The randomised trial
The OPERA trial was a cluster randomised trial including 78 care homes in England; the unit of randomisation was the individual care home. Protocols with detailed descriptions of the methods and the intervention are available elsewhere [7, 10]. The trial took place from 2008 to 2011 and we recruited care homes from the Midlands (Coventry and Warwickshire) and North East London, UK. The physical activity intervention involved both activities aimed at changing the culture of the home and, most importantly, twice-weekly, moderate intensity, progressive group exercise sessions led by a physiotherapist. Activities were aimed at changing the culture of the homes so that residents would be supported and encouraged to be more active. These included individual physiotherapy assessments and exercise prescriptions for all residents, advice for staff on ways to safely increase the mobility of the residents, the provision of simple aids to maximise individuals’ mobility, and formal care home staff training on recognising depression and the potential importance of promoting physical activity in residents. We planned to introduce a home ‘champion’ into each intervention home who would be a member of staff who would encourage and reinforce safe mobility and ‘activity’ when the physiotherapist was not there. Control homes only received formal staff training on depression awareness without any specific mention of promoting physical activity.
The primary outcomes of the trial were the prevalence of depression in participants able to complete assessments 12 months after randomisation, and the change in the number of depressive symptoms in all participating residents at 12 months after randomisation plus the change in number of depressive symptoms 6 months after randomisation in those participating who were depressed at baseline (based on the 15-item Geriatric Depression Scale (GDS-15) ). Secondary outcomes included remission of depression, cognitive function (Mini Mental State Examination (MMSE) ), health-related quality of life , mobility (Short Physical Performance Battery (SPPB) [14, 15]) and exercise tolerance, pain, fear of falling, social engagement, health utility, falls and mortality rates.
We have reported full details of our methods elsewhere . Briefly, we used a mixed methods approach combining quantitative data from all the study homes and quantitative and qualitative data from a purposive sample of eight case study homes (two control and six intervention) that were studied in depth. We used a sampling frame representing the different type of homes within the study (for example, size, location and ownership) to select the eight case study homes .
Across all the homes in the study we collected quantitative data including size, occupancy, and home facilities from the Care Quality Commission website (http://www.cqc.org.uk/cqcdata) and directly from care home managers prior to randomisation via our field researchers. We also collected trial process data from the control and intervention arms, on the staff training and, in the intervention homes only, on the delivery and receipt of the intervention. We evaluated the staff training by means of questionnaires handed out to staff at the end of each training session. They were asked to rate five statements on a five-point scale, anchored ‘strongly disagree’ (1) to ‘strongly agree’ (5). The statements were: ‘the session was relevant to my job’, ‘I learned something new from the session’, ‘I am glad I attended this session’ and ‘the session was about the right length’. In addition, there was space for comments or suggestions. We also attempted to obtain a long-term evaluation of the staff training by means of a second questionnaire mailed out to staff who had attended the training approximately 3 months after the training. This questionnaire asked staff to reflect on the training delivered and the materials that went with it and to state if they found them useful or not. Respondents were also asked to comment on the components they found most and least useful and whether they had become more aware of depression among residents since the training. Finally, they were asked if they required more information about depression, how they would like this delivered, and for any comments and suggestions. There was a slight difference in the questionnaires sent to staff in intervention homes, as this had an additional section asking about the mobility advice that was given in these homes during training.
We assessed the quality and fidelity of the intervention delivery during a site visit by the study’s lead physiotherapist to each physiotherapist involved in the delivery of the intervention at one of their assigned care homes on at least two occasions: at 6 weeks and at 6 months after the intervention started in that home. At these visits the delivery of the exercise group and use of the whole home approach were assessed by observation, and a sample of the intervention data collection and clinical record forms and registers were checked (unless any major issues were identified, in which case all were checked). An observation tool developed for this purpose was used to ensure a consistent approach. Each item was scored on a strict three-level criterion (not achieved, partially achieved, or satisfactorily achieved). Additionally, a senior researcher not involved in the delivery of the intervention (DRE) observed 21 exercise classes in the case study care homes. Sampling of classes for observation included seven in the early stages after their introduction (between weeks 3 and 5), six at about the midpoint (between weeks 24 and 27) and eight at the end of the 12 months (between weeks 45 and 48). Observation included how the session fitted into the day, how the residents were reacting, how the physiotherapist was interacting with the residents, care staff/home involvement and what happened when it ended.
In the eight case study homes the experienced qualitative researcher (DRE) conducted face to face, semistructured interviews with managers, and a sample of staff, residents and their relatives; at baseline (within the first 2 weeks) and in towards the end of the trial (between weeks 46 and 52). A small number of managers were interviewed at approximately the midpoint (between weeks 24 and 27). DRE also conducted a detailed ethnographic study in these homes observing routine daily life. We adopted a phenomenological approach for all qualitative work [16, 17]. Focus groups were held with the intervention physiotherapists and the recruiting team at the end of the study to ensure their input.
Finally, we collected quantitative, descriptive data on activity within the case study homes at three timepoints (baseline, around 6 months and around 12 months) using the Behaviour Category Codes (BCC) instrument [18, 19]. This involved completing a checklist of what residents were doing, including interactions with staff and others, at regular intervals in the day. Observational data sweeps occurred every 15 minutes, for a 90-minute period (that is, six sweeps in 90 minutes). Each sweep recorded the total number of residents within each public area of the home and the number of residents engaged in particular behaviours at that timepoint. DRE carried out observations starting at different times but covering a whole day over a number of visits. Behaviours were collapsed into seven key ‘activity’ behaviours for analysis (active social interaction, eating/drinking, recreational activity (not exercise), exercise, passive social interaction, socially inactive, walking/wandering). The trial was completed prior to activity data analysis to avoid influencing the data collection.
Ethical review for the trial and its process evaluation was provided by the Joint University College London/University College London Hospital Committees on the Ethics of Human Research (Committee A), now known as Central London REC 4. The REC reference for the study is 07/Q0505/56. All participants provided written informed consent to participate.
Process data analysis
We digitally recorded interviews, subject to permission of each participant, and where appropriate, transcribed verbatim after anonymisation. Transcripts were managed using NVivo 7 QSR International Pty Ltd. Version 7, 2006. Researcher bias was minimised through regular crosschecking of data and findings by the members of research team. To ensure reliability, another member of the team (CS) coded a sample of 10% of transcripts. These two perspectives and subsequent discussions aided in the development of the coding scheme. The analysis was thematic and we adopted the framework method described by Ritchie and Spencer  and Pope et al. . We use quotations as exemplars of key themes. In the quotations, an assigned code identifies the respondent, their role, and timepoint within the study of the interview (BL = baseline, FU = follow-up).
Quantitative data were analysed using the statistical package SPSS (Version 18; SPSS, Chicago, IL, USA). Analysis was largely exploratory, not hypothesis driven and necessarily unadjusted for baseline covariates. Data from the activity sweeps in the case study homes are summarised in tables (see below) with activities collapsed into groups. We based the mean percentages quoted on 12 to 14 observation sweeps of 90 minutes each covering the period from 10.30 am in the morning to 17.30 pm in each of the 8 case study homes at baseline and end of study. Change is follow-up minus baseline.
As the results were examined, a number of post hoc subgroup analyses were carried to looking at the effect key baseline variables (age, cognition, physical frailty and depression) on the number of groups a participant would attend. We present descriptive statistics for these including both means (SD) and medians (ranges). As these data were not normally distributed, the tests carried out were non-parametric.