Surgical Intervention in Patients Receiving Pediatric Palliative Care Services

We performed a retrospective and prospective cohort analysis of surgical interventions performed on children enrolled in an ongoing, multicenter, prospective cohort study (PPC Research Network SHARE Study) being conducted at 7 children’s hospitals (Children’s Hospital of Philadelphia, Boston Children’s/Dana-Farber Cancer Institute, Seattle Children’s, Children’s Hospital and Clinics of Minnesota, Akron Children’s Hospital, Children’s of Alabama, and Texas Children’s). Patients aged <30 years receiving PPC services were eligible for inclusion in the study. Data, including demographic information and patient and primary caregiver symptoms, were collected between April 2017 and March 2021. Interventions undergone by these patients were determined using the Pediatric Health Information System (PHIS) database, which captures utilization data, including diagnoses and surgical interventions, across 50 children’s hospitals.

The overall SHARE study and this analysis was approved by the institutional review boards of all participating children’s hospitals.

As described previously,¹⁶  all International Classification of Diseases, Ninth and 10th Revisions, Clinical Modification (ICD-CM), diagnostic and surgical intervention codes were reviewed and classified by body systems involved and likely purpose. Body systems (adapted from the Accreditation Council for Graduate Medical Education’s Residency Review Committee for Surgery) classified the interventions on the basis of the primary operative site. The 7 likely purposes previously identified were: insertion or adjustments of hardware or catheters, attempts to rescue patient from mortality, exploration, biopsy, resection of tissue, care for cardiac congenital conditions, and care for noncardiac congenital conditions. Patients’ ICD-CM diagnoses were classified into CCC categories.¹⁷ 

At the time of study enrollment, participants provided consent to gather direct-report demographic and clinical data and to allow the linkage to the participating child’s administrative hospitalization data record contained in the PHIS (Children's Hospital Association, Lenexa, KS), both retrospectively and prospectively, during their participation in the study. Patients’ ICD-CM codes, upon which the identification of surgical interventions and CCC status were based, were obtained from their PHIS records. PHIS contains date information of the encounters associated with each ICD-CM code, thus enabling the construction of the cohort’s temporal sequence of events.

Our analyses focused on the incidence, type, and likely purpose of surgical interventions performed after the initiation of PPC services. We calculated frequencies, means, medians, and interquartile ranges (IQRs). Covariates used in the analyses included patients’ age at study entry, sex, parent-reported race and ethnicity, and whether the patient had any of the 12 CCC categories and the total number of CCC categories. To assess the likelihood of undergoing any surgical intervention post-PPC initiation, we fit a model of the dichotomous outcome “any surgical intervention” status predicted by the patient’s number of CCCs and other covariates as predictors. To assess the number of surgical interventions per patient after PPC initiation, we fit a negative binomial model with the outcome of the number of post-PPC surgical interventions predicted by the number of CCCs and the other covariates.

At the time of study entry (Table 1), the cohort of children receiving PPC (n = 602) were, on average, aged 7.1 years (SD, 6.9), ranging from 0 to 27.6 years. About half (46.7%) were female, and most were reported by their parents to be white (72.7%) and non-Hispanic (84.7%). Almost all children had at least 1 CCC (99.0%) and averaged 5.1 CCC categories of diagnoses (SD, 2.1; range, 0–10).

On average, these patients had first started to receive PPC 422 days (SD, 865) before enrolling in this study, ranging from 0 to 6762 days. The total time under observation (ie, the date of PPC initiation to the last date of PHIS follow-up) ranged from 260 to 7316 days, and the median observational period was 1174 days (IQR, 814–1578).

After initiation of PPC, 81.1% (n = 488) of patients had undergone at least 1 surgical intervention (range, 1–71) with a median of 4 interventions (IQR, 1–9). Across the entire sample, there were an average of 1.85 surgical interventions per patient-year. Among those patients with any surgical intervention after PPC initiation, patients had a median of 5 surgical interventions (IQR, 2.5–11) and mean of 8.4 (SD, 9.2), with a rate of 2.2 surgical interventions per patient-year.

The most frequent surgical interventions (Table 2) were feeding tube placement or exchange, endoscopic biopsy, tracheostomy placement or exchange, bone marrow biopsy, tunneled catheter placement or removal, bronchoscopic lavage or biopsy, and chest tube placement. These were followed by more invasive interventions, including sternum closure, abdominal wall closure, atrial septal defect repair, heart transplantation, and ventricular septal defect repair.

The body systems most likely to be the primary site of intervention (Fig 1) were the gastrointestinal (42.1%), thoracic (18.1%), and vascular/lymphatic systems (7.3%). The most common likely purposes of interventions (Fig 2) were hardware and catheters (43.1%) and biopsy (25.0%), followed by resection (9.2%), exploration (6.8%), repair of congenital conditions (cardiac, 4.9%; noncardiac, 3.2%), and rescue operations (2.9%) (Fig 2). Patients were equally likely to undergo hardware placement, repair of congenital noncardiac anomalies, and exploration at any age. Patients aged <1 year were more likely to undergo congenital cardiac operations and less likely to undergo biopsies or resections, whereas patients aged >18 years were more likely to undergo rescue operations.

Regarding patients’ specific demographic characteristics, controlling for the other characteristics, the adjusted probability of undergoing a surgical intervention after PPC decreased with age, from 0.84 (95% confidence interval [CI], 0.81–0.88) at 1 year of age to 0.72 (95% CI, 0.64–0.81) by 20 years of age. Compared with females (0.76; 95% CI, 0.71–0.81), males had an increased adjusted probability of undergoing a surgical intervention (0.85 [95% CI, 0.81–0.89]). Finally, patients identified as non-Hispanic were more likely to undergo a surgical intervention (0.83 probability; 95% CI, 0.80–0.86) compared with those identified as Hispanic (0.71 probability; 95% CI, 0.61–0.81).

Regarding clinical patient characteristics, no CCC types were associated with the odds of undergoing a surgery post-PPC initiation.

In unadjusted analysis, each additional CCC was associated with an increased rate of surgical interventions (incidence rate ratio [IRR], 1.26; 95% CI, 1.20–1.31; P < .001).

In adjusted analysis, the number of CCCs was associated with more surgical interventions (IRR, 1.56; 95% CI, 1.1–2.17; P = .001). Compared with children identified as white, Asian American children underwent a high rate of surgeries (IRR, 1.62; 95% CI, 1.03–2.55; P = .03). Additionally, there was variability by type of CCC: having a neuromuscular CCC (IRR, 0.56; 95% CI, 0.39–0.79; P = .001), renal CCC (IRR, 0.68; 95% CI, 0.46–0.99; P = .04), or hematologic CCC (IRR, 0.65; 95% CI, 0.42–0.99; P = .042) was associated with a reduced rate of post-PPC surgical interventions, compared with not having these conditions. Additionally, technology dependence did not change the association of surgical intervention with age or other demographic characteristics.

Children who underwent surgical interventions were statistically less likely to die while receiving PPC services (29% vs 40%, P < .03), even though patients who underwent surgical interventions had spent more time (albeit not to a statistically significant degree) receiving PPC services (429 days versus 394 days, P = .70) (Table 1). Moreover, the majority (60%) of those who did undergo intervention were alive at 5 years after their first intervention (Fig 3), and there was greater survivorship among those who underwent intervention compared with those who did not (Fig 3) (71% vs 61%, P = .03).

Although surgical patients are less likely than nonsurgical patients to receive palliative care consultation,¹⁸  our findings demonstrate that a high proportion of the PPC population (81.6%) undergoes surgical intervention. Furthermore, undergoing surgical intervention is not a solitary event: half of those patients who experienced a surgical intervention did so 5 or more times. Although many of the surgical interventions were likely performed for the purpose of maintenance of implanted hardware or diagnosis (hardware and catheters and biopsy accounted for 68% of surgical interventions), the majority of the remaining 30% were likely performed with the intention of repair or rescue. Notably, undergoing surgical intervention was not associated with an increase in mortality; in fact, most patients who had undergone an intervention were alive 5 years after the first surgical intervention. These data suggest that the role of surgical care in the PPC population is substantial in terms of the incidence and varied nature of the surgical interventions performed and does not appear to be futile or harmful.

Our findings indicate a high prevalence of surgical interventions among children receiving PPC, and that patients with more medical complexity (as indicated by a greater number of CCC categories) undergo more interventions. Almost all the children in this study had at least 1 CCC, which may explain the high proportion of surgery likely performed for the maintenance of implanted hardware or diagnostic purposes. In a study assessing the health care utilization of children with CCCs during their terminal hospitalizations, one-third underwent surgical intervention.¹⁹  In our sample, having more CCCs was associated with undergoing more surgical interventions. These findings suggest that children with greater medical complexity may require even more collaboration among palliative care clinicians, pediatric surgeons, intensivists, anesthesia, and nursing. This may be especially true for patients with cardiovascular, respiratory, gastrointestinal, genitourinary, metabolic, malignant CCCs, or who are technology dependent, because those diagnoses were each associated with having at least 1 surgical intervention.

Finally, older children, female children, and Hispanic children were less likely to undergo intervention after PPC initiation. Each of these populations have distinct plausible explanations for this observation. Regarding age, this finding most likely reflects the patient’s primary CCC, since certain CCCs that are more amenable to surgical intervention (such as congenital heart disease or other congenital malformations) tend to present at birth or younger ages. Regarding sex, published data are mixed regarding sex-based differences in pediatric surgical outcomes, but some studies have shown increased mortality among female patients,²⁰  perhaps affecting perceptions of surgical risk-to-benefit. Regarding race and ethnicity, numerous studies have shown that white patients are more likely to undergo surgery, and that nonwhite patients are more likely to experience operative morbidity and mortality.^21,22  However, the relationship between race and number of interventions that may explain why Asian American children had a higher incidence of interventions compared with white children is an unclear one, and warrants further investigation.

In this study, we intentionally used the term “interventions” to refer broadly to both surgical interventions and surgical operations. Although these vary greatly in terms of risk and invasiveness, as the adage goes, and is worth remembering regarding surgical interventions, nothing is free. Most surgical interventions require general anesthesia and confer some risk of complications, such that even “small” interventions in an already medically vulnerable population ought to be approached in conjunction with discussions about patient and family goals and the extent to which the intended purpose of the intervention is concordant. Although these conversations may be initiated and led by whoever raises the possibility of surgical intervention, this is an area in which pediatric surgeons, who have a keen and nuanced understanding of not only the risks and benefits but also the impact of these interventions, may be able to provide “primary palliative care,” which is to say, the provision of palliative care services and skills by nonspecialty clinicians. Despite extensive exposure to seriously ill and dying patients, surgical trainees consistently report deficits in training on how to navigate communication and decision-making for patients with serious illness and the tangible possibility of death, though they agree that palliative skills are important for surgeons and support the integration of these skills into surgical training.²³ 

Though this study did not include a marker for illness severity, the proportion of the PPC population made up of children with acute, serious illness such as malignancy may undergo operative intervention for different purposes. A retrospective review of children who had an operation within 6 months of their deaths, arguably a more acutely ill population, showed that 20% were performed with curative intent, and notably that over half were performed urgently and almost 20% in spite of a do-not-resuscitate order.¹⁵  An analysis of another very seriously ill population, children during terminal hospitalizations, revealed that half of children underwent surgery, with over half (53%) performed for the likely purpose of repair or rescue.¹⁶  The context of a child nearing the end of life or even admitted in a terminal hospitalization may provoke more explicitly end-of-life rather than palliative care, but even as the degree of illness acuity seems to shift the purpose of intervention, there is nonetheless a high incidence of interventions.

Of considerable note, surgical intervention in this study was not associated with a decrease in survival duration. Three-quarters of patients who underwent intervention in this study were alive 5 years after the first surgery, which suggests that operative intervention in the palliative care population is not futile. Further supporting this conclusion is that children who underwent intervention had greater survivorship at 5 years compared with those who did not. Although this later finding may be attributable in part to a justifiable selection bias (whereby children with acutely life-threatening conditions were less likely to be offered surgical intervention or their parents were less likely to choose such intervention, and thus patients expected to live long enough to benefit from receiving surgical interventions were more likely to do so), taken together, these findings challenge the viewpoint that surgical interventions for patients receiving palliative care are futile or harmful and tend to hasten mortality. They also reinforce the value of involving PPC services early.²⁴  Furthermore, these findings underscore the importance of ensuring that racial and ethnic minority groups receive equitable and effective surgical intervention counseling regarding the potential benefits of surgical intervention.

This study has 4 main limitations warranting discussion. First, interventions were classified on the basis of ICD-CM, Ninth and 10th Revision codes, which lack clinical context. Accordingly, the likely purpose classifications for surgical interventions may not reflect the intent in having performed them. Second, we were unable to determine goal concordance or the impact on quality of life of these interventions. Third and relatedly, SHARE data do not include illness severity or all code status changes; consequently, we were unable to ascertain whether children who did not undergo surgical interventions were more critically ill or had do-not-resuscitate or do-not-intubate orders active, which may have been associated with lower odds of undergoing surgery. Future directions for research include determining via patient- or parent-report the family’s intended purpose and associated goal concordance of surgical interventions, and the role of near-term mortality prognosis in surgical decision-making. Finally, our findings are limited by the fact that our study population overrepresents white, non-Hispanic patients (63.1% of our sample compared with 49.9% nationally).²⁵ 

Most children receiving PPC services will undergo at least 1 surgical intervention, and many will undergo multiple interventions. Although most of these interventions are non- or minimally invasive surgical interventions likely related to hardware maintenance, catheter placement, or biopsy, a notable proportion include invasive operations for rescue or resection with substantial surgical risk. Undergoing intervention is not futile, because surgical intervention is associated with increased survival. Various patient populations that are more likely to undergo surgical intervention (such as those with multiple CCCs) or less likely to undergo intervention (such as female children, older children, and ethnic minorities) warrant specific focus as methods to enhance counseling and periprocedural care around surgical interventions are developed for the PPC population.

We thank all the parents who participated in this study. We also thank all members of the Pediatric Palliative Care Research Network (PPCRN) SHARE Project team, including Sarah Friebert, Tammy Kang, Emily Johnston, Ross Hays, Jori Borgetz, Stephan Friedrichsdorf, Kris Catrine (site investigators), Douglas Hill, Pamela S. Hinds, Matt Hall (collaborating investigators), Karen Crew (project coordinator), Hannah Katkoff and Heather Griffis (database design and management), and Rae Xiao (statistical analysis), and Tatiana Arevalo-Soriano, Leah Beight, Madeline Bilodeau, Jennifer Chapman, Porag Jeet Das, Gabby Helton, Rachel Jenkins, Ali Kolste, Deborah Maglionico, Isaac Martinez, Amanda Mercer, Ashley Morris, Shimei Nelapati, Rachel Porth, Nicole Etsekson Sherr, Kelly Shipman, and Namrata Walia (study site research coordinators).

CCC

complex chronic condition

CI

confidence interval

ICD-CM

International Classification of Diseases Clinical Modification

IQR

interquartile range

IRR

incidence rate ratio

PHIS

Pediatric Health Information System

PPC

pediatric palliative care