Priorities for enhancing nurses' and social workers’ competence and confidence in helping families support dependent children through parental death. A classic-Delphi survey

In this classic-Delphi survey, we aimed to establish the key issues faced by N&SWs working in the UK, when supporting families with dependent children through parental death from any non-communicable, life-limiting condition. We also sought consensus on the priorities for enhancing their confidence and competence in providing effective support. Our objectives were to develop research findings that would inform employers and care providers in middle to high-income nations about unmet workforce needs that could be addressed through continuing professional education.

We addressed issues and priorities for both N&SWs as national 16-20 and international [16‐18, 21‐23] peer-reviewed literature highlights similar issues for these professions. Previous research findings [16‐23] highlight that N&SWs experience caring for parents with dependent children when one parent has a life-limiting illness as challenging. However, there is limited knowledge of specific issues these professionals face in assisting families to support children, or of the priorities for enhancing N&SWs’ confidence and competence in doing so.

The survey was conducted with a purposively selected panel of topic experts based in the UK. See Table 1. for participants’ inclusion and exclusion criteria.

There is a lack of consensus over the ideal number of experts needed for a Delphi survey panel [37]. Largely, this depends on the degree of homogeneity or heterogeneity of panellists’ expertise [37]. In selecting the sample, we were guided by the systematic review by Boulkedid et al. [37] asserting that heterogeneous panels of topic experts can enhance research credibility and the stance of Baker et al. [38] who position that heterogeneous panels should comprise at least 20 people. We identified a panel of 55 topic experts who collectively had the expertise for this consensus-building exercise. Because we selected panellists for their expertise rather than geographical distribution, 54 of those selected were from England and one was from Scotland.

Eighty-four per cent (46/55) of our HSCPs sample were identified based on our knowledge of their expertise in clinical practice, research, and education; with the other 16% recruited by the authors’ links within clinical practice. We purposively sampled panellists who were national and local leaders (n=46) working in practice, education, and research. Participants worked in acute hospitals, hospices, the tertiary sector, and education. Using non-probability techniques [39] we sampled 46 UK registered HSCPs with contemporary real-world knowledge including N&SWs; doctors; clinical psychologists; bereavement counsellors, play, drama, and occupational therapists. Additionally, academics and directors of independent organisations who were registered HSCPs were included.

Additionally, we sampled a subset (n=9) of people living with bereavement. This subset comprised parents (n=6) living in the UK, whose partner had died whilst parenting dependent children, and young UK adults (n=3) aged 16-25, who were bereaved as dependent children and had the lived experiences of issues encountered when preparing for parental death.

Akard et al. [40] describe how children's involvement in grief studies might be emotionally challenging in the first year following a family member's death. Therefore, young adults — and bereaved parents — were accessed via the lead for a UK specialist support charity. This individual supported these participants during the research and was on hand should they become distressed through participating in the study.

Figure 1 illustrates the four stages of this three-round Delphi survey, the process of developing survey instruments and data analysis procedures at each stage.

Hasson, Keeney, and McKenna [35] posit that, assuming enough questions are asked, three rounds of a classic Delphi survey are sufficient to reach consensus and or to achieve meaningful results. Further, a systematic review by Diamond et al. [41] reported that of the 100 Delphi studies they reviewed, 98% aimed to achieve consensus. Most (71%) specified a priori the number of survey rounds to be conducted, two-thirds stated an a priori level of consensus sought [41]. However, there was no clear definition of consensus [41]. Of the studies setting a percentage threshold a priori, the most common was 75% (range: 50-97%) [41]; where 75% of panellists agreed or strongly agreed with items when measured using a Likert scale.

In keeping with Keeney et al. [42] we designed a three-round Delphi survey, using a Likert scale and setting a percentage threshold of 75% agree or strongly agree to statements.

Our Delphi survey comprised three iterative rounds during which a survey questionnaire was completed by panellists during each round. The first-round questionnaire (Additional file 1) was developed from findings reported in a previously conducted qualitative review and thematic synthesis concerning HSCPs’ experiences of supporting parents and their dependent children through parental death [17]. Table 2. shows the sections and questions presented to participants in round one of the survey.

In round two, the survey questionnaire (see Additional file 2) was derived from content analysis of panellists’ reporting on the issues, challenges, and priorities in round one.

The third-round survey questionnaire (see Additional file 3) was based on the findings from round two. In round three, panellists completed the same questionnaire as in round two concerning the issues and challenges faced by N&SWs. Before completing the third-round questionnaire, each panellist was provided with their own scoring from round two alongside the group Median (Mdn.) for round two. This allowed panellists to consider their responses in light of the group average (Mdn.) (see Fig. 1). Additionally, in round three, we asked panellists to rank the priorities for enhancing N&SWs support to families and children before and after the death of a parent in their preferred order.

All survey questionnaires were administered electronically, hosted by a UK-based online survey platform [43]. The survey questionnaires were piloted and reviewed before circulation to participants. Panellists were offered paper versions if requested — none were. Potential panellists were sent an email, containing a link to the online version of the survey. Before completing each round, they were asked to read the Participant Information Sheet informing them of the purpose of the survey and why they were asked to take part. Before participating they completed an online consent form.

We sent three email reminders (two weeks apart) to panellists not responding to rounds one or two. Because in round three they were asked to consider their responses to the items scored in round two, panellists not participating in this round were not sent the round three questionnaire.

Content analysis [34] (Table 3) was used to analyse panellists’ ‘free-text’ responses to the round one survey questions (Table 2).

.

Two authors (PF and AA) independently analysed survey responses and then compared findings and developed statements for the following round. When phrasing statements for inclusion in the second and third-round questionnaires, we tried to stay close to words and phrases used in panellists’ original written responses [34].

In round two, in keeping with analytical methods reported in previous research [30], we used the median (Mdn.) to measure the group’s average level of agreement and the Mean Absolute Deviation from the Median. (MADM.) to measure the extent of agreement for each statement rated by panellists. Statements for rating were related to issues N&SWs face and priorities for enhancing their support to families and children before and after the death of a parent. Median scores of 4.00 and above were deemed high-level agreement, \(\ge\) 3.00 \(\le\) 4.00 moderate, and 1.00 \(-\le\) 3.00 low. The median is a good metric for this purpose as it is not influenced by extreme values [44]. Further, studies on consensus development [45‐49] advocate reporting the MADM. to depict the extent of agreement. Like the median, the MADM. is a useful measurement of variance in non-parametric studies as it disregards outliers [45, 49]. A MADM. of <1 demonstrates that most data values are close to the median and > 1 indicates wide variation [49]. The lower the MADM., the stronger the consensus [30, 49]. We used SPSS [50] to calculate the MADM. We chose the MADM. over the interquartile range (IQR.) because the MADM. is purported to have more sensitivity [30, 49]. After the third survey questionnaire, we tested relationships between panellists’ ratings, in rounds two and three, regarding issues faced by N&SWs.

Based on previously reported methods [24, 29, 41, 42] we set our threshold for consensus at \(\ge\) 75% (range 50-100%) of topic experts agreeing with the statements rated in rounds two and three achieving high-level agreement (agree, agree strongly), with ratings remaining stable across rounds two and three. Further, we investigated stability of ratings across rounds two and three as Crisp et al. [51] advocate the importance of stability. Therefore, in round three, we asked panellists to place the priorities for enhancing N&SWs' ability to provide support for families with children aged under eighteen years in their preferred order. The relative importance of priorities was ranked according to the inverse sum of weighted averages [52]. First, each priority was assigned a value (weighted rank), then the frequency of panellists’ (n=28) preferred ordering was multiplied by this value. The results of ranking for each priority were then summed and the sum of ranks presented in inverse order of importance (lowest = most important; highest = least important).

Seventy-nine per cent (34/43) of participating panellists responded to the first questionnaire (see Table 2). Of the 34 responding to the round one questionnaire, five were bereaved parents, and one was a young adult who had been bereaved in childhood. Response rates varied from 47% (n=16/34) for question (Q) six, ‘…is there anything else you would like to add?’ to 100% (n=34/34) for Q1 seeking panellists’ opinions on the challenges for N&SWs ‘…when establishing relationships and in connecting with families about how to support children aged under eighteen years, before a parent dies.

Responses to round one of the survey generated five categories related to issues and priorities for N&SWs and gave rise to 27 statements (items) for rating by panellists in round two.

Round two of the survey attained an 84% (36/43) response rate. Five bereaved parents and one young adult participated in this round. Table 5 presents the results of panellists’ ratings in response to the round two survey.

A high-level agreement (agree, Mdn. =4) was reached for each of the issues rated in round two. The extent of agreement was strong (MADM.= <1) for each issue rated. Each priority (S5.1-5.7) rated in round two scored the maximum (strongly agree, Mdn. =5).

Table 6 shows the (round three) results from panellists' ratings of the same issues for N&SWs as presented in round two.

The same five bereaved parents and one young adult participated in this round. The response rate for round three was 86% (31/36). The level of agreement remained high for 18/20 (90%) of the issues rated in round three. The MADM. remained the same or decreased, between rounds two and three, for 15/20 (75%) of these issues. Ninety per cent (28/31) of the panellists responding to the round three questionnaire ranked the priorities for enhancing the provision of support to families and their children by N&SWs before, and after, the death of a parent. Notably, because each priority (S5.1-5.7) rated in round two scored the maximum level of agreement (Mdn. =5), rather than rating these again (as consensus was clear) in round three, panellists were asked to place these priorities in their preferred order (see Table 7).

The inverse sum of weighted averages for each priority ranged widely from 57 for the highest-ranked—(S5.1) ‘Training in opening conversations with families about children’s needs, before the death of a parent’, to 167 for the lowest—(S5.6) ‘Increasing knowledge of existing sources of information (written, online and audio-visual materials) to help them to support families, after the death of a parent.’

This survey is the first to measure consensus on the issues and priorities for N&SWs in assisting families to support children through parental death. Arguably, a lack of confidence and competence is fuelled by insufficient access to appropriate training and resources to equip N&SWs with knowledge and skills to ask about the presence of dependent children.

There was consensus regarding the importance of supporting N&SWs in managing their own and patients’ family members’ strong emotions when preparing families for bereavement. We suggest findings could inform the development of national as well as local policies, interventions, and mechanisms for supporting N&SWs to connect and engage with family members when a parent has a life-limiting illness and dependent children. However, based on our findings, there is a need for further research specifying which N&SWs, i.e., those working in oncology, palliative care, or as generalists, need training, supervision and support, within their work settings (hospitals, hospices, and the community), and the form it should take. Moreover, there is a need to identify the commonalities and differences in issues faced by N&SWs.

Delphi surveys are useful in helping topic-experts to engage objectively with the views of others, to reflexively nuance their opinions and, if appropriate, change these in response to the group average levels of agreement [35]. However, using a Delphi survey to gather qualitative data, as in round one of this survey, leaves the technique open to accusations of bias [35]. Arguably, the results of a Delphi survey are only as valid as the expert panel and the questions they are asked. There is debate over what constitutes a topic expert [38], and panel selection can influence the rigour of findings [39]. In our research, we used the definition by Keeney, Hasson, and McKenna. [42] of topic experts as informed individuals, specialists in the field or those with subject-specific knowledge. We selected the panellists from different health and social care disciplines in the acute and hospice sectors. Further, we measured the views of professionals and service users together. Consensus was achieved regarding 100% of the priorities. However, findings might have differed had we, as in Cox et al. [27] analysed service users’ views separately. Maybe doing so would have achieved more nuanced results.

We believe that selecting a heterogeneous sample supported the development of priorities in this Delphi survey. We included young adults because we were mindful of previous research findings identifying that bereaved children’s opinions are poorly represented [65]. However, only three young adults were recruited, and one participated, meaning their perspectives were underrepresented. We had thought that presenting young adults with a survey questionnaire, rather than asking them to participate in individual interviews, would be time efficient as they were likely to have busy lives. This may not have been the case. Their low engagement may also be an artefact of the questions in round one, and statements in subsequent rounds, being insufficiently relevant to their experiences. However, we tried to ensure this was not the case. Importantly, the specialist support charity lead involved in supporting the young adults reported that none were distressed by their participation. So, this appeared not to be a factor in their lower response rate.