The Pediatric Palliative Care Quality Network: Palliative Care Consultation and Patient Outcomes Free

This was a cross-sectional retrospective data analysis of PedPCQN hospital encounters for children aged 0 to 18 years between January 2016 and September 2020.

The PedPCQN was a membership-based, volunteer PPC registry and quality improvement (QI) organization that featured real-time, standardized, patient-level data collection and reporting. PedPCQN was developed as the pediatric platform companion to the national Palliative Care Quality Network (PCQN), which was initiated in 2009.¹⁹  PPC programs affiliated with any of the active PCQN sites were invited to join the PedPCQN, and sites were self-selecting. Pediatric data elements were chosen by a team of PPC experts who considered the characteristics, processes of care, treatment, and patient outcomes that are most relevant to children while promoting ease of collaboration, comparison and benchmarking, and data collection by building off of the existing adult PCQN system.¹⁸  For the registry, data collection was performed manually or via automated abstraction and uploaded from the electronic medical record to the PedPCQN registry and maintained by information technology teams on secure servers at the host university. Staff from participating programs were formally trained by the University of California San Francisco (UCSF) PCQN program manager via remote video conferencing or in person on how to read the PedPCQN data dictionary, how to enter and report data, and how to navigate the registry to ensure data collection consistency across sites. Data quality checks were performed periodically by the UCSF PCQN data management team. The goal of the PedPCQN was not research, but rather to capture real-world practices and to facilitate ongoing QI efforts; therefore, clinicians were encouraged, but not required, to enter data for each of the data elements.

The PedPCQN included standardized data elements across 3 domains: structure (patient and diagnostic characteristics), processes of care outcomes (consult and treatment), and patient-level outcomes (symptom assessments, palliative care needs, goals of care/advance care planning [GOC/ACP], and discharge).¹⁸  We examined characteristics, processes, and outcomes of care, including reasons for referral, source of referral (unit), primary diagnosis associated with PPC referral, hospital admission date, initial PPC consult date, hospital discharge date, ACP (including Physician Orders for Life Sustaining Treatment [POLST]) documented in the chart at initial PPC consult (yes or no), code status at initial consult (full code, partial code, do not resuscitate, or do not intubate), PPC team member involvement by discipline (eg, physician and chaplain were counted as different disciplines), patient age (years), and sex. Notably, GOC/ACP were combined into a single “reason for referral” category in the adult and PedPCQN because, in practice in the hospital setting, they are typically overlapping elements of serious illness communication. We also examined other processes of PPC, including screening for psychosocial needs, spiritual needs, pain symptoms, and non-pain symptoms (yes, no, or unknown) and whether a supportive intervention targeting each identified concern (yes or no) was implemented. Ratings of self-reported, observed, and caregiver proxy pain and non-pain symptoms (none, mild, moderate, or severe) were recorded by the PPC team at initial and subsequent visits.

To gather information about the structure and characteristics of PPC programs in the PedPCQN registry, a survey was emailed to participating sites in April 2022. The survey included questions about location, institution type, number of hospital beds, number of years the PPC program had been in service, number of hospice referrals per month, the approximate percentage of patients receiving state or federal medical assistance, age groups cared for (pediatric only or combined pediatric and adult), PPC service lines (eg, perinatal), service availability (eg, Monday through Friday), monthly patient census, and total PPC staff full-time equivalents (FTE), including physicians nurses, social workers, chaplains, psychologists, and pharmacists.

Analyses were performed by using IBM SPSS version 28.²⁰  To characterize the patients included in this study, age, sex, and primary diagnosis leading to PPC referral were reported at the patient level because they are likely to be stable across encounters and give context about the patient population. Because our primary interest was processes of care received during the hospitalization, irrespective of the patient’s past experience with PPC, we chose to analyze data at the encounter level, which was defined as a unique hospital admission involving PPC referral (ie, a patient who was admitted and referred to PPC 2 times during the analytic window was counted 2 times in the analysis). Because no information was available regarding the number of hospital admissions patients had before or after the analytic timeframe, it is possible that patients with only 1 admission had previous or subsequent admissions outside of the analytic window. A sub-analysis was conducted to determine if there were clinically meaningful differences between children with <3 versus ≥3 admissions at the first admission. We found that children with <3 admissions were more likely to be referred to PPC by a neonatal critical care unit (76/387, 20% vs 5/85, 6%, P < .01), more likely to be <1 year of age (239/663, 36% vs 28/141, 20%, P < .001), and more likely to die before discharge (155/662, 23% vs 0/141, 0%, P < .001). Descriptive analyses were run for all data elements (percentages, means, medians, SD, interquartile ranges [IQR], and ranges [shown as minimum-maximum values]). The changes in symptom scores were calculated from the first to second assessment for patients reporting moderate or severe symptoms at the initial assessment who also received their second assessment within 72 hours. A clinically meaningful change in the rating of symptom distress was defined as a change by 1 category (eg, severe to moderate). Unsuccessful attempts to assess symptoms (eg, patient was sleeping) were documented by clinicians, accounting for 27% to 48% of missing symptom scores at T1.

We analyzed differences between children who died and those who were discharged alive for days from hospitalization to PPC referral, PPC consultation, and discharge using Mann-Whitney U tests. If assessments were attempted but PPC clinicians were unable to assess symptoms for any reason, they documented it, and the data were counted as missing in the analysis. No adjustments or imputations were made for missing data, resulting in different Ns for each analysis. Notably, in 2019 PCQN and PedPCQN joined with other palliative care registries to form the national Palliative Care Quality Collaborative (PCQC). The PCQN was institutional review board-approved.

Five PPC programs participated in PedPCQN and were included in the analysis (Table 1). Programs were located across the US, including states in the West (2), Midwest (1), South (1), and East (1). Four (80%) programs cared for children only. Programs contributed between 152 and 663 encounters, except for the combined adult-pediatric palliative care program located in a nondedicated children’s hospital, which contributed only 21. All were located within teaching hospitals and were predominantly nonprofit (n = 4, 80%). Hospitals varied in size, from 84 to 309 inpatient beds, and all offered multispecialty pediatric services (eg, hematology-oncology, neurology). As of 2022, the average number of years PPC programs were in service was 13 (SD = 4.8, range 6–17). The mean total number of FTE PPC team members (all disciplines) was 5.8 (SD = 2.0, range 3.0–7.6). All PPC programs had at least 1 physician (mean = 1.7, SD = 0.8) and nurse FTE (registered nurse [RN], clinical nurse specialist [CNS], or advanced practice registered nurse [APRN]) (mean = 2.1, SD = 1.0) on staff. All programs had a dedicated chaplain or psychosocial clinician on staff (mean FTE = 1.4, SD = 1.0). Programs cared for a median of 35 (IQR 20–85, range 15–110) patients per month.

There were 1587 hospital encounters with a referral for PPC services, 1517 (96%) of which resulted in a PPC consult. Reasons provided by PPC teams for not completing an initial consult included discharge before being seen (n = 14), parents being unavailable (n = 3), canceled referral request (n = 1), being transferred to a different hospital (n = 1), parent refusal (n = 1), patient death (n = 2), and reason not recorded (n = 48). Patient-level coding was available for 1440/1587 (90.7%) encounters, which made it possible to identify how many hospital encounters each patient had during the analytic window. Of the 1440 encounters with this information, 804 were of unique patients, most of whom had only 1 (n = 562, 69.9%) or 2 (n = 101, 12.6%) encounters.

More than one-half of patients were male (n = 446/804, 55.5%; Table 2), with a median age of 4.0 years (IQR 0.42–12, range 0–18; Fig 1). At the encounter level, the top 5 primary diagnoses were solid tumor cancer (28.6%), congenital or chromosomal disease (13.9%), neurologic conditions (12.2%), pulmonary disease (10.2%), and complex chronic conditions (ie, several diseases without 1 primary condition; 9.3%). The median number of days from hospitalization to PPC consult was 2.0 (IQR 1–8, range 0–303). The reasons given for requesting a PPC consultation by referring providers varied, and more than one-half were for multiple reasons (n = 882, 60.3%). The top reasons included to provide support for the patient and family (ie, the patient or family needed additional information, advocacy, or support; n = 1166, 77.6%), to facilitate GOC/ACP discussions (n = 630, 41.9%), to address non-pain symptoms (n = 508, 33.8%), and to manage pain (n = 318, 21.2%). An advance directive or POLST was documented in the medical record for 229 (14.4%) encounters at some point during the PPC consultation period (Table 2). Referral locations generally mirrored top primary diagnoses (eg, solid tumor cancer was the most common primary diagnosis and hematology-oncology units were the most common referring locations). The total hospital length of stay was a median of 12 days (IQR 4–36, range 0–597), of which a median of 7 days (IQR 2–23, range 0–574) were spent being followed by the PPC team.

At the end of the encounter period, children representing 1395 (88.0%) encounters were alive. Among children who died in the hospital before discharge (n = 191 or 12.0% of encounters), the top referral diagnoses were solid tumor cancer (n = 38, 21.1%), congenital or chromosomal conditions (n = 25, 13.9%), and vascular disease (n = 25, 13.9%). Compared with children discharged alive, children who died had longer hospital length of stay (median = 24.0 vs 11.0 days, P < .001). These patients spent more time in the hospital before PPC referral (median = 6.0 vs 2.0 days, P < .001) and received PPC services longer (median = 10.0 vs 6.5 days, P < .05). Among the children who died, the most common referral settings were the pediatric critical care unit (n = 34, 29.8%) or neonatal critical care unit (n = 31, 27.2%).

Care was received from clinicians representing 3 or more different disciplines less than one-half of the time during the PPC consultation period (n = 611, 47.5%; Table 3). Moreover, more than one-half (n = 675, 52.4%) of patient encounters involved care from only 1 or 2 different disciplines throughout the consulting period. PPC teams screened patients across 5 core PPC domains of care, and the most common needs identified were psychosocial support (n = 1323, 92.7%), care for non-pain symptoms (n = 1147, 80.2%), and GOC/ACP (n = 903, 71.4%). For each domain, patients who screened positive during a PPC assessment received a targeted intervention at least 90% of the time. The screening needs identified by PPC clinicians were generally in line with the primary reasons given for referral by referring providers.

The median total number of symptom assessments per encounter was 2 (IQR 1–5, range 0–133). The median time between the initial and second assessment by the PPC team was 1 day (IQR 1–4, range 1–78). At the initial assessment, moderate-severe symptoms were reported for feeding issues (n = 301, 36.8%), followed by fatigue (n = 200, 30.5%), dyspnea (n = 261, 28%), and pain (n = 201, 20.4%). For those with moderate-severe symptom scores who also had a second assessment within 72 hours, there was an improvement in 67/94 (71%) children who reported pain, 54/106 (51%) who reported dyspnea, 33/72 (46%) who reported fatigue, and 51/132 (39%) who reported feeding issues (Fig 2).

This multisite retrospective study of >1500 PPC encounters across 5 PPC programs in the United States provides new insights into the current inpatient PPC referral processes, interdisciplinary PPC team involvement in care delivery, processes of care, and impact on patient outcomes among children with serious illness and their families. Although more than two-thirds of patients had noncancer diagnoses, the largest single category of referrals to PPC by location and diagnosis were from oncology units for children with solid tumor cancer diagnoses. Although more than one-half of referrals were made for multiple reasons, the majority of PPC referrals were made by referring providers for psychosocial support for the patient and family, which was also the most common PPC need identified by PPC clinicians during their screening at the initial PPC consultation. The same alignment was found for GOC/ACP discussions and non-pain symptom management, which were the second and third most common. Needs identified at the time of screening were intervened on 90% of the time or more across all PPC domains, and for the subset of patients with data, improvements were seen for pain and the top 3 non-pain symptoms by the second assessment.

We found that the psychosocial needs of children and their families were highly prevalent. Despite the need, approximately one-half (52%) of encounters involved only 1 or 2 different PPC disciplines, which is concerning given the need for interdisciplinary care to address complex needs. To adequately meet psychosocial needs, there should be mechanisms in place to specifically focus on issues that are important to patients and families, such as communication, direct caregiving, and emotional needs.²¹  In addition, a sizeable minority of children saw only 1 PPC team member (22%) during the consultation period, which is at odds with national guidelines.^15,22  These data suggest that PPC teams may be understaffed because appropriately staffed teams would include a more diverse range of medical and psychosocial providers throughout the patient’s hospital stay. This aligns with a recent national survey of 54 PPC teams in the United States that revealed that only 37% of interdisciplinary teams met staffing standards, and as a result, 60% of teams were unable to meet clinical demand.^12,15  Similarly, a state-wide survey of PPC programs in California revealed that pediatric hospitals cared for an average of 168 children per year with only 1.8 FTE staff dedicated to PPC.²³  A challenge commonly faced by hospital-based PPC programs is that, although they are highly valued by their organizations, they are primarily funded through philanthropy and restricted hospital budgets, making it challenging for programs to retain interdisciplinary teams and sustain programs able to meet inpatient care demands, particularly around nonbillable services such as psychosocial supports.^12,13,24  This has important policy implications as the PPC community and its institutional and community partners work to determine how best to advocate with federal, state, local, and institutional leadership to meet the needs of children and their families.

In line with existing literature, our study findings suggest a high prevalence of moderately to severely distressing symptoms in children receiving PPC.^25–28  Although pain symptoms were often reported, other non-pain symptoms, such as feeding issues, dyspnea, and fatigue, were common. The presence of multiple symptoms is similar to findings in the PPC literature during the past decade for children with serious illness and palliative care needs.^26,27,29,30  A recent study of parent-rated symptoms in children with complex chronic conditions revealed substantial polysymptomatology,²⁵  with children having a median of 7 symptoms, the most prevalent being pain, lack of energy, irritability, drowsiness, and shortness of breath. Multifaceted symptoms require multimodal treatment. Yet multimodal care provided by disciplines including massage therapy, acupuncture, psychology, physical therapy, pharmacy, and grief counseling is often limited, despite the importance of these services and the growing evidence of their benefits.^31,32  Clearly, there is a need for focused attention on pain and non-pain symptoms in ways that acknowledge both pharmacologic and nonpharmacologic treatments, as well as the complex interplay of physical and psychosocial needs.

There are important limitations to this study. First, the PedPCQN registry did not capture functional or disease status metrics at the initial PPC consultation, so we were unable to include this in our analysis, although we know whether children died before discharge. Second, participating PedPCQN programs were self-selected and few in number, and thus, may not be generalizable to all PPC programs. Sites were encouraged by PedPCQN to report all encounters, but it is unknown what percentage of total possible encounters from each site were reported. Adjustments were not made in the analysis to account for patients who had >1 encounter during the data collection window because we wanted to understand the processes of each encounter. In addition, 1 site was missing unique patient identifiers for part of their encounters, making it impossible to identify patient counts for these encounters. Additional limitations include those inherent to the PedPCQN quality indicators and data collection process. For example, children may have had multiple or overlapping diagnoses (eg, congenital condition with neurologic involvement), but sites were instructed to enter only the primary diagnosis associated with the PPC referral. PedPCQN data could be entered in real-time or whenever teams were available, which could lead to recall bias. PedPCQN data were self-reported by PPC clinicians and, therefore, may not accurately represent the actual care that was provided, although processes and outcomes were not always favorable, suggesting teams were forthright with data collection. Although interventions were performed at least 90% of the time when a screening need was identified, we do not know what they were nor who performed them. However, for the patient and family, the most important issue is that needs are addressed and not who addresses them, and it is likely that PPC involvement influenced the provision of interventions, even when they did not provide them directly. Finally, PedPCQN data were collected by clinicians in the course of clinical care, and we did not follow up with programs to collect missing data; therefore, we report actual Ns for each process and outcome. Given the variable number of encounters at each site, we did not conduct any site-level analyses aside from staffing and service provision differences. Still, the large number of PPC encounters across 5 diverse sites provides valuable information about the care provided to seriously ill children and their families.

This review of >1500 PPC encounters revealed that non-cancer diagnoses account for the majority of referrals, although children with cancer constitute the largest single category of referrals. PPC referrals were often made for multiple reasons, the most common being for psychosocial support. PPC teams intervened 90% of the time or more for unmet needs, and distress scores in many children with moderate-severe symptoms generally improved. These outcomes were achieved although more than one-half of children received care from clinicians representing only 1 or 2 PPC disciplines. These findings point to important areas for further PPC QI, including bolstering interdisciplinary staffing to ensure optimal psychosocial support. Findings from this study can inform future work within PCQC to provide a more detailed understanding of gaps in care and to support efforts to close those gaps.

We wish to thank the Pediatric Palliative Improvement Network leadership team, including Rachel Thienprayoon, MD, Emma Jones, MD, and Conrad Williams, MD, for their expertise and initial feedback on PedPCQN data structure, elements, and definitions. We are grateful for the time and effort contributed by the UCSF PCQN core team, including Jessica Lin. We are thankful to each of the participating PedPCQN member sites for their commitment and contributions which made this report possible.