Table 1 Summary of planned and actual study design and conduct

 Village sampling frame

Sampling frame would be complete except for new villages opening during the study period and additional units opening in existing villages .

Some villages duplicated in list; and some villages listed were out of area.

Villages closed units for renovation, or added units/apartments

Facilities were listed that were not retirement villages.

The details were updated during the course of the Survey phase.

 Village selection

No selection of villages was planned. All villages would be approached, and we expected any decline in participation to be minor.

Slower than anticipated study progress led to recognition that not all villages could be included, A change in policy saw villages being recruited in random order after Dec 2017.

 Village recruitment

Gate-keeping role of village managers was recognised, but the extent research team were not given permission to recruit at all was not anticipated.

A high proportion of managers declined participation early in study. This will be detailed in a separate paper.

 Unit sampling frame

Within each village we anticipated that a list of all units/apartments (without names of residents) would be provided by village manager, enabling random selection of units, together with permission and security access for door-knocks and letter-drops allowing recruitment to proceed.

Some Managers declined to provide a list of units and also to permit access to doors, making door-knocks or even letter-drops impossible. Instead, these managers permitted study nurses to present an outline of the study at a residents’ meeting, and seek volunteers for participation. This will be detailed in a separate paper as above.

 Call-backs

Where at door-knock there was no response, the researchers were to call back up to 5 times, in accordance with Good Practice in Conduct and Reporting of Survey Research [17]. to avoid bias towards those who are frequently at home.

After several months of fieldwork, 10 call-backs was deemed too high a number for the yield provided, and call-backs were reduced to 3.

 Resident eligibility criteria

All residents in randomly-selected units would be invited to participate. Further, every resident under an LTC contract (i.e. receiving formal publicly-funded LTC services) would be included.

All residents in each randomly-selected unit/apartment were invited. In villages where random selection was not undertaken, all residents were invited.

Few residents were in receipt of LTC contracts, and the numbers did not justify continuing this line of enquiry.

 Resident sampling

Where more than one resident was in a unit/apartment, and both met resident inclusion criteria, then both were to be invited, to enable assessment of support from co-residence and inter-dependence.

This went to plan.

 Informed consent

All residents were eligible to participate regardless of health or occupancy status. For those without capacity to consent but with a legal proxy available, we planned to use the combination of respondent and proxy.

On the basis of the NZ legislation, the ethics committee approval for the study did not permit recruitment of residents who did not have capacity to consent. This will be detailed in a separate paper as above.

 Survey instrument(s)

Both feasibility and pilot studies used questionnaires based on our prior research in residential care facilities, based on Booth questionnaire, BRIGHT study and ARCHUS tools.

Mandatory introduction of the interRAI tool nationally was deemed more relevant, appropriate and thorough than our instruments, so the interRAI CHA was added. We selected the interRAI CHA tool (previously unused in NZ) as our main study instrument, and adapted the trialled survey tool as a supplementary survey questionnaire to capture additional items such as reasons for moving into a RV.

 Timeline

We planned to begin the Survey phase recruitment on 06June 2016 and complete it by August 2018.

We began Survey phase recruitment on 27 July 2016 and completed in late August 2018.

 Sample size and Power calculations

Refer to text.

Refer to text.

 Data availability and collection

Survey data were expected to be collected online and immediately available for use in monitoring progress, conducting follow-up and to determine eligibility for the RCT.

Once the decision was made to use an interRAI instrument (where data are captured online and available only occasionally to researchers), it was impossible to use interRAI data directly for determining eligibility for Phase 3. Selected variables were therefore entered into the survey database for this purpose.

 Timeline

From June 2016 to August 2018

From July 2016 to January 2020

 Length of follow-up

Average 1 year

Average 2.5 years

 Endpoints of interest

Primary endpoint: First acute hospitalisations,

Secondary endpoints: ED presentation, LTC admissions, mortality

Primary endpoint: First acute hospitalisations,

Secondary endpoints: number of all acute hospitalisations, LTC admissions, all-cause mortality

 Power calculations

Refer to text

Refer to text

 Eligibility criteria

Initially it was planned to recruit those from the Survey phase sample deemed most ‘vulnerable’ (to adverse outcomes) based on selection criteria employed in our ARCHUS study, including the number of medications prescribed.

Once the interRAI tool was chosen, interRAI based criteria were used to select residents, including triggering of 3 or more CAPS, (LTC institutionalisation risk) scores of 3 or more or at the discretion of the GNS. The number of medications reported in the interRAI tool included over-the-counter medications. On investigation, very few residents were being missed if number of medication criteria was dropped, so that inclusion criterion was removed. We retained from the ARCHUS study the validated criterion of the GNS’s impression of vulnerability/medical acuity.

 Pre-randomisation checks

All eligible for the trial were to be checked by a study nurse before randomisation to ensure those no longer eligible for the trial were excluded from randomisation.

Survey phase recruitment issues meant that resources to conduct pre-randomisation checks were limited to database checks and short contact with residents in the active arm only, i.e. after randomisation.

 Sample size

Refer to text

Refer to text

 Timeline

We originally planned to complete the RCT in all villages by 28 Feb 2019

We completed RCT randomisation for in mid-January 2019 and completed the final MDT on 14 March 2019.

 Interval between survey and randomisation

The interval between survey (and interRAI assessment) was anticipated as being 3 months, so the assessment would reasonably represent health status at time of randomisation.

An average interval of 9 months between survey and randomisation was achieved. Consequently, some assessments will misrepresent health status at time of randomisation.

 Length of follow-up

Planned follow up possibly at 1, 2 & 4 years post-intervention.

Planned follow up 1.5 years post intervention MoH data. The 4th years post-intervention follow-up is conditional on obtaining additional funding

 Endpoints of interest

Primary endpoint: First cute hospitalisation

Secondary endpoints; All acute hospitalisations, LTC admissions, mortality

Primary endpoint: First acute hospitalisation

Secondary endpoints: number of all acute hospitalisation, LTC admissions, mortality

Power analyses are based on the primary outcome. We did consider changing the primary outcome to ED presentations, but decided against this as, in NZ, many people who are acutely hospitalised are not admitted via ED but go via GP referral to acute admitting units.

 Power calculations

That residents were clustered into villages led to power calculations including an adjustment for a cluster effect.

Clustering was catered for by stratified random sampling, and adjustment for cluster effect was unnecessary.