In this chapter, we report on the following feasibility outcomes:
- recruitment of participants with ID and volunteers
- the number of participants with ID who were successfully matched
- the number of participants who completed the follow-up assessment at 6 months and withdrew from the trial
- exploratory analysis of the health and social outcomes
- exploratory data on the analysis of cost-effectiveness.
Overview of issues encountered in recruitment
The study began in July 2018 and was funded for a 2-year period. In the original study Gantt chart, we had aimed to obtain all the necessary approvals (i.e. approval of the protocol by Priment Clinical Trials Unit, sponsorship and ethics approval, exchange of site contracts and approval of the study database) and start recruitment within 4 months of the trial start date. This target proved to be unrealistic and, as a result of delays in obtaining the necessary approvals, recruitment did not begin until April 2019, 6 months later than intended. Recruitment was slow, and the number of participants with ID who were eligible or willing to take part was smaller than anticipated, as was the number of volunteers expressing an interest in taking part in the study. One issue that we encountered with recruitment through the Suffolk scheme was that it was not considered to be appropriate (or acceptable) by the scheme to change the current standard of practice, because of concerns over randomisation and the potential allocation of participants to the control group. This issue was raised after the protocol was approved and sponsorship approval had been given. It was agreed that the scheme could offer individuals the option of entering the study or remaining on the waiting list for a befriender. Owing to the uncertainty of whether service users would be allocated a befriender or be in the control group, those who were approached chose to remain on the waiting list until a volunteer was allocated through the usual route, even if it meant waiting for a longer period. Staff at the service also expressed their discomfort at the possibility of participants being left in the control group without a befriender for a year. To manage this situation, we agreed to make changes to the protocol to permit participants in the control group to be allocated a befriender after they completed their 6-month follow-up. It was not possible to change the design of the study (e.g. to a waiting list control group) because of the time frame and delays to the study, as this would have required significant changes to the protocol and approval from the funders.
A number of changes to the protocol were made to improve recruitment (see Protocol changes and amendments), such as recruiting student and staff volunteers directly through University College London and the study website, and recruiting participants through advertising on posters, but these strategies did not contribute to an overall improvement in recruitment. Further changes to the study protocol were also considered, such as expanding the number of befriending services taking part in the study and introducing patient identification sites, but these changes were not implemented as the study funders did not approve extending the recruitment period, which would have also required a costed extension to the trial duration and study end date. In September 2019, the funders communicated to the chief investigator that recruitment should close and that preparations should be made to end the study in June 2020, which was the expected end date. The chairperson of the Trial Steering Committee was consulted for advice. With this time frame, it was not possible to complete the 12-month post-randomisation follow-up outcome assessments, and the protocol was updated to reflect this change. Owing to the COVID-19 pandemic, there were some delays in conducting the qualitative interviews and, therefore, a no-cost extension of 2 months was granted.
Recruitment and retention of participants
The flow of participants with ID in the pilot trial is shown in the Consolidated Standards of Reporting Trials (CONSORT) flow diagram (Figure 2). Twenty-four potential participants were referred to the study team. Fifteen referrals were from NELFT clinicians, six were from Outward (Hackney) and three were from The Befriending Scheme (Suffolk) (Table 1). The number of participants with ID and volunteers who were recruited each month is shown in Table 2. On average, just under three participants (2.7) and two (1.7) volunteers were recruited each month into the study. Of the 24 participants with ID referred to the study, 21 were assessed and five were excluded (1) because they did not meet the inclusion criteria (n = 3), (2) because they declined to take part (n = 1) or (3) because of challenging behaviour that may have put the safety of volunteers at risk (n = 1). Sixteen participants agreed to take part in the trial and were randomised to either the intervention group (befriending, resource booklet and usual care; n = 8) or the control group (resource booklet and usual care; n = 8).
FIGURE 2 The CONSORT flow diagram for participants with ID. (Figure not reproduced here; see the original article.)
| Referral | Referral source | ||
|---|---|---|---|
| NELFT | Hackney scheme (Outward) | Suffolk scheme (The Befriending Scheme) | |
| Participants with ID | |||
| Total number of referrals | 15 | 6 | 3 |
| Number eligible | 14 | 3 | 1 |
| Number randomised | 13 | 2 | 1 |
| Volunteers | |||
| Total number of referrals | N/A | 9 | 3 |
| Total number enrolled into study | N/A | 8 | 2 |
| Month | Number of participants recruited | |
|---|---|---|
| People with ID | Volunteers | |
| April | 2 | 3 |
| May | 2 | 2 |
| June | 2 | 1 |
| July | 6 | 0 |
| August | 1 | 1 |
| September | 3 | 1 |
| October | 0 | 2a |
Of the eight participants who were randomised to the intervention group, six received the befriending intervention. It was not possible to match two of the participants with a volunteer; the reasons for this were geographical distance (no local volunteer could be identified) and, in one case, deterioration in the mental state of the participant prior to matching.
Of the six participants who were matched, only four had a volunteer for the whole 6-month period. One participant was matched with another volunteer after the first volunteer dropped out because of work commitments. Of the two participants who did not have a volunteer for the 6-month period, in one case the family carer was not satisfied with the match, as they felt that the interests of the participant and the volunteer were not compatible, and in the other case the volunteer experienced difficulties working with the participant because of interference from the participant’s paid carers, who were based in the participant’s supported living accommodation.
All 16 trial participants completed the 6-month follow-up, but one participant in the control group only partially completed the outcome measures owing to a deterioration in their mental state and refusal to continue the assessment.
The flow of volunteers recruited to the study is shown in Figure 3. Twelve volunteers expressed an interest in taking part in the study (nine from the Hackney site and three from the Suffolk site). Two dropped out before being enrolled in the study (one did not complete the training and one could not be contacted for their baseline assessment). Ten volunteers completed the baseline assessment, but one dropped out shortly afterwards (no reason given). Of the nine remaining volunteers, seven were matched with a participant with ID; however, one volunteer who had been matched dropped out of the study because of work commitments. Two volunteers in Suffolk were not matched, as no participants with ID were allocated to the intervention group (the only participant to be randomised was allocated to the control group). All eight remaining volunteers completed the 6-month follow-up assessments.
FIGURE 3 Volunteer flow chart. (Figure not reproduced here; see the original article.)
Baseline characteristics of participants
The average age of the 16 participants who took part in the trial was 41.6 (SD 16.7) years; nine were female (56%), eight were of white ethnicity (50%) and nine (56%) were living in supported living accommodation. The majority of the participants had a mild ID (81%), and the average IQ was 55.3 (SD 7.6). Ten participants (63%) had a clinical diagnosis of depression and five (31%) had a diagnosis of anxiety disorder. Nine participants (56%) were taking an antidepressant.
Table 3 displays the participants’ demographic characteristics according to allocated group. The participants in the intervention group were older than those in the control group [average age 48.9 (SD 17.2) years, compared with 34.3 (SD 13.4) years, respectively], were mainly female [6 (75%) compared with 3 (38%), respectively], were less likely to have a diagnosis of depression [4 (50%) compared with 6 (75%), respectively] and there were fewer multiple mental health diagnoses [2 (25%) compared with 5 (62.5%), respectively].
| Demographic | Group | |
|---|---|---|
| Control (N = 8) | Intervention (N = 8) | |
| Age at randomisation (years), mean (SD) | 34.4 (13.4) | 48.9 (17.2) |
| Gender, n (%) | ||
| Male | 5 (63) | 2 (25) |
| Female | 3 (38) | 6 (75) |
| Ethnicity, n (%) | ||
| White | 4 (50) | 4 (50) |
| Asian | 1 (13) | 3 (38) |
| Black | 0 (0) | 1 (13) |
| Mixed | 3 (38) | 0 (0) |
| Living arrangements, n (%) | ||
| Alone | 1 (13) | 3 (38) |
| With family | 1 (13) | 1 (13) |
| Supported living (< 24 hours) | 0 (0) | 1 (13) |
| Supported living (24 hours) | 5 (63) | 3 (38) |
| Residential care | 1 (13) | 0 (0) |
| IQ score: mean (SD) | 55.6 (10.3) | 54.9 (4.1) |
| Degree of ID, n (%) | ||
| Mild | 6 (75) | 7 (88) |
| Moderate | 2 (25) | 1 (13) |
| Mental health diagnoses, n (%) | ||
| Depression | 6 (75) | 4 (50) |
| Psychosis/schizophrenia | 1 (13) | 2 (25) |
| Bipolar affective disorder | 1 (13) | 0 (0) |
| Anxiety disorder | 3 (38) | 2 (25) |
| Autism | 2 (25) | 0 (0) |
| ADHD | 2 (25) | 0 (0) |
| Other | 1 (13) | 1 (13) |
| Two or more mental health diagnoses, n (%) | 5 (62.5) | 2 (25.0) |
| Taking antidepressants: yes, n (%) | 5 (63) | 6 (75) |
| Concomitant epilepsy: yes, n (%) | 1 (13) | 0 (0) |
| Mobility, n (%) | ||
| Mobilises independently | 8 (100) | 5 (63) |
| Mobilises with walking stick or frame | 0 (0) | 3 (38) |
| Travels independently: yes, n (%) | 8 (100) | 6 (75) |
Table 4 summarises the outcome measures at baseline according to group allocation. Depressive symptoms at baseline were lower in the intervention group than in the control group, with mean GDS-LD scores of 18.6 (SD 5.7) and 21.4 (SD 9.8), respectively.
| Outcome | Group | |
|---|---|---|
| Control (N = 8) | Intervention (N = 8) | |
| Depressive symptoms (GDS-LD), mean (SD) | 21.4 (9.8) | 18.6 (5.7) |
| Self-esteem (adapted Rosenberg self-esteem scale), mean (SD) | 20.8 (4.5) | 22.4 (4.1) |
| Quality of life, mean (SD) | ||
| MANS-LD | 64.1 (10.3) | 73.0 (9.4) |
| WHOQOL-8 | 23.8 (4.7) | 30.0 (4.4) |
| Loneliness and social satisfaction (MWLQ), mean (SD) | ||
| Aloneness | 7.6 (2.8) | 8.4 (2.3) |
| Social dissatisfaction | 6.8 (3.4) | 7.9 (3.5) |
| Social support (SSSR), median (IQR) | ||
| Family | 5.0 (0.5–6.5) | 5.5 (2.5–6.5) |
| Staff | 5.5 (5.0–7.0) | 5.5 (3.5–7.5) |
| Friends | 5.5 (3.5–7.5) | 3.5 (1.5–6.0) |
| Partner | 0 (0–7.5) | 0 (0–3.5) |
| Social participation (GCPLA), mean (SD) | 73.9 (25.9) | 57.5 (23.4) |
Self-esteem, as measured by the adapted Rosenberg scale, was higher in the intervention group (mean 22.4; SD 4.1) than in the control group (mean 20.8; SD 4.5), as was quality of life, as measured by the MANS-LD [mean 73.0 (SD 9.4) and mean 64.1 (SD 10.3), respectively] and the WHOQOL-8 [mean 30.0 (SD 4.4) and mean 23.8 (SD 4.7), respectively]. Social support, as measured by the SSSR, was similar across both groups, although the control group reported more contacts with friends [median 5.5 contacts, interquartile range (IQR) 3.5–7.5 contacts] than the intervention group (median 3.5 contacts, IQR 1.5–6.0 contacts). Social participation, as measured by the GCPLA, was 16.4 units higher in the control group (mean 73.9 units, SD 25.9 units) than in the intervention group (mean 57.5 units, SD 23.4 units).
The baseline characteristics of the volunteers are summarised in Table 5. The average age of volunteers was 33.2 (SD 11.2) years. Seven (70%) volunteers were female and seven (70%) volunteers were of white ethnicity. Six (60%) volunteers were single, seven (70%) were employed and the majority had A Level (Advanced Level) qualifications or a higher level of educational qualification (n = 7; 70%). Four (30%) volunteers had prior experience of working with someone with ID and three (30%) had experience of working with someone with mental health problems. The baseline scores on the outcome measures of self-esteem, psychological well-being, loneliness and attitudes towards people with ID are shown in Table 6.
| Demographic | Befriending volunteers (N = 10) |
|---|---|
| Age at entry (years), mean (SD) | 33.2 (11.2) |
| Gender, n (%) | |
| Male | 3 (30) |
| Female | 7 (70) |
| Ethnicity, n (%) | |
| White | 7 (70) |
| Asian | 2 (20) |
| Other | 1 (10) |
| Marital status, n (%) | |
| Married or cohabiting | 4 (40) |
| Single | 6 (60) |
| Highest completed level of education, n (%) | |
| No qualification | 1 (10) |
| GCSE or equivalent | 1 (10) |
| NVQ or equivalent | 1 (10) |
| A Level or equivalent | 2 (20) |
| Degree or higher degree | 5 (50) |
| Employment status, n (%) | |
| Employed | 7 (70) |
| Unemployed, not seeking work | 1 (10) |
| Student | 2 (20) |
| Previous experience of working with, n (%) | |
| People with ID | 4 (40) |
| People with mental health problems | 3 (30) |
| Outcome | Time point, mean (SD) score | |
|---|---|---|
| Baseline (n = 10) | Follow-up (n = 8) | |
| Self-esteem (Rosenberg self-esteem scale) | 28.3 (2.3) | 32.0 (4.6) |
| Psychological well-being (WEMWBS) | 50.6 (6.9) | 53.0 (6.0) |
| Loneliness (UCLA Loneliness Scale) | 55.5 (3.9) | 56.0 (4.0) |
| Attitudes (ATTID) | ||
| Discomfort | 30.5 (10.6) | 29.9 (9.3) |
| Knowledge of capacity and rights | 46.6 (6.6) | 42.9 (6.9) |
| Interaction | 26.6 (8.7) | 24.4 (8.6) |
| Sensibility or tenderness | 16.5 (2.9) | 16.6 (3.5) |
| Knowledge of causes of ID | 18.8 (6.4) | 19.4 (7.0) |
Adverse events
There were no serious adverse events associated with the intervention for participants with ID. In the case of volunteers, there was one incident involving a disagreement between a paid carer at a supported living placement and a volunteer. The volunteer was unfairly accused of buying the participant an unhealthy meal and disregarding his care plan. This disagreement led to a breakdown in the befriending relationship and was upsetting for the volunteer. The volunteer reported this incident to the volunteer co-ordinator, who raised the issue with the manager of the placement. The manager acknowledged that there was a misunderstanding between the paid carer and volunteer.
Other incidents
A volunteer raised a safeguarding alert after their befriendee reported feeling bullied by a member of staff at their supported living placement. The volunteer supported the individual to report the incident. The incident was investigated by the local authority and the situation was satisfactorily resolved.
Outcome assessments at 6 months’ follow-up
Table 7 shows the scores on the outcome measures for participants with ID at 6 months’ follow-up. Depressive symptoms at 6 months’ follow-up were lower in the intervention group than in the control group, with mean GDS-LD scores of 12.9 (SD 6.7) and 17.5 (SD 6.5), respectively. Compared with baseline scores, there was a reduction in the GDS-LD score in both groups (3.9 units in the control group and 5.7 units in the intervention group). There was a slight increase in mean self-esteem scores in both groups: an increase of 0.6 units in the control group, to 21.4 units (SD 3.5 units), and an increase of 1.5 units in the intervention group, to 23.9 units (SD 3.4 units). Mean quality-of-life scores on the MANS-LD also increased in both groups: an increase of 7.9 units in the control group, to 72 (SD 5.5) units, and an increase of 7.0 units for the intervention group, to 80.0 (SD 9.9) units. The mean WHOQOL-8 scores in the control group increased by 4.6 units, to 28.4 (SD 3.5) units, but the scores in the intervention group decreased by 0.6 units, to 29.4 (SD 5.3) units. There were only small changes in the scores for loneliness and social dissatisfaction on the MWLQ. The median social support scores increased slightly in the control group in relation to family (an increase of 2 units, to 7.0 units, IQR 4.0–9.0 units), compared with no change in the intervention group, and in relation to friends (an increase of 1.5 units to 7.0 units, IQR 4.0–9.0 units), compared with an increase of 0.5 units in the intervention group (4.0 units, IQR 3–6.5 units). Between baseline and follow-up, the social participation score increased by 3.8 units in the intervention group (to a mean of 61.8; SD 22.5) and by 0.5 units in the control group (to a mean of 74.4; SD 20.4).
| Outcome | Group | |
|---|---|---|
| Control (n = 8) | Intervention (n = 8) | |
| Depressive symptoms (GDS-LD), mean (SD) | 17.5 (6.5) | 12.9 (6.7) |
| Self-esteem (adapted Rosenberg self-esteem scale), mean (SD) | 21.4 (3.5) | 23.9 (3.4) |
| Quality of life, mean (SD) | ||
| MANS-LD | 72.0 (5.5) | 80.0 (9.9) |
| WHOQOL-8 | 28.4 (3.5) | 29.4 (5.3) |
| Loneliness and social satisfaction (MWLQ), mean (SD) | ||
| Aloneness | 7.0 (3.7) | 8.6 (3.5) |
| Social dissatisfaction | 8.3 (2.7) | 7.6 (2.1) |
| Social support (SSSR), median (IQR) | ||
| Family | 7.0 (4.0–9.0) | 5.5 (1.5–7.0) |
| Staff | 6.0 (5.0–8.0) | 6.0 (2.5–7.5) |
| Friends | 7.0 (4.0–9.0) | 4.0 (3.0–6.5) |
| Partner | 0 (0–0) | 0 (0–5.0) |
| Social participation (GCPLA), mean (SD) | 74.4 (20.4) | 61.3 (22.5) |
Regression analysis of main outcome (depression)
After adjustment for depressive symptoms at baseline, depressive symptoms were lower in the intervention group. We estimated a four-point difference in the GDS-LD depressive symptoms score in the intervention group compared with the control group (adjusted mean difference –4.0, 95% CI –11.2 to 3.2). This difference was equivalent to a standardised effect size of 0.5 SDs.
Outcome assessments in volunteers
At 6 months’ follow-up, there was an increase in the mean scores for self-esteem (an increase of 3.7 units to 32.0 units; SD 4.6 units) and psychological well-being for volunteers (an increase of 2.4 units to 53.0 units; SD 6.0 units; see Table 6) from baseline. There was a slight increase in mean loneliness scores at 6 months’ follow-up (a 0.5-unit increase to 56.0 units; SD 4.0 units). Scores on the measure of attitudes towards ID changed very little, apart from a slight improvement in the mean scores for knowledge of capacity and rights (a decrease in score of 3.7 units to 42.9 units; SD 6.9 units).
Health economic data
The total cost of intervention implementation was £5647.38 (see Appendix 2, Table 12), which includes the cost of recruitment (staff time); DBS and reference checks; the cost of staff time and resources to deliver training, supervision, monitoring calls and visits; and volunteer expenses. The mean cost per participant was £701.00.
It was possible to obtain NHS resource data from carers or from health professionals in the majority of cases, but three participants (who lived independently and had minimal support) provided self-report data. Data were available for all 16 participants at baseline and follow-up. No significant difficulties were reported by carers or participants with ID, although participants with ID may have had difficulty recalling the number of contacts with health services over the previous 6 months. Participants were encouraged to use their diaries to record this information.
The costs of NHS resource use and EQ-5D-Y health state values at baseline and at 6 months post randomisation are provided in Appendix 2, Tables 13 and 14. Health-care resource use costs at baseline were substantially lower in the intervention group (mean £1387; 95% CI £359 to £2415) than in the control group (mean £4949; 95% CI £351 to £9548). In both groups, resource use costs at 6 months’ follow-up were lower than those at baseline (mean £855 in the intervention group and mean £4014 in the control group). The mean difference in resource use costs between the two groups was –£3159 (95% CI –£6250 to –£73), but this difference does not take into account baseline cost differences. After including the intervention cost, the mean difference was –£2458 (95% CI –£5549 to £628), but was not statistically significant.
The EQ-5D-Y and EQ VAS were completed by individuals with ID. The EQ-5D-Y was completed by all participants at baseline, but data were missing for one participant at 6-month follow-up. The EQ VAS was completed by almost all of the participants (data were missing for one participant at baseline and follow-up). No issues were reported in the completion of the EQ-5D-Y descriptive system, but some participants found the instructions for the EQ VAS difficult to understand, which may explain the discrepancies in the scores between the EQ-5D-Y and EQ VAS (see below).
At baseline, the EQ-5D-Y utility value was slightly higher in the intervention group (0.90; SD 0.10) than in the control group (0.88; SD 0.08). At 6 months, it was slightly higher in the control group (0.90; SD 0.13) than in the intervention group (0.88; SD 0.11). At baseline, the EQ VAS score was substantially higher in the intervention group (75.0; SD 23.2) than in the control group (45,6; SD 31.5), but, at 6 months’ follow-up, the EQ VAS score appeared to decrease in the intervention group (62.3; SD 40.8) and increased slightly in the control group (50.7; SD 27.6). The CIs for the EQ VAS were large, and the scores did not appear to correspond well to the EQ-5D-Y utility values, suggesting that the EQ VAS may be difficult for people with ID to complete.
We have explored further the responses to the EQ-5D-Y questionnaires that are displayed in Appendix 2, Tables 15 and 16, splitting the EQ-5D-Y levels into ‘no problem’ (level 1) and ‘any problems’ (levels 2 and 3), changing the profile into frequencies of reported problems. This is helpful, as it describes the changes in health over time. Although an overall summary is preferred, this approach shows that there were improvements in usual activities; pain and discomfort; and feeling worried, sad or unhappy in the control group; and improvements in participants looking after themselves in the intervention group.
Owing to the small sample size and discrepancies in the scores, QALYs were not calculated. Therefore, the incremental cost per change in the primary health outcome (depression scores on the GDS-LD) between the intervention group and control group was estimated. After adjustment for baseline values, the ICER was –£1810 per 1-point change in GDS-LD score.
The befriending intervention has a 35% probability of being cost-effective at a willingness to pay of £20,000 (see Appendix 5, Figure 5). These results are illustrative only and should be interpreted with caution because of the small sample size.
Protocol changes and amendments
Protocol changes and amendments were approved by the Public Health Research programme manager and submitted for sponsorship and ethics approval. The main changes were as follows:
- Amendment 1 (approved on 21 June 2019): changes to recruitment to include recruitment of volunteers via the trial website, recruitment of undergraduate students at University College London and use of posters to recruit participants with ID from ID services in NELFT. Changes to the matching process were also introduced, following recommendations from the Trial Steering Committee, enabling participants to be rematched if a volunteer dropped out, and for more than one volunteer to be matched to a participant.
- Amendment 2 (approved on 25 March 2020): the 12-month follow-up was removed because of significant delays to the study. Focus groups with volunteers, carers and volunteer co-ordinators were changed to interviews to assist data collection.