Cluster randomised clinical trial results: sites and patients

Participants: sites

Site selection

Expressions of interest to participate in the POPPI cluster RCT were received from 118 NHS adult, general critical care units across England, Wales and Northern Ireland. Potential sites were asked to complete a site feasibility questionnaire, of which 86 completed questionnaires and were subsequently assessed for eligibility by the ICNARC CTU.

Relative to the target of 24 participating sites – of 86 potential sites completing a questionnaire – 32 sites did not meet inclusion criteria (e.g. had insufficient resources/capacity to deliver the trial or would see too few patients over the trial period). The final 24 sites were selected from the remaining 54 sites, with reasons for selection including good research track-record and display of a high level of enthusiasm. A small number of sites were already delivering early psychological support and so were not selected. In addition to the 24 sites, we selected eight reserve sites to be approached in the event that a selected site was no longer able to participate.

One site that was initially selected lost eligibility to take part as a result of inability to implement use of the CAM-ICU68 to assess delirium prior to the start of the trial (see Chapter 4 for the inclusion criteria for sites). A site from the reserve list was rapidly set up as a replacement.

Site set-up

A total of 24 sites (22 in England, one in Wales and one in Northern Ireland) obtained local NHS permissions/approvals and opened to recruitment in three groups of eight sites, at 2-month intervals. Sites allocated to the first group opened to recruitment on 1 September 2015. Groups two and three opened on 2 November 2015 and 4 January 2016, respectively. One site in group three opened 3 days late, owing to delays in obtaining local NHS permissions through the Christmas/New Year period. Site initiation visits were carried out at all participating sites prior to start of patient screening and recruitment. Sites were randomised in their second month of recruitment (during the baseline period). The flow of sites (clusters) through the cluster RCT is presented in Figure 12, according to the CONSORT extension for cluster trials.80

FIGURE 12 The CONSORT flow of sites (clusters) and patients. ICU, intensive care unit. Reproduced with permission from JAMA 2019;321(7):665–75.71 Copyright © 2019 American Medical Association. All rights reserved. (Figure not reproduced here; see the original article.)

Overall, as planned, 24 sites participated in the POPPI cluster RCT. One site from the second group of sites closed in December 2016 prior to the end of the recruitment period because key staff were scheduled onto other projects when the recruitment period was extended. The remaining 23 sites remained open to screening and recruitment until the end of recruitment in January 2017.

Intervention site set-up

Following site randomisation, each intervention group site identified three self-selected POPPI nurses (plus one reserve), according to the person specification. Each POPPI nurse signed a form confirming their commitment to undertake the POPPI nurse role. Intervention site initiation visits were carried at all intervention group sites ahead of the 3-day training course and prior to commencement of the transition period.

In total, 38 POPPI nurses were trained across the series of three central 3-day training courses held in February, April and June 2016. For two sites, the reserve POPPI nurses was also trained – one because an original POPPI nurse went on a career break and the other to help provide increased cover for delivery of the stress support sessions at one of the larger participating intervention group sites.

Characteristics of participating sites

The characteristics of the 24 adult, general critical care units that participated in the POPPI cluster RCT compared with all other adult, general critical care units in the CMP (n = 191) are presented in Table 6. Overall, a representative mix of university and non-university/university-affiliated hospitals, geographically spread across England, Wales and Northern Ireland, took part in the cluster RCT. A higher proportion of larger units participated in the POPPI cluster RCT compared with all other units participating in the CMP. Use of diaries in the critical care units was reported at the start of the study at nine intervention and seven control sites.

Participants: patients

Screening

Table 7 summarises data from screening and enrolment logs maintained by participating sites, stratified by treatment group (usual care or intervention) and time period (baseline, transition and intervention).

Of the patients who were deemed to be potentially eligible for the trial, the numbers of patients who were reported as being missed appeared to increase over time, in both usual care and intervention sites. The most commonly reported reason was that there was a relatively short time frame from a patient regaining capacity and being discharged from the critical care unit (the initial approach for consent was to be made in the unit). In addition, patients were often also missed because of a lack of resources to enable screening and recruitment in the evenings and at weekends. A similar trend was observed with regards to refusals of consent, with increasing refusals over time in both treatment groups, but higher rates of refusals being observed in patients at intervention group sites during the intervention period. The reported reasons for declining consent varied (but were consistent across treatment groups and time periods); often, patients felt too tired/fatigued to take part, others felt that they were not ‘up to’ completing questionnaires and that taking part in a research study would be burdensome to them. Some patients were already taking part in other studies and therefore did not want to consider a further study (even when co-enrolment was permitted between the studies).

Recruitment

A total of 1458 participants were recruited between 1 September 2015 and 3 February 2017. Figure 13 shows the cumulative patient recruitment against the anticipated pre-trial sample size and the revised minimum target. There was variation across the 24 participating sites in the rate of recruitment; the overall mean monthly recruitment rate was 4.1 (range 1.8–6.6) patients per site per month.

FIGURE 13 Patient recruitment. (Figure not reproduced here; see the original article.)

Patients were generally recruited into the POPPI cluster RCT during weekdays (Monday to Friday). Less than 10% of patients were recruited on either a Saturday or a Sunday (Figure 14). Most participating sites reported having insufficient resources to enable screening and recruitment outside usual office hours. Five participants (0.3%) withdrew consent, resulting in 1453 participants being included in the final sample. The 100 participants recruited during the transition (training) period were excluded from the primary analyses, as planned.

FIGURE 14 Percentage of patients recruited by day of the week. (Figure not reproduced here; see the original article.)

Patient characteristics

The groups were well matched across treatment groups and time periods (Tables 8–10). The mean age of patients was similar (baseline period: 57.2 years in the usual care group and 59.5 years in the intervention group; intervention period: 57.2 years in the usual care group and 60.4 years in the intervention group). The sample was predominantly white in ethnicity and ≈60% of recruited participants were male (this proportion was slightly lower at intervention group sites during the intervention period: 55% male).

The majority of participants were medical admissions to the critical care unit, with a slightly higher proportion of planned surgical admissions at usual care group sites (than in intervention group sites). Patients’ severity of illness, assessed using the APACHE II scoring system and based on physiology variables during the first 24 hours of unit admission, was similar across treatment groups and time periods (median score of 16, and slightly higher at intervention group sites during the intervention period, with a median score of 17).

During the baseline period, patients were consented, on average, 9.6 (usual care group) and 10.5 (intervention group) days after admission to the critical care unit; compared with 11.9 and 13.2 days, respectively, during the intervention period. The proportion of patients consented in the unit, as opposed to on the wards after critical care discharge, was lower at intervention group sites during the intervention period than during all other trial time periods. At the time of consent, median STAI-6 and quality-of-life health thermometer scores were the same across treatment groups and time periods.

To assess case-mix representativeness, patients in the cluster RCT were compared with a best approximation of the POPPI eligibility criteria applied to the CMP database for both POPPI cluster RCT units and all other adult, general, critical care units participating in the CMP (non-POPPI units; Table 11). Patients who were recruited to the cluster RCT were representative of similar patients in the CMP, but were more likely to be white and not from the most deprived areas (according to quintile of IMD 2015).76

Patient follow-up

At 6 months, all participants were successfully followed up for mortality status. Of those surviving to 6 months, 79.3% completed the follow-up questionnaire (range 78.4–79.9% across treatment groups and time periods; Table 12). There were no differences in response rates between groups or time periods. The majority of questionnaires were completed on paper and returned via post.

Of returned completed questionnaires, data completeness for the primary outcome measure (PSS-SR score) was high (> 99%). Data completeness for the HADS67 and EQ-5D-5L56 were also high (> 96%), but less complete responses were received for the health services questionnaire (note that this questionnaire appeared last in the questionnaire booklet).

For participants alive at 6 months, response rates varied by patient characteristics (Table 13). Response rates were lowest in the youngest age group, and were generally lower for females than for males, for non-white ethnicities than for those of white ethnicity, and for patients with higher levels of deprivation. However, there appeared to be no clear relationship between response rate and either baseline acute severity of illness or anxiety and HRQoL at the time of consent. For patients receiving the intervention, the response rate increased with the number of stress support sessions received.

Multiple imputation

Table 14 reports all of the variables considered for multiple imputation, and for each variable, the number of missing values, and the imputation model chosen.