Research burden in pediatric cancer: A mixed methods study from the PediQUEST Response Trial

Abstract

Context:

Research burden in pediatric advanced cancer is understudied.

Objectives:

To analyze study burden among parents and children participating in the PediQUEST Response Trial, and explore associated factors, including belonging to historically marginalized (HM) backgrounds or facing economic hardship (EH).

Methods:

Mixed-methods secondary analysis of data collected for an RCT. Children ≥2 years old with advanced cancer and a parent were enrolled for 18-weeks across five US pediatric cancer centers. Trough thematic analysis, we examined all parents and child (≥8yo) interviews to identify burden sources. We created a six-level “degree of burden” variable that integrated qualitative data (burden sources, levels 0–3) and quantitative data (dropout status, levels 4–5). Social disparity was grouped as: neither HM nor EH, either HM or EH, and both. We used ordinal logistic regression to explore burden-disparity associations.

Results:

Among 154 randomized dyads, 149 had HM/EH data; 102 families completed interviews (102 parents, 44 children). Two burden themes were identified: study-related and daily life stressors. Most parents (62.4%) and children (48.8%) had burden levels 0–3; high levels (4–5) affected about 20% of families. Social disparity levels were 47.7% HM−/EH−, 30.9% HM+ or EH+, and 21.5% HM+/EH+. Parent burden increased with social disparity (HM+ or EH+ vs. HM−/EH−: cumulative OR 2.3, 95%CI 1.1, 5.1; HM+/EH+ vs. HM−/EH− OR 4.9, 95%CI 1.9, 12.2); no associations found for child’s burden.

Conclusion:

Study burden was mostly low to moderate but higher in families endorsing social disparities. Enhanced support strategies may promote diversity in pediatric advanced cancer research. Clinicaltrials.gov NCT03408314

Key Message:

This mixed-methods study analyzed burden in the context of pediatric advanced cancer research and its relationship with families’ sociodemographic characteristics. We found low to moderate burden levels that in most cases did not affect study completion. However, results uncover disparities in burden distribution related to race/ethnicity and economic hardship.

INTRODUCTION

Investigation related to improving the lives of children with advanced cancer and their families has long been a national priority.^1–3 However, research in supportive, palliative, and psychosocial care intervention studies faces multiple challenges,^4,5 including study burden (the perceived effort, time commitment, and stress related to research participation),^6–9 a known barrier to recruitment in pediatrics.^5,10,11 A recent systematic review highlighted the scarcity of evidence on study burden in pediatric palliative care research: of the 24 included studies, only 17 addressed burden and most were conducted at end-of-life or involved bereaved parents, and identified the emergence of strong emotions, time constraints, and the child’s deteriorating health condition as the main sources of burden.⁸ Little is known about the impact of research participation among families of seriously ill children earlier in the illness trajectory, and whether this impact differs for underrepresented groups,^12–14 such as those from historically marginalized (HM) backgrounds or facing economic hardship (EH).

Caring for a child with cancer imposes significant economic impacts on caregivers, especially for marginalized families, leaving them more vulnerable to study burden.^15–17 These caregivers are also less frequently invited to participate in research.¹⁸ While cultural competence and rapport with study staff may improve recruitment and retention of HM patients,^19–22 caregiving responsibilities, work, or financial constraints may still hinder sustained participation.^23,24

Using a mixed-methods approach, this study analyzes the frequency, intensity, and sources of study burden, and the association of the level of study burden in families with racially or ethnically HM backgrounds and EH, among parents and children participating in the PediQUEST Response Trial, a palliative care randomized controlled trial.²⁵

METHODS

Design, Setting, and Population

This mixed-methods secondary analysis used data collected for the PediQUEST Response Trial (NCT03408314). We provide a brief trial description to portray study burden potential. The PediQUEST Response intervention combined regular feedback of electronic patient-reported outcomes using the PediQUEST system and consultation with specialized pediatric palliative care (SPPC). Across five major US pediatric cancer centers, we enrolled children ≥2 years old with advanced cancer, not receiving regular SPPC care, who were English/Spanish speakers, along with one parent. Study protocol was approved by local IRBs, with full methods published elsewhere.²⁵

Data Collection

From a mixed methods perspective, the PediQUEST Response Trial used an explanatory sequential design (QUAN → qual), see Figure 1.²⁶ The study included a 2-week run-in to identify responders, i.e. dyads (child and parent) who completed ≥2 of three assigned surveys, who were randomized to the PQ-Response or control arm and followed for 16 weeks. All parents and patients ≥5 years old completed age-adapted PediQUEST-Weekly surveys on symptom burden and health-related quality-of-life. Parents also completed baseline and monthly surveys on family characteristics, parent distress, and childcare. Surveys were completed through the PediQUEST system, via the PQ-App or online. Intervention families received feedback reports sharing PediQUEST-Weekly survey results and ongoing SPPC consultation (initial visit + at least monthly follow-up). Primary and SPPC teams of intervention participants also received PQ-reports. Dyads in the control arm continued with usual care, completed all study surveys, and did not receive feedback reports. All parents, and children able to answer surveys for ≥8 years old, were invited to participate in semi-structured exit interviews at study’s end or upon dropout conducted by trained site staff.

Figure 1: The PediQUEST Response RCT study procedures.

The Figure summarizes participation details for the dyads throughout the 18-week follow-up period. The study followed an explanatory sequential design with an initial quantitative phase followed by a qualitative phase (as indicated above).

*During the 2-week run-in period we identified and excluded non-responders (i.e. dyads who answered fewer than 2 of the 3 assigned PQ-Weekly surveys).

^∤Only children able to answer surveys for ≥8 yo invited to interview.

Abbreviations: SPPC, specialized pediatric palliative care team; PQ, PediQUEST; HRQoL, Health-related Quality of Life.

Retention and adherence strategies included establishing staff-participant rapport, flexibility regarding survey assignment day and communication means, automated email reminders, and close follow-up after missed surveys. Children and parents each received gift cards as a token of our appreciation ($20 for the run-in period and $40 per month of participation).

Mixed-Methods Analytic Approach

We included dyads with data on parental race, ethnicity, and economic hardship. We integrated quantitative data on study participation and socio-demographic factors with qualitative findings on study burden. Data sources included the participant’s tracking database, baseline surveys, and exit interviews. We describe below data treatment, merging, and representation. While presented in three phases, the process was iterative.

Quantitative Analysis

Information on subject participation included child and parent survey response rates, study arm assignment, and end-of-study status categorized as: completed follow-up, died while on study, or dropout. We distinguished three dropout reasons: (i) unanswered surveys (when ≥1 in the dyad failed to answer ≥ 4 consecutive surveys despite intense follow up, we ceased study procedures assuming this implied unwillingness to continue); (ii) study burden (when dyads voluntarily stopped participation endorsing that research procedures/interactions were burdensome); (iii) other reasons unrelated to burden (e.g. transfer of care).

Type of cancer was extracted from medical records. Child and parent demographics (gender, age at enrollment, race, and ethnicity), parent education level, and family household material hardship were parent-reported at baseline. We created a combined race and ethnicity binary variable, HM− (non-Hispanic White) vs. HM+ (Hispanic, Black/African American, Asian, Pacific Islander, or multiple races) in agreement with existing recommendations.²⁷ Economic hardship was evaluated using an adapted version of the 13-item Household Material Hardship framework proposed by Bona et al.,²⁸ which assembled previously validated items covering three domains: energy, housing, and food. We used four questions asking whether in the last six months: (i) the energy company threatened disconnection (yes/no); (ii) the family had financial difficulties paying rent/mortgage on time (yes/no); and (iii) they worried about running out of food, or (iv) ran out of food. Answering “often true” or “sometimes true” to either food question was considered a “yes.” Economic hardship (EH) was defined as a binary variable, coded as positive (EH+) if a participant answered “yes” to either the energy, housing or food domains, and negative (EH−) otherwise.

Recognizing that HM and EH are closely intertwined in shaping health outcomes,²⁷ we created a composite variable, “social disparity level,” which integrates these indicators into three levels: HM−/EH− (none), HM+ or EH+ (low to moderate), and HM+/EH+ (high).

To compare socio-demographic characteristics and study participation of parents and children who participated in the exit interview against those who did not, we used chi-squared or Fisher’s exact tests for categorical and t-tests for continuous variables. To examine the relationship between number of surveys completed across social disparity levels, we fitted a linear model, and report Bonferroni adjusted pairwise comparisons. All analysis were conducted in Stata v18.0 software.²⁹

Qualitative Analysis

Study burden was identified through reflexive thematic analysis of exit interviews while doing the trial’s process measures analysis.^30,31 The interview guides did not include specific questions about burden. They aimed to capture participants’ study experience, such as overall likes and dislikes, thoughts about surveys, and among intervention arm participants, opinions about PQ-reports and SPPC team interactions. The analysis team was multidisciplinary and had extensive qualitative methods experience (MA, IJEU, JW, VD, MLR). Using MAXQDA software,³² at least two independent coders re-analyzed each interview (audio-recorded, de-identified and transcribed verbatim) focusing on study burden (ET, DF, AM, MLR, MA, IJEU, OA). Through inductive coding we examined burden sources and grouped them into themes, revised and refined during weekly meetings. To contextualize the grounded data, we conducted a literature review close to analysis completion. We report findings per the Standards for Reporting Qualitative Research.³³

Mixed-Data Integration

We merged quantitative and qualitative datasets using data transformation³⁴ and created an ordinal “degree of burden” variable. Anchors were: “no burden reported” (participants who did not mention burden in interviews and either completed the study, dropped out for non-burden reasons, or died) and “maximum burden” (those who dropped out citing study burden). Participants who reported burden in interviews and completed the study, dropped out for non-burden reasons, or died, were assigned low-to-moderate burden levels based on the number of burden sources endorsed. Finally, we assumed that those who dropped out due to unanswered surveys experienced higher-than-moderate burden, but less than the maximum. For dyads falling in the two dropout categories, we contrasted date of dropout with disease/clinical status.

To contextualize study burden for children and parents, we developed a data matrix heat map that triangulated each dyad’s social disparity level, study arm, end-of-study status, burden degree, and completion of exit interview. Additionally, we fitted an ordinal logistic regression to explore the hypothesis that burden increased with social disparity.

RESULTS

Of the 154 randomized parent-child dyads, 149 with complete data on parental reported race, ethnicity, and EH, were included in the analysis. Among the five excluded dyads, four were assigned to the intervention; two dropped out because of unanswered surveys, two completed follow-up, and one child died during the study. Exit interviews were completed by 102 families (102 parents (68.5%) and 44 children (52.4% of 84 eligible)). Table 1 presents participant characteristics, comparing those with and without parent exit interviews, and across social disparity levels. Children were on average 10.8 years old, half were girls, and 45.6% had brain tumors. Parents were predominantly White non-Hispanic English-speaking mothers, averaging 42 years old, and mostly married or living with their partner. Regarding social disparity, 47.7% were HM−/EH−, 30.9% HM+ or EH+, and 21.5% HM+/EH+. Parents in the HM−/EH− group had significantly higher education attainments than those at greater risk for social disparity (71.8% of HM−/EH− held a bachelor’s degree or higher, compared to 30.4% (HM+ or EH+) and 34.4% (HM+/EH+), p <0.001). In terms of study adherence, 75.8% of dyads completed follow-up and 22.8% dropped out (17.4% due to unanswered surveys, 2.7% due to study burden, and 2.7% for other reasons); Two children (1.3%) died during the study.

Table 1.

Characteristics of Participants Included in Analysis^a (n=149) by Interview Status and Social Disparity Level

Interview Status
	Children				Parents
Characteristics	Full sample (n=149) N (%)	With Parent Interviewsb (n=102) N (%)c	Without Parent interviewsb (n=47) N (%)c	p-valued	Full sample (n=149) N (%)	With Parent Interviewsb (n=102) N (%)c	Without Parent interviewsb (n=47) N (%)c	p-valued
Age at enrollment
Mean (SD)	10.8 (6.1)	11.4 (6.2)	9.6 (5.7)	0.109	42.2 (8.6)	43.0 (8.7)	40.7 (8.2)	0.135
Gender
Female	74 (49.7)	47 (63.5)	27 (36.5)	0.220	125 (83.9)	85 (68.0)	40 (32.0)	0.999
Male	75 (50.3)	55 (73.3)	20 (26.7)		24 (16.1)	17 (70.8)	7 (29.2)
Race/Ethnicity combined e
Non-historically marginalized (HM−)					100 (67.1)	76 (76.0)	24 (24.0)	0.008
Historically marginalized (HM+)					49 (32.9)	26 (53.1)	22 (46.9)
Economic Hardship (EH) f
None (EH−)					88 (59.1)	66 (75.0)	22 (25.0)	0.049
At least one (EH+)					61 (40.9)	36 (59.0)	25 (41.0)
Social Disparity g
None (HM−/EH−)					71 (47.7)	55 (77.5)	16 (22.5)	0.009
Moderate (HM+ or EH+)					46 (30.9)	32 (69.6)	14 (30.4)
High (HM+/EH+)					32 (21.5)	15 (46.9)	17 (53.1)
Marital Status
Married or living with partner					117 (78.5)	82 (70.1)	35 (29.9)
Separated or divorced					16 (10.7)	13 (81.3)	3 (18.8)	0.069
Single or widowed					16 (10.7)	7 (43.8)	9 (56.3)
Education
Up to associate degree					73 (49.0)	44 (60.3)	29 (39.7)
Bachelor’s degree and higher					76 (51.0)	58 (76.3)	18 (23.7)	0.052
Language spoken at home
English					127 (85.2)	88 (69.3)	39 (30.7)
Spanish					9 (6.0)	5 (55.6)	4 (44.4)	0.650
Otherh					13 (8.7)	9 (69.2)	4 (30.8)
Tumor Type
Brain Tumor	68 (45.6)	49 (72.1)	19 (27.9)
Solid Tumor	58 (38.9)	38 (65.5)	20 (34.5)	0.724
Hematological Cancer	23 (15.4)	15 (65.2)	8 (34.8)
Study arm
Usual Cancer Care	79 (53.0)	51 (64.6)	28 (35.4)	0.295
PQ-Response intervention	70 (47.0)	51 (72.9)	19 (27.1)
End of study status
Completed 18 weeks follow-up	113 (75.8)	90 (79.6)	23 (20.4)
Died while on study	2 (1.3)	1 (50.0)	1 (50.0)
Dropout - unanswered surveys	26 (17.4)	7 (26.9)	19 (73.1)	<0.001
Dropout - study burden	4 (2.7)	2 (50.0)	2 (50.0)
Dropout - other	4 (2.7)	2 (50.0)	2 (50.0)
# Surveys completed i
Weekly - Mean (SD)	11.7 (5.7)	12.9 (5.4)	8.8 (5.4)	0.001	13.5 (4.7)	14.8 (4.0)	10.7 (4.8)	<0.001
Monthly - Mean (SD)					4.1 (1.3)	4.5 (1.0)	3.3 (1.4)	<0.001
Social Disparity Level
	Children				Parents
	HM−/EH− N=71 n (%)	HM+ or EH+ N=46 n (%)	HM+/EH+ N=31 n (%)	p-valuej	HM−/EH− N=71 n (%)	HM+ or EH+ N=46 n (%)	HM+/EH+ N=31 n (%)	p-valuej
End of study status
Completed 18 weeks follow-up	56 (78.9)	35 (76.1)	22 (68.8)
Died while on study	1 (1.4)	0 (0.0)	1 (3.1)
Dropout - unanswered surveys	8 (11.3)	9 (19.6)	9 (28.1)	0.378
Dropout - study burden	3 (4.2)	1 (2.2)	0 (0.0)
Dropout - other	3 (4.2)	1 (2.2)	0 (0.0)
# Surveys completed i
Weekly - Mean (SD)	13.3 (4.8)	11.3 (6.3)	8.5 (5.3)	0.003	14.7 (4.5)	13.0 (4.4)	11.3 (4.7	0.002
Monthly - Mean (SD)					4.3 (1.1)	3.9 (1.4)	3.8 (1.3)	0.052

^aSubjects with parent race/ethnicity and economic hardship data.

^bNumber of dyads with parent interviews within levels of each factor. Data for children with interviews is not shown to increase readability. Patterns are similar.

^cRow percentages.

^dt-test for continuous variables, Chi-squared or Fisher’s exact test (all levels, excluding missing) for categorical variables.

^eRace/ethnicity combined: binary variable including non-historically marginalized (White non-Hispanic) and historically marginalized (Hispanic, Black, or African American, Asian, and others).

^fEconomic Hardship (EH): positive answer to ≥1 of the household material hardship domains assessed (energy, housing, and food).

^gSocial Disparity Level: three-level ordinal variable: HM−/EH−: neither historically marginalized nor economic hardship; HM+ or EH+: either historically marginalized or economic hardship; HM+/EH+: historically marginalized and economic hardship.

^hOther Languages spoken at home (in order of frequency): Chinese, Polish, Portuguese, Arabic, French, German, Hindi, Pujarati, Romanian, Urdu, Vietnamese.

ⁱMaximum number of weekly surveys: 17; maximum number of monthly surveys: 5.

^jFisher’s exact test for end of study status and linear regression for survey completed.

Parents with exit interviews (n=102) differed from those without interviews (n=47) in several aspects (Table 1). Those from the HM+/EH+ group and those with lower educational levels were significantly less likely to participate in interviews. Further, dyads participating in exit interviews were more likely to complete study follow-up and answered, on average, more surveys compared to those without interviews. Study arm was not associated with parent participation in interviews. Children from HM+/EH+ families were also less likely to participate in interviews (data not shown).

Finally, Table 1 shows the relationship between social disparity level, end of study status, and number of surveys completed. Social disparity was not associated with reaching the end of follow-up. However, the average number of weekly surveys answered by both children and parents decreased as social disparity increased, with significant differences between the HM−/EH− and HM+/EH+ levels in adjusted pairwise comparisons (children: 13.3 vs 8.5, p=0.001 and parents 14.7 vs 11.3 p<0.001).

Sources of Burden Themes

Through our qualitative analysis, we identified two emerging burden themes: 1) study stressors, defined as those specifically related to RCT participation, i.e., fulfilling study procedures and/or receiving the intervention; and 2) daily life stressors, the additional demands that study participation imposed on routine activities. Each theme included various sub-themes (burden sources) and their dimensions. Table 2 depicts these relationships with supporting participant quotes.

Table 2.

Emerging Themes and Sources of Burden among Children (n=44) and Parents (n=102) who Provided Exit Interviews

Theme	Burden Source	Dimension	Quote #	Quote + Participant information1
Study Stressors	Survey related	Survey Window	1	It wasn’t so much that it was weekly but the time to get [the survey] done. Because all day Tuesday I’m in clinic with him and so I thought I’d be able to get it done. And then Monday we’re trying get things ready for clinic and run errands and get everything ready. So, it was the turnaround for that sometimes was just a little challenging.  Mother, Child 2 yo with solid tumor, HM-/EH-, Intervention, Complete
			2	Sometimes I was at school or something, so I’d have [the survey] on, and it wouldn’t work because the connection wouldn’t work. So, once I finally got done with a day or two, it would be like, oh, I guess I can’t answer it now.  Patient (M) 15 yo with brain tumor, HM-/EH-, Intervention, Complete
		Survey Length	3	I knew that [the survey] would take 15–20 minutes of my life, and sometimes it was too much. Streamlining it would be better.  Father, Child 19 yo with brain tumor, HM-/EH+, Intervention, Complete
			4	I don’t think there was necessarily anything that really bothered me except for the length of the survey which I’m not sure can necessarily be changed. It did feel a little long and tedious.  Patient (F) 18 yo with brain tumor, HM-/EH-, Intervention, Dropout (transfer care)
		Survey Completion	5	Initially, [the survey] was okay, but then after a while it kind of felt like homework.  Mother, Child 8 yo with brain tumor, HM+/EH+, Control, Complete
			6	Sometimes, I can honestly say [the survey] was annoying to do sometimes. Not the first thing that I would choose to do.  Patient (F) 17 yo with solid tumor, HM-/EH-, Control, Complete
		Language Barrier	7	There were some words that I didn’t understand. I know there was something like made me nervous and [inaudible]. It’s not like I don’t like that part, but I was kind of confused a little bit at the beginning, but then to get the meaning and then oh, there’s a new word right there.  Mother, Child 2 yo with solid tumor, HM+/EH+, Control, Complete
		Reflections triggered by surveys	8	Well, I just don’t want to think too hard. Some questions are hard for me. Some kind of purposely to seek more, don’t think about it in my day-to-day life.  Mother, Child 11 yo with brain tumor, HM+/EH-, Intervention, Complete
			9	I think the only thing that I disliked about it had to mostly be the timing of when I joined because I was [inaudible] PediQUEST surveys post treatment. And so, I think it just had to do with like, gosh darn it, I don’t really wanna be thinking about cancer right now. And then having a survey to do every week. I guess that was honestly the only thing that really was a bummer.  Patient (F) 19 yo with hematological cancer, HM-/EH-, Control, Complete
	Study Duration		10	I mean [the study duration] was a very long time. It took a – the timeframe to do that study just seemed like kind of unnecessary for it to take as long as it did.  Mother, Child 14 yo with brain tumor, HM-/EH+, Control, Complete
			11	I mean 18 weeks, maybe just it was a little bit – it seemed like a long time.  Patient (F) 14 yo with brain tumor, HM-/EH+, Intervention, Complete
	Palliative Care Team Interactions		12	It got a little overwhelming when I had to meet up with people. I was kind of hoping we would fall on the side where you just answer questions, and you don’t have to talk to people. Not that I didn’t like the people, because they were really nice and there was nothing— I had nothing against them, it’s just, I don’t like meeting new people.  Mother, Child 8 yo with solid tumor, HM-/EH-, Intervention, Dropout (burden)
			13	I know it’s hard because of scheduling on my end with chemo and sometimes I’m here and sometimes I’m not. But it would be nice to have an idea for how often I’m going to see them and when I’m going to see them. […] Just because it was kind of spontaneous a lot of the time, and I wasn’t always up for seeing another doctor.  Patient (F) 17 yo with solid tumor, HM+/EH-, Intervention, Complete
Daily Life Stressors	Competing Responsibilities	Adding to Preexisting Heavy Workload	14	Because when parents are dealing with kids with cancer, they’ve just got so much extra on their plate besides everything else in life. This is from – dealing with kids with cancer and everything that comes with it, it’s a lot to deal with physically, emotionally, mentally, financially, spiritually. And then when you’ve got this thing that you’re feeling stressed about getting done on top of all that when it’s – I think it can be a little bit too much for some people.  Mother, Child 14 yo with brain tumor, HM-/EH+, Control, Complete
			15	Usually I would get the notification [to complete the survey] and I’d be in the middle of something and not really have time to sit down and do a 20-minute survey or whatever. […] or I would forget to complete them – even though I could have probably figured out a way. But that was like the last thought on my mind at that time.  Patient (M)18 yo with solid tumor, HM-/EH+, Intervention, Complete
		Conflicts with Work /School	16	I work snow removal and everything so a lot of times that I had to work late, and stuff and things would end up – I’d forgetting or – everything would expire and I’m like, oh, okay, sorry.  Father, Child 14 yo with brain tumor, HM-/EH+, Control, Complete
			17	Sometimes I was really busy doing homework or something and I’d get the alert, and I continued doing my homework because I needed to do that. And I forget entirely about the [survey].  Patient (M) 14 yo with brain tumor, HM-/EH+, Control, Complete
		Competing Family Demands	18	Because honestly you go to sit down and complete the survey and then like I said, mother-in-law would come downstairs and you know, then I’m dealing with her for the next two hours because she’s got early-stage dementia. Things would just consume the day.  Mother, Child 12 yo with brain tumor, HM+/EH+, Intervention, Complete
	Child Clinical Context		19	But we spent all days in the hospital with him or tried to take care of him, tried to get him to visit, to help him with something, and this took all the time. I didn’t check email or do anything else, and by the time we were able to get back to normal a little bit I checked my email, or what not, [and] the time to submit the questions had passed. Father, Child 6 yo with brain tumor, HM+/EH-, Control, Complete
			20	When I get my hard chemos, Saturdays around the ten-day mark, which is when I’m at my absolute lowest in terms of blood counts, hemoglobin mainly, and everything else. So, there’s a chance that I could’ve been in the emergency room then, or in the hospital receiving chemo. […] So that probably would be the other component of [missing surveys], along with just forgetting.  Patient (M) 20 yo with solid tumor, HM-/EH-, Control, Complete

¹Participant information include participant role (father, mother, patient), child’s age, tumor type (brain, solid, or hematological malignancy), HM/EH status, study arm (intervention or control), end of study status (complete or dropout)

Abbreviations: F, female, M, male, HM-, non-historically marginalized (White non-Hispanic); HM+, historically marginalized (Hispanic, Black, or African American, Asian, and others); EH-, no economic hardship; EH+, economic hardship (positive answer to ≥1 of the household material hardship domains assessed (energy, housing, and food)).

Study Stressors

Participants identified weekly surveys and interactions with a new team as sources of burden. Parents reported challenges in answering surveys within the 54-hour allotted window due to competing caregiving tasks, whereas children noted they forgot or prioritized schoolwork (Table 2, quotes 1–2). Surveys were described as “too long” (approximately 15–20 minutes), “tedious,” or exceeding their attention span (Table 2, quotes 3–4). Some found the weekly frequency “annoying,” likening it to “homework” or “a chore” (Table 2, quotes 5–6), particularly when symptom burden was perceived as low, and week-to-week responses changing little.

Survey comprehensibility and the reflections prompted by answering were also mentioned as sources of stress. One parent felt “nervous” and “confused” by unfamiliar words (Table 2, quote 7), while others noted that questions confronted them with issues they typically tried to avoid, such as the cancer itself or treatment effects (Table 2, quotes 8–9). Other participants found the 18-week study duration “long” (Table 2, quotes 10–11).

In the intervention arm, interactions with the SPPC team could create additional challenges. Some found incorporating new providers to be overwhelming, even while describing the team as “very nice.” One parent voiced that meeting the SPPC team reminded them about the lack of progress in their child’s treatment (Table 2, quotes 12–13).

Daily Life Stressors

Participants cited aspects of daily life, such as work responsibilities, caregiving, and numerous appointments, as adding to their burden during study participation. While the study was described as manageable, some felt “already busy” or “overwhelmed with other things” (Table 2, quotes 14–15). Parents and children reflected that work, or school, took precedence over study participation (Table 2, quotes 16–17), and parents highlighted caring for other family members as an added conflicting responsibility (Table 2, quote 18). During periods of child clinical instability, such as high symptom burden or hospital admissions, surveys added stress and were frequently neglected (Table 2, quotes 19–20).

Degree of burden

Categorized into six levels, ranging from 0 to 5: level 0 indicated no burden, levels 1–3 reflected low to moderate burden based on the number of sources mentioned in the interview (of the five sub-themes identified, no participant endorsed more than three), level 4 denoted high burden (drop out due to unanswered surveys); and level 5 indicated maximum burden (dropout because of study burden). Most dyads completed the study with no, low or moderate burden levels: 42.3% of parents reported no burden and 20.1% low to moderate burden. Among 84 eligible children, 28.6% reported no burden, 20.2% low to moderate burden, and for 29.8% we did not have burden information. High burden was observed for 17.4% of parents and 17.9% of children, while 2.7% of parents and 3.6% of children were classified with maximum burden. Of note, burden levels did not differ by study arm. We lacked burden information for 17.4% of the sample (n=26); of these, 23 completed follow-up, one patient died, and two dyads dropped out for other reasons.

Among the thirty dyads that dropped out (study burden levels 4–5), we observed two distinct patterns of clinical changes occurring close to the time they stopped answering: 1) remission or regression with low symptom scores (n=20), and 2) disease progression, major surgery, daily radiotherapy, stem cell transplant, or end-of-life (n=10).

Contextualizing study burden

We created a heatmap to display degree of burden alongside other study variables for all dyads (Figure 2). Study arm was not associated with parent or child burden. Of the four maximum burden cases (level 5), three occurred in HM−/EH− families. Figure 3 presents burden distribution within social disparity categories by respondent type with burden collapsed into three levels: none, low to moderate (levels 1–3), and high (levels 4–5) burden. For parents, burden was significantly associated with social disparity (p=0.004). An ordinal logistic regression was fitted for cases with burden information (none, low/moderate, high), showing that burden increased with social disparity (HM+ or EH+ vs. HM−/EH−: cumulative OR 2.3, 95%CI 1.1, 5.1, p=0.034; HM+/EH+ vs. HM−/EH− cumulative OR 4.9, 95%CI 1.9, 12.2, p=0.001). No significant association was found in children (p=0.155); notably, 47.4% of HM+/EH+ children lacked burden information.

Figure 2: Study Burden Heat Map for Dyads Participating in the PediQUEST Response RCT by Social Disparity Level and Study Arm (n=149).

This heat map presents the study burden level of participants. Each box represents a single participant. Dyads are grouped by social disparity level, with parent in the top cell and their child below. The vertical axis separates control and intervention arm participants. Dyads are ordered by burden degree (from higher to lower) from the center outwards. If cell is not colored, indicates no information on burden. Besides degree of burden, each cell indicates interview status and end of study status. If end of study status is not indicated, the dyad completed the follow-up. See references for more details.

Figure 3: Distribution of Study Burden by Social Disparity Level for Parents and Children in the PediQUEST Response RCT.

The figure shows that the distribution of study burden (including no information) varies across social disparity levels for the 149 parents and 84 children included in the analysis (* Fisher’s exact test). Ordinal logistic regression confirms that burden level (no, low/moderate, high) increases with social disparity.

Social Disparity Levels: HM−/EH−, neither historically marginalized nor economic hardship; HM+ or EH+, either historically marginalized or economic hardship; HM+/EH+, historically marginalized and economic hardship.

DISCUSSION

Main findings

This study aimed to investigate parent and child study burden in pediatric advanced cancer while participating in a palliative care RCT. Most dyads finished the study and reported low to moderate burden with no significant differences by arm. However, research participation in the context of advanced cancer involved a non-negligible effort. We identified two burden themes: study-related and daily life stressors. Our results suggest that families from marginalized backgrounds experienced higher study burden than their counterparts and that burden was higher as the risk for social disparity augmented. We show that study burden exists on a spectrum, infrequently reported by researchers, but critical for understanding the full range of experiences and identifying ways to improve participation.

What this study adds

The high proportion of dyads completing follow-up suggests that potential threats to study participation may have been outweighed by potential benefits,^35,36 such as feeling altruistic,^37–39 addressing secondary issues often neglected by the care team,³⁷ or new thinking and reflection.^38,39 Also, threats were possibly minimized by the extensive retention strategies and incentives used,^25,40 which are aligned with existing recomendations.^41,42 However, facing daily life stressors and study demands while dealing with pediatric advanced cancer treatment can be taxing. Thus, it is essential to involve parents and children from design to implementation to ensure that research is pertinent and a personalized worthwhile experience.^43–45

Understanding study burden as a barrier to participation is critical to address inequalities in pediatric cancer research.^46,47 This study uncovers that burden distributes unevenly across socio-demographic groups, with those facing historical marginalization or economic hardship probably bearing the heaviest load. Interestingly, families with social disparities tended to dropout due to unanswered surveys (level 4) rather than because of study burden (level 5). This raises questions about whether these families, who also had lower education levels, felt less empowered to openly leave the study or might have stayed had we posed fewer participation demands. Reducing study burden would benefit participants from marginalized backgrounds the most, improving sample diversity and enhancing the generalizability of results as the US population continues to diversify.^48–50

Patient’s clinical context seems to affect perception of study burden and consequently, participation. Dyads with children with low disease/symptom burden may have stopped answering because the study felt less relevant. Conversely, some families reported that high clinical demands limited their time or willingness to engage in study procedures, leading some to discontinue participation. Since these patients may greatly benefit from palliative care interventions, adopting more pragmatic and flexible approaches regarding the survey component, such as offering alternative administration methods (e.g. telehealth or face-to-face) or allowing parent responses only (by proxy), may help mitigate study burden and improve retention.

Finally, even though a few parents found the SPPC component demanding, the lack of association between burden and study arm suggests that this early palliative care intervention was not inherently burdensome. This finding is supported by existing literature for timely integration of SPPC.^51–53

Strengths and limitations

Although not specifically designed to examine study burden, several strengths stand out. Burden was identified while analyzing process measures. Our decision to report on it reflects an intent to collectively address these challenges. Interviews included participants who dropped out and a bereaved parent, broadening the range of experiences. By integrating qualitative and quantitative data, our mixed-methods captured both participants’ expressions and actions, amplifying the voices of those who did not directly mention burden.⁵⁴

However, the burden degree variable has important limitations. Except for level 5 participants, who explicitly cited burden as their reason for dropping out, the remaining levels were inferred from verbal and behavioral data. We treated burden sources as cumulative (3>1), and assumed survey non-response reflected less burden than dropout-burden. While ordinally consistent, these inferences may not fully capture participants’ perceived burden. Future studies should directly assess burden, acknowledging its subjective nature.

Other study limitations include potential misclassification, selection bias in exit interviews, and constraints related to sample size. HM and EH were researcher-defined categories which may not fully reflect how participants perceived themselves. Also, the underrepresentation of HM+/EH+ families in exit interviews may have limited our ability to capture burden specific to this group. Employing community-based participatory research designs could engage participants in co-developing relevant measures and enhance enrollment and retention rates.⁵⁵ Additionally, the decision to collapse race and ethnicity into a binary variable was driven by sample limitations. While this condensation may blur the complex identities and experiences within these broader categories, possibly leading to omission or overgeneralization,^56–58 it also reflects the structural advantages historically afforded to the non-Hispanic White group.^59–61

CONCLUSION

This study explores sources and degree of study burden and its relationship with social disparities among families of children with advanced cancer. Our findings suggest that race, ethnicity, and economic hardship impact the experience of participating in a palliative care RCT. Future research should explore ways to support and minimize demands^15,39 placed on these already strained families during clinical research participation.

Abbreviations:

EH

economic hardship

HM

historically marginalized

HM−/EH−

neither historically marginalized nor economic hardship

HM+ or EH+

either historically marginalized or economic hardship

HM+/EH+

historically marginalized and economic hardship

PediQUEST

Pediatric Quality of Life and Evaluation of Symptoms Technology

Data Management and Sharing

The data will be made available upon reasonable request from the corresponding author (vdussel@mgh.harvard.edu).