Resource Utilization and Safety of Outpatient Management Following Intensive Induction or Salvage Chemotherapy for Acute Myeloid Leukemia or Myelodysplastic Syndrome: A Nonrandomized Clinical Comparative Analysis

Importance  Adults with acute myeloid leukemia (AML) or myelodysplastic syndrome (MDS) typically remain hospitalized after induction or salvage chemotherapy until blood cell count recovery, with resulting prolonged inpatient stays being a primary driver of health care costs. Pilot studies suggest that outpatient management following chemotherapy might be safe and could reduce costs for these patients.

Objective  To compare safety, resource utilization, infections, and costs between adults discharged early following AML or MDS induction or salvage chemotherapy and inpatient controls.

Design  Nonrandomized, phase 2, single-center study conducted at the University of Washington Medical Center. Over a 43-month period (January 1, 2011, through July 31, 2014), 178 adults receiving intensive AML or MDS chemotherapy were enrolled. After completion of chemotherapy, 107 patients met predesignated medical and logistical criteria for early discharge, while 29 met medical criteria only and served as inpatient controls.

Interventions  Early-discharge patients were released from the hospital at the completion of chemotherapy, and supportive care was provided in the outpatient setting until blood cell count recovery (median, 21 days; range, 2-45 days). Controls received inpatient supportive care (median, 16 days; range, 3-42 days).

Main Outcomes and Measures  We analyzed differences in early mortality, resource utilization including intensive care unit (ICU) days, transfusions per study day, and use of intravenous (IV) antibiotics per study day), numbers of infections, and total and inpatient charges per study day among early-discharge patients vs controls.

Results  Four of the 107 early-discharge patients and none of the 29 control patients died within 30 days of enrollment (P = .58). Nine early-discharge patients (8%) but no controls required ICU-level care (P = .20). No differences were noted in the median daily number of transfused red blood cell units (0.27 vs 0.29; P = .55) or number of transfused platelet units (0.26 vs 0.29; P = .31). Early-discharge patients had more positive blood cultures (37 [35%] vs 4 [14%]; P = .04) but required fewer IV antibiotic days per study day (0.48 vs 0.71; P = .01). Overall, daily charges among early-discharge patients were significantly lower than for inpatients (median, $3840 vs $5852; P < .001) despite increased charges per inpatient day when readmitted (median, $7405 vs $5852; P < .001).

Conclusions and Relevance  Early discharge following intensive AML or MDS chemotherapy can reduce costs and use of IV antibiotics, but attention should be paid to complications that may occur in the outpatient setting.

Adults with newly diagnosed or relapsed or refractory acute myeloid leukemia (AML) or high-risk myelodysplastic syndrome (MDS) commonly require intensive chemotherapy to achieve disease remission.^1,2 In many countries, standard practice dictates that these patients remain hospitalized until blood cell count recovery owing to the risk of overwhelming infections and bleeding during pancytopenia. This policy requires hospitalization for an average of 3 to 4 weeks after completion of chemotherapy.³ Several cost analyses have shown that the resulting prolonged inpatient stays are a driver of exceedingly high costs of leukemia care.^4-6 These costs are only expected to rise owing to the continued introduction of expensive diagnostic tools and therapeutic modalities. To offset these expenditures, it would be desirable to reduce the time that patients with AML or MDS spend in the hospital during remission induction.

Over the last several years, the use of oral prophylactic antimicrobial agents has increased, and transfusion support of outpatients has become routine. These advances could support a shift from inpatient to outpatient management for some patients with hematologic malignant conditions undergoing intensive therapies. Indeed, an increasing number of studies has documented the feasibility, safety, and potential cost savings of treating patients undergoing autologous or allogeneic hematopoietic cell transplantation in the outpatient setting during much of their therapy.^7-19 However, only a few retrospective and noncontrolled prospective studies,^20-25 including one conducted at our institution,²⁶ have investigated whether selected patients with AML or MDS can be safely discharged after completion of induction chemotherapy. Our group’s pilot study²⁶ involving 15 early-discharge patients and 5 controls suggested that early hospital discharge is safe and could significantly reduce health care charges. This observation motivated the present larger phase 2 trial to compare the safety, resource utilization, and health care charges in adults discharged early following AML induction or salvage chemotherapy with those of inpatient controls.

Box Section Ref ID

Patients aged 18 to 75 years were eligible for study participation if they had begun (or were to begin within the next week) intensive AML-like chemotherapy (eg, with “7 + 3”—7 days of standard-dose cytarabine, and 3 days of an anthracycline antibiotic or an anthracenedione—or a regimen of similar or higher intensity) for untreated or relapsed or refractory AML or MDS, excluding acute promyelocytic leukemia. Patients with significant hypersensitivities to prophylactic antimicrobial agents were excluded. The institutional review board at the Fred Hutchinson Cancer Research Center approved this protocol, and written informed consent was obtained from patients in accordance with the Declaration of Helsinki.²⁷

Eligibility for early hospital discharge after completion of chemotherapy depended on medical and logistical criteria at the time of contemplated discharge. The medical criteria were Eastern Cooperative Oncology Group (ECOG) performance status of 0 to 1, bilirubin level less than or equal to 3 times the upper limit of normal, glomerular filtration rate at least 25% of the lower limit of normal, and no clinical signs of heart failure or uncontrolled bleeding. Need for intravenous (IV) antimicrobial agents did not preclude early hospital discharge. The logistical criteria were residency within 60 minutes of the study center (University of Washington Fred Hutchinson Cancer Research Center), a reliable caregiver, and willingness to frequently follow up at the primary outpatient care facility (Seattle Cancer Care Alliance). Patients meeting both medical and logistical criteria were discharged; if readmitted, subsequent early hospital discharge was possible if all medical and logistical criteria were again met. Patients who met medical but not logistical criteria after completion of chemotherapy remained hospitalized and served as inpatient controls.

All patients were prescribed prophylactic antimicrobial agents (levofloxacin, fluconazole, and acyclovir or medications with similar antimicrobial coverage) until peripheral blood cell count recovery. Asymptomatic patients received red blood cell transfusions for a hematocrit level less than 26% and platelet transfusions for a platelet count below 10 × 10⁹/L; the same transfusion triggers were used for both inpatients and outpatients. Patients who had fever (as defined by the Infectious Diseases Society of America²⁸) and an absolute neutrophil count (ANC) below 1000/µL received IV antibiotics in the hospital. Management of neutropenic fever generally consisted of a third-generation cephalosporin (unless not tolerated) with or without additional gram-positive coverage if indicated. Data regarding vancomycin-resistant Enterococcus colonization and history of methicillin-resistant Staphylococcus aureus infections were highlighted in the electronic medical record to allow for tailoring of empirical gram-positive coverage. The inpatient team determined the duration of IV antibiotic treatment and appropriateness to transition back to oral antimicrobial agents.

Paralleling our institution’s approach to the outpatient care of patients with AML or MDS following postremission chemotherapy, the study protocol suggested that outpatients be seen 3 times weekly by an oncology nurse and weekly by the primary outpatient oncologist or an advanced practice clinician if asymptomatic; patients were seen more frequently if mandated by medical problems (eg, nausea, dehydration, or transfusion needs). Patients and caregivers had 24-hour access to phone consultation with a physician.

Information on medical complications (infections) and use of medical resources (transfusions, IV antibiotics, ICU admissions) was collected from electronic medical records. Transfusion requirements were analyzed as number of units of either packed red blood cells or platelets per study day. Differences in infectious risk between the 2 patient cohorts were assessed based on clinically relevant bloodstream and Clostridium difficile infections. Bloodstream infections were diagnosed for any positive blood culture findings except for coagulase-negative S aureus when only 1 of 2 bottles was reported positive. Clostridium difficile infection was assumed when the toxin B gene was detected via polymerase chain reaction during the study period. Professional and facility gross fees or charges associated with inpatient and outpatient management were captured using electronic billing information. Charges (in US dollars) were measured per study day.

For early-discharge patients, follow-up time began the day after hospital discharge, which had to occur within 72 hours of chemotherapy completion, and ended when the patient (1) received additional chemotherapy, (2) attained an ANC of 0.5 × 10⁹/L or higher and self-sustained platelet count of 20 × 10⁹/L or higher, (3) sought medical attention elsewhere, or (4) 45 days had elapsed from discharge, whichever occurred first. For inpatient controls, follow-up time began the first day the patient met the medical criteria for discharge (but within 72 hours of completion of chemotherapy) and ended when they (1) were discharged from the hospital, (2) achieved the same blood cell counts needed for study termination of early-discharge patients, or (3) received additional chemotherapy. Chemotherapy regimens were categorized as standard-intensity regimens (7 + 3 or 7 + 3–like therapy) and high-intensity regimens (containing cytarabine doses of >1 g/m²). Characteristics of early-discharge patients and inpatient controls were compared with Fisher exact (categorical characteristics) and Wilcoxon Mann-Whitney tests (continuous characteristics). Seven patients were enrolled twice in the study: 5 patients were in the early-discharge cohort twice, and 2 patients were included once each in the early-discharge cohort and the control cohort. Analyses accounted for the potential correlation from patients enrolled twice. Outcome analyses accounted for days in the study. Linear (quantitative outcomes) and logistic (binary outcomes) regression analysis was used to evaluate prognostic variables for the proportion of time spent as an inpatient after discharge, the need for hospital readmission, and the risk of bacteremia.

Between January 1, 2011, and July 31, 2014, 178 patients with MDS or AML were enrolled before or during intensive induction or salvage chemotherapy. Forty-two patients (23.6%) were deemed medically ineligible for discharge by the time they finished chemotherapy infusion, primarily due to an ECOG performance score greater than 1, which was associated with the development of an acute medical problem (most commonly clinically significant fever or infection and acute toxic effects from chemotherapy such as nausea, mucositis, or diarrhea). A total of 107 patients (60.1%) met both medical and logistical criteria and were discharged early from the hospital, within 72 hours of completing chemotherapy. The remaining 29 patients (16.3%) met the medical but not the logistical criteria and therefore remained in the hospital, serving as inpatient controls. Demographic and treatment-related characteristics of early-discharge and control patients are summarized in Table 1. More early-discharge patients were female and received therapy for relapsed or refractory AML or MDS. From the date of cohort assignment after completion of chemotherapy, early-discharge and control patients spent a median of 21 days (range, 2-45 days) and 16 days (range, 3-42 days), respectively, in the study.

While in the study, 93 of the 107 early-discharge patients were readmitted; 19 patients were readmitted twice; 5 were readmitted 3 times; and 1 was readmitted 4 times. Causes for readmission were neutropenic fever (n = 108), bacteremia and/or sepsis (n = 7), localized infection (n = 2), and nausea and/or vomiting (n = 2). One patient each was readmitted for upper gastrointestinal tract bleeding, diarrhea, facial swelling, hypotension, mucositis, and monitoring after a motor vehicle accident. Early-discharge patients spent a median of 13 days (range, 0-42 days) as outpatients. The median total number of days per early-discharge patient spent in the hospital was 8 (range, 0-33); thus, early-discharge patients spent a median of 61% (range, 0%-100%) of the time from discharge until removal from the study as outpatients (Table 2). Neither age (>55 vs ≤55 years), type of treatment (7 + 3–like vs higher intensity), nor type of disease (newly diagnosed vs relapsed or refractory) were associated with readmissions of early-discharge patients. Early-discharge patients older than 55 years spent on average 11% less time as inpatients than younger early-discharge patients, but this difference was not statistically significant (a median of 56% vs 67% of study time; P = .64). There was also no statistically significant difference in the proportion of time spent as inpatients among early-discharge patients who received7 + 3–like treatment vs higher-intensity chemotherapy (a median of 67% of total study time for 7 + 3-like treatment vs 61% for high-intensity treatment; P = .97) or among patients with newly diagnosed vs relapsed or refractory disease (a median of 61% vs 61% of total study time; P = .95).

Nine (8%) of the early-discharge patients required a median of 2 (range, 1-6) days of ICU care (8 patients for sepsis, 1 for upper gastrointestinal tract bleeding). There were no patients who required ICU care in the control population (P = .20). Four early-discharge patients (4%) but no control patients died within 30 days of study follow-up (P = .58). These patients were all older than 50 years (aged 52, 57, 62, and 70 years). Two of these 4 patients were being treated for relapsed or refractory AML; 1 patient had secondary AML after being previously diagnosed with MDS; and 1 patient developed AML after being previously treated for chronic lymphocytic leukemia with a fludarabine-containing regimen. Causes of death were sepsis in 3 patients and invasive fungal sinusitis (mucormycosis) in the patient previously treated with fludarabine.

As detailed in Table 2, the number of bloodstream infections was statistically higher among early-discharge patients (P = .04). On review of the microbiological data, we found 28 gram-positive and 8 gram-negative bacterial infections as well as 5 fungal infections (4 patients had >1 positive culture during follow-up) among early-discharge patients, while 4 bacteremias due to gram-positive organisms were noted in inpatient controls. In multivariable analysis, receiving treatment for relapsed or refractory AML or MDS rather than newly diagnosed disease was independently associated with the diagnosis of bacteremia (odds ratio [OR], 2.66; P = .02), whereas treatment arm (ie, early discharge vs inpatient) (OR, 2.71, P = .09), age (OR, 0.999 per year; P = .94), , and type of chemotherapy (7 + 3–like vs higher-intensity regimen OR, 1.79, P = .18), were not. Despite this increase in documented bloodstream infections, early-discharge patients overall received IV antibiotics (given as inpatient and, in many cases, as outpatient after discharge) for less of the study period (median, 0.48 vs 0.71 days of IV antibiotics per study day; P = .01) (Table 2). However, there was no statistically significant difference in the number of days of IV antibiotics per study day among early-discharge patients vs controls who had positive cultures (0.56 early discharge vs 0.71 inpatient; P = .36). There was a slightly, statistically nonsignificant increased risk of C difficile infection among early-discharge patients (n = 10 [9%]) compared with controls (n = 0) (P = .12).

As detailed in Table 2, there were no differences between the 2 cohorts in the median number of red blood cell units transfused per study day (P = .55) or the number of platelet units transfused per study day (P = .31).

The median charges per study day among early-discharge patients were significantly lower than those of inpatient controls (median $3840 vs $5852; P < .001) (Table 3 and Figure). This was true despite higher daily inpatient charges for early-discharge patients once they were readmitted compared with the daily inpatient charges incurred by controls (median $7405 vs $5852; P < .001). Neither age, regimen intensity, nor disease status was associated with daily charges incurred (median of $4670 for ≤55 years vs $4599 for >55 years, P = .68; median of $4470 for 7 + 3–like vs $4670 for higher-intensity chemotherapy, P = .38; and median of $4934 for newly diagnosed vs $4234 for relapsed or refractory disease, P = .19).

The delivery of effective supportive care for patients with AML or MDS has increased substantially in the past 1 to 2 decades with the introduction of oral broad-spectrum antimicrobial agents and routine use of transfusions. Nonetheless, in many countries including the United States, patients receiving induction regimens such as 7 + 3 continue to spend many weeks preemptively as inpatients for close monitoring at large financial and possibly psychological detriment. Our group therefore conducted a pilot study (NCT00844441) between April 2009 and April 2010 to explore early discharge of adult patients with AML or MDS.²⁶ Among 39 enrollees, 15 were discharged early and 5 patients served as inpatient controls. No patient died within 30 days of enrollment, and patients discharged early incurred lower median daily charges ($3270 vs $5467; P = .01) than controls.²⁶ The present study extends our experience to a much larger patient population. Consistent with the data from the pilot study, early-discharge patients spent a median of more than 50% of their time as outpatients, demonstrating the feasibility of this care strategy. Moreover, this follow-up study indicates that a policy of early hospital discharge following intensive chemotherapy for AML or MDS decreases health care charges (Figure) without significantly increasing treatment-related mortality.

Over the last 2 decades, the rates of early death (treatment-related mortality) have significantly decreased in patients with AML or MDS treated with intensive induction chemotherapy,²⁹ likely as a result of improved supportive care measures. Indeed, we observed only 4 deaths within the first 30 days after completion of chemotherapy among the 136 patients in the present study (2.9%), and only 9 patients required ICU-level care while in the study. However, all patients who required ICU-level care or who experienced early death had been allocated to the early-discharge arm. The rates of ICU-level care or early death were not statistically significantly different between early-discharge and control patients (P = .58 and P = .20, respectively), and they appeared well within the expected range, suggesting that early discharge is safe. Nevertheless, the rarity of these outcomes does not exclude the possibility of increased risk of life-threatening complications after early discharge following intensive chemotherapy. Thus, this finding supports a cautious approach to this care strategy, and our recommendation is that patients be monitored very closely in well-equipped outpatient facilities with immediate access to a hospital.

The rising costs associated with cancer care pose a significant burden to individual patients and to society in general.^30-33 It is well established that patients with hematologic malignant conditions such as AML or MDS incur some of the highest costs.^34,35 Thus, care strategies aimed at reducing these costs are of high economic interest. As health care costs for patients with AML are largely driven by inpatient care,³ we posited that a policy of early hospital discharge might lead to a reduction in health care expenditures. Consistent with the findings from our group’s pilot study,²⁶ the results from this phase 2 study strongly support this idea by demonstrating that hospital discharge soon after completion of intensive chemotherapy for AML or high-risk MDS is associated with reduced daily costs compared with standard practice, and this despite similar use of red blood cell and/or platelet transfusions between groups and higher inpatient costs for early-discharge patients who were rehospitalized. To our knowledge, this is the first large prospective study designed to determine the fiscal impact of these 2 approaches. The median difference in charges accrued between our inpatient and early-discharge populations during the supportive phase of care (approximately $2000/study day) is similar to what was seen in our group’s pilot study,²⁶ arguing for the validity of our findings.

While several methods for studying health care costs have been used,^4-6 we chose to use hospital charges, which are appreciably higher than the actual amount reimbursed by payers on behalf of the patient. This allowed us to avoid inconsistencies in reimbursement policies negotiated between the institution and various payers and perform a valid comparison between the study groups; however, we would caution against using our results to estimate the direct costs of AML or MDS care. Moreover, while we avoided this discrepancy by performing all inpatient and outpatient care within a hospital-based clinic facility, hospital-based health care system charges may be greater than non–hospital-based outpatient clinic charges owing to the larger amount of overhead required to operate large health care facilities. This possibility suggests that even greater savings may be realized if early-discharge policies could be implemented in non–hospital-based outpatient clinics.

One objective of increased outpatient patient management is to reduce the morbidity and/or mortality from nosocomial or treatment-related infections. Indeed, a retrospective study demonstrated that septicemias were reduced when the practice of early discharge was implemented, although an infection shift from gram-negative to gram-positive organisms was noted in the outpatient cohort.²⁴ In contrast, in our study, early-discharge patients experienced a higher number of bacteremias than inpatient controls, with most bacteremias in early-discharge patients being due to gram-positive organisms. Consistent with the findings of the study by Halim et al,²⁴ this observation suggests that with increased use of outpatient care, there might be an increase in the frequency of gram-positive infections. The large number of staphylococcal and streptococcal infections in the early-discharge patient population suggests the value of closer attention to patient education regarding care of indwelling catheters. It may also be influenced by the use of levofloxacin, which has broad gram-negative coverage, as outpatient prophylaxis, although the approach to antimicrobial prophylaxis did not differ between our 2 groups.

Because of the size of our cohort and the occurrence of only 41 documented bloodstream infections in our study, we were relatively limited in our ability to identify the covariates associated with bloodstream infection. However, in our multivariate analysis, treatment for relapsed or refractory AML or MDS was independently associated with the diagnosis of bloodstream infection. It is therefore also possible that the increased number of bacteremias was due to an imbalance in the stage of treatment (newly diagnosed vs relapsed or refractory disease) (Table 1) rather than the setting of treatment (inpatient vs outpatient). Despite more positive blood cultures, early-discharge patients received fewer days of IV antibiotics during the study period. Since the days of IV antibiotics did not differ between early-discharge patients and controls when positive blood cultures were found, it is interesting to speculate that the overall reduction in IV antibiotic days during the study period for early-discharge patients was driven by differences in the management of culture-negative neutropenic fevers and/or a tendency toward more rapid conversion from IV to oral antibiotics after clinical stabilization in patients who were considered for early hospital discharge.

Although our early-discharge patients spent a median of more than 60% of the study time as outpatients, hospital readmissions were common. While we restricted study eligibility to a group of medically fit patients, it is plausible that a subset could be identified that is at particularly high risk of readmission and medical complications and thus not well suited for early hospital discharge. To begin addressing this question, we investigated the relationship between rehospitalization time and older age, treatment intensity, and disease stage (newly diagnosed vs relapsed or refractory disease), but none of these factors was associated with increased time spent as inpatient or higher treatment charges, suggesting that early hospital discharge should not be withheld based on any of these factors.

One limitation of our study was the inability to account for the influence of certain disease- and patient-related factors on early mortality, given the limited number of events. As a second limitation, we did not collect data on certain expenses incurred such as lodging, transportation, caregiver time, home care costs (including costs for home administration of antimicrobial agents), prescription costs, and nursing and child care expenses (such costs are higher for outpatients than inpatients and may disproportionally burden outpatients directly because they may not be covered by their health care insurance).

Finally, the nonrandomized nature of this study should be acknowledged as a limitation. Theoretically, a randomized assessment comparing an early hospital discharge strategy with conventional hospital care for selected patients after completion of induction or reinduction chemotherapy for MDS or AML is feasible and would be ideal to further test this care concept in this patient population In addition to assessing the safety and direct health care utilization and associated costs, such a study could investigate effects on quality of life and indirect costs (eg, lodging, transportation) and their immediate financial effects on patients and caregivers. However, such a trial might be difficult to conduct: in our experience, most patients with AML or MDS were highly interested in an outpatient care approach if there were no logistical hurdles, such as lack of caregiver or housing, and would not have been willing to remain hospitalized if randomized to the control group. Moreover, some physicians may strongly prefer one care approach over the other. With this in mind, we suspect that a randomized study including a large number of eligible patients may be challenging to complete and potentially subject to significant selection bias.

Our findings suggest that an early-discharge policy following intensive AML or MDS chemotherapy in selected adult patients allows a shift toward a greater portion of care being delivered in the outpatient setting. This care strategy may reduce health care costs and the duration of use of IV antibiotics; however, close follow-up in well-equipped and well-staffed outpatient facilities is required, and specific outpatient support and readmission procedures should be put in place to ensure maximum patient safety.

Accepted for Publication: July 2, 2015.

Corresponding Author: Jennifer E. Vaughn, MD, MSPH, Blue Ridge Cancer Care, Department of Medicine, Virginia Tech Carilion School of Medicine and Research Institute, 2013 S Jefferson St, Roanoke, VA 24014 (jvaughn@vtc.vt.edu).

Version: July 2, 2015.

Published Online: September 10, 2015. doi:10.1001/jamaoncol.2015.2969.

Author Contributions: Dr Vaughn had full access to all of the data in the study and takes responsibility for the integrity of the data and accuracy of the data analysis.

Study concept and design: Vaughn, Estey, Walter.

Acquisition, analysis, or interpretation of data: Vaughn, Othus, Powell, Gardner, Rizzuto, Hendrie, Becker, Pottinger, Walter.

Drafting of the manuscript: Vaughn, Othus, Walter.

Critical revision of the manuscript for important intellectual content: Vaughn, Othus, Powell, Gardner, Rizzuto, Hendrie, Becker, Pottinger, Estey, Walter.

Statistical analysis: Vaughn, Othus, Rizzuto, Estey, Walter.

Administrative, technical, or material support: Walter.

Study supervision: Walter.

Conflict of Interest Disclosures: None reported.

Funding/Support: Dr Vaughn was supported by fellowship training grant T32-HL007093 from the National Heart, Lung, and Blood Institute, National Institutes of Health; Dr Walter is a Leukemia & Lymphoma Society Scholar in Clinical Research.

Role of the Funder/Sponsor: The supporting agencies played no role in the design and conduct of the study; collection, management, analysis, and interpretation of the data; preparation, review, or approval of the manuscript; and decision to submit the manuscript for publication.