OBJECTIVE. Invasive aspergillosis (IA) is the most common filamentous fungal infection observed in immunocompromised patients. The incidence of invasive aspergillosis has increased significantly in recent decades in parallel with the increasing number and improved survival of immunocompromised patients. IA in adults has been well characterized; however, only a few small studies of IA in children have been reported. Therefore, the objective of this study was to describe the incidence and outcomes of children with IA
METHODS. We performed a retrospective cohort study using the 2000 Kids Inpatient Database, a national database of hospital inpatient stays during 2000. IA was defined as aspergillosis that occurred in a child with malignancy (solid tumor, leukemia, or lymphoma), hematologic/immunologic deficiency, or transplant (bone marrow or solid organ). Discharge weighting was applied to the data to obtain nationally representative estimates of disease.
RESULTS. During 2000, there were an estimated 666 pediatric cases of IA among 152231 immunocompromised children, yielding an annual incidence of 437/100000 (0.4%) among hospitalized immunocompromised children. Children with malignancy accounted for the majority (74%) of cases of IA. The highest incidence of IA was seen in children who had undergone allogeneic bone marrow transplantation (4.5%) and those with acute myelogenous leukemia (4%). The overall in-hospital mortality of immunocompromised children with IA was 18%. Children with malignancy and IA were at higher risk for death than children with malignancy and without IA. Pediatric patients with IA had a significantly longer median length of hospital stay (16 days) than immunocompromised children without IA (3 days). The median total hospital charges for patients with IA were $49309 compared with immunocompromised children without IA ($9035).
CONCLUSIONS. The impact of IA on increases in mortality, length of hospital stay, and the burden of cost in the hospital setting underscores the need for improved means of diagnosis, prevention, and treatment of IA in immunocompromised children.
Invasive aspergillosis (IA) is the most common filamentous fungal infection in immunocompromised patients. The incidence of IA1,2 has increased significantly in recent decades corresponding with the increase in the number of immunocompromised patients.3,4 At 1 of the largest bone marrow transplant (BMT) centers in the United States, the incidence of proven or probable IA increased from 7.9% in 1992 to 16.9% in 1998.4 Data from the Centers for Disease Control and Prevention report that the mortality associated with IA has increased 357% since 1980.2 IA is also associated with significant health care costs.5
The epidemiology and outcome of IA in adults have been well characterized1–4; however, only a few small studies of IA in children have been reported to date.6–11 Similar to many other disease processes, studies of adult patients with invasive fungal infections may not be relevant in pediatric patients.12 For example, with invasive candidiasis, there are differences in isolated Candida species and antifungal susceptibilities,13 risk factors,14,15 treatment,16 and outcomes14,17 between pediatric and adult patients. Little is known about the epidemiology, risk factors, outcome, and costs of IA in a large cohort of pediatric patients. We therefore conducted a retrospective cohort study of the national epidemiology and outcomes of IA in immunocompromised children using an Agency for Healthcare Research and Quality (AHRQ)-sponsored database of hospital discharges during the year 2000.
This retrospective cohort study was performed using the 2000 Kids' Inpatient Database (KID), which contains hospital inpatient information from US states that partnered with the AHRQ on the federally sponsored Healthcare Cost and Utilization Project18 (detailed information on the KID database is available at www.hcup-us.ahrq.gov). The data set reflects information on inpatient stays from short-term, nonfederal, nonrehabilitation general and specialty hospitals across 27 states and comprises a 10% sample of uncomplicated, in-hospital births and an 80% sample of all other pediatric admissions for patients ≤20 years of age. We excluded all normal, uncomplicated births and restricted pediatric admissions on the basis of the US Census definition of adulthood to any patient younger than 18 years at admission, yielding a data set of 1.9 million records from the total 2.5 million records contained in the KID 2000.
Immunocompromised Children and Invasive Aspergillosis
We defined IA as any patient with an International Classification of Diseases, Ninth Revision, Clinical Modification (ICD-9-CM) code of 117.3 (aspergillosis) and any 1 of the following immunocompromising conditions:
Malignancy (solid tumor, leukemia/lymphoma, and/or other malignancy not otherwise specified)
Hematologic/immunologic disorder (chronic granulomatous disease [CGD], Wiskott-Aldrich syndrome, HIV, combined immunodeficiency, congenital hypogammaglobulinemia, and aplastic anemia)
Transplant (bone marrow or solid organ)
Clinical and Demographic Data
We examined IA-associated hospitalizations by age, gender, race/ethnicity, location of hospital (standard US Census regions of Northeast, Midwest, South, and West), National Association of Children's Hospitals and Related Institution hospital type (rural, urban nonteaching, and urban teaching), comorbid diagnoses, and clinical procedure. The type of underlying malignancy and the presence of graft-versus-host disease (GVHD) were identified using ICD-9-CM codes for all solid tumor or hematologic malignancies, acquired or inherited immunologic deficiencies, and hematologic disorders. Solid organ and bone marrow transplantation were identified using the Clinical Classification Software, a tool developed by AHRQ for grouping patient diagnoses and procedures into a manageable number of clinically meaningful categories. (More detailed information on Clinical Classification Software is available at www.ahrq.gov/data/hcup.) All comorbidity variables were considered separately and were not integrated into any summary scoring system.
We used the Healthcare Cost and Utilization Project weighting method to generate national estimates of all reported characteristics. We report the frequency of IA as the number of estimated cases per 100000 hospital admissions. Survey methods were used to account for the sampling design and clustering of observations at specific reporting centers. Continuous data are reported as medians with interquartile ranges (IQRs). Mann-Whitney rank-sum tests were used in univariate analyses of continuous data comparing patients with IA with patients without IA. Categorical data were reported as frequencies and percentages. The χ2 test was used for unadjusted comparisons of patients with and without IA with respect to categorical variables. A 2-tailed P < .05 was considered significant for all statistical tests. All statistical procedures were performed using SAS version 9.1 (SAS Corp, Cary, NC).
Human Subjects Oversight
The conduct of this study was approved by the Committees for the Protection of Human Subjects at the Children's Hospital of Philadelphia.
We reviewed ∼1.9 million hospital discharge records that were in the database. On the basis of our inclusion criteria, we found 152231 immunocompromised children who were at risk for developing IA. Among those patients, 666 immunocompromised children had IA. Therefore, the incidence of IA among immunocompromised patients was 437/100000 (0.4%) pediatric admissions. The median (IQR) age of the immunocompromised children with IA was 13 years (8–15), whereas the median age of the immunocompromised children without IA was 7 (3–13; P < .001). The median age of children who had acute myelogenous leukemia (AML) and therefore were at high risk for IA was 8 (2–14). Fifty-five percent of the children with IA were male. Most immunocompromised children with reported race were white (52%), followed by Hispanic (14%) and black (4%). As expected, cases of IA were reported from each of the 4 major census regions in the United States, and the majority of patients were hospitalized at larger, urban hospitals.
The incidence of IA for children with specific underlying clinical conditions is presented in Table 1. Children with malignancy accounted for the majority (74%) of cases of IA, followed by those with a hematologic (28%) or immunologic (18%) disorder. BMT (15%) or solid organ transplant (1%) recipients composed the remainder of patients with IA.
The highest incidence of IA was seen in patients with Wiskott-Aldrich (30%), followed by CGD (6.5%), allogeneic BMT (4.5%), and AML (3.7%). Among allogeneic BMT patients, those with GVHD did not have a significantly higher incidence of IA than those without GVHD. When compared with patients who underwent autologous BMT, those who underwent allogeneic BMT had a significantly greater risk for developing IA (relative risk [RR]: 11.9; 95% confidence interval [CI]: 3.7–37.8). The incidence of IA in patients with AML was significantly greater than the incidence in patients with acute lymphocytic leukemia (ALL; RR: 5.6; 95% CI: 4.6–7.0).
Overall, 18% (122 of 666) of children with IA died in the hospital, compared with 1% (1736 of 151537) of similarly immunocompromised children without IA (P < .001). When patients were matched for underlying medical conditions, the patients with IA had significantly higher crude mortality (Table 2). The mortality rate for children with all types of malignancies and IA (21%) was far greater than that in children with malignancy but without IA (1%), yielding an RR of 13.5 (95% CI: 10.9–16.8). Certain malignancies deserve specific mention because the risk for death from IA in these patient groups was found to be much greater than without IA; among these patients were those with central nervous system (CNS) malignancies (RR: 21.6; 95% CI: 9.1–51.0), ALL (RR: 14.9; 95% CI: 10.2–21.7), and lymphoma (RR: 13.5; 95% CI: 6.7–27.1). Mortality rates for allogeneic BMT recipients with GVHD were similar to those of BMT recipients without GVHD (44% vs 52%; P = .80).
Immunocompromised children with IA had a longer median length of hospital stay (16 days; IQR: 3–38) than did immunocompromised children without IA and similar underlying disorders (3 days; IQR: 2–6; P < .001). The length of hospital stay for children with IA and CGD16 was shorter than that for children with BMT (42 days) or AML (23 days) and IA (P < .05). The median per-patient hospital charges for immunocompromised children with IA were $49309 (IQR: $7975–$189579) compared with immunocompromised children without IA ($9035; IQR: $4774–$19656; P < .01).
Our nationally representative analysis of IA in the United States provides a robust estimate of the incidence of pediatric IA in immunocompromised children and the impact of IA on these patients. To our knowledge, this is the first nationally representative study to report the incidence of IA in children with specific underlying conditions and the substantial attributable mortality of IA in immunocompromised children.
The incidence rate of IA among immunocompromised children in the United States was 0.5% in 2000. This rate of IA in hospitalized patients may be considered an approximation of the true incidence in the United States because almost all children with IA are hospitalized. In addition, we believe that we have reasonably estimated the attributable mortality for IA by comparing mortality rates of similarly immunocompromised children with and without IA. Comparing patients within these diagnostic subgroups allows for some adjustment of underlying risk for mortality on the basis of similar clinical status. We believe that we adequately captured the population at risk for IA because the number of patients who were reported to have undergone a bone marrow transplantation in the KID 2000 was approximately equal to the number of cases reported by the Children's Oncology Group for that year (Children's Oncology Group, data on file).
Previous epidemiologic studies of pediatric IA that determined incidence rates of IA have been limited to single-center studies.6–9,11 A retrospective cohort study of 485 pediatric BMT patients at Duke University reported a frequency of aspergillosis of 4.79%, with the highest incidence reported in recipients of allogeneic BMTs.7 A Finnish study of 148 pediatric BMT recipients similarly reported an IA incidence of 5%.6 The incidence of IA in BMT patients from these 2 centers is also similar to the incidence reported in our study. Marr et al4 reported an incidence of 13% in patients who were younger than 19 years. Differences in the definition of IA (the inclusion of probable cases) and differences in institutional experiences among single-center studies may partially explain the differences in incidence rates between studies.
Another study of IA in children was a review of 66 proven cases from ∼9500 children treated who were from 1962 to 1996 at St Jude's Children's Hospital in Memphis.9 The investigators reported the incidence of IA in specific immunocompromised children and found that 8% (2 of 25) of the patients with myelodysplastic syndrome had IA, followed by an IA incidence of 7% (1 of 14) in patients with CGD, 6% (1 of 16) in choriocarcinoma, 4.6% (2 of 43) in aplastic anemia, 4% (26 of 647) in AML, 4% (1 of 24) in chronic myelogenous leukemia, and 1% (29 of 2659) in ALL. Although we did not determine incidence rates for all of the aforementioned conditions, the incidence rates of IA in patients with CGD (coded as functional disease of neutrophils), ALL, and AML were similar to those reported above. Investigators from the Hospital for Sick Children in Toronto reviewed 39 cases of pediatric IA from 1979 to 19888 and found that 74% of patients with IA had a hematologic malignancy. In our study, we found a similar percentage of patients who had IA and a hematologic malignancy (71%).8 The high incidence of IA in patients with Wiskott-Aldrich syndrome was surprising, although the majority of patients with Wiskott-Aldrich undergo bone marrow transplantation. There are no data on IA in Wiskott-Aldrich syndrome. Additional data on IA in these patients are needed to confirm our findings.
Few studies have described the overall mortality of IA or the mortality of IA in specific populations of immunocompromised children. A review of IA case fatality rates pooled outcomes of 1941 patients from 1995 to 1999 stratified by decade of life,19 whereby the youngest cohort (≤20 years) had a case fatality rate of 68.2% (15 of 22).19 The 2 other studies of IA from St Jude's Hospital, the Hospital for Sick Children, and University Hospital of Frankfurt/Main reported mortality rates of 85%, 77%, and 69%, respectively.8,9,11 The case fatality in a single-center study of pediatric BMT patients was reported to be 87%.6 Our data revealed a mortality rate of 44% in children who had IA and received a BMT.
We postulate that the nationally representative sample frame of this study mitigated a potential bias in the published literature, which may preferentially contain reports from single institutions where IA is associated with an above-average mortality rate. If such a bias exists, then the use of nationally representative data sets may allow for much greater generalizability of these findings than those of previous single-institution studies. As was expected, children who had IA and had not undergone bone marrow transplantation had a lower overall mortality rate. To our knowledge, there are no other data on specific malignancy or immunodeficiency mortality rates for IA. Consistent with previous reports, we did not find a difference between BMT patients with and without GVHD.4 This is probably because of other transplant-specific variables, such as mismatched or unrelated donor status, that maybe overwhelming the effect of GVHD. We were unable to analyze our data for these transplant-specific variables using this database.
Overall, children with malignancy had a 13-fold higher risk for death when they developed IA. This rate varied between types of malignancies, with higher RRs identified in patients with ALL, lymphoma, and CNS and bone malignancies. The RR was lower in patients with AML, which likely reflects the higher mortality of the underlying disease even uncomplicated by IA and that death becomes a competing risk in the setting of high mortality rates.
A unique feature of our study was the use of KID 2000, a national pediatric inpatient database, to examine IA hospitalizations among children in the United States. KID is an all-payer hospital discharge database that was designed specifically to generate national estimates of pediatric hospitalizations of both common and rare childhood diseases. Use of this administrative database offers the unique advantage of allowing for the generation of national estimates of pediatric IA rates. Administrative data are limited, however, with specific regard to the possibility of miscoded or inaccurate information. Although the ICD-9-CM used in our study is the only code for aspergillosis, we are unaware of any analysis that has determined the sensitivity and the specificity of this particular ICD-9-CM code in detecting cases compared with a thorough review of all medical records. In general, health services researchers believe that the use of ICD-9-CM codes to identify cases in administrative databases has high specificity (eg, few instances in which patients did not in fact receive a diagnosis of the condition) but may be lower in sensitivity (ie, the administrative diagnosis may fail to detect all true cases). However, in the case of IA, physicians may tend to overdiagnose for fear of undertreating a potentially fatal infection in patients who are at high risk. Many if not most cases of IA are diagnosed empirically on the basis of clinical or radiologic findings that may not be diagnostic for IA.20 Therefore, if these cases are captured by the ICD-9 coding for IA, then we may overestimate the incidence. However, the incidence in this study is similar to the incidence reported in single-center chart review studies, making a gross overestimation unlikely.
The use of a case identification procedure with imperfect specificity and lower sensitivity would have led to an underestimation of the frequency of IA but is unlikely to have influenced the attributable outcome estimates in our study. Any resulting bias would be in the direction of more conservative, less dramatic findings than what in truth might be occurring. For example, it is possible, although unlikely given the underlying diseases of this population, that other forms of aspergillosis (eg, allergic bronchopulmonary aspergillosis [ABPA]) were miscoded as IA and thus included as cases. We identified only 3 patients in our cohort of 666 immunocompromised children who had a code for ABPA, whereas within the KID database, there were 11124 children with a code of cystic fibrosis, a leading underlying disease with ABPA. Of these children with cystic fibrosis, only 5 had a diagnosis of IA, including 1 patient who also had aplastic anemia and 1 who underwent solid-organ transplantation. There were no patients with the third and most obscure code for aspergillosis, 495.4 (Malt Worker's lung, alveolitis caused by Aspergillus clavatus). This misclassification of noninvasive forms of aspergillosis as IA should have minimal effect on our findings, and any bias likely would have resulted in an underestimation of the attributable mortality because those noninvasive forms of aspergillosis do not often result in death, thus creating a conservative bias in our results.
This is the first nationally representative report to date of large-scale epidemiology and attributable mortality of IA in specific underlying immunocompromising conditions in children. These results should incite additional exploration into new strategies to prevent and treat IA in specific high-risk pediatric populations. The total cost of hospitalization of >$25 million in 2000 for pediatric patients with IA should act as additional incentive for attempting to reduce the rate of IA. Early detection and initiation of appropriate treatment likely will reduce morbidity and mortality related to these infections among both pediatric and adult populations. The results of this study will be useful in the design and implementation of future interventions for early detection, prevention, and treatment of IA in children.
Dr Zaoutis was supported by grant K23 AI0629753-01 from the National Institutes of Health and grant U18-HS10399 from the Agency for Healthcare Research and Quality, Centers for Education and Research on Therapeutics.
- Accepted September 27, 2005.
- Address correspondence to Theoklis E. Zaoutis, MD, MSCE, Department of Pediatrics, Center for Clinical Epidemiology and Biostatistics, University of Pennsylvania School of Medicine, Division of Infectious Diseases, Children's Hospital of Philadelphia, 3535 Market St, Room 1527, Philadelphia, PA 19104. E-mail:
The authors have indicated they have no financial relationships relevant to this article to disclose.
- ↵McNeil MM, Nash SL, Hajjeh RA, et al. Trends in mortality due to invasive mycotic diseases in the United States, 1980–1997. Clin Infect Dis.2001;33 :641– 647
- ↵Denning DW. Invasive aspergillosis. Clin Infect Dis.1998;26 :781– 805
- ↵Marr KA, Carter RA, Crippa F, Wald A, Corey L. Epidemiology and outcome of mould infections in hematopoietic stem cell transplant recipients. Clin Infect Dis.2002;34 :909– 917
- ↵Dasbach EJ, Davies GM, Teutsch SM. Burden of aspergillosis-related hospitalizations in the United States. Clin Infect Dis.2000;31 :1524– 1528
- ↵Walsh TJ, Gonzales C, Lyman CA, Chanock S, Pizzo P. Invasive fungal infections in children: recent advances in diagnosis and treatment. Adv Pediatr Infect Dis.1995;11 :187– 290
- ↵Pfaller MA, Diekema DJ, Jones RN, Messer SA, Hollis RJ. Trends in antifungal susceptibility of Candida Spp. isolated from pediatric and adult patients with bloodstream infections: SENTRY antimicrobial surveillance program, 1997 to 2000. J Clin Microbiol.2002;40 :852– 856
- ↵Gonzalez CE, Venzon D, Lee S, Mueller BU, Pizzo PA, Walsh TJ. Risk factors for fungemia in children infected with human immunodeficiency virus: a case-control study. Clin Infect Dis.1996;23 :515– 521
- ↵Pappas PG, Rex JH, Lee J, et al. A prospective observational study of candidemia: epidemiology, therapy, and influences on mortality in hospitalized adult and pediatric patients. Clin Infect Dis.2003;37 :634– 643
- ↵Project HaCU. Kids' Inpatient Database (KID), 2000 [CD-ROM]. Rockville, MD: Agency for Healthcare Research and Quality; 2002
- ↵Ascioglu S, Rex JH, de Pauw B, et al. Defining opportunistic invasive fungal infections in immunocompromised patients with cancer and hematopoietic stem cell transplants: an international consensus. Clin Infect Dis.2002;34 :7– 14
- Copyright © 2006 by the American Academy of Pediatrics