Netfilter: iptables

Ten substantial amendments to the protocol and associated trial documentation were made during the trial. Internal pilot

Initially, two homes were recruited to the study early to allow internal piloting and review of trial processes, procedures, measures and tools ahead of recruitment of further care homes. Data from these homes were included in the trial. Changes to the original project protocol, implemented following this pilot, are reported in detail in the published protocol,98_{in Table 3 and Appendix 2.}

Design change

Our original sample size estimation, to detect a clinically important difference of 3 points (SD 7.5 points) in the primary end point of agitation using the CMAI questionnaire, assumed a 25% loss to follow-up 16 months after care home randomisation. If loss to follow-up was higher than anticipated (but no greater than 35%), our intended sample size of 750 residents still provided more than 85% power at a two-sided 5% significance level to detect a moderate effect size, equating to 0.4 SDs.

Through monitoring loss to follow-up within the trial, we determined by November 2015 that the rate would exceed our lower limit of 25%. Using data from care homes randomised into the trial up to 27 November 2015, we predicted that loss to follow-up at 16 months would be in the range of 32% to 48% (see Appendix 3, Figures 10 and 11). As such, continuing the trial as planned would not provide sufficient power for statistical analysis of the primary end point. An amendment to the trial design was required to ensure that the results of the trial were robust and generalisable. After considering all of the available options, we proposed recruiting additional residents at follow-up (i.e. a move to an open-cohort design) (see Appendix 3). All those consenting to take part (residents already participating in the trial and consented at baseline, as well as additional residents consenting at 16 months) provided data at 16 months.

The key impact of this design change was to increase the size of the cohort at follow-up to maintain the power of the trial and its ability to detect the effect size of 0.4 with 90% power (see Appendix 3, Table 67). Sample size calculations

With an estimated 48% loss to follow-up, we expected to lose 360 residents before the 16-month follow-up, resulting in data at all three time points from 388 residents. All of the other parameters (significance level,

two-sided test and an ICC of 0.1) remained the same. Consideration was given to recruiting only a proportion of eligible residents at each home at 16 months.

Three possible scenarios of additional recruitment were considered (an average of three additional residents per care home, recruiting 35% of residents lost to follow-up in each care home or replacing only 25% of residents lost to follow-up in each care home) and all provided sufficient power to detect the effect size of 0.4 (89%, 91% and 90% power, respectively). The TMG, the oversight committees (TSC and DMEC) and the funder agreed that imposing a recruitment ceiling at 16 months would be open to selection bias and that statistical power and the ability to generalise could be limited. Recruitment processes could also be protracted as a result of allowing time for decision-making via a Personal Consultee (i.e. should this be a refusal to take part, further resident–consultee dyads would then need to be approached, thereby considerably lengthening the recruitment process, researcher workload and thus cost).

Researchers were therefore instructed to recruit as many residents as possible to minimise bias. Numbers were monitored to ensure that at least three extra residents from each remaining care home were recruited.

TABLE 3 Summary of substantial amendments to the protocol and associated trial documentation

Amendment

number Date Summary of amendment

SA1 10 January 2014 Modification to method and content of health resource data to be collected, including from medical records and NHS Digital (previously Health and Social Care Information Centre)

SA2 22 April 2014 Modifications to care home information sheet to improve clarity and provide additional information following review by the PPI panel

SA3 26 June 2014 Modifications to care home recruitment process; resident, staff and relative eligibility criteria; screening of proxy informants; translation of trial documentation; process for completion of independent assessments; monitoring of DCM implementation; relative/friend withdrawal; resident safety monitoring; and information included on participant information sheets and consent forms (for mappers, staff proxy and residents including consultees). In addition, clarification of mutually exclusive staff roles, amendment of assessment measures to be used, establishment of a DMEC and development of a short form of the resident information sheet

SA4 10 September 2014 Personal Consultee introductory letter and reminder letter, and relative/friend proxy informant introductory letter approved

SA5 15 January 2015 GP letter to accompany guidance on antipsychotic prescribing approved SA6 15 January 2015 Change of sponsor, modification to care home eligibility criteria, modification to

resident eligibility criteria and modification to randomisation stratification criteria SA7 22 October 2015 Modification to requirements for witnessing resident consent, addition of SMS

reminders for mappers and modifications to participant information sheets and consent forms

SA8 4 February 2016 Detail added to the protocol on conduct of the process evaluation, modifications to staff measures booklet, modification to continued attempts to recruit relative/friend proxy informants post baseline and modifications to participant information sheets SA9 15 April 2016 Change to open-cohort design, with additional recruitment of resident participants

at the 16-month follow-up and associated changes to trial documentation approved; modification to staff proxy informant consent processes; modification regarding requirements to check care home indemnity insurance; introduction/ modification of documents to support process evaluation and to proposed process evaluation methods and processes; and modification to process for assessing ongoing capacity of Personal Consultees

SA10 25 July 2016 Modification to data collected during process evaluation and additional text messages to remind mappers about mutually exclusive staff roles

The benefits of the design change were:

l An ability to detect intervention effects at the care home level (as the intervention is aimed at the whole care home).

l Conclusions could be generalised to a broader population of residents (i.e. not just to those still residing in the care home 16 months following randomisation).

l We would be able to analyse the data for a cross-sectional (i.e. open-cohort) and closed-cohort (longitudinal) design.

l We minimised selection bias by providing an objective criterion for inclusion (all eligible consenting