Skip Navigation

Outcomes Attributable to Neonatal Candidiasis

  1. Theoklis E. Zaoutis1,2,3,4,6,
  2. Kateri Heydon1,
  3. Russell Localio5,6,
  4. Thomas J. Walsh8, and
  5. Chris Feudtner1,7
  1. 1Division of Infectious Diseases, Philadelphia
  2. 2Pediatric Generalists Research Group, Division of General Pediatrics, The Children's Hospital of Philadelphia, Philadelphia
  3. 3Departments of Pediatrics, Philadelphia
  4. 4Departments of Epidemiology, Philadelphia
  5. 5Departments of Biostatistics, Philadelphia
  6. 6Departments of Center for Epidemiology and Biostatistics, Philadelphia
  7. 7Departments of Leonard Davis Institute of Health Economics, University of Pennsylvania School of Medicine, Philadelphia
  8. 8Immunocompromised Host Section, Pediatric Oncology Branch, National Cancer Institute, Bethesda, Maryland
  1. Reprints or correspondence: Dr. Theoklis E. Zaoutis, Div. of Infectious Diseases, The Children's Hospital of Philadelphia, CHOP North, 3535 Market St., Ste. 1527, Philadelphia, PA 19104 (zaoutis{at}email.chop.edu).
 
Next Section

Abstract

Background. The incidence of candidiasis has increased in neonatal intensive care units, and invasive candidiasis is associated with significant morbidity and mortality. However, few data exist on outcomes directly attributable to neonatal candidiasis.

Methods. We estimated the incidence of systemic candidiasis in hospitalized neonates within the United States and determined the attributable mortality, length of hospital stay, and associated costs. We used the 2003 Kid's Inpatient Database from the Healthcare Cost and Utilization Project. Systemic candidiasis and comorbidities were defined by International Classification of Diseases, Ninth Revision, Clinical Modification codes. Neonates with uncomplicated births and neonates who died within the first 3 days of life were excluded. We used propensity score methods to balance covariates between the neonates with and neonates without candidiasis. Attributable outcomes were calculated between propensity score—matched neonates with and neonates without candidiasis. Because of the known confounding effect of birth weight, we performed separate propensity score analyses for extremely low birth weight (ELBW) neonates (i.e., neonates weighing <1000 g).

Results. The overall incidence of invasive candidiasis in neonates is 15 cases per 10,000 neonatal admissions (95% confidence interval [CI], 13–16 cases per 10,000 neonatal admissions). ELBW neonates with invasive candidiasis were 2 times more likely to die (odds ratio, 2.2; 95% CI, 1.4–3.5) than propensity-matched ELBW neonates without candidiasis. The propensity score—adjusted mortality rate attributable to candidiasis among ELBW neonates was 11.9%. Candidiasis in ELBW infants was not associated with an increase in length of hospital stay but was associated with a mean increase in total charges of $39,045 (95% CI, $1374–$76,715). Among infants with a birth weight ⩾1000 g, those who had candidiasis did not experience a significant increase in mortality, compared with infants without candidiasis. However, the propensity score—adjusted length of stay and charges attributable to candidiasis among neonates with a birth weight ⩾1000 g were 16 days (95% CI, 8–24 days) and $122,302 (95% CI, $80,457–$164,148), respectively.

Conclusions. Invasive candidiasis is associated with a significantly increased risk of death and excess hospital charges in ELBW neonates and with excess hospital stay and excess hospital charges in neonates with a birth weight ⩾1000 g.

Candida species are the leading cause of invasive fungal infection in the neonatal intensive care unit (NICU) and are the third most common blood culture isolates recovered from cases of late-onset sepsis in the NICU. Candidiasis is frequently associated with dissemination and resultant end-organ damage. The incidence of neonatal candidiasis (candidemia and/or disseminated candidiasis) in extremely low birth weight (ELBW) infants (defined as infants with a birth weight <1000 g) is 7%–20%, and it decreases with increasing birth weight to <1% in neonates with a birth weight >1500 g [1–4].

Neonatal candidiasis is associated with significant morbidity and mortality, and previous epidemiologic outcome studies of neonatal candidiasis have reported crude mortality rates of 30%–60%, with ELBW infants experiencing the highest mortality rates [2, 5–9]. Determining the health impact of infection due to Candida species on premature neonates is an important, yet difficult, task. Attributable outcomes in premature neonates are difficult to determine because of the potential confounding effect of comorbid conditions related to prematurity that predispose individuals both to candidiasis and to poor outcomes. Although the high crude mortality rates associated with candidiasis are well documented, the proportion of neonatal mortality and other health care—related outcomes that is attributable specifically to candidiasis is unknown. Therefore, we conducted a retrospective cohort study of neonatal candidiasis and used propensity score analyses to determine the outcomes attributable to neonatal candidiasis with use of a nationally representative database of hospital discharges, prepared by the Agency for Healthcare Research and Quality.

Previous SectionNext Section

Methods

Data sources. This study was performed using the 2003 Kids' Inpatient Database (KID 2003; AHRQ). The KID 2003 dataset contains hospital discharge information from US states that were partners of the Agency for Healthcare Research and Quality in the federally sponsored Healthcare Cost and Utilization Project (detailed information on the KID 2003 database is available at the Healthcare Cost and Utilization Project Web site [10]). The KID 2003 contains data from 3438 hospitals and includes >2.9 million pediatric hospital discharge records from 36 states. The dataset reflects hospital discharge information from short-term, nonfederal, nonrehabilitation general and specialty hospitals. KID 2003 data comprise a 10% sample of uncomplicated, in-hospital births and an 80% sample of all other pediatric admissions for subjects <20 years of age.

The KID 2003 includes demographic information, admission type and source, diagnostic codes, procedure codes, payer data, total charges, length of hospital stay (LOS), and hospital disposition. A KID 2003 hospital data file with facilities characteristics can be linked to the patient core data. The KID 2003 does not contain physiological or laboratory data. This study was approved by the Committees for the Protection of Human Subjects at The Children's Hospital of Philadelphia.

Study population. For this study, eligibility was limited to children admitted to the hospital on their day of birth. Children who died within 3 days after birth were excluded, because neonates are not considered to be at risk for candidiasis until after the third day of life [2, 11]. Our study cohort was derived from the 80% KID 2003 sample of all pediatric admissions. Children with uncomplicated births were excluded. This yielded a dataset of an estimated 294,783 unique patient records (on the basis of the KID 2003 weighting methodology) from the 7 million hospital records contained in the KID 2003.

Exposure classification. Neonatal candidiasis during a hospital stay was identified by the International Classification of Diseases, Ninth Revision, Clinical Modification (ICD-9-CM) code for disseminated and/or systemic candidiasis (112.5) or candidiasis site necrotizing enterocolitis (112.89). These codes may have been listed as any of the 15 diagnoses recorded on the hospital discharge record. Patients with diagnosis codes for mucosal candidiasis at a variety of body sites were not considered to have had neonatal candidiasis.

Validation of ICD-9-CM codes for the diagnosis of neonatal candidiasis. The ICD-9-CM coding system for identifying patients with neonatal candidiasis was validated using a cohort of patients hospitalized in the NICU at the Children's Hospital of Philadelphia (Philadelphia, PA) between October 2003 and June 2005. Patients were identified as having neonatal candidiasis by (1) infection-control surveillance records that included data on all health care—associated infections in the NICU or (2) microbiology laboratory confirmation of a blood culture positive for Candida species. A code of 112.5 or 112.89 was present in the discharge records of 14 of 16 Children's Hospital of Philadelphia patients with confirmed neonatal candidiasis, yielding a positive predictive value of 74% (95% CI, 49%–91%) and a sensitivity of 88% (95% CI, 62%–98%).

Clinical and demographic data. We examined hospitalizations by age group, sex, race and/or ethnicity, location of hospital (using standard 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), hospital size (small, medium, and large), comorbid diagnoses (up to 15 possible diagnoses), and clinical procedures (up to 15 possible procedures).

Birth weight data is collected as a variable in the KID 2003 dataset. For patients for whom birth weight was not recorded in the discharge record, we used existing ICD-9-CM diagnostic categories to define accepted categories of birth weight. Infants were divided into 2 groups: infants with a birth weight <1000 g (ELBW infants) and infants with a birth weight ⩾1000 g (non-ELBW infants).

The presence of concurrent chronic illness was assessed using a diagnostic classification system for pediatric complex chronic conditions, which was partitioned into the following 9 diagnostic categories: neuromuscular, cardiovascular, respiratory, hematological or immunological, metabolic, malignant, and genetic or other congenital defect conditions [12]. These categories were generated using ICD-9-CM codes.

In-hospital procedures were classified using Clinical Classification Software, a tool developed by the Agency for Healthcare Research and Quality for grouping patient diagnoses and procedures into a manageable number of clinically meaningful categories. These included vascular catheterization, mechanical ventilation, enteral or parenteral nutrition, and any gastrointestinal procedure. All variables were considered separately and were not integrated into any kind of summary comorbidity scoring system.

Outcome. The primary outcome in the propensity analysis was in-hospital mortality attributable to neonatal candidiasis. We defined the calculation for mortality attributable to candidiaisis as (mortality rate of patients with candidiasis) - (mortality rate of propensity-matched patients without candidiasis). Secondary outcomes included the LOS and total hospital charges that could be attributed to candidiasis and were calculated from the difference in LOS or hospital charges between patients with candidiasis and propensity-matched control subjects.

Analysis. Summary statistics were constructed using frequencies and proportions for categorical data elements and means and medians for continuous variables. The χ2 test was used for unadjusted comparisons between patients with and patients without candidiasis with respect to categorical variables. The dataset contains a weight variable for each observation, such that weighted analyses can generate national estimates (with 95% CIs) of total US hospital discharges for specific diagnoses and procedures. We report the frequency of candidiasis as the number of estimated cases per 10,000 neonatal hospital admissions.

Propensity score analysis. Propensity score analysis attempts to identify patients who are similar except for their treatment or exposure status; propensity scores reflect the probability, based on the patient's observed covariates, that the patient is exposed (e.g., has candidiasis) [13, 14]. The propensity analysis was conducted as follows and included only patients with available birth weight data. First, the probability that any patient would develop candidiasis during hospitalization (i.e., the propensity score) was estimated using a multivariable logistic regression model that incorporated available demographic, comorbidity, and procedural variables, irrespective of their presumed clinical relevance. Because of the known association between birth weight and both the exposure variable (candidiasis) and the outcome variables, we derived 2 separate datasets and performed separate analyses for ELBW infants and non-ELBW infants. Patients for whom birth weight could not be ascertained, either by the birth weight variable or ICD-9-CM code, were excluded.

Next, for each patient with candidiasis, we implemented nearest-neighbor matching to find the 2 unexposed patients with the closest propensity scores [15]. This matching was performed using SAS software, version 9.1 (SAS) and the SAS macro gmatch, obtained from the Mayo Clinic [16]. We then used a conditional logistic regression model, stratifying on the matched sets of 3 observations, to estimate the associations of outcome (death) and exposure (candidiasis).

We performed a sensitivity analysis to measure the potential influence an unmeasured covariate (e.g., marker of severity of illness) might have on the OR estimates of the association between outcome and candidiasis. We also performed a sensitivity analysis to measure the potential influence of misclassification bias [17]. The inclusion of ICD-9-CM code 112.89 may represent infants with focal, nonsystemic, or mucosal candidiasis. We examined the association between mortality and propensity score with and without the inclusion of ICD-9-CM code 112.89. Finally, we performed a sensitivity analysis to measure the influence of excluding patients on the basis of the day of death. For our primary analysis, we excluded children who died within 3 days after birth, because neonates are not considered to be at risk for candidiasis until after day 3 of life. We performed additional sensitivity analysis, varying the exclusion criteria across several levels: excluding patients with day of death <3, <7, or <10 days after birth. All statistical analyses were performed using Stata Statistical software, version 8.2 (Stata) and SAS software, version 9.1 (SAS).

Previous SectionNext Section

Results

There were a total of 294,783 estimated neonatal hospital admissions that met inclusion criteria for the study, including 9746 ELBW neonates and 164,793 non-ELBW neonates (table 1). During 2003, there were an estimated 433 cases of candidiasis in the United States, yielding an overall incidence of 0.15% (95% CI, 0.13%–0.16%). Data on birth weight were available for 385 of the neonates with candidiasis, and, therefore, all analyses that included birth weight as a variable were conducted using this subset of the 433 total cases. Among ELBW infants and non-ELBW infants, the incidence of candidiasis was 2.6% (95% CI, 2.3%–2.9%) and 0.08% (95% CI, 0.07%–0.1%), respectively. Overall, 252 (65%) of 385 cases of candidiasis occurred in ELBW infants.

Table 1
View larger version:
Table 1

Incidence of candidiasis among neonatal patients, by birth weight.

Outcomes in unmatched analysis. The crude mortality rate among all patients with candidiasis was 19%, compared with a mortality rate of 0.8% in all patients without candidiasis. Among ELBW infants with candidiasis, the crude mortality rate was 26%, compared with 13% among ELBW infants without candidiasis. Among non-ELBW infants with candidiasis, the crude mortality rate was 2%, compared with a mortality rate of 0.4% among non-ELBW infants without candidiasis.

The mean LOS and hospital charges among ELBW infants with candidiasis were 85 days and $374,481, respectively, compared with 67 days and $273,365, respectively, among ELBW infants without candidiasis. The mean LOS and hospital charges among non-ELBW infants with candidiasis were 62 days and $306,194, respectively, compared with 11 days and $34,642, respectively, among non-ELBW infants without candidiasis.

Outcomes in propensity-matched analysis. The candidiasis-exposed and candidiasis-unexposed patients, matched by propensity score, were similar overall with respect to observed demographic and clinical characteristics (table 2). After matching by propensity scores (table 3), candidiasis in ELBW infants was significantly associated with mortality; ELBW infants with candidiasis had a mortality rate of 26%, compared with a mortality rate among matched patients without candidiasis of 14%, yielding an absolute 11.9% (95% CI, 5.4%–18.3%) mortality rate increase attributable to candidiasis. Candidiasis among ELBW infants was not associated with a significant increase in LOS but was associated with a mean increase in total charges of $39,045.45 (95% CI, $1374–$76,715). After propensity score matching, non-ELBW infants with candidiasis did not experience a significant increase in mortality, compared with non-ELBW infants without candidiasis. However, candidiasis in non-ELBW infants resulted in a mean increase in LOS of 16 days and a mean increase in hospital charges of $122,302 (95% CI, $80,457–$164,148).

Table 2
View larger version:
Table 2

Demographic and clinical characteristics for propensity score—matched neonatal patients with and patients without candidiasis.

Table 3
View larger version:
Table 3

Outcomes attributable to neonatal candidiasis.

Sensitivity analyses. The association between candidiasis and death did not differ significantly when ICD-9-CM code 112.89 was excluded from the analysis. Furthermore, varying the exclusion criteria for day of death from day 3 of life to day 7 of life did not affect the results (table 4). When neonates who died within the first 10 days of hospitalization were excluded from the analysis, the OR for the association between candidiasis and mortality increased to 3.16 (95% CI, 1.95–5.22), and the attributable mortality increased to 14.7% (95% CI, 8.7–20.8). For the statistically significant association between candidiasis and death to disappear, there would have to be (1) a 5-fold difference in association of an omitted confounding variable and death among patients with candidiasis and patients without candidiasis, (2) a 2-fold difference in prevalence of the omitted confounding variable among patients with candidiasis and patients without candidiasis, and (3) a prevalence of the omitted variable of 20%–40%.

Table 4
View larger version:
Table 4

Sensitivity analysis of mortality attributable to candidemia in neonates.

Previous SectionNext Section

Discussion

We found that the mortality rate attributable to neonatal candidiasis in ELBW infants was 11.9%, whereas there was no significant increase in mortality rate among non-ELBW infants with candidiasis. We also found that candidiasis in ELBW infants had no effect on LOS and hospital charges but that, in non-ELBW infants, candidiasis was associated with a statistically significant increase in LOS and charges. Finally, we found that candidiasis is a disease that is seen almost exclusively in ELBW infants

Our estimate of a 26% crude mortality rate associated with candidiasis in ELBW infants is similar to prospective data collected by the National Institute of Child Health and Human Development—sponsored Neonatal Research Network (32%) and is almost identical to the findings of a previous retrospective study [2, 18]. Comparing our attributable mortality findings among ELBW infants with findings from previous studies is more difficult, because these studies included non-ELBW patients. In a prospective registry study of infants with birth weights <1500 g that was conducted by the National Institute of Child Health and Human Development Neonatal Research Network, the unadjusted mortality difference attributable to neonatal candidiasis was 22% [19]. The inclusion of neonates with greater birth weight may have biased the attributable mortality rate towards the null hypothesis, because larger infants have been reported to have a lower overall rate of mortality attributable to candidiasis. Our more conservative estimates of attributable mortality rate are likely caused by the adjustment of confounding effects on clinical outcomes by propensity-score matching for comorbidities and inpatient procedures. Our estimate of the attributable mortality rate associated with candidiasis among ELBW infants is similar to that reported by Benjamin et al. [7]. In our primary analysis, we excluded children who died within 3 days after birth, because neonates are not considered to be at risk for candidiasis until after day 3 of life. Our estimates of the attributable mortality increased when the children who died within 10 days after birth were excluded. We believe that this reflects a reduction over time in the overall likelihood of death among neonates, which causes the effect of candidemia to be more pronounced.

Our finding that candidiasis in ELBW infants had a relatively small effect on LOS and hospital charges may be attributable to censoring as a result of death, because more patients in the candidiasis group died and, therefore, had a shorter LOS than would otherwise have been the case, and LOS is directly related to hospital charges. In addition, the LOS of ELBW infants is likely to be driven by issues of gestational age and other issues of prematurity and not by neonatal candidiasis. This hypothesis is supported by the finding that, among non-ELBW infants, neonatal candidiasis was associated with significant increases in LOS and hospital charges; therefore, it was more likely for candidiasis to prolong the LOS of neonates who otherwise would not have remained hospitalized on the basis of gestational age.

Our incidence estimate for neonatal candidiasis among ELBW neonates is lower than estimates reported in previous studies [2, 11, 19]. In a recent study that involved 128 NICUs participating in the National Nosocomial Infections Surveillance system, the incidence of candidiasis among ELBW infants between 2000 and 2004 was 5% [8]. The lower incidence reported in our study may be attributable to the lower sensitivity of ICD-9-CM codes for the diagnosis of candidiasis. Our incidence estimate for candidiasis among non-ELBW neonates is low and is consistent with previous reports [1, 8].

Use of these national administrative databases offers the unique advantage of allowing for the generation of nationwide estimates of candidiasis rates. Administrative data are limited, however, with specific regard to the possibility of miscoded or inaccurate information. Although 112.5 is the only ICD-9-CM code that explicitly describes systemic disease and has been used in a previous study of candidiasis that used administrative data, we are unaware of any analysis that has determined the sensitivity and specificity of this particular ICD-9-CM code for detecting cases, compared with, for example, a thorough review of all medical charts. In our review of the medical records of neonates with documented candidiasis at our center, we identified an additional code used for patients with candidiasis, 112.89, which resulted in improved sensitivity and positive predictive value. The addition of this code did not alter the association between candidiasis and death in ELBW infants. We believe that the coding practices for candidiasis at the other institutions are unlikely to be significantly different, given the rarity of the disease, as well as the unambiguous case definition of a patient with a blood or tissue culture that grows Candida species (compared with “sepsis,” which is not defined by the result of a definitive test). Our finding of high specificity and low sensitivity is consistent with the findings of previous studies that have used ICD-9-CM codes to identify cases in administrative databases; although high in specificity (i.e., resulting in few instances in which patients did not, in fact, receive a diagnosis of the condition), this method may be low in sensitivity (i.e., the administrative diagnosis may fail to detect all true cases) [20–23].

Any analysis of a potential cause-and-effect relationship between candidiasis and clinical outcome would be strengthened by more information than this study and its datasets can provide regarding the temporal sequence of events and the severity of illness before the development of candidiasis. As with all observational studies, propensity analyses cannot completely control for the effect of confounding, because they can only adjust for factors that were measured in the cohort, and residual confounding, therefore, remains a possibility. Overall, the propensity-score matching served as an adequate and robust method for controlling a large number of factors, as shown by the lack of statistically significant differences among almost all of the observed variables. Finally, our sensitivity analysis revealed that an omitted confounding factor would have to be very common and be associated with a 5-fold higher risk of death in the candidiasis group than in the noncandidiasis group for the significant result to disappear, which is an unlikely scenario.

In summary, the attributable mortality of candidiasis among ELBW infants is substantial, whereas non-ELBW infants do not seem to experience excess mortality as a result of candidiasis but do have increased use of health care resources. Our results suggest that, in this population, 1 life would be saved for every 8 ELBW infants in whom candidiasis can be prevented. One single-center, randomized clinical trial has supported the use of antifungal prophylaxis for the prevention of candidiasis in ELBW infants; however, the incidence of candidiasis at that center was extremely high [4]. Given the overall low incidence of candidiasis among ELBW infants, validation of clinical prediction rules that identify subsets of ELBW infants at particularly high risk for candidiasis, with subsequent intervention studies focused on this subgroup of infants, is critically needed. Previous investigators have suggested that preventative strategies should be targeted to populations with a baseline rate of candidiasis of >10% [24]. We hope that these findings will be useful in the design and implementation of future interventions.

Previous SectionNext Section

Acknowledgments

Financial support. National Institutes of Health (1K23 AI0629753-01 to T.E.Z.).

Potential conflicts of interest. All authors: no conflicts.

  • Received October 2, 2006.
  • Accepted January 15, 2007.
Previous Section
 

References

  1. ↵
    1. Smith PB,
    2. Steinbach WJ,
    3. Benjamin DK Jr
    . Neonatal candidiasis. Infect Dis Clin North Am 2005;19:603-15.
    CrossRefMedlineWeb of Science
  2. ↵
    1. Benjamin DK Jr,
    2. Stoll BJ,
    3. Fanaroff AA,
    4. et al
    . Neonatal candidiasis among extremely low birth weight infants: risk factors, mortality rates, and neurodevelopmental outcomes at 18 to 22 months. Pediatrics 2006;117:84-92.
    Abstract/FREE Full Text
    1. Manzoni P,
    2. Arisio R,
    3. Mostert M,
    4. et al
    . Prophylactic fluconazole is effective in preventing fungal colonization and fungal systemic infections in preterm neonates: a single-center, 6-year, retrospective cohort study. Pediatrics 2006;117:e22-32.
    Abstract/FREE Full Text
  4. ↵
    1. Kaufman D,
    2. Boyle R,
    3. Hazen KC,
    4. Patrie JT,
    5. Robinson M,
    6. Donowitz LG
    . Fluconazole prophylaxis against fungal colonization and infection in preterm infants. N Engl J Med 2001;345:1660-6.
    CrossRefMedlineWeb of Science
  5. ↵
    1. Saiman L,
    2. Ludington E,
    3. Pfaller M,
    4. et al
    . Risk factors for candidemia in neonatal intensive care unit patients. The National Epidemiology of Mycosis Survey study group. Pediatr Infect Dis J 2000;19:319-24.
    CrossRefMedlineWeb of Science
    1. Friedman S,
    2. Richardson SE,
    3. Jacobs SE,
    4. O'Brien K
    . Systemic Candida infection in extremely low birth weight infants: short term morbidity and long term neurodevelopmental outcome. Pediatr Infect Dis J 2000;19:499-504.
    CrossRefMedlineWeb of Science
  7. ↵
    1. Benjamin DK,
    2. DeLong E,
    3. Cotten CM,
    4. Garges HP,
    5. Steinbach WJ,
    6. Clark RH
    . Mortality following blood culture in premature infants: increased with gram-negative bacteremia and candidemia, but not gram-positive bacteremia. J Perinatol 2004;24:175-80.
    CrossRefMedline
  8. ↵
    1. Fridkin SK,
    2. Kaufman D,
    3. Edwards JR,
    4. Shetty S,
    5. Horan T
    . Changing incidence of Candida bloodstream infections among NICU patients in the United States: 1995–2004. Pediatrics 2006;117:1680-7.
    Abstract/FREE Full Text
  9. ↵
    1. Lee BE,
    2. Cheung PY,
    3. Robinson JL,
    4. Evanochko C,
    5. Robertson CM
    . Comparative study of mortality and morbidity in premature infants (birth weight, <1250 g) with candidemia or candidal meningitis. Clin Infect Dis 1998;27:559-65.
    Abstract/FREE Full Text
  10. ↵
    Healthcare Cost and Utilization Project Web site. Available at: http://www.hcup-us.ahrq.gov. Accessed 9 March 2007.
  11. ↵
    1. Benjamin DK Jr,
    2. DeLong ER,
    3. Steinbach WJ,
    4. Cotton CM,
    5. Walsh TJ,
    6. Clark RH
    . Empirical therapy for neonatal candidemia in very low birth weight infants. Pediatrics 2003;112:543-7.
    Abstract/FREE Full Text
  12. ↵
    1. Feudtner C,
    2. Christakis DA,
    3. Connell FA
    . Pediatric deaths attributable to complex chronic conditions: a population-based study of Washington State, 1980–1997. Pediatrics 2000;106:205-9.
    MedlineWeb of Science
  13. ↵
    1. Weitzen S,
    2. Lapane KL,
    3. Toledano AY,
    4. Hume AL,
    5. Mor V
    . Principles for modeling propensity scores in medical research: a systematic literature review. Pharmacoepidemiol Drug Saf 2004;13:841-53.
    CrossRefMedlineWeb of Science
  14. ↵
    1. D'Agostino RB Jr
    . Propensity score methods for bias reduction in the comparison of a treatment to a non-randomized control group. Stat Med 1998;17:2265-81.
    CrossRefMedlineWeb of Science
  15. ↵
    1. Rosenbaum PR,
    2. Rubin DB
    . RPaR. Constructing a control group using multivariate matched sampling methods that incorporate the propensity score. Am Stat 1985;39:33-8.
    CrossRef
  16. ↵
    Mayo Clinic College of Medicine Division of Biostatistics Web site. Available at: http://cancercenter.mayo.edu/mayo/research/biostat/sasmacros.cfm. Accessed 9 March 2007.
  17. ↵
    1. Lin DY,
    2. Psaty BM,
    3. Kronmal RA
    . Assessing the sensitivity of regression results to unmeasured confounders in observational studies. Biometrics 1998;54:948-63.
    CrossRefMedlineWeb of Science
  18. ↵
    1. Kossoff EH,
    2. Buescher ES,
    3. Karlowicz MG
    . Candidemia in a neonatal intensive care unit: trends during fifteen years and clinical features of 111 cases. Pediatr Infect Dis J 1998;17:504-8.
    CrossRefMedlineWeb of Science
  19. ↵
    1. Stoll BJ,
    2. Hansen N,
    3. Fanaroff AA,
    4. et al
    . Late-onset sepsis in very low birth weight neonates: the experience of the NICHD Neonatal Research Network. Pediatrics 2002;110:285-91.
    Abstract/FREE Full Text
  20. ↵
    1. Shay DK,
    2. Holman RC,
    3. Newman RD,
    4. Liu LL,
    5. Stout JW,
    6. Anderson LJ
    . Bronchiolitis-associated hospitalizations among US children, 1980–1996. JAMA 1999;282:1440-6.
    Abstract/FREE Full Text
    1. Thompson WW,
    2. Shay DK,
    3. Weintraub E,
    4. et al
    . Influenza-associated hospitalizations in the United States. JAMA 2004;292:1333-40.
    Abstract/FREE Full Text
    1. Zhan C,
    2. Miller MR
    . Excess length of stay, charges, and mortality attributable to medical injuries during hospitalization. JAMA 2003;290:1868-74.
    Abstract/FREE Full Text
  23. ↵
    1. Rosamond WD,
    2. Chambless LE,
    3. Sorlie PD,
    4. et al
    . Trends in the sensitivity, positive predictive value, false-positive rate, and comparability ratio of hospital discharge diagnosis codes for acute myocardial infarction in four US communities, 1987–2000. Am J Epidemiol 2004;160:1137-46.
    Abstract/FREE Full Text
  24. ↵
    1. Rex JH,
    2. Sobel JD
    . Prophylactic antifungal therapy in the intensive care unit. Clin Infect Dis 2001;32:1191-200.
    CrossRefMedlineWeb of Science
| Table of Contents

This Article

  1. Clin Infect Dis. (2007) 44 (9): 1187-1193. doi: 10.1086/513196
  1. AbstractFree
  2. » Full Text (HTML)Free
  3. Full Text (PDF)Free

Classifications

Share

  1. Email this article

Navigate This Article

Current Issue

  1. March 1, 2012 54 (5)
  1. Clinical Infectious Diseases
  1. Alert me to new issues

Most