“She’s gone now.” A mixed methods analysis of the experiences and perceptions around the deaths of children who died unexpectedly in health care facilities in Cape Town, South Africa

Purpose The sudden death of a child is a catastrophic event for both the family and the healthcare workers involved. Confidential enquiries provide a biomedical depiction of the processes and quality of care delivered and drive improvements in care. However, these rarely include an assessment of the patient/caregiver experience which is increasingly regarded as a key measure of quality of care. Methods A parallel convergent mixed methods design was used to compare and contrast medically-assessed clinical quality of care with caregiver perceptions of quality and care in a cohort of sudden childhood deaths in emergency facilities in Cape Town, South Africa. Results Amongst the 29 sudden childhood deaths, clinical quality of care was assessed as poor in 11 (38%) and the death was considered avoidable or potentially avoidable in 16 (55%). The main themes identified from the caregivers were their perception of the quality of care delivered (driven by perceived healthcare worker effort, empathy and promptness), the way the family was dealt with during the final resuscitation, and communications at the time of and after the death. Ten (35%) caregivers were predominantly negative about the care delivered, of whom four received fair clinical quality of care; 13 (49%) of caregivers had predominantly positive experiences, one of whom received poor clinical quality of care. Conclusions Caregivers’ experiences of the healthcare service around their child’s death are influenced largely by the way healthcare workers communicate with them, as well as the perceived clinical effort. This is not always concordant with the clinically assessed quality of care. Simple interventions such as protocols and education of healthcare workers in dealing with families of a dying or deceased child could improve families’ experiences at a time when they are most vulnerable.

Introduction Despite significant improvements in the last decade, South Africa failed to meet the Millennium Development Goals relating to neonatal and under-5 mortality [1]. Neonatal conditions, respiratory tract infection, gastro-enteritis and malnutrition continue to account for the majority of deaths, with HIV being an important co-morbidity. In older children, trauma and unnatural causes of death are also common [2]. Confidential enquiries are a well-established method to investigate child deaths, and have been used in South Africa since the late 1990's to identify modifiable factors that contributed to death [3]. This methodology has been shown to improve outcomes in high as well as low-resource settings [4,5]. However, these methods typically focus on a biomedical approach to assessment of quality of care and avoidability of adverse events, with findings commenting on cause of death, age group, and points of delay from the perspective of healthcare workers. Generally, such methods do not attempt to assess the experience or perceptions of care of the family of the deceased.
Increasingly the quality delivered by a healthcare system is understood to include the patient experience, along with patient safety and clinical effectiveness [6]. A systematic review of the links between these quality measures, all in high income countries, showed consistent positive associations-meaning that we need to take patient experience seriously [7]. Reports of patient experience in lower income settings such as South Africa are scarce, and largely limited to specific long term treatment outcomes and satisfaction with TB and HIV management [8,9], yet the patient experience is well recognized as a priority [10,11].
Understanding and monitoring the determinants of patient experience is increasingly a focus of quality of care assessment [12]. Such assessment is particularly important at times when patients and their families are most vulnerable. When children die, it becomes important to understand the experience of their families, and particularly those of the caregivers (usually the parents). Priorities of parents during end-of-life care from a US study [13] were shown to include regular and honest communication, supportive staff and preserving the parent-child relationship. However, their experiences are also difficult to capture: the final resuscitation, the management of the death itself, and management of the family is emotive. Death in the emergency context is usually sudden and unexpected, making evaluation of the experience very difficult; further, the experience of death (particularly that of a child) is a personal, and potentially very private experience so accessing that information can only be done with great sensitivity and with the trust of the people concerned.
As part of a larger study involving caregivers of critically ill and injured children in Cape Town, South Africa [14], interviews were conducted with caregivers whose children died in healthcare facilities; together with the medical records of these deaths, we are able to put together a detailed analysis of the care during the child's death. The aim of this study was to identify practices which influenced the caregivers' experience and satisfaction with the care of their child around the time of death, as well as to compare caregivers' experience with medical assessment of care quality.

Setting and participants
This study used mixed methods to evaluate the quality of care and the experiences of healthcare of the caregivers of a cohort of children who died in health facilities in the Cape Town Metro area. Data were drawn from the Pathways to Care Project, a study conducted from the main tertiary paediatric referral hospital in Cape Town, the Red Cross War Memorial Children's Hospital (RCWMCH) over 12 months from November 2011 to October 2012 [14]. This study used an extension of the child death review (confidential enquiry) concept which is well described elsewhere [1,2,15,16]. Cases of critically ill and injured children under 13 years old were reviewed to uncover the entire health care referral pathway for their critical event, by collecting health records from every facility involved in the acute care pathway, interviewing the caregiver of each child, and conducting an expert review for each child. Study participants were recruited from two sources: the majority were critically ill and injured children admitted to the RCWMCH Paediatric Intensive Care Unit (PICU) with an emergency healthcare episode-these children were sampled by alternate week recruitment throughout the year-long study; the second group were children who died suddenly in healthcare facilities in the RCWMCH metropolitan drainage area, or who died in RCWMCH Emergency Centre. This study has been well described in two publications, one focused primarily on the quality of clinical care along the care pathway, looking specifically for modifiable factors and avoidable events [14], and the other a qualitative paper with a focus on the experiences of caregivers in seeking care for their children [17].
Neither of the prior publications focused on the cohort of children who died within the healthcare services prior to PICU admission, which is the focus of this paper. We excluded children without a full interview with the caregivers, those with chronic medical conditions, those admitted to any hospital for more than 5 days prior to death, deaths following a palliative care decision, neonates direct from obstetric services, and those without signs of life on arrival at the healthcare facility. All eligible deaths documented over the year-long period were included. Semi-structured in-depth interviews were conducted by one of three trained interviewers within two weeks of the death of the child. All interviews were conducted in the home language of each caregiver, recorded, transcribed and then translated into English by the interviewer. An interview guide developed by the authors and refined after a pilot of 20 caregivers, was used to draw details on the referral pathway, timeline and the caregivers' perceptions and observations on the care received (S1 Table), described in detail by Jones et al. [17]. All data were collected with the intention of using mixed methods to analyse and understand the two datasets together-representing essentially a clinical, consensus-based assessment, as well as the caregivers' observations and perceptions of the care. Mixed methods analysis enables deeper understanding than qualitative or quantitative analysis alone, and is a useful tool to enhance the value of a dataset, and to explore deeper issues around a phenomenon such as patient experience [18].

Data analysis
Quantitative analysis. Quantitative data included quality of care (QOC) as assessed by consensus of an expert panel, following detailed review of the medical records and caregiver interview for each child. The expert panel comprised a paediatric intensivist, emergency physician and a primary health care expert, all senior clinicians in this health system, but none had dealt directly with the patients described. The primary assessments of QOC were: a) the avoidability of death by interventions in the final acute care pathway (avoidable, potentially avoidable or not avoidable); and b) global QOC across the referral pathway (good, fair or poor) which looked at the overall care delivered in the acute pathway (this often involved several facilities/ambulances, with potentially variable individually assessed QOC). These assessments were primarily made from a clinician's point of view, although with the insights of the caregivers' experience, and represent a predominantly biomedical assessment of the QOC. Further details are published elsewhere [14].
Qualitative analysis. Qualitative data analysis was based on grounded theory using an inductive approach to the analysis. Thematic analysis of the caregiver interviews was conducted. Two researchers (PH & JP) read all the transcripts for children who had died prior to PICU, and together developed a coding scheme. Both researchers coded all the data, and checked one-another, deliberating any contentious issues. Both PH and JP are clinicians with experience in the Cape Town healthcare system (one with in depth insights to the study and data collection (PH), and the other naïve to the quantitative outcomes (JP)). The coding scheme and emergent themes were discussed with other authors (AA & SR), reassessed and simplified into several coding matrices to represent the [19] key issues identified. Cases were searched for confirmatory as well as outlier material. Following immersion and analysis of each interview, an overall judgement was made by the two primary investigators (PH & JP) as to the caregivers' expressed level of satisfaction with the healthcare delivered to each child (positive, neutral or negative).
Mixed methods analysis. A convergent, parallel mixed methods design was used, with quantitative and qualitative data collected and analysed separately, and findings then triangulated using a mixed methods matrix (Fig 1) [19]. Equal priority was given to qualitative and quantitative data. The matrix allowed themes and categories from both analyses to be visualized and compared for each case, and overall, to integrate the findings, and to search for patterns and meaning.
Ethical considerations. The study was approved by the University of Cape Town Human Research Ethics Committee (HREC 211/2011), by the Oxford Tropical Research Ethics Committee, Oxford University (OXTREC 29-11), and site approvals to collect data were obtained from the Western Cape Department of Health and the City of Cape Town. Caregivers gave written consent to participate and for their interviews to be recorded and transcribed.

Quantitative analysis
Over the 12-month study, there were 29 children who died prior to PICU and all were included. The demographics of these children and their caregivers (mothers were interviewed in all cases, with fathers also present for 2 interviews) are shown in Table 1.
There were 15 deaths at healthcare facilities prior to arrival at RCWMCH, and 14 deaths in the Emergency Centre of RCWMCH ( Table 2). Ages of children at time of death were similar in both groups (mean age of deaths prior to RCWMCH 34 months; IQR 4, 37 and mean age of deaths at RCWMCH 40 months; IQR 8, 70). Infection was the likely aetiology of the deaths in a third of cases (10, 34.5%), and just over a quarter died due to trauma (8,27.6%). The reported delay from onset of illness to first presentation for these cases was less than a day for 19 (65.5%) cases, suggesting a rapid (or instant for trauma) disease progression. More than three quarters presented after hours. All had a brief pathway, with 18 (62.1%) visiting only the facility where they died, and only 12 (41.4%) being transferred by ambulance. There was no clear correlation between child's age at death and any of the outcome measures.
Following an expert panel confidential enquiry approach on each case, it was considered that death was avoidable or potentially avoidable in more than half of cases (16, 55.2%) and global QOC was considered poor in 37.9%. The major modifiable factors identified in this cohort using categories as described in the prior quantitative publication [14], included: (a) initial assessment: inadequate assessment or interpretation of severity, missing key clinical findings; (b) management: resuscitation not done or inadequate for shocked patient, circulatory issues, delay in critical management decisions, referral delay; (c) ambulance services: response time delay). Poor communication with caregivers was also identified as a modifiable factor in many cases.

Qualitative results
Regarding caregivers' experiences of the healthcare system and their satisfaction with the care, three main themes emerged. The themes and subthemes are depicted in Fig 2 and discussed below. An overall assessment of the caregiver impression of the care received showed 13 caregivers to be positive, six neutral, and 10 negative about the care that they had received.  Impression of healthcare worker effort. Caregivers' perception of healthcare worker effort was very mixed: impression of effort seemed related to both what was done (visible procedures e.g. putting up drips, giving antibiotics, connecting to cardiac monitors etc.), the number of people called to work on the child and whether they were taken away from other patients to attend to their child: Promptness/Delays in being attended to or transferred. In general, caregivers were satisfied that children were attended to promptly on arrival. All the trauma cases were seen very rapidly at facilities, and for medical cases generally, these children arrived in extremis and were seen rapidly, with some exceptions. Many caregivers in Cape Town perceive referral to the tertiary paediatric hospital, RCWMCH as the panacea for any critically ill child and often questioned why their child wasn't referred. In most cases the experiences were positive at RCWMCH, but not always.

Whilst still waiting outside I also sneaked and went to peep but my baby was concealed by a curtain and could not see what was happening. I could see that they were inserting drips and there were also some pipes inserted but another man told me to sit down and not peep and I did sit down. [GOOD; NA; Neutral]
Others didn't like needles and so were happy to wait outside, or found it very uncomfortable watching procedures on their child. Cause of death explained/understood:. The cause of death was not well understood in general. Reasons for this were mixed: some were never told; some recalled that they were told but they were too distressed/confused to really take in the information at the time.  Tables 3 and 4, the children of four caregivers who had very negative experiences were actually assessed as having received a fair clinical quality of care. All four were cases with a poor prognosis, regardless of treatment, and it was deemed that nothing else could have been done to save the children, though there were elements of delay and some referral issues noted. However, caregivers felt they weren't seen rapidly enough, and felt that more could and should have been done when healthcare workers decided to terminate the resuscitation. Clearly the process and communication as perceived by the caregiver was inadequate.

Mixed methods results. As shown in
One caregiver seemed to have had a relatively positive experience, yet was assessed as having received poor quality of care. This was an older child with a severe asthma attack, badly treated from a medical perspective, with delays, indecision and errors across two facilities, yet However the majority of quantitative and qualitative assessments were concordant, as were the caregivers' responses when asked if there was anything that could have been done better. There did not seem to be specific correlations to link any assessed theme (quantitative or qualitative) in terms of maternal age, delay to onset, number of steps in the pathway, or even whether caregivers were present at the final resuscitation. There were more dissatisfied caregivers of children with medical illness as opposed to trauma. Mixed methods analysis of the sudden deaths of children in health care facilities

Discussion
Our results show quite clearly that in this setting the caregivers' experience is not necessarily related to the clinically assessed quality of clinical care delivered. We identified the main themes which caregivers related from both positive and negative perspectives and how they seem to have influenced the overall experience of their child's death, regardless of the clinical quality of care. In most cases there was some association between the clinical assessment and the caregiver's experience, in that there were more unavoidable deaths amongst the group with positive perceptions, and more avoidable or potentially avoidable deaths in the group with negative perceptions. But paradoxically in a few cases we see evidence of what medical experts deemed poor care, with an avoidable death, yet the caregiver appeared satisfied with the care they received and had accepted the death. This is perhaps due to some element of perceived care or communication by the healthcare team that was not captured by the research, perhaps a culturally mediated phenomenon, or the result of a complex power dynamic in the patienthealth care worker dynamic. This is in accordance with other research which suggest there is a complex process of developing meaning to grieving and death in mothers, especially in a multi-cultural context such as in Cape Town [20][21][22][23].

Family perception of effort
Triage is an effective, well implemented strategy that has been adopted by most health facilities, as evidenced by the prompt treatment received and the perceptions of caregivers. Many caregivers in this study anticipated (and accepted) delays and waiting as part of the healthcare experience, to the extent that any sense of urgency and rapid care was memorable and remarked upon. In contrast, caregivers seldom mentioned lack of care and urgency (as well as lack of empathy) which perhaps demonstrated their low expectations. The quantitative results did show profound delays (median delay from first presentation to PICU admission 12.3 hours), even unexpected ones such as at the tertiary hospital (median delay from arrival at the tertiary hospital to PICU admission of 5.0 hours), and there is clearly much room for improvement across the system despite the evidence for some well-intentioned individual clinician urgency [14].
There was a marked distinction in different caregivers' acceptance or closure around the death of their children. Some caregivers seemed to accept the death according to their own expectations of whether the injury or illness was survivable, while those with least closure blamed delays and errors in the healthcare system. Although many caregivers desperately wanted referral of their child to a tertiary hospital, and saw this as the preferred outcome, our small sample does not show any particular association between caregiver satisfaction and the children who died at peripheral sites versus RCWMCH.

Family presence
There is substantial research and evidence from other countries looking at the process of dealing with a child and the family of a child who dies suddenly in a healthcare facility, including whether the family is allowed or encouraged to be present during the final resuscitation [24][25][26][27][28][29]. There is sufficient evidence to say that some (most) parents would prefer to be present with their child during resuscitations (as borne out by the majority of caregivers in this study) and this is likely to aid their acceptance and coping with the death. Many healthcare workers are uncomfortable with family presence and discourage it, but our results show that caregivers are clear on their desire and would like the opportunity to be present [26,[28][29][30]. Yet when given the opportunity to stay with their child during invasive procedures, several parents expressed their discomfort at witnessing this, and may in fact have been expressing a felt need to be with their child rather than the reality of observing the resuscitation. This is an important point for healthcare workers who should consider the benefit of parental presence in any resuscitation-not only from a clinical perspective as caregivers often provide crucial clinically beneficial parts of the history, but also given the impact on the caregiver's experience as we can seldom predict when a terminal resuscitation will occur. As other researchers have pointed out, caregivers in this situation need to be supported by a staff member, able to explain the procedures, liaise with the clinicians, and individualize the caregiver's experience [28].

Communicating and supporting caregivers when their child dies
The key needs of families are communication, empathy, and concern which are often missing in burnt out health care professionals [13]. Work by Brysiewicz conducted in a South African setting not unlike that of this study, has suggested a model to deal with sudden death in the emergency unit-encompassing planning, resources and culture before the event, optimal care of the patient and family during the death, and assistance not only to the family after the death, but also to the health care professionals involved [31]. Ganca et al in Cape Town [32] and Hassankhani et al. in Iran [33], point out the specific communications skills required by health care professionals in this situation. Furthermore, cross cultural interactions add additional challenges and require specific skills that are not necessarily part of an emergency clinician's skill set, although desirable to deliver high-quality care [34].
Breaking the news of death to family members is always an uncomfortable and stressful procedure for healthcare workers, and even more so when a child is involved [21,27,35]. While our data shows some examples of good practice, such as regular and open communications with caregivers from a senior clinician, it also highlights that many frontline clinicians avoid formally and professionally breaking the news to family members, and instead break the news informally at the bedside or in a working environment, and do it hurriedly, often using euphemisms and with little further explanation or confirmation of understanding to the family. Although this must be seen in the context of a resource strained healthcare system, with health care workers stretched for time to spend with family during and after a resuscitation, it nevertheless may point to a lack of knowledge and the need for inclusion in training curricula as suggested by a recent survey of healthcare workers in Cape Town [32], as well as the need for clear policies and protocols for such times. Follow up support and communication with grieving parents needs to be part of the continuity of care following such a death, perhaps through community health and support services.
O'Malley et al refer to a "good death" which encompasses acknowledging that not all deaths are avoidable; caring for survivors (by retaining trust, giving them some control and decision making, and allowing presence and "goodbyes"); and having a clear protocol in place [36]. There were no policies in place, even in the tertiary RCWMCH for supporting caregivers. Most deaths occur outside of office hours, and in this system that usually means without the presence of senior clinicians, consultants, or any immediately available counselling service, which needs to be taken into account. Beyond palliative care, and tertiary intensive care settings, the death of a child is a rare event, and most healthcare workers will only occasionally be involved, usually unexpectedly, and it occurs more often after-hours with limited support resources in place. This is a challenge both in terms of educating health care workers, and making specific resources such as senior staff and counsellors available.

Strengths & limitations
This study is unique in including a significant number of sudden and unexpected deaths of children, in a setting where reliable data collection is possible, and in being able to correlate biomedical assessment of the quality of care as well as the caregiver's perception and experience.
Limitations include the differences between the three interviewers and the depth of information they were able to obtain from caregivers, as well as the cultural difference in caregivers' expectations and willingness to share their experiences of the health care system which were not explored and differentiated. All caregivers were interviewed within two weeks of the death of their child, but clearly their bereavement processing and the circumstances around the death may have influenced their recall, and their perception of the entire process is likely clouded by the terminal event. Although just mothers were present for most interviews, the father was also present in two of the interviews which may have affected the mother's responses and this was not analysed, yet did not seem to be the case from the transcripts. The meaning, and some expressions (as well as non-verbal cues) may have been lost in the translation (such as euphemism for death and dying), although having the interviewers translate the transcripts themselves soon after the interview helped to mitigate this. Having the transcripts translated by the interviewer may have introduced bias, but they were all experienced researchers with no involvement in any aspect of the study other than conducting their own interviews. Due to our sampling, there were few neonatal deaths which is not representative-South African literature suggests almost half of childhood deaths are in the neonatal period [1]. And the assessment of the quality of care used was a global one for the entire referral process which may have entailed steps of varying quality, perhaps skewed by the places where major events took place (as would be the case for the caregivers' experience).

Conclusion & recommendations
In this study, which represents a significant number of childhood deaths over a year long period across a spectrum of healthcare facilities and presentations, we show there is considerable discrepancy between different measures of quality of care. 40% of this cohort were assessed to have had clinically poor care and 45% of these deaths were assessed as either avoidable or potentially avoidable. It should therefore be no surprise that around half of the caregivers were unimpressed by the care delivered. However, the drivers behind caregivers' perceptions of quality of care are different from the criteria used to determine clinical quality of care, as highlighted by the discordant cases. Recognising this discrepancy and attending to those features which impact caregivers' experiences could have enormous effect on the families of children who die suddenly, and likely on the clinicians who manage such cases. Relatively simple and high value interventions could be adopted to address many of these, such as better training of healthcare workers in dealing with death and dying in emergency healthcare settings, policies and protocols in place for how to break the news and support bereaved family (including systems that work after-hours), follow up with families after a death to offer ongoing support and counselling and creating spaces in institutions to talk to families. Processes to routinely investigate deaths and consider underlying factors whilst in place, might be expanded to include an evaluation of the family's experience of care, and those factors which most affected their perceptions of the service they received.
Supporting information S1 what were often painful experiences with the research team in the hopes of improving the health service for others.
We also acknowledge the research team who helped to collect and analyse this data, knowing that these interviews were the most difficult and traumatic for the whole team. We hope it was justified by the outcomes.