An analysis of the resource use and costs of febrile neutropenia events in pediatric cancer patients in Australia

Febrile neutropenia (FN) in children with cancer generally requires in‐hospital care, but low‐risk patients may be successfully managed in an outpatient setting, potentially reducing the overall healthcare costs. Updated data on the costs of FN care are lacking.


INTRODUCTION
Febrile neutropenia (FN) events are a common complication of treatment of childhood cancer.These events are typically managed in-hospital using intravenous antibiotics, with the duration depending on blood culture results, fever resolution, a rising neutrophil count, and underlying cancer treatment status. 1 Available data indicate unplanned admissions due to FN represent a high burden in terms of their impact on patient and caregiver quality of life 2 and an increased use of healthcare resources. 3tients experiencing an FN event, including in the pediatric setting, are heterogeneous and have different risk factors impacting prognosis and outcomes. 1This has led to the development and validation of several clinical decision rules (CDRs) that identify patients who are at low risk of developing severe infections or complications. 4Overall, the aim of these CDRs is to reliably guide clinicians to determine whether a patient with FN may be managed at home, thereby reducing hospital length of stay (LOS) or avoiding admission alltogether.This recognizes that hospitalization itself poses a risk for nosocomial infections, that the use of repeated exposure to broad-spectrum antibiotics increases the risk of antibiotic resistance, prolonged illness, and unfavorable outcomes, and ultimately, increases the burden to hospitals in terms of the utilization of resources and overall costs of health care. 5Accordingly, outpatient management of low-risk FN has been found to be effective and is recommended as part of clinical care. 6evious studies have shown that an increase in hospital LOS is the main driver accounting for the overall cost of managing an FN event. 7ile treating FN in the outpatient setting will reduce some hospital costs, others will be transferred from the inpatient to the outpatient setting, mainly with respect to the use of antibiotics, medical practitioner consultations, and neutrophil monitoring.Of particular interest is the use of antibiotics, a critical component in both the inpatient and outpatient setting, and how costs may differ across patients by the type of antibiotic used, treatment duration, route of administration (oral or intravenous), and delivery mode (inpatient or outpatient).Overall, the magnitude and structure of differences across patients in terms of the composition of healthcare use and the associated costs for the treatment of FN in the inpatient and outpatient settings are unknown. 8e objective of this study is to describe the use of healthcare resources and estimate the overall costs associated with FN events in pediatric patients classified as high and low risk for FN.A formal low-risk FN program was not in use during the study period, and FN episodes were managed according to state-based hospital FN guidelines that were in keeping with international recommendations. 9Cessation of antibiotics and hospital discharge was typically considered in patients with neutrophil recovery, negative cultures, and at least a 24-hour period of clinical stability and absence of fever.Antibacterial prophylaxis (excluding for Pneumocystis jirovecii pneumonia) was not routinely used.

Data
Each FN episode was stratified using the internationally derived 10 and locally validated 4 PICNICC rule.The PICNICC rule uses a series of weighted variables (malignancy type, temperature, clinically "severely unwell," hemoglobin, white cell count, and absolute monocyte count) to estimate the risk of infection, stratified as being low (<.The analysis that assessed the utilization of pharmaceuticals grouped drugs via the Anatomical Therapeutic Chemical (ATC) classification and health services via the MBS item descriptor (for service type) and code.
Outpatient data on healthcare utilization were used to conduct a bottom-up microcosting analysis allowing for an assessment of potential differences in FN events risk classification.Total costs for an FN event comprised the sum of fees to government for medical, pharmaceutical, and hospital services, and out-of-pocket payments to patients for MBS and PBS listed items.Thus, the analysis took a quasi-societal perspective, which does not consider the broader economic impact this may have on families in terms of productive capacity.Costs were attributed to an FN event, if they were incurred from the date of FN diagnosis until the resolution of fever and last observed date of antibiotic use.Healthcare use during the FN event (from FN diagnosis until end of antibiotic therapy) was compared with use prior to the event (30 days prior to the FN diagnosis) and post event (30 days after the end of the FN event).A comparison of utilization of outpatient pharmaceuticals during and 30 days after the end of the event was also explored.
The mean PBS and MBS costs and corresponding bias corrected 95% CIs were estimated by bootstrapping the data, with 1000 replications.A generalized linear model (GLM) was applied to estimate the mean total cost per FN episode, taking into account that individuals may have experienced multiple FN events during the study period.Based on the non-normally distributed nature of the data (being non-negative and positive-skewed), a GLM for a gamma distribution with a logarithmic link function was used. 12,11The model controlled for gender, age at cancer diagnosis, cancer diagnosis, previous hematopoietic stem cell transplant (HSCT), antibiotic treatment duration, hospital LOS, and FN risk status.Mean costs per low-and high-risk groups were reported using the post-estimation margins command.Finally, outpatients Medicare data were analyzed to assess whether the bulk of its use occurred at a particular time post FN onset.Statistical analyses were undertaken using STATA version 16. 12 This study has been approved by the Royal Children's Hospital Melbourne Human Research Ethics Committee (RCH HREC 36040A), and an ethics ratification was obtained by the University of Technology Sydney (HREC ETH17-1128).Parents were consented prior to the child's enrolment in the study to allow access to their Medicare data.

RESULTS
The Australian PICNICC study reported 858 FN events occurring in 462 pediatric cancer patients.Overall, 28% of patients experienced one event, 49% experienced two to three events, and the remaining 23% experienced more than four events (range: four to nine).Of all events, 703 (81.9%) and 155 (18.1%) were categorized as high and low risk as per the PICNICC rule, respectively. 10In-hospital cost data were provided by seven out of eight hospitals for a total of 757 (88%) events corresponding to 416 (90%) patients.The disaggregated in-hospital costs were available for six out of the eight hospitals, and accounted for 710 of the 858 FN events.A small proportion of these events (2.1%) were handled as hospital in the home (HITH), hence the results from this study reflect the management of FN mainly in the hospital setting.
Medicare data (MBS and PBS) were available for 429 FN events in 212 (51%) patients.

In-hospital utilization of resources and costs
The mean costs of medical services delivered in-hospital, stratified by risk group, is shown in Table 1.Overall, no substantial differences were observed in terms of the pattern of utilization of resources when comparing FN events classified as high and low risk.The cost categories that contributed most to the total hospitalization cost were in-hospital services, mainly including the cost of wards, and hospital consultations (i.e., specialist consultations such as medical oncologist).
For both high-and low-risk patients, these two categories contributed most to the total cost, accounting to approximately 70% of the total cost.The cost category that showed the largest difference across highand low-risk events was therapeutics, accounting for 11.9% and 9.4% of total costs, respectively.Overall, patients classified as high risk had a higher mean therapeutic cost per event compared to low-risk events ($4208 and $1536, respectively).Statistically significant differences were observed between the two risk groups in terms of LOS where, on average, patients categorized as high risk and low risk remained in hospital for 11 and 6 days, respectively (mean difference of 4.8 days; 95% CI: 2.2-7.3;p < .0002).
However, this did not translate into a significant difference in mean cost per day between the two risk groups ($2748 vs. $2793 per day for high and low risk, respectively) (Table 1).
Despite no significant differences in the pattern of resource use, results of the GLM analysis for the hospital costs per FN event show that, on average, FN events were associated with statistically TA B L E 1 Costs of managing a febrile neutropenia event considering the in-hospital and outpatient setting.  1 and Table S1).The mean hospital cost for high-risk events was $17,685 (95% CI: 17,013-18,356) and for low-risk events was $10,222 (95% CI: 9417-11,026).The variables that had a statistically significant impact on the total cost were the risk status, hospital LOS, and duration of the FN event (see Supporting Information).

Outpatient resources and costs
MBS and PBS data were available for 429 FN events (of which 357 were classified as high risk and 72 as low risk).For the period from diagnosis with FN and up to 30 days after the end of the FN event, a total of 723 drugs (PBS funded) were prescribed on an outpatient basis F I G U R E 1 Utilization of pharmaceuticals as per ATC name prior, during, and up to 30 days after the end of the FN event.Figure does not include utilization of drugs classified as dermatologicals, sensory organs, various +, cardiovascular system, respiratory system, musculo-skeletal system, and genito urinary system and sex due to insignificant observations (n = 14 observations).The x-axis represents the proportion of utilization corresponding to each ATC name.Numbers in the table reflect frequency of utilization (i.e., proportion of use out of total number of drugs prescribed).Small numbers were registered for other types of pharmaceuticals, which are not presented in this figure (dermatological, musculo-skeletal system, genito urinary system and sex, respiratory system, cardiovascular system, and sensory organs).ATC, Anatomical Therapeutic Chemical; FN, febrile neutropenia.
(mean = 1.7 drugs per event) and 19,578 MBS services were delivered (mean = 45.6 services per event).Of these, 72 drugs (less than one per event) and 7289 MBS services (mean per event = 17) were delivered while the event was ongoing, and 651 drugs (mean per event = 1.5) and 12,289 services were delivered after the event had ended (mean per event = 30).
Figure 1 shows the outpatient use of pharmaceuticals classified by ATC during the period prior to the FN diagnosis, during the FN event, and 30 days after the end of the FN event.Across all three periods (pre event, during event, and post event) utilization was dominated by the use of drugs classified as "antineoplasics and immunomodulating" agents, "anti-infectives for systemic use," and "alimentary and tract metabolism."The use of antineoplastics and immunomodulators was dominated by various antineoplastic agents (94%) and some colony-stimulating factors (6%).The use of anti-infectives predominantly included antibiotics (72.0%) and antifungals (27.6%).While there were no substantial differences in the type and use of drugs between events classified as high or low risk, more pathology services were delivered (Table 2).Noting some differences between groups in the proportion of patients receiving cytarabine, cyclophosphamide, doxorubicin, and vincristine were used post FN.No other substantive changes in the pattern of use were observed pre and post FN event, nor between high-and low-risk patients.When we explored the time point as to when the utilization occurred, no differences were observed in any of the service categories analyzed, except for pathology services, where we noticed that the increase in use was concentrated in the fourth (last) week of the 30-day period after the FN event (data not shown).
As can be observed from the results in Table 1, mean total Medicare costs (MBS and PBS costs) did not differ between FN events categorized as high risk and those categorized as low risk, either during 30 days after the end of the FN event.For both risk groups, mean total Medicare costs were higher in 30 days after the event (Table 1).
In the GLM analysis, FN events were associated with significantly higher costs when patients were categorized as high risk compared to low risk for both hospital costs and overall total costs (p < .000)(Table 1).The mean overall cost per event (i.e., in-hospital and outpatient costs) in high-risk patients was $17,827 and in low-risk patients was $10,574.For both patient groups, the bulk of the total cost came from the in-hospital setting (99.0%).The variables that had a statistically significant impact on the total cost were the risk status and the hospital LOS (see Supporting Information).

DISCUSSION
Our study provides an in-depth understanding of the cost structure associated with pediatric FN management in Australia and is the first study to incorporate outpatient data in this analysis.Overall, the total cost of FN was due to the use of in-hospital resources (99.0%), and events classified as high-risk are significantly more costly compared to those classified as low risk.Factors significantly impacting total cost were the risk status, hospital LOS, and duration of the FN event.
Excluding antineoplastics, antimicrobials are the most commonly used medications in the inpatient and outpatient setting and in the periods before, during, and after the FN event.Our analyses of healthcare resource across the FN event trajectory (30 days prior, during the event, and 30 days post) found, not surprisingly, that during the event, there is a substantial drop in utilization of outpatient pharmaceutical and medical/diagnostic services.When we compared the pattern of use 30 days prior to the event with the period 30 days after the end of FN event, we also observed a slight reduction in most anti-infectives and antineoplastics.A noteworthy exception was the use of amoxicillin/clavulanic acid that showed a trend to increase.
Although a formal low-risk FN program was not implemented during the study period, this increase in amoxicillin/clavulanic acid, frequently used to complete FN treatment for lower risk patients, suggests that some centers may have adopted this approach, albeit on an ad hoc basis.Other changes in utilization, such as the reduction in the use of trimethoprim plus sulfamethoxazole, may reflect the practice to wait to neutrophil recovery prior to re-commencement.
The type of drugs (i.e., anti-infectives and antineoplastics) used to manage patients in the inpatient (hospital) setting were not available from the disaggregated hospital cost data.Rather, utilization on the use of inpatient antibiotics was collected as part of the PICNICC study.These data show variation in the use of inpatient antibiotics, which may reflect a lack of consensus in antibiotic treatment in this patient population. 1,13Of the drugs used in the outpatient setting, antiinfectives accounted for the second highest usage across all periods, after antineoplastics.This represents a previously unrecognised pattern of usage, and highlights the importance of including outpatient anti-infectives in antimicrobial stewardship activities.
[16][17][18] In fact, the studies suggest that the outpatient management is dominant leading to a reduction in costs and an improvement in terms of health outcomes for both patients and carers. 5,8,19However, none of these studies provide a comprehensive understanding of both in-hospital and outpatient cost structure and the potential for costoffsets if an intervention that categorizes patients according to risk status is effectively implemented.In general, the available cost studies or cost-effectiveness analyses addressing the different settings where an FN event can be managed have used strong assumptions regarding the outpatient management of FN.For example, one costeffectiveness analysis that compared inpatient with outpatient management, assumed that all low-risk patients managed in an outpatient setting received the maximum dose of levofloxacin (750 mg/day). 5Similarly, a cost analysis of FN in adult patients conducted in Australia, assumed all patients were treated with oral antibiotics on that basis that there was a lack of evidence on the use of parenteral antibiotics in the outpatient setting. 8While these assumptions may be reasonable, it does not reflect the actual use of resources and differential management that these events are likely to have in the outpatient setting.More recently, a cost-effectiveness analysis was conducted to assess the impact of a pilot pediatric low-risk FN program. 20This program enabled children, identified as low risk using a validated risk stratification tool, to complete FN treatment at home after a brief (<48 hours) inpatient admission.In this study, patients received blood on healthcare utilization were collected alongside the Australian Predicting Infectious Complications of Neutropenic sepsis In Children with Cancer (PICNICC) study, a prospective observational study of FN in pediatric patients recruited across eight tertiary hospitals in Australia from December 1, 2016 to January 31, 2018.Detailed methodology is described elsewhere. 7Data on consecutive episodes of FN in children (age < 18 years) with cancer or hematological malignancy were prospectively collected.Episodes were included if they had a documented fever and neutropenia, and excluded if FN treatment commenced at a non-participating site, the patient had undergone a hematopoietic stem cell transplant within the preceding 3 months or the episode occurred while they were receiving concurrent intravenous or oral antibiotics (excluding prophylaxis).

TA B L E 2
Disaggregated utilization of PBS and MBS services across the three periods studied.

High risk (N = 623) Low risk (N = 134) Mean cost ($AUD) [95% Confidence interval] Mean cost ($AUD) [95% Confidence interval]
Table shows in-hospital disaggregated costs.Outpatients costs include PBS and MBS for the period of 30 days prior to the event, the period considering the duration of the event (since diagnosis until last antibiotic), and 30 days after the event.A generalized linear model was conducted to estimate the total cost.
significant higher costs for high-risk compared to low-risk patients (p < .000)(Table

30 days prior to FN event FN event (date of diagnosis and AB stop) 30 days after the end of FN event High risk (n = 357) Low risk (n = 72) High risk (n
Abbreviations: AB, antibiotic; FN, febrile neutropenia; Freq, frequency; MBS, Medicare Benefits Scheme; PBS, Pharmaceutical Benefits Scheme.