1. Recruitment of health centres (clusters)
Original health centre recruitment plan
This pragmatic trial was designed as a public health programme effectiveness study  and the interventions were therefore made available through the established health system under standard operating procedures, without the use of resource-intensive extraordinary efforts to secure uptake.
There are 94 health centres in the Santiago Metropolitan area. The study is aimed at low to medium socio-economic status older people, and the nine health centres in high socio-economic status municipalities were therefore excluded. The average number of persons aged 65.0–67.9 per health centre was derived from 1992 census data for each municipality, and small municipalities (those with less than 400 persons of this age group per health centre) were excluded (n = 49) in order to facilitate data collection.
The geographic catchment areas of the remaining 36 health centres in low to medium socio-economic status areas with sufficient older persons were mapped, and following further discussions and site inspections, 15 health centres were excluded because of insufficient interest among health centre staff (self-perception of having heavy burden of work), or infrastructure limitations, for example lack of space to store the nutritional supplement, or no private room to conduct the interviews. Of the 21 remaining health centres, 20 were randomly assigned, using stratified random sampling, to one of the four treatment arms. The remaining health centre was kept in reserve in the event of the withdrawal of a health centre.
Revised health centre recruitment plan
Following the change to the sample size requirement outlined below [see Revised sample size calculations for the nutrition intervention (PACAM)], which required a further 8 clusters to be included, the reserve health centre and the 15 health centres originally excluded were reconsidered for participation in the study. An additional eight health centres agreed to take part and could support the limited data collection required (see Data collection among additional eight clusters). The eight health centres were randomly assigned, using stratified random sampling, to either the nutrition intervention arm or the control arm of the trial.
2. Recruitment of participants
Following the random assignment of the selected health centres to one of the four trial arms, project field staff initiated the recruitment of potential participants living in the health centre catchment areas. Households within the selected catchment areas, randomly sampled from detailed maps of city blocks, were visited by project field staff, and all potentially eligible older people within selected households were identified. Potential participants were aged 65.0–67.9 years at baseline: individuals in this age-group are currently not eligible to receive PACAM.
A team of 20 project field staff initiated the process of randomly sampling houses in the selected clusters on 4th May 2005, and after two months the team had sampled more than 6000 houses and managed to identify 601 potentially eligible individuals. Of these, 351 individuals stated that they were willing to attend a further appointment, the majority of the remainder stated that they had no time (65%) or no interest (35%). Of the 351 individuals expressing an interest, 242 were recruited to the study. The concomitant availability of 2002 census data made it clear that adults aged 65–67.9 were spread thinly and were also unevenly geographically distributed in the selected clusters. It was therefore decided that a new sampling strategy within clusters was required. The revised recruitment strategy was based on the gathered evidence which indicated that 95% of those who had been identified as potentially eligible via the random sampling procedure, were also on health centre registries. It was therefore agreed that the names of older people on registries held in health centres would be used as the basis for recruitment.
Potential participants identified by either of these strategies received, on contact, an information sheet outlining the nature and importance of the study, and were asked for contact details. The information sheet explicitly stated which intervention the potential participant would receive on providing consent to join the study. This method of randomising clusters prior to identifying participants has been shown to result in potential recruitment bias , and extra efforts, such as demonstrations by health workers and public meetings at the health centres were made to ensure that acceptance rates in the four study arms were comparable.
Potential participants were then contacted by telephone, or in person, by a member of the project field staff a minimum of two days after receiving the information sheets. If the potential participant expressed an interest in the study, they were asked if they were ambulatory (being able to walk independently and thus complete the physical assessment required for the exercise intervention), if they had sought medical advice due to unexplained weight loss of greater than 3 kg in the past three months (which may suggest an underlying pathology), or if they intended to move house (to a different health centre catchment area) within the next three months. Individuals who report that they were non-ambulatory, and/or had sought medical advice due to unexplained weight loss, and/or intended to move within the next three months were not invited to take part in the study and were thanked for their time and co-operation.
Remaining potential participants were invited to attend an appointment at a local community centre. Individuals who were unable to attend the appointment at the community centre but remained interested in the study were visited at their homes by a member of the project field staff. Potential participants were assured (as stated on the information sheet) that if they did not wish to take part in the study, it would not prejudice the quality of health care provision from their health centre. In addition, potential participants on health centre registries were invited to informal group meetings to receive information about the study.
At the baseline appointment (approximately one week later), potential participants were provided with detailed information sheets and fully informed by the project field staff about the nature and relevance of the trial, and exactly what will be involved if they agreed to take part. Potential participants were then asked to give consent to undergo a brief cognitive screen, a short version of the Mini Mental State Examination (MMSE). The MMSE is an easy to administer test that has been widely validated as a screen for dementia , and was used to exclude participants with low cognitive status. Participants with an MMSE short-version score of less than 13 (out of a maximum of 19), suggesting possible cognitive impairment, were further assessed using the 11-item Pfeffer screen . Potential participants scoring six or greater in the Pfeffer screen were not included in the study, as a short-form MMSE score of less than 13 combined with a Pfeffer screen score of six or greater has been validated in Chile as a marker for possible dementia in adults . Individuals with an MMSE short version score below 13 and a Pfeffer screen score of 6 or greater were thanked for their time and co-operation.
Potential participants who were not excluded for possible dementia were invited to consent to take part in the study. Individuals unwilling to participate further were thanked for their time and co-operation. Individuals giving informed consent were enrolled in the trial.
All consenting participants were then first assessed by completing questionnaires in their local community centre or their homes. They were then invited to attend a local community centre where they underwent anthropometric assessment, blood pressure measurement and two brief mobility tests (see Data Collection). The baseline assessment took approximately 90 minutes to complete and the anthropometric assessment, blood pressure measurement and mobility tests took approximately 30 minutes to complete for participants from the original 20 clusters. It took approximately 45 minutes to collect the limited data required for the pneumonia outcome in the additional eight clusters. At this time, participants were informed about health care provision at their local health centre and the potential benefits of registration with the National Health Service. Participants were given contact details for the study manager, and thanked for their time and co-operation.