Shepherding parents to prepare for end-of-life decision-making: a critical phenomenological study of the communication approach of paediatricians caring for children with life-limiting conditions in Australia


Study design

A critical phenomenological study design,28 involving individual semistructured, post-simulation interviews, was chosen to elicit richly detailed data. Interviews were conducted following a simulation with both paediatricians and the two medically trained actors who participated in the simulation. Critical phenomenology was particularly well suited for this study, because how we communicate with others about death is influenced by our cultural attitudes and the meanings we assign to their reactions.29 Critical phenomenology maps out how first-person and third-person accounts of experiences are possible, analyses their function and identifies new possibilities that can influence development of meaning of an experience.28 Further detail of the rationale underpinning the research question with the choice of methodology and research design has been published elsewhere.30 Paediatricians who care for children with life-limiting neurodisability in Victoria (Australia) were eligible to participate. Purposive sampling recruited paediatricians across acute intensive care and long-term care outpatient settings.

Clinical simulation

The simulation prior to the interview was matched to the clinical context of the participating paediatrician. The simulated scenario was either: (1) meeting with anxious parents of a child with life-limiting neurodisability currently admitted to the intensive care unit (ICU) with their first respiratory deterioration requiring non-invasive ventilatory support, or (2) the first outpatient clinic review two weeks following discharge from the ICU admission. Two internationally based paediatricians confirmed that each scenario portrayed a child sufficiently vulnerable to a shortened life that any paediatrician would consider initiating discussions to share clinical concern with the child’s parents but not feel compelled to reorientate the goals of care for the child to only comfort at this time. Each simulation was pilot tested by a relevant subject matter expert (neurodevelopmental paediatrician and paediatric intensivist). The same two highly experienced medically trained actors were involved in all simulations across the study to allow for comparison of different communication approaches: they played the role of parents in each simulation and were interviewed together after each simulation drawing on their expertise in communication.

Patient and public involvement

The characters played by the medically trained actors were developed during a full-day workshop, in which five bereaved parents of children with severe neurodisability shared their experiences with the actors. Further details of this workshop, including the attention to ethical considerations when involving bereaved parents in research design, have been published elsewhere.31

Data collection

The study was conducted between April and May 2021. Data were collected from semistructured post-simulation interviews with paediatricians and the parent-actors. Each 30-minute simulation was directly observed by the research team (SV, JO, JH, LG). Interview guides were developed by the research team (online supplemental file 1), and further refined and particularised based on the research team’s observation of each simulation. Paediatrician interviews, conducted by SV (a paediatric palliative care physician and trained qualitative researcher, who was known to all participants), were of 30 to 60 minutes in duration. Actor interviews, ranging from 25 to 45 minutes in duration, were conducted following each simulation by JO (a bereaved mother, clinical nurse consultant involved in the care of children with neurodisability and trained postdoctoral qualitative researcher). The interviews were undertaken approximately 15 to 20 minutes after each simulation. All interviews were audio-recorded, professionally transcribed verbatim and checked.

Supplemental material

Data analysis

After completion of all data collection, thematic analysis was undertaken. Figure 1 illustrates the process of data analysis conducted by SV, JO, JH (paediatric palliative care physician and postdoctoral qualitative researcher) and LG (clinical ethicist and postdoctoral qualitative researcher). Transcripts from the paediatrician interviews were analysed first to identify the paediatricians’ experiences. The final agreed coding schema for the paediatrician data set informed but did not determine coding of the actor transcripts. The actor data set was analysed independently. Once the coding of the two data sets was finalised, codes were grouped together across the data sets, and common themes in relation to the research objectives were identified.

Figure 1
Figure 1

Process of thematic analysis.

Rigour was maintained in several ways: triangulation of coding between paediatrician and actor transcripts; prolonged engagement32 with the data over 11 months; contemporaneous notes on the robust, critical discussions in the interdisciplinary research team meetings; and attention paid to reflexivity (using field notes and regular debriefing during the analytical process). Data were managed using both hardcopy and electronic NVivo files.33


A number of important findings were identified in this study, including variation in the approaches and perceptions of paediatricians, and their feelings of not always being in control. This discussion, however, will focus on articulating the construct that draws together the common themes we have identified in paediatricians’ description of their practices. We propose the term ‘shepherding’ to characterise this. The construct of shepherding explains how paediatricians enact their aim to lead parents and their preference to control discussions without being overtly directive. Shepherding involves finding a balance between being too subtle and too explicit in the words carrying guidance or direction. Finding the balance between too subtle and too explicit is a general communication challenge, but is brought into sharp relief in this setting. Our results suggest that paediatricians aim to gently lead parents on a pathway of reflective discussions to help coach them in how to think about their child’s situation before EOLDM occurs. Although not overt about it, they do this to help parents develop an understanding of their child’s quality of life, suffering, risk of death, and the concept of risks and benefits of treatments. In so doing, paediatricians can advance parental thinking about emergent decisions concerning life-prolonging treatments while normalising decisions not to pursue such treatments. This reflective process may enable parents, over time, to cognitively reach a place where they can make the treatment decisions that paediatricians anticipate will need to be made. Later in this process, paediatricians provide parents with their rationale for treatment recommendations, hoping this might assist decision-making when circumstances become more time pressured and emotionally charged. To our knowledge, shepherding has not previously been described in the literature, although its strategies are akin to motivational interviewing. However, unlike motivational interviewing, which is a communication intervention targeted towards behavioural lifestyle change,34 35 shepherding is a construct of what occurs in the practice of the paediatricians who participated in this study. Shepherding is significant as it intends to influence parents by framing information and guiding parental understanding in advance of decision-making. In what follows, we will consider how this newly identified preparatory approach differs from previously recognised forms of physician influence in decision-making and discuss its ethical status and implications.

Paediatrician influence in therapeutic alliances is well described.36 37 This influence is essential to avoid a purely transactional therapeutic relationship.38 Beauchamp and Childress39 recognised three primary categories of influence existing along a spectrum of increasing ethical permissibility, running from clearly unethical coercion (using threats), through manipulation (managing information to influence decisions covertly) and then to ethically appropriate persuasion (giving explicit reasons for a particular treatment decision). Where does shepherding fit on this spectrum? Shepherding is not coercive as it does not involve paediatricians threatening parents to control decision-making.39 It also differs from persuasion, as it is more subtle than explicit and starts well before the discussion and reasoning that occurs when paediatricians provide specific treatment recommendations.27 39–41 Notably, shepherding is more delicate and less specific than the persuasive communication strategies identified by Popejoy et al,27 which include overt discussion of treatment choices preferred by paediatricians, early and repeated emphasis on how such choices would be in the child’s best interests, and, at times, explicit discussion of the limits of parental decisional responsibility.27 In contrast, shepherding strategies identified in this study include pacing of reflection, encouragement of thinking about quality of life and discussions over time. No specific treatment decision is referred to.

This leaves the possibility that shepherding could be considered manipulation. Manipulation is the deliberate management of information to alter patients’ understanding of a situation with the intent to influence decision-making.39 One major way in which manipulation is distinguished from persuasion is that it operates under the surface, rather than in the open. The intention of shepherding, as preparation for more overt EOLDM discussions in the child’s future, is under the surface, in that it is not necessarily made clear to parents when commenced. A second feature of manipulation is that it is achieved by imposing physicians’ values paternalistically, leaving patients ignorant of other values or options. Manipulation therefore undermines patient autonomy.38 We believe shepherding is different because its strategies leave space for parents’ own thinking. Shepherding does not guarantee parents will think in the way that paediatricians hope they will or make treatment decisions that paediatricians prefer. Furthermore, parents maintain influence around the pace and content of the conversation, and may react to information differently than expected and have their own agenda.42 Shepherding, we suggest, is not ethically problematic because parents retain control in assigning meaning and weight to the concepts and concerns that paediatricians raise.

We argue in fact that shepherding is an important and ethically legitimate part of the paediatrician’s role, if done well. It involves assessing the child’s situation, raising parental awareness of important considerations and then listening to the value parents place on these considerations. This is a way of enhancing SDM and is analogous to the underlying premise of the deliberative physician–patient decision-making model described by Emanuel and Emanuel.38 In this model, physicians assume responsibility for explicitly raising the patient’s awareness of the weight that the physician places on the specific values that ultimately contribute to what they see as the preferred course of action. This tends to occur at the time of decision-making. Shepherding has the same ethos but is more subtle and commences well before a decision-making point is reached. Regardless of what parents choose, the practice of shepherding can enhance parental autonomy by preparing them to make the best possible decisions for their child and family, through a process of carefully paced reflection, thinking and provision of information.43 44

The lack of explicitness in shepherding may raise ethical concerns, which need to be addressed. Parental preferences for communication at this time are well-known; they prefer active collaboration,9 21 22 45 and contextualisation to their child and family circumstance.46 47 Even though shepherding is not out in the open, we suggest that it can still satisfy these preferences by facilitating shared responsibility for decisions made. If paediatricians were more explicit in the practice of shepherding, this may risk emphasising the hierarchy within the therapeutic alliance and create an impression that they are telling parents what to do. The opposite risk in shepherding is that paediatricians may be too subtle, such that parents do not hear what they are obliquely trying to convey. This may hinder a parent’s ability to make decisions aligned with their own values. This complexity is reflected in the variations in parent-actor reactions to explicit or non-explicit communication approaches.

There is a further tension between candour and subtlety, with the latter potentially being misconceived by paediatricians as a method to preserve parental hope. This may, in part, explain the delicacy of shepherding and worries about taking parents ‘too far’. Paediatricians may use vague language when discussing death,48 citing concerns about taking away hope.48–50 A hopeful approach is valued by many parents and is perceived as an essential part of a paediatrician’s ability to relate to, and support, parents’ feelings.50 However, hope and clarity of information are not mutually exclusive. It has been found that clear prognostic disclosure may not compromise parental hope51; it may actually enhance it.52

Dealing with the risks of subtlety is where courage comes in: paediatricians need courage to face parents’ and their own emotions, their own fears about being incorrect in their prognostication and their concerns about potential conflict with parents. Some paediatricians in this study worried that without courage, they would remain in a superficial therapeutic alliance with parents by avoiding meaningful substantive discussions around EOLDM. Janvier et al53 earlier identified this need for courage in communication: to discuss death and to provide treatment recommendations at times of clinical uncertainty.

The key clinical implication of recognising shepherding related to how clinicians think about their practice is it provides a shared language for clinicians to critically reflect, both individually and with colleagues, on the challenges and subtleties of preparing parents for EOLDM. It also provides a frame of reference for further research to look at variations in practice and practical and ethical challenges. One important ethical question requiring further attention is whether there is any material or ethical difference between shepherding parents with the intention of optimising their process of decision-making (wherever it leads), and shepherding with the intention of steering them towards a particular decision. The latter would bring shepherding closer to being a form of ethically questionable manipulation. The shepherding strategies described in this study could be used for either intention. Other important questions here are: what do paediatricians think is their goal? and how do parents perceive this process? Our results do not provide clarity on these issues, so further study is needed to facilitate ethical debate about this aspect of shepherding.


We used simulation to prime interviews to study this communication because of our assessment of the ethical and logistical challenges of observing this communication in real life. Over the course of many years of medical care, it is difficult to predict when these communication events will occur, and it would be a significant intrusion into the privacy of the doctor–parent relationship to audio record a large number of consultations. Feasibility was an additional concern. While there are inherent differences between reality and simulation, we believe our approach has yielded meaningful and trustworthy data. Simulation has been successfully used previously in studies of communication practices11 54–56 and research involving qualitative inquiry.57–60 There is good evidence that clinicians find simulations believable enough to elicit realistic communication from them.54 61 It might appear that the actors’ lack of personal experience of EOLDM is a limitation. However, we did not interview them as proxy parents, but rather as experts in communication. The actors were able to provide observations and reflections based on their many years of experience in clinical communication skills training.

We acknowledge that paediatricians’ responses in the interviews may have been influenced by their familiarity with SV. While this may have prompted them to frame their responses to suit perceived expectations of the researchers, it most likely also promoted trust, enabling clinicians to be franker than with an interviewer not known to them. Finally, caution must be exercised when considering the generalisability of these findings. Given this study was conducted in a single sociolegal jurisdiction, we are uncertain of the extent of influence the sociolegal context of clinical practice has on these research findings. Applicability to other settings needs to be assessed rather than assumed.

This post was originally published on