clinical consequences of this disease. The average lifetime risk in a 50-year-old person to experience an osteoporotic fracture has been estimated at 40–50% for women and at 13–22% for men [2]. Osteoporotic fractures are associated with significant morbidity, and both hip and vertebral osteoporotic fractures are also associated with excess mortality [3]. Hip fracture incidence has been shown to increase exponentially with age, and the worldwide hip fracture incidence has been estimated to increase dramatically within the next decades, mainly due to the increasing number of elderly people [4].

However, in the recent past, there has also been increasing evidence that in some populations, this secular trend is no longer prevailing. The rise in hip fracture incidence in Finland between the early 1970s and the late 1990s has been followed by a clear decrease [5]. Similar changes have been observed for populations in Sweden, Norway, Denmark, Switzerland, Ontario (Canada), and some North American regions [610].

Austria, located in the southern part of Central Europe, counted some 8.3 million inhabitants at the beginning of 2008. The present day age pyramid shows a narrow base due to a reduction in birth rates. However, similar to other countries in the European Union, the percentage of the senior population, 50 years of age and older, is increasing. Within the elderly Austrian population, the number of women clearly exceeds the number of men which is due not only to the higher life expectancy of women but also to the large number of men who died in World War II [11].

The epidemiology of hip fractures in Austria between 1994 and 2006 has been investigated recently by Mann et al. [12]. However, the methodology has three limitations. Firstly, the number of hip fractures has been derived from the Austrian Hospital Discharge Register (AHDR) but without having considered a correction factor for several-fold registrations for the same diagnosis. Secondly, the time period under consideration is relatively short. Thirdly, the measures of change used—average annual changes (incidence rate ratios (IRRs))—were calculated from the data for the entire period, making the analysis susceptible to large changes in any given time period, thus potentially limiting the use of the analysis to make inferences about changes in trend. The Appendix in Mann et al. shows that the age-standardized incidence (fractures per 100,000 person years) in women increased from 557 in 1994 to 630 in 1999 and fluctuated between 613 and 643 thereafter, with the last observation (2006) at 623, indicating that this may be a problem, potentially invalidating their conclusion of no off or downward trend in hip fracture incidence in the Austrian elderly population.

The aim of the present study was to assess the incidence of hip fractures in the Austrian population aged 50 years and above, over a period of two decades, using a correction factor for multiple registrations of the same diagnosis in the AHDR and to estimate the trend in average annual change.

Materials and methods

It has been mandatory for all hospitals within Austria to record discharge diagnoses by using the code classes of the International Classification of Diseases (ICD) from 1989. This information is registered into the AHDR, which is maintained by Statistics Austria [13].

For the purposes of this study, hip fracture events were obtained from the AHDR for the entire Austrian population ≥50 years of age between 1989 and 2008. For identification of hip fractures, the code classes 820 (ICD9, 1989–2000) and S72.0, S72.1, and S72.2 (ICD10, 2001–2008) have been applied. The data derived from the AHDR provides information on the number of fractures in 5-year intervals (50–54, 55–59, 60–64, 65–69, 70–74, 75–79, 80–84, 85–90, 90–95, and 95-99) per year stratified by sex and province (Upper Austria, Lower Austria, Styria, Carinthia, Salzburg, Tyrol, Vorarlberg, Burgenland, and Vienna).

In the AHDR, as in many other registries, it is possible that patients have multiple registrations for the same diagnosis. The number of patients with multiple registrations in the register cannot be reviewed given that the data are presented in aggregated form. To assess the proportion of diagnoses with multiple registrations, the medical records of hip fracture patients admitted to the trauma units of two large Austrian hospitals (the University Hospital of Graz and the Hanusch Hospital of Vienna) were reviewed. These two units have been randomly chosen from a total of 67 trauma units in Austria. The number of hip fractures treated in these trauma units represent roughly 5% of all hip fractures treated in Austria per year. The proportion of diagnoses with multiple registrations found in these two hospitals might therefore provide a reasonable estimate of the percentage of multiple registrations nationwide.

Overall, it was found in this review that ∼10% of all the diagnoses had multiple registrations. This estimate is similar to what has been found in a recent German study [14]. Based on this, a correction factor of 0.9 was used to adjust the incidence calculations. Furthermore, in order to estimate the potential instability of the correction factor over time, medical records over a period of five consecutive years were reviewed for the Hanusch Hospital of Vienna (2004–2008) and four consecutive years were reviewed for the University Hospital of Graz (2004–2007), respectively. Over the periods investigated, the percentage of multiple registrations remained fairly stable for both centers (9.67 ± 1.18; mean ± SD).

Statistical analysis

Crude incidence rates (cases per 100,000) of hip fracture were calculated for both sexes in 5-year age intervals based on the number of fractures from AHDR and the annual midyear populations for the years and regions under consideration obtained from Statistics Austria (2009). Age- and sex-standardized incidence rates were estimated using the direct standardization method with the average population over the period used as standard.

In order to analyze the overall change in hip fracture for the period, average annual change, expressed as IRRs, was calculated using regression models based on count data stratified by sex–age class, province, and year. Analyses were conducted for the entire population and stratified by sex and sex–age class, controlling for age and sex as appropriate. In an analyses conducted for the entire population, an interaction between sex and annual change was introduced to control for the difference in annual changes between men and women. Generally, IRRs are calculated using Poisson models. However, one assumption underlying the Poisson model is that the mean and variance are of the same size, and given that the variance was significantly larger than the mean in the present dataset, a negative binomial regression model—a model used to analyze count data which do not assume equal mean and variance—was used.

In order to assess how the change in hip fractures developed over the period, the linear trend in the average annual change was estimated using regression models. The models were based on first differences in count data stratified by sex–age class, state, and year. Analyses were conducted for the entire population and stratified by sex, controlling for first differences in population data, year, and age as appropriate. In this analysis, an accelerating trend would result in a positive year coefficient, a stable trend in a year coefficient equal to zero, and an attenuating trend in a negative year coefficient. Ordinary least square regression models adjusted with robust standard errors were used, given that the data showed signs of heteroscedasticity. Furthermore, in order to assess whether the decline in 2006 observed in the data presented by Mann et al. have continued, a dummy variable for the last 3 years (2006 to 2008) in the dataset at hand was introduced in the negative binomial regression models discussed above.

All statistical tests were two sided with the level of significance set at 5%. Statistical analyses were conducted with STATA 9.0.

Results

Figure 1 shows the number of registered hip fracture discharge diagnoses from 1989 to 2008. Over the period, the number of hip fracture discharges in women and men increased from 7,528 to 11,493 and from 2,044 to 4,122, respectively. When adjusting for multiple registrations for the same fracture, the crude incidence per 100,000 person years in women and men increased from 493 to 642 and from 192 to 280, respectively.
https://static-content.springer.com/image/art%3A10.1007%2Fs00198-010-1271-9/MediaObjects/198_2010_1271_Fig1_HTML.gif
Fig. 1

Number of annual hip fracture discharges in Austrian population aged >50

The annual age-standardized incidence rates per 100,000 person years (1989 to 2008; adjusted for multiple registrations per fracture) are shown in Fig. 2. The age-standardized incidence rates for women and men increased from 515 to 605 and from 193 to 261, respectively. Figure 2 indicates that the age-standardized incidence in females increased until 1999, dipped slightly in 2000, and remained fairly stable from 2001 until 2005 whereafter it decreased. Furthermore, Fig. 2 indicates that the age-standardized incidence rate in men increased at a measured pace until 2006, declined in 2007, and rebounded slightly in 2008. For the population as a whole (also exhibited in Fig. 2), the trend in age-standardized incidence rates for the period largely mirrors the trend in women, reflecting the larger number of women and the higher risk in women. The sex ratio of age-standardized incidence for hip fractures remained relatively stable at c. 2.5–2.7 from 1989 to 1999 whereafter it declined to c. 2.2 in 2008.
https://static-content.springer.com/image/art%3A10.1007%2Fs00198-010-1271-9/MediaObjects/198_2010_1271_Fig2_HTML.gif
Fig. 2

Age-standardized hip fracture incidence rates (per 100,000) for the Austrian population aged ≥50

The overall increase in the age-standardized incidence rates are reflected in the analysis of the IRRs: The multiple negative binomial regression models show a small but statistically significant average annual increase of hip fracture incidence of 1.008 (1.005–1.010, ρ < 0.01) for women and 1.014 (1.011–1.017, ρ < 0.01) for men. However, increase in average annual incidence partially hides the change in trends observed in Fig. 2, and the leveling off is supported by statistical analysis: In the total population, the increase leveled off by 0.16 (ρ < 0.01), driven by an attenuation of 0.29 in women (ρ < 0.01), whereas the leveling off in men was non-significant. Furthermore, the decline in the incidence of hip fracture observed in the last 3 years (2006 to 2008) in Fig. 2 was supported by statistical analysis. In the total population, the IRR for the last 3 years was significantly below the IRR for the first 17 years (0.94, ρ < 0.01), driven by a lower IRR in women (0.91, ρ < 0.01) and to a lesser extent by a lower IRR (not statistically significant) in men (0.96, ρ < 0.15; Table 1).
Table 1

Incidence rates of hip fractures and trends in Austria (1989 to 2008)

 

1989

1990

1991

1992

1993

1994

1995

1996

1997

1998

1999

2000

2001

2002

2003

2004

2005

2006

2007

2008

∆IRR p.a.a

95% CI

Population ≥50 (k)

2,334

2,377

2,419

2,450

2,472

2,494

2,505

2,512

2,536

2,569

2,601

2,631

2,664

2,694

2,719

2,752

2,790

2,836

2,884

2,934

  

Number of hip fractures

9,572

9,901

10,808

11,180

11,486

11,649

12,362

12,829

13,421

13,561

13,877

13,735

14,423

14,663

15,129

14,891

15,697

15,941

15,427

15,615

  

Age-standardized incidencesb

 Men

193

197

222

224

230

223

239

237

247

243

243

246

251

251

262

260

274

280

256

261

1.014

1.011–1.017

 Women

515

522

557

573

581

589

604

630

656

659

667

638

664

660

674

654

667

53

617

605

1.008

1.005–1.01

 Total

376

381

412

422

429

430

446

460

479

479

483

468

485

483

495

483

496

491

461

456

1.010

1.008–1.012

Incidencesb (men)

 Age

  50–54

79

70

67

69

78

72

71

75

67

55

58

78

67

59

57

53

59

59

54

48

0.981

0.975–0.988

  55–59

84

79

95

100

98

85

94

82

86

84

92

89

76

83

77

79

84

91

84

75

0.993

0.987–1

  60–64

108

112

113

123

135

115

133

128

110

122

126

106

113

112

112

97

114

135

129

113

0.999

0.993–1.005

  65–69

135

144

177

169

182

161

187

158

187

153

170

163

145

146

182

187

188

191

184

184

1.010

1.004–1.016

  70–74

240

190

236

260

264

248

292

292

313

306

261

311

264

280

272

306

278

312

243

251

1.007

1–1.013

  75–79

384

411

483

418

418

471

409

439

519

488

525

516

531

514

526

502

526

532

469

508

1.014

1.008–1.019

  80–84

542

698

711

713

784

738

841

906

887

909

862

812

966

1,059

1,066

976

1,120

1,192

950

974

1.028

1.021–1.034

  85–89

1,228

1,140

1,268

1,410

1,289

1,384

1,436

1,411

1,517

1,635

1,542

1,713

1,634

1,499

1,710

1,926

2,150

1,787

1,998

2,162

1.030

1.024–1.037

  90–94

1,319

1,734

2,246

2,174

2,085

2,040

2,568

2,386

2,314

2,445

2,628

2,196

3,554

3,133

3,478

3,349

2,827

3,314

2,923

3,406

1.036

1.028–1.044

  95+

2,365

2,703

2,539

2,260

2,636

3,887

2,413

2,969

2,659

3,459

3,236

3,573

4,751

5,573

5,092

4,641

6,021

4,137

6,461

5,793

1.056

1.042–1.069

Incidencesb (women)

 Age

  50–54

51

45

48

52

50

41

52

56

52

36

47

51

37

39

33

31

47

42

41

35

0.983

0.975–0.992

  55–59

94

75

96

83

89

75

76

74

74

78

83

74

73

81

76

86

76

90

72

83

0.996

0.989–1.003

  60–64

126

142

148

146

150

137

123

140

138

130

130

121

119

120

129

113

119

150

119

113

0.991

0.984–0.997

  65–69

239

238

259

258

277

263

269

276

257

253

279

237

230

187

250

227

231

246

235

185

0.991

0.986–0.997

  70–74

431

424

422

482

445

477

524

549

528

529

537

520

471

464

476

457

464

568

396

408

1.002

0.997–1.007

  75–79

807

817

893

890

945

966

913

948

991

1,067

1,108

1,069

910

939

991

918

1,029

1,048

905

863

1.007

1.002–1.011

  80–84

1,405

1,486

1,598

1,603

1,651

1,662

1,719

1,899

2,069

2,008

2,014

1,812

2,158

2,170

2,057

1,873

1,868

2,011

1,724

1,726

1.016

1.011–1.021

  85–89

2,461

2,490

2,618

2,744

2,726

2,767

3,084

2,963

3,127

3,160

3,125

3,167

3,089

3,106

3,293

3,623

3,715

2,998

3,458

3,365

1.020

1.016–1.025

  90–94

3,239

3,259

3,397

3,786

3,727

4,011

3,719

4,127

4,413

4,586

4,312

4,218

5,352

5,257

5,162

4,967

4,541

4,098

4,295

4,632

1.024

1.018–1.03

  95+

3,500

3,573

3,996

3,302

3,353

3,901

3,985

3,732

3,941

3,729

4,229

4,572

6,957

6,430

6,610

6,174

6,321

4,008

6,779

6,062

1.020

1.008–1.031

aIncidence rate ratio per annum (measures average annual change in the incidence of hip fractures), from negative binomial regression models

bAdjusted for multiple registrations for the same fracture

Discussion

Studies conducted in North America and the Nordic countries from the 1930s to the 1980s consistently reported increasing age-adjusted incidence rates of hip fractures in both men and women [15]. However, in more recent years, the trend has been reversed for both sexes: In Northern USA and Canada, age-adjusted incidence rates have decreased since the 1980s, and data from the Nordic countries also exhibit a reversal of the trend, albeit from a later date [510].

Published studies on the age-adjusted incidence in Middle European countries have generally been conducted on relatively recent data and exhibit divergent trends. In Switzerland, the incidence decreased in women—driven by a decrease in the incidence in institution-dwelling women—but not for men [16, 17]. In Germany, small increases were reported for men and elderly (>74 years of age) women [14].

The present study indicates that while hip fracture incidence in Austria increased over the period 1989–2008, the incidence peaked at the turn of the century and has since remained stable and even decreased in recent years. This trend was driven by a leveling off and subsequent decrease in fracture incidence in women, whereas the incidence in men has continued to rise at a measured pace.

These findings contrast to conclusions drawn by Mann et al. who do not observe a leveling off in the increase in the age-adjusted hip fractures incidence in Austria [12]. The discrepancy partially reflects the difference in time periods under consideration (1989 to 2008 vs 1994 to 2006) but also reflects that Mann et al. only considers change aggregated over the whole period (average annual changes and changes over the whole study period), whereas the present study also analyzes changes in trend.

Several explanations for the reversal of the secular trend in age-adjusted hip fractures incidence have been put forward. Firstly, a cohort effect resulting from changes in early life risk factors such as perinatal nutrition may have affected the fracture risk in the elderly population [18]. Secondly, behavioral changes such as smoking cessation may have decreased the risk of fractures [19, 20]. Thirdly, the increase in average body mass index seen throughout the Western world in recent years may have contributed to a decreased risk [21]. A fourth explanation could be the increasing treatment of osteoporosis with, e.g., calcium and vitamin D supplements as well as bone-specific drugs which result in lower fracture risks. While such interventions are efficacious, only a minority of osteoporotic patients are treated, implying that the reductions are unlikely to be explained by medication use [2224]. The discrepancy in secular trends of hip fracture incidence between men and women observed in Middle Europe suggests that a change in a female specific risk factor may play a role. A secular increase in the average number of reproductive years and exposure to circulating endogenous hormones which have been reported in females would be such a potential contender [25].

While the coming years will show whether the decreasing trend continues or is reversed, the absolute number of hip fractures is likely to continue to increase in the future, reflecting the rapid and continuous aging of the population. Consequently, effective fracture-preventing measures and strategies will continue to play an important role, especially in the light of the large treatment gap observed in osteoporosis.

This study has several strengths: it is based on national population data spanning 20 years and includes a correction factor for repeated hospitalizations for the same fracture. However, it also has limitations. Although administrative data should reliably identify hip fractures, the change in ICD coding in 2000 may have impacted how fractures are reported, but this change should not affect the underlying trend except for the year in which the change occurred. Furthermore, it might be assumed that the AHDR cannot cover all incident hip fractures. However, the Austrian health care system stipulates that any person exhibiting radiological or clinical evidence of hip fracture will immediately be admitted to the nearest hospital. Since it is mandatory for all Austrian hospitals to document all discharge diagnoses, it is very unlikely that any hip fracture might not be covered by the AHDR. Potential instability of the correction factor used in this study could be another limitation. However, it has been shown in a small substudy that the percentage of multiple registrations for the same diagnosis has been fairly stable over time, and finally, another limitation of the prevailing study lies in the fact that the dataset does not contain data on risk factors which could explain the change in fracture incidence trend.

Conclusion

The present study indicates that since 2006, age-standardized incidence of hip fractures has been declining in the Austrian population aged 50 years and above. This reversal in the secular trend has primarily been driven by a decrease in hip fracture incidence in women. A similar trend has been observed in other populations within the recent past, and several factors have been considered as potential explanations. To the extent that further evidence on the reversal of trends in hip fractures in various areas become available, the timing of the break in trend in different regions could be compared to assess whether common causes appear to exist. While such an approach would not determine causality, it could indicate where explanations of the reversal (or non-reversal) of the secular trend seen in this study and in similar studies may be sought.

Acknowledgments

The study was funded by a medical research fund from the Medical University of Graz. We would like to thank Dr. Erika Baldaszti and Mag. Barbara Leitner from the Statistics Austria for providing us with the data from the Austrian Hospital Discharge Register. We are grateful to Dr. Marlene Klingspiegl, from the Steiermärkische Krankenanstalten Ges.m.b.H., and Ingrid Lackner, from the Hanusch Krankenhaus, Vienna, Fourth Medical Department, for extracting the data file from the hospital register.

Conflicts of interest

None.

Copyright information

© International Osteoporosis Foundation and National Osteoporosis Foundation 2010