Intervention
Theory basis
The present intervention draws on the habit-formation model [30, 32]. Habitual behaviours are controlled by impulses that are generated automatically when encountering situations in which the behaviour has been performed in the past [29]. Habitual behaviours are thought to self-perpetuate, and can prompt behaviour even where conscious motivation to do the behaviour is low [35]. Habit-formation has thus been proposed as a potential mechanism for sustaining new behaviours over time, by shielding them against common long-term losses in motivation that threaten to reverse short-term behaviour gains [36]. Habit forms when a given behaviour is repeated in a particular context, because this strengthens the mental association between context and behaviour [23]. Habit-formation advice is based on repeating a chosen behaviour in a setting until it becomes automatic [30]. It differs from non-habit-based advice in specifying not only which behaviour should be adopted, but also how it might be performed so as to aid maintenance [23]. Forming a habit requires the motivation and volitional skills and resources to sustain behavioural repetition until the behaviour becomes automatic [32]. Repetition is best facilitated by pursuing behaviours that are manageable and realistic [37], and by self-monitoring behaviour [32]. Habit is thought to form more quickly for simple actions [31]. Habit-formation interventions may therefore be most effective where paired with a ‘small changes’ approach to behaviour change, based on making minor modifications to existing practices rather than major changes [30].
Lally and Gardner’s [32] habit-formation framework was used to guide the selection of behaviour change techniques for the intervention. Specifically, techniques were chosen to: enhance motivation to reduce SB and increase PA; facilitate the translation of motivation into action; and promote and sustain repeated performance of PA, or disruption of SB, in consistent contexts. The intervention was co-designed by a panel of 15 experts (covering sports and exercise science, ageing, geriatrics, general practice, psychology, physiology, and physiotherapy), and two independent panels (Ns = 17 and 23) of self-reportedly inactive (< 30 mins leisure time MVPA per week) and sedentary (> 6 leisure time hours spent sitting per day) retired adults aged 60–75 years. Further intervention development detail has been provided elsewhere [25].
Intervention content
The present intervention comprised an information booklet, outlining the health risks of SB and benefits of PA, and offering tips and rationale for undertaking PA in a way that would reduce SB or build PA habits, as supplemented by a set of tick-sheets to record adherence to each tip (for both intervention and data collection purposes). The tips were designed to promote all four recommended forms of PA in older adulthood (aerobic, balance, flexibility, muscle-strengthening) and reduce SB. Where possible, tips specified an everyday contextual cue (e.g. ‘during breaks between TV programmes…’) and recommended a behaviour for enactment in the presence of the cue (‘…stand up and walk around’), with justification relating to health or wellbeing (‘this will stop your joints from seizing up’). This format was used to promote motivation to perform the action and the context-dependent repetition necessary for habit to form [32]. ‘Handy hints’ were offered with most tips to provide instructions, offer less or more intensive variants of the recommended activity, or suggest preparatory or supplementary actions likely to increase likelihood of enactment (e.g. ‘leave the remote control by the TV so that you have to get up to change channel’). Text at the end of the booklet described behaviour change techniques conducive to habit formation (e.g. ‘plan ahead’ [action planning], ‘track your progress’ [self-monitoring], ‘start low, go slow’ [graded tasks]) [32]. The booklet recommended adhering to as many tips as possible, while not attempting anything that felt uncomfortable or that a physician had advised against. An extensive description of intervention content, coded according to component behaviour change techniques [38], is provided in Additional file 1: Table S1.
The intervention was presented in an information-only leaflet-based format because, while such interventions are often assumed to be ineffective [39], a leaflet providing habit-formation advice has previously shown efficacy for changing behaviour [26, 27]. This suggests that information content, not delivery method, may determine effectiveness. Previous studies have shown written advice on context-consistent repetition of simple actions to be novel, motivating and acceptable to participants [27, 40].
Intervention cost
Excluding an initial charge of £1050 (~US $1600) for visual design and typesetting, each intervention booklet as supplemented by 8 ticksheets cost £4.50 (~US $7), which covered printing costs only. No other intervention costs were incurred.
Study design
An uncontrolled (pre-post) intervention design was used, with two independent samples, with three data collection time points over an 8-week study period (baseline, 4-weeks, 8-weeks). All study procedures were undertaken with Sample 1 by a registered post-doctoral health psychologist, and with Sample 2 by a fully-trained MSc Health Psychology student. Ethical approval was provided by the University College London Research Ethics Committee.
Procedure
Eligibility criteria
In both samples, participants were eligible only where they reported being aged 60–75 yearsFootnote 1, able to speak and understand English, and with no physical impairments precluding engagement in light PA. No explicit PA or SB criteria were imposed on Sample 1. In Sample 2, participants were only eligible where they self-reported ≥ 6 h of mean daily sitting time, and < 150 total weekly minutes of MVPA, over a typical week. These criteria were used to ensure that all participants in Sample 2 were sedentary and insufficiently active according to national guidelines, characteristics assumed of Sample 1 based on previous literature [41]. Participants who self-reported any mental health problem were excluded, because the researchers were not qualified to assess capacity to provide informed consent. Eligibility was based on participant self-report only; no further screening (e.g. physical health) took place.
Recruitment
Sample 1 was recruited from sheltered housing sites, which are self-contained flats within a larger building, with communal areas for socialization, and warden assistance, for adults aged 55 or over who are less able to live independently. This group was purposefully selected because sheltered housing residents tend to have higher rates of physical inactivity than those living independently [41], and so were targeted to reflect the least active subgroup of older adults, who would derive most benefit from intervention. Sample 2 was recruited from community settings, purposefully selected on the basis of SB, and physical inactivity according to national guidelines.
Two recruitment methods were used. Sample 1 was recruited at sheltered housing sites in London, between November 2013 and January 2014. A housing trust, responsible for multiple London sheltered housing sites, permitted access to managers of local sites. Managers were informed of the study purpose and told that we were seeking older adults who sit for long periods and do little PA. Managers at five sites agreed to allow access to residents. Recruitment procedures at each site varied according to preferences of the site manager. At one site, the site manager gave a talk about the study to a group of potentially eligible participants, and those who were interested consented to participate at a subsequent face-to-face visit from the researcher. In the remaining four sites, managers suggested potential participants, based on the manager’s perceptions of their low PA and high SB, to the researcher, who approached them individually, explained the study and consented interested individuals. Sample 1 refusal rates were calculated by recording how many of those informed about the study (including potentially ineligible adults) did not wish to participate.
Sample 2 was recruited between March and June 2014, through written advertising materials and talks advertising the study at community and faith centres, and via a notice in an Age UK South London newsletter. Refusal rates could not therefore be calculated. Interested potential participants were pre-screened, by phone, using items from the International Physical Activity Questionnaire (IPAQ [42]), and a validated SB questionnaire [43] to establish their typical weekly MVPA and SB.
Data collection
Data were collected from Sample 1 between December 2013 and March 2014, and from Sample 2 between March and August 2014. Data collection took place at the participant’s home (Samples 1 and 2) or another location convenient to them (Sample 2). At the baseline session (Time 1; T1), participants completed a questionnaire of (quantitative) study measures, and then received the ‘On Your Feet to Earn Your Seat’ intervention booklet, together with nine tick-sheets to record adherence to the intervention tipsFootnote 2. The researcher explained to each participant the content and purpose of the booklet and tick-sheet, and how to complete the tick-sheet. Specifically, participants were told of the potentially independent health risks of SB and PA, and advised to follow the recommendations provided in the booklet on how to integrate more PA into everyday routines, while reducing SB. They were told that completing the tick-sheet could help them to monitor their progress in changing their PA and SB (and would enable the research team to monitor adherence to tips). No further advice or counselling was provided in either sample. Four and eight weeks post-baseline (T2 and T3 respectively), participants were visited again and asked to complete further quantitative measures. At T3, a face-to-face semi-structured interview was also conducted to capture participants’ views towards the intervention, and all tick-sheets were collected. T2 and T3 sessions were undertaken for measurement purposes only; no further active intervention was delivered, though participants were able to request clarifications of information provided to them at T1. Records were not kept of whether or what information was requested in this way.
Quantitative measures were self-administered where possible, or by the researcher at the participant’s request. All participants received a £10 (~US$15) shopping voucher at each of the three data collection points.
Measures
Quantitative data
All quantitative data were self-reported. Demographics (gender, age, ethnicity, marital status, education) and health status variables were recorded at T1 only, for sample description purposes. Ethnicity was reported using UK census categories [44]. Marital status was reported using a single item (‘are you: single/married/widowed/divorced or separated?’). Education was recorded using two items, capturing the age at which participants left school, and whether they had completed a university degree (yes/no). Health status was assessed by a single item about long-term illness (‘have you any long-standing illness, disability or infirmity?’ [yes, please state/no]) [45].
Sedentary behaviour was assessed at T1-T3 using two measurements. One was an item derived from the IPAQ, assessing the total time spent sitting ‘while at home, when outdoors, or during leisure time’ (including ‘time spent sitting at a desk, visiting friends, reading, travelling on a bus, or sitting or lying down to watch television’) over the preceding seven days. IPAQ sitting items have been shown to have test-retest reliability, and to correlate with objectively measured inactivity [42, 46]. The second was the Measure of Older Adults’ Sedentary Time (MOST) [43], a multi-item questionnaire recording total time spent in seven common sedentary activities over the prior seven days (watching television, using the computer, reading, socializing, transportation, hobbies, ‘other activities’). The MOST has been validated against accelerometer step count readings, and shown to have test-retest reliability and be sensitive to changes in SB [43]. Data were summed across the seven activities, such that values denote total sitting time. MOST data were treated as missing where none of the seven items was completed, and eligible for analysis where at least one item was completed.
Physical activity was measured at T1-T3, using three single-items derived from the IPAQ relating to time spent walking, or in moderate or vigorous PA respectively, over the previous seven days (‘How much time in total did you spend [walking/doing vigorous physical activities/doing moderate physical activities] in the last 7 days?’). Vigorous activities were defined as those ‘that take hard physical effort and make you breathe much harder than normal’, and moderate as those ‘that take moderate physical effort and make you breathe somewhat harder than normal’. For both items, participants were asked to consider only ‘those physical activities that you did for at least 10 min at a time’. IPAQ PA items have been shown to have test-retest reliability, and to correspond with multiple objectively measured PA indicators [42, 46]. Responses to all IPAQ SB and PA items were provided in hours and minutes, and converted to total minutes for the purpose of analysis.
SB and PA habit were each measured using the Self-Report Behavioural Automaticity Index (SRBAI) [47], a four-item subscale of the Self-Report Habit Index (SRHI) [48]. The SRBAI focuses on the automaticity that characterises habitual responses [29, 49]. Both the SRHI and SRBAI show sensitivity to theorised effects of habit on action, and convergence with implicit association-based habit measures [47]. A systematic re-analysis of previous SRHI applications showed the SRBAI to have consistently strong internal reliability, and convergent validity with its parent index [47]. SRHI/SRBAI applications with greater contextual specificity are likely to minimize respondent interpretation error [50], and so item wordings specified a behaviour (‘sitting…’) and a context (‘…during my leisure time’). For SB habit, items followed the stem ‘Sitting during my leisure time is something…’, and PA habit items followed the stem ‘Physical activity during my leisure time is something…’ (‘…I do automatically’, ‘…I do without thinking’, ‘…I do without having to consciously remember’, ‘…I start doing before I realize I’m doing it’; 1 = strongly disagree, 7 = strongly agree). Mean scores were generated for each index, with higher scores indicating stronger habit. Reliability was good at all time points (SB habit, α range Sample 1: .77-.95, Sample 2: .88-.96; PA habit, α range Sample 1: .86-.97, Sample 2: .86-.95).
Adherence to tips was assessed via 7-day tick-sheets. Participants were asked to record a tick on each day on which they completed each recommended activity, for the study duration. For one tip, which recommended setting a manageable walking target (see Additional file 1: Table S1), participants were asked to record their daily target and whether it had been achieved.
Qualitative data
Semi-structured interviews covered five topics: experiences of using the leaflet, barriers to adherence, habit-formation, whether further support was required, and suggestions for improvement. Participants’ responses shaped progression through topics. Audio recordings of interviews were transcribed verbatim.
Analysis
Quantitative data
Refusal rates (Sample 1 only) and attrition rates were summarized using descriptive statistics. Rates of adherence to each tip were calculated using the seven tick-sheets for which full (7-day) data were available (i.e., Weeks 2–8; see Endnote 2). Weekly adherence to each tip was calculated by summing, for each tip, the number of ticks recorded in that week and dividing the total by seven (i.e. 7 days). Mean total adherence to each tip was calculated by summing all ticks for each tip across all seven tick-sheets and dividing by 49 (i.e. 7 days × 7 weeks). Global mean total adherence across all tips was calculated by summing the mean total adherence to each of the 16 tips and dividing by 16. A supplementary analysis was undertaken of weekly adherence across all tips, as calculated by summing all ticks for all tips in each week and dividing by 112 (i.e. 16 tips × 7 days). All rates were multiplied by 100, to allow expression as percentages.
The purpose of reporting SB and PA behaviour and habit strength across the three timepoints was to investigate behavioural responding as an indicator of acceptability, with trends towards decreased SB and/or increased PA, regardless of statistical significance, seen to reflect intervention acceptability. Nonetheless, inferential statistical tests were run, and p values reported, for completeness. Behaviour and habit changes were tested using repeated-measures ANOVA. Where normality and equality of variance assumptions were not met, Friedman’s two-way ANOVA for non-parametric data were used. A supplementary analysis was run to document the number of participants showing changes from baseline, at either follow-up point, on SB and PA behaviour and habit indices. Quantitative data were assessed for each sample in isolation, and, with the exception of attrition analyses, run only for those who completed all three study timepoints.
Quantitative analyses were run to describe trends observed in available data, rather than to investigate intervention effects. Thus, missing data were handled using pairwise deletion for descriptive purposes, and listwise deletion for inferential statistical tests.
Qualitative data
Interview transcripts were analysed by two coders, using thematic analysis [51]. Themes were inductively developed and iteratively refined by one coder, and verified through discussions with a second coder. Disagreements were resolved through discussion between coders. Only excerpts relating to intervention content were eligible for analysis; responses relating to visual presentation were used to refine materials, but are not presented here. Although thematic analysis is not well-suited to formal comparisons between groups, coders did not observe differences between the two samples in the content or fit of each theme, and so analyses are reported for both samples combined.