What are the risk factors for admission to the intensive care unit (ICU) among patients with acute myeloid leukemia, and what is the effect of ICU care on mortality and use of health care resources?
In this cohort study, factors associated with admission to the ICU, mortality, and cost among 43 249 hospitalized patients with acute myeloid leukemia included both fixed demographic and systemic factors and modifiable clinical variables. Admission to the ICU was associated with high mortality and costs that increased proportionally with comorbidity burden.
Identification of modifiable risk factors for ICU use and mortality allows for testing of primary prevention and intervention strategies.
Adults with acute myeloid leukemia (AML) commonly require support in the intensive care unit (ICU), but risk factors for admission to the ICU and adverse outcomes remain poorly defined.
To examine risk factors, mortality, length of stay, and cost associated with admission to the ICU for patients with AML.
Design, Setting, and Participants
This study extracted information from the University HealthSystem Consortium database on patients 18 years or older with AML who were hospitalized for any cause between January 1, 2004, and December 31, 2012. The University HealthSystem Consortium database contains demographic, clinical, and cost variables prospectively abstracted by certified coders from discharge summaries. Outcomes were analyzed using univariate and multivariable statistical techniques. Data analysis was performed from November 15, 2013, to August 15, 2016.
Main Outcomes and Measures
Primary outcomes were admission to the ICU and inpatient mortality among patients requiring ICU care. Secondary outcomes included length of stay in the ICU, total hospitalization length of stay, and cost.
Of the 43 249 patients with AML (mean [SD] age, 59.5 [16.6] years; 23 939 men and 19 310 women), 11 277 (26.1%) were admitted to the ICU. On multivariable analysis (with results reported as odds ratios [95% CIs]), independent risk factors for admission to the ICU included age younger than 80 years (1.56 [1.42-1.70]), hospitalization in the South (1.81 [1.71-1.92]), hospitalization at a low- or medium-volume hospital (1.25 [1.19-1.31]), number of comorbidities (10.64 [8.89-12.62] for 5 vs none), sepsis (4.61 [4.34-4.89]), invasive fungal infection (1.24 [1.11-1.39]), and pneumonia (1.73 [1.63-1.82]). In-hospital mortality was higher for patients requiring ICU care (4857 of 11 277 [43.1%] vs 2959 of 31 972 [9.3%]). On multivariable analysis, independent risk factors for death in patients requiring ICU care included age 60 years or older (1.16 [1.06-1.26]), nonwhite race/ethnicity (1.18 [1.07-1.30]), hospitalization on the West coast (1.19 [1.06-1.34]), number of comorbidities (18.76 [13.7-25.67] for 5 vs none), sepsis (2.94 [2.70-3.21]), invasive fungal infection (1.20 [1.02-1.42]), and pneumonia (1.13 [1.04-1.24]). Mean costs of hospitalization were higher for patients requiring ICU care ($83 354 vs $41 973) and increased with each comorbidity, from $50 543 for patients with no comorbidities to $124 820 for those with 5 or more comorbidities.
Conclusions and Relevance
Admission to the ICU is associated with high mortality and cost that increase proportionally with the comorbidity burden in adults with AML. Several demographic factors and medical characteristics identify patients at risk for admission to the ICU and mortality and provide an opportunity for testing primary prevention strategies.
The survival of adults with acute myeloid leukemia (AML) has gradually improved during the last 4 decades, largely owing to advances in supportive care and hematopoietic cell transplantation, with some progress made in therapeutic strategies.1-3 Although they have contributed to improved survival, contemporary high-intensity therapies for AML bear the risk of complications that require intensive care unit (ICU)-level care. Despite advances in ICU management strategies for hematologic malignant neoplasms,4-6 ICU care remains associated with substantial mortality and long-term sequelae.7-10
Few studies have examined risk factors for admission to the ICU in patients with AML, with limited data pointing to a role of age, comorbidities, infection, and therapeutic regimens.10,11 Conversely, retrospective studies have identified a larger number of possible risk factors for mortality after admission to the ICU in patients with a wider range of hematologic malignancies, including performance status, Acute Physiology and Chronic Health Evaluation score, organ failure, need for mechanical ventilation, and use of vasopressors.4,9,10,12-15 However, most of these studies were small, single-institution investigations, limiting generalizability given variability across ICUs.16 Furthermore, few studies have focused specifically on patients with AML, although some evidence indicates that outcomes in such patients are particularly poor.13,14 Finally, while of great interest from a medical and economic perspective, to our knowledge, the effects of ICU care on use of health care resources, length of stay (LOS), and costs have not been evaluated.
Understanding relevant risk factors for admission to the ICU and short-term mortality, the primary goal of this study, is an essential step in identifying patients with AML at high risk for adverse events. This understanding would allow for the development of preemptive strategies aimed at optimizing treatment outcomes and reducing the economic consequences of AML therapy.
The University HealthSystem Consortium constitutes a collaboration between academic and affiliated health institutions for the purposes of research and clinical practice analysis.17 Currently, the University HealthSystem Consortium includes 117 academic centers with 338 affiliated hospitals throughout the United States, representing 43 states. The consortium established a longitudinal hospitalization database comprising a set of prespecified clinical variables (not including laboratory, radiologic, or pathologic data or cause of death) prospectively abstracted by certified coders from discharge summaries, and cost charges generated by consortium institutions. Data collection is closely monitored and strictly quality controlled to ensure completeness of data. For this analysis, no direct patient or institutional identifiers were provided to the investigators. Institutional review board waiver was granted by Duke University, Durham, North Carolina, where the original analysis was conducted, and completed at the Fred Hutchinson Cancer Research Center, Seattle, Washington. A waiver was granted on the basis that no patient identifiers were used. No informed consent was required.
The study population consisted of all adults 18 years or older with newly diagnosed, relapsed, or refractory AML who were hospitalized for any cause between January 1, 2004, and December 31, 2012, at University HealthSystem Consortium member hospitals. To identify patients of interest, inclusion criteria were developed based on International Classification of Diseases, Ninth Revision, Clinical Modification (ICD-9-CM) codes. All claims had to contain at least 1 diagnosis for active AML based on ICD-9-CM criteria (codes 205.00, 205.02, 205.30, 205.32, 206.00, 206.02, 207.00, and 207.02), which included patients admitted for therapy for AML or complications from AML or its treatment. Patients with a concurrent code for myelodysplastic syndrome were included. Admissions of patients with AML in remission, those who had undergone hematopoietic cell transplantation on the current or prior admission, and patients with other cancers were not included. For patients with more than 1 admission during the observation period, 1 random hospitalization was selected for analysis, as was done in prior similar studies using this database.18
Study Outcomes and Independent Variables
Primary outcomes for analysis were admission to the ICU and inpatient mortality among those requiring ICU care. Secondary outcomes included LOS in the ICU, total hospitalization LOS, and cost. Independent variables included age, sex, race/ethnicity (white, black, Hispanic, Asian, and other or unknown), hospitalization year, geographical location, hospital volume, comorbidities, infections, and procedures. Hospital volume was extrapolated from the number of patients with cancer admitted per year, defined as low (<3000 patients with cancer admitted per year), medium (3000-6000 patients with cancer admitted per year), and high (>6000 patients with cancer admitted per year). Comorbidities, infections, and procedures were examined based on 99 available ICD-9-CM codes. Comorbidities included congestive heart failure, other heart diseases, lung disease, liver disease, renal disease, cerebrovascular disease, peripheral vascular disease, type 1 or type 2 diabetes, deep venous thrombosis, and pulmonary embolism. Infections included type and site of infection, encompassing bacterial infections, fungal infections, invasive Candida infections, invasive aspergillosis, pneumonia, urinary tract infection, indwelling catheter infection, and sepsis. Procedures of interest were transfusions of red blood cells or platelets and in-hospital chemotherapy. Total costs per admission were provided by the consortium, adjusted for inflation, and converted to 2014 US dollars using the US Department of Labor’s Consumer Price Index Inflation calculator.19
Data analysis was performed from November 15, 2013, to August 15, 2016. For univariate analysis of binary outcomes, risk categories were compared using unadjusted odds ratios. The age variable was split into separate categories increasing by 10 years, starting at 40 years. The final cut points were chosen based on the proportion of ICU admissions by age category and mortality rates by age category. For continuous variables (LOS and cost), linear regression was used to compare differences between categories after natural logarithm transformation of both outcomes. A multivariable logistic regression model was used to evaluate associations of covariates with risk of admission to the ICU and mortality among patients requiring ICU care. For each model separately, the covariates were first screened by stepwise regression (P = .05 required for model entry and elimination). The final model included the demographic and clinical variables that were associated with increased risk during the selection process. The SEs and associated 95% CIs were calculated using the Wald method. In addition, to adjust for correlation between observations from the same hospital, robust SEs and 95% CIs were calculated using generalized estimating equation methods. Wald 95% CIs are presented within the text. Statistical analysis was performed using SAS, version 9.4 (SAS Institute Inc).
A total of 126 076 hospitalizations were reported for 49 530 patients with AML at 229 hospitals during the study period, with 26 102 patients (52.7%) admitted more than once. We excluded 6281 patients with a history of hematopoietic cell transplantation (5996 allogeneic and 285 autologous), leaving 43 249 for the final analysis. Table 1 presents the demographic characteristics, major comorbidities, and infections for the 11 277 patients (26.1%) who were admitted to the ICU and the 31 972 patients (73.9%) who were not admitted to the ICU during hospitalization.
Risk Factors for Admission to the ICU
On univariate analysis (with results reported as odds ratios [95% CIs]), risk factors for admission to the ICU included age younger than 80 years (1.41 [1.3-1.53]), black race/ethnicity (1.23 [1.15-1.32]), hospitalization in the South (1.58 [1.50-1.66]), hospitalization in a low- or medium-volume hospital (1.17 [1.12-1.23]), 1 or more comorbidity (3.58 [3.34-3.83]), sepsis (7.13 [6.76-7.53]), pneumonia (2.88 [2.75-3.02]), and invasive fungal infection (2.32 [2.11-2.55]; all P < .001). Figure, A, depicts the association between comorbidity burden and risk of ICU admission. Comorbidities that most affected risk of ICU admission included cerebrovascular, hepatic, and lung diseases (1452 of 2650 [54.8%], 1410 of 3007 [46.9%], and 6890 of 14 819 [46.5%] of patients with each comorbidity admitted to the ICU). Receipt of inpatient chemotherapy had minimal effect on the risk of ICU admission: 6203 of 23 145 patients (26.8%) who received chemotherapy required ICU admission vs 5074 of 20 104 patients (25.2%) who did not. On multivariable analysis (with results reported as odds ratios [95% CIs]), independent risk factors for ICU admission included age younger than 80 years (1.56 [1.42-1.70]), hospitalization in the South (1.81 [1.71-1.92]), hospitalization in a low- or medium-volume hospital (1.25 [1.19-1.31]), 1 or more comorbidity (10.64 [8.89-12.62] for 5 vs none), sepsis (4.61 [4.34-4.89]), invasive fungal infection (1.24 [1.11-1.39]), and pneumonia (1.73 [1.63-1.82]) (Table 2).
Risk Factors for Inpatient Mortality Among Patients Requiring ICU Care
Overall in-hospital mortality was 18.1% (n = 7816), but was significantly higher for patients requiring ICU care (4857 of 11 277 [43.1%] vs 2959 of 31 972 [9.3%]; P < .001). On univariate analysis (with results reported as odds ratios [95% CIs]), risk factors for in-hospital mortality for patients admitted to the ICU included age 60 years or older (1.41 [1.31-1.52]; P < .001), nonwhite race/ethnicity (1.23 [1.13-1.33]; P < .001), hospitalization at a high-volume hospital (1.15 [1.06-1.24]; P = .001), hospitalization on the West coast (1.30 [1.17-1.44]; P < .001), sepsis (4.19 [3.87-4.54]; P < .001), pneumonia (1.85 [1.72-2.00]; P < .001), and invasive fungal infection (1.74 [1.50-2.01]; P < .001). The number of major comorbid conditions for patients in the ICU was also highly associated with mortality risk. Specifically, for patients requiring ICU care, 61 of 1063 patients (5.7%) with no comorbidities died while in the hospital, whereas 423 of 1993 patients (21.2%) with 1 comorbidity and 432 of 649 patients (66.6%) with 5 or more comorbidities died while in the hospital (P < .001 for trend). Not all comorbid conditions were equally associated with mortality, with the highest risk noted for cerebrovascular and hepatic disease (930 of 1452 [64%] and 895 of 1410 [63.5%] mortality, respectively). Infection was an important contributor to risk, and high mortality rates were found in patients in the ICU with sepsis (2914 of 4603 [63.3%]), invasive fungal infections (446 of 799 [55.8%]), and pneumonia (2337 of 4479 [52.2%]). On multivariable analysis (with results reported as odds ratios [95% CIs]), independent risk factors for mortality in patients requiring ICU care were age 60 years or older (1.16 [1.06-1.26]), nonwhite race/ethnicity (1.18 [1.07-1.30]), hospitalization on the West coast (1.19 [1.06-1.34]), number of comorbidities (18.76 [13.7-25.67] for 5 vs none), sepsis (2.94 [2.70-3.21]), invasive fungal infection (1.20 [1.02-1.42]), and pneumonia (1.13 [1.04-1.24]) (Table 3).
The mean (median) LOS for patients who were not admitted to the ICU was 15.3 (8.0) days, compared with 22.4 (18.0) days for patients requiring ICU admission. On univariate analysis, other factors associated with prolonged hospitalization included age, geographical region, hospital volume, infection, and comorbid conditions (Table 4). For example, the LOS for those requiring ICU admission was longest at high-volume hospitals (23.8 [20.0] vs 19.4 [13.0] days for low-volume hospitals; P < .001). Length of stay was longest on the West coast, at 25.3 (21.0) days compared with 22.0 (17.0) days for other regions (P < .001).
The mean (median) total cost of hospitalization for the entire patient cohort was $52 833 ($29 487). These cost estimates were significantly higher for patients requiring ICU care than for patients not admitted to the ICU ($83 354 [$61 172] vs $41 973 [$21 751]; P < .001). For patients admitted to the ICU, costs varied considerably by demographic and clinical variables (Table 4), with the highest costs for high-volume hospitals vs low-volume hospitals ($96 979 [$72 468] vs $67 195 [$42 021]; P < .001 for trend). Comorbidities also had a significant effect on cost (Figure, B). Patients requiring ICU care without comorbidities had mean (median) hospitalization costs of $50 543 ($33 408), increasing to $124 820 ($99 292) for those with 5 or more comorbidities (P < .001 for trend). Among ICU patients with infections, those with invasive fungal infections ($150 413 [$126 415]) and catheter-associated infections ($127 691 [$104 302]) incurred the highest costs. Geography also significantly affected cost for patients requiring ICU care, with total mean (median) hospitalization costs highest on the West coast ($108 918 [$80 586]) and lowest in the South ($72 588 [$49 214]; P < .001). Finally, 765 patients requiring ICU admission were eventually discharged to hospice. The mean (median) hospitalization cost for patients transferred to hospice was slightly lower than for the other patients ($78 070 [$48 994] vs $83 740 [$61 942]; P = .01).
Admission rates to the ICU decreased slightly from 978 of 3730 patients (26.2%) in 2004 to 1630 of 7007 (23.3%) in 2012 (P < .001 for trend). Although in-hospital mortality decreased over time for the entire patient cohort (727 of 3730 [19.5%] in 2004 to 1013 of 7007 [14.5%] in 2012; P < .001), this decrease was driven by those not requiring ICU care (342 of 2752 [12.4%] in 2004 to 364 of 5377 [6.8%] in 2012; P < .001). In contrast, in-hospital mortality for patients requiring ICU care remained relatively constant during the observation period (385 of 978 [39.4%] in 2004 vs 649 of 1630 [39.8%] in 2012; P = .44 for trend). Mean total LOS remained unchanged for patients not admitted to the ICU (16.0 days in 2004 vs 15.9 days in 2012; P = .80 for trend) and those admitted to the ICU (23.2 days in 2004 vs 23.2 days in 2012; P = .27 for trend). Mean ICU days, however, decreased from 11.0 to 7.8 days between 2004 and 2012 (P < .001). Despite slightly decreased rates of admission to the ICU, decreased ICU LOS, and no change in overall LOS, costs after adjustment for inflation rose for patients not admitted to the ICU and those admitted to the ICU, although more notably for those admitted to the ICU ($83 771 in 2004 to $89 673 in 2012; P < .001 for trend).
Treatment-related mortality in patients with AML has decreased during the past 2 decades.2,20 Consistent with this fact, we found a decrease in mortality during the study period in our whole patient cohort. However, this decrease was almost entirely driven by reduced early mortality in patients not requiring ICU care, whereas ICU admission rates have only slightly decreased and ICU stays remained associated with persistently high mortality and rising costs during the past decade.
Although models are available to estimate the risk of treatment-related mortality after initial intensive chemotherapy for newly diagnosed AML21 or for patients undergoing hematopoietic cell transplantation,22 to our knowledge, no tools exist to reliably recognize patients at risk for ICU admission and subsequent short-term mortality. Here, we identify demographic variables, systemic factors, and potentially modifiable medical characteristics (eg, comorbidities and infections) associated with ICU admission, in-hospital mortality, and resource use in patients with AML. These findings could spawn the development of prediction models for ICU use and outcome, which might provide a tool for risk stratification, facilitate patient counseling, and could ultimately lead to new primary prevention and intervention strategies aimed at decreasing morbidity and mortality. Through targeting of modifiable risk factors, such strategies may be useful for improving patient outcomes.
Intensive care unit use and cost have risen in the United States since the 1990s, and critical care services now represent a higher proportion than ever before of the gross domestic product.23 Consistent with this trend, cost estimates in our analysis were more than double for patients with AML requiring ICU admission compared with those who did not, and increased over time despite stable mortality rates, stable total hospital LOS, and decreased ICU LOS. The specific reasons for this disproportionate rise in cost remain unknown, as detailed epidemiologic data encompassing the hospital care and resource use of patients with AML are lacking,24 but are likely interrelated with the increase in health care costs in general and in cancer care and ICU costs in particular.25 These costs are fueled by increasingly expensive therapies as well as expansion in the extent of care offered and the use and availability of technology in the oncology and critical care settings.26 An important step in addressing this rise in costs of cancer care includes the identification of contributing, modifiable factors, as revealed in our study. This knowledge provides the basis for future investigations on how these risk factors could be mitigated to reduce ICU use in patients with AML.
Another factor contributing to high costs are readmissions,27 which are common in patients with AML for both disease- and treatment-related complications. Notably, 23 428 patients (47.3%) in our database had only a single admission. In analyzing this subgroup, we found several potential contributing factors, including death on this first admission (6649 [13.4%]) as well as a higher rate of single admissions in the final year of the study period. Furthermore, these patients were more likely to be admitted to hospitals with lower volume, and thus possibly transferred to tertiary care centers not in the consortium for further care, and were older, and therefore potentially less likely to be treated and readmitted for complications. An upcoming study will examine risk factors for readmissions in patients with AML.
Evidence has demonstrated that hospital volume can affect mortality of patients undergoing surgical or medical cancer treatments.28 For patients with AML, previous studies indicated that mortality rates while undergoing chemotherapy are lower in high-volume vs low-volume centers despite similar mean LOS and cost.29 In this study, we found that hospital volume was not independently associated with mortality in patients with AML requiring ICU care, but that mean LOS and costs were significantly higher at high-volume hospitals. Referral bias of sicker patients to hospitals with higher volumes, and increased availability, and therefore use of, resources in hospitals with higher volumes may partly explain this observation, a notion supported by data indicating that ICU bed supply increases ICU bed use and health care spending, even after controlling for severity of illness.30 This difference may account for greater spending on critical care services in the United States compared with other countries30 and offers a rational target for intervention.
The use of a preexisting database offers several advantages, including the economy of data on a large scale and a large sample size, aspects uncommon to studies of patients with AML. In addition, as many recommendations for management of AML are based on findings from selected study patients that may not be applicable to the general population with AML,31,32 the use of real-world patient data may provide more generalizable insight. In addition, by including diverse hospitals across the United States, our approach removes the population bias of single-center studies. Still, a retrospective database analysis has several inherent limitations.
First and foremost, the University HealthSystem Consortium database relies on administrative coding and captures only hospital-based care. However, these limitations are unlikely to significantly affect our findings, as coding for AML and comorbidities has been validated in prior reports and found to be accurate,33,34 and most resource-intense care management of patients with AML remains hospital-based. In addition, depending on the coder, some patients in remission but receiving consolidation therapy for their disease likely were included as having active AML; inclusion of these patients, who may be considered less sick, may have lowered our estimates of poor outcomes and resource use. Furthermore, the variables available for analysis were predetermined and did not include data on disease characteristics (eg, cytogenetic profile), treatment phase (induction vs consolidation), response to therapy, reason for hospital and ICU admission, use of prophylactic antimicrobials, and cause of death. To address some of these limitations, we performed exploratory analyses investigating the effect of some of these factors by identifying patients with relapsed disease and those with a concurrent diagnosis of myelodysplastic syndrome, potentially a surrogate for AML with an antecedent hematologic disorder. Although patients with relapsed AML had a slightly lower risk of ICU admission than other patients, which remained significant in multivariable analysis (791 of 3383 [23.4%] vs 10 486 of 39 866 [26.3%]; odds ratio, 0.73; 95% CI, 0.67-0.80; P < .001), relapsed disease was not independently associated with mortality in multivariable analysis. Furthermore, concomitant diagnosis of myelodysplastic syndrome did not independently affect the risk of ICU admission or mortality in multivariable analysis. Another limitation is that the random selection of 1 hospitalization for patients admitted more than once during the study period may have led to an underestimate of true resource use, as patients with AML are often readmitted in short time spans for treatment-related complications, such as neutropenic fever, infections, or adverse effects on organs. Finally, the criteria for ICU admission vary among institutions, introducing heterogeneity into our study cohort. For example, admission of less sick patients to certain ICUs may affect mortality and estimates of resource use but would be expected to yield smaller, rather than larger, differences between patients requiring and not requiring ICU care.
Our findings demonstrate that ICU care remains common in patients with AML and is associated with substantial mortality and an increasing demand on health care resources over time. The identification of several factors that are significantly associated with ICU admission, mortality, and cost provide the basis for the development of tools for personalized, informed decisions in the management of AML that can assist in optimizing resource use for patients with AML and improving treatment outcomes.
Accepted for Publication: August 25, 2016.
Corresponding Author: Anna B. Halpern, MD, Hematology/Oncology Fellowship Program, Fred Hutchinson Cancer Research Center/University of Washington, 1100 Fairview Ave N, Suite D5-100, PO Box 91024, Seattle, WA 98109 (firstname.lastname@example.org).
Published Online: November 10, 2016. doi:10.1001/jamaoncol.2016.4858
Author Contributions: Dr Halpern had full access to all the data in the study and takes responsibility for the integrity of the data and the accuracy of the data analysis.
Study concept and design: Halpern, Lyman.
Acquisition, analysis, or interpretation of data: All authors.
Drafting of the manuscript: Halpern, Culakova, Lyman.
Critical revision of the manuscript for important intellectual content: All authors.
Statistical analysis: Culakova.
Study supervision: Lyman.
Conflict of Interest Disclosures: Drs Culakova and Lyman reported receiving salary support from research grant 214735 from Amgen for their institution. Dr Walter reported serving in a consulting and advisory role for Amphivena Therapeutics, Covagen AG, AstraZeneca, Seattle Genetics, and Pfizer, and reported receiving research funding from Seattle Genetics, Amgen, Celator, CSL Behring, Seattle Genetics, Amphivena Therapeutics, and Abbvie. No other disclosures were reported.
Funding/Support: Dr Halpern is 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 funding sources had 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.
Surveillance, Epidemiology, and End Results Program, National Cancer Institute. Previous version: SEER cancer statistics review, 1975-2012. http://seer.cancer.gov/archive/csr/1975_2012/
. Updated November 18, 2015. Accessed August 15, 2016.
et al. Declining rates of treatment-related mortality in patients with newly diagnosed AML given ‘intense’ induction regimens: a report from SWOG and MD Anderson. Leukemia
. 2014;28(2):289-292.PubMedGoogle ScholarCrossref
et al. Reduced mortality after allogeneic hematopoietic-cell transplantation. N Engl J Med
. 2010;363(22):2091-2101.PubMedGoogle ScholarCrossref
et al. Improved survival in cancer patients requiring mechanical ventilatory support: impact of noninvasive mechanical ventilatory support. Crit Care Med
. 2001;29(3):519-525.PubMedGoogle ScholarCrossref
et al. Noninvasive ventilation in immunosuppressed patients with pulmonary infiltrates, fever, and acute respiratory failure. N Engl J Med
. 2001;344(7):481-487.PubMedGoogle ScholarCrossref
et al; Early Goal-Directed Therapy Collaborative Group. Early goal-directed therapy in the treatment of severe sepsis and septic shock. N Engl J Med
. 2001;345(19):1368-1377.PubMedGoogle ScholarCrossref
et al. Outcomes and prognostic factors for patients with acute myeloid leukemia admitted to the intensive care unit. Leuk Lymphoma
. 2014;55(1):97-104.PubMedGoogle ScholarCrossref
Roze des Ordons
SM. Clinical characteristics and outcomes of patients with acute myelogenous leukemia admitted to intensive care: a case-control study. BMC Cancer
. 2010;10:516.PubMedGoogle ScholarCrossref
et al. Survival and predictors of outcome in patients with acute leukemia admitted to the intensive care unit. Cancer
. 2008;112(10):2233-2240.PubMedGoogle ScholarCrossref
et al. Prognostic factors for intensive care unit admission, intensive care outcome, and post-intensive care survival in patients with de novo acute myeloid leukemia: a single center experience. Haematologica
. 2011;96(2):231-237.PubMedGoogle ScholarCrossref
et al. Outcomes for older patients with acute myeloid leukemia admitted to the intensive care unit. Blood
. 2015;126(23):2104.Google Scholar
PC. Outcomes and prognostic factors in patients with haematological malignancy admitted to a specialist cancer intensive care unit: a 5 yr study. Br J Anaesth
. 2012;108(3):452-459.PubMedGoogle ScholarCrossref
et al. Prognostic factors in critically ill patients with hematologic malignancies admitted to the intensive care unit. J Crit Care
. 2012;27(6):739.e1-739.e6.PubMedGoogle ScholarCrossref
et al. Prognosis of patients with acute myeloid leukaemia admitted to intensive care. Br J Haematol
. 2005;129(3):350-357.PubMedGoogle ScholarCrossref
et al. Improved intensive care unit survival for critically ill allogeneic haematopoietic stem cell transplant recipients following reduced intensity conditioning. Br J Haematol
. 2013;161(4):578-586.PubMedGoogle ScholarCrossref
et al. Variability in outcome and resource use in intensive care units. Intensive Care Med
. 2007;33(8):1329-1336.PubMedGoogle ScholarCrossref
GH. Mortality, morbidity, and cost associated with febrile neutropenia in adult cancer patients. Cancer
. 2006;106(10):2258-2266.PubMedGoogle ScholarCrossref
CA. Improvements in the early death rate among 9380 patients with acute myeloid leukemia after initial therapy: A SEER database analysis. Cancer
. 2015;121(12):2004-2012.PubMedGoogle ScholarCrossref
et al. Prediction of early death after induction therapy for newly diagnosed acute myeloid leukemia with pretreatment risk scores: a novel paradigm for treatment assignment. J Clin Oncol
. 2011;29(33):4417-4423.PubMedGoogle ScholarCrossref
et al. Hematopoietic cell transplantation (HCT)–specific comorbidity index: a new tool for risk assessment before allogeneic HCT. Blood
. 2005;106(8):2912-2919.PubMedGoogle ScholarCrossref
SM. Critical care medicine in the United States 2000-2005: an analysis of bed numbers, occupancy rates, payer mix, and costs. Crit Care Med
. 2010;38(1):65-71.PubMedGoogle ScholarCrossref
HU. Resource use and outcome in critically ill patients with hematological malignancy: a retrospective cohort study. Crit Care
. 2008;12(3):R75.PubMedGoogle ScholarCrossref
E. Addressing the American health-care cost crisis: role of the oncology community. J Natl Cancer Inst
. 2013;105(23):1777-1781.PubMedGoogle ScholarCrossref
EA. Rehospitalizations among patients in the Medicare fee-for-service program. N Engl J Med
. 2009;360(14):1418-1428.PubMedGoogle ScholarCrossref
CE. Hospital and physician volume or specialization and outcomes in cancer treatment: importance in quality of cancer care. J Clin Oncol
. 2000;18(11):2327-2340.PubMedGoogle ScholarCrossref
MG. Impact of hospital volume on outcomes of patients undergoing chemotherapy for acute myeloid leukemia: a matched cohort study. Blood
. 2015;125(21):3359-3360.PubMedGoogle ScholarCrossref
JM. ICU bed supply, utilization, and health care spending: an example of demand elasticity. JAMA
. 2014;311(6):567-568.PubMedGoogle ScholarCrossref
et al. Do results from clinical trials in acute myeloid leukemia reflect clinical reality? a Danish national cohort study of 813 patients. Blood
. 2012;120(21):1477.Google Scholar
AK. Applicability of a “Pick a Winner” trial design to acute myeloid leukemia. Blood
. 2011;118(9):2389-2394.PubMedGoogle ScholarCrossref
WA. Validity of information on comorbidity derived from ICD-9-CCM
administrative data. Med Care
. 2002;40(8):675-685.PubMedGoogle ScholarCrossref
et al. Validation of the University HealthSystem Consortium administrative dataset: concordance and discordance with patient-level institutional data. J Surg Res
. 2014;190(2):484-490.PubMedGoogle ScholarCrossref