Health-protective behaviours and risk of fall-related hip fractures: a population-based case–control study

Abstract

Background: fall-related hip fractures are one of the most common causes of disability and mortality in older age. The study aimed to quantify the relationship between lifestyle behaviours and the risk of fall-related hip fracture in community-dwelling older people. The purpose was to contribute evidence for the promotion of healthy ageing as a population-based intervention for falls injury prevention.

Methods: a case–control study was conducted with 387 participants, with a case–control ratio of 1:2. Incident cases of fall-related hip fracture in people aged 65 and over were recruited from six hospital sites in Brisbane, Australia, in 2003–04. Community-based controls, matched by age, sex and postcode, were recruited via electoral roll sampling. A questionnaire designed to assess lifestyle risk factors, identified as determinants of healthy ageing, was administered at face-to-face interviews.

Results: behavioural factors which had a significant independent protective effect on the risk of hip fracture included never smoking [adjusted odds ratio (AOR): 0.33 (0.12–0.88)], moderate alcohol consumption in mid- and older age [AOR: 0.49 (0.25–0.95)], not losing weight between mid- and older age [AOR: 0.36 (0.20–0.65)], playing sport in older age [AOR: 0.49 (0.29–0.83)] and practising a greater number of preventive medical care [AOR: 0.54 (0.32–0.94)] and self-health behaviours [AOR: 0.56 (0.33–0.94)].

Conclusion: with universal exposures, clear associations and modifiable behavioural factors, this study has contributed evidence to reduce the major public health burden of fall-related hip fractures using readily implemented population-based healthy ageing strategies.

Introduction

Hip fractures are one of the most common causes of disability and mortality in older age, and the frequency is increasing by 1–3% per year in most areas of the world because of population ageing [1]. Such injuries are a major public health problem and will present a serious challenge to health-care systems in future years [2].

Over 90% of hip fractures are the result of direct trauma to the hip following a fall [3]. Numerous studies have identified risk factors for falls and fall injuries, often classified according to the contribution of factors both intrinsic and extrinsic to the host [4]. Based on this conceptual model, multi-factorial interventions target high-risk individuals and are aimed at ameliorating both intrinsic and extrinsic risk factors present at the time of the fall [5].

In contrast to the high-risk strategy, population strategies aim to change the distribution of a risk variable in the whole population [6]. However, such approaches to falls injury prevention are rare [7], due, in part, to lack of evidence of population-wide risk factors, which can be addressed by community-based strategies. A comprehensive hierarchical health determinants model allows the examination of more distal factors that influence health outcomes over the life course and contribute to falls risk. Such a model also has the potential to address population-based risk factors and tailor interventions appropriate to life stages using healthy ageing strategies.

Using a health determinants model, the aim of the present study was to examine the relationship between modifiable lifestyle behaviours, identified as determinants of healthy ageing [8], and risk of fall-related hip fracture in community-dwelling older people. The purpose was to contribute to the evidence for the promotion of healthy ageing as a population-based intervention for falls injury prevention.

Methods

Study design and setting

A case–control study was conducted in Brisbane, capital of Queensland on the eastern coast of Australia. The sample size calculation was based on an estimated exposure rate of 60% to at least one of the three lifestyle risk factors (smoking, no exercise and an inadequate diet) for the Australian population aged 75 and over [9]. To detect an odds ratio (OR) of 2.0, with 80% power and an alpha risk of 5%, where the proportion of controls exposed to lifestyle risk factors could be 60%, a sample size of 366 (122 cases and 244 controls) was required.

Study participants

Eligible cases were aged 65 years or over and admitted to one of the six metropolitan hospitals for acute treatment for a fall-related fracture of the proximal femur during the period May 2003–October 2004. Ineligible cases were those in receipt of high care in residential institutions at the time of the fall. Controls were selected from the electoral roll, the most up-to-date and comprehensive database of Australian citizens [10]. For each case, at least two controls were randomly selected, after matching on sex, age (within a five-year age range) and postcode of residence. As for cases, controls were ineligible if they were in residential care.

Ethical approval

Ethical approval for the study was obtained from the University of Queensland and Queensland Health Ethics Committees. After receiving information about the project and an invitation to participate, all subjects (or their proxy health carers) gave written consent to participate.

Data collection

A questionnaire designed to assess lifestyle behaviours, identified from literature review as determinants of healthy ageing [8], was administered at face-to-face interviews. A proxy was sought when the score on a comprehension sub-test administered at interview was below the cut-off of 7 out of 10 correct answers to questions assessing orientation to person, place and time.

Study factors

Smoking status

Smoking history included whether ever or never smoked as well as age started and ceased, if applicable. The variable was categorised as current, past or never smoker and, based on previous studies [11], past smokers were subdivided by the number of years since cessation.

Alcohol consumption

Alcohol consumption was recorded over three life stages: older age (65 and over), middle age (40–64) and young adulthood (15–39). Consumption was categorised as none (non-drinker), low (drink on special occasions only), moderate (drinking no more than two standard drinks per day in older age and no more than four drinks per day in middle and younger ages for males and half these measures for females) and high (more than moderate level). The definition of moderate consumption was based on recommended levels to achieve a health benefit [12].

Body weight

Self-reported body weight and height were converted to body mass index (BMI) and classified as underweight (<18.5), normal (18.5–24.9), overweight (25.0–29.9) and obese (≥30.0) [13]. Information also included perceptions of weight as underweight, normal or overweight over the life stages.

Diet

Intake of fruit, vegetables and fluid as well as breakfasting daily in older age, middle age and as a young adult was ascertained. Classification of adequate intake was based on recommended guidelines [13].

Physical activity

Calculations of sufficient or insufficient physical activity were based on minutes per week spent walking, as well as doing moderate and/or vigorous activity in an average week in the last 6 months [14]. Sport involvement over the life stages was also recorded.

Other health-protective behaviours

Other health-protective behaviours associated with healthy ageing [15, 16] were grouped as preventive medical care, self-health, home safety and reproductive health practices (for women only). Preventive medical care practices included having regular medical, dental, eyesight and skin examinations, as well as an annual influenza vaccination. Self-health practices included having an enduring power of attorney, having 7–8 h of sleep at night, not being idle during the day, spending free time outdoors, keeping a pet and taking sunscreen precautions when outdoors. Home safety practices included keeping emergency telephone numbers handy, having smoke detectors installed and grab rails in the bathroom/toilet. Reproductive health variables included taking hormone replacement therapy, breastfeeding children and having regular Pap smears and mammograms. Summary measures for each of these health behaviours (preventive medical care, self-health, home safety and reproductive health practices) were calculated by addition of positive responses to the questions within the categories.

Demographics

Demographic information included age, sex, education, financial status and country of birth.

Potential confounders

Health-status measures identified as risk factors for falls [17] included self-report of medical conditions, sensorimotor impairments, number of prescribed medications, limitations in activities of daily living and falls history.

Analysis

Stata 9.0 (StataCorp College Station, TX, 2005) was used in conditional logistic regression analysis to produce ORs with 95% confidence intervals (CIs) for the effect of exposure measures on the outcome of hip fracture [18]. Multivariate models were developed for behavioural factors by adding, in step-wise progression, factors significant in univariate analysis. At each stage in the modelling process, alternative models were compared using, as appropriate, the likelihood ratio chi-square statistic, pseudo R-square, as well as Akaike’s and Bayesian information criteria. A probability (P) value of <0.05 was the criterion of significance.

Two-way interactions between the behavioural factors and with age and sex were tested by adding to the model one at a time. Potential confounders, significant in univariate analysis, were also added to the model. The model was adjusted if the ORs of behavioural factors were altered by >10%.

Results

In total, 126 cases and 261 controls completed an interviewer-administered questionnaire. The overall participation rate of eligible subjects was 85%. Chi-square tests showed there were no significant differences between participants and non-participants on sex or age-group distribution. Proxies assisted with 25 (6.5%) of the interviews, with no significant differences between cases and controls on proxy use, nor on demographic variables. The distribution of demographic characteristics in the study population is shown in Table 1.

The behavioural factors significant in univariate analysis are summarised in Table 2.

The final behavioural model, adjusted for confounding effects of health-status measures, is shown in Table 3. The pseudo R-squared value of 23% indicated the percentage predictability of the model.

In this model, behavioural factors having a significant independent protective effect on hip fracture risk included never smoking [adjusted odds ratio (AOR): 0.33 (0.12–0.88)], moderate alcohol consumption in mid- and older age [AOR: 0.49 (0.25–0.95)], not losing weight between mid- and older age [AOR: 0.36 (0.20–0.65)], playing sport in older age [AOR: 0.49 (0.29–0.83)] and practising a greater number of preventive medical care [AOR: 0.54 (0.32–0.94)] and self-health behaviours [AOR: 0.56 (0.33–0.94)]. Dietary variables were not independent predictors. Reproductive health behaviour, which was protective of hip fractures in women, was not retained in the final model in favour of a model inclusive of both females and males and because of co-linearity between this factor and preventive medical care practices. Other significant associations between behavioural factors in the model were playing sport in older age with both moderate alcohol consumption and a greater number of self-health behaviour practises. Despite these associations, the factors remained independent predictors in the multivariate model.

When added to the behavioural factors model, the health-status measures, which remained significant independent predictors of hip fracture, included having a history of respiratory disease, hearing problems, limitations in activities of daily living and previous hip fracture. Inclusion of these factors improved the predictive capability of the model (pseudo R-square of 31%). However, only hearing problems had a confounding relationship with behavioural factors, lowering the ORs for past and never smokers and resulting in the adjustment of the behavioural factors model. While interactions produced some effect in the modification for different levels of the behavioural factors, their inclusion did not markedly improve the model. Furthermore, they did not change the importance of the protective effects of the behavioural factors established in the best-fit model.

In the final behavioural model, two cases (1.6% of cases) and three controls (1.1% of controls) had missing data on at least one of the six factors. An examination of the records showed that the missing data appeared to be random and, had the data been available, unlikely to have changed the results observed.

Discussion

There have been few studies which examined as an a priori aim the association between lifestyle behavioural factors and fall-related hip fracture, particularly of protective factors over the life course. The major findings of this study are that the behavioural factors (never smoking, moderate alcohol consumption, not losing weight between mid- and older age, playing sport and practising a greater number of preventive medical care and self-health behaviours) independently confer a protective effect on the risk of fall-related hip fracture in older age.

Consistent with the results of this research, studies of tobacco use and risk of fracture [19, 20] have reported that hip fracture is a major adverse effect of smoking, because of substantial cumulative excess bone loss. The risk decreased for former smokers, depending on years after cessation. Excessive alcohol intake is also a known cause of osteoporosis and increased risk of hip fracture [21]. High intakes in the present study were not significantly associated with increased risk, but this result may have been because of low numbers in the study population classified in this category. Evidence for the effects of moderate consumption on hip fracture risk is less well documented [21], although, as in the present study, a weak inverse association of moderate alcohol intake with hip fracture risk has been reported [22].

The relationship between low BMI and hip fracture risk in this study is consistent with other results [23], as is the finding that weight loss between mid- and older age is also an important risk indicator [24]. Weight loss is a marker of frailty that may increase the risk of hip fracture through association with several indicators of poor health, including physical disability and low level of physical activity [25]. Consistent with reviews of the studies on physical activity [26, 27], this study showed a reduced risk of hip fracture for those engaging in physical activity in leisure in older age.

There has been little research to date of an association between preventive medical care and other self-health practices on the risk of fall-related hip fracture, despite the fact that comprehensive multidisciplinary health/environmental assessments have been advocated for falls prevention [5]. Such health-protective behaviours may have no direct biological explanation for an association with hip fractures but may act indirectly through an association with better health outcomes in older age.

While the effect of behavioural factors on hip fracture risk is thought to be mediated mainly through lowered bone mineral density (BMD) [19, 21, 23, 27], a diagnosis of osteoporosis (a proxy measure of lowered BMD) was not found to be an important intermediary or confounder in this study. The health-promoting behaviours (never smoking, moderate alcohol consumption, maintaining healthy weight, playing sport and undertaking preventive medical and self-health care) confer a protective effect on the risk of hip fracture, independent of current medical conditions and other health-status measures.

The strengths of the study are the rigours of the methodology; in particular, the population-based case and control groups with a high response rate, low level of missing data and with adequate control for confounding. The evidence from this research is consistent with the findings from cohort and case–control studies reported in the reviews of behavioural factors and fracture risk [19–21, 23, 26, 27], and the magnitude of effects observed are large. The temporal relationship whereby exposures preceded outcome and a dose–response gradient for different levels of a variable (e.g. smoking, alcohol consumption and BMI) were also evident in the results obtained. Limitations of the study are those associated with case–control study designs including recall bias and reliance on self-reported retrospective data. While it has been shown that older people are able to give reproducible answers to questions about a wide range of exposures for many decades in the past [28], a limitation of cross-sectional data still remains that perceptions of events may well be influenced by outcome status.

Conclusion

The health determinants model of healthy ageing has enabled the identification of health-protective behaviours associated with fall-related hip fracture. The major challenge is to discover the best ways to apply this research about behaviour change to encourage populations to adopt healthy actions. Evidence is beginning to emerge from trials of community-based programmes [29, 30] that lifestyle interventions for falls and osteoporosis prevention are reducing fracture incidence. The results of this study contribute to the evidence that promoting healthy ageing has the potential to reduce the major public health burden of fall-related hip fractures.

Key points

• There have been few previous studies examining the association between lifestyle behavioural factors and fall-related hip fracture, particularly with respect to the protective factors over the life course.

• Behavioural factors of not smoking, moderate alcohol consumption, not losing weight between mid- and older age, playing sport in older age and practising preventive medical care and self-health behaviours had a significant independent protective effect on the risk of hip fracture.

• The study provides evidence for addressing falls injury prevention among community-dwelling older people using readily implemented population-based healthy ageing strategies.

Author contributions

Dr Nancye Peel was responsible for the study concept and design, recruitment of subjects, data collection, analysis and interpretation of the data and preparation of this article. Professor Roderick McClure was responsible for the study concept and design, interpretation of the data and preparation of this article. Joan Hendrikz was responsible for statistical advice and analysis and preparation of this article.

Other contributions

Professor Helen Bartlett (Australasian Centre on Ageing, University of Queensland), as one of the principal investigators, was responsible for conceptual design and study overview. Professor Gail Williams (School of Population Health, University of Queensland) provided a statistical overview of the research. Cecilia Wilson (Centre of National Research on Disability and Rehabilitation Medicine, University of Queensland) provided research assistance with data collection. Dr Janene Suttie (Australasian Centre on Ageing, University of Queensland) provided research assistance with data collection. Hospital staff at the participating hospitals (geriatricians and orthopaedic physiotherapists) identified cases for the research project.

Declaration of sources of funding

The research was supported by an RM Gibson Scientific Research Fund Grant from the Australian Association of Gerontology. The funders played no role in the design, conduct, analysis or interpretation of the data, nor in the writing of this article.

Conflicts of interest statement

There are no conflicts of interest to declare.

References