Original Article | Published:

Prospective cohort study of body mass index and the risk of hospitalisation: findings from 246 361 participants in the 45 and Up Study

International Journal of Obesity volume 37, pages 790799 (2013) | Download Citation

Abstract

Objective:

To quantify the risk of hospital admission in relation to fine increments in body mass index (BMI).

Design, setting and participants:

Population-based prospective cohort study of 246 361 individuals aged 45 years, from New South Wales, Australia, recruited from 2006–2009. Self-reported data on BMI and potential confounding/mediating factors were linked to hospital admission and death data.

Main outcomes:

Cox-models were used to estimate the relative risk (RR) of incident all-cause and diagnosis-specific hospital admission (excluding same day) in relation to BMI.

Results:

There were 61 583 incident hospitalisations over 479 769 person-years (py) of observation. In men, hospitalisation rates were lowest for BMI 20–<25 kg m−2 (age-standardised rate:120/1000 py) and in women for BMI 18.5–<25 kg m−2 (102/1000 py); above these levels, rates increased steadily with increasing BMI; rates were 203 and 183/1000 py, for men and women with BMI 35–50 kg m−2, respectively. This pattern was observed regardless of baseline health status, smoking status and physical activity levels. After adjustment, the RRs (95% confidence interval) per 1 kg m−2 increase in BMI from 20 kg m−2 were 1.04(1.03–1.04) for men and 1.04(1.04–1.05) for women aged 45–64; corresponding RRs for ages 65–79 were 1.03(1.02–1.03) and 1.03(1.03–1.04); and for ages 80 years, 1.01(1.00–1.01) and 1.01(1.01–1.02). Hospitalisation risks were elevated for a large range of diagnoses, including a number of circulatory, digestive, musculoskeletal and respiratory diseases, while being protective for just two—fracture and hernia.

Conclusions:

Above normal BMI, the RR of hospitalisation increases with even small increases in BMI, less so in the elderly. Even a small downward shift in BMI, among those who are overweight not just those who are obese, could result in a substantial reduction in the risk of hospitalisation.

Introduction

Obesity rates have doubled or tripled in many countries over the past three decades, and in almost half of all Organisation for Economic Co-operation and Development countries 50% or more of the population is now overweight or obese.1 It is known that obesity, and to a lesser extent overweight, increases the risk of mortality2, 3 and many chronic diseases.4 However, while pooled analyses have provided reliable evidence of increasing mortality with incremental increases in body mass index (BMI) above the normal range;2, 3 there is more limited information on the precise relationship between BMI and other important measures of disease and health service use, such as hospitalisation.5, 6, 7, 8, 9, 10, 11 Moreover, while there is considerable evidence supporting the link between obesity and these outcomes (particularly total hospital use), the evidence on the risk of outcomes in people who are overweight but not obese is less clear. Further, there is little on the relationship between incremental changes in BMI and risk of being hospitalised, and how this might vary according to factors such as age (particularly in older age), pre-existing disease and other personal characteristics, and on the differential risks for hospitalisation for specific diagnoses. Studies of the gradient in risk add to the evidence on the health consequences of above-normal BMI relating to both fatal and non-fatal diseases and the likely optimal BMI.

In this study, we aim to quantify the relative risks (RRs) of hospitalisation according to BMI. We report on the relationship of increasing increments of BMI to the risk of hospitalisation, including how the relationship varies by sex and age (particularly older age groups), across levels of smoking and physical activity, and for hospitalisation for a range of diagnoses.

Materials and methods

We used data from the 45 and Up Study, an Australian cohort involving 266 724 men and women aged 45 and over from New South Wales (NSW), Australia. Potential participants were randomly sampled from the database of Australia’s universal health insurance provider, Medicare Australia, which provides virtually complete coverage of the general population, and were invited to take part. There were no exclusion criteria and the Study oversampled, by a factor of two, individuals aged 80 years and over and people resident in rural areas. Participants joining the study completed a baseline questionnaire (between January 2006 and April 2009) and gave signed consent for follow-up and linkage of their information to a number of health-related databases. The overall response rate was 18% meaning that 10% of the entire NSW population aged 45 and over were included. Further details of the Study are described in a separate publication,12 and questionnaires can be viewed at http://www.45andup.org.au.

Baseline survey data from the study participants were linked to individual death data from the NSW Registry of Births, Deaths and Marriages (to June 2010), hospital data from the NSW Admitted Patient Data Collection (July 2000–June 2010), and cancer data from the NSW Central Cancer Registry (Jan 2000–date of baseline survey). The NSW Admitted Patient Data Collection includes records of all hospitalisations in NSW, dates of admission and discharge and the primary and additional reasons for admission (coded using the International Statistical Classification of Diseases and Related Health Problems, Tenth Revision, Australian Modification (ICD-10-AM)).13 The NSW Central Cancer Registry contains information on type of cancer diagnosis and date of diagnosis, and the NSW Registry of Births, Deaths and Marriages includes information on the date of death. Data were linked probabilistically by the Centre for Health Record linkage using personal information (including full name, date of birth, sex and address), and the quality assurance data show false positive and negative rates for data linkage of <0.5% and <0.1%, respectively.14

The main exposure, BMI, was calculated from weight and height as self-reported on the baseline survey. Because of the increased risk of measurement error in extreme BMI categories, and consistent with previous well-known studies (for example, Berrington et al.2 and Whitlock et al.3), participants with BMI>50 kg m−2 or BMI<15 kg m−2 were excluded. BMI was then categorised using the following cut-points (WHO weight classification15 in brackets): 15– (underweight); 18.5–, 20– and 22.5– (normal weight); 25– and 27.5– (overweight); 30– and 32.5– (obese class I); and 35– (obese class II–III). Data on potential confounding and modifying factors were derived from self-reported data from the baseline questionnaire.

The main outcome of interest was incident hospitalisation, defined as the first hospital admission after recruitment into the study, which involved at least one night in hospital (that is, day admissions to hospital were excluded), ascertained through linkage to the Admitted Patient Data Collection. We also examined incident hospitalisation for specific diagnoses, defined according to final primary diagnosis at discharge, based on the ICD-10-AM code (specific codes listed in Table 3). Where the admission was part of a transfer, the primary diagnosis of the first admission in the transfer was selected. The diagnoses examined in the analysis were selected based on a priori knowledge of conditions known to be associated with obesity,4 or to be major contributors to the burden of disease in Australia,16 or if they were among the top 20 diagnoses in this study (excluding rehabilitation and surgical follow-up care and complications following a procedure).

Survival methods were used to analyse the data, using age as the underlying time variable. Participants were followed from the date of recruitment to either the date of first hospital admission (any or cause-specific) or death, or 30 June 2010 (the last date to which hospital data were available), whichever occurred first. Incident hospitalisation rates, standardised using 5-year age groups, were calculated according to baseline BMI. Rates were stratified by sex and age group (45–64, 65–79 and 80 years). We also examined rates stratified by smoking (never, past and current) given its strong association with both BMI and health outcomes, and by prior serious illness and tertile of physical activity given the complex relationship between these factors and BMI, including that they may lie on the causal pathway between BMI and health outcomes. Prior history of serious illness was defined as at least one of the following: a cancer diagnosis in the 5 years before enrolment (excluding non-melanoma skin cancer), ascertained from the cancer register; a hospital admission within the 5 years before enrolment that included a diagnosis of circulatory disease, diabetes or chronic obstructive pulmonary disease, ascertained from the hospital admission data; or a doctor diagnosis of either cancer, heart disease, stroke or diabetes self-reported on the questionnaire. Physical activity was assessed with the Active Australia questionnaire,17 which has documented reliability and validity.17, 18 A score is generated based on the number of weekly sessions of walking, moderate and vigorous activity, weighted for intensity. This was categorised into tertiles. Interaction effects were tested using the likelihood-ratio test, comparing models with and without the interaction terms.

Cox regression was used to estimate the hazard ratios (referred to henceforth as relative risks (RRs)) for incident hospitalisation according to BMI. For all-cause hospitalisations, the analyses were performed separately for men and women and were stratified by age group (45–64, 65–79 and 80 years). The reference BMI group was 20–<22.5 kg m−2 and tests for linear trend were also performed excluding those with a BMI of <20 kg m−2. Adjustment was made for potential confounding variables, including area of residence (major city, regional or remote based on the Accessibility/ Remoteness Index of Australia Plus,19 derived from postcode of residence), education (six categories from ‘did not complete secondary school’ to ‘university degree’), pre-tax annual household income (seven income brackets, from AUD ‘<$5000’ to ‘$70 000 or more per year’), smoking (never, past and current), alcohol intake (0/1–14/15 or more drinks per week), and private health insurance (additional to universal health insurance). Participants with any missing values on any of these variables were assigned to a separate category for that variable. In addition, we performed a sensitivity analysis to examine the possible impact of any potential reverse causality, that is, the effect of illness on baseline BMI, by excluding all py and hospitalisations in the first year of follow-up. We also examined the effects of additional adjustment for country of birth, classified into 13 regions, with a separate category for ‘Australian-born’ based on a modified version of that used in the Global Burden of Disease Study.20

For hospitalisations for specific diagnoses, we stratified by age group, but combined males and females, and used broader BMI categories, comparing rates in overweight (25–<30 kg m−2) and obese (30–50 kg m−2) to normal weight (18.5–<25 kg m−2), in order to retain statistical power, while adjusting for all confounders including sex. The proportional hazards assumption was satisfied for all models. All analyses were performed using Stata version 12.0.21

Ethics approval for this project was obtained from the NSW Population and Health Services Research Ethics Committee and the Australian National University Human Research Ethics Committee.

Results

After excluding those with BMI <15 kg m−2 (n=303) or >50 kg m−2 (n=2463) or missing (n=17 600), we followed 246 361 participants (92.4%), with the average time between baseline survey and end of the study period (or death) being 2.3 years (range: 1 day–4.5 years).

The characteristics of the sample are shown in Table 1. Just under half the sample (47%) comprised men, with the majority of participants (62%) aged 45–64 years and 10% aged 80 years or older. Regarding BMI, 62% of the sample were overweight or obese (40% and 22%, respectively), two-thirds of men (47% overweight and 22% obese) and just over half of the women (33% overweight and 23% obese); this proportion was considerably lower in those who were aged 80 years or older (48% overweight/obese).

Table 1: Characteristics of the study population (%)a by category of BMI

There were 61 583 incident hospitalisations over 479 769 py of observation; a rate of 128 per 1000 py (95% confidence interval (CI): 127–129) overall, 143 (95% CI: 142–145) in males and 115 (95% CI: 114–117) in females. Age-adjusted incident hospitalisation rates according to BMI are presented in Figure 1. Among men, incident hospitalisation rates were lowest in those with a BMI between 20 and 25 kg m−2 (age-adjusted rate of 120 per 1000 py) and among women with a BMI of between 18.5 and 25 kg m−2 (102 per 1000 py); rates were highest in the highest BMI category (35–50 kg m−2) for both men and women (rates of 203 and 183 per 1000 py, respectively). Excluding those with a BMI<20 kg m−2, there was a significant linear trend of increasing rate of hospitalisation with increasing BMI (P<0.001 for both males and females).

Figure 1
Figure 1

Age-standardised incident hospitalisation rates by BMI. Diamonds and squares correspond to the median BMI in each category. Test for trend where BMI20 kg m−2. P<0.001 for both males and females.

When examined separately by age group (Figure 2), while the absolute rates of hospitalisation rose with increasing age, as expected, the pattern of rising relative rates of hospitalisation with increasing BMI was attenuated with increasing age. After adjustment for potential confounding variables (see Table 2), the trend of increasing hospitalisation with increasing above-normal BMI remained apparent with trends in both men and women stronger in the youngest age group (RR 1.04 per 1 kg m−2 increase) and weaker in the 80 age group (RR 1.01 per 1 kg m−2 increase). Among those aged 45–64, the risk of hospitalisation for the severely obese group (BMI 35–50 kg m−2) was around twice that of the group with normal BMI (male RR:1.86, 95% CI:1.69–2.05; female RR:2.09; 95% CI:1.96–2.23), and among those aged 65–79 years the corresponding risks were around 50–60% higher (male RR:1.48, 95% CI:1.34–1.64; female RR:1.61; 95% CI:1.47–1.76). In people aged 80 years or older, although risks were elevated in those with a BMI of 27.5 kg m−2 or higher and the overall trend was significant, there was considerable overlap in the confidence intervals of these risk estimates. Additional adjustment for region of birth made no material difference to the results (not shown). The sensitivity analysis excluding the first year of follow-up (n=242 829; 39 821 events over 282 243 py) showed little change in the RRs (see Supplementary Table).

Figure 2
Figure 2

Age-standardised incident hospitalisation rates (with 95% CI) by BMI, stratified by sex and age group. Squares correspond to the median BMI in each category. Tests for trend are for BMI 20 kg m−2. Different scales are used for different age groups.

Table 2: Adjusted RRa of hospitalisation by BMI, stratified by sex and age group

To allow for closer examination of the effect of prior serious illness, smoking status and physical activity on the relationship between BMI and hospitalisation, age-adjusted rates are also presented separately for those with and without serious illness, by current, past and never smokers and by tertile of physical activity (Figure 3). These show that the trend of increasing rate of hospitalisation as BMI increased from normal weight to severely obese was apparent regardless of history of prior serious illness, smoking status or physical activity level. These trends remained in the Cox regression analyses, following adjustment for potential confounding factors, and there was no significant difference in the relationship between BMI and hospitalisation in any sex/age group examined, except among males aged 45–64 (P(interaction)=0.002), where the strength of the relationship was weaker among current smokers than past or never smokers (RRs per 1 kg m−2 increase in BMI where BMI>20 kg m−2 of 1.02 (1.01–1.03), 1.04 (1.03–1.05) and 1.04 (1.04–1.05), respectively), and among males aged 80, where the effect size was stronger in those with high than lower levels of physical activity (RRs per 1 kg m−2 increase in BMI where BMI>20 kg m−2 of 1.00 (0.98–1.01), 1.02 (1.01–1.03) and 1.04 (1.02–1.05), in lowest, middle and top tertile of physical activity, respectively).

Figure 3
Figure 3

Age-standardised hospitalisation rates (with 95% CI) by BMI, stratified by (a) prior history of serious illness; (b) smoking status; and (c) tertile of physical activity. Diamonds, squares and triangles correspond to the median BMI in each category. Prior serious illness includes circulatory disease, diabetes, chronic obstructive pulmonary disease and cancer. Test for trend where BMI20 kg m−2; P<0.001 for all strata of illness, smoking and physical activity, in both males and females.

The rates of hospitalisation due to specific diagnoses in overweight and obese individuals, relative to those of normal weight (18–<25 kg m−2), are presented in Table 3 for the three age groups (males and females combined). For most diagnoses there was an increased risk of hospitalisation associated with obesity and to a lesser extent overweight, with effect sizes generally weaker with increasing age. The diagnoses for which these relationships were particularly strong—that is, where the effect sizes were at least moderate, showed a gradient from overweight to obese and were evident in both younger and older age groups—included diabetes, ischaemic heart disease, chest pain, diverticular disease, gallbladder disease, osteoarthritis, asthma (excluding 65–79 year age group), sleep apnoea and cellulitis. Risks were also elevated in the 45–64 year old age group for atrial fibrillation and cerebrovascular disease, as well as for depression (a non-significant reverse trend was shown in the older people) and prolapsed uterus. Overweight/obesity appeared to be protective for two conditions—hernia and fracture (particularly in the 80 age group).

Table 3: Adjusted relative risks of diagnosis-specific hospitalisation by BMI category, by age group, males and females combined

Discussion

The risk of hospitalisation rises gradually with increasing above-normal BMI, in both men and women. This pattern of increasing risk is evident among past, current and never smokers and among both physically active and inactive people. The RRs of hospitalisation associated with above-normal BMI are highest among younger mid-age adults (45–64 years), with the risk of hospitalisation among those with a BMI>35 kg m−2 approximately double that of those with BMI 20–<22.5 kg m−2. RRs are not as high in the older mid-age people (65–79 years) but are still substantial, while the pattern of increasing RR of hospitalisation with increasing above-normal BMI is fairly weak in older people (80 years and older). The increased risk of hospitalisation associated with overweight and, to a greater extent, obesity is evident for a large range of conditions in all the age groups, including those aged 80 years or more.

Causality between BMI and hospitalisation cannot be confirmed in this study. However, the dose–response relationship between BMI and hospitalisation among those with a BMI25 kg m−2, which remained after adjusting for a range of important confounders and after stratification for smoking and physical activity, and which was evident for a large range of causes consistent with previous evidence,4 supports a causal relationship. Together with the knowledge that hospitalisation is a costly and not uncommon event, and that there is a high prevalence of overweight and obesity in the population (in Australia, 40% of people aged 45 and over are overweight and 30% are obese22), the findings indicate that overweight, not just obesity, is a substantial contributor to health service use and health costs in the population.

The findings of increasing risk of hospitalisation with increasing BMI are broadly consistent with other studies, which have shown that obesity, but less consistently overweight, is associated with increased rates of hospitalisation, in either the shorter or longer term.5, 6, 7, 8, 9, 10, 11 It is also broadly consistent with findings of stronger associations among mid-age than older people in some studies,9, 11 although others have found no clear patterns with respect to age.6 While caution should be applied in interpreting the absolute value of BMI at which risk increases, the level at which risk of hospitalisation was lowest in this study—20–<25 in men and 18.5–<25 kg m−2 in women—is also consistent with these previous studies, as well as with studies on BMI and incidence of disease4 and all-cause mortality.2, 3 That risks were lowest in these ranges is also concordant with the WHO definition of 18.5–<25 kg m−2 being in the ‘normal’ range, with the exception that in our study risk was actually elevated in the low normal weight range (18.5–<20 kg m−2) among the younger men. This may partly reflect reverse causality due to pre-existing disease, as hospitalisation rates were not elevated in this low normal weight range among people without serious pre-existing disease.

We found overweight, and to a greater extent, obesity is associated with an increased risk of hospitalisation for a heterogeneous range of conditions. These included potentially life-threatening diseases such as cardiovascular disease, stroke (in younger obese men), diabetes and asthma, and also common, high-cost and high-morbidity conditions, such as osteoarthritis, diverticular disease, dorsopathy, gallbladder disease, sleep apnoea and uterine prolapse. Although we also found above-normal BMI is associated with a reduced risk of hernia and fracture, particularly in the older age group, the sum total of the relationship between BMI and hospitalisation is adverse. While the association between obesity and most of these conditions is fairly well established,4, 23, 24, 25, 26, 27, 28, 29, 30, 31 their investigation in a single population-based data set is relatively novel and allows both a general perspective on the conditions associated with both obesity and overweight, as well as insights into the diagnoses underlying the observed elevation in overall hospitalisation rates.

The 45 and Up Study is an order of magnitude larger than previous studies investigating the relationship of BMI to hospitalisation, enabling the risks associated with small incremental changes in BMI to be examined. This is particularly useful as dose–response relationships can provide a good indication of the magnitude of risk and likely causality. The inclusion of individuals aged over 80 also meant that we had the power to examine variations in risk in the very old, an increasingly important group as the population ages, yet one which is often excluded (or at best aggregated) in population health studies. Another advantage of the study was the ability to link to administrative records, allowing virtually complete, and objective, ascertainment of outcomes over time. Use of hospitalisation records meant we could capture the risks of both fatal and non-fatal serious diseases associated with BMI. The ability to link these records to survey data meant that we were able to adjust for a range of factors other than BMI that might influence hospitalisation.

A limitation of our study is that BMI was the only anthropometric measure available and this measurement was based on self-reported weight and height at recruitment. It is well established that people tend to underestimate their weight and overestimate their height (and hence underestimate their BMI).32 However, consistent with other studies, a validation study involving participants in the 45 and Up Study revealed that the mean difference between self-reported and measured BMI was not large (on average −0.74 kg m−2) and correlations between self-reported and measured height and weight were 0.95 and 0.99, respectively.33 This means that self-reported data are useful for research investigating the RRs of disease and other health outcomes comparing incremental changes in BMI, and although bias in the RR estimates cannot be ruled out, a recent study using National Health and Nutrition Examination Survey data showed that mortality RRs based on self-reported BMI data were for the most part not significantly different from those based on measured BMI.34 Further, BMI was only measured at one point in time, which does not take into account duration of obesity—this may enhance risk estimates over and above BMI level alone.35 Another limitation is that, like most population-based cohort studies, response rates were relatively low and the potential for a ‘healthy cohort effect’ means that the estimates of RRs shown here are likely to be conservative. We estimated the age-adjusted total hospitalisation rates (excluding day admissions) in the 45 and Up cohort (233 and 217 per 1000 py for males and females, respectively) and found these were only 11–12% lower than for the rates in the Australian population of the same age (respectively, 263 and 246 per 1000 py).36 Moreover, it is important to note that RRs, comparing groups within the cohort, remain valid and can be generalised more broadly.37, 38

It is evident from this and previous research that overweight and obesity are associated with an increase in the risk of hospitalisation across a large range of diagnoses. While it is reasonable to attribute at least part of the increased risk to increased adiposity, some caution should be applied. First, BMI is an imperfect measurement of adiposity and in different populations such as young fit men with high muscle mass or the elderly, BMI alone may not be a good measure of adiposity. Second, although longitudinal in design, we cannot totally disentangle the effects of complex disease pathways, and in particular reverse causality. The effect of disease on subsequent weight loss is especially an issue for cancer-related diagnoses, although the relationship between BMI and hospitalisation remained even when those with a diagnosis of cancer and other serious illness before entry into the study were excluded, and also when the first year of follow-up was excluded. Any such reverse causality may lead to an underestimate of the effect of above-normal weight on hospitalisation. It can also lead to an overestimate of the effect of underweight and these effects are observed in studies with longer follow-up.2, 3 In some cases, reverse causality may also work in the opposite direction, leading to an overestimation of effects. For example, with sleep apnoea not only has obesity been linked to an increased risk of developing this condition, but sleep apnoea can also lead to an increase in obesity.28 Even the relationship between obesity and cardiovascular disease is complex and not limited to the standard risk factors such as hypertension, dyslipidemia and diabetes.10 Some studies on BMI and hospitalisation adjust for pre-existing disease (such as cardiovascular disease and diabetes) to limit the possibility of reverse causality, but as pre-existing disease may lie on the causal pathway between above-normal BMI and hospitalisation, adjusting for these diseases could constitute overadjustment and inappropriately attenuate risks.

While previous studies have adjusted for smoking, there are concerns that because smoking is so strongly associated with illness, and also BMI, and the amount of smoking may be imperfectly measured, that simple adjustment for smoking is insufficient. By stratifying our results according to smoking, we were able to show clearly the risks of hospitalisation associated with BMI independently from smoking status. Similarly, the relationship between physical activity, BMI and hospitalisation is complex, and by stratifying by physical activity we were able to show the effect of BMI on hospitalisation was evident regardless of physical activity level. That this finding was shown to hold even in the inactive population is important, as BMI as a measure of adiposity is most valid in this population.39

It is unclear why the strength of the association between BMI and hospitalisation declines with increasing age; there are a number of possible explanations. First, in elderly people, BMI may not be a good measure of adiposity because as one ages, height decreases, lean body mass decreases, and adipose tissue increases without weight gain, with the possible development of sarcopenic obesity.40 There is also a greater probability of reverse causality because older people have a greater number of health conditions that result in weight loss. Second, a likely contribution to the decline in the association between BMI and hospitalisation with age is the survival effect, as those with higher BMIs and/or those most susceptible to the effects of BMI on health may have already died. Third, it may be possible that being overweight in later life confers a genuine health benefit, as frailty and underweight become increasingly important contributors to ill health. Finally, the number of people in the oldest age group, particularly those with high BMIs, is limited, resulting in reduced power in the elderly. Importantly it should be kept in mind that while RRs were lower in older than younger people, this may not necessarily translate to a lower absolute burden of hospitalisation attributable to overweight/obesity in older people than younger people; burden must also take into account the age-specific prevalence of overweight and obesity in the population (which is lowest in those aged 80) and the absolute rates of hospitalisation (which are highest amongst those aged 80).

There has been some uncertainty about the relationship between BMI and the risk of disease and hospitalisation, in particular regarding the level of BMI at which risk starts to increase in the population and whether the risk associated with obesity persists in the elderly. The incremental nature of the increasing risk of hospitalisation above the normal range of BMI found in this study shows that at the population level, even mild overweight increases risk, especially among those aged below 80 years of age. That a dose–response relationship was found after adjustment for a range of important confounders gives strength to the argument that above-normal BMI is causally linked to many diseases. The findings suggest that, all else being equal, populations with even a minor difference in the BMI distribution towards lower values could have substantially lower hospitalisation rates.

References

  1. 1.

    Organisation for Economic Co-operation and Development. Obesity and the Economics of Prevention: Fit not Fat. OECD: Paris, 2010.

  2. 2.

    , , , , , et al. Body-mass index and mortality among 1.46 million white adults. N Engl J Med 2011; 363: 2211–2219.

  3. 3.

    , , , , , et al. Body-mass index and cause-specific mortality in 900 000 adults: collaborative analyses of 57 prospective studies. Lancet 2009; 373: 1083–1096.

  4. 4.

    , , , , , . The incidence of co-morbidities related to obesity and overweight: a systematic review and meta-analysis. BMC Public Health 2009; 9: 88.

  5. 5.

    , , , , , . Impact of overweight and obesity on hospitalization: race and gender differences. Int J Obes (Lond) 2009; 33: 249–256.

  6. 6.

    , , , . Obesity and use of acute hospital services in participants of the Renfrew/Paisley study. J Public Health (Oxf) 2007; 29: 53–56.

  7. 7.

    , . Obesity and hospitalization over the adult life course: does duration of exposure increase use? J Health Soc Behav 2007; 48: 434–449.

  8. 8.

    , , , , , . Obesity, survival, and hospital costs-findings from a screening project in Sweden. Value Health 2005; 8: 562–571.

  9. 9.

    , , , , . Body mass index and hospitalization in the elderly. J Am Geriatr Soc 2003; 51: 1615–1620.

  10. 10.

    , , , , , et al. Midlife body mass index and hospitalization and mortality in older age. JAMA 2006; 295: 190–198.

  11. 11.

    , , . Obesity, health services use, and health care costs among members of a health maintenance organization. Arch Intern Med 1998; 158: 466–472.

  12. 12.

    , , , , , et al. Cohort profile: the 45 and up study. Int J Epidemiol 2008; 37: 941–947.

  13. 13.

    National Centre for Classification in Health. The International Statistical Classification Of Diseases And Related Health Problems, 10th revision, Australian Modification (ICD-10-AM) 3rd edn. University of Sydney: Sydney, 2002.

  14. 14.

    CHeReL. Centre for Health Record Linkage. < >, 2011.

  15. 15.

    World Health Organization. WHO: Global database on body mass index < >, 2006.

  16. 16.

    , , , , , . The burden of disease and injury in Australia 2003. AIHW cat. no. PHE 82. Australian Institute of Health and Welfare: Canberra, 2007.

  17. 17.

    Australian Institute of Health and Welfare. The Active Australia Survey: A Guide and Manual for Implementation, Analysis and Reporting. Canberra: AIHW, 2003.

  18. 18.

    , , , . Reliability and validity of a modified self-administered version of the Active Australia physical activity survey in a sample of mid-age women. Aust NZ J Public Health 2008; 32: 535–541.

  19. 19.

    Australian Institute of Health and Welfare. Rural, regional and remote health: a guide to remoteness classifications. Rural health series No. 4. AIHW cat. no. PHE 53. AIHW: Canberra, 2004.

  20. 20.

    Harvard University Institute for Health Metrics and Evaluation at the University of Washington Johns Hopkins University, University of Queensland World Health Organization. The Global Burden of Diseases, Injuries, and Risk Factors Study, Operations Manual Final Draft 2009.

  21. 21.

    Stata statistical software: release 10.1 [program].. StataCorp LP: College Station, TX, 2009.

  22. 22.

    Australian Bureau of Statistics. National Health Survey: Summary of Results 2007–08. ABS cat. no. 4364.0. ABS: Canberra, 2009.

  23. 23.

    , , , , . Obesity increases the risks of diverticulitis and diverticular bleeding. Gastroenterology 2009; 136: 115–122 e1.

  24. 24.

    , , . Excess weight and sleep-disordered breathing. J Appl Physiol 2005; 99: 1592–1599.

  25. 25.

    , , , , , et al. Bidirectional association between depression and obesity in middle-aged and older women. Int J Obes (Lond) 2012; 36: 595–602.

  26. 26.

    , , , , , et al. Factors predisposing to acute and recurrent bacterial non-necrotizing cellulitis in hospitalized patients: a prospective case-control study. Clin Microbiol Infect 16: 729–734.

  27. 27.

    , . Risk factors for inguinal hernia among adults in the US population. Am J Epidemiol 2007; 165: 1154–1161.

  28. 28.

    , . Sleep & the metabolic syndrome. Indian J Med Res 2010; 131: 206–216.

  29. 29.

    , , , . Obesity is a risk factor for fracture in children but is protective against fracture in adults: A paradox. Bone 2011; 48: 189–196.

  30. 30.

    , , . Pelvic organ prolapse. Lancet 2007; 369: 1027–1038.

  31. 31.

    , , , , . Body mass index and cardiovascular disease in the Asia-Pacific Region: an overview of 33 cohorts involving 310 000 participants. Int J Epidemiol 2004; 33: 751–758.

  32. 32.

    , , , . A comparison of direct vs self-report measures for assessing height, weight and body mass index: a systematic review. Obes Rev 2007; 8: 307–326.

  33. 33.

    , , , , , et al. Validity of self-reported height and weight and derived body mass index in middle-aged and elderly individuals in Australia. Australian and NZJ of Public Health 2011; 35: 557–563.

  34. 34.

    , , , , . Use of self-reported height and weight biases the body mass index-mortality association. Int J Obes 2011; 35: 401–408.

  35. 35.

    , , , , , et al. The number of years lived with obesity and the risk of all-cause and cause-specific mortality. Int J Epidemiol 2011; 40: 985–996.

  36. 36.

    Australian Institute of Health and Welfare. Separation statistics by principal diagnosis in ICD-10-AM, Australia, 2008-09 to 2009-10 < >AIHW 2011.

  37. 37.

    , , , , , . Investigation of relative risk estimates from studies of the same population with contrasting response rates and designs. BMC Med Res Methodol 2010; 10: 26.

  38. 38.

    , , . Is this finding relevant? Generalisation and epidemiology. Aust NZJ Public Health 1996; 20: 54–56.

  39. 39.

    World Health Organization. Physical Status: The Use and Interpretation of Anthropometry. WHO Technical Report Series 854: 9 WHO: Geneva, 1995.

  40. 40.

    , . Sarcopenia and obesity. Clin Geriatr Med 2011; 27: 401–421.

Download references

Acknowledgements

We thank the men and women participating in the 45 and Up Study. The 45 and Up Study is managed by The Sax Institute in collaboration with major partner Cancer Council New South Wales; and partners the National Heart Foundation of Australia (New South Wales Division); New South Wales Department of Health; beyondblue: the national depression initiative; Ageing, Disability and Home Care, Department of Human Services New South Wales; and Uniting Care Ageing. We also acknowledge the support of the Centre for Health Record Linkage. This specific project was supported by Australian National Health and Medical Research Council (NHMRC) Project Grant number 585402 and arose as an initiative of the MBF Policy in Action Roundtable, funded solely by the Bupa Health Foundation. Funding sources had no role in study design; in the collection, analysis, and interpretation of data; in the writing of the report; nor in the decision to submit the article for publication. Emily Banks and Bette Liu are supported by the NHMRC.

Data sharing: The 45 and Up Study is an accessible data resource for approved research projects; see www.45andUp.org.au for details.

Author information

Affiliations

  1. National Centre for Epidemiology and Population Health, The Australian National University, Canberra, Australian Capital Territory, Australia

    • R J Korda
    •  & E Banks
  2. Australian Centre for Economic Research on Health, The Australian National University, Canberra, Australian Capital Territory, Australia

    • R J Korda
  3. The Kirby Institute, The University of New South Wales, Sydney, New South Wales, Australia

    • B Liu
  4. Department of Medical Epidemiology and Biostatistics, Karolinska Institutet, Sweden

    • M S Clements
  5. Sydney School of Public Health, The University of Sydney, Sydney, New South Wales, Australia

    • A E Bauman
  6. School of Medicine, University of Western Sydney, Sydney, New South Wales, Australia

    • L R Jorm
    •  & H J Bambrick
  7. The Sax Institute, Sydney, New South Wales, Australia

    • L R Jorm
    •  & E Banks

Authors

  1. Search for R J Korda in:

  2. Search for B Liu in:

  3. Search for M S Clements in:

  4. Search for A E Bauman in:

  5. Search for L R Jorm in:

  6. Search for H J Bambrick in:

  7. Search for E Banks in:

Competing interests

The authors declare no conflict of interest.

Corresponding author

Correspondence to R J Korda.

Supplementary information

About this article

Publication history

Received

Revised

Accepted

Published

DOI

https://doi.org/10.1038/ijo.2012.155

Supplementary Information accompanies the paper on International Journal of Obesity website (http://www.nature.com/ijo)

Further reading