Differences in Advance Care Planning and Circumstances of Death for Pediatric Patients Who Do and Do Not Receive Palliative Care Consults: A Single-Center Retrospective Review of All Pediatric Deaths from 2012 to 2016

Statistical analysis

Analyses were conducted using Stata 15 software. ¹⁸ Patient demographics, clinical characteristics, and GOC and ACP preferences were explained using descriptive statistics. To understand differences in demographic and clinical characteristics between patients who received PPC consultation and those who did not, Pearson's chi-square tests were performed. Multivariable logistic regression models were developed to examine associations using adjusted odds ratios (ORs) between the outcomes of interest with patient demographic and clinical characteristics, including palliative care status, age at death, diagnosis, primary unit, documentation of religion, RUCA code, and year of death. The outcomes were separated into two categories: ACP (including GOC and POLST) and circumstances of death (including do-not-resuscitate [DNR] code status, hospice involvement, and location of death) as binary measures. A significance level of p < 0.05 was used for all models.

Clinical and demographic characteristics

From 2012 to 2016, 256 of the 558 pediatric deaths (45.9%) received PPC, the remaining 306 (54.1%) did not (Table 1). There was a significant relationship between PPC and non-PPC over time throughout the study period (p = 0.028). In 2012, PPC was involved in 37% of the deaths that occurred that year, which increased to 57% in 2016 (Fig. 2), which was also the first year that PPC was involved in caring for most of the decedents. When examined by clinical and demographic variables, the largest number of deaths occurred among those 0–30 days old, and PPC involvement was highest among this age group. The most common diagnoses in children who did not receive palliative care were trauma (n = 56) and “other” diagnoses (n = 46) (Table 1 and Fig. 3). Conversely, deaths from malignancy and respiratory diseases had the highest PPC utilization. The most common places of care before death were the PICU (n = 150) and NICU (n = 142). Patients whose primary place of care was outpatient or “other” experienced more deaths without PPC, whereas patients whose primary place of care was inpatient had more deaths with PPC involvement (Table 1 and Fig. 4). Overall, 62.2% (n = 347) of patients were from urban areas, 13.3% (n = 74) were from large rural areas, another 13.3% (n = 74) were from small rural areas, and 11.3% (n = 63) were from isolated areas (Table 1). Given hereunder are the results of the multivariable models by each outcome.

OPEN IN VIEWER

FIG. 2.

Decedent patients by year of death from 2012 to 2016 by palliative care status and total.^{a a}Number of decedent patients by year of death from 2012 to 2016; patients were divided into groups based on palliative care versus nonpalliative care status.

OPEN IN VIEWER

FIG. 3.

Patient primary diagnosis by palliative care status.^{a,b,c a}Patients were categorized by primary diagnosis and subdivided into palliative care and nonpalliative care status. ^bCategory “Other” includes endocrine, gastrointestinal, genitourinary, hematologic, multiorgan system failure, renal complications, and unknown diagnoses. ^cStatistically significant at p < 0.05; reference value of neurologic diagnosis.

OPEN IN VIEWER

FIG. 4.

Primary place of care by palliative care status.^{a,b,c a}Primary place of care was identified for all patients; each patient was subdivided into palliative care and nonpalliative care status. ^bCategory “Other” includes emergency department, labor and delivery, home, and unknown primary unit. ^cStatistically significant at p < 0.05; reference value of outpatient unit.

Goals of care

Patients with PPC involvement were significantly associated with having their GOC documented (OR = 96.93), after adjusting for age at death, diagnosis, unit, religion, RUCA, and year of death (Table 2). By age category, those who were 5–14 years of age were less likely to have their GOC documented after adjusting for the remaining variables in the model (5–9 years, OR = 0.21; 10–14 years, OR = 0.27). Patients with cardiac conditions (OR = 0.26), infection (OR = 0.15), and those who experienced trauma (OR = 0.17) as their primary cause of death were also less likely to have their GOC documented.

POLST and code status

The likelihood of having a POLST documented at the time of death increased over the study period. Patients receiving PPC were more likely to have a POLST form (OR = 24.06) or inpatient DNR status (OR = 7.71). In the multivariate model, patients receiving PPC in the nonintensive care inpatient environment (OR = 13.95) and outpatient environment (OR = 16.32) were more likely to have a completed POLST at time of death compared with patients receiving their care in the NICU or PICU (Table 2). Patients with cardiac disease (OR = 0.14), congenital heart disease (OR = 0.15), and infection (OR = 0.13) were less likely to have DNR code status at the time of death compared with patients with neurologic diseases in the multivariate model (Table 2). All patients in the sample who were >30 days old at the time of death were less likely (OR = 0.16–0.37) to have DNR as their code status compared with infants who were <30 days old.

Hospice involvement

Patients who received PPC were more likely to have had hospice care at the time of death (OR = 11.7). Patients who received care in an inpatient (OR = 7.67) or outpatient (OR = 16.36) setting were more likely to have been under the care of hospice at time of death compared with patients receiving their care in the PICU (Table 2). Those treated for malignancy were more likely to have hospice involved with their care (OR = 2.48) compared with those with neurologic conditions. However, patients with respiratory conditions were less likely to have hospice involved with their care before death (OR = 0.13).

Location of death

After adjusting for the clinical and demographic covariates in the model, palliative care patients were more likely (OR = 2.23) to die at home (Table 2). Those patients who died at home were more likely to have had their primary place of care in an inpatient setting (OR = 20.61), outpatient setting (OR = 60.33), or other (OR = 25.93) compared with their counterparts seen in the PICU. Results presented in Figure 5.

OPEN IN VIEWER

FIG. 5.

Location of death by palliative care status.^{a,b,c a}Location of death was identified for all patients; each patient was subdivided into palliative care and nonpalliative care status. ^bCategory “Other” includes inpatient hospice, accident scene, long-term care setting, and unknown location of death. ^cStatistically significant at p < 0.05; reference value of pediatric intensive care unit.

Palliative care involvement

Although there is variation across the United States regarding the role of the PPC provider on the care team, timing of consultation, services provided, and credentials and training of providers, agreement has been reached within the context of comprehensive pediatric care that this specialty fills a gap in care and provides much-needed support to patients and their families. ¹ Variation in the role of PPC on the clinical care team of pediatric patients has been discussed in the literature.^{2–4,19–21} In this study, PPC was provided by a dedicated team of physicians, nurses, and advance practice providers with certification in palliative care, in collaboration with hospital-based services providing spiritual services, social work and care coordination support, music therapy, pharmacy, and child life services. The PPC team was established 2 years before the start of the study; however, there was growth of the team during the study period, which might partially explain the increased proportion of PPC involvement over time.

Deaths from malignancy and respiratory issues had the highest utilization of PPC, possibly because of the nature of those diseases, which are life threatening and whose trajectories have natural opportunities for PPC to be consulted. In contrast, patients with trauma diagnoses were less likely to have PPC involvement, which may be expected because of the acute nature of those conditions. This is also likely the case for the “other” category, which were often either poorly understood, highly uncertain, or rapidly deteriorating, making it difficult for the primary teams involved to decide if and when to consult palliative care. The diversity in disease types represented in the study emphasizes the benefit of early integration of PPC for children with serious illness to avoid “unexpected” deaths of patients, with underlying medical conditions, who could have benefited from PPC support but did not receive it. ²² Many deaths from diagnoses other than trauma would be consistent with what Feudtner et al. ¹⁶ refer to as chronic complex conditions, where early involvement of PPC allows for support of quality of life and ACP that can impact the circumstances at death.

Advance care planning

Facilitating ACP is one of the important tasks of palliative care. Discussing GOC is an early step that helps align the treatment plan with patient and family preferences and priorities. We found that GOC were much more likely to be documented in patients followed by PPC, as it is part of every consult. As may be expected, patients who experienced trauma were less likely to have their GOC documented before death, likely because the goal remained full disease-directed treatment because of the acute nature of their clinical needs. Patients with acute cardiac conditions less frequently had completed POLST documentation compared with those with neurologic diseases, which may also be expected because of the acute versus chronic trajectory of illness. Furthermore, those who received the bulk of their care in an inpatient (non-ICU) or outpatient setting were more likely to have a completed POLST form. These findings are similar to studies by Marcus et al. ¹⁰ and Morell et al., ²³ and perhaps are illustrative of efforts by the PPC team to offer POLST documentation to patients with serious illness before discharge or in the outpatient setting.

Circumstances of death

Patients <31 days of age were more likely than all other age groups to have a code status of DNR at death, after adjusting for the other variables in the model. This may be because of perinatal counseling that allowed some families to choose comfort care for their infants with life-threatening illness, and a change in code status if treatment was initially focused on life prolongation but later redirected to a comfort focus.

Patients with PPC involvement were significantly more likely to have hospice involved and their location of death at home compared with those without PPC involved in their care. Patients with PPC involvement were significantly more likely to die at home after adjusting for the remaining covariates in the model. Of note, those treated in non-ICU settings (i.e., inpatient, outpatient, and other) were significantly more likely to die at home after adjusting for PPC and the other covariates. This may be because of logistical challenges in discharging a patient from the ICU in a rural state (e.g., setting up oxygen for transport and use at home). In addition, some of the deaths at home could have been considered “unexpected” even if there was a life-threatening illness present. Furthermore, patients who did not receive PPC may have had hospice support at home or in the hospital before death at the referral of their primary care provider or specialty team. This discrepancy in hospice enrollment and PPC involvement highlights that hospice care does not necessarily equal palliative care, and vice versa. Pediatric patients can, and often do, receive palliative care alongside curative therapies.

Rural disparities

There is little research that has examined PPC delivery and its effect on outcomes, particularly ACP, in rural areas. Although the adult literature shows that rural hospice patients rate their care very highly, ²⁴ it is unknown if the same is true for pediatric patients. The location of a pediatric patient's death has been examined in two studies, with conflicting findings. The first study by Feudtner et al. (2006), reported on patients in Washington state who died between 1989–2002. It indicated that patients who lived far away from a tertiary care center were more likely to die in the hospital. ²⁵ Conversely, in 2015, Jamorabo et al. found that a patient's distance away from the hospital has little, if any, bearing on where they die or the care they receive. ²⁶ Because of these conflicting study results in combination with our findings, additional study of rural disparities for children is needed to understand the rural differential for PPC and how to use the most precise mechanism to measure or define rural areas (e.g., zip code or Census tract level measurements instead of county-based measurement).

The rural nature of a patient's home did influence their location of death. In the multivariable model, patients residing in a large rural (or micropolitan) area were less likely to die at home than their urban counterparts. Given the previous work showing that pediatric patients with chronic disease and life-limiting conditions are shifting where they die,^27–29 these data suggest that further work needs to be carried out to ascertain why patients from urban areas are more likely to die at home than their large rural (or micropolitan) counterparts. This study underscores the importance of future research to address rural disparities through effective policy change.

Limitations

This study is not without limitations. First, this retrospective chart review relies on accuracy and completeness of the medical record; however, documentation may be variable from provider to provider. It is possible that comprehensive discussions of GOC may have occurred but were not documented. Second, because there is no uniform place in the medical record to document GOC conversations, such data can be challenging to extract. The electronic medical record is not necessarily designed or intended with research in mind, and thus the meaning of available data might be unclear. ³⁰ The research team member abstracting data from the medical record was not blinded and this could have biased our results. ³⁰ However, to address potential bias in determining the primary diagnosis category, clarification of the primary diagnosis category was discussed if needed.