Cohort Profile: HABITAT—a longitudinal multilevel study of physical activity, sedentary behaviour and health and functioning in mid-to-late adulthood

Abstract

The benefits of physical activity in reducing the risk of non-communicable diseases are well documented.1,2 Physical inactivity contributes to 6–10% of the burden of coronary heart disease, type 2 diabetes, and breast and colon cancers.1,3 Physical activity helps reduce waist circumference, blood pressure and cholesterol,2,4 and may play a key role in the prevention and management of poor mental health.5 Recent evidence demonstrates that regular physical activity is particularly important for healthy ageing.6–8 Physical activity at older ages reduces the risk of falls, musculoskeletal conditions, disability and functional/cognitive decline, anxiety and depression; and promotes longevity, health-related quality of life and wellbeing.9 Thus, understanding the patterns, prevalence and determinants of physical activity participation is key to understanding population health and healthy ageing. Ecological models of health behaviour posit that there are multiple levels of influence on physical activity, including individual (e.g. biological, psychological), social (e.g. social support and norms), organizational (e.g. social institutions), environmental (e.g. neighbourhood walkability, recreational facilities) and policy (e.g. legislation).10 These models provide a framework for understanding the multiple and interacting determinants of physical activity, which in turn can inform comprehensive interventions that target individuals and the environments in which they live.11,12 Few studies have been designed to prospectively assess the multilevel determinants of physical activity (i.e. simultaneously examining area-, group- and individual-level effects on physical activity outcomes).13–15 Those that do rely primarily on secondary (i.e. published) data sources, thus lacking the data to study specific physical activities and the environments and contexts relevant for those activities.16 Further, although other large prospective studies include measures of physical activity, few have investigated the influence of multilevel factors in the context of change in people’s activity levels over time as they age.17–19 The HABITAT Study commenced in 2007 with a cohort of men and women aged 40–65 years (n = 11 035) living in 200 neighbourhoods in the Brisbane Local Government Area, Australia. Funded by two nationally competitive project grants, the overarching aim of Phase One of HABITAT was to examine change in physical activity, and investigate the relative contributions of environmental, social, psychological and sociodemographic factors, to these changes (Figure 1). Additional funding awarded in 2013 allowed an expansion of focus (Phase Two) to include objective assessment of physical activity, sedentary behaviour and physical functioning. Sedentary behaviours are defined as any waking activities characterized by an energy expenditure ≤1.5 METs in a sitting or reclining posture.20 The MET (metabolic equivalent of task) is a measure of energy expenditure whilst engaging in an activity relative to energy expenditure at rest. Sedentary behaviours have synergistic effects with physical activity and have been associated with a range of outcomes relevant to healthy ageing21–25 The primary objectives of Phase Two were to assess the role of physical activity and sedentary behaviours in preventing, delaying or accelerating declines in physical function as people age; and to examine how associations between physical activity, sedentary behaviour and trajectories of physical functioning are influenced by environmental, social, psychological and sociodemographic factors. HABITAT study conceptual framework Initial sample selection used a two-stage design, whereby study areas were selected first, and individuals chosen subsequently (see Burton et al.26 for more detail). Study areas were Australian Bureau of Statistics (ABS) Census Collection Districts (CCDs), each of which typically contains ∼200 dwellings in urban areas.27 CCDs were ranked into deciles using the ABS’ Index of Relative Socioeconomic Disadvantage (IRSD),28 with 20 areas per decile randomly selected (n = 200). A CCD’s IRSD score reflects the area’s overall level of disadvantage measured based on multiple social and economic components including education, occupation, income, unemployment, household structure and motor vehicle availability. Using systematic probability sampling without replacement and proportional to the number of households per CCD with at least one person aged 40–65 years, an average of 85 households per area were selected, and one person per household was randomly chosen to participate. A structured self-administered questionnaire was developed (available at https://cur.org.au/project/habitat/) and copies were sent to 17 000 potentially eligible participants in May 2007 using a mail survey methodology developed by Dillman.29 Completed questionnaires were returned by 11 035 eligible participants, with 841 refusals and 4251 non-responders (response = 68.4%). Compared with 2006 census data, the sociodemographic characteristics of the HABITAT cohort at baseline were broadly representative of the Brisbane population aged 40–65 years (Table 1). Socio-demographic profile of the HABITAT Study cohort at baseline (2007) compared with the Brisbane population aged 40–65 years (2006); and the socio-demographic profile of the HABITAT Study cohort in 2016 (sample participants aged 49–74 years) Based on Australian Bureau of Statistics 2006 Census data (i.e. closest Census to baseline data collection). The survey question pertaining to education was only asked at baseline (2007). Number of participants with valid education data in 2016 was 5081: those who returned a completed survey in 2016 and who were not the sampled participant in 2007 (n = 106) were excluded from the percentage calculation as their education level was unknown. Category includes the retired, home duties, unemployed and permanently unable to work. Quintile 1 contains the 20% least disadvantaged neighbourhoods (CCDs) and quintile 5 contains the 20% most disadvantaged neighbourhoods. Since the baseline data collection in 2007, the cohort (i.e. those who responded in 2007 and who had not actively withdrawn) have been approached to complete four follow-up surveys (2009, 2011, 2013 and 2016). Surveys were sent out in winter (May to July) each time to minimize potential seasonal effects on physical activity.30–32 Participants who relocated from their baseline address to elsewhere in Brisbane or Australia (‘movers’) remained eligible to participate, and as at 2016, 17.4% (n = 1916) of the HABITAT cohort has changed address on at least one occasion. Multiple strategies were used to optimize cohort maintenance, including: personalized communication; the collection of contact details about a family member or friend who did not live with the participant in case the participant moved or contact was lost; a study newsletter and Christmas card; the inclusion of a change-of-address card with most correspondence; a study website, email address and free-call phone number; and the use of survey front-covers that were customized for each suburb to orient participants to provide data about their local area rather than Brisbane in general. At Wave 5 in 2016, 77.8% (n = 8588) of the baseline participants remained in the study and 22.2% (n = 2447) were classified as ‘withdrawn’ for a range of reasons: voluntarily withdrew (n = 1474), deceased (n = 311), moved overseas (n = 75), physical or cognitive incapacity (n = 49), language impairment (n = 8) and lost to follow-up/uncontactable (n = 530). The proportion of participants who did not return a survey after five attempts to contact them via postal letter (classified as ‘non-respondents’) has risen incrementally across each successive follow-up wave (Table 2). Analysis of study attrition shows that loss to follow-up has been higher for older persons, the least educated, blue collar workers and persons not in the labour force, and members of lower income households. HABITAT Study participant attrition: 2007–16 As part of Phase Two, a sub-sample was randomly selected and invited to take part in a physical function sub-study in 2014. The sub-study assessments were conducted individually at a time and location negotiated with participants, and typically occurred at their residence. Assessments were conducted from July to February, with participants interviewed at approximately the same time of year at each wave of assessment. From the 1559 HABITAT respondents invited to participate, 154 were deemed ineligible (n = 49 no longer living in selected study area; n = 96 no valid phone number/unable to contact; n = 3 deceased; n = 6 unable to stand or walk without assistance/language difficulties) and 767 were assessed in 2014/2015 (response = 54.6%). Follow-up assessments were completed in 2016/17 (n = 606, response = 79.0%). Socio-demographic characteristics of the sub-sample at baseline and first follow-up are in Table 3 Socio-demographic characteristics of the HABITAT clinical sub-study participants: 2014 (baseline) and predictors of loss to follow-up between 2014 and 2016 Modelled using multilevel logistic regression, with model parameters [expressed as odds ratios (OR) and 95% credible intervals (CrI)] estimated by Markov chain Monte Carlo simulation using MLwiN software.33 Age and sex adjustment only. Simultaneous adjustment for all sociodemographic characteristics. Quintile 1 contains the 20% least disadvantaged neighbourhoods (CCDs) and quintile 5 contains the 20% most disadvantaged neighbourhoods. Baseline survey measures were pilot tested and assessed for test-retest reliability,34 and new items used in follow-up surveys are based on validated measures (where possible). In addition to assessing multilevel determinants specific to physical activity, other items included in the survey relate to healthy ageing, socio-economic disadvantage, social determinants of health, life events, psychological wellbeing, transportation and health (physical and mental). Given the complexity of physical activity behaviour, several domain-specific measures were incorporated in the survey. Items from the Active Australia survey assess the frequency and total time spent during the previous week (i) walking for recreation, exercise or to get to or from places, (ii) doing vigorous gardening or heavy work around the yard, (iii) doing vigorous physical activity (e.g. jogging) and (iv) other more moderate physical activity (e.g. slow swimming).35 These items have acceptable levels of reliability and validity and have been recommended for use in population-based monitoring of physical activity in Australia.36,37 In addition, we measured frequency of participation in each of 15 specific active recreation pursuits (e.g. running, tennis, etc.) which were derived from the Exercise, Recreation and Sport Survey,38 and the total time spent in the previous week walking for transport, cycling for transport, walking for recreation, and cycling for recreation. Sedentary behaviour is assessed as sitting time (hours, minutes) on a usual weekday and weekend day across four domains: whilst travelling, watching television (including gaming), in general leisure, and using a computer at home. This measure has been shown to be more reliable and valid for weekdays than weekends, and more valid for assessment of sitting at work, watching television, and computer use at home, than for other domains.39 In 2013 and 2016 self-reported physical functioning was assessed using the 10 item Physical Functioning Scale (PF-10), a component of the Short Form 36 Health Survey.40 The PF-10 has been extensively validated41 and measures a hierarchical range of difficulties, from vigorous activities such as lifting heavy objects to bathing and dressing. Participants who moved to a different address between data collection waves were sent a survey that contained additional questions to those included in the non-movers survey. The questions asked about the reasons for moving (e.g. to buy a bigger home; commence a new job; relationship breakdown) and the reasons for choosing the new address (e.g. closeness to work, childcare, public transport, plus others). These data were used to measure residential self-selection, which we defined as moving to a neighbourhood that was consistent with ones’ preferences, life-stage, circumstances or socio-demographic charactersitics. Self-selection is a potential confounder of the association between environmental factors and health, and represents one of the biggest threats to claims of causal inference,42 hence measuring self-selection allowed this to be accounted for in analyses. At each of the five time-points corresponding to the survey data collections, a suite of objective environmental measures was generated using a Geographic Information System (GIS). The built environment measures (i.e. residential density, street connectivity, land-use mix, street lights, bikeways and parks) were generated at four scales: (i) 1 km Euclidean (straight-line) buffer around each participant’s home; (ii) 1 km road network buffer around each participant’s home; (iii) CCD; and (iv) suburb. GIS was also used to create road network distances from the participant’s home to the Brisbane Central Business District, the Brisbane River and coast, the closest public transport node (i.e. bus stop, train station, ferry terminal), shop, public open space, and CityCycle station (public bicycle-hire scheme). In addition, at each of the four built environment scales, we generated measures of crime, topography and traffic density, and for CCDs we derived an area-level measure of disadvantage using the ABS’ IRSD.28 Individual-level data in the main cohort study have been linked to mortality records. As part of the HABITAT sub-study we measured blood pressure and resting heart rate, height, weight and waist circumference. Physical functioning was measured based on static balance using the Short Physical Performance Battery measure; grip strength; and functional fitness using the Seniors Fitness Test (upper and lower body muscular strength, aerobic endurance, upper and lower body flexibility, agility and dynamic balance).43–46 All measurements were taken by research assistants trained to follow a standard protocol. Sub-study participants also wore an Actigraph GT3X-BT accelerometer and QStarz BT-Q1000XT Global Positioning Systems (GPS) device during waking hours for 7 days. A list of publications arising from the HABITAT study to date is available online (https://cur.org.au/project/habitat/). Baseline evidence shows that residents of socio-economically disadvantaged neighbourhoods report lower levels of total physical activity, general walking, and moderate and vigorous activity; however, they are more likely to walk for transport.47 Propensity to walk for transport declines with age; however, the declines are more precipitous for older persons, members of lower income households and residents of disadvantaged neighbourhoods.48 Higher levels of walking for transport in disadvantaged neighbourhoods are associated with living in a built environment more conducive to walking (i.e. greater street connectivity, more diverse land use mix).49 In the mid-to-older-age population, walking for transport at levels consistent with physical activity recommendations is more likely in neighbourhoods characterized by greater residential density, access to bikeways, proximity to public transport and shops, and living in a well-lit area.50 Moreover, compared with traditional suburban developments, transport walking (and other active modes such as cycling and public transport use) are more likely in ‘Transit Oriented Developments’, which are urban forms that integrate mixed land use, a relatively dense built environment, well connected street networks, and pedestrian-friendly infrastructure around a transport node (e.g. train station, bus transit centre).51 Recent longitudinal findings show that investments in changing the built environment to be more walkable are associated with increased walking for transport.52 At the individual-level, our work shows that social and group contexts influence propensity to engage in recreational physical activity;53–55 e.g. older people prefer activities with others of a similar age, but are less likely than their younger counterparts to express a preference for fixed-time and structured activity sessions; and persons from low income households are more likely to express a preference for low-cost and team-based activities.56 Among the mid-to-older age population in Brisbane in 2007, ∼20% cycled for recreation and 4% for transport.57 A diverse range of environmental (built and social) and individual-level factors (e.g. socio-demographics, perceptions) are associated with cycling, and the determinants of recreational and transport cycling are often different.58 Cross-sectional evidence shows that sedentary leisure is largely independent of physical activity level and does not preclude meeting physical activity recommendations.59 Those who report longer sitting times (especially watching TV) are likely to be male, single and living alone, experience health problems, be less educated, not in paid employment, and be overweight.60 Longitudinal findings show that overall total between 2007 and were in (e.g. home computer use, and in (e.g. women and the These were by declines in sitting to urban and mental health and is and in and mental health Using a longitudinal we to this evidence by that of in neighbourhood were associated with in mental health over 3 persons who a in a in Among mid-to-older aged Brisbane use is higher with and residents of more socio-economically are more likely to physical activity than and of neighbourhood and more time doing vigorous activity and engage in more activity than who use or for physical activity, and those who live walking to a or the coast, are more likely to express support for as measured by to for the environment (e.g. in standard of and higher and to environmental assessments of body show that consistent use of active (i.e. walking and is associated with lower and that changes in transport from to active (e.g. motor vehicle use to are associated with in to we have no evidence that moving to a new neighbourhood is associated with change in or that weight changes are associated with area-level Cross-sectional multilevel show that residents of socio-economically disadvantaged neighbourhoods are more likely to with type 2 diabetes, heart and and lower levels of physical disadvantage was associated with these outcomes after adjustment for in individual-level (i.e. education, occupation, household that of the neighbourhood environment are associated with these The main of HABITAT include longitudinal study and focus on physical activity, a key health-related behaviour relevant to physical and psychological health HABITAT includes a range of and assessed determinants of physical activity, measured at multiple scales, and most measured at five time-points between 2007 and All of GIS data have been at each of the survey thus to examine how changing environments influence change in behaviour and The population-based sample and relatively has findings that are representative and to the Brisbane population of HABITAT is an cohort study based primarily on data and hence to all the and in this As is the case with most cohort response have over attrition has to an of higher socio-economic designed as a longitudinal multilevel study of physical activity, the psychological and social on recreational activity and may not to other activity (e.g. active Further, measuring sedentary behaviour as an HABITAT was not to examine the multilevel determinants of this in on data with the can be to the The benefits of physical activity in reducing the risk of non-communicable and healthy are well however, do not engage in activity to health and activity to with HABITAT was to our understanding of the and determinants of physical activity (and sedentary as the for policy designed to population health and support healthy ageing. HABITAT is a multilevel prospective study of change in physical activity and sedentary behaviour and associated health and the relative contributions of environmental, social, psychological and sociodemographic factors to these The HABITAT study is conducted in the Local Government of Australia. Baseline data were in 2007 from 200 neighbourhoods (n = 11 035 participants, response The baseline sample was aged 40–65 years and representative of the Brisbane population this Since follow-up has questionnaires in (n = response (n = 2013 (n = and 2016 (n = by mail survey to the main and two clinical assessments of a in 2014 (n = response and 2016/17 (n = 606, 79.0%). of data include physical activity and sedentary physical health and psychological physical risk transport psychological and social on physical activity; neighbourhood environment and (e.g. residential population measures using and in with the HABITAT and the data contact the 1 and 2 of HABITAT were by Australian Health and is by an

Topics & Concepts