The full protocol for this feasibility cluster trial with nested qualitative data has been previously reported in detail.30 The objectives were i) to determine the feasibility of conducting a randomized cluster trial in terms of recruitment and retention; ii) to assess the feasibility of implementing the 9-cell bereavement tool; iii) to determine whether there would be contamination between the clusters; iv) to assess the acceptability and completeness of measures and data; v) to identify trial participants’ views and experience of the intervention and its mechanisms of action; vi) to estimate potential effect size; and vii) to determine whether a full trial is warranted. We summarise the methods below. They were conducted as per protocol.
This feasibility trial took place in Chitungwiza, a high-density dormitory town (population 456,000) situated approximately 30 kilometers from the capital city of Harare, Zimbabwe. We selected two comparable suburbs as study sites, 8 kilometers apart to reduce risk of contamination. Island Hospice and Healthcare Zimbabwe (IHH) collaborated on the study given their longstanding involvement in Chitungwiza, and connections to the local community groups, churches, local government structures. Lay community workers were recruited amongst the community leaders that Island Hospice and Healthcare Zimbabwe had been working with in other programs in Chitungwiza.
The 9-cell bereavement tool’s structure i) draws on participants’ existing knowledge, models an open-minded, non-judgmental approach, recognizes the diversity of grief within individuals, genders, families, cultures, and faiths, and encourages participants to listen to others while breaking down previously held beliefs about how grief can be expressed. It draws on Stroebe and Schut's oscillation model (1998) 31, which focuses on people’s oscillation between the process of grief in itself, together with re-engaging with a life transformed by the loss32. The intervention uses 9 ‘cells’ to help an individual identify i) personal feelings in relation to their bereavement; ii) judgmental attitudes, religious tenets, lack of understanding, and iii) effects of family and community support. A 9-cell table (see Figure 1) is constructed, the horizontal line representing three time points after the loss of a loved one: i) the immediate, ii) a little while later, and, iii) a long time after the event. The vertical divisions examine i) the individual’s feelings, ii) how these are outwardly expressed, and, iii) what is culturally permissible.
Discussions comparing the different cells in the 9-cell table and introducing evidence-based alternative points for consideration, act as the intervention. They help participants realize the gap between what they feel and what is permissible, and then work together to develop a personal bereavement approach. Discussion can take place in either direction, first examining feelings along a timeline, or exploring how expressions differ between immediate feelings, what is outwardly expressed and what is permissible. The facilitator merely asks for thoughts or personal experiences to be identified in each cell, probing for further details, differences between experiences in the group and, in particular, contrasting thoughts relating to rituals, religious teachings and individual interpretations of these, while offering optional ideas for reflection.
There were three levels of recruitment: 1) Community leaders were identified, 2) Community leaders then recruited lay community members to participate in the one-day intervention training (i.e. interventionists), 3) Interventionists then recruited trial participants and delivered the intervention to them post baseline (T0) data collection.
Week 1: The recruitment process began with study engagement using existing relationships that Island Hospice and Healthcare has with community leaders (CL) from the selected two communities in Chitungwiza. The rationale, goals and intended procedures of the study were explained, community leader permission was sought to undertake the research, and support from the community leaders requested to deliver the study.
Community leaders were tasked to invite 25 people (i.e. potential interventionists) from their respective communities or clusters, who they knew to have suffered loss in the past 6 to 18 months to attend a meeting. This allowed for a total of 50 potential interventionists representing both clusters.
Week 2: Separately, the potential interventionists, from each of the two clusters, were provided with information about the study and asked to give informed consent to indicate their willingness to participate in the study. They completed a short, self-administered questionnaire that assessed their socio-demographic background and bereavement history. Trained surveyors provided instruction to the interventionists, and were available to assist with clarification of any questions where required. None of the two clusters were aware which of them was the intervention and which was the control group.
The interventionists from each community were tasked with identifying 2 to 3 people according to the following criteria: a) at least 18 years old, b) resident within their neighbourhoods, c) someone with whom they interacted with on a daily basis, d) whom they know to have been bereaved in the past 6 months, e) would have the ability to either verbally consent or be able to provide written consent, f) be expected to attend and participate in the study.
Week 3: Potential participants who met all of these criteria attended a meeting where information regarding the study was shared. Before the interventionists started the recruitment process, a focus group discussion with a sub-sample of them, was conducted to assess the feasibility of them identifying potential trial participants and inviting them to meet with the researchers to learn more about the study.
Following written informed consent, research assistants collected sociodemographic information and bereavement history from trial participants. Trial participants also completed three psychosocial measures at baseline, then at 3 months (T1) and 6 months (T2) post randomization: the Shona Symptom Questionnaire (SSQ)33 a screening tool for mental health, the Medical Outcomes Study - Social Support Survey (MOS-SSS)34 which measures social support and the Texas Revised Inventory of Grief (TRIG) 35,36which measures intensity of a person’s grief. A question also asked if the participant had contact with anyone within the other geographical cluster.
The two study sites were randomized following T0 data collection by a statistician (independent of the study) at Kings College London to intervention or wait-list control.
Focus group data
Focus group discussions were held with interventionists to examine their experience and use of the training material, implementation and processes within the intervention. Discussions were captured through detailed notes captured by a note-taker and translated from the vernacular (Shona) to English.
Past research has recommended sample sizes of 24 and 50 37-40 . For this feasibility study, sample sizes for community leaders were set at 25 in total, and for interventionists, at 25 per cluster; with the trial participants set at 75. The target was 50, however a reach of 75 was set, to allow for any challenges in recruitment and for possible attrition from the study.
The analysis was conducted as follows, and in line with the study objectives.
1. To determine the feasibility of conducting a randomized cluster trial in terms of recruitment and retention;
Numbers of both interventionists and trial participants recruited, and participating in the trial were recorded at baseline (T0), midline (T1) and endline (T2). At the end of each data collection phase, researchers analysed the flow of work to assess what worked and what needed adjustment in allowing a conducive environment for optimum data collection from the participants. Where any process may have affected the resultant figures, these were documented in short reports to be used as reference points for the end of trial report.
2. To assess the feasibility of implementing the 9-cell bereavement tool;
This was established in part through qualitative observation of the discussions that emerged during intervention implementation and through a discussion with the facilitators themselves post-study to assess (a) their experience in implementing the intervention, (b) their assessment of the experience of the interventionists as they participated in the intervention, and (c) to highlight what made the process feasible. Framework analysis was used to analyse the qualitative data collected from the discussion.
3. To determine whether there would be contamination between the clusters;
Contamination questions were provided to both the intervention and the control group, to assess whether any of the participants had engaged, visited or spoken to either party in between the data collection dates. Questions were centered on whether they had visited the intervention community, and/or been in contact with a participant from the intervention group who may have shared with them, information about the study. Responses were manually assessed to see if any had contact and had been ‘contaminated’.
4. To assess the acceptability and completeness of measures and data;
Questionnaire data from quantitative data collected at baseline (T0), midline (T1) and endline (T2), were entered into an Excel spread sheet. These were then imported into Stata V15 software for analysis 41. All participants’ data were analysed according to the community in which they were recruited and randomised.
We tabulated all the variables for each measure used from baseline, midline and endline and each variable with missing data is recorded with reasons provided. We summed up the number of participants who completed measures at baseline, midline and endline.
5. To identify trial participants’ views and experience of the intervention and its mechanisms of action;
Focus group discussions were held with the trial participants at the end of the trial. Framework analysis with emerging themes around specific questions was used to analyse their responses.
6. To estimate potential effect size
We calculated baseline and final scores for outcome measures (SSQ, TRIG, and MOS-SSS) and summarised within-group changes for each outcome measures and for differences between communities at the final timepoint.
We performed longitudinal analysis using multilevel modelling for repeated measures with generalised linear latent and mixed models (GLAMM) which accounts for correlated or clustered data over time in analysing categorical data. Each outcome was divided into quartiles, as GLAMM operates more successfully with fewer categories of the dependent variable, and was adjusted by baseline score. This enabled comparison of the effect of the intervention on all dependent variables.
7. To determine whether a full trial is warranted.
The combined results from the above analysis would warrant whether a full trial were possible, with the ability of conducting a full RCT in terms of recruitment and retention; successful implementation of the intervention; little or absent contamination between the clusters; acceptability and completeness and measures and data; identification of trial participant’s views and experience of the intervention and its mechanisms of action; and the ability to estimate potential effect size.
Ethical approvals were obtained from the Zimbabwe National Medical Research Council of Zimbabwe (MRCZ/A/2230) and from King’s College London (HR-17/18-5415). In addition, the study sought clearance from the local community police and researchers carried all clearance documents with them throughout the study.