An official website of the United States government
NCBI Bookshelf. A service of the National Library of Medicine, National Institutes of Health.
Kelly D. Getz, PhD, MPH, Julia E. Szymczak, PhD, Farah Contractor, BS, Brian T. Fisher, DO, MSCE, and Richard Aplenc, MD, PhD.
Author Information and AffiliationsPediatric acute myeloid leukemia (AML) requires multiple courses of intensive chemotherapy. Each course of chemotherapy destroys a child's white blood cells, called neutrophils, leaving the child in a state of neutropenia for 2 or more weeks. During neutropenia, the child is at significant risk for life-threatening infection; such toxicities can result in delays in starting the next course of chemotherapy and significant morbidity and mortality. Current supportive care guidelines recommend hospitalization after chemotherapy completion until neutrophil recovery. Although ~70% of US treatment centers adhere to this recommendation, there are little data to support inpatient over outpatient management during neutrophil recovery.
We compared clinical outcomes (aim 1), patient experiences (aim 2), and patient health-related quality of life (HRQOL) (aim 3) between the 2 strategies (outpatient management vs inpatient management) in a nationally representative sample of patients <19 years old with newly diagnosed AML.
For aim 1, we used standardized medical abstraction for patients with AML treated between 2011 and 2019 at 17 centers in the United States. The unit of analysis was chemotherapy courses; the primary exposure was inpatient vs outpatient management, and the primary outcomes were bacteremia and time-to-next-treatment course. Only courses in which the patient met “discharge-eligible” criteria were included. Log-binomial regressions compared bacteremia incidence, and generalized linear regression compared times to next course. Generalized estimating equations accounted for correlation between courses from the same patient.
For aim 2, patients/families at 9 centers were enrolled from November 2015 to February 2017 and underwent semistructured qualitative interviews that were analyzed using a hybrid inductive-deductive modified grounded theory approach.
For aim 3, participants were enrolled from June 2016 through May 2019. Parent-proxies at 14 centers completed a validated assessment of patient HRQOL called the Pediatric Quality of Life (PedsQL) 4.0 Generic Core Scales at the start of chemotherapy and then after neutrophil recovery in a single treatment course. Only “discharge-eligible” patients were included. Follow-up scores on the parent-proxy PedsQL Generic Core Scales were compared by management strategy using an analysis of covariance accounting for baseline scores.
Distributions of sex, ethnicity, diagnosis year, central line type, and individual antibacterial prophylaxis were generally comparable between patients who were managed as outpatients and those managed as inpatients. We used propensity score methods to control for demographic, clinical, and hospital-level factors that we determined to be potential confounders. For aim 1, we found that overall rates of bacteremia were not significantly different in patients receiving outpatient vs inpatient management (23.8% for outpatient vs 29.0% for inpatient; adjusted relative risk, 0.73; 95% CI, 0.56-1.06; P = .082). Additionally, there were no delays to the start of the next chemotherapy course for patients managed in the outpatient setting vs in the hospital.
For aim 2, 86% of families receiving inpatient management expressed satisfaction, and 85% of those receiving outpatient management expressed satisfaction. Dissatisfaction with inpatient management was driven by concerns for hospital-acquired infections and separation from family. Dissatisfaction with outpatient management stemmed from the stress of caring for a neutropenic child at home. Patients/families reporting satisfaction with outpatient management emphasized that the approach would not be appropriate for all families.
For aim 3, the mean (SD) parent-proxy follow-up PedsQL total score did not differ between outpatient (70.1 ± 18.9) and inpatient management (68.7 ± 19.4), with an adjusted mean difference of −2.8 (SD, −11.2 to 5.6).
Our results are susceptible to confounding, which was mitigated by the capture of detailed demographic, diagnostic, clinical, and hospital-level factors to enable control for such bias. Patient HRQOL was only assessed at a single treatment course, which limited the ability to explore potential heterogeneity by course. Although individual details of site-directed protocols for monitoring for infection at home were not captured, differential underascertainment of bacteremia was minimized by capturing results of all cultures drawn during inpatient admissions and outpatient clinic encounters as well as those submitted to treating hospitals by transferring institutions.
Rates of bacteremia during outpatient management were not significantly different from those observed during inpatient management, with no delay to the start of the next chemotherapy course or meaningful differences in patient HRQOL. Semistructured interviews revealed strong alignment between patient/family satisfaction and center discharge practice. However, families experiencing outpatient management noted that this strategy would not be suitable for all families. These clinical and patient-centered results suggest that outpatient management during neutropenia is a viable approach without excess risk for children with AML. However, implementation studies are needed to identify patient/family characteristics that portend a positive experience with an outpatient strategy.
Acute myeloid leukemia (AML) is the second most common pediatric hematologic malignancy, with approximately 600 new cases per year among persons <20 years of age in the United States. Although it accounts for only about 20% of leukemias in children, it is responsible for more than half of pediatric leukemia deaths.1 The prognosis of children with AML has improved greatly over the past 30 years, with complete remission and overall survival rates as high as 80% to 90% and 50% to 60%, respectively.2-4 These improvements are largely attributable to intensification and standardization of chemotherapy regimens.
All pediatric patients with newly diagnosed AML receive multiple consecutive courses of intensive myelosuppressive chemotherapy. Treatment schedules (ie, total number of courses, agents used, and timing of initiation of each course) are planned to maximize the likelihood of achieving and sustaining a remission.5 Each course of chemotherapy destroys a child's white blood cells, called neutrophils, leaving the child in a state of neutropenia for 2 or more weeks. During neutropenia, the child is at significant risk for life-threatening infection, which can result in delays in starting the next course of chemotherapy. Such delays could potentially lead to remission induction failure or a higher likelihood of relapse. Previous reports have found that 57% to 80% of febrile neutropenia episodes among pediatric patients with AML are compromised by at least 1 microbiologically documented infection,6,7 with bacteremia constituting the most prevalent infection.8 These infectious complications remain a major cause of therapy-associated morbidity and mortality in children with AML.9,10
Published pediatric neutropenia guidelines state “there are no validated schemas for defining those patients at high risk of developing complications of fever and neutropenia.” As a result, the guidelines have refrained from making specific recommendations for patients with AML.11 Therefore, clinicians are left to decide between keeping a child with AML in the hospital for close observation until neutropenia resolves, a median time of 35 days from the start of chemotherapy in the given course, and discharging patients to outpatient management within a few days of chemotherapy completion with instructions to return if symptoms of infection develop. Physicians who elect to observe children with neutropenia in the hospital do so under the assumption that hospital observation will reduce the risk of serious infection and thereby reduce delays in starting the next course of chemotherapy. Furthermore, this approach assumes that the potential reduction in infection risk outweighs the potential negative consequences of a prolonged inpatient stay, namely separation from family, reduced quality of life (QOL), increased exposure to multiresistant nosocomial infections, and increased health care cost. There is documented variation in practice across institutions on inpatient vs outpatient management of neutropenia, with approximately 60% of Children's Oncology Group (COG) institutions reporting a policy to always keep patients hospitalized during severely neutropenic periods and the remaining 40% of hospitals reporting a policy of home management some or all of the time.12-14 This variation in practice highlights the need for data on additional clinician- and patient-centered outcomes to appropriately guide the management of neutropenia in pediatric patients with AML.
The limited literature on the clinical consequences of outpatient vs inpatient management of neutropenia in AML is predominantly focused on the experience of adult patients. Adult patients discharged early to outpatient supportive care consistently had shorter cumulative lengths of stay than inpatients.15-21 Additionally, early discharge of adult patients with AML receiving chemotherapy resulted in fewer and shorter febrile episodes,19,22 a better response to first-line antibiotics, and shorter duration of intravenous antibiotic administration.17,19,21,22 Although these adult studies provide some reassurance that outpatient management may be safe and feasible, they are limited, as most included data from only a single institution, had very small sample sizes, or lacked an appropriate inpatient reference population because the studies were not restricted to include only patients who would have been clinically stable enough for discharge.
Published literature evaluating outpatient management of neutropenia in pediatric AML is limited. A single study of only 13 patients from 1 hospital found similar rates of relapse and mortality for outpatient vs inpatient management of neutropenia.23 In our own preliminary analyses based on administrative resource use data from 43 freestanding children's hospitals in the United States, we found that patients who were discharged early to outpatient management following AML induction and intensification chemotherapy courses incurred fewer cumulative days of hospitalization, but were frequently readmitted and had higher rates of antibiotic, vasopressor, and supplemental oxygen use than patients who remained hospitalized during the entirety of their neutropenia.12 In the absence of clinical data and laboratory confirmation, it is unclear whether these observed increases in resource use are an accurate proxy for a greater incidence of infection or more severe infection in the early discharge patients.
There is also a dearth of literature on child and parent/guardian perspectives on outpatient vs inpatient management of neutropenia24-26 or the impact of neutropenia management strategy on QOL in pediatric patients with cancer.27 The single qualitative study on parent preferences for neutropenia management in pediatric cancer found that most respondents preferred hospital-based treatment once a patient developed a fever during neutropenia.28 Factors that influenced parent preference for febrile neutropenia management included convenience and disruptiveness to family life, concerns about the child's physical health (including infection), and concerns about the child's emotional well-being. Although this study was a good first step in identifying patient-centered outcomes for febrile neutropenia management in pediatric cancer, it has some notable limitations: it was conducted at a single center in Canada, children were not interviewed, and it involved the elicitation of parent preferences via hypothetical scenarios rather than more in-depth study of the actual experience of neutropenia management in the hospital or the home.
Decisions regarding supportive care strategies are intended to not only improve clinical outcomes but also to impact patient QOL outcomes, such as minimizing the psychological, social, and spiritual problems related to chemotherapy.29 The limited published data that do exist regarding patient-oriented outcomes relevant to neutropenia management suggest that patients and caregivers may hold different perceptions of and place a different value on the risk-benefit profile of any particular intervention.25,30 Thus, it is also imperative to determine the impact of a chosen neutropenia management strategy on the child's QOL and interactions with his or her broader social environment, including impact on the family.
We aimed to establish important comparative effectiveness data on physician-directed outcomes for outpatient vs inpatient management of AML neutropenia. We also used an innovative mixed-methods design to identify outcomes and perceptions regarding neutropenia management that are important to children and caregivers and subsequently compared these patient-centered outcomes for outpatient and inpatient neutropenia management strategies. Through the 3 specific aims, we sought to provide a more complete risk-benefit profile to inform physicians, patients, and caregivers in their selection of the most appropriate strategy of neutropenia management.
Specific aim 1: To compare the effectiveness of inpatient vs outpatient management of neutropenia in children with AML.
Specific aim 2: To identify outcomes related to the management of neutropenia that are most important to children with AML and their parents/guardians.
Specific aim 3: To compare patient and caregiver reported QOL and other patient-centered outcomes identified in specific aim 2 (ie, patient sleep disruption, parental stress, and financial toxicity) for inpatient vs outpatient management of neutropenia in children with AML.
Figure 1 depicts the specific aims and the numbers of participants analyzed for each.
Specific Aims.
We partnered with several patient and caregiver stakeholder research partners to ensure that the planning, execution, and dissemination of this study and its findings are truly patient-centered. Darlene Barkman, MA, is the co-Chair of the Children's Hospital of Philadelphia (CHOP) Family Advisory Council (FAC) and is a family consultant at CHOP. She was directly involved in the research, contributing to the recruitment plan, the design of interview guides, and the patient-centered outcomes survey. She assisted with the language used in all informed consent documents and semistructured interview guides. She also served as a liaison between our investigative team and the CHOP FAC.
The CHOP FAC leadership facilitated the creation of an AML-specific subpanel of its membership composed of 4 parents and 1 patient who have direct experience with AML therapy. Through a combination of in-person meetings and email correspondence, the CHOP AML FAC shared perspectives informed by their experiences and provided feedback on key documents and study procedures during developmental stages and while the study was ongoing. They also shared many suggestions for disseminating the final research findings. For example, the FAC provided helpful suggestions on how to prepare families for the aim 2 interviews, particularly the separate interviews of the patient from the parent. Specifically, the FAC suggested that we give parents the list of questions beforehand, talk with parents first to set expectations of what the interview would look like, and avoid beginning the patient interviews with medical questions. The team adapted the aim 2 methodology to include these suggestions.
We also partnered with Alex's Lemonade Stand Foundation (ALSF), a large, nationwide 501(c)3 charity that funds research and raises awareness for childhood cancer. We primarily worked with Lisa Towry, Director of Programs and Services at ALSF. In August 2016, Ms Towry helped identify members for the Caregiver Dissemination Board who are parents of children diagnosed with leukemia. They participated by sharing feedback on the best way to communicate, disseminate, and translate our research findings into a format that is accessible and useful to patients and their caregivers. Ms Towry also shared the results of a survey series called My Childhood Cancer on the topic of where families seek information on childhood cancer. We will use the survey results, along with the advice from our other stakeholders, to create a dissemination plan for our study findings.
In March 2016, the team presented the study to the COG Patient Advisory Council (PAC), consisting of children with cancer and their caregivers. At this initial meeting, we presented our study protocol and elicited feedback on the overall study design, recruitment, and data collection methods. The COG PAC had suggestions for aim 2 interview strategies and aim 3 data elements, such as the location of the interview and discharge teaching practices. For example, the PAC strongly encouraged conducting the aim 2 interviews with patients separate from parents. One cancer survivor said that she felt substantially freer to talk about her feelings when her parents were not present because then she did not feel the need to protect them. The PAC also strongly encouraged interviewing patients/caregivers in the home. We incorporated this guidance directly into the design and conduct of aim 2. We kept the COG PAC updated on our study progress via email and, later, elicited their suggestions for dissemination of the research findings.
In June 2018, we began having joint stakeholder meetings, which included the CHOP FAC, ALSF members, and the COG PAC. These meetings resulted in essential feedback for the implementation of aims 2 and 3. All stakeholders shared similar advice on the timing of the approach of families for aim 3 consent and subsequent assessments. They recommended that it would be best to approach families in the afternoons once the morning rounds had been completed, and that evenings might be better because parents tend to be more available then. The stakeholders emphasized that flexibility and following up with families was key to increasing the consent rates and retention in the study. We maintained stakeholder engagement by providing monthly updates on progress via email. We plan to have additional stakeholder meetings to develop a formal implementation study and to further plan the dissemination of our final research findings.
We used a bidirectional observational cohort study design to evaluate the comparative effectiveness of outpatient and inpatient postchemotherapy neutropenia management in children with AML who received frontline treatment at 17 participating institutions across the United States. Data were obtained from standardized abstraction of patients' electronic medical records (EMRs) for both retrospectively and prospectively identified participants.
The source population included patients newly diagnosed with AML who received standard intensive frontline chemotherapy at 17 pediatric institutions across the United States between January 2011 and July 2019. Waivers of consent for medical record abstraction were obtained at all participating sites, allowing for complete capture of the target population. Patients were identified by investigators at each institution from their local cancer registries and prospectively at a weekly oncology case conference where new AML diagnoses are discussed. Patients were ineligible if they had acute promyelocytic leukemia, were >18 years at diagnosis, received nonstandard reduced intensity chemotherapy, or received only a hematopoietic stem cell transplantation at the participating site. Patients contributed data for each of their frontline chemotherapy courses. As indicated previously, standard treatment for newly diagnosed AML involved multiple consecutive courses of intensive chemotherapy in 2 phases. The first phase includes 2 courses of chemotherapy (ie, induction I, induction II) aimed at killing the leukemia cells in the blood and bone marrow to induce disease remission. The second phase of treatment begins after a remission is achieved and includes up to 3 courses of chemotherapy (ie, intensification I, intensification II, intensification III). The goal of intensification is to kill any remaining leukemia cells that may not be active but could begin to regrow and cause a relapse. The study population was identified from the source population as patients who survived and were determined to be eligible for early discharge to outpatient management at the completion of course-specific chemotherapy. Patients were considered eligible for discharge if there was no evidence of fever (ie, temperature of 38.4 °C or higher), microbiologically documented infection (ie, positive blood culture), or intensive care unit–level requirements (ie, vasoactive infusions, supplemental oxygen, or dialysis) within 3 days of the last dose of chemotherapy in the given course. Patients not meeting the discharge eligibility criteria were excluded from the analytic study population in an effort to make the comparison groups (outpatient and inpatient) comparable with respect to baseline infection risk.
The timing of discharge relative to the last day of course-specific chemotherapy was calculated as the difference between the initial date of discharge following chemotherapy completion and the date of the last chemotherapy administration. These data were used to classify patients according to discharge status into 1 of 2 exposure groups (outpatient vs inpatient management). Patients discharged within the 3 days after chemotherapy completion were categorized as early discharge to outpatient neutropenia management. Patients meeting discharge eligibility criteria but remaining in the hospital >3 days after chemotherapy completion were categorized as inpatient management. The threshold for defining the outpatient management group was chosen based on preliminary data on the distribution of timing of discharge relative to the completion of course-specific chemotherapy and represents a period of time within which recovery from neutropenia after AML chemotherapy is improbable. Discharge strategy was determined based on the site standard practice—each participating institution used 1 discharge strategy or another. As such, patient characteristics did not inform the discharge strategy, nor did patients/families have a choice as to which management strategy would be used.
The primary outcomes of interest were course-specific occurrence of bacteremia during postchemotherapy neutropenia and time to the initiation of the next chemotherapy course. Bacteremia was chosen as a primary outcome because it is the most common form of infection in children with AML, and it contributes substantially to chemotherapy-related mortality. Time to initiation of subsequent chemotherapy course was selected as a primary outcome because long chemotherapy delays could potentially result in remission induction failure or a greater chance for relapse. Follow-up for identification of bacteremia began 3 days after course-specific chemotherapy completion and continued until the earliest of neutrophil count recovery, the start of the next course of chemotherapy, 50 days from the start of chemotherapy in the course, or death. The occurrence of bacteremia was defined as a single positive blood culture for a bacterial pathogen unless the bacterium was considered a common commensal organism by the National Healthcare Safety Network. In the case of common commensal organisms, 2 separate positive blood cultures within 3 consecutive days were required for classification as bacteremia.30 Viridans group streptococci were considered as true pathogens, because they are a common cause of bacteremia in patients with AML. Time to the start of the next chemotherapy course was measured as the difference in days from the first day of systemic chemotherapy in the next course and 3 days after chemotherapy completion in the preceding course. For the subpopulation of patients who were discharged early, rates and timing of first readmission relative to the start of follow-up were also computed.
We expected to identify a total of 533 patients newly diagnosed with AML across the participating centers. Assuming 90% of identified patients would be eligible for early discharge, the anticipated study population was 480 patients, of which 27% were expected to be managed as outpatients and the remaining managed as inpatients during neutropenia. Assuming each patient contributes 1 treatment course and a baseline risk of bacteremia of 60%, the proposed size of the 2 study groups would provide 81.3% power to detect a 15% change (increase or decrease) in bacteremia rates among patients managed as outpatients vs those managed as inpatients, with a significance level of α = .05. If we assumed that each patient would contribute 2 courses and a within-subject correlation of 0.15, we would have 96.8% power to detect a 15% difference in bacteremia rates between groups.
For time to next chemotherapy course, we used a 2-sample t test in the power calculation to be conservative (ie, statistical analysis using a generalized linear model with correct distribution will result in increased power over a t test). Assuming each patient would contribute 1 treatment course and an average time to next chemotherapy course of 29 days and an SD of 10 (based on preliminary data), our study would have 83.0% power to detect a difference of 3 days between the 2 groups, with a significance level of α = .05. Assuming each patient would contribute 2 courses and a within-subject correlation of 0.15, then we would have 98.1% power to detect a 3-day difference.
Study personnel from the coordinating institution (CHOP) were formally trained using a detailed medical record abstraction manual developed specifically for the current study by site investigators, and they were required to achieve >95% concordance with their trainer on multiple practice abstractions before commencing study EMR reviews. The trained study personnel traveled to each participating institution to complete the manual medical record abstraction of specific information on demographics (sex, age, race, ethnicity, insurance status at each course); diagnosis details (year of diagnosis, AML subtype, risk classification); treatment information (clinical trial enrollment, course-specific chemotherapy regimens, including dates of administration, central line type); antimicrobial prophylaxis; dates for the inpatient admissions and outpatient clinic encounters; and dates and results for blood cultures performed at the treating institution during inpatient admissions and outpatient clinic encounters, as well as those submitted from outside institutions. All data were entered directly into electronic case report forms. Measurement occurred before the determination of the discharge strategy. Following each round of medical record abstraction, collected data were extracted and evaluated for quality and completeness using programs designed in collaboration with the clinical team to prevent and monitor missing data and to identify potentially erroneous data entries. All generated queries were verified against the patient medical records. The resolutions for each interrogation or confirmation of the accuracy of abstracted data upon verification against the EMR were recorded in the query forms and archived. This standardized approach to medical record abstraction was chosen in an effort to maintain comparable quality and consistency in data collection across all participating institutions and across all years of the study and to avoid differential misclassification often associated with retrospective designs.31 Furthermore, this practice resulted in the complete capture of exposure, outcome, and relevant potential confounders of the associations of interest. All patients were accounted for in the reporting, and the nature and degree of missing information was evaluated by exposure to determine whether analytic methods to account for missing data were required.
Given that the exposure of interest (outpatient vs inpatient management) was determined by institutional practice, there was nearly unavoidable confounding by other site-level practices related to management strategy and the outcomes of interest. In an attempt to mitigate such confounding, we captured detailed information on hospital-level covariates via a survey to site investigators to acquire data on institutional standard practice with respect to infection control (ie, the use of systemic antimicrobial prophylaxis and chlorhexidine gluconate bathing) and central line management (ie, antibiotic and ethanol catheter lock therapy).
Log-binomial regression was used to estimate risk ratios with 95% CIs, comparing the incidence of bacteremia following an outpatient vs inpatient neutropenia management strategy. Before statistical model fitting, the normality of the distribution of the time-to-next-chemotherapy course was assessed. Given that the normality assumption was not violated, traditional generalized linear regression models (normal distribution with an identity link) were used to compare time to next course for outpatient vs inpatient management.
Propensity score analyses were employed to adjust for potential confounding by baseline covariates. Propensity scores were derived from the predicted probabilities estimated from regression models of the use of outpatient vs inpatient management during neutropenia, conditional on all baseline factors determined from bivariate analyses to be true confounders (ie, those associated with both exposure and outcome with P < .2) and those determined to be potential confounders (ie, those associated only with the outcome interests). Patients were then stratified into 5 groups using quintiles of the estimated propensity score. The balance of covariates between outpatient and inpatient management groups was assessed before and after the application of the generated propensity score. Control for confounding was accomplished through adjustment for the propensity score quintiles as well as any remaining unbalanced patient- or hospital-level confounders. All course-specific comparisons employed methods to account for nonindependence of observations from patients from the same institution, specifically general estimating equations (GEEs).
As indicated below, we evaluated the associations across treatment courses to assess for any potential effect heterogeneity by course. Summary measures of the associations across courses were computed using GEE methods with an autoregressive correlation matrix to account for nonindependence between courses contributed by the same patient.
The original research plan for aim 1 included medical record abstraction for only a single chemotherapy course (induction II). In 2015, the protocol was modified to allow for the abstraction of all intensive frontline chemotherapy courses, thus enabling us to evaluate the primary associations of interest at each treatment course and to assess whether observed associations are consistent across courses, which will be important information for clinicians, patients, and families. The protocol was also revised such that the medical record abstractions at each participating institution would be performed by CHOP personnel rather than training staff at each site to do them.
In August 2016, the sample size for aim 1 was changed from 385 to 480 in order to increase our study power, as fewer sites used outpatient management than initially projected. To accomplish this, we recruited 5 sites that confirmed use of outpatient management (University of Mississippi Medical Center, Jackson, MS; Lucile Packard Children's Hospital Stanford, Palo Alto, CA; Children's Hospital of Michigan, Detroit; Seattle Children's Hospital, WA [uses both inpatient and outpatient management]; and St. Jude Children's Research Hospital, Memphis, TN). In December 2017, the aim 1 IRB protocol was modified to allow for the identification of a maximum of 600 eligible patients with AML to accommodate the necessary continued enrollment on aim 3, because aim 1 medical record abstraction is the mechanism by which we ascertained all exposure and covariate information for aim 3 participants.
Study aim 2 was to identify the outcomes related to the management of neutropenia. These outcomes, which are most important to children with AML and their caregivers, could be incorporated into aim 3 to further augment the physician-generated outcomes (bacteremia and time to initiation of next round of chemotherapy) for the comparative effectiveness analysis of inpatient vs outpatient management of neutropenia. To achieve aim 2, we systematically interviewed children with AML (and their caregivers) who were either hospitalized for the duration of their neutropenic period or who were discharged to recover at home. Patient and caregiver participants were enrolled between November 2015 and February 2017.
We recruited children and caregivers who were actively receiving treatment and those who had completed treatment for AML to participate in an in-depth, semistructured interview. We recruited participants from 9 of the study sites (Table 1). Hospital sites were selected based on their geographic location and primary neutropenia management strategy (inpatient vs outpatient).
Hospital Sites From Which Aim 2 Patients and Caregivers Were Recruited.
We purposefully sampled interview respondents at each hospital site to introduce variation by child age, type of neutropenia management strategy experienced (outpatient vs inpatient, dictated by hospital site), and length of time from the completion of chemotherapy. At each site, we recruited children diagnosed with AML and their primary caregivers (eg, parents or grandparents). To be eligible for inclusion, a child and their family had to have received their AML therapy at the hospital site and be within 6 to 12 months of the completion of their second course of chemotherapy or up to 3 years out following the completion of all their frontline AML chemotherapy. Patients were eligible to participate in an interview if they were between 8 and 21 years of age. Parents of children <8 years of age who met the above criteria were also eligible to participate. We excluded the parents of children who had died because a distinct methodological and semistructured interview technique is needed to ensure sensitive, minimally burdensome, and valid data, requirements that were determined to be beyond the scope of the present study.32
Study coordinators at each institution identified all patients who met our inclusion criteria. Then, we requested that a member of the child's care team (eg, primary oncologist, social worker, nurse practitioner) contact the family to introduce the study and to ask whether the family would be willing to be contacted by a CHOP researcher. We allowed each study site to determine the best way to do this (eg, by phone, via email, at a scheduled clinic visit). Families were told that the purpose of the study was to better understand the child and family experience of the neutropenic period. If the parent and child agreed to be contacted by the CHOP qualitative data collection team, the study coordinator provided us with their name and phone number. We contacted the family and invited the child (if eligible) and the parent to participate in an interview at a later date. Questions about the study were answered, and if both parent and child expressed interest, an interview was scheduled. We offered families a choice of either in-person or telephone interviews. In-person interviews were conducted by the CHOP research team at 1-week site visits to each children's hospital. Telephone interviews were also conducted by the CHOP research team at the family's convenience.
In conjunction with one of our patient research partners, Ms Darlene Barkman, we created 2 semistructured interview guides: one for children and the other for their caregivers. We asked a series of open-ended questions to elicit responses with minimal specific prompting. These questions focused on having the child and caregiver reflect, in detail, upon their past experiences with neutropenia management. The goal was to elicit their detailed thoughts about the impact of the child's illness and treatment on family life, beliefs about the risks and benefits of different neutropenia management strategies, preferences related to the implementation of a neutropenia management strategy within the child's overall cancer care, and opinions about which outcomes pertaining to neutropenia management are most important. The interview guide included open-ended questions intended to elicit detailed narratives, thoughts, and perceptions about the experience of neutropenia with minimal prompting (see Table 2 and Table 3 for our interview guide questions for caregivers and children).
Interview Questions for Caregivers.
Interview Questions for Children.
Our interview guides underwent extensive pilot testing before data collection. First, the guides were pilot tested with members of the CHOP FAC. We presented our interview guides to the full FAC membership at one of their monthly meetings and solicited feedback pertaining to question formation, language, and comprehensibility of concepts. We worked with CHOP's Child Life Specialist Team to provide feedback on the child interview guide, to ensure that the language used was developmentally appropriate for a range of children with AML between the ages of 8 and 18 years. Second, the guide was pilot tested with 3 caregiver-child dyads undergoing therapy for AML at CHOP. The interview guides were revised based on these pilot interviews.
Interviews were conducted by coinvestigator Dr Julia E. Szymczak (J.E.S.), a medical sociologist, in-person during our study team's site visit to each hospital site or by a trained research assistant by telephone based on family preference. The interviewer gave the child and parent the choice of whether to be interviewed separately or together. For adolescents, we requested that the child be interviewed separately from the parent, because there is frequently a discrepancy in the reports that parents and adolescents give about the illness experience.6 However, we deferred to the child and parent level of comfort with the interview configuration. The interviewer asked the child and parent if they gave permission for the interview to be audio recorded. Once all open-ended questions were asked, the interviewer asked the respondent a series of demographic questions. At the conclusion of the interview, each respondent (child and parent) was given $25 for participating.
Our approach to sample size adequacy in aim 2 followed best practices in qualitative methodology, which suggest that respondents be recruited until thematic saturation, the state in which increasing sample size is unlikely to produce new insights, is achieved.7 During data collection, we monitored for saturation to determine when to stop recruiting participants by listening to the audio of each interview within a week of its completion and recording themes in a matrix that corresponded to domains of our interview guide.8 Once we had exhausted the identification of new themes in each domain we determined that our sample was an adequate size.
Analysis was performed by 2 coders (J.E.S. and a research assistant) using a hybrid inductive-deductive modified grounded theory approach that occurred in 2 phases.33,34 First, we created a series of codes that corresponded to domains in the interview guide. Second, we read through all the transcripts and identified groups of salient, repeated themes. These themes were defined by consensus and created as codes in NVivo 11 (QSR International). After the code list was developed and refined, the research assistant reviewed all transcripts to determine which codes fit the concepts suggested by the data.
J.E.S. double-coded every fifth interview transcript, and intercoder reliability was assessed. We used NVivo 11 to calculate a coefficient of reliability (coding agreements/total number of coding decisions) for that double-coded transcript. We considered a coefficient of >0.95 to be our cutoff for proceeding with the coding. If our coefficient was <0.95, we reviewed the coding discrepancies and refined our code definitions to ensure reliability. After the second evaluation (10 transcripts), each successive check resulted in a coefficient >0.95. Discrepancies or questions that could not be resolved in this meeting were brought to the larger meeting of the entire study team. Once the entire data set was coded, we performed queries in NVivo 11.
We took many precautions to minimize the bias in our methods for aim 2, recognizing that this aim was not a comparison of treatments and was instead designed as a process to inform the subsequent comparative effectiveness study detailed in aim 3. First, to avoid leading question bias, we designed our interview guide to start with neutral, open-ended questions in order to encourage respondents to express, in their own words, which aspects of neutropenia management are important to them. This was not only important to reduce bias but also to fulfill one of the great promises of qualitative methodology: to identify features of human experience that the investigator did not think to ask about at the start of the research.
Second, we minimized sampling bias by making an effort to include participants with diverse characteristics in our sample. Sampling in qualitative research is driven by the goal of selecting cases that can best help the investigator understand the issue under study. Cases were not randomly selected; rather, they were chosen for in-depth study because they were thought to be information rich and analytically useful. We cannot claim representativeness to the entire population of children with AML and their caregivers across the United States; however, we tried to create a sample with an adequate range of critically important dimensions, including child age, sex, race, socioeconomic status, site of care, and geographic location. We decided to sample children and caregivers who had already had experience with AML therapy and neutropenia management because it allowed us to gather more nuanced and valid insights than if we had asked children and caregivers about what they anticipated they will would about neutropenia management and its impact on their QOL.
We reduced our target sample size from 132 respondents to 60 respondents. This modification was justified because we achieved thematic saturation with the first 37 respondents. Moreover, the reviewers of our original proposal indicated that a sample size of 132 was large, and perhaps excessive, given the focused nature of our research question. The results showing thematic saturation from the 37 respondents confirmed this critique. However, we realized that our studied population was almost exclusively White and was skewed toward higher socioeconomic and educational status. We revised the original protocol to allow for either telephone interviews in addition to in-person (at home or clinic) interviews. We then performed an additional 23 interviews to confirm thematic saturation with a cohort that included diverse patient and family characteristics (with respect to race/ethnicity and socioeconomic status). Given our methodological framework, we determined that 23 interviews were sufficient to capture additional novel insights from this group based on the nature of our research question and insights generated in previous interviews.35
We used a prospective observational cohort design to compare patient health-related quality of life (HRQOL), as well as additional patient-centered outcomes identified through aim 2 as important to patients with AML and their families, including parental stress, patient sleep disturbance, and financial toxicity. Participants were prospectively enrolled as patient-caregiver dyads during treatment for newly diagnosed AML between June 2016 and May 2019. Data were obtained through a series of questionnaire assessments administered during 2 visits within a single chemotherapy cycle. Details on the specific assessments and the timing and frequency of survey administration are provided in the “Study Outcomes” and “Data Collection and Sources” sections below.
Patients with AML (aged 2-18 years) receiving frontline chemotherapy at 15 pediatric hospitals across the United States and their caregivers were included in this study aim. Fourteen of the hospitals were a subset of the 17 institutions contributing to aim 1, plus 1 additional institution that did not contribute to aim 1 due to an ongoing local study with overlapping objectives.
Eligible patients were required to be <19 years of age at initial AML diagnosis, receiving standard intensive AML frontline chemotherapy, and able to read English or Spanish if they were ≥8 years of age. Eligible caregivers were the legal guardian of the patient and were required to read English or Spanish. Patients being treated for relapsed AML, diagnosed with acute promyelocytic leukemia, undergoing stem cell transplantation, or receiving reduced intensity frontline chemotherapy were not eligible.
Ascertainment of potential participants occurred weekly by either study coordinator attendance at the weekly Leukemia Team Clinical Conference or review of the Leukemia Team Clinical Conference minutes. At sites without a weekly Leukemia Team Clinical Meeting, the study coordinator contacted the attending physician or fellows on the oncology service to identify new patients. Once identified, each patient was reviewed against the study inclusion and exclusion criteria using an electronic screening tool. Once it was determined that a patient met an exclusion criterion, a notification of ineligibility was generated, and the reason for ineligibility was saved along with the patient's study ID. Site coordinators approached each eligible patient for consent before the end of chemotherapy in the study course and completed the documentation of consent forms via the electronic screening eligibility tool to record the date of consent, or, in cases of nonconsent, the reason for refusal to participate. Site personnel informed the coordinating center about eligible patients and their consent through biweekly updates and regular email correspondence throughout the week. Regular contact between the coordinating center and site personnel at each institution via a combination of biweekly email updates and monthly calls with personnel from all institutions allowed for consistent monitoring of eligibility and progress of enrolled participants throughout the study. To show our appreciation for the patient's and caregiver's participation, we gave them $25 gift cards after they completed each of the baseline and follow-up assessments, for a total of $50.
The comparators of interest were the same as for aim 1; specifically, early discharge to outpatient neutropenia management vs inpatient management. Information on hospital admissions and discharges was manually abstracted from the EMR (for additional detail, see the “Data Collection and Sources” subsection in the “Specific Aim 2 Methods” section).
The primary outcome of interest was patient HRQOL measured based on parent-proxy responses using the acute Pediatric Quality of Life (PedsQL) 4.0 Generic Core Scales.36 These scales use a 7-day time frame for measurement. The multidimensional assessment includes items in 4 domains: (1) physical functioning, (2) emotional functioning, (3) social functioning, and (4) school functioning. Respondents document responses to each question using a 5-point Likert scale anchored by “never a problem” (0) to “almost always a problem” (4). These scales demonstrate internal reliability acceptable for group comparisons (PedsQL Generic Core Total Scale score, Cronbach α = .93 for parent report).36 These scales have also been validated and shown to be sensitive to change over time in children with cancer.37
The secondary outcomes of interest included (1) scores of difficulty with events faced by parents of children with serious illness as measured by the Pediatric Inventory for Parents-Difficulty (PIP-D) assessment38; (2) scores for patient sleep disruption as measured based on parent-proxy responses using the Sleep Disturbance Scale for Children–Disorders of Initiating and Maintaining Sleep (SDCS-DIMS) domain39; and (3) scores for financial distress as measured by the modified COmprehensive Score for financial Toxicity (COST).40 The PedsQL Generic Core Scales, PIP-D, SDCS-DIMS, and COST assessments have all been validated. The PedsQL, PIP-D, and SDCS-DIMS were specifically validated in pediatric populations/families of pediatric patients with critical illness. The COST assessment was developed and validated in adult patients undergoing cancer therapy. We adapted it for use by parents of children receiving cancer treatment. In addition to these validated measures, we also collected exploratory information on sources of financial difficulties and parenting, emotional, domestic, and financial support.
The PIP-D assesses stress-related difficulty with events faced by parents of children with serious illness across 4 domains: (1) communication (eg, with child, partner, or health care team); (2) emotional functioning (eg, impact of illness on sleeping and mood); (3) child's medical care (eg, carrying out medical regimen); and (4) role functioning (eg, impact of illness on parent's ability to work and care for other children). All items are rated on a 5-point Likert scale from 1 (not at all difficult) to 5 (extremely difficult). Internal consistency and reliability for the PIP was demonstrated to be high (α = .80-.96), and scores were significantly correlated with measures of anxiety and parenting stress, demonstrating construct validity.36
The SDSC-DIMS assesses sleep indices such as latency and duration, night awakenings, and reluctance to go to bed. All items are measured on a 5-point Likert scale from 1 (never) to 5 (always). The SDSC was reported to have high internal consistency among both healthy (α = .79) and sleep-disordered participants (α = .71), as well as high test-retest reliability (r = 0.71).39
The COST assessment includes 11 statements about the financial situation of the caregiver/family in relation to the child's treatment adapted from existing literature. A 5-point Likert response scale is used for the parent to indicate the degree to which they agree with each statement, from 0 (not at all) to 4 (very much). The COST measure has been shown to have internal consistency and test-retest reliability, and scores were found to be correlated with income and psychosocial distress.40
We incorporated these secondary outcome assessments after the completion of aim 2 and rolled them out continuously as they were developed; the IRB protocols incorporating the changes were approved by participating sites. Therefore, not every aim 3 participant was administered all of the secondary outcome assessments.
The study population was anticipated to be 120 patients, with approximately 40% being managed as outpatients (n = 48) and the remaining managed as inpatients (n = 72) during neutropenia. Based on the literature, the expected SD for the PedsQL total score was 20 (on a scale of 0-100) and a minimal clinically important difference in change scores between groups was 7 to 10 points. The power calculation was conducted under the framework of a multivariate general linear hypothesis for general linear models with a significance level of α = .05 (Wilks λ test).41,42 Table 4 shows the minimal effect size and corresponding difference in the mean change scores between groups (outpatient vs inpatient) able to be detected with 80% power for a range of within-subject correlations. If there were no correlations, we anticipated having 80% power to detect an effect size of 0.75, or a 15.0-point difference in the mean change scores between the 2 groups; if the correlation was as high as 0.8, we anticipated having 80% power to detect a smaller effect size of 0.34, or a 6.8-point difference in the mean change scores between the 2 groups.
Aim 3 Minimal Detectable Effect at 80% Power.
Figure 2 illustrates the timing of aim 3 assessments. Baseline PedsQL assessments and the COST assessment were administered at 2 points: at the start of the study-contributed chemotherapy course before the patient became neutropenic (baseline) and again within the period between neutropenia resolution (absolute neutrophil count [ANC], >500/μL) and the start of the subsequent course of chemotherapy (follow-up). The PIP-D and SDCS-DIMS were administered only at the follow-up time point.
Aim 3 Study Diagram.
Information on relevant demographic, clinical, and hospital covariates were obtained using the medical record abstraction process outlined for aim 1, as aim 3 participants were a subset of aim 1 participants. In addition, a brief questionnaire was completed by parents/caregivers at baseline to capture socioeconomic information that would not be available in the medical record.
PedsQL items were reverse scored and linearly transformed to a scale of 0 to 100 such that higher scores reflect better HRQOL. The total score was calculated as the sum of all item-specific scores divided by the number of answered items.36,43 In addition to the total score, a physical health subscore was similarly computed using the physical functioning domain items, and a psychosocial health subscore was created using the emotional and social domain items. Analysis of covariance was used to test the association between neutropenia management strategy and PedsQL scores (and COST scores). Control for confounding was accomplished through adjustment for baseline scores and propensity score quintiles using methods comparable to those outlined for aim 1.
Responses to each of the items on the PIP-D were summed to obtain a total score reflecting the amount of difficulty handling events experienced by parents of children with serious illness. Higher scores indicate greater difficulty and increased pediatric parenting stress. We summed responses to each of the items on the SDSC-DIMS subscale and separately for the COST assessment to obtain a total score.
Generalized linear models were used to compare follow-up PIP-D scores (normal distribution), SDSC-DIMS scores (ɣ distribution), and change in COST scores (normal distribution) for outpatient vs inpatient management. Again, control for confounding was accomplished through adjustment for propensity score quintile and any remaining unbalanced covariates. Balance was assessed before and after application of the propensity score.
In the process of reviewing the listings of retrospective AML diagnoses to be included in aim 1 for each of the collaborating institutions, we found that the projected average number of diagnoses per year provided by some sites in their letters of support were overestimates. Similar to aim 1, these overestimates and the lower than expected use of outpatient management negatively impacted our sample size and power projections. Therefore, we opened the aim 3 protocol at the 5 sites mentioned in the section “Specific Aim 1 Methods” under “Changes to the Original Study Protocol” to expand the source population for outpatient management. Given IRB delays that resulted in slower than anticipated accrual for aim 3, in 2017, we also added 3 institutions that practiced inpatient management, namely Children's Hospital of Colorado, Alfred I. duPont Hospital for Children, and Dana-Farber Cancer Institute, to further increase the likelihood of meeting our sample size target. Allowing patients to contribute any postinduction I frontline chemotherapy course to aim 3 offered additional efficiency gains. Financial toxicity was added as a secondary outcome of interest midway through the aim 3 study period.
Last, there was an extension to the original period of performance to allow for additional time to complete recruitment, and the original research project end date was extended by several months.
A total of 18 institutions contributed to at least 1 study aim: 11 (61%) institutions used inpatient management as standard practice, and 7 (39%) used outpatient management. Institutions that used outpatient management were qualitatively more likely to use antibacterial prophylaxis (57% vs 36%), antifungal prophylaxis (100% vs 82%), and antibacterial bathing (57% vs 45%) compared with institutions that used inpatient management. Although the use of anti-infective prophylaxis varied by institutional neutropenia management strategy, sites that did routinely use prophylaxis generally administered it for the duration of neutropenia in each course. There were no meaningful differences in the use of line lock therapy (18% outpatient vs 14% inpatient), and none of the participating sites routinely used granulocyte colony-stimulating factor (GCSF) prophylaxis.
A flow chart depicting the assembly of the study population is presented in Figure 3. The full abstraction cohort included 610 patients with new-onset AML treated at 1 of the 17 participating institutions. Of these patients, approximately 90.8% (n = 554) met all early discharge eligibility criteria for inclusion in the current analyses for at least 1 treatment course and contributed a total of 1196 postinduction I chemotherapy courses. Induction I was excluded a priori from comparisons of outpatient vs inpatient management, because sites that practiced outpatient management were less likely to use that strategy in the first treatment course given the generally high acuity of patients at initial presentation. Because the standard treatment for pediatric AML during the majority of the study period included only 4 courses of therapy, there were too few intensification III courses to complete course-specific analyses.
Assembly of Final Analytic Study Population (N = 554) From Overall Medical Record Abstraction Cohort of Patients Newly Diagnosed With Pediatric AML.
Figure 4 illustrates the bimodal distribution of the timing of discharge relative to the completion of chemotherapy at each course. A subset of patients were discharged within a few days of their last chemotherapy administration at each course, but most patients remained hospitalized with median times to discharge ranging from 17 to 24 days postchemotherapy completion across treatment courses (Table 5). At each course, a small number of patients (ie, 2.3% to 6.5% across courses) remained hospitalized through the start of the next course. Rates of early discharge to outpatient management varied by course, ranging from 15.9% to 27.8%. Readmission rates following early discharge to outpatient management were high and ranged from 73.2% to 85.7% across courses, with median times to first readmission ranging from approximately 7 to 9 days.
Distribution of Timing of Discharge Relative to the Last Day of Chemotherapy for Each Course.
Days to Initial Discharge, Frequency of Early Discharge, and Rate and Timing of Readmission After Early Discharge by Treatment Course.
Table 6 presents the course-specific distributions on patient demographic, clinical, and hospital characteristics by outpatient and inpatient management strategies. Standardized medical record abstraction resulted in near complete ascertainment of relevant covariate information, with the exception of a small number of patients (<5%) who did not have their race or course-specific insurance documented in the EMR; these factors were included in propensity score construction as a separate category.
Baseline Demographic, Clinical, and Hospital-Level Characteristics for Outpatient vs Inpatient Management for Course-Specific Study Populations.
In general, patients discharged early to outpatient management were less likely to be <2 years of age, high-risk patients (induction II), receiving Pneumocystis pneumonia (PCP) prophylaxis, or treated at institutions that practiced line lock therapy as standard of care compared with those that managed inpatients. They were also more likely not to be Black, White, or Asian race; to be treated in a St. Jude trial; to be publicly insured or have undocumented insurance status; or to be treated at an institution that used chlorhexidine bathing as standard of care.
Course-specific rates of bacteremia during postchemotherapy neutropenia increased over the course of frontline therapy, ranging from 21% during induction II to 43% during intensification II. Although there were no statistically significant differences in the risk for bacteremia in patients managed in the outpatient vs inpatient setting, those managed as outpatients consistently had a qualitatively lower rate of bacteremia across courses (Table 7). Specifically, outpatient management was associated with a 15%, 24%, and 26% lower risk for bacteremia relative to inpatient management during induction II, intensification I, and intensification II, respectively.
Comparisons of the Incidence of Bacteremia During Postchemotherapy Neutropenia for Outpatient vs Inpatient Neutropenia Management.
Times to next course of chemotherapy ranged from 31 to 40 days overall. Adjusted differences in time to next chemotherapy course were small and varied by course, ranging from 3 days shorter in induction II (mean difference = −3.1; 95% CI, −5.2 to −1.0; P = .003), 1.5 days shorter in intensification II (mean difference = −1.5; 95% CI; −12.8 to 9.8; P = .792), and 1 day shorter in intensification I (mean difference = −1.0; 95% CI; −4.2 to 2.1; P = .514) for outpatient management compared with inpatient; only the difference in induction II was statistically significant (Table 8).
Comparisons of Time to Next Chemotherapy Course for Outpatient vs Inpatient Neutropenia Management.
We interviewed 86 respondents from 57 families with a child with AML. Of the 57 patients in our study, 39 were cared for as inpatients while neutropenic, and 18 were managed at home. Of the 18 managed at home, 2 stayed at the Ronald McDonald House during the neutropenic period. Three patients experienced relapse. We interviewed 54 parents (44 mothers, 8 fathers and 2 grandparents) and 32 children. The mean (SD) age of the patient was 14.4 years (6.54 years) and 27 of the patients (47%) were male. Interviews ranged in length from 10 to 118 minutes, with a mean (SD) length of 39.6 minutes (22.15 minutes). There were 25 parent-child dyads in our study and 2 triads composed of interviews with both parents and the child. Within these dyads and triads, 21 children requested interviews separate from their parents.
Here, we present the results of our primary analysis, identification of novel patient- and caregiver-centered outcomes repeated across the sample (see the “Novel Patient- and Caregiver-Centered Outcomes” section below) and a detailed subanalysis that describes patient and caregiver perceptions of their experiences being hospitalized or being at home while undergoing neutropenia management (see the “Patient and Caregiver Perceptions of Their Experiences During Neutropenia” section below).
We identified 3 novel patient- and caregiver-centered outcomes related to neutropenia management in pediatric AML through our qualitative data: impact of hospitalization on the patient's siblings, parent anxiety, and child sleep quality (see Table 9 for distribution of repeated themes).
Themes Related to Neutropenia Management.
These outcomes were derived from primary themes that emerged across respondents, including the following:
Based on this analysis, we determined that we needed to incorporate measures that captured the impact of the neutropenic period on parental distress and financial difficulties and on the patient's sleep disruption in aim 3.
Table 10a and Table 10b summarize the perceptions described in this section.
Perceptions of Neutropenia Management Strategy: Inpatient Management.
Perceptions of Neutropenia Management Strategy: Outpatient Management.
A total of 57 respondents (32 parents and 25 children) who experienced inpatient management expressed satisfaction with being hospitalized during neutropenia. The primary reason why these respondents reported being satisfied is that they perceived the hospital to be the safest place for a child with AML during neutropenia. Both parents and children attributed the safety of the hospital to close proximity to emergency care, the reassurance of constant medical surveillance, and a reduced risk of infection. There was a perception among this group that children with AML could become acutely ill without much warning. As such, timely access to emergency care was identified as a priority:
At the hospital he got the help he needed. If things go downhill. There was a point where he was done with his chemo and his ANC went down as expected and then started climbing up. He was feeling just fine. He was eating. He was drinking. He was playing. And then <snap> a fever. And he needed a cooling blanket and wool socks and oxygen. He went to bed one night totally fine and woke up with a 104° fever just a few hours later. It's just that quick, within 4 hours quick. And he was on the verge of going to the ICU. So yeah, now think of your child being at home, sitting on the couch and that happens? Do you call an ambulance? Do you get in the car and run stop signs to get to the hospital? Or would you rather already be at the hospital? — Father (inpatient management)
Forty-eight respondents (30 parents and 18 children) who experienced inpatient management reported that they felt there was a lower risk for infection in the hospital than at home; in fact, 10 parents described the hospital environmental using the word “sterile.” In contrast, only 8 mentioned the risk of hospital-acquired infection. Thirty-two parents whose children were cared for as inpatients explained that at home it was difficult to control “inputs” that could put their child at risk for infection such as school-aged children, visitors, and the behavior of pets. As a result of the perception that the hospital was a safe environment, 28 parents described feeling that, in general, being in the hospital during their child's neutropenic period is less stressful than being at home. These parents envisioned that caring for a neutropenic child at home would be anxiety ridden:
And I think it would just stress me out to be home and be like, “Is that normal? Should I call somebody? Should I not?” At what point do you become an annoying mother who calls too much, vs, there's really something wrong. You know what I mean? — Mother (inpatient management)
Respondents who were satisfied with inpatient management said that while being in the hospital for long periods of time could be difficult, there were positive aspects, including engagement with child life staff who provided fun activities; the development of meaningful relationships with clinical caregivers, which many respondents described as being “like family”; and the importance of hospitalization to the development of a sense of patient identity. All 27 children we interviewed who experienced inpatient management described fond memories of humorous stories, pranks, and creative games that they developed to pass the time. For children whose home lives were chaotic, the hospital was described as calm and restful:
Nine months, yes ma'am. Children's Hospital, I was at Children's Hospital for 9 months. The first month or 2, maybe the first month and a half, it wasn't—it didn't feel like home, but once I got used to it, you know, being around the environment and the sweet people and the nurses and doctors and everything and knew around the place, it was okay. It felt like I was home. It felt like I was around family, like anything I needed they got me and anything I wanted they got me … I stopped noticing that I was in the hospital and I just felt like I was at home and, you know, like I was around sweet people, like I don't have to worry about yelling anymore, I don't have to worry about people lying, I don't have to worry about nothing bad at all. It was all good vibes and good things. — Patient, 17 years old (inpatient management)
The 9 respondents dissatisfied with being hospitalized during neutropenia included 7 parents and 2 children. The primary reason the 2 children gave for why they wished they could have been home was separation from family and friends and feeling a lack of freedom. Of the parents who expressed dissatisfaction, 6 were employed in health care–related positions. The primary reason these parents gave for being dissatisfied with their child's hospitalization was infection. Six parents described infections their child acquired in the hospital that they believed would have been avoided if they had been at home.
We interviewed 5 children of the 7 parents who said they would have preferred to be home. Four out of those 5 children disagreed with their parent that being home was preferable to being hospitalized. The children perceived their home environment differently, describing a high risk of infection or a stressful, chaotic environment:
To be honest, after going through it all, I actually would've rather stayed at the hospital, because the noise at home, after being perfectly quiet in the hospital room and the only thing happening was the beeping of the pump. That doesn't match how many kids are in our house yelling and hitting each other daily. So, I would actually kind of rather stayed—because the noise itself just caused such a lot of stress for me, because I just hated that, compared to the nice, perfectly quiet hospital room. — Patient, 9 years old (inpatient management)
Seventeen respondents (12 parents and 5 children) who experienced home management described satisfaction with their experience. These respondents suggested that the primary benefit of being home was the psychosocial advantage conveyed to the child. Specific features of being home that were seen to confer these benefits included the ability of the child to sleep in their own bed, ability to eat home-cooked foods, having something to look forward to during chemotherapy, having an increased sense of privacy, minimization of family separation, and reduced boredom.
Thirteen respondents (11 parents and 2 children) who experienced home management and were satisfied described the benefits of being home with the caveat that it may not be right for every family. Respondents who felt comfortable being home during neutropenia explained that they felt home management worked for them because they had certain circumstances that made it a successful experience:
If we would've had older kids, I would've really been worried. I think she probably would've gotten sicker, because older kids are going to be going to school, they're going to be bringing those germs home, so we didn't have any other kids that we exposed her to and think that that helped. But I can imagine for families that have older kids or also have kids in school that maybe staying in the hospital is the only option. But staying in the hospital for weeks on end, especially with a toddler, is very difficult. — Mother (home management)
First, all 11 parents satisfied with home management described the presence of extended family willing to help them manage the challenges of being home during neutropenia. For example, in 1 multichild family, the grandmother moved into the family's home during the child's neutropenic period to assist with managing the patient's siblings. This was described as helpful in minimizing the risk of spreading infection between siblings in the home.
Second, 7 parents and 1 child described how they felt comfortable being home during neutropenia because they believed that their care team trusted them to provide safe care at home, to quickly react if there was a change in the child's condition, and to follow instructions to minimize risk. Parents who discussed this sense of trust believed that it accrued over time as their child's care team observed the family and their interactions.
Third, 6 parents suggested that they felt especially comfortable performing medical tasks such as central line care or monitoring for acute changes in status because they or a close family member worked in health care–related positions. These respondents had backgrounds in medicine, nursing, medical assisting, and the pharmaceutical industry.
Fourth, all 11 satisfied parents and 2 children explained that they felt comfortable at home because they were confident in the line of communication that they had with their care team. Specific features of good communication included knowing the clinician on the other end of the line, feeling confident that concerns of any type would be taken seriously, and trusting that the care team had round the clock availability.
The 3 respondents dissatisfied with outpatient management were mothers with limited financial means, who spoke English as a second language, had no caregiving support at home, had other young children, and/or had limited confidence in their ability to safely care for their child. Their experience of the neutropenic period was characterized by anxiety about preventing infection, managing central line care, and being able to get to the hospital quickly in case of an emergency:
I would have much rather been in the hospital the whole time. I loved it. I loved it because it was stressless. It's better than being home ‘cause when I was home I would always shiver – I was so scared. But when I was there, I never shivered. When I was home I didn't get any help and I couldn't sleep ‘cause I was always watching her. That's why I liked going to the hospital more. I could sleep in the hospital. And that was on a couch. I'd find that room to be like a castle. Like, I could get some rest. And when they said it was time to go home, I'd be happy a little ‘cause of my other kids, but at the same time, from inside I'm not. — Mother (home management)
We interviewed the child of one of these mothers, who disagreed with her about preferring to be in the hospital. He expressed more confidence than his mother in managing the risks of being home and less fear about bad things happening:
Mom would have wanted a live-in nurse if we couldn't stay in the hospital. For me, it wasn't much of a problem. Eventually, after the first 3 weeks, we got the hang of it. It was just bam, bam, bam, fall asleep, bam, bam, bam, fall asleep, stuff like that. She was a little bit cautious. But living at home, you can do anything again.” — Patient, 17 years old (home management)
When asked what they would pick if given the choice again, respondents stated a clear preference for either inpatient or home management. Of those respondents who experienced inpatient management, 57 (86.4%) said they would choose to be hospitalized. Of those who experienced home management, 17 (85%) said they would choose to go home. In describing their preferred strategy, many respondents gave a nuanced explanation that decision-making about care during neutropenia for pediatric AML is complex and that the “right” choice depends on the specific circumstances, characteristics of the child, and priorities of each individual patient and their family.
Ultimately, all parents said that their primary concern was survival and that any impact on their or their family's lives apart from that would be tolerated in order to ensure their child had the best outcome. In reflecting on the idea of choice surrounding neutropenia management, we found that 15 respondents who were satisfied with inpatient management expressed apprehension that in the future, children with AML would be required to go home, suggesting that it was not fair, and they expressed gratitude that they were not made to go home:
When given the option in the very beginning to go home for 3 or 4 days, my first question was, “Can we just stay? We'd rather do that.” We didn't want to take this fragile flower home, and have the petals fall off on the way there because we've never taken care of a cancer patient. We were so grateful we got to stay here. — Father (inpatient management)
Figure 5 depicts the study population completing each of the primary and secondary assessments. A total of 223 eligible patient-caregiver dyads were identified. In total, 154 (69.1%) were consented and of those 120 (77.9%) completed baseline and follow-up parent-proxy PedsQL patient HRQOL assessments.
Assembly of Final Analytic Study Population From Overall Medical Record Abstraction Cohort of Patients Newly Diagnosed With Pediatric AML.
Of the 120 participants who completed both baseline and follow-up assessments, 97 participants (81%) met all early discharge eligibility criteria to be included in the current analyses: 22 participants were discharged early to outpatient management (22.7%), and 75 participants (77.3%) were managed as inpatient. Table 11 presents the distribution of patient and caregiver characteristics for the study population by management strategy. Again, patients discharged early to outpatient management were less likely to be <2 years of age. They were also more likely to be non-White race, to be treated in a St. Jude trial, to be publicly insured or have undocumented insurance status, and to have a caregiver with a high school education or less.
Demographic, Clinical, and Hospital-Level Characteristics for Outpatient vs Inpatient Management by Course for the Full Aim 3 Study Population.
Baseline PedsQL scores were low for all patients with AML, regardless of discharge strategy. Although scores were qualitatively higher (better HRQOL) for outpatient than inpatient management, differences were not statistically significant (Table 12).
Comparison of Baseline PedsQL Scores for Outpatient vs Inpatient Management.
In addition, follow-up scores accounting for baseline score did not differ significantly by management strategy, with mean differences ranging from 2.2 to 2.8 points lower for outpatient management (Table 13).
Crude and Adjusted Comparison of Follow-Up PedsQL Scores for Outpatient vs Inpatient Management.
Of the 106 participants who completed the SDSC-DIMS, 84 (80.2%) met all early discharge eligibility criteria to be included in the current analyses: 22 were discharged early to outpatient management (25.9%), and 62 (74.1%) were managed as inpatient. Table 14 presents the distribution of patient and caregiver characteristics for this aim 3 subpopulation by management strategy. The majority of participants were in their second course of AML chemotherapy, regardless of discharge strategy. Again, patients discharged early to outpatient management were less likely to be <2 years of age or to have siblings. They were also more likely to be non-White race, to be treated in a St. Jude trial, and to have a caregiver with a high school education or less.
Demographic, Clinical, and Hospital-Level Characteristics for Outpatient vs Inpatient Management by Course for the Aim 3 Subpopulation With SDSC-DIMS Assessments.
Patients discharged early to outpatient management during neutropenia had significantly lower SDSC-DIMS T-scores than patients managed in the hospital (Table 15). In addition, a larger proportion of patients managed in the hospital had T-scores meeting the threshold for pathological disorders of sleep initiation and maintenance compared with those receiving outpatient management (22.2% vs 9.1%; P = .220).
Crude and Adjusted Comparison of SDSC-DIMS T-Scores for Outpatient vs Inpatient Management.
The 85 participant caregivers (22 outpatients, 63 inpatients) who completed the SDSC-DIMS assessment also completed the PIP-D assessment of stress-related difficulty with events faced by parents of children with serious illness. Overall, parents of children managed as outpatients reported lower levels of parental stress based on total PIP-D scores than did parents of those managed as inpatients, but the difference was not statistically significant (Table 16).
Crude and Adjusted Comparison of PIP-D Scores for Outpatient vs Inpatient Management.
Of the 63 participants who completed the COST financial distress assessment, 48 (76.2%) met all early discharge eligibility criteria to be included in the current analyses; 8 (16.7%) were discharged early to outpatient management, and 63 (83.3%) were managed in the hospital. Table 17 presents the distribution of patient and caregiver characteristics for this aim 3 subpopulation by management strategy. Patients discharged early to outpatient management were less likely to be female, <2 years of age, or White race. They were also more likely to be treated in a St. Jude trial and to have a caregiver with a high school education or less.
Demographic, Clinical, and Hospital-level Characteristics for Outpatient vs Inpatient Management by Course for the Aim 3 Subpopulation With Modified COST Assessments.
Financial distress was similar for outpatient and inpatient management as evidenced by comparable mean COST scores both at the start (baseline) and end of the study-contributing treatment course (follow-up) (Table 18).
Comparison of Baseline and Follow-Up Modified COST Score for Outpatient vs Inpatient Management.
Likewise, there was no significant difference in the change in parental financial distress from the start of the treatment course (baseline) to neutrophil count recovery (follow-up) between patients discharged to outpatient management and those managed in the hospital (Table 19). That said, >95% of caregivers of children being treated for AML noted financial difficulties regardless of discharge strategy (Table 20). Although difficulties buying food, paying automobile expenses, paying for household utilities, and saving for the future appeared similar for the 2 discharge strategies, difficulties paying for mortgage/rent were more commonly reported for outpatient management than inpatient management (59% vs 25%, P = .008) (Table 21).
Comparison of Baseline and Follow-Up Modified PIP-D Score for Outpatient vs Inpatient Management.
Crude and Adjusted Comparison of Change in COST Scores for Outpatient vs Inpatient Management.
Summary of the Prevalence of Financial Difficulties Experienced by Families of Children Treated for AML, and Crude Comparison by Outpatient vs Inpatient Management.
Our mixed-methods patient-centered outcomes research study found that pediatric patients with AML who were discharged early to outpatient management during neutropenia did not have higher rates of bacteremia during course-specific periods of postchemotherapy neutropenia, and despite frequent readmission, had no delay in progression to subsequent treatment courses compared with those treated with inpatient management. Although patient HRQOL was low for both groups, it did not differ significantly between those in outpatient and inpatient management. Furthermore, overall parental distress in caring for a child with acute illness was not worse among those managed at home. Yet, while the overall degree of financial distress during treatment was similar for the 2 management strategies, there were specific financial difficulties that appeared to differentially impact families experiencing outpatient management. Qualitative interviews revealed that 85% of families who experienced outpatient management and 86% of families who experienced impatient management were satisfied with the neutropenia management strategy that their child received. As with financial difficulties, the reasons of dissatisfaction differed between the 2 groups, with those dissatisfied with outpatient management highlighting limited financial means and social support. Together these findings suggest that outpatient management during neutropenia following AML chemotherapy may be safe and feasible, but that implementation studies are needed to identify the patients and families that would benefit from outpatient management and the resources required to support successful management at home.
Our results align with a few small studies in pediatric AML populations. One pediatric study found similar rates of mortality for outpatient vs inpatient management of neutropenia following AML induction chemotherapy.23 Another study evaluating outpatient vs inpatient AML intensification chemotherapy found fewer days of febrile neutropenia and less antimicrobial use among outpatients than among inpatients.44 It should be noted that these studies had many limitations that may have resulted in biased associations: most included data from only a single institution, had small sample sizes, lacked an internal reference or used an inpatient reference population that included patients too ill to be discharged, or did not adjust resource use comparisons for differences in duration of hospitalization between compared groups.
Consistent with the current results, our prior multicenter analysis of administrative data from the Pediatric Health Information System (PHIS),12 which accounted for the limitations of these previous studies, demonstrated that pediatric patients with AML who were discharged early to outpatient neutropenia management following induction and intensification chemotherapy courses had similar course-specific survival and no apparent delay to subsequent chemotherapy compared with patients managed in the hospital. However, patients who were discharged to outpatient management had higher rates of antibiotic, vasopressor, and supplemental oxygen use when readmitted to the hospital for suspected bacteremia than patients who remained hospitalized. Taken together with our current results, outpatient management may have a similar or lower overall incidence of bacteremia as inpatient management, but the severity of the infections may be greater.
Furthermore, our results, with respect to qualitatively worse sleep disruption in the inpatient setting, are consistent with the work of others who demonstrate that persistently elevated sound levels and abrupt increases in sound intensity throughout the night are not conducive to restful nighttime sleep and may serve as an additional source of physiologic and psychologic stress for hospitalized children with cancer.45
The current study evaluated both clinical and patient-centered outcomes in the same study population, thus maximizing the validity and granularity of the resulting risk-benefit profile. Such an accurate risk-benefit profile will be useful and effective in informing physicians, patients, and caregivers in their selection of the most appropriate strategy of neutropenia management. The mixed-methods approach allowed patients and families to directly contribute to the research their vivid firsthand accounts of experiences during and perspectives following AML therapy—voices that are largely absent from clinical research. In addition, these accounts informed assessments that may not have otherwise been implemented in the prospective phase of our study. The value and importance of seeking insight directly from patients/families is underscored by the fact that the outcomes identified through the qualitative interviews, namely patient sleep disturbance and parental distress, were 2 domains that may be most impacted by choice of discharge strategy. Furthermore, the families provided a rich pragmatic context within which to interpret our quantitative findings and better plan for implementation. Our analyses were conducted using a large nationally representative cohort of pediatric patients with AML, using standardized data collection directly from the EMR and thus increasing the generalizability and robustness of our results over small, single-institution studies. Finally, unlike most previous evaluations, we restricted our comparison of outpatient and inpatient management to include only patients with AML who met clinical eligibility criteria for early discharge at the time of completion of course-specific chemotherapy. In the absence of this restriction, the inpatient reference population would likely include patients with poorer baseline risk than those discharged to outpatient management, which would bias comparisons in favor of outpatient management.
Despite the strengths of the study, some limitations should be acknowledged. First, discharge strategy was determined based on each site's standard practice; therefore, this exposure of interest did not vary within a given institution. As a result, there was the potential for unavoidable confounding by other site-level practices that would be associated with exposure and outcome. We captured and controlled for hospital-level practices with respect to infection control and central line management, and we used statistical methods to account for hospital-level clustering. While we believe this would have controlled for much of the site-level confounding, there is a possibility of residual confounding by unmeasured individual-level factors. For example, the additional data collected in aim 3 reveal that the distribution of parental educational attainment, income, and family size differed by discharge strategy. Specifically, those who experienced outpatient management were more likely to have lower education and a lower annual income than those who experienced inpatient management, yet these covariates were not measured in aim 1, as they are not routinely captured in the medical record. The direction of the potential confounding by these factors on the association between outpatient management and bacteremia occurrence would depend on their associations with infection. If poorer, less educated, and larger families also have a higher baseline risk for infection, then the observed point estimate suggesting a potential protective association between outpatient management and bacteremia may be biased toward the null, and a corrected point estimate would suggest a larger protective effect. In contrast, these factors would have to be associated with early discharge and bacteremia in opposing directions with minimum relative risks of 2.8 (E-value = 2.8) and 0.36 (inverse of E-value)46 to fully explain away the borderline significant observed overall association between early discharge to outpatient management and bacteremia occurrence.
Second, while clinical outcomes were evaluated across all courses of frontline chemotherapy, patient HRQOL, parental distress, and financial toxicity were only assessed at a single treatment course. Thus, it is possible that there may be some heterogeneity by course or temporal trends that we were unable to observe. Relatedly, baseline for the prospectively evaluated patient-centered outcomes was defined as the start of the treatment course that the participant contributed to the study rather than at initial diagnosis. Therefore, initial measures of HRQOL and financial distress may not be reflective of their status before the start of AML treatment, and, assuming they would have worsened from before the cancer diagnosis, the observed changes in these measures may be underestimated. Further, the PedsQL assessment measures HRQOL over the previous 7 days, which may introduce modest measurement error.
Third, although the medical record abstraction was quite extensive, feasibility prevented the acquisition of daily supportive care requirements. Thus, while outpatient management does not appear to be associated with an increased incidence of bacteremia relative to inpatient management, there may be differences in the severity of the infections. Given the importance of understanding both occurrence and severity, we plan to merge the established cohort with an administrative data resource, which includes data from the majority of participating institutions to obtain daily resource use data.
Fourth, medical record abstraction was performed only at the treating institution; therefore, we may be missing information on admissions to another institution. This pattern of missing data is likely differential, leading to a greater underestimate of resource use in the discharged patients, and it could explain a modestly lower risk for bacteremia with outpatient management. However, the medical record abstraction did capture whether a patient was seen at an outside hospital; such emergency department visits or admissions occurred infrequently, suggesting that the likelihood of bias is low. Likewise, while we did not capture individual details of site-directed protocols for monitoring for infection at home, these invariably included frequent outpatient clinic visits each week during neutropenia, which we captured in the detailed medical record abstraction, and presumably included instructions to return to the treating institution with onset of fever. We minimized a differential underascertainment of bacteremia by capturing results of all cultures drawn during inpatient admission and outpatient clinic encounters as well as those submitted to treating institutions by transferring institutions. Confidence intervals for the associations between outpatient management and bacteremia occurrence were moderately wide, suggesting our study may have been underpowered to detect such differences.
The aim 2 study had limitations as well. Although we included respondents from a range of pediatric hospitals, the use of a qualitative design limits the generalizability of our findings. It is possible that our respondents possessed systematically different characteristics that influenced their willingness to participate compared with those who declined to participate. We did not gather information about those who declined to participate, so we cannot assess selection bias. Further, we cannot exclude unmeasured differences affecting participation decisions related to specific positive or negative experiences during treatment. Last, although we captured perspectives of patients and families, we did not capture clinician perspectives, which will be important to ascertain in order to maximize success and satisfaction in implementing outpatient management more broadly.
Plans for future research include merging our study data with PHIS to assess intensive supportive care requirements by discharge status as a proxy for infection acuity. We are also in the process of geocoding the residential addresses for participants, which will allow us to assess whether differences in the distance to treating institution mediate any potential differences in the severity of bacteremia for patients managed as outpatients. Linkage of geo-referenced data will also allow for “theme mapping” that could potentially identify threats to generalizability or potential subpopulations that will require adaptation in broader implementation of outpatient management.
Data from all 3 study aims indicate that a formal implementation study to evaluate broader implementation of outpatient neutropenia management is warranted. Effective implementation may decrease bacteremia risk, improve sleep disorders, and decrease family stress. However, satisfaction generally aligned with the neutropenia management strategy each family experienced, and most institutions practice strictly inpatient management. Thus, successful implementation will require changing institutional perspectives by identifying barriers to clinicians transitioning patients to outpatient management, accurately communicating associated risks and benefits, and establishing clear and distinct approaches to management both for patients eligible for discharge and those who are ineligible. Our results additionally demonstrate that considerations of discharge eligibility must go beyond physiologic measures of clinical status to include evaluations of family financial context and available sources of support.
Outpatient neutropenia management for patients with pediatric AML was not associated with higher bacteremia rates compared with inpatient management or with delays to the start of the next chemotherapy course. Additionally, there were no meaningful differences in patient HRQOL between the 2 strategies, and a possible decrease in the occurrence of disorders of initiating and maintaining sleep for outpatient compared with inpatient management. Additionally, the assessments of parental stress and financial distress provided highlighted areas that may require additional attention when deciding on the most appropriate management strategy. Semistructured interviews revealed strong alignment between patient/family satisfaction and center discharge practice. However, families experiencing outpatient management noted high stress in caring for profoundly neutropenic patients at home, and they noted that this strategy would not be suitable for all families. These clinical and patient-centered results suggest that outpatient management during neutropenia is a viable approach without excess risk for children with AML. However, implementation studies are needed to identify patient/family characteristics that portend a positive experience with an outpatient strategy.
First and foremost, we are grateful to the patients and families who contributed valuable data and insights to this research. We also gratefully acknowledge the support of the principal investigators at each of the 17 participating children's hospitals, including Drs. Anupam Verma, Maria Gramatges, Staci Arnold, Rajen Mody, Naomi Winick, Amir Mian, Ellen Morgan, Meret Henry, Catherine Aftandilian, Jeffrey Rubnitz, Jennifer Wilkes, Kelly Maloney, Emi Caywood, Kira Bona, Anderson Collier, Craig Lotterman, and Jennifer Yu. We further acknowledge the efforts of the nursing and other allied health professionals who contributed to this study, and Rachel Madding and Mitchelle Matesva who contributed to cross-site study coordination and data collection. Last, we acknowledge the medical student abstractors who traveled to sites to complete the required medical record abstractions.
Research reported in this report was funded through a Patient-Centered Outcomes Research Institute® (PCORI®) Award (#CER-1409-22827). Further information available at: https://www.pcori.org/research-results/2015/comparing-chemotherapy-recovery-home-versus-hospital-children-acute-myeloid
Getz KD, Szymczak JE, Contractor F, Fisher BT, Aplenc R. (2021). Comparing Chemotherapy Recovery at Home versus in the Hospital for Children with Acute Myeloid Leukemia. Patient-Centered Outcomes Research Institute (PCORI). https://doi.org/10.25302/01.2021.CER.140922827
The [views, statements, opinions] presented in this report are solely the responsibility of the author(s) and do not necessarily represent the views of the Patient-Centered Outcomes Research Institute® (PCORI®), its Board of Governors or Methodology Committee.
This book is distributed under the terms of the Creative Commons Attribution-NonCommercial-NoDerivs License which permits noncommercial use and distribution provided the original author(s) and source are credited. (See https://creativecommons.org/licenses/by-nc-nd/4.0/
Your browsing activity is empty.
Activity recording is turned off.
See more...
