Background
Group interventions in healthcare
Evaluation of group interventions
Objectives
Methods
Case studies
Study | Population (target) | Group intervention [additional sessions] | Facilitator (per session) | Method and timing of recruitment | Type and timing of randomisation | Outcome data collection |
---|---|---|---|---|---|---|
LM [2] | 65+ years (prevention: general) | 16 × weekly sessions: 2-h face-to-face occupational therapy [4 × one-to-one sessions] | Two NHS Band 4 equivalent staff. Recruited on university contracts specifically to deliver the intervention Training: two- day face-to-face course | Mail-out via GPs, healthcare referral and self-referral from study promotion (including researchers visiting dementia cafes and other groups) | Individual. Done by central team immediately after consent and baseline data collection | Central research assistants; blinded outcome assessor; follow-up anchored to randomisation |
PLINY [3] | 75+ years (prevention: loneliness) | 12 × weekly sessions: 1-h telephone friendship [6 × one-to-one telephone calls before group] | One volunteer from a community organisation. Training: 4 × 1 h sessions via telephone | Mail-out via GPs and to research cohort | Individual. Done by central team immediately after consent and baseline data collection | Central research assistants; blinded outcome assessor; follow-up anchored to randomisation |
REPOSE [4] | 18+ years. Type I diabetes (therapy: self-care education) | 5 × daily sessions: full-day face-to-face Education (total approx. 38 h). [1 optional × group follow-up session] | Two diabetes specialist nurses/dieticians. Training: five- day observation, three-day face-to-face workshop, peer-reviewed delivery of five-day course and one-day workshop (105 h) | Referral via care team in person or via mail-out | Cluster by course in pairs. Delayed randomisation: after groups were filled, 6 weeks before first course; baseline taken after randomisation | Facilitator at clinic visits; unblinded outcome assessor; follow-up anchored to group attendance |
STEPWISE [5] | 18+ years. First episode psychosis + schizophrenia (prevention: cardiovascular) | 4 × weekly sessions: 2.5-h face-to-face. [3 × quarterly booster group sessions and fortnightly 1:1 support calls between booster sessions] | Two NHS staff (mental health staff; occupation therapists and dieticians). Training: three-day face-to-face course plus one-day booster session training | Referral via care team | Individual. Done immediately after consent and baseline data collection | CMHT staff or research nurses; blinded outcome assessor; follow-up anchored to randomisation |
JtD [1] | 18+ years. Dementia (prevention: dependency) | 12 × weekly sessions: Approx. 2-h face-to-face psychosocial education [4 × one-to-one sessions] | Two NHS staff. Training: two-day face-to-face course, plus online resources | Mail-out via GPs/care teams, mail-out to research cohort by research team, referral via care team, or self-referral from study promotion (including researchers visiting dementia cafes and other groups) | Individual. Delayed randomisation: after collection of baseline data < 2 months before intervention | Central and local site research assistants; blinded outcome assessor; follow-up anchored to randomisation |
Results
Participant recruitment and attrition
Invited to take part (mail-out only) | Response rate to mail out (% of sent) | Screened | Eligible (% of screened) | Consented (% of screened) | Randomised (% of screened) | Attended at least one group session (% of randomised) | |
---|---|---|---|---|---|---|---|
Randomisation concurrent with consent | |||||||
LM [2] | 9330 a | 385 (4.1) a | 313 | 294 (93.9) | 288 (92.0) | 288 (92.0) | 123/145 (84.8) |
PLINY [3] | 9579 b | 275 (2.9) b | 178 | 159 (89.3) | 157 (88.2) | 157 (88.2) | 21/78 (26.9) |
STEPWISE [5] | N/A | N/A | 1223 | 989 (80.9) | 423 (34.6) | 414 (33.9) | 171/208 (82.2) |
Delayed randomisation | |||||||
REPOSE [4] | 1278 | 885 (69.2) | 362 | 334 (92.3) | 321 (88.7) | 317 (87.6) | 267/317 c (84.2) |
JtD [1] | 958 a | 68 (7.1) a | 1183 | 521 (44.1) | 520 (43.9) | 480 (40.6) | 217/239 (90.8) |
Setting group dates
Attrition
Participant demand and facilitator supply
Facilitator training
PLINY case study: facilitator supply did not meet participant demand
Therapeutic dose
Minimum number of sessions for ‘therapeutic dose’ (or per protocol population) | Median (range) sessions attended | Number of participants achieving ‘therapeutic dose’ (% of those receiving any intervention) | Number attending all planned sessions (% of those receiving any intervention) | |
---|---|---|---|---|
LM [2] | 12/16 (75%) | 12 (1–16) | 71/123 (57.7%) | 5/123 (4.1%) |
PLINY [3] | 9/12 (75%) | 11 (2–12) | 9/21 (42.9%) | 10/21 (47.6%) |
REPOSE [4] | 4/5 (80%) | 5 (1–5) | 261/267 (97.8%) | 250/267 (93.6%) |
STEPWISE [5] | 2/7 a (28.6%) | Foundation: 3 (0–4) Booster: 2 (0–3) | 161/171 a (94.2%) | 47/171 a (27.5%) |
JtD [1] | 10/16 b (62.5%) | 1:1 sessions: 4 (1–4) Groups: 9 (0–12) | 168/217 b (77.4%) | 36/217 b (16.6%) |
Group size
Ideal group size | Mean number of participants per group (SD) a | Median number of participants per group (IQR) a | Minimum and maximum group size a | Number of sessions delivered with correct group size (% of number of sessions delivered a) | Number of sessions with 0 or 1 person recorded (% of planned sessions) | |
---|---|---|---|---|---|---|
LM [2] | 8–16 | 7.7 (2.42) | 7 (6–9) | 2–15 | 80/173 (46.2%) | 13/186 (6.9%) |
PLINY [3] | 6–8 | 5.1 (1.17) | 6 (4–6) | 3–6 | 24/41 (58.5%) | 1/42 (2.4%) |
REPOSE [4] | 5–8 | 5.8 (1.38) | 6 (5–7) | 3–8 | 36/46 (78.3%) | 0/46 (0%) |
STEPWISE [5] | 6–8 | Foundation: 4.2 (1.69) | 4 (3–5) | 2–9 | 25/129 (18.9%) | 3/132 (2.3%) |
Booster: 3.5 (1.41) | 3 (2–4) | 2–8 | 7/75 (7.1%) | 23/98 (23.5%) | ||
JtD [1] | 8–12 | 5.3 (1.71) | 6 (4–6) | 2–9 | 32/331 (9.7%) | 5/336 (1.5%) |
Process evaluation
Training | Delivery of treatment | Receipt of treatment | Enactment of skills | Findings a | |
---|---|---|---|---|---|
LM [2] | Standardised training and materials; direct observation of training, checklist used by two researchers | Visual recording of a purposive sample of sessions; assessed by two people using delivery checklist; facilitator supervision; attendance registers | Assessed through qualitative research with a subset of participants at 6 months after randomisation; attendance registers | Assessed through qualitative research with a subset of participants at 6 and 24 months after randomisation | Reported as categorical for fidelity: Satisfactory (based on a continuous fidelity score of 61%–70%) in 7/8 recordings for both participant and facilitator performance |
PLINY [3] | Standardised training and materials; audio recording of purposive sample of training sessions; checklist used by two researchers; checklist for facilitator skill acquisition | Audio recording of sample of groups at three timepoints; assessed by two people using delivery checklist; facilitator supervision; attendance registers | Checklist for treatment included elements of receipt and assessed through qualitative research with a subset of participants; attendance registers | As for receipt, as enactment of skills should have taken place in the group sessions in this intervention | Reported percentage scores for fidelity: Training scores were high but overall fidelity scores for facilitators were low |
REPOSE [4] | Standardised training and materials | Direct observation of particular sessions in intervention arm; existing quality assurance for standard DAFNE course; facilitator supervision + feedback from fidelity assessment; attendance registers | Assessed through qualitative research with a subset of participants immediately after intervention; attendance registers | Assessed through qualitative research with a subset of participants at 6 months after intervention | Achievement of essential outcomes assessed as: Yes, Partial (with a percentage) or Not Observed; appear to be delivered as planned; fed back to facilitators |
STEPWISE [5] | Standardised training and materials | Direct observation of a sample of 30 sessions in 10 sites using a checklist and an interaction observation tool; also explored in qualitative interviews with a subset of participants, facilitators and intervention developers; facilitator supervision; attendance registers | Assessed through qualitative interviews with a subset of participants and facilitators; checklist for session included elements of receipt and enactment; attendance registers | Reported percent of facilitator talk time and the percent of time in positive and negative behaviours; poor intervention fidelity reported | |
JtD [1] | Standardised training and materials; direct observation of training, checklist used by two researchers; checklist for facilitator skill acquisition | Direct observation of two sessions of four groups at four sites; checklist used by two independent observers; self-report checklist by facilitator at each session; facilitator supervision and supervision checklists; attendance registers | Assessed through qualitative interviews with a subset of participants; attendance registers | Assessed through qualitative interviews with a subset of participants | The fidelity data have not yet been analysed or interpreted |
Clustering concerns
Couple recruitment
More than one facilitator
Discussion
Principal findings
Participant recruitment and attrition
Facilitator training and attrition
Therapeutic dose
Group size
Process evaluation
Clustering issues
Challenges and solutions for group intervention implementation
Challenges in group intervention implementation | Potential solutions |
---|---|
Participant recruitment: A number of participants (likely to be a minimum of 8 in each arm to allow for drop out) need to be recruited before a group intervention can commence Participant recruitment rates can be affected by the relative timing of screening, randomisation and initiation of group, but there is no clear signal as to the best strategy | Recruitment projection and the timing of intervention delivery need to be considered together at the design stage. Those planning group intervention trials should consider demand-forecasting procedures, like those used in clinical settings characterised by surges and slumps and should aim for the maximum group number before commencing the group sessions to allow for attrition. Consider the pros and cons of randomising patients closer to consent (potentially better for accrual rates) or to group initiation (potentially better for retention rates) |
Participant attrition: Attrition may be greater where individuals are required to wait longer before starting the group sessions (due to the requirement of recruiting enough people per group) | Delaying randomisation until there are enough recruited participants to run the group may lead to attrition between consent and randomisation but reduce post-randomisation, pre-intervention attrition, which is important to maintain statistical power Maintaining contact with participants before randomisation and the setting up of group sessions may help to reduce pre-randomisation attrition If randomising a number of participants at the same time, trialists should consider and plan for the impact on the follow-up data collection, these participants will need to be followed-up at the same time |
Setting group dates: Deciding when to set the group sessions can be challenging. Day/dates can be set before recruitment or once all participants needed for a group session are recruited | Our data did not suggest that either method is superior Those planning group intervention trials should plan groups around the recruitment projection and allow for some flexibility if recruitment does not go as planned |
Facilitator training and attrition: Two facilitators are often needed to run group sessions. Recruiting facilitators can be challenging Group facilitators will be lost over the course of the intervention delivery – our data show 70% attrition of trained facilitators in one trial | Allow enough time to recruit and train facilitators prior to the start of recruitment Plan to train replacement facilitators at each site, and/or plan for training sessions throughout the project to account for facilitator attrition or re-training facilitators Group dynamics are important to group interventions; any change in facilitator should be recorded and investigated as part of the process evaluation and through multi-level modelling for analysis where appropriate |
Therapeutic dose: This can be difficult to determine for complex interventions but is required for a per-protocol analysis. This may be more difficult for group interventions, as there is less control over what people are exposed to than in one-to-one sessions In our experience, investigators define a therapeutic dose by a threshold number of sessions attended | Defining the per-protocol population should be undertaken by expert consensus, with oversight from the project steering committee. Time should be reserved for this purpose during protocol development For group interventions, ‘therapeutic dose’ may relate to certain intervention criteria being delivered rather than the number of sessions attended and this should be investigated as part of the process evaluation |
Group size: An ideal group size will be applicable to the intervention but may be difficult to achieve for all group sessions. Groups may have to run with fewer participants than the ideal. There may be reluctance to amend the group membership (e.g. by adding new participants) once running due to the impact of group dynamics | The impact of group size on the effectiveness of the intervention and must be considered in fidelity assessments and on outcomes Protocol development should include discussions about what happens in the event of small groups and should specify if any number of participants is too few for intervention delivery. Can groups be combined or can new participants or non-participants be added? Consider whether the group size or the maintenance of group dynamics is more important to the intervention |
Process evaluation: Assumes interventions work at an individual level meaning some constructs may need adapting for assessing group interventions | Recruitment and ‘dose delivered’ can be assessed at the group level whereas ‘dose received’ can be assessed at the individual level; fidelity can be assessed at the group (delivery) or individual level (receipt and enactment of skills) Process evaluations should include components of the intervention specific to group processes, such as facilitation techniques, group dynamics and development and inter-personal change processes |
Clustering issues: Couples recruited to trials and participants that receive the same intervention from the same facilitators are likely to have more similar outcomes than if this was not the case. RCTs may not be powered to use multi-level modelling | This needs to be accounted for in the sample size calculation and made clear when interpreting the findings |