Prospective observational study of physical functioning, physical activity, and time outdoors and the risk of hip fracture: A population-based cohort study of 158,057 older adults in the 45 and up study

Authors

  • Jeffrey K C Lai,

    Corresponding author
    1. National Centre for Epidemiology and Population Health, The Australian National University, Canberra, Australia
    • Address correspondence to: Jeffrey K C Lai, PhB (Hons), MBBS (Hons), National Centre for Epidemiology and Population Health, The Australian National University, Canberra, ACT, 0200, Australia. E-mail: jeffrey.lai@anu.edu.au

    Search for more papers by this author
  • Robyn M Lucas,

    1. National Centre for Epidemiology and Population Health, The Australian National University, Canberra, Australia
    Search for more papers by this author
  • Miranda Armstrong,

    1. Cancer Epidemiology Unit, University of Oxford, Oxford, United Kingdom
    Search for more papers by this author
  • Emily Banks

    1. National Centre for Epidemiology and Population Health, The Australian National University, Canberra, Australia
    Search for more papers by this author

ABSTRACT

Low levels of physical activity or sun exposure and limitations to physical functioning (or disability) have been identified as possible risk factors for hip fracture. However, these factors are closely related, and data on their independent and joint association with risk of hip fracture are limited. A total of 158,057 individuals aged ≥45 years sampled from the general population of New South Wales, Australia, from the prospective 45 and Up Study completed a baseline postal questionnaire in 2006 to 2009 including data on physical activity (Active Australia questionnaire); sun exposure (usual time outdoors); and physical functioning (Medical Outcomes Score-Physical Functioning; scored 0 to 100). Incident first hip fractures were ascertained by linkage to administrative hospital data (n = 293; average follow-up 2.3 years). The relative risk (RR) of hip fracture was estimated using Cox proportional hazards. Poorer physical functioning, lower physical activity, and less time outdoors were positively related to each other at baseline and individually associated with significantly increased hip fracture risk. However, physical activity and time outdoors were not significantly related to hip fracture risk after adjustment for baseline physical functioning or when analysis was restricted to those with no or mild baseline physical limitation. In contrast, physical functioning remained strongly related to hip fracture risk after adjustment for the other two factors; compared with the group without limitation (100), the RR of hip fracture among those with mild (75–95), moderate (50–70), severe (25–45), and greatest (0–20) level of physical limitation was 1.38 (95% confidence interval [CI] 0.88–2.14), 2.14 (1.29–3.53), 3.87 (2.31–6.44), and 5.61 (3.33–9.42), respectively. The findings suggest that limitation in physical functioning, but not physical activity or time outdoors, is strongly related to hip fracture risk. The apparent increased risk of hip fracture previously described for low physical activity or sun exposure may be, at least in part due to uncontrolled confounding. © 2013 American Society for Bone and Mineral Research.

Introduction

Hip fracture is a serious and debilitating injury associated with significant morbidity and mortality.[1] The incidence of hip fracture has been well studied and is predicted to rise in line with population aging[2] with an increasing burden to be borne by developing countries experiencing demographic shifts.[3] The identification of risk factors that may lead to interventions to reduce hip fracture incidence remains a research priority.

Higher levels of physical activity,[4, 5] physical functioning,[6, 7] and sun exposure[8] have been separately shown to be associated with decreasing risk of hip fracture. Yet these three factors are related to each other and to vitamin D status,[9] which is known to be important to bone health. Each of these factors—physical activity, sun exposure, and physical functioning—are likely to reflect an individual's overall health status and frailty, which, in turn, are strongly related to hip fracture risk.[5] To our knowledge, no previous study has investigated all three of these factors together in relation to hip fracture risk. Yet, without clearly delineating the overlapping contributions of these factors, it is difficult to determine their individual contributions and thus to target appropriate interventions.

The present study uses data from the prospective cohort 45 and Up Study. Previous studies have reported results on physical activity[8, 9] and physical limitation[8] from this large cohort. However, this is the first study to link data on such factors to hospital admissions for hip fractures. The aim of this article is to describe the separate and combined relationships of physical activity, sun exposure, and physical functioning to the risk of hip fracture in older adults.

Materials and Methods

Participants and data

The 45 and Up Study is a prospective large-scale Australian cohort study of individuals 45 years and older, described in detail elsewhere.[10] Briefly, from January 2006 to April 2009, more than 250,000 participants were randomly sampled from the general population in New South Wales (NSW), Australia, through the Medicare Australia enrollment database, which provides virtually complete coverage of the general population. Individuals were mailed study enrollment materials and the study questionnaire, which contained questions on a range of exposures and outcomes of public health importance for an aging population. Participants joined the study by returning the self-administered questionnaire along with written consent for follow-up and linkage of data to population health databases. Overall, the 45 and Up Study covers around one-tenth of the general population in NSW in the target age range.

Personal information including full name, date of birth, sex, and address was used as part of the Master Linkage Key at the NSW Centre for Health Record Linkage (CHeReL) to link baseline 45 and Up Study data to the NSW Admitted Patient Data Collection and deaths registrations. Linkages were made using a probabilistic algorithm with manual clerical review of indeterminate matches. Once linked, personal identifiers were removed.

The Admitted Patient Data Collection is a census of all admitted patient hospital services provided in NSW. Details of hospital admissions were available for participants up until 30 June 2010. All diagnoses are coded according to the International Statistical Classification of Diseases and Related Health Problems, Tenth Revision, Australian Modification (ICD-10-AM). For this analysis, incident hip fractures were defined as the first hospital admission after recruitment into the 45 and Up Study with a primary diagnosis of fracture of the neck of femur (ICD-10 AM codes S72.0–S72.2).

Physical functioning, physical activity, time outdoors, and other factors

Physical functioning was measured using the Medical Outcomes Score-Physical Functioning (MOS-PF),[11] which is equivalent to items from the physical functioning scale (PF-10) of the SF-36 health survey.[12] The PF-10 has been well validated as a measure of physical functioning in terms of hierarchical structure, unidimensionality, and reproducibility across a wide range of patient groups varying by age, sex, and comorbidities.[13] It consists of 10 questionnaire items targeted at varying levels of physical functioning, in response to the question: “Does your health now limit you in any of the following activities?” The items are “vigorous activities,” “moderate activities,” “lifting or carrying shopping,” “climbing several flights of stairs,” “climbing one flight of stairs,” “walking one kilometer,” “walking half a kilometer,” “walking 100 meters,” “bending, kneeling, or stooping,” and “bathing or dressing yourself.” For each item, participants answer “yes, limited a lot,” “yes, limited a little,” or “no, not limited at all,” with scoring of 0 (yes, limited a lot), 50 (yes, limited a little), or 100 (no, not limited at all). An overall physical functioning score was calculated from the average of scores from all 10 items, and participants were classified into the following categories: 0–20, 25–45, 50–70, 75–95, and 100, considered as very severe impairment, severe impairment, moderate impairment, mild impairment, and no impairment, respectively. In certain cases where the questionnaire data were incomplete, some values were imputed. For example, if there was no limitation in walking one kilometer, then it was assumed there was no limitation in walking half a kilometer, or 100 meters. Similarly, participants who were limited a lot in bathing or dressing, the least strenuous physical activity item, were placed into the lowest category of 0–20 for overall physical functioning score if the data from all other items were not otherwise complete.

Questionnaire items on physical activity were derived from the Active Australia Survey.[14] Study participants were asked about the frequency of physical activity at recruitment by the question: “How many times did you do each of these activities last week?” with the following three items: “walking continuously, for at least 10 minutes (for recreation or exercise or to get to or from places),” “vigorous physical activity (that made you breathe harder or puff and pant, like jogging, cycling, aerobics, competitive tennis, but not household chores or gardening),” and “moderate physical activity (like gentle swimming, social tennis, vigorous gardening, or work around the house)” Participants were required to respond to each item with a numerical answer.

From the questionnaire data, the metabolic equivalent intensity level (MET-adjusted) number of sessions of physical activity over the last week was calculated as the sum of the total number of times walking (W), doing moderate activity (M), and two times the number of times doing vigorous activities (V), ie, W + M + 2V.[9] Participants were classified into the following categories: ≤5, >5–9, >9–14, >14 MET-adjusted sessions per week.

Sun exposure was determined by the question: “About how many hours a day would you usually spend outdoors on a weekday and on a weekend?” Participants were asked to provide a numerical answer for both weekdays and weekends. From these values, the total number of hours spent outdoors over a week was obtained from the weighted sum of weekdays and weekends. Participants were classified into the following categories: ≤12, >12–20, >20–30, and >30 hours per week.

Height and weight were self-reported in the questionnaire and used to calculate body mass index (BMI). Women were classified as postmenopausal if they self-reported having been through menopause, were over 55 years old, or had had a bilateral oophorectomy. Women were classified as premenopausal if they self-reported not having been through menopause, were 55 years or younger, and were not on menopausal hormone therapy (MHT). All other women were classified as having unknown menopausal status. This group included women aged 55 or younger reporting hysterectomy without bilateral oophorectomy, and current MHT users who were self-reported as premenopausal, as MHT can mask menopausal status. Past users of MHT were grouped together with non-current users of MHT because of the transience of the effect of MHT on bone health.[15] From this information, menopause and MHT status were categorized as: premenopausal and not currently using MHT; postmenopausal and not currently using MHT; postmenopausal and currently using MHT; unknown menopausal status and currently using MHT; and unknown menopausal status and not currently using MHT.

Exclusion criteria

Participants were excluded from the analysis if they had reported a previous hip fracture, had a hip fracture recorded in the Admitted Patients Data Collection before recruitment into the study, or had been diagnosed with cancer before recruitment (except non-melanoma skin cancer) according to the questions: “Have you had a broken/fractured bone in the past five years?” and “If yes, which bones were broken?” with response “yes” and “hip,” respectively, and “Has a doctor ever told you that you have: melanoma, prostate cancer, breast cancer, or other cancer” with response of “yes.” Participants with missing data on physical activity, physical functioning, or time outdoors were also excluded from the analysis.

Statistical analysis

Baseline characteristics of participants in relation to categories of physical activity, time spent outdoors, and physical functioning were summarized with mean and standard deviations or number and proportions.

Cox proportional hazards models were used to estimate the relative risk and associated 95% confidence intervals (CI) for incident hip fractures according to physical functioning, physical activity, time outdoors, and other variables. Age was used as the underlying time variable. Person-years at risk were calculated from enrollment into the 45 and Up Study through to time of admission to hospital for first hip fracture, death, or 30 June 2010, whichever occurred earlier. All analyses were done in Stata version 9.0.[16]

The proportional hazards assumption was examined using Schoenfeld residuals with global chi-square tests for each model. For individual predictor variables, non-zero slopes of residuals were investigated using plots of Schoenfeld residuals against the logarithm of time with lowess curves.

In the minimally adjusted models, results were adjusted for age and sex only. Fully adjusted models were additionally adjusted for alcohol consumption (none, 1–14 standard drinks per week, ≥15 standard drinks per week), smoking (never, past, current), education level (no school certificate, school certificate, higher school certificate, trade or apprenticeship, certificate or diploma, university degree or higher), height (in cm), BMI (in kg/m2 where defined) and, in women, menopause/MHT status (in categories described previously). A category for missing values was used where necessary.

Additional adjustment was performed only for those confounders that altered the relative risk of physical activity, time outdoors, or physical functioning score by at least 5%. The only significant additional variable was previous (non-hip) fracture in the past 5 years (yes, no). Variables found to contribute negligibly to the model and not included were: socioeconomic status (by Socio-Economic Indexes for Areas [SEIFA] quintiles), comorbidities (diabetes, heart disease, blood clots, or stroke), skin color (fair, light olive, dark olive, brown, black), skin tannability (very tanned, moderately tanned, mildly tanned, never tan), and use of bisphosphonates and multivitamins.

Fully Adjusted Model A examined the association of hip fracture to one main exposure, ie, it modeled the relationship of physical activity, time outdoors, or physical functioning score to hip fracture, individually. Fully Adjusted Model B accounted additionally for all main exposures and hence included all three of these variables.

Several additional sensitivity analyses were done. These included restricting to participants who self-reported: i) never previously being diagnosed by a doctor as having heart disease, diabetes, blood clots, or stroke; ii) no prior fracture in the past 5 years, treatment for arthritis, or bisphosphonate use; and, iii) mild or no physical impairment (physical functioning score ≥75). In addition, adjustment for physical functioning was investigated by using a stratified Cox model according to physical functioning scores (0–20, 25–45, 50–70, 75–95, 100).

To examine whether physical activity altered the risk of hip fracture according to the baseline level of physical functioning, data were stratified according to the following cutoffs: physical functioning score <75 versus ≥75, and categories of MET-adjusted physical activity as previously defined. Relative risks were obtained with covariates used in the fully adjusted Cox model, including time outdoors. Interaction between physical activity and physical functioning score was tested using the likelihood ratio test comparing the models with and without the interaction terms.

Results were considered statistically significant if the two-sided p value was <0.05. Graphs of relative risks were produced with the R computing package (version 2.5.1; 2007, available at http://www.cran.r-project.org).[17]

Results

A total of 158,057 participants including 84,532 women and 73,525 men were eligible for inclusion in the analysis. Total follow-up time was 362,368 person-years, an average of 837 days or 2.3 years per participant. During the follow-up period, there were 293 first admissions to hospital with a primary diagnosis of hip fracture. This equated to an overall incidence rate of hip fracture of 80 per 100,000 person-years.

Baseline characteristics of participants in relation to physical activity, time spent outdoors, and physical functioning are reported in Table 1. There were strong positive associations between baseline number of physical activity sessions, time spent outdoors, and physical functioning score for the study population (Fig. 1). Physical activity, time outdoors, and physical functioning were all inversely associated with falls over the past 12 months. Associations with other variables were less consistent. Physical activity was positively associated with alcohol consumption, multivitamin use, and fewer educational qualifications, and inversely associated with BMI and proportion of subjects with comorbidities. Time outdoors was positively associated with height, alcohol consumption, proportion of current smokers, being classified within the lowest socioeconomic status (SES) quintile, fewer educational qualifications, and having at least one comorbidity, and inversely associated with multivitamin and bisphosphonate use and proportion of females. Physical functioning was higher with greater height, and lower with increasing age, increasing bisphosphonate use, previous non-hip fracture, current smoking, lower SES, fewer educational qualifications, and having at least one comorbidity.

Table 1. Baseline Characteristics of Participants in Relation to Physical Activity, Time Spent Outdoors, and Physical Functioning
Characteristics at recruitmentPhysical activityTime outdoorsPhysical functioningTotal
MET-adjusted sessions per weekHours per weekPhysical functioning score
 ≤5>5–9>9–14>14≤12>12–20>20–30>300–2025–4550–7075–95100 
Age (years), mean (SD)60.4 (10.9)60.7 (10.3)61.1 (10.2)59.7 (9.5)59.5 (10.8)59.7 (10.0)61.5 (10.2)61.1 (9.8)69.4 (13.1)66.6 (12.0)65.2 (11.2)61.7 (9.9)56.6 (8.0)60.4 (10.2)
Female, % (n)52.2 (18,192)55.8 (21,149)56.7 (22,338)49.7 (22,338)69.3 (25,703)60.8 (26,143)50.9 (18,526)34.1 (14,160)58.0 (2992)60.1 (4124)59.8 (8507)50.8 (34,959)53.9 (33,950)53.5 (84,532)
BMI (kg/m2), mean (SD)28.1 (5.5)27.0 (4.8)26.5 (4.6)26.2 (4.4)27.0 (5.3)26.8 (4.8)26.8 (4.6)27.1 (4.5)29.1 (6.8)29.2 (6.2)28.6 (5.6)27.2 (4.7)25.8 (4.1)26.9 (4.8)
Height (cm), mean (SD)168.4 (11.9)168.1 (11.8)168.0 (12.0)169.4 (12.0)166.3 (11.5)167.8 (11.8)168.8 (11.9)171.0 (12.1)165.8 (12.8)166.3 (12.2)166.8 (12.0)168.9 (12.1)169.0 (11.6)168.5 (12.0)
Alcohol standard drinks/week, mean (SD)6.2 (9.8)6.8 (9.2)7.2 (9.1)8.4 (10.5)5.4 (8.0)6.7 (8.7)7.6 (9.7)9.1 (11.5)4.7 (10.1)5.6 (10.3)6.2 (10.3)7.7 (9.9)7.3 (9.1)7.2 (9.7)
Current smokers, % (n)8.9 (3102)6.6 (2488)6.3 (2538)7.4 (3318)5.5 (2024)6.4 (2759)7.2 (2605)9.8 (4058)11.5 (594)10.7 (733)9.0 (1286)7.0 (4821)6.4 (4012)7.2 (11,446)
Socioeconomic status: lowest fifth, % (n)19.0 (6625)17.3 (6541)17.6 (7079)17.7 (7962)15.0 (5576)16.1 (6912)18.1 (6595)22.0 (9124)25.6 (1319)23.8 (1635)22.0 (3120)17.3 (11,909)16.3 (10,224)17.9 (28,207)
Education: no higher school certificate, % (n)34.0 (11,710)30.0 (11,265) 29.5 (11,775)25.4 (11,335)26.5 (9748)27.2 (11,592)29.8 (10,729)34.2 (14,016)54.3 (2731)47.3 (3195)40.6 (5696)27.9 (19,038)24.7 (15,425)29.5 (46,085)
Current users of menopausal hormone therapy in females, % (n)10.3 (1848)10.6 (2222)10.6 (2386)11.5 (2535)10.8 (2749)10.9 (2815)10.7 (1946)10.6 (1481)9.0 (262)12.3 (497)12.0 (1002)11.1 (3839)10.1 (3391)10.8 (8991)
Previous non-hip fracture in past 5 years, % (n)10.6 (3633)9.9 (3662)10.3 (4087)11.0 (4870)11.0 (4017)10.5 (4431)9.9 (3543)10.5 (4261)21.7 (1089)16.3 (1092)14.7 (2046)9.8 (6601)8.7 (5424)10.5 (16,252)
Bisphosphonate use, % (n)2.5 (880)2.4 (918)2.6 (1034)1.9 (839)3.0 (1101)2.4 (1042)2.3 (835)1.7 (693)6.9 (356)5.4 (372)4.8 (681)2.3 (1566)1.1 (696)2.3 (3671)
Multivitamin use, % (n)23.2 (8075)24.9 (9426)26.4 (10,643)27.7 (12,441)28.9 (10,713)27.0 (11,615)24.9 (9075)22.1 (9182)20.9 (1077)23.2 (1594)25.8 (3678)25.4 (17,508)26.6 (16,728)25.7 (40,585)
Fall over past year, % (n)20.7 (7224)18.1 (6849)17.8 (7157)17.1 (7698)20.2 (7505)18.3 (7863)17.7 (6452)17.1 (7104)52.7 (2720)39.9 (2734)29.6 (4207)16.8 (11544)12.3 (7723)18.3 (28928)
At least one comorbidity, % (n)35.5 (12,386)31.3 (11,847)30.3 (12,212)29.1 (13,082)27.0 (10,011)28.0 (12,055)32.9 (11,971)37.3 (15,490)60.1 (3099)53.5 (3667)45.1 (6419)33.8 (23317)20.7 (13025)31.3 (49,527)
Physical activity, mean MET-adjusted sessions/week (SD)    9.8 (10.4)12.0 (14.2)13.6 (17.5)16.7 (20.9)6.7 (12.5)9.8 (15.5)11.1 (14.2)12.8 (15.0)14.8 (18.6)13.1 (16.5)
Time outdoors, mean hours/week (SD)21.0 (18.0)23.2 (17.8)25.3 (18.4)29.3 (20.3)    18.9 (19.7)22.1 (19.0)23.6 (18.6)25.2 (18.3)25.9 (19.6)25.0 (19.0)
Physical limitation score, mean (SD)77.2 (29.0)85.9 (20.3)88.1 (18.0)90.7 (16.5)82.0 (26.0)86.6 (20.6)86.9 (19.7)87.8 (19.4)     85.9 (21.6)
Figure 1.

Relationship between physical activity, time spent outdoors, and physical functioning. Means of each variable with 95% CI are plotted against each category of (A) physical activity, (B) time spent outdoors, and (C) physical limitation score.

The proportion of missing responses for physical activity, time spent outdoors, and physical functioning score varied. Within participants that would otherwise have been eligible for inclusion in the analysis except for missing or incomplete baseline responses (n = 222,879), the proportions of missing data were: physical activity (30,232; 13.6%); time spent outdoors (16,865; 7.6%); and physical functioning score (34,516, 15.5%).

Results from the Cox proportional hazards model are reported in Fig. 2. The proportional hazards assumption was satisfied in all main models.

Figure 2.

Relative risks (RR) of hip fracture in relation to physical activity, time spent outdoors, and physical functioning. Minimally adjusted model is adjusted for age and sex only. Fully adjusted models are additionally adjusted for menopause/menopausal hormone therapy status, smoking status, alcohol consumption, education level, BMI, height, and previous non-hip fracture. Fully Adjusted Model A shows associations for one main exposure, ie, one of physical activity, time spent outdoors, or physical functioning score. Fully Adjusted Model B accounts additionally for all main exposures, ie, includes physical activity, time spent outdoors, and physical functioning score.

Using minimally adjusted models, lower physical activity, time spent outdoors, and physical functioning score were all significantly individually associated with increased hip fracture risk. In the case of physical functioning, risk increased gradually with increasing levels of physical limitation (p for trend < 0.01), with the highest risk in those with the most severe limitation (physical functioning score of 0–20; RR = 5.42, 95% CI 3.37–8.74) compared with no physical limitation. Similarly, there was a dose-response relationship seen in time spent outdoors with those in the highest group (>30 hours per week) having the lowest risk of hip fracture (RR = 0.54, 95% CI 0.37–0.78) compared with those with ≤12 hours outdoors per week (p for trend < 0.01). Hip fracture risk was also lower in those with higher levels of physical activity compared with those with lower levels (p for trend < 0.01), although all groups completing >5 sessions per week had a similar reduction in hip fracture risk (p = 0.59, X22 = 1.07).

In the fully adjusted model examining each main exposure in relation to hip fracture (Fully Adjusted Model A), results were similar, with significant trend of increasing hip fracture risk with increasing physical limitation (p for trend < 0.01), reduced physical activity (p for trend < 0.01), and reduced time outdoors (p for trend < 0.01). However, within the fully adjusted model additionally accounting for all main exposures (Fully Adjusted Model B), only physical limitation was significantly related to an increase in hip fracture risk (Fig. 2). Physical functioning scores of 50–70 (RR = 2.14, 95% CI 1.29–3.53), 25–45 (RR = 3.87, 95% CI 2.31–6.44), and 0–20 (RR = 5.61, 95% CI 3.33–9.42) were associated with an increased risk of hip fracture compared with the no physical limitation group. In all models, the risk of hip fracture within the physical functioning score 75–95 group was not statistically different from the risk of hip fracture within the reference group (physical functioning = 100).

Notably, relative risks were attenuated substantially and physical activity and time outdoors were not significantly associated with hip fracture risk if physical functioning was included in the model. This highlights the lack of any individual effect of these factors on hip fracture risk in this data set, although power constraints should be borne in mind. In contrast, the association between physical functioning and hip fracture risk persisted in the fully adjusted model accounting for all exposures, with estimates similar compared with the minimally adjusted, or adjusted for one main exposure, models.

Results of sensitivity analyses are shown in Table 2. Similar results on the interplay between physical functioning, physical activity, and time spent outdoors were observed when the analysis was restricted to those participants with no medical comorbidities or no previous fracture, treatment for arthritis, or use of bisphosphonates. In all cases, increasing physical limitation was associated with a significant increase in the risk of hip fracture that persisted after adjustment for physical activity and time spent outdoors.

Table 2. Sensitivity Analysis
 Participants with no comorbiditiesParticipants with no previous fractures, treatment for arthritis, or use of bisphosphonatesParticipants with mild and no physical impairmentModel fitted with stratification by physical functioning levels
Events/NRR95% CIEvents/NRR95% CIEvents/NRR95% CIEvents/NRR95% CI
  1. Fully adjusted models account for physical activity, time outdoors, and physical functioning fitted after restricting to participants with i) no comorbidities (heart disease, diabetes, blood clots, or stroke); ii) no previous fractures, treatment for arthritis, or use of bisphosphonates; and iii) mild and no physical impairment. Additionally, physical functioning was investigated by using stratification according to physical functioning scores (0–20, 25–45, 50–70, 75–95, 100).
MET-adjusted physical activity sessions
≤5/week55/26,4131.00 55/28,4441.00 17/24,9001.00 113/34,8871.00 
>5–9/week40/30,0250.89(0.57, 1.38)42/31,4621.04(0.68, 1.61)31/31,5381.01(0.55, 1.84)63/37,8710.80(0.57, 1.10)
>9–14/week43/32,2860.97(0.61, 1.51)46/33,4351.16(0.75, 1.79)36/34,8590.96(0.53, 1.74)70/40,3260.92(0.65, 1.28)
>14/week34/37,1280.91(0.56, 1.49)25/37,6000.74(0.43, 1.24)31/40,5100.88(0.48, 1.62)47/44,9730.77(0.53, 1.13)
   p-trend = 0.813  p-trend = 0.431  p-trend = 0.607  p-trend = 0.282
Time outdoors
≤12/week59/29,6051.00 55/30,0471.00 22/28,8511.00 106/37,1151.00 
>12–20/week49/34,8531.09(0.72, 1.63)47/35,6121.04(0.69, 1.56)33/36,2621.14(0.66, 1.96)83/43,0151.08(0.80, 1.47)
>20–30/week36/28,7160.88(0.55, 1.37)34/30,3750.84(0.53, 1.33)30/30,8731.00(0.57, 1.76)59/36,3950.91(0.65, 1.28)
>30/week28/32,6780.74(0.44, 1.23)32/34,9070.88(0.54, 1.43)30/35,8210.97(0.54, 1.74)45/41,5320.77(0.52, 1.14)
   p-trend = 0.193  p-trend = 0.432  p-trend = 0.764  p-trend = 0.165
Physical functioning
0–2037/24504.39(2.34, 8.25)37/28767.43(3.80, 14.50)      
25–4527/37673.39(1.83, 6.27)27/41124.67(2.40, 9.07)      
50–7018/91591.15(0.60, 2.19)25/96262.39(1.24, 4.58)      
75–9565/54,2451.26(0.77, 2.05)61/57,9551.62(0.93, 2.81)      
10025/56,2311.00 18/56,3721.00       
   p-trend < 0.01  p-trend < 0.01      

In analysis restricted to participants who had high levels of physical functioning at baseline, and stratified Cox model according to physical functioning scores, neither physical activity nor time spent outdoors was significantly related to hip fracture risk. Thus, neither of these factors was significantly related to hip fracture risk if physical functioning was adequately controlled for within the analysis.

The joint relationship of physical functioning and physical activity to hip fracture is shown in Table 3. Within the group of participants with physical functioning score ≥75, physical activity had no significant association with hip fracture risk. Within participants with physical functioning score <75, all physical activity levels had a similarly increased risk of hip fracture compared with the reference group (physical functioning score ≥75 and >14 physical activity sessions per week). In the model with no interaction terms, physical functioning score <75 was associated with an increased risk of hip fracture (p < 0.01, t1 = 6.22) compared with physical functioning score ≥75, but all levels of physical activity were statistically equivalent (p = 0.08, X32 = 6.78). The overall p value for the interaction between physical functioning and physical activity using the likelihood ratio test was p = 0.41 (X32 = 2.86). This suggests that there was no significant variation in the risk of hip fracture according to physical activity for different levels of physical functioning.

Table 3. Joint Consideration of Physical Functioning and Physical Activity
Physical functioningMET-adjusted physical activityRR95% CI
  1. Participants were stratified according to categories of MET-adjusted physical activity (≥5, 5–9, 9–14, >14 sessions/week) and physical functioning score (<75 versus ≥75). Models were adjusted for age, sex, menopause/menopausal hormone therapy status, smoking status, alcohol consumption, education level, BMI, height, previous non-hip fracture, and time outdoors.
Physical functioning score ≥75>141.00 
 >9–141.11(0.68, 1.80)
 >5–91.13(0.68, 1.86)
 ≤51.08(0.59, 1.97)
Physical functioning score <75>142.12(1.14, 3.92)
 >9–142.76(1.66, 4.59)
 >5–92.14(1.27, 3.60)
 ≤53.67(2.33, 5.79)

Discussion

In this large prospective study, physical activity, time outdoors, and physical functioning scores were positively related to one another and, when considered individually, were inversely associated with hip fracture risk. However, only physical functioning had a significant association with hip fracture risk when all three factors were considered together. Those participants with the most severe physical limitation (physical functioning score 0–20) had more than five times the risk of hip fracture compared with those with no physical limitation. In particular, physical activity was not significantly prospectively associated with hip fracture risk amongst those without significant physical limitation, ie, with physical functioning scores of 75–100. This is consistent with our a priori hypothesis that the previously observed protective association of increasing physical activity on hip fracture risk could be, at least in part, because of the confounding effects of disability/frailty.

The subgroup with physical functioning scores of 75–100 represents a group that is relatively homogenous with respect to functional limitation (the 75–95 group had equivalent hip fracture risk to the 100 group in all models). In contrast, it is difficult to draw conclusions in more impaired, frail populations because in such populations physical activity diminishes gradually with increasing frailty in a way that cannot be accounted for completely with statistical adjustment.

To our knowledge, no previous study has investigated all three main exposures of physical limitation, physical activity, and time outdoors together in relation to hip fracture risk. Many previous studies have focused on physical limitation[6, 7] or physical activity[4, 18] individually, without investigating their combined effects. Our results are in keeping with those reported by Stel and colleagues,[19] who found that functional limitation was an independent risk factor for fractures but that physical activity had no association with fracture risk after adjustment for confounders. However, that study included fractures at all sites and a total of only 24 hip fractures.

Strengths of this current study include the size of the study population and the study design. The 45 and Up Study is the largest cohort study of its type in Australia and is a sample of middle-aged to elderly persons drawn from the general population with comprehensive linkage to other databases.[10] We have used a well-defined, objective outcome of incident hip fracture that is almost completely captured within hospital admissions. Linked data from hospital-admitted patient data and the deaths registry were used to ensure accurate censoring from event outcomes and deaths. Finally, this was also a prospective study using baseline questionnaire data covering a number of health, socioeconomic, and demographic factors that were all collected before event outcomes.

The results of this study are likely to be generalizable to the broader Australian population because data regarding relative risks from cohort studies of this nature have been shown to be robust and not dependent on population representativeness.[20] Our age-standardized hip fracture rates were 200 per 100,000 for females and 80 per 100,000 for males compared with around 200 and 115 per 100,000, respectively, across the Australian population over a similar period.[21] Comparisons with other international studies are difficult because study populations are heterogeneous and hip fracture rates have been noted to vary as much as 15-fold between countries.[22]

This study is limited by the relatively short follow-up time, the reliability and scope of the self-reported baseline questionnaire data, and the number of hip fractures observed. For example, data were available for selected measures of physical functioning, physical activity, and time outdoors, but not for other domains such as cognitive ability or serological vitamin D status. Furthermore, missing data were in the range of 7% to 15% for the variables of interest, although this is consistent with that reported in other similar cohort studies (9%[18, 19] to 30%[23]).

Overall health status is a key risk factor for osteoporotic fractures and a potential confounder in assessing the relationship of other factors to fracture risk.[5] Physical activity, time outdoors, and physical functioning are each strongly related to health status and physical well-being because healthier persons are likely to be more active, spend more time outdoors, and have fewer functional limitations. Frailty and disability associated with physical impairment lead to reduced physical activity and time outdoors.

Frail individuals are at particular risk for hip fracture, with one study showing a 40% higher risk in frail (defined as having at least three of: unintentional weight loss, weakness, self-reported poor energy, slow walking speed, and low physical activity) compared with non-frail individuals.[24] Frailty overlaps with disability and comorbid illnesses,[25] with associated mechanisms including inactivity, weight loss, and decreased strength,[26] with consequences seen in levels of physical activity, time outdoors, and physical functioning. Here, we show that, of these, physical functioning is the measure with most utility for risk stratification and prediction of hip fracture risk.

Physical functioning encompasses an individual's ability to perform everyday tasks and reflects overall muscle functioning, balance, coordination, and fitness. Other studies have also found physical functioning to be a major contributor to the risk of hip fracture[6, 7, 27] and a useful characteristic in clustering functional outcomes in post-hip fracture patients. In this study, we used the Physical Functioning component of the Medical Outcomes Study that summarizes many domains of overall health status as it indicates an individual's actual ability, albeit self-reported, to perform everyday tasks.[11] This contrasts with physical activity and time outdoors, which are also measures of an individual's lifestyle and habits.

An association of higher levels of physical activity with lower hip fracture risk has been demonstrated in several observational studies.[18, 28] In one large prospective study of 61,200 postmenopausal women, hip fracture risk was reduced by 6% for the equivalent of every 1 hour walking per week.[4] Plausible mechanisms include maintenance of bone mineral density and muscle mass, improved balance, decreased risk of falls, and enhanced vitamin D status.[29, 30] Such findings have prompted some commentators to suggest physical inactivity as “the most salient explanatory factor for the increasingly high hip fracture rates”[30]

Our results show the primary importance of physical functioning and suggest that the apparent association between physical activity and hip fracture risk could be at least in part, because of residual confounding through failure to account for physical functioning. Such a hypothesis has support from experimental evidence. Although physical activity interventions, such as group exercise classes and home-based prescriptions, have been found to improve physical functioning (including strength, balance, and reduced falls) in older adults,[27, 31] such interventions have not been shown to reduce hip fracture risk,[5, 27, 32] although there may be a slight beneficial effect on bone mineral density (BMD).[33] This highlights that unless potential confounders are adequately included in any analysis, misleading conclusions may result when predicting outcomes with complex contributing factors such as hip fractures. When interpreted in the context of presumed causality, this may lead to the implementation of potentially ineffective interventions.

Time outdoors and hip fracture have been less thoroughly studied. As a proxy for sunlight exposure, time outdoors is hypothesized to reduce hip fracture risk through stimulating endogenous production of vitamin D and thereby improving vitamin D status. The 45 and Up Study self-reported questions on time outdoors used in this study have been validated and found to be significantly correlated with current sun exposure as measured by personal electronic dosimeters, but interestingly, not serum 25(OH)D levels.[34] This contrasts with another study that found measures of time outdoors to be a predictor of 25(OH)D levels.[35] Sunlight exposure has been reported to reduce hip fracture risk in a meta-analysis of randomized trials involving various subgroups with neurological diseases, including Alzheimer's, stroke, and Parkinson's.[36]

Interestingly, in a similar population, self-reported physical activity sessions as measured in the 45 and Up Study questionnaire have been shown to be the best predictor of serum 25(OH)D levels.[34] Given these relationships, both increasing physical activity and time outdoors, and thus higher vitamin D status, could potentially be linked to risk of hip fracture. Nevertheless, the role of vitamin D in hip fracture risk remains uncertain despite several studies on the topic.[37-39] Our previous meta-analysis showed that observational and experimental studies had conflicting results,[39] in that randomized trials showed no significant beneficial effect of vitamin D supplementation on hip fracture, whereas observational studies showed lower 25(OH) D levels in individuals with hip fracture versus those without; these results would generally be expected to be consistent if there was a true causal relationship between vitamin D and hip fracture risk. The current study shows that any association between time outdoors or physical activity and hip fracture risk disappears when more detailed characteristics of overall health status, namely physical functioning, are included in models. It is plausible that the apparent association between higher vitamin D status and reduced hip fractures found in observational studies is also the result of residual confounding, through failure to fully control for characteristics of overall health status.

A potential way forward could include high-quality, long-term randomized controlled trials of interventions designed to improve physical functioning, with hip fractures as an outcome. This will require strict accounting for baseline physical functioning levels and a particular focus on lower functioning groups. Interventions could include various modalities of physical activity and outdoor interventions as well as other factors such as nutrition, vitamin D status, cognitive training, and provision of exercise aids. Experimental evidence is likely to be particularly important in delineating the interrelationships between different facets of overall health status and in the identification of any potential effect modifiers. This will allow the targeting of specific groups at greatest risk and, hopefully, the prevention of hip fractures.

In conclusion, physical functional limitation or disability is an important independent risk factor for hip fracture amongst middle-aged and older adults. Physical activity and time spent outdoors, both strongly correlated with physical functioning, were not independently significantly associated with hip fracture risk in our study.

Disclosures

All authors state that they have no conflicts of interest.

Acknowledgments

The authors thank Dr Rosemary Korda for her help in obtaining and preparing the data.

RML and EB are supported by the Australian National Health and Medical Research Council (NHMRC). MA is supported by the UK Medical Research Council.

Authors' roles: Study design: JL, RML, and EB. Data collection: EB. Data analysis: JL. Data interpretation: JL, RML, MA, and EB. Drafting manuscript: JL. Revising manuscript content: JL, RML, MA, and EB. Approving final version of manuscript: JL, RML, MA, and EB. JL takes responsibility for the integrity of the data analysis.

Ancillary